diff options
Diffstat (limited to 'kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr')
66 files changed, 6739 insertions, 5705 deletions
diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/Kbuild b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/Kbuild index 2e1b92f71..9ad0d0e78 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/Kbuild +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/Kbuild @@ -1,21 +1,8 @@ -nvkm-y += nvkm/engine/gr/ctxnv40.o -nvkm-y += nvkm/engine/gr/ctxnv50.o -nvkm-y += nvkm/engine/gr/ctxgf100.o -nvkm-y += nvkm/engine/gr/ctxgf108.o -nvkm-y += nvkm/engine/gr/ctxgf104.o -nvkm-y += nvkm/engine/gr/ctxgf110.o -nvkm-y += nvkm/engine/gr/ctxgf117.o -nvkm-y += nvkm/engine/gr/ctxgf119.o -nvkm-y += nvkm/engine/gr/ctxgk104.o -nvkm-y += nvkm/engine/gr/ctxgk20a.o -nvkm-y += nvkm/engine/gr/ctxgk110.o -nvkm-y += nvkm/engine/gr/ctxgk110b.o -nvkm-y += nvkm/engine/gr/ctxgk208.o -nvkm-y += nvkm/engine/gr/ctxgm107.o -nvkm-y += nvkm/engine/gr/ctxgm204.o -nvkm-y += nvkm/engine/gr/ctxgm206.o +nvkm-y += nvkm/engine/gr/base.o nvkm-y += nvkm/engine/gr/nv04.o nvkm-y += nvkm/engine/gr/nv10.o +nvkm-y += nvkm/engine/gr/nv15.o +nvkm-y += nvkm/engine/gr/nv17.o nvkm-y += nvkm/engine/gr/nv20.o nvkm-y += nvkm/engine/gr/nv25.o nvkm-y += nvkm/engine/gr/nv2a.o @@ -23,18 +10,43 @@ nvkm-y += nvkm/engine/gr/nv30.o nvkm-y += nvkm/engine/gr/nv34.o nvkm-y += nvkm/engine/gr/nv35.o nvkm-y += nvkm/engine/gr/nv40.o +nvkm-y += nvkm/engine/gr/nv44.o nvkm-y += nvkm/engine/gr/nv50.o +nvkm-y += nvkm/engine/gr/g84.o +nvkm-y += nvkm/engine/gr/gt200.o +nvkm-y += nvkm/engine/gr/mcp79.o +nvkm-y += nvkm/engine/gr/gt215.o +nvkm-y += nvkm/engine/gr/mcp89.o nvkm-y += nvkm/engine/gr/gf100.o -nvkm-y += nvkm/engine/gr/gf108.o nvkm-y += nvkm/engine/gr/gf104.o +nvkm-y += nvkm/engine/gr/gf108.o nvkm-y += nvkm/engine/gr/gf110.o nvkm-y += nvkm/engine/gr/gf117.o nvkm-y += nvkm/engine/gr/gf119.o nvkm-y += nvkm/engine/gr/gk104.o -nvkm-y += nvkm/engine/gr/gk20a.o nvkm-y += nvkm/engine/gr/gk110.o nvkm-y += nvkm/engine/gr/gk110b.o nvkm-y += nvkm/engine/gr/gk208.o +nvkm-y += nvkm/engine/gr/gk20a.o nvkm-y += nvkm/engine/gr/gm107.o nvkm-y += nvkm/engine/gr/gm204.o nvkm-y += nvkm/engine/gr/gm206.o +nvkm-y += nvkm/engine/gr/gm20b.o + +nvkm-y += nvkm/engine/gr/ctxnv40.o +nvkm-y += nvkm/engine/gr/ctxnv50.o +nvkm-y += nvkm/engine/gr/ctxgf100.o +nvkm-y += nvkm/engine/gr/ctxgf104.o +nvkm-y += nvkm/engine/gr/ctxgf108.o +nvkm-y += nvkm/engine/gr/ctxgf110.o +nvkm-y += nvkm/engine/gr/ctxgf117.o +nvkm-y += nvkm/engine/gr/ctxgf119.o +nvkm-y += nvkm/engine/gr/ctxgk104.o +nvkm-y += nvkm/engine/gr/ctxgk110.o +nvkm-y += nvkm/engine/gr/ctxgk110b.o +nvkm-y += nvkm/engine/gr/ctxgk208.o +nvkm-y += nvkm/engine/gr/ctxgk20a.o +nvkm-y += nvkm/engine/gr/ctxgm107.o +nvkm-y += nvkm/engine/gr/ctxgm204.o +nvkm-y += nvkm/engine/gr/ctxgm206.o +nvkm-y += nvkm/engine/gr/ctxgm20b.o diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/base.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/base.c new file mode 100644 index 000000000..090765ff0 --- /dev/null +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/base.c @@ -0,0 +1,136 @@ +/* + * Copyright 2015 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs <bskeggs@redhat.com> + */ +#include "priv.h" + +#include <engine/fifo.h> + +static void +nvkm_gr_tile(struct nvkm_engine *engine, int region, struct nvkm_fb_tile *tile) +{ + struct nvkm_gr *gr = nvkm_gr(engine); + if (gr->func->tile) + gr->func->tile(gr, region, tile); +} + +u64 +nvkm_gr_units(struct nvkm_gr *gr) +{ + if (gr->func->units) + return gr->func->units(gr); + return 0; +} + +int +nvkm_gr_tlb_flush(struct nvkm_gr *gr) +{ + if (gr->func->tlb_flush) + return gr->func->tlb_flush(gr); + return -ENODEV; +} + +static int +nvkm_gr_oclass_get(struct nvkm_oclass *oclass, int index) +{ + struct nvkm_gr *gr = nvkm_gr(oclass->engine); + int c = 0; + + if (gr->func->object_get) { + int ret = gr->func->object_get(gr, index, &oclass->base); + if (oclass->base.oclass) + return index; + return ret; + } + + while (gr->func->sclass[c].oclass) { + if (c++ == index) { + oclass->base = gr->func->sclass[index]; + return index; + } + } + + return c; +} + +static int +nvkm_gr_cclass_new(struct nvkm_fifo_chan *chan, + const struct nvkm_oclass *oclass, + struct nvkm_object **pobject) +{ + struct nvkm_gr *gr = nvkm_gr(oclass->engine); + if (gr->func->chan_new) + return gr->func->chan_new(gr, chan, oclass, pobject); + return 0; +} + +static void +nvkm_gr_intr(struct nvkm_engine *engine) +{ + struct nvkm_gr *gr = nvkm_gr(engine); + gr->func->intr(gr); +} + +static int +nvkm_gr_oneinit(struct nvkm_engine *engine) +{ + struct nvkm_gr *gr = nvkm_gr(engine); + if (gr->func->oneinit) + return gr->func->oneinit(gr); + return 0; +} + +static int +nvkm_gr_init(struct nvkm_engine *engine) +{ + struct nvkm_gr *gr = nvkm_gr(engine); + return gr->func->init(gr); +} + +static void * +nvkm_gr_dtor(struct nvkm_engine *engine) +{ + struct nvkm_gr *gr = nvkm_gr(engine); + if (gr->func->dtor) + return gr->func->dtor(gr); + return gr; +} + +static const struct nvkm_engine_func +nvkm_gr = { + .dtor = nvkm_gr_dtor, + .oneinit = nvkm_gr_oneinit, + .init = nvkm_gr_init, + .intr = nvkm_gr_intr, + .tile = nvkm_gr_tile, + .fifo.cclass = nvkm_gr_cclass_new, + .fifo.sclass = nvkm_gr_oclass_get, +}; + +int +nvkm_gr_ctor(const struct nvkm_gr_func *func, struct nvkm_device *device, + int index, u32 pmc_enable, bool enable, struct nvkm_gr *gr) +{ + gr->func = func; + return nvkm_engine_ctor(&nvkm_gr, device, index, pmc_enable, + enable, &gr->engine); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.c index 57e2c5b13..56f392d3d 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.c @@ -23,7 +23,6 @@ */ #include "ctxgf100.h" -#include <subdev/bar.h> #include <subdev/fb.h> #include <subdev/mc.h> #include <subdev/timer.h> @@ -1005,6 +1004,7 @@ void gf100_grctx_mmio_item(struct gf100_grctx *info, u32 addr, u32 data, int shift, int buffer) { + struct nvkm_device *device = info->gr->base.engine.subdev.device; if (info->data) { if (shift >= 0) { info->mmio->addr = addr; @@ -1021,29 +1021,29 @@ gf100_grctx_mmio_item(struct gf100_grctx *info, u32 addr, u32 data, return; } - nv_wr32(info->priv, addr, data); + nvkm_wr32(device, addr, data); } void gf100_grctx_generate_bundle(struct gf100_grctx *info) { - const struct gf100_grctx_oclass *impl = gf100_grctx_impl(info->priv); + const struct gf100_grctx_func *grctx = info->gr->func->grctx; const u32 access = NV_MEM_ACCESS_RW | NV_MEM_ACCESS_SYS; const int s = 8; - const int b = mmio_vram(info, impl->bundle_size, (1 << s), access); + const int b = mmio_vram(info, grctx->bundle_size, (1 << s), access); mmio_refn(info, 0x408004, 0x00000000, s, b); - mmio_wr32(info, 0x408008, 0x80000000 | (impl->bundle_size >> s)); + mmio_wr32(info, 0x408008, 0x80000000 | (grctx->bundle_size >> s)); mmio_refn(info, 0x418808, 0x00000000, s, b); - mmio_wr32(info, 0x41880c, 0x80000000 | (impl->bundle_size >> s)); + mmio_wr32(info, 0x41880c, 0x80000000 | (grctx->bundle_size >> s)); } void gf100_grctx_generate_pagepool(struct gf100_grctx *info) { - const struct gf100_grctx_oclass *impl = gf100_grctx_impl(info->priv); + const struct gf100_grctx_func *grctx = info->gr->func->grctx; const u32 access = NV_MEM_ACCESS_RW | NV_MEM_ACCESS_SYS; const int s = 8; - const int b = mmio_vram(info, impl->pagepool_size, (1 << s), access); + const int b = mmio_vram(info, grctx->pagepool_size, (1 << s), access); mmio_refn(info, 0x40800c, 0x00000000, s, b); mmio_wr32(info, 0x408010, 0x80000000); mmio_refn(info, 0x419004, 0x00000000, s, b); @@ -1053,13 +1053,13 @@ gf100_grctx_generate_pagepool(struct gf100_grctx *info) void gf100_grctx_generate_attrib(struct gf100_grctx *info) { - struct gf100_gr_priv *priv = info->priv; - const struct gf100_grctx_oclass *impl = gf100_grctx_impl(priv); - const u32 attrib = impl->attrib_nr; - const u32 size = 0x20 * (impl->attrib_nr_max + impl->alpha_nr_max); + struct gf100_gr *gr = info->gr; + const struct gf100_grctx_func *grctx = gr->func->grctx; + const u32 attrib = grctx->attrib_nr; + const u32 size = 0x20 * (grctx->attrib_nr_max + grctx->alpha_nr_max); const u32 access = NV_MEM_ACCESS_RW; const int s = 12; - const int b = mmio_vram(info, size * priv->tpc_total, (1 << s), access); + const int b = mmio_vram(info, size * gr->tpc_total, (1 << s), access); int gpc, tpc; u32 bo = 0; @@ -1067,91 +1067,95 @@ gf100_grctx_generate_attrib(struct gf100_grctx *info) mmio_refn(info, 0x419848, 0x10000000, s, b); mmio_wr32(info, 0x405830, (attrib << 16)); - for (gpc = 0; gpc < priv->gpc_nr; gpc++) { - for (tpc = 0; tpc < priv->tpc_nr[gpc]; tpc++) { + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + for (tpc = 0; tpc < gr->tpc_nr[gpc]; tpc++) { const u32 o = TPC_UNIT(gpc, tpc, 0x0520); mmio_skip(info, o, (attrib << 16) | ++bo); mmio_wr32(info, o, (attrib << 16) | --bo); - bo += impl->attrib_nr_max; + bo += grctx->attrib_nr_max; } } } void -gf100_grctx_generate_unkn(struct gf100_gr_priv *priv) +gf100_grctx_generate_unkn(struct gf100_gr *gr) { } void -gf100_grctx_generate_tpcid(struct gf100_gr_priv *priv) +gf100_grctx_generate_tpcid(struct gf100_gr *gr) { + struct nvkm_device *device = gr->base.engine.subdev.device; int gpc, tpc, id; for (tpc = 0, id = 0; tpc < 4; tpc++) { - for (gpc = 0; gpc < priv->gpc_nr; gpc++) { - if (tpc < priv->tpc_nr[gpc]) { - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x698), id); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x4e8), id); - nv_wr32(priv, GPC_UNIT(gpc, 0x0c10 + tpc * 4), id); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x088), id); + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + if (tpc < gr->tpc_nr[gpc]) { + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x698), id); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x4e8), id); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0c10 + tpc * 4), id); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x088), id); id++; } - nv_wr32(priv, GPC_UNIT(gpc, 0x0c08), priv->tpc_nr[gpc]); - nv_wr32(priv, GPC_UNIT(gpc, 0x0c8c), priv->tpc_nr[gpc]); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0c08), gr->tpc_nr[gpc]); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0c8c), gr->tpc_nr[gpc]); } } } void -gf100_grctx_generate_r406028(struct gf100_gr_priv *priv) +gf100_grctx_generate_r406028(struct gf100_gr *gr) { + struct nvkm_device *device = gr->base.engine.subdev.device; u32 tmp[GPC_MAX / 8] = {}, i = 0; - for (i = 0; i < priv->gpc_nr; i++) - tmp[i / 8] |= priv->tpc_nr[i] << ((i % 8) * 4); + for (i = 0; i < gr->gpc_nr; i++) + tmp[i / 8] |= gr->tpc_nr[i] << ((i % 8) * 4); for (i = 0; i < 4; i++) { - nv_wr32(priv, 0x406028 + (i * 4), tmp[i]); - nv_wr32(priv, 0x405870 + (i * 4), tmp[i]); + nvkm_wr32(device, 0x406028 + (i * 4), tmp[i]); + nvkm_wr32(device, 0x405870 + (i * 4), tmp[i]); } } void -gf100_grctx_generate_r4060a8(struct gf100_gr_priv *priv) +gf100_grctx_generate_r4060a8(struct gf100_gr *gr) { + struct nvkm_device *device = gr->base.engine.subdev.device; u8 tpcnr[GPC_MAX], data[TPC_MAX]; int gpc, tpc, i; - memcpy(tpcnr, priv->tpc_nr, sizeof(priv->tpc_nr)); + memcpy(tpcnr, gr->tpc_nr, sizeof(gr->tpc_nr)); memset(data, 0x1f, sizeof(data)); gpc = -1; - for (tpc = 0; tpc < priv->tpc_total; tpc++) { + for (tpc = 0; tpc < gr->tpc_total; tpc++) { do { - gpc = (gpc + 1) % priv->gpc_nr; + gpc = (gpc + 1) % gr->gpc_nr; } while (!tpcnr[gpc]); tpcnr[gpc]--; data[tpc] = gpc; } for (i = 0; i < 4; i++) - nv_wr32(priv, 0x4060a8 + (i * 4), ((u32 *)data)[i]); + nvkm_wr32(device, 0x4060a8 + (i * 4), ((u32 *)data)[i]); } void -gf100_grctx_generate_r418bb8(struct gf100_gr_priv *priv) +gf100_grctx_generate_r418bb8(struct gf100_gr *gr) { + struct nvkm_device *device = gr->base.engine.subdev.device; u32 data[6] = {}, data2[2] = {}; u8 tpcnr[GPC_MAX]; u8 shift, ntpcv; int gpc, tpc, i; /* calculate first set of magics */ - memcpy(tpcnr, priv->tpc_nr, sizeof(priv->tpc_nr)); + memcpy(tpcnr, gr->tpc_nr, sizeof(gr->tpc_nr)); gpc = -1; - for (tpc = 0; tpc < priv->tpc_total; tpc++) { + for (tpc = 0; tpc < gr->tpc_total; tpc++) { do { - gpc = (gpc + 1) % priv->gpc_nr; + gpc = (gpc + 1) % gr->gpc_nr; } while (!tpcnr[gpc]); tpcnr[gpc]--; @@ -1163,7 +1167,7 @@ gf100_grctx_generate_r418bb8(struct gf100_gr_priv *priv) /* and the second... */ shift = 0; - ntpcv = priv->tpc_total; + ntpcv = gr->tpc_total; while (!(ntpcv & (1 << 4))) { ntpcv <<= 1; shift++; @@ -1176,202 +1180,211 @@ gf100_grctx_generate_r418bb8(struct gf100_gr_priv *priv) data2[1] |= ((1 << (i + 5)) % ntpcv) << ((i - 1) * 5); /* GPC_BROADCAST */ - nv_wr32(priv, 0x418bb8, (priv->tpc_total << 8) | - priv->magic_not_rop_nr); + nvkm_wr32(device, 0x418bb8, (gr->tpc_total << 8) | + gr->magic_not_rop_nr); for (i = 0; i < 6; i++) - nv_wr32(priv, 0x418b08 + (i * 4), data[i]); + nvkm_wr32(device, 0x418b08 + (i * 4), data[i]); /* GPC_BROADCAST.TP_BROADCAST */ - nv_wr32(priv, 0x419bd0, (priv->tpc_total << 8) | - priv->magic_not_rop_nr | data2[0]); - nv_wr32(priv, 0x419be4, data2[1]); + nvkm_wr32(device, 0x419bd0, (gr->tpc_total << 8) | + gr->magic_not_rop_nr | data2[0]); + nvkm_wr32(device, 0x419be4, data2[1]); for (i = 0; i < 6; i++) - nv_wr32(priv, 0x419b00 + (i * 4), data[i]); + nvkm_wr32(device, 0x419b00 + (i * 4), data[i]); /* UNK78xx */ - nv_wr32(priv, 0x4078bc, (priv->tpc_total << 8) | - priv->magic_not_rop_nr); + nvkm_wr32(device, 0x4078bc, (gr->tpc_total << 8) | + gr->magic_not_rop_nr); for (i = 0; i < 6; i++) - nv_wr32(priv, 0x40780c + (i * 4), data[i]); + nvkm_wr32(device, 0x40780c + (i * 4), data[i]); } void -gf100_grctx_generate_r406800(struct gf100_gr_priv *priv) +gf100_grctx_generate_r406800(struct gf100_gr *gr) { + struct nvkm_device *device = gr->base.engine.subdev.device; u64 tpc_mask = 0, tpc_set = 0; u8 tpcnr[GPC_MAX]; int gpc, tpc; int i, a, b; - memcpy(tpcnr, priv->tpc_nr, sizeof(priv->tpc_nr)); - for (gpc = 0; gpc < priv->gpc_nr; gpc++) - tpc_mask |= ((1ULL << priv->tpc_nr[gpc]) - 1) << (gpc * 8); + memcpy(tpcnr, gr->tpc_nr, sizeof(gr->tpc_nr)); + for (gpc = 0; gpc < gr->gpc_nr; gpc++) + tpc_mask |= ((1ULL << gr->tpc_nr[gpc]) - 1) << (gpc * 8); for (i = 0, gpc = -1, b = -1; i < 32; i++) { - a = (i * (priv->tpc_total - 1)) / 32; + a = (i * (gr->tpc_total - 1)) / 32; if (a != b) { b = a; do { - gpc = (gpc + 1) % priv->gpc_nr; + gpc = (gpc + 1) % gr->gpc_nr; } while (!tpcnr[gpc]); - tpc = priv->tpc_nr[gpc] - tpcnr[gpc]--; + tpc = gr->tpc_nr[gpc] - tpcnr[gpc]--; tpc_set |= 1ULL << ((gpc * 8) + tpc); } - nv_wr32(priv, 0x406800 + (i * 0x20), lower_32_bits(tpc_set)); - nv_wr32(priv, 0x406c00 + (i * 0x20), lower_32_bits(tpc_set ^ tpc_mask)); - if (priv->gpc_nr > 4) { - nv_wr32(priv, 0x406804 + (i * 0x20), upper_32_bits(tpc_set)); - nv_wr32(priv, 0x406c04 + (i * 0x20), upper_32_bits(tpc_set ^ tpc_mask)); + nvkm_wr32(device, 0x406800 + (i * 0x20), lower_32_bits(tpc_set)); + nvkm_wr32(device, 0x406c00 + (i * 0x20), lower_32_bits(tpc_set ^ tpc_mask)); + if (gr->gpc_nr > 4) { + nvkm_wr32(device, 0x406804 + (i * 0x20), upper_32_bits(tpc_set)); + nvkm_wr32(device, 0x406c04 + (i * 0x20), upper_32_bits(tpc_set ^ tpc_mask)); } } } void -gf100_grctx_generate_main(struct gf100_gr_priv *priv, struct gf100_grctx *info) +gf100_grctx_generate_main(struct gf100_gr *gr, struct gf100_grctx *info) { - struct gf100_grctx_oclass *oclass = (void *)nv_engine(priv)->cclass; - - nvkm_mc(priv)->unk260(nvkm_mc(priv), 0); - - gf100_gr_mmio(priv, oclass->hub); - gf100_gr_mmio(priv, oclass->gpc); - gf100_gr_mmio(priv, oclass->zcull); - gf100_gr_mmio(priv, oclass->tpc); - gf100_gr_mmio(priv, oclass->ppc); - - nv_wr32(priv, 0x404154, 0x00000000); - - oclass->bundle(info); - oclass->pagepool(info); - oclass->attrib(info); - oclass->unkn(priv); - - gf100_grctx_generate_tpcid(priv); - gf100_grctx_generate_r406028(priv); - gf100_grctx_generate_r4060a8(priv); - gf100_grctx_generate_r418bb8(priv); - gf100_grctx_generate_r406800(priv); - - gf100_gr_icmd(priv, oclass->icmd); - nv_wr32(priv, 0x404154, 0x00000400); - gf100_gr_mthd(priv, oclass->mthd); - nvkm_mc(priv)->unk260(nvkm_mc(priv), 1); + struct nvkm_device *device = gr->base.engine.subdev.device; + const struct gf100_grctx_func *grctx = gr->func->grctx; + + nvkm_mc_unk260(device->mc, 0); + + gf100_gr_mmio(gr, grctx->hub); + gf100_gr_mmio(gr, grctx->gpc); + gf100_gr_mmio(gr, grctx->zcull); + gf100_gr_mmio(gr, grctx->tpc); + gf100_gr_mmio(gr, grctx->ppc); + + nvkm_wr32(device, 0x404154, 0x00000000); + + grctx->bundle(info); + grctx->pagepool(info); + grctx->attrib(info); + grctx->unkn(gr); + + gf100_grctx_generate_tpcid(gr); + gf100_grctx_generate_r406028(gr); + gf100_grctx_generate_r4060a8(gr); + gf100_grctx_generate_r418bb8(gr); + gf100_grctx_generate_r406800(gr); + + gf100_gr_icmd(gr, grctx->icmd); + nvkm_wr32(device, 0x404154, 0x00000400); + gf100_gr_mthd(gr, grctx->mthd); + nvkm_mc_unk260(device->mc, 1); } int -gf100_grctx_generate(struct gf100_gr_priv *priv) +gf100_grctx_generate(struct gf100_gr *gr) { - struct gf100_grctx_oclass *oclass = (void *)nv_engine(priv)->cclass; - struct nvkm_bar *bar = nvkm_bar(priv); - struct nvkm_gpuobj *chan; + const struct gf100_grctx_func *grctx = gr->func->grctx; + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + struct nvkm_memory *chan; struct gf100_grctx info; int ret, i; + u64 addr; /* allocate memory to for a "channel", which we'll use to generate * the default context values */ - ret = nvkm_gpuobj_new(nv_object(priv), NULL, 0x80000 + priv->size, - 0x1000, NVOBJ_FLAG_ZERO_ALLOC, &chan); + ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, 0x80000 + gr->size, + 0x1000, true, &chan); if (ret) { - nv_error(priv, "failed to allocate channel memory, %d\n", ret); + nvkm_error(subdev, "failed to allocate chan memory, %d\n", ret); return ret; } + addr = nvkm_memory_addr(chan); + /* PGD pointer */ - nv_wo32(chan, 0x0200, lower_32_bits(chan->addr + 0x1000)); - nv_wo32(chan, 0x0204, upper_32_bits(chan->addr + 0x1000)); - nv_wo32(chan, 0x0208, 0xffffffff); - nv_wo32(chan, 0x020c, 0x000000ff); + nvkm_kmap(chan); + nvkm_wo32(chan, 0x0200, lower_32_bits(addr + 0x1000)); + nvkm_wo32(chan, 0x0204, upper_32_bits(addr + 0x1000)); + nvkm_wo32(chan, 0x0208, 0xffffffff); + nvkm_wo32(chan, 0x020c, 0x000000ff); /* PGT[0] pointer */ - nv_wo32(chan, 0x1000, 0x00000000); - nv_wo32(chan, 0x1004, 0x00000001 | (chan->addr + 0x2000) >> 8); + nvkm_wo32(chan, 0x1000, 0x00000000); + nvkm_wo32(chan, 0x1004, 0x00000001 | (addr + 0x2000) >> 8); /* identity-map the whole "channel" into its own vm */ - for (i = 0; i < chan->size / 4096; i++) { - u64 addr = ((chan->addr + (i * 4096)) >> 8) | 1; - nv_wo32(chan, 0x2000 + (i * 8), lower_32_bits(addr)); - nv_wo32(chan, 0x2004 + (i * 8), upper_32_bits(addr)); + for (i = 0; i < nvkm_memory_size(chan) / 4096; i++) { + u64 addr = ((nvkm_memory_addr(chan) + (i * 4096)) >> 8) | 1; + nvkm_wo32(chan, 0x2000 + (i * 8), lower_32_bits(addr)); + nvkm_wo32(chan, 0x2004 + (i * 8), upper_32_bits(addr)); } /* context pointer (virt) */ - nv_wo32(chan, 0x0210, 0x00080004); - nv_wo32(chan, 0x0214, 0x00000000); + nvkm_wo32(chan, 0x0210, 0x00080004); + nvkm_wo32(chan, 0x0214, 0x00000000); + nvkm_done(chan); - bar->flush(bar); - - nv_wr32(priv, 0x100cb8, (chan->addr + 0x1000) >> 8); - nv_wr32(priv, 0x100cbc, 0x80000001); - nv_wait(priv, 0x100c80, 0x00008000, 0x00008000); + nvkm_wr32(device, 0x100cb8, (addr + 0x1000) >> 8); + nvkm_wr32(device, 0x100cbc, 0x80000001); + nvkm_msec(device, 2000, + if (nvkm_rd32(device, 0x100c80) & 0x00008000) + break; + ); /* setup default state for mmio list construction */ - info.priv = priv; - info.data = priv->mmio_data; - info.mmio = priv->mmio_list; + info.gr = gr; + info.data = gr->mmio_data; + info.mmio = gr->mmio_list; info.addr = 0x2000 + (i * 8); info.buffer_nr = 0; /* make channel current */ - if (priv->firmware) { - nv_wr32(priv, 0x409840, 0x00000030); - nv_wr32(priv, 0x409500, 0x80000000 | chan->addr >> 12); - nv_wr32(priv, 0x409504, 0x00000003); - if (!nv_wait(priv, 0x409800, 0x00000010, 0x00000010)) - nv_error(priv, "load_ctx timeout\n"); - - nv_wo32(chan, 0x8001c, 1); - nv_wo32(chan, 0x80020, 0); - nv_wo32(chan, 0x80028, 0); - nv_wo32(chan, 0x8002c, 0); - bar->flush(bar); + if (gr->firmware) { + nvkm_wr32(device, 0x409840, 0x00000030); + nvkm_wr32(device, 0x409500, 0x80000000 | addr >> 12); + nvkm_wr32(device, 0x409504, 0x00000003); + nvkm_msec(device, 2000, + if (nvkm_rd32(device, 0x409800) & 0x00000010) + break; + ); + + nvkm_kmap(chan); + nvkm_wo32(chan, 0x8001c, 1); + nvkm_wo32(chan, 0x80020, 0); + nvkm_wo32(chan, 0x80028, 0); + nvkm_wo32(chan, 0x8002c, 0); + nvkm_done(chan); } else { - nv_wr32(priv, 0x409840, 0x80000000); - nv_wr32(priv, 0x409500, 0x80000000 | chan->addr >> 12); - nv_wr32(priv, 0x409504, 0x00000001); - if (!nv_wait(priv, 0x409800, 0x80000000, 0x80000000)) - nv_error(priv, "HUB_SET_CHAN timeout\n"); + nvkm_wr32(device, 0x409840, 0x80000000); + nvkm_wr32(device, 0x409500, 0x80000000 | addr >> 12); + nvkm_wr32(device, 0x409504, 0x00000001); + nvkm_msec(device, 2000, + if (nvkm_rd32(device, 0x409800) & 0x80000000) + break; + ); } - oclass->main(priv, &info); + grctx->main(gr, &info); /* trigger a context unload by unsetting the "next channel valid" bit * and faking a context switch interrupt */ - nv_mask(priv, 0x409b04, 0x80000000, 0x00000000); - nv_wr32(priv, 0x409000, 0x00000100); - if (!nv_wait(priv, 0x409b00, 0x80000000, 0x00000000)) { - nv_error(priv, "grctx template channel unload timeout\n"); + nvkm_mask(device, 0x409b04, 0x80000000, 0x00000000); + nvkm_wr32(device, 0x409000, 0x00000100); + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x409b00) & 0x80000000)) + break; + ) < 0) { ret = -EBUSY; goto done; } - priv->data = kmalloc(priv->size, GFP_KERNEL); - if (priv->data) { - for (i = 0; i < priv->size; i += 4) - priv->data[i / 4] = nv_ro32(chan, 0x80000 + i); + gr->data = kmalloc(gr->size, GFP_KERNEL); + if (gr->data) { + nvkm_kmap(chan); + for (i = 0; i < gr->size; i += 4) + gr->data[i / 4] = nvkm_ro32(chan, 0x80000 + i); + nvkm_done(chan); ret = 0; } else { ret = -ENOMEM; } done: - nvkm_gpuobj_ref(NULL, &chan); + nvkm_memory_del(&chan); return ret; } -struct nvkm_oclass * -gf100_grctx_oclass = &(struct gf100_grctx_oclass) { - .base.handle = NV_ENGCTX(GR, 0xc0), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gf100_gr_context_ctor, - .dtor = gf100_gr_context_dtor, - .init = _nvkm_gr_context_init, - .fini = _nvkm_gr_context_fini, - .rd32 = _nvkm_gr_context_rd32, - .wr32 = _nvkm_gr_context_wr32, - }, +const struct gf100_grctx_func +gf100_grctx = { .main = gf100_grctx_generate_main, .unkn = gf100_grctx_generate_unkn, .hub = gf100_grctx_pack_hub, @@ -1387,4 +1400,4 @@ gf100_grctx_oclass = &(struct gf100_grctx_oclass) { .attrib = gf100_grctx_generate_attrib, .attrib_nr_max = 0x324, .attrib_nr = 0x218, -}.base; +}; diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.h b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.h index 3676a3342..3c64040ec 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.h +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf100.h @@ -3,7 +3,7 @@ #include "gf100.h" struct gf100_grctx { - struct gf100_gr_priv *priv; + struct gf100_gr *gr; struct gf100_gr_data *data; struct gf100_gr_mmio *mmio; int buffer_nr; @@ -19,12 +19,11 @@ void gf100_grctx_mmio_item(struct gf100_grctx *, u32 addr, u32 data, int s, int) #define mmio_skip(a,b,c) mmio_refn((a), (b), (c), -1, -1) #define mmio_wr32(a,b,c) mmio_refn((a), (b), (c), 0, -1) -struct gf100_grctx_oclass { - struct nvkm_oclass base; +struct gf100_grctx_func { /* main context generation function */ - void (*main)(struct gf100_gr_priv *, struct gf100_grctx *); + void (*main)(struct gf100_gr *, struct gf100_grctx *); /* context-specific modify-on-first-load list generation function */ - void (*unkn)(struct gf100_gr_priv *); + void (*unkn)(struct gf100_gr *); /* mmio context data */ const struct gf100_gr_pack *hub; const struct gf100_gr_pack *gpc; @@ -50,60 +49,61 @@ struct gf100_grctx_oclass { u32 alpha_nr; }; -static inline const struct gf100_grctx_oclass * -gf100_grctx_impl(struct gf100_gr_priv *priv) -{ - return (void *)nv_engine(priv)->cclass; -} - -extern struct nvkm_oclass *gf100_grctx_oclass; -int gf100_grctx_generate(struct gf100_gr_priv *); -void gf100_grctx_generate_main(struct gf100_gr_priv *, struct gf100_grctx *); +extern const struct gf100_grctx_func gf100_grctx; +int gf100_grctx_generate(struct gf100_gr *); +void gf100_grctx_generate_main(struct gf100_gr *, struct gf100_grctx *); void gf100_grctx_generate_bundle(struct gf100_grctx *); void gf100_grctx_generate_pagepool(struct gf100_grctx *); void gf100_grctx_generate_attrib(struct gf100_grctx *); -void gf100_grctx_generate_unkn(struct gf100_gr_priv *); -void gf100_grctx_generate_tpcid(struct gf100_gr_priv *); -void gf100_grctx_generate_r406028(struct gf100_gr_priv *); -void gf100_grctx_generate_r4060a8(struct gf100_gr_priv *); -void gf100_grctx_generate_r418bb8(struct gf100_gr_priv *); -void gf100_grctx_generate_r406800(struct gf100_gr_priv *); - -extern struct nvkm_oclass *gf108_grctx_oclass; +void gf100_grctx_generate_unkn(struct gf100_gr *); +void gf100_grctx_generate_tpcid(struct gf100_gr *); +void gf100_grctx_generate_r406028(struct gf100_gr *); +void gf100_grctx_generate_r4060a8(struct gf100_gr *); +void gf100_grctx_generate_r418bb8(struct gf100_gr *); +void gf100_grctx_generate_r406800(struct gf100_gr *); + +extern const struct gf100_grctx_func gf108_grctx; void gf108_grctx_generate_attrib(struct gf100_grctx *); -void gf108_grctx_generate_unkn(struct gf100_gr_priv *); +void gf108_grctx_generate_unkn(struct gf100_gr *); -extern struct nvkm_oclass *gf104_grctx_oclass; -extern struct nvkm_oclass *gf110_grctx_oclass; +extern const struct gf100_grctx_func gf104_grctx; +extern const struct gf100_grctx_func gf110_grctx; -extern struct nvkm_oclass *gf117_grctx_oclass; +extern const struct gf100_grctx_func gf117_grctx; void gf117_grctx_generate_attrib(struct gf100_grctx *); -extern struct nvkm_oclass *gf119_grctx_oclass; +extern const struct gf100_grctx_func gf119_grctx; -extern struct nvkm_oclass *gk104_grctx_oclass; -extern struct nvkm_oclass *gk20a_grctx_oclass; -void gk104_grctx_generate_main(struct gf100_gr_priv *, struct gf100_grctx *); +extern const struct gf100_grctx_func gk104_grctx; +extern const struct gf100_grctx_func gk20a_grctx; +void gk104_grctx_generate_main(struct gf100_gr *, struct gf100_grctx *); void gk104_grctx_generate_bundle(struct gf100_grctx *); void gk104_grctx_generate_pagepool(struct gf100_grctx *); -void gk104_grctx_generate_unkn(struct gf100_gr_priv *); -void gk104_grctx_generate_r418bb8(struct gf100_gr_priv *); -void gk104_grctx_generate_rop_active_fbps(struct gf100_gr_priv *); +void gk104_grctx_generate_unkn(struct gf100_gr *); +void gk104_grctx_generate_r418bb8(struct gf100_gr *); +void gk104_grctx_generate_rop_active_fbps(struct gf100_gr *); + +void gm107_grctx_generate_bundle(struct gf100_grctx *); +void gm107_grctx_generate_pagepool(struct gf100_grctx *); +void gm107_grctx_generate_attrib(struct gf100_grctx *); -extern struct nvkm_oclass *gk110_grctx_oclass; -extern struct nvkm_oclass *gk110b_grctx_oclass; -extern struct nvkm_oclass *gk208_grctx_oclass; +extern const struct gf100_grctx_func gk110_grctx; +extern const struct gf100_grctx_func gk110b_grctx; +extern const struct gf100_grctx_func gk208_grctx; -extern struct nvkm_oclass *gm107_grctx_oclass; +extern const struct gf100_grctx_func gm107_grctx; void gm107_grctx_generate_bundle(struct gf100_grctx *); void gm107_grctx_generate_pagepool(struct gf100_grctx *); void gm107_grctx_generate_attrib(struct gf100_grctx *); -extern struct nvkm_oclass *gm204_grctx_oclass; -void gm204_grctx_generate_main(struct gf100_gr_priv *, struct gf100_grctx *); +extern const struct gf100_grctx_func gm204_grctx; +void gm204_grctx_generate_main(struct gf100_gr *, struct gf100_grctx *); +void gm204_grctx_generate_tpcid(struct gf100_gr *); +void gm204_grctx_generate_405b60(struct gf100_gr *); -extern struct nvkm_oclass *gm206_grctx_oclass; +extern const struct gf100_grctx_func gm206_grctx; +extern const struct gf100_grctx_func gm20b_grctx; /* context init value lists */ diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf104.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf104.c index c5a8d55e2..54fd74e9c 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf104.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf104.c @@ -79,17 +79,8 @@ gf104_grctx_pack_tpc[] = { * PGRAPH context implementation ******************************************************************************/ -struct nvkm_oclass * -gf104_grctx_oclass = &(struct gf100_grctx_oclass) { - .base.handle = NV_ENGCTX(GR, 0xc3), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gf100_gr_context_ctor, - .dtor = gf100_gr_context_dtor, - .init = _nvkm_gr_context_init, - .fini = _nvkm_gr_context_fini, - .rd32 = _nvkm_gr_context_rd32, - .wr32 = _nvkm_gr_context_wr32, - }, +const struct gf100_grctx_func +gf104_grctx = { .main = gf100_grctx_generate_main, .unkn = gf100_grctx_generate_unkn, .hub = gf100_grctx_pack_hub, @@ -105,4 +96,4 @@ gf104_grctx_oclass = &(struct gf100_grctx_oclass) { .attrib = gf100_grctx_generate_attrib, .attrib_nr_max = 0x324, .attrib_nr = 0x218, -}.base; +}; diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf108.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf108.c index 87c844a5f..505cdcbfc 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf108.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf108.c @@ -730,18 +730,18 @@ gf108_grctx_pack_tpc[] = { void gf108_grctx_generate_attrib(struct gf100_grctx *info) { - struct gf100_gr_priv *priv = info->priv; - const struct gf100_grctx_oclass *impl = gf100_grctx_impl(priv); - const u32 alpha = impl->alpha_nr; - const u32 beta = impl->attrib_nr; - const u32 size = 0x20 * (impl->attrib_nr_max + impl->alpha_nr_max); + struct gf100_gr *gr = info->gr; + const struct gf100_grctx_func *grctx = gr->func->grctx; + const u32 alpha = grctx->alpha_nr; + const u32 beta = grctx->attrib_nr; + const u32 size = 0x20 * (grctx->attrib_nr_max + grctx->alpha_nr_max); const u32 access = NV_MEM_ACCESS_RW; const int s = 12; - const int b = mmio_vram(info, size * priv->tpc_total, (1 << s), access); + const int b = mmio_vram(info, size * gr->tpc_total, (1 << s), access); const int timeslice_mode = 1; const int max_batches = 0xffff; u32 bo = 0; - u32 ao = bo + impl->attrib_nr_max * priv->tpc_total; + u32 ao = bo + grctx->attrib_nr_max * gr->tpc_total; int gpc, tpc; mmio_refn(info, 0x418810, 0x80000000, s, b); @@ -749,43 +749,35 @@ gf108_grctx_generate_attrib(struct gf100_grctx *info) mmio_wr32(info, 0x405830, (beta << 16) | alpha); mmio_wr32(info, 0x4064c4, ((alpha / 4) << 16) | max_batches); - for (gpc = 0; gpc < priv->gpc_nr; gpc++) { - for (tpc = 0; tpc < priv->tpc_nr[gpc]; tpc++) { + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + for (tpc = 0; tpc < gr->tpc_nr[gpc]; tpc++) { const u32 a = alpha; const u32 b = beta; const u32 t = timeslice_mode; const u32 o = TPC_UNIT(gpc, tpc, 0x500); mmio_skip(info, o + 0x20, (t << 28) | (b << 16) | ++bo); mmio_wr32(info, o + 0x20, (t << 28) | (b << 16) | --bo); - bo += impl->attrib_nr_max; + bo += grctx->attrib_nr_max; mmio_wr32(info, o + 0x44, (a << 16) | ao); - ao += impl->alpha_nr_max; + ao += grctx->alpha_nr_max; } } } void -gf108_grctx_generate_unkn(struct gf100_gr_priv *priv) +gf108_grctx_generate_unkn(struct gf100_gr *gr) { - nv_mask(priv, 0x418c6c, 0x00000001, 0x00000001); - nv_mask(priv, 0x41980c, 0x00000010, 0x00000010); - nv_mask(priv, 0x419814, 0x00000004, 0x00000004); - nv_mask(priv, 0x4064c0, 0x80000000, 0x80000000); - nv_mask(priv, 0x405800, 0x08000000, 0x08000000); - nv_mask(priv, 0x419c00, 0x00000008, 0x00000008); + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_mask(device, 0x418c6c, 0x00000001, 0x00000001); + nvkm_mask(device, 0x41980c, 0x00000010, 0x00000010); + nvkm_mask(device, 0x419814, 0x00000004, 0x00000004); + nvkm_mask(device, 0x4064c0, 0x80000000, 0x80000000); + nvkm_mask(device, 0x405800, 0x08000000, 0x08000000); + nvkm_mask(device, 0x419c00, 0x00000008, 0x00000008); } -struct nvkm_oclass * -gf108_grctx_oclass = &(struct gf100_grctx_oclass) { - .base.handle = NV_ENGCTX(GR, 0xc1), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gf100_gr_context_ctor, - .dtor = gf100_gr_context_dtor, - .init = _nvkm_gr_context_init, - .fini = _nvkm_gr_context_fini, - .rd32 = _nvkm_gr_context_rd32, - .wr32 = _nvkm_gr_context_wr32, - }, +const struct gf100_grctx_func +gf108_grctx = { .main = gf100_grctx_generate_main, .unkn = gf108_grctx_generate_unkn, .hub = gf108_grctx_pack_hub, @@ -803,4 +795,4 @@ gf108_grctx_oclass = &(struct gf100_grctx_oclass) { .attrib_nr = 0x218, .alpha_nr_max = 0x324, .alpha_nr = 0x218, -}.base; +}; diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf110.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf110.c index b3acd931b..7df398b53 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf110.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf110.c @@ -330,17 +330,8 @@ gf110_grctx_pack_gpc[] = { * PGRAPH context implementation ******************************************************************************/ -struct nvkm_oclass * -gf110_grctx_oclass = &(struct gf100_grctx_oclass) { - .base.handle = NV_ENGCTX(GR, 0xc8), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gf100_gr_context_ctor, - .dtor = gf100_gr_context_dtor, - .init = _nvkm_gr_context_init, - .fini = _nvkm_gr_context_fini, - .rd32 = _nvkm_gr_context_rd32, - .wr32 = _nvkm_gr_context_wr32, - }, +const struct gf100_grctx_func +gf110_grctx = { .main = gf100_grctx_generate_main, .unkn = gf100_grctx_generate_unkn, .hub = gf100_grctx_pack_hub, @@ -356,4 +347,4 @@ gf110_grctx_oclass = &(struct gf100_grctx_oclass) { .attrib = gf100_grctx_generate_attrib, .attrib_nr_max = 0x324, .attrib_nr = 0x218, -}.base; +}; diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf117.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf117.c index 9bbe2c975..74de7a96c 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf117.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf117.c @@ -182,18 +182,18 @@ gf117_grctx_pack_ppc[] = { void gf117_grctx_generate_attrib(struct gf100_grctx *info) { - struct gf100_gr_priv *priv = info->priv; - const struct gf100_grctx_oclass *impl = gf100_grctx_impl(priv); - const u32 alpha = impl->alpha_nr; - const u32 beta = impl->attrib_nr; - const u32 size = 0x20 * (impl->attrib_nr_max + impl->alpha_nr_max); + struct gf100_gr *gr = info->gr; + const struct gf100_grctx_func *grctx = gr->func->grctx; + const u32 alpha = grctx->alpha_nr; + const u32 beta = grctx->attrib_nr; + const u32 size = 0x20 * (grctx->attrib_nr_max + grctx->alpha_nr_max); const u32 access = NV_MEM_ACCESS_RW; const int s = 12; - const int b = mmio_vram(info, size * priv->tpc_total, (1 << s), access); + const int b = mmio_vram(info, size * gr->tpc_total, (1 << s), access); const int timeslice_mode = 1; const int max_batches = 0xffff; u32 bo = 0; - u32 ao = bo + impl->attrib_nr_max * priv->tpc_total; + u32 ao = bo + grctx->attrib_nr_max * gr->tpc_total; int gpc, ppc; mmio_refn(info, 0x418810, 0x80000000, s, b); @@ -201,68 +201,62 @@ gf117_grctx_generate_attrib(struct gf100_grctx *info) mmio_wr32(info, 0x405830, (beta << 16) | alpha); mmio_wr32(info, 0x4064c4, ((alpha / 4) << 16) | max_batches); - for (gpc = 0; gpc < priv->gpc_nr; gpc++) { - for (ppc = 0; ppc < priv->ppc_nr[gpc]; ppc++) { - const u32 a = alpha * priv->ppc_tpc_nr[gpc][ppc]; - const u32 b = beta * priv->ppc_tpc_nr[gpc][ppc]; + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + for (ppc = 0; ppc < gr->ppc_nr[gpc]; ppc++) { + const u32 a = alpha * gr->ppc_tpc_nr[gpc][ppc]; + const u32 b = beta * gr->ppc_tpc_nr[gpc][ppc]; const u32 t = timeslice_mode; const u32 o = PPC_UNIT(gpc, ppc, 0); + if (!(gr->ppc_mask[gpc] & (1 << ppc))) + continue; mmio_skip(info, o + 0xc0, (t << 28) | (b << 16) | ++bo); mmio_wr32(info, o + 0xc0, (t << 28) | (b << 16) | --bo); - bo += impl->attrib_nr_max * priv->ppc_tpc_nr[gpc][ppc]; + bo += grctx->attrib_nr_max * gr->ppc_tpc_nr[gpc][ppc]; mmio_wr32(info, o + 0xe4, (a << 16) | ao); - ao += impl->alpha_nr_max * priv->ppc_tpc_nr[gpc][ppc]; + ao += grctx->alpha_nr_max * gr->ppc_tpc_nr[gpc][ppc]; } } } void -gf117_grctx_generate_main(struct gf100_gr_priv *priv, struct gf100_grctx *info) +gf117_grctx_generate_main(struct gf100_gr *gr, struct gf100_grctx *info) { - struct gf100_grctx_oclass *oclass = (void *)nv_engine(priv)->cclass; + struct nvkm_device *device = gr->base.engine.subdev.device; + const struct gf100_grctx_func *grctx = gr->func->grctx; int i; - nvkm_mc(priv)->unk260(nvkm_mc(priv), 0); + nvkm_mc_unk260(device->mc, 0); - gf100_gr_mmio(priv, oclass->hub); - gf100_gr_mmio(priv, oclass->gpc); - gf100_gr_mmio(priv, oclass->zcull); - gf100_gr_mmio(priv, oclass->tpc); - gf100_gr_mmio(priv, oclass->ppc); + gf100_gr_mmio(gr, grctx->hub); + gf100_gr_mmio(gr, grctx->gpc); + gf100_gr_mmio(gr, grctx->zcull); + gf100_gr_mmio(gr, grctx->tpc); + gf100_gr_mmio(gr, grctx->ppc); - nv_wr32(priv, 0x404154, 0x00000000); + nvkm_wr32(device, 0x404154, 0x00000000); - oclass->bundle(info); - oclass->pagepool(info); - oclass->attrib(info); - oclass->unkn(priv); + grctx->bundle(info); + grctx->pagepool(info); + grctx->attrib(info); + grctx->unkn(gr); - gf100_grctx_generate_tpcid(priv); - gf100_grctx_generate_r406028(priv); - gf100_grctx_generate_r4060a8(priv); - gk104_grctx_generate_r418bb8(priv); - gf100_grctx_generate_r406800(priv); + gf100_grctx_generate_tpcid(gr); + gf100_grctx_generate_r406028(gr); + gf100_grctx_generate_r4060a8(gr); + gk104_grctx_generate_r418bb8(gr); + gf100_grctx_generate_r406800(gr); for (i = 0; i < 8; i++) - nv_wr32(priv, 0x4064d0 + (i * 0x04), 0x00000000); + nvkm_wr32(device, 0x4064d0 + (i * 0x04), 0x00000000); - gf100_gr_icmd(priv, oclass->icmd); - nv_wr32(priv, 0x404154, 0x00000400); - gf100_gr_mthd(priv, oclass->mthd); - nvkm_mc(priv)->unk260(nvkm_mc(priv), 1); + gf100_gr_icmd(gr, grctx->icmd); + nvkm_wr32(device, 0x404154, 0x00000400); + gf100_gr_mthd(gr, grctx->mthd); + nvkm_mc_unk260(device->mc, 1); } -struct nvkm_oclass * -gf117_grctx_oclass = &(struct gf100_grctx_oclass) { - .base.handle = NV_ENGCTX(GR, 0xd7), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gf100_gr_context_ctor, - .dtor = gf100_gr_context_dtor, - .init = _nvkm_gr_context_init, - .fini = _nvkm_gr_context_fini, - .rd32 = _nvkm_gr_context_rd32, - .wr32 = _nvkm_gr_context_wr32, - }, +const struct gf100_grctx_func +gf117_grctx = { .main = gf117_grctx_generate_main, .unkn = gk104_grctx_generate_unkn, .hub = gf117_grctx_pack_hub, @@ -281,4 +275,4 @@ gf117_grctx_oclass = &(struct gf100_grctx_oclass) { .attrib_nr = 0x218, .alpha_nr_max = 0x7ff, .alpha_nr = 0x324, -}.base; +}; diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf119.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf119.c index 8d8761443..605185b07 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf119.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgf119.c @@ -498,17 +498,8 @@ gf119_grctx_pack_tpc[] = { * PGRAPH context implementation ******************************************************************************/ -struct nvkm_oclass * -gf119_grctx_oclass = &(struct gf100_grctx_oclass) { - .base.handle = NV_ENGCTX(GR, 0xd9), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gf100_gr_context_ctor, - .dtor = gf100_gr_context_dtor, - .init = _nvkm_gr_context_init, - .fini = _nvkm_gr_context_fini, - .rd32 = _nvkm_gr_context_rd32, - .wr32 = _nvkm_gr_context_wr32, - }, +const struct gf100_grctx_func +gf119_grctx = { .main = gf100_grctx_generate_main, .unkn = gf108_grctx_generate_unkn, .hub = gf119_grctx_pack_hub, @@ -526,4 +517,4 @@ gf119_grctx_oclass = &(struct gf100_grctx_oclass) { .attrib_nr = 0x218, .alpha_nr_max = 0x324, .alpha_nr = 0x218, -}.base; +}; diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk104.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk104.c index b12f6a9fd..a843e3689 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk104.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk104.c @@ -843,27 +843,27 @@ gk104_grctx_pack_ppc[] = { void gk104_grctx_generate_bundle(struct gf100_grctx *info) { - const struct gf100_grctx_oclass *impl = gf100_grctx_impl(info->priv); - const u32 state_limit = min(impl->bundle_min_gpm_fifo_depth, - impl->bundle_size / 0x20); - const u32 token_limit = impl->bundle_token_limit; + const struct gf100_grctx_func *grctx = info->gr->func->grctx; + const u32 state_limit = min(grctx->bundle_min_gpm_fifo_depth, + grctx->bundle_size / 0x20); + const u32 token_limit = grctx->bundle_token_limit; const u32 access = NV_MEM_ACCESS_RW | NV_MEM_ACCESS_SYS; const int s = 8; - const int b = mmio_vram(info, impl->bundle_size, (1 << s), access); + const int b = mmio_vram(info, grctx->bundle_size, (1 << s), access); mmio_refn(info, 0x408004, 0x00000000, s, b); - mmio_wr32(info, 0x408008, 0x80000000 | (impl->bundle_size >> s)); + mmio_wr32(info, 0x408008, 0x80000000 | (grctx->bundle_size >> s)); mmio_refn(info, 0x418808, 0x00000000, s, b); - mmio_wr32(info, 0x41880c, 0x80000000 | (impl->bundle_size >> s)); + mmio_wr32(info, 0x41880c, 0x80000000 | (grctx->bundle_size >> s)); mmio_wr32(info, 0x4064c8, (state_limit << 16) | token_limit); } void gk104_grctx_generate_pagepool(struct gf100_grctx *info) { - const struct gf100_grctx_oclass *impl = gf100_grctx_impl(info->priv); + const struct gf100_grctx_func *grctx = info->gr->func->grctx; const u32 access = NV_MEM_ACCESS_RW | NV_MEM_ACCESS_SYS; const int s = 8; - const int b = mmio_vram(info, impl->pagepool_size, (1 << s), access); + const int b = mmio_vram(info, grctx->pagepool_size, (1 << s), access); mmio_refn(info, 0x40800c, 0x00000000, s, b); mmio_wr32(info, 0x408010, 0x80000000); mmio_refn(info, 0x419004, 0x00000000, s, b); @@ -872,31 +872,33 @@ gk104_grctx_generate_pagepool(struct gf100_grctx *info) } void -gk104_grctx_generate_unkn(struct gf100_gr_priv *priv) +gk104_grctx_generate_unkn(struct gf100_gr *gr) { - nv_mask(priv, 0x418c6c, 0x00000001, 0x00000001); - nv_mask(priv, 0x41980c, 0x00000010, 0x00000010); - nv_mask(priv, 0x41be08, 0x00000004, 0x00000004); - nv_mask(priv, 0x4064c0, 0x80000000, 0x80000000); - nv_mask(priv, 0x405800, 0x08000000, 0x08000000); - nv_mask(priv, 0x419c00, 0x00000008, 0x00000008); + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_mask(device, 0x418c6c, 0x00000001, 0x00000001); + nvkm_mask(device, 0x41980c, 0x00000010, 0x00000010); + nvkm_mask(device, 0x41be08, 0x00000004, 0x00000004); + nvkm_mask(device, 0x4064c0, 0x80000000, 0x80000000); + nvkm_mask(device, 0x405800, 0x08000000, 0x08000000); + nvkm_mask(device, 0x419c00, 0x00000008, 0x00000008); } void -gk104_grctx_generate_r418bb8(struct gf100_gr_priv *priv) +gk104_grctx_generate_r418bb8(struct gf100_gr *gr) { + struct nvkm_device *device = gr->base.engine.subdev.device; u32 data[6] = {}, data2[2] = {}; u8 tpcnr[GPC_MAX]; u8 shift, ntpcv; int gpc, tpc, i; /* calculate first set of magics */ - memcpy(tpcnr, priv->tpc_nr, sizeof(priv->tpc_nr)); + memcpy(tpcnr, gr->tpc_nr, sizeof(gr->tpc_nr)); gpc = -1; - for (tpc = 0; tpc < priv->tpc_total; tpc++) { + for (tpc = 0; tpc < gr->tpc_total; tpc++) { do { - gpc = (gpc + 1) % priv->gpc_nr; + gpc = (gpc + 1) % gr->gpc_nr; } while (!tpcnr[gpc]); tpcnr[gpc]--; @@ -908,7 +910,7 @@ gk104_grctx_generate_r418bb8(struct gf100_gr_priv *priv) /* and the second... */ shift = 0; - ntpcv = priv->tpc_total; + ntpcv = gr->tpc_total; while (!(ntpcv & (1 << 4))) { ntpcv <<= 1; shift++; @@ -921,86 +923,79 @@ gk104_grctx_generate_r418bb8(struct gf100_gr_priv *priv) data2[1] |= ((1 << (i + 5)) % ntpcv) << ((i - 1) * 5); /* GPC_BROADCAST */ - nv_wr32(priv, 0x418bb8, (priv->tpc_total << 8) | - priv->magic_not_rop_nr); + nvkm_wr32(device, 0x418bb8, (gr->tpc_total << 8) | + gr->magic_not_rop_nr); for (i = 0; i < 6; i++) - nv_wr32(priv, 0x418b08 + (i * 4), data[i]); + nvkm_wr32(device, 0x418b08 + (i * 4), data[i]); /* GPC_BROADCAST.TP_BROADCAST */ - nv_wr32(priv, 0x41bfd0, (priv->tpc_total << 8) | - priv->magic_not_rop_nr | data2[0]); - nv_wr32(priv, 0x41bfe4, data2[1]); + nvkm_wr32(device, 0x41bfd0, (gr->tpc_total << 8) | + gr->magic_not_rop_nr | data2[0]); + nvkm_wr32(device, 0x41bfe4, data2[1]); for (i = 0; i < 6; i++) - nv_wr32(priv, 0x41bf00 + (i * 4), data[i]); + nvkm_wr32(device, 0x41bf00 + (i * 4), data[i]); /* UNK78xx */ - nv_wr32(priv, 0x4078bc, (priv->tpc_total << 8) | - priv->magic_not_rop_nr); + nvkm_wr32(device, 0x4078bc, (gr->tpc_total << 8) | + gr->magic_not_rop_nr); for (i = 0; i < 6; i++) - nv_wr32(priv, 0x40780c + (i * 4), data[i]); + nvkm_wr32(device, 0x40780c + (i * 4), data[i]); } void -gk104_grctx_generate_rop_active_fbps(struct gf100_gr_priv *priv) +gk104_grctx_generate_rop_active_fbps(struct gf100_gr *gr) { - const u32 fbp_count = nv_rd32(priv, 0x120074); - nv_mask(priv, 0x408850, 0x0000000f, fbp_count); /* zrop */ - nv_mask(priv, 0x408958, 0x0000000f, fbp_count); /* crop */ + struct nvkm_device *device = gr->base.engine.subdev.device; + const u32 fbp_count = nvkm_rd32(device, 0x120074); + nvkm_mask(device, 0x408850, 0x0000000f, fbp_count); /* zrop */ + nvkm_mask(device, 0x408958, 0x0000000f, fbp_count); /* crop */ } void -gk104_grctx_generate_main(struct gf100_gr_priv *priv, struct gf100_grctx *info) +gk104_grctx_generate_main(struct gf100_gr *gr, struct gf100_grctx *info) { - struct gf100_grctx_oclass *oclass = (void *)nv_engine(priv)->cclass; + struct nvkm_device *device = gr->base.engine.subdev.device; + const struct gf100_grctx_func *grctx = gr->func->grctx; int i; - nvkm_mc(priv)->unk260(nvkm_mc(priv), 0); + nvkm_mc_unk260(device->mc, 0); - gf100_gr_mmio(priv, oclass->hub); - gf100_gr_mmio(priv, oclass->gpc); - gf100_gr_mmio(priv, oclass->zcull); - gf100_gr_mmio(priv, oclass->tpc); - gf100_gr_mmio(priv, oclass->ppc); + gf100_gr_mmio(gr, grctx->hub); + gf100_gr_mmio(gr, grctx->gpc); + gf100_gr_mmio(gr, grctx->zcull); + gf100_gr_mmio(gr, grctx->tpc); + gf100_gr_mmio(gr, grctx->ppc); - nv_wr32(priv, 0x404154, 0x00000000); + nvkm_wr32(device, 0x404154, 0x00000000); - oclass->bundle(info); - oclass->pagepool(info); - oclass->attrib(info); - oclass->unkn(priv); + grctx->bundle(info); + grctx->pagepool(info); + grctx->attrib(info); + grctx->unkn(gr); - gf100_grctx_generate_tpcid(priv); - gf100_grctx_generate_r406028(priv); - gk104_grctx_generate_r418bb8(priv); - gf100_grctx_generate_r406800(priv); + gf100_grctx_generate_tpcid(gr); + gf100_grctx_generate_r406028(gr); + gk104_grctx_generate_r418bb8(gr); + gf100_grctx_generate_r406800(gr); for (i = 0; i < 8; i++) - nv_wr32(priv, 0x4064d0 + (i * 0x04), 0x00000000); + nvkm_wr32(device, 0x4064d0 + (i * 0x04), 0x00000000); - nv_wr32(priv, 0x405b00, (priv->tpc_total << 8) | priv->gpc_nr); - gk104_grctx_generate_rop_active_fbps(priv); - nv_mask(priv, 0x419f78, 0x00000001, 0x00000000); + nvkm_wr32(device, 0x405b00, (gr->tpc_total << 8) | gr->gpc_nr); + gk104_grctx_generate_rop_active_fbps(gr); + nvkm_mask(device, 0x419f78, 0x00000001, 0x00000000); - gf100_gr_icmd(priv, oclass->icmd); - nv_wr32(priv, 0x404154, 0x00000400); - gf100_gr_mthd(priv, oclass->mthd); - nvkm_mc(priv)->unk260(nvkm_mc(priv), 1); + gf100_gr_icmd(gr, grctx->icmd); + nvkm_wr32(device, 0x404154, 0x00000400); + gf100_gr_mthd(gr, grctx->mthd); + nvkm_mc_unk260(device->mc, 1); - nv_mask(priv, 0x418800, 0x00200000, 0x00200000); - nv_mask(priv, 0x41be10, 0x00800000, 0x00800000); + nvkm_mask(device, 0x418800, 0x00200000, 0x00200000); + nvkm_mask(device, 0x41be10, 0x00800000, 0x00800000); } -struct nvkm_oclass * -gk104_grctx_oclass = &(struct gf100_grctx_oclass) { - .base.handle = NV_ENGCTX(GR, 0xe4), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gf100_gr_context_ctor, - .dtor = gf100_gr_context_dtor, - .init = _nvkm_gr_context_init, - .fini = _nvkm_gr_context_fini, - .rd32 = _nvkm_gr_context_rd32, - .wr32 = _nvkm_gr_context_wr32, - }, +const struct gf100_grctx_func +gk104_grctx = { .main = gk104_grctx_generate_main, .unkn = gk104_grctx_generate_unkn, .hub = gk104_grctx_pack_hub, @@ -1021,4 +1016,4 @@ gk104_grctx_oclass = &(struct gf100_grctx_oclass) { .attrib_nr = 0x218, .alpha_nr_max = 0x7ff, .alpha_nr = 0x648, -}.base; +}; diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk110.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk110.c index b3f58be04..7b95ec2fe 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk110.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk110.c @@ -808,17 +808,8 @@ gk110_grctx_pack_ppc[] = { * PGRAPH context implementation ******************************************************************************/ -struct nvkm_oclass * -gk110_grctx_oclass = &(struct gf100_grctx_oclass) { - .base.handle = NV_ENGCTX(GR, 0xf0), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gf100_gr_context_ctor, - .dtor = gf100_gr_context_dtor, - .init = _nvkm_gr_context_init, - .fini = _nvkm_gr_context_fini, - .rd32 = _nvkm_gr_context_rd32, - .wr32 = _nvkm_gr_context_wr32, - }, +const struct gf100_grctx_func +gk110_grctx = { .main = gk104_grctx_generate_main, .unkn = gk104_grctx_generate_unkn, .hub = gk110_grctx_pack_hub, @@ -839,4 +830,4 @@ gk110_grctx_oclass = &(struct gf100_grctx_oclass) { .attrib_nr = 0x218, .alpha_nr_max = 0x7ff, .alpha_nr = 0x648, -}.base; +}; diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk110b.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk110b.c index b11c26794..048b1152d 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk110b.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk110b.c @@ -69,17 +69,8 @@ gk110b_grctx_pack_tpc[] = { * PGRAPH context implementation ******************************************************************************/ -struct nvkm_oclass * -gk110b_grctx_oclass = &(struct gf100_grctx_oclass) { - .base.handle = NV_ENGCTX(GR, 0xf1), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gf100_gr_context_ctor, - .dtor = gf100_gr_context_dtor, - .init = _nvkm_gr_context_init, - .fini = _nvkm_gr_context_fini, - .rd32 = _nvkm_gr_context_rd32, - .wr32 = _nvkm_gr_context_wr32, - }, +const struct gf100_grctx_func +gk110b_grctx = { .main = gk104_grctx_generate_main, .unkn = gk104_grctx_generate_unkn, .hub = gk110_grctx_pack_hub, @@ -100,4 +91,4 @@ gk110b_grctx_oclass = &(struct gf100_grctx_oclass) { .attrib_nr = 0x218, .alpha_nr_max = 0x7ff, .alpha_nr = 0x648, -}.base; +}; diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk208.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk208.c index 6e8ce9fc3..67b7a1b43 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk208.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk208.c @@ -530,17 +530,8 @@ gk208_grctx_pack_ppc[] = { * PGRAPH context implementation ******************************************************************************/ -struct nvkm_oclass * -gk208_grctx_oclass = &(struct gf100_grctx_oclass) { - .base.handle = NV_ENGCTX(GR, 0x08), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gf100_gr_context_ctor, - .dtor = gf100_gr_context_dtor, - .init = _nvkm_gr_context_init, - .fini = _nvkm_gr_context_fini, - .rd32 = _nvkm_gr_context_rd32, - .wr32 = _nvkm_gr_context_wr32, - }, +const struct gf100_grctx_func +gk208_grctx = { .main = gk104_grctx_generate_main, .unkn = gk104_grctx_generate_unkn, .hub = gk208_grctx_pack_hub, @@ -561,4 +552,4 @@ gk208_grctx_oclass = &(struct gf100_grctx_oclass) { .attrib_nr = 0x218, .alpha_nr_max = 0x7ff, .alpha_nr = 0x648, -}.base; +}; diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk20a.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk20a.c index 2f241f6f0..ddaa16a71 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk20a.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgk20a.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2014, NVIDIA CORPORATION. All rights reserved. + * Copyright (c) 2014-2015, NVIDIA CORPORATION. All rights reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -20,34 +20,60 @@ * DEALINGS IN THE SOFTWARE. */ #include "ctxgf100.h" +#include "gf100.h" -static const struct gf100_gr_pack -gk20a_grctx_pack_mthd[] = { - { gk104_grctx_init_a097_0, 0xa297 }, - { gf100_grctx_init_902d_0, 0x902d }, - {} -}; +#include <subdev/mc.h> + +static void +gk20a_grctx_generate_main(struct gf100_gr *gr, struct gf100_grctx *info) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + const struct gf100_grctx_func *grctx = gr->func->grctx; + int idle_timeout_save; + int i; + + gf100_gr_mmio(gr, gr->fuc_sw_ctx); + + gf100_gr_wait_idle(gr); + + idle_timeout_save = nvkm_rd32(device, 0x404154); + nvkm_wr32(device, 0x404154, 0x00000000); + + grctx->attrib(info); + + grctx->unkn(gr); + + gf100_grctx_generate_tpcid(gr); + gf100_grctx_generate_r406028(gr); + gk104_grctx_generate_r418bb8(gr); + gf100_grctx_generate_r406800(gr); + + for (i = 0; i < 8; i++) + nvkm_wr32(device, 0x4064d0 + (i * 0x04), 0x00000000); -struct nvkm_oclass * -gk20a_grctx_oclass = &(struct gf100_grctx_oclass) { - .base.handle = NV_ENGCTX(GR, 0xea), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gf100_gr_context_ctor, - .dtor = gf100_gr_context_dtor, - .init = _nvkm_gr_context_init, - .fini = _nvkm_gr_context_fini, - .rd32 = _nvkm_gr_context_rd32, - .wr32 = _nvkm_gr_context_wr32, - }, - .main = gk104_grctx_generate_main, + nvkm_wr32(device, 0x405b00, (gr->tpc_total << 8) | gr->gpc_nr); + + gk104_grctx_generate_rop_active_fbps(gr); + + nvkm_mask(device, 0x5044b0, 0x8000000, 0x8000000); + + gf100_gr_wait_idle(gr); + + nvkm_wr32(device, 0x404154, idle_timeout_save); + gf100_gr_wait_idle(gr); + + gf100_gr_mthd(gr, gr->fuc_method); + gf100_gr_wait_idle(gr); + + gf100_gr_icmd(gr, gr->fuc_bundle); + grctx->pagepool(info); + grctx->bundle(info); +} + +const struct gf100_grctx_func +gk20a_grctx = { + .main = gk20a_grctx_generate_main, .unkn = gk104_grctx_generate_unkn, - .hub = gk104_grctx_pack_hub, - .gpc = gk104_grctx_pack_gpc, - .zcull = gf100_grctx_pack_zcull, - .tpc = gk104_grctx_pack_tpc, - .ppc = gk104_grctx_pack_ppc, - .icmd = gk104_grctx_pack_icmd, - .mthd = gk20a_grctx_pack_mthd, .bundle = gk104_grctx_generate_bundle, .bundle_size = 0x1800, .bundle_min_gpm_fifo_depth = 0x62, @@ -59,4 +85,4 @@ gk20a_grctx_oclass = &(struct gf100_grctx_oclass) { .attrib_nr = 0x240, .alpha_nr_max = 0x648 + (0x648 / 2), .alpha_nr = 0x648, -}.base; +}; diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm107.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm107.c index fbeaae3ae..95f59e316 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm107.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm107.c @@ -863,27 +863,27 @@ gm107_grctx_pack_ppc[] = { void gm107_grctx_generate_bundle(struct gf100_grctx *info) { - const struct gf100_grctx_oclass *impl = gf100_grctx_impl(info->priv); - const u32 state_limit = min(impl->bundle_min_gpm_fifo_depth, - impl->bundle_size / 0x20); - const u32 token_limit = impl->bundle_token_limit; + const struct gf100_grctx_func *grctx = info->gr->func->grctx; + const u32 state_limit = min(grctx->bundle_min_gpm_fifo_depth, + grctx->bundle_size / 0x20); + const u32 token_limit = grctx->bundle_token_limit; const u32 access = NV_MEM_ACCESS_RW | NV_MEM_ACCESS_SYS; const int s = 8; - const int b = mmio_vram(info, impl->bundle_size, (1 << s), access); + const int b = mmio_vram(info, grctx->bundle_size, (1 << s), access); mmio_refn(info, 0x408004, 0x00000000, s, b); - mmio_wr32(info, 0x408008, 0x80000000 | (impl->bundle_size >> s)); + mmio_wr32(info, 0x408008, 0x80000000 | (grctx->bundle_size >> s)); mmio_refn(info, 0x418e24, 0x00000000, s, b); - mmio_wr32(info, 0x418e28, 0x80000000 | (impl->bundle_size >> s)); + mmio_wr32(info, 0x418e28, 0x80000000 | (grctx->bundle_size >> s)); mmio_wr32(info, 0x4064c8, (state_limit << 16) | token_limit); } void gm107_grctx_generate_pagepool(struct gf100_grctx *info) { - const struct gf100_grctx_oclass *impl = gf100_grctx_impl(info->priv); + const struct gf100_grctx_func *grctx = info->gr->func->grctx; const u32 access = NV_MEM_ACCESS_RW | NV_MEM_ACCESS_SYS; const int s = 8; - const int b = mmio_vram(info, impl->pagepool_size, (1 << s), access); + const int b = mmio_vram(info, grctx->pagepool_size, (1 << s), access); mmio_refn(info, 0x40800c, 0x00000000, s, b); mmio_wr32(info, 0x408010, 0x80000000); mmio_refn(info, 0x419004, 0x00000000, s, b); @@ -895,17 +895,17 @@ gm107_grctx_generate_pagepool(struct gf100_grctx *info) void gm107_grctx_generate_attrib(struct gf100_grctx *info) { - struct gf100_gr_priv *priv = info->priv; - const struct gf100_grctx_oclass *impl = (void *)gf100_grctx_impl(priv); - const u32 alpha = impl->alpha_nr; - const u32 attrib = impl->attrib_nr; - const u32 size = 0x20 * (impl->attrib_nr_max + impl->alpha_nr_max); + struct gf100_gr *gr = info->gr; + const struct gf100_grctx_func *grctx = gr->func->grctx; + const u32 alpha = grctx->alpha_nr; + const u32 attrib = grctx->attrib_nr; + const u32 size = 0x20 * (grctx->attrib_nr_max + grctx->alpha_nr_max); const u32 access = NV_MEM_ACCESS_RW; const int s = 12; - const int b = mmio_vram(info, size * priv->tpc_total, (1 << s), access); + const int b = mmio_vram(info, size * gr->tpc_total, (1 << s), access); const int max_batches = 0xffff; u32 bo = 0; - u32 ao = bo + impl->attrib_nr_max * priv->tpc_total; + u32 ao = bo + grctx->attrib_nr_max * gr->tpc_total; int gpc, ppc, n = 0; mmio_refn(info, 0x418810, 0x80000000, s, b); @@ -914,97 +914,90 @@ gm107_grctx_generate_attrib(struct gf100_grctx *info) mmio_wr32(info, 0x405830, (attrib << 16) | alpha); mmio_wr32(info, 0x4064c4, ((alpha / 4) << 16) | max_batches); - for (gpc = 0; gpc < priv->gpc_nr; gpc++) { - for (ppc = 0; ppc < priv->ppc_nr[gpc]; ppc++, n++) { - const u32 as = alpha * priv->ppc_tpc_nr[gpc][ppc]; - const u32 bs = attrib * priv->ppc_tpc_nr[gpc][ppc]; + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + for (ppc = 0; ppc < gr->ppc_nr[gpc]; ppc++, n++) { + const u32 as = alpha * gr->ppc_tpc_nr[gpc][ppc]; + const u32 bs = attrib * gr->ppc_tpc_nr[gpc][ppc]; const u32 u = 0x418ea0 + (n * 0x04); const u32 o = PPC_UNIT(gpc, ppc, 0); mmio_wr32(info, o + 0xc0, bs); mmio_wr32(info, o + 0xf4, bo); - bo += impl->attrib_nr_max * priv->ppc_tpc_nr[gpc][ppc]; + bo += grctx->attrib_nr_max * gr->ppc_tpc_nr[gpc][ppc]; mmio_wr32(info, o + 0xe4, as); mmio_wr32(info, o + 0xf8, ao); - ao += impl->alpha_nr_max * priv->ppc_tpc_nr[gpc][ppc]; + ao += grctx->alpha_nr_max * gr->ppc_tpc_nr[gpc][ppc]; mmio_wr32(info, u, ((bs / 3 /*XXX*/) << 16) | bs); } } } -static void -gm107_grctx_generate_tpcid(struct gf100_gr_priv *priv) +void +gm107_grctx_generate_tpcid(struct gf100_gr *gr) { + struct nvkm_device *device = gr->base.engine.subdev.device; int gpc, tpc, id; for (tpc = 0, id = 0; tpc < 4; tpc++) { - for (gpc = 0; gpc < priv->gpc_nr; gpc++) { - if (tpc < priv->tpc_nr[gpc]) { - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x698), id); - nv_wr32(priv, GPC_UNIT(gpc, 0x0c10 + tpc * 4), id); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x088), id); + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + if (tpc < gr->tpc_nr[gpc]) { + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x698), id); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0c10 + tpc * 4), id); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x088), id); id++; } - nv_wr32(priv, GPC_UNIT(gpc, 0x0c08), priv->tpc_nr[gpc]); - nv_wr32(priv, GPC_UNIT(gpc, 0x0c8c), priv->tpc_nr[gpc]); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0c08), gr->tpc_nr[gpc]); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0c8c), gr->tpc_nr[gpc]); } } } static void -gm107_grctx_generate_main(struct gf100_gr_priv *priv, struct gf100_grctx *info) +gm107_grctx_generate_main(struct gf100_gr *gr, struct gf100_grctx *info) { - struct gf100_grctx_oclass *oclass = (void *)nv_engine(priv)->cclass; + struct nvkm_device *device = gr->base.engine.subdev.device; + const struct gf100_grctx_func *grctx = gr->func->grctx; int i; - gf100_gr_mmio(priv, oclass->hub); - gf100_gr_mmio(priv, oclass->gpc); - gf100_gr_mmio(priv, oclass->zcull); - gf100_gr_mmio(priv, oclass->tpc); - gf100_gr_mmio(priv, oclass->ppc); + gf100_gr_mmio(gr, grctx->hub); + gf100_gr_mmio(gr, grctx->gpc); + gf100_gr_mmio(gr, grctx->zcull); + gf100_gr_mmio(gr, grctx->tpc); + gf100_gr_mmio(gr, grctx->ppc); - nv_wr32(priv, 0x404154, 0x00000000); + nvkm_wr32(device, 0x404154, 0x00000000); - oclass->bundle(info); - oclass->pagepool(info); - oclass->attrib(info); - oclass->unkn(priv); + grctx->bundle(info); + grctx->pagepool(info); + grctx->attrib(info); + grctx->unkn(gr); - gm107_grctx_generate_tpcid(priv); - gf100_grctx_generate_r406028(priv); - gk104_grctx_generate_r418bb8(priv); - gf100_grctx_generate_r406800(priv); + gm107_grctx_generate_tpcid(gr); + gf100_grctx_generate_r406028(gr); + gk104_grctx_generate_r418bb8(gr); + gf100_grctx_generate_r406800(gr); - nv_wr32(priv, 0x4064d0, 0x00000001); + nvkm_wr32(device, 0x4064d0, 0x00000001); for (i = 1; i < 8; i++) - nv_wr32(priv, 0x4064d0 + (i * 0x04), 0x00000000); - nv_wr32(priv, 0x406500, 0x00000001); + nvkm_wr32(device, 0x4064d0 + (i * 0x04), 0x00000000); + nvkm_wr32(device, 0x406500, 0x00000001); - nv_wr32(priv, 0x405b00, (priv->tpc_total << 8) | priv->gpc_nr); + nvkm_wr32(device, 0x405b00, (gr->tpc_total << 8) | gr->gpc_nr); - gk104_grctx_generate_rop_active_fbps(priv); + gk104_grctx_generate_rop_active_fbps(gr); - gf100_gr_icmd(priv, oclass->icmd); - nv_wr32(priv, 0x404154, 0x00000400); - gf100_gr_mthd(priv, oclass->mthd); + gf100_gr_icmd(gr, grctx->icmd); + nvkm_wr32(device, 0x404154, 0x00000400); + gf100_gr_mthd(gr, grctx->mthd); - nv_mask(priv, 0x419e00, 0x00808080, 0x00808080); - nv_mask(priv, 0x419ccc, 0x80000000, 0x80000000); - nv_mask(priv, 0x419f80, 0x80000000, 0x80000000); - nv_mask(priv, 0x419f88, 0x80000000, 0x80000000); + nvkm_mask(device, 0x419e00, 0x00808080, 0x00808080); + nvkm_mask(device, 0x419ccc, 0x80000000, 0x80000000); + nvkm_mask(device, 0x419f80, 0x80000000, 0x80000000); + nvkm_mask(device, 0x419f88, 0x80000000, 0x80000000); } -struct nvkm_oclass * -gm107_grctx_oclass = &(struct gf100_grctx_oclass) { - .base.handle = NV_ENGCTX(GR, 0x08), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gf100_gr_context_ctor, - .dtor = gf100_gr_context_dtor, - .init = _nvkm_gr_context_init, - .fini = _nvkm_gr_context_fini, - .rd32 = _nvkm_gr_context_rd32, - .wr32 = _nvkm_gr_context_wr32, - }, +const struct gf100_grctx_func +gm107_grctx = { .main = gm107_grctx_generate_main, .unkn = gk104_grctx_generate_unkn, .hub = gm107_grctx_pack_hub, @@ -1025,4 +1018,4 @@ gm107_grctx_oclass = &(struct gf100_grctx_oclass) { .attrib_nr = 0xaa0, .alpha_nr_max = 0x1800, .alpha_nr = 0x1000, -}.base; +}; diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm204.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm204.c index ea8e66151..170cbfdbe 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm204.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm204.c @@ -918,17 +918,18 @@ gm204_grctx_pack_ppc[] = { * PGRAPH context implementation ******************************************************************************/ -static void -gm204_grctx_generate_tpcid(struct gf100_gr_priv *priv) +void +gm204_grctx_generate_tpcid(struct gf100_gr *gr) { + struct nvkm_device *device = gr->base.engine.subdev.device; int gpc, tpc, id; for (tpc = 0, id = 0; tpc < 4; tpc++) { - for (gpc = 0; gpc < priv->gpc_nr; gpc++) { - if (tpc < priv->tpc_nr[gpc]) { - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x698), id); - nv_wr32(priv, GPC_UNIT(gpc, 0x0c10 + tpc * 4), id); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x088), id); + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + if (tpc < gr->tpc_nr[gpc]) { + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x698), id); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0c10 + tpc * 4), id); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x088), id); id++; } } @@ -936,101 +937,95 @@ gm204_grctx_generate_tpcid(struct gf100_gr_priv *priv) } static void -gm204_grctx_generate_rop_active_fbps(struct gf100_gr_priv *priv) +gm204_grctx_generate_rop_active_fbps(struct gf100_gr *gr) { - const u32 fbp_count = nv_rd32(priv, 0x12006c); - nv_mask(priv, 0x408850, 0x0000000f, fbp_count); /* zrop */ - nv_mask(priv, 0x408958, 0x0000000f, fbp_count); /* crop */ + struct nvkm_device *device = gr->base.engine.subdev.device; + const u32 fbp_count = nvkm_rd32(device, 0x12006c); + nvkm_mask(device, 0x408850, 0x0000000f, fbp_count); /* zrop */ + nvkm_mask(device, 0x408958, 0x0000000f, fbp_count); /* crop */ } -static void -gm204_grctx_generate_405b60(struct gf100_gr_priv *priv) +void +gm204_grctx_generate_405b60(struct gf100_gr *gr) { - const u32 dist_nr = DIV_ROUND_UP(priv->tpc_total, 4); - u32 dist[TPC_MAX] = {}; + struct nvkm_device *device = gr->base.engine.subdev.device; + const u32 dist_nr = DIV_ROUND_UP(gr->tpc_total, 4); + u32 dist[TPC_MAX / 4] = {}; u32 gpcs[GPC_MAX] = {}; u8 tpcnr[GPC_MAX]; int tpc, gpc, i; - memcpy(tpcnr, priv->tpc_nr, sizeof(priv->tpc_nr)); + memcpy(tpcnr, gr->tpc_nr, sizeof(gr->tpc_nr)); /* won't result in the same distribution as the binary driver where * some of the gpcs have more tpcs than others, but this shall do * for the moment. the code for earlier gpus has this issue too. */ - for (gpc = -1, i = 0; i < priv->tpc_total; i++) { + for (gpc = -1, i = 0; i < gr->tpc_total; i++) { do { - gpc = (gpc + 1) % priv->gpc_nr; + gpc = (gpc + 1) % gr->gpc_nr; } while(!tpcnr[gpc]); - tpc = priv->tpc_nr[gpc] - tpcnr[gpc]--; + tpc = gr->tpc_nr[gpc] - tpcnr[gpc]--; dist[i / 4] |= ((gpc << 4) | tpc) << ((i % 4) * 8); gpcs[gpc] |= i << (tpc * 8); } for (i = 0; i < dist_nr; i++) - nv_wr32(priv, 0x405b60 + (i * 4), dist[i]); - for (i = 0; i < priv->gpc_nr; i++) - nv_wr32(priv, 0x405ba0 + (i * 4), gpcs[i]); + nvkm_wr32(device, 0x405b60 + (i * 4), dist[i]); + for (i = 0; i < gr->gpc_nr; i++) + nvkm_wr32(device, 0x405ba0 + (i * 4), gpcs[i]); } void -gm204_grctx_generate_main(struct gf100_gr_priv *priv, struct gf100_grctx *info) +gm204_grctx_generate_main(struct gf100_gr *gr, struct gf100_grctx *info) { - struct gf100_grctx_oclass *oclass = (void *)nv_engine(priv)->cclass; + struct nvkm_device *device = gr->base.engine.subdev.device; + const struct gf100_grctx_func *grctx = gr->func->grctx; u32 tmp; int i; - gf100_gr_mmio(priv, oclass->hub); - gf100_gr_mmio(priv, oclass->gpc); - gf100_gr_mmio(priv, oclass->zcull); - gf100_gr_mmio(priv, oclass->tpc); - gf100_gr_mmio(priv, oclass->ppc); + gf100_gr_mmio(gr, grctx->hub); + gf100_gr_mmio(gr, grctx->gpc); + gf100_gr_mmio(gr, grctx->zcull); + gf100_gr_mmio(gr, grctx->tpc); + gf100_gr_mmio(gr, grctx->ppc); - nv_wr32(priv, 0x404154, 0x00000000); + nvkm_wr32(device, 0x404154, 0x00000000); - oclass->bundle(info); - oclass->pagepool(info); - oclass->attrib(info); - oclass->unkn(priv); + grctx->bundle(info); + grctx->pagepool(info); + grctx->attrib(info); + grctx->unkn(gr); - gm204_grctx_generate_tpcid(priv); - gf100_grctx_generate_r406028(priv); - gk104_grctx_generate_r418bb8(priv); + gm204_grctx_generate_tpcid(gr); + gf100_grctx_generate_r406028(gr); + gk104_grctx_generate_r418bb8(gr); for (i = 0; i < 8; i++) - nv_wr32(priv, 0x4064d0 + (i * 0x04), 0x00000000); - nv_wr32(priv, 0x406500, 0x00000000); + nvkm_wr32(device, 0x4064d0 + (i * 0x04), 0x00000000); + nvkm_wr32(device, 0x406500, 0x00000000); - nv_wr32(priv, 0x405b00, (priv->tpc_total << 8) | priv->gpc_nr); + nvkm_wr32(device, 0x405b00, (gr->tpc_total << 8) | gr->gpc_nr); - gm204_grctx_generate_rop_active_fbps(priv); + gm204_grctx_generate_rop_active_fbps(gr); - for (tmp = 0, i = 0; i < priv->gpc_nr; i++) - tmp |= ((1 << priv->tpc_nr[i]) - 1) << (i * 4); - nv_wr32(priv, 0x4041c4, tmp); + for (tmp = 0, i = 0; i < gr->gpc_nr; i++) + tmp |= ((1 << gr->tpc_nr[i]) - 1) << (i * 4); + nvkm_wr32(device, 0x4041c4, tmp); - gm204_grctx_generate_405b60(priv); + gm204_grctx_generate_405b60(gr); - gf100_gr_icmd(priv, oclass->icmd); - nv_wr32(priv, 0x404154, 0x00000800); - gf100_gr_mthd(priv, oclass->mthd); + gf100_gr_icmd(gr, grctx->icmd); + nvkm_wr32(device, 0x404154, 0x00000800); + gf100_gr_mthd(gr, grctx->mthd); - nv_mask(priv, 0x418e94, 0xffffffff, 0xc4230000); - nv_mask(priv, 0x418e4c, 0xffffffff, 0x70000000); + nvkm_mask(device, 0x418e94, 0xffffffff, 0xc4230000); + nvkm_mask(device, 0x418e4c, 0xffffffff, 0x70000000); } -struct nvkm_oclass * -gm204_grctx_oclass = &(struct gf100_grctx_oclass) { - .base.handle = NV_ENGCTX(GR, 0x24), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gf100_gr_context_ctor, - .dtor = gf100_gr_context_dtor, - .init = _nvkm_gr_context_init, - .fini = _nvkm_gr_context_fini, - .rd32 = _nvkm_gr_context_rd32, - .wr32 = _nvkm_gr_context_wr32, - }, +const struct gf100_grctx_func +gm204_grctx = { .main = gm204_grctx_generate_main, .unkn = gk104_grctx_generate_unkn, .hub = gm204_grctx_pack_hub, @@ -1051,4 +1046,4 @@ gm204_grctx_oclass = &(struct gf100_grctx_oclass) { .attrib_nr = 0x400, .alpha_nr_max = 0x1800, .alpha_nr = 0x1000, -}.base; +}; diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm206.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm206.c index 91ec41617..d6be6034c 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm206.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm206.c @@ -49,17 +49,8 @@ gm206_grctx_pack_gpc[] = { {} }; -struct nvkm_oclass * -gm206_grctx_oclass = &(struct gf100_grctx_oclass) { - .base.handle = NV_ENGCTX(GR, 0x26), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gf100_gr_context_ctor, - .dtor = gf100_gr_context_dtor, - .init = _nvkm_gr_context_init, - .fini = _nvkm_gr_context_fini, - .rd32 = _nvkm_gr_context_rd32, - .wr32 = _nvkm_gr_context_wr32, - }, +const struct gf100_grctx_func +gm206_grctx = { .main = gm204_grctx_generate_main, .unkn = gk104_grctx_generate_unkn, .hub = gm204_grctx_pack_hub, @@ -80,4 +71,4 @@ gm206_grctx_oclass = &(struct gf100_grctx_oclass) { .attrib_nr = 0x400, .alpha_nr_max = 0x1800, .alpha_nr = 0x1000, -}.base; +}; diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm20b.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm20b.c new file mode 100644 index 000000000..670260402 --- /dev/null +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxgm20b.c @@ -0,0 +1,103 @@ +/* + * Copyright (c) 2015, NVIDIA CORPORATION. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ +#include "ctxgf100.h" + +static void +gm20b_grctx_generate_r406028(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + u32 tpc_per_gpc = 0; + int i; + + for (i = 0; i < gr->gpc_nr; i++) + tpc_per_gpc |= gr->tpc_nr[i] << (4 * i); + + nvkm_wr32(device, 0x406028, tpc_per_gpc); + nvkm_wr32(device, 0x405870, tpc_per_gpc); +} + +static void +gm20b_grctx_generate_main(struct gf100_gr *gr, struct gf100_grctx *info) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + const struct gf100_grctx_func *grctx = gr->func->grctx; + int idle_timeout_save; + int i, tmp; + + gf100_gr_mmio(gr, gr->fuc_sw_ctx); + + gf100_gr_wait_idle(gr); + + idle_timeout_save = nvkm_rd32(device, 0x404154); + nvkm_wr32(device, 0x404154, 0x00000000); + + grctx->attrib(info); + + grctx->unkn(gr); + + gm204_grctx_generate_tpcid(gr); + gm20b_grctx_generate_r406028(gr); + gk104_grctx_generate_r418bb8(gr); + + for (i = 0; i < 8; i++) + nvkm_wr32(device, 0x4064d0 + (i * 0x04), 0x00000000); + + nvkm_wr32(device, 0x405b00, (gr->tpc_total << 8) | gr->gpc_nr); + + gk104_grctx_generate_rop_active_fbps(gr); + nvkm_wr32(device, 0x408908, nvkm_rd32(device, 0x410108) | 0x80000000); + + for (tmp = 0, i = 0; i < gr->gpc_nr; i++) + tmp |= ((1 << gr->tpc_nr[i]) - 1) << (i * 4); + nvkm_wr32(device, 0x4041c4, tmp); + + gm204_grctx_generate_405b60(gr); + + gf100_gr_wait_idle(gr); + + nvkm_wr32(device, 0x404154, idle_timeout_save); + gf100_gr_wait_idle(gr); + + gf100_gr_mthd(gr, gr->fuc_method); + gf100_gr_wait_idle(gr); + + gf100_gr_icmd(gr, gr->fuc_bundle); + grctx->pagepool(info); + grctx->bundle(info); +} + +const struct gf100_grctx_func +gm20b_grctx = { + .main = gm20b_grctx_generate_main, + .unkn = gk104_grctx_generate_unkn, + .bundle = gm107_grctx_generate_bundle, + .bundle_size = 0x1800, + .bundle_min_gpm_fifo_depth = 0x182, + .bundle_token_limit = 0x1c0, + .pagepool = gm107_grctx_generate_pagepool, + .pagepool_size = 0x8000, + .attrib = gm107_grctx_generate_attrib, + .attrib_nr_max = 0x600, + .attrib_nr = 0x400, + .alpha_nr_max = 0xc00, + .alpha_nr = 0x800, +}; diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv40.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv40.c index dc31462af..80a6b017a 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv40.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv40.c @@ -111,7 +111,6 @@ #include "ctxnv40.h" #include "nv40.h" -#include <core/device.h> /* TODO: * - get vs count from 0x1540 @@ -583,13 +582,13 @@ nv40_gr_construct_shader(struct nvkm_grctx *ctx) offset += 0x0280/4; for (i = 0; i < 16; i++, offset += 2) - nv_wo32(obj, offset * 4, 0x3f800000); + nvkm_wo32(obj, offset * 4, 0x3f800000); for (vs = 0; vs < vs_nr; vs++, offset += vs_len) { for (i = 0; i < vs_nr_b0 * 6; i += 6) - nv_wo32(obj, (offset + b0_offset + i) * 4, 0x00000001); + nvkm_wo32(obj, (offset + b0_offset + i) * 4, 0x00000001); for (i = 0; i < vs_nr_b1 * 4; i += 4) - nv_wo32(obj, (offset + b1_offset + i) * 4, 0x3f800000); + nvkm_wo32(obj, (offset + b1_offset + i) * 4, 0x3f800000); } } @@ -675,7 +674,7 @@ nv40_grctx_init(struct nvkm_device *device, u32 *size) struct nvkm_grctx ctx = { .device = device, .mode = NVKM_GRCTX_PROG, - .data = ctxprog, + .ucode = ctxprog, .ctxprog_max = 256, }; @@ -684,9 +683,9 @@ nv40_grctx_init(struct nvkm_device *device, u32 *size) nv40_grctx_generate(&ctx); - nv_wr32(device, 0x400324, 0); + nvkm_wr32(device, 0x400324, 0); for (i = 0; i < ctx.ctxprog_len; i++) - nv_wr32(device, 0x400328, ctxprog[i]); + nvkm_wr32(device, 0x400328, ctxprog[i]); *size = ctx.ctxvals_pos * 4; kfree(ctxprog); diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv40.h b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv40.h index 8a8996195..50e808e9f 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv40.h +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv40.h @@ -9,7 +9,8 @@ struct nvkm_grctx { NVKM_GRCTX_PROG, NVKM_GRCTX_VALS } mode; - void *data; + u32 *ucode; + struct nvkm_gpuobj *data; u32 ctxprog_max; u32 ctxprog_len; @@ -22,7 +23,7 @@ struct nvkm_grctx { static inline void cp_out(struct nvkm_grctx *ctx, u32 inst) { - u32 *ctxprog = ctx->data; + u32 *ctxprog = ctx->ucode; if (ctx->mode != NVKM_GRCTX_PROG) return; @@ -56,7 +57,7 @@ cp_ctx(struct nvkm_grctx *ctx, u32 reg, u32 length) static inline void cp_name(struct nvkm_grctx *ctx, int name) { - u32 *ctxprog = ctx->data; + u32 *ctxprog = ctx->ucode; int i; if (ctx->mode != NVKM_GRCTX_PROG) @@ -124,6 +125,6 @@ gr_def(struct nvkm_grctx *ctx, u32 reg, u32 val) reg = (reg - 0x00400000) / 4; reg = (reg - ctx->ctxprog_reg) + ctx->ctxvals_base; - nv_wo32(ctx->data, reg * 4, val); + nvkm_wo32(ctx->data, reg * 4, val); } #endif diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv50.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv50.c index 9c9528d2c..1e13278cf 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv50.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/ctxnv50.c @@ -107,7 +107,6 @@ #include "ctxnv40.h" -#include <core/device.h> #include <subdev/fb.h> #define IS_NVA3F(x) (((x) > 0xa0 && (x) < 0xaa) || (x) == 0xaf) @@ -269,7 +268,7 @@ nv50_grctx_init(struct nvkm_device *device, u32 *size) struct nvkm_grctx ctx = { .device = device, .mode = NVKM_GRCTX_PROG, - .data = ctxprog, + .ucode = ctxprog, .ctxprog_max = 512, }; @@ -277,9 +276,9 @@ nv50_grctx_init(struct nvkm_device *device, u32 *size) return -ENOMEM; nv50_grctx_generate(&ctx); - nv_wr32(device, 0x400324, 0); + nvkm_wr32(device, 0x400324, 0); for (i = 0; i < ctx.ctxprog_len; i++) - nv_wr32(device, 0x400328, ctxprog[i]); + nvkm_wr32(device, 0x400328, ctxprog[i]); *size = ctx.ctxvals_pos * 4; kfree(ctxprog); return 0; @@ -299,7 +298,7 @@ nv50_gr_construct_mmio(struct nvkm_grctx *ctx) struct nvkm_device *device = ctx->device; int i, j; int offset, base; - u32 units = nv_rd32 (ctx->device, 0x1540); + u32 units = nvkm_rd32(device, 0x1540); /* 0800: DISPATCH */ cp_ctx(ctx, 0x400808, 7); @@ -570,7 +569,7 @@ nv50_gr_construct_mmio(struct nvkm_grctx *ctx) else if (device->chipset < 0xa0) gr_def(ctx, 0x407d08, 0x00390040); else { - if (nvkm_fb(device)->ram->type != NV_MEM_TYPE_GDDR5) + if (device->fb->ram->type != NVKM_RAM_TYPE_GDDR5) gr_def(ctx, 0x407d08, 0x003d0040); else gr_def(ctx, 0x407d08, 0x003c0040); @@ -784,9 +783,10 @@ nv50_gr_construct_mmio(struct nvkm_grctx *ctx) static void dd_emit(struct nvkm_grctx *ctx, int num, u32 val) { int i; - if (val && ctx->mode == NVKM_GRCTX_VALS) + if (val && ctx->mode == NVKM_GRCTX_VALS) { for (i = 0; i < num; i++) - nv_wo32(ctx->data, 4 * (ctx->ctxvals_pos + i), val); + nvkm_wo32(ctx->data, 4 * (ctx->ctxvals_pos + i), val); + } ctx->ctxvals_pos += num; } @@ -1156,9 +1156,10 @@ nv50_gr_construct_mmio_ddata(struct nvkm_grctx *ctx) static void xf_emit(struct nvkm_grctx *ctx, int num, u32 val) { int i; - if (val && ctx->mode == NVKM_GRCTX_VALS) + if (val && ctx->mode == NVKM_GRCTX_VALS) { for (i = 0; i < num; i++) - nv_wo32(ctx->data, 4 * (ctx->ctxvals_pos + (i << 3)), val); + nvkm_wo32(ctx->data, 4 * (ctx->ctxvals_pos + (i << 3)), val); + } ctx->ctxvals_pos += num << 3; } @@ -1190,7 +1191,7 @@ nv50_gr_construct_xfer1(struct nvkm_grctx *ctx) int i; int offset; int size = 0; - u32 units = nv_rd32 (ctx->device, 0x1540); + u32 units = nvkm_rd32(device, 0x1540); offset = (ctx->ctxvals_pos+0x3f)&~0x3f; ctx->ctxvals_base = offset; @@ -3273,7 +3274,7 @@ nv50_gr_construct_xfer2(struct nvkm_grctx *ctx) struct nvkm_device *device = ctx->device; int i; u32 offset; - u32 units = nv_rd32 (ctx->device, 0x1540); + u32 units = nvkm_rd32(device, 0x1540); int size = 0; offset = (ctx->ctxvals_pos+0x3f)&~0x3f; diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpc.fuc b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpc.fuc index 194afe910..7dacb3cc0 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpc.fuc +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpc.fuc @@ -52,10 +52,12 @@ mmio_list_base: #endif #ifdef INCLUDE_CODE +#define gpc_addr(reg,addr) /* +*/ imm32(reg,addr) /* +*/ or reg NV_PGRAPH_GPCX_GPCCS_MMIO_CTRL_BASE_ENABLE #define gpc_wr32(addr,reg) /* +*/ gpc_addr($r14,addr) /* */ mov b32 $r15 reg /* -*/ imm32($r14, addr) /* -*/ or $r14 NV_PGRAPH_GPCX_GPCCS_MMIO_CTRL_BASE_ENABLE /* */ call(nv_wr32) // reports an exception to the host @@ -161,7 +163,7 @@ init: #if NV_PGRAPH_GPCX_UNK__SIZE > 0 // figure out which, and how many, UNKs are actually present - imm32($r14, 0x500c30) + gpc_addr($r14, 0x500c30) clear b32 $r2 clear b32 $r3 clear b32 $r4 diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf117.fuc3.h b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf117.fuc3.h index 64d07df4b..bb820ff28 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf117.fuc3.h +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgf117.fuc3.h @@ -314,7 +314,7 @@ uint32_t gf117_grgpc_code[] = { 0x03f01200, 0x0002d000, 0x17f104bd, - 0x10fe0542, + 0x10fe0545, 0x0007f100, 0x0003f007, 0xbd0000d0, @@ -338,184 +338,184 @@ uint32_t gf117_grgpc_code[] = { 0x02d00103, 0xf104bd00, 0xf00c30e7, - 0x24bd50e3, - 0x44bd34bd, -/* 0x0430: init_unk_loop */ - 0xb06821f4, - 0x0bf400f6, - 0x01f7f00f, - 0xfd04f2bb, - 0x30b6054f, -/* 0x0445: init_unk_next */ - 0x0120b601, - 0xb004e0b6, - 0x1bf40126, -/* 0x0451: init_unk_done */ - 0x070380e2, - 0xf1080480, - 0xf0010027, - 0x22cf0223, - 0x9534bd00, - 0x07f10825, - 0x03f0c000, - 0x0005d001, - 0x07f104bd, - 0x03f0c100, - 0x0005d001, - 0x0e9804bd, - 0x010f9800, - 0x015021f5, - 0xbb002fbb, - 0x0e98003f, - 0x020f9801, - 0x015021f5, - 0xfd050e98, - 0x2ebb00ef, - 0x003ebb00, - 0x98020e98, - 0x21f5030f, - 0x0e980150, - 0x00effd07, - 0xbb002ebb, - 0x35b6003e, - 0x0007f102, - 0x0103f0d3, - 0xbd0003d0, - 0x0825b604, - 0xb60635b6, - 0x30b60120, - 0x0824b601, - 0xb90834b6, - 0x21f5022f, - 0x2fbb02d3, - 0x003fbb00, - 0x010007f1, - 0xd00203f0, + 0xe5f050e3, + 0xbd24bd01, +/* 0x0433: init_unk_loop */ + 0xf444bd34, + 0xf6b06821, + 0x0f0bf400, + 0xbb01f7f0, + 0x4ffd04f2, + 0x0130b605, +/* 0x0448: init_unk_next */ + 0xb60120b6, + 0x26b004e0, + 0xe21bf401, +/* 0x0454: init_unk_done */ + 0x80070380, + 0x27f10804, + 0x23f00100, + 0x0022cf02, + 0x259534bd, + 0x0007f108, + 0x0103f0c0, + 0xbd0005d0, + 0x0007f104, + 0x0103f0c1, + 0xbd0005d0, + 0x000e9804, + 0xf5010f98, + 0xbb015021, + 0x3fbb002f, + 0x010e9800, + 0xf5020f98, + 0x98015021, + 0xeffd050e, + 0x002ebb00, + 0x98003ebb, + 0x0f98020e, + 0x5021f503, + 0x070e9801, + 0xbb00effd, + 0x3ebb002e, + 0x0235b600, + 0xd30007f1, + 0xd00103f0, 0x04bd0003, - 0x29f024bd, - 0x0007f11f, - 0x0203f008, - 0xbd0002d0, -/* 0x0505: main */ - 0x0031f404, - 0xf00028f4, - 0x21f424d7, - 0xf401f439, - 0xf404e4b0, - 0x81fe1e18, - 0x0627f001, - 0x12fd20bd, - 0x01e4b604, - 0xfe051efd, - 0x21f50018, - 0x0ef405fa, -/* 0x0535: main_not_ctx_xfer */ - 0x10ef94d3, - 0xf501f5f0, - 0xf4037e21, -/* 0x0542: ih */ - 0x80f9c60e, - 0xf90188fe, - 0xf990f980, - 0xf9b0f9a0, - 0xf9e0f9d0, - 0xf104bdf0, - 0xf00200a7, - 0xaacf00a3, - 0x04abc400, - 0xf02c0bf4, - 0xe7f124d7, - 0xe3f01a00, - 0x00eecf00, - 0x1900f7f1, - 0xcf00f3f0, - 0x21f400ff, - 0x01e7f004, - 0x1d0007f1, - 0xd00003f0, - 0x04bd000e, -/* 0x0590: ih_no_fifo */ - 0x010007f1, - 0xd00003f0, - 0x04bd000a, - 0xe0fcf0fc, - 0xb0fcd0fc, - 0x90fca0fc, - 0x88fe80fc, - 0xf480fc00, - 0x01f80032, -/* 0x05b4: hub_barrier_done */ - 0x9801f7f0, - 0xfebb040e, - 0x02ffb904, - 0x9418e7f1, - 0xf440e3f0, - 0x00f89d21, -/* 0x05cc: ctx_redswitch */ - 0xf120f7f0, + 0xb60825b6, + 0x20b60635, + 0x0130b601, + 0xb60824b6, + 0x2fb90834, + 0xd321f502, + 0x002fbb02, + 0xf1003fbb, + 0xf0010007, + 0x03d00203, + 0xbd04bd00, + 0x1f29f024, + 0x080007f1, + 0xd00203f0, + 0x04bd0002, +/* 0x0508: main */ + 0xf40031f4, + 0xd7f00028, + 0x3921f424, + 0xb0f401f4, + 0x18f404e4, + 0x0181fe1e, + 0xbd0627f0, + 0x0412fd20, + 0xfd01e4b6, + 0x18fe051e, + 0xfd21f500, + 0xd30ef405, +/* 0x0538: main_not_ctx_xfer */ + 0xf010ef94, + 0x21f501f5, + 0x0ef4037e, +/* 0x0545: ih */ + 0xfe80f9c6, + 0x80f90188, + 0xa0f990f9, + 0xd0f9b0f9, + 0xf0f9e0f9, + 0xa7f104bd, + 0xa3f00200, + 0x00aacf00, + 0xf404abc4, + 0xd7f02c0b, + 0x00e7f124, + 0x00e3f01a, + 0xf100eecf, + 0xf01900f7, + 0xffcf00f3, + 0x0421f400, + 0xf101e7f0, + 0xf01d0007, + 0x0ed00003, +/* 0x0593: ih_no_fifo */ + 0xf104bd00, + 0xf0010007, + 0x0ad00003, + 0xfc04bd00, + 0xfce0fcf0, + 0xfcb0fcd0, + 0xfc90fca0, + 0x0088fe80, + 0x32f480fc, +/* 0x05b7: hub_barrier_done */ + 0xf001f800, + 0x0e9801f7, + 0x04febb04, + 0xf102ffb9, + 0xf09418e7, + 0x21f440e3, +/* 0x05cf: ctx_redswitch */ + 0xf000f89d, + 0x07f120f7, + 0x03f08500, + 0x000fd001, + 0xe7f004bd, +/* 0x05e1: ctx_redswitch_delay */ + 0x01e2b608, + 0xf1fd1bf4, + 0xf10800f5, + 0xf10200f5, 0xf0850007, 0x0fd00103, - 0xf004bd00, -/* 0x05de: ctx_redswitch_delay */ - 0xe2b608e7, - 0xfd1bf401, - 0x0800f5f1, - 0x0200f5f1, - 0x850007f1, - 0xd00103f0, - 0x04bd000f, -/* 0x05fa: ctx_xfer */ - 0x07f100f8, - 0x03f08100, - 0x000fd002, - 0x11f404bd, - 0xcc21f507, -/* 0x060d: ctx_xfer_not_load */ - 0x6a21f505, - 0xf124bd02, - 0xf047fc07, - 0x02d00203, - 0xf004bd00, - 0x20b6012c, - 0xfc07f103, - 0x0203f04a, - 0xbd0002d0, - 0x01acf004, - 0xf102a5f0, - 0xf00000b7, - 0x0c9850b3, - 0x0fc4b604, - 0x9800bcbb, - 0x0d98000c, - 0x00e7f001, - 0x016f21f5, - 0xf101acf0, - 0xf04000b7, - 0x0c9850b3, - 0x0fc4b604, - 0x9800bcbb, - 0x0d98010c, - 0x060f9802, - 0x0800e7f1, - 0x016f21f5, + 0xf804bd00, +/* 0x05fd: ctx_xfer */ + 0x0007f100, + 0x0203f081, + 0xbd000fd0, + 0x0711f404, + 0x05cf21f5, +/* 0x0610: ctx_xfer_not_load */ + 0x026a21f5, + 0x07f124bd, + 0x03f047fc, + 0x0002d002, + 0x2cf004bd, + 0x0320b601, + 0x4afc07f1, + 0xd00203f0, + 0x04bd0002, 0xf001acf0, - 0xb7f104a5, - 0xb3f03000, + 0xb7f102a5, + 0xb3f00000, 0x040c9850, 0xbb0fc4b6, 0x0c9800bc, - 0x030d9802, - 0xf1080f98, - 0xf50200e7, - 0xf5016f21, - 0xf4025e21, - 0x12f40601, -/* 0x06a9: ctx_xfer_post */ - 0x7f21f507, -/* 0x06ad: ctx_xfer_done */ - 0xb421f502, - 0x0000f805, - 0x00000000, + 0x010d9800, + 0xf500e7f0, + 0xf0016f21, + 0xb7f101ac, + 0xb3f04000, + 0x040c9850, + 0xbb0fc4b6, + 0x0c9800bc, + 0x020d9801, + 0xf1060f98, + 0xf50800e7, + 0xf0016f21, + 0xa5f001ac, + 0x00b7f104, + 0x50b3f030, + 0xb6040c98, + 0xbcbb0fc4, + 0x020c9800, + 0x98030d98, + 0xe7f1080f, + 0x21f50200, + 0x21f5016f, + 0x01f4025e, + 0x0712f406, +/* 0x06ac: ctx_xfer_post */ + 0x027f21f5, +/* 0x06b0: ctx_xfer_done */ + 0x05b721f5, + 0x000000f8, 0x00000000, 0x00000000, 0x00000000, diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk104.fuc3.h b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk104.fuc3.h index 2f596433c..911976d20 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk104.fuc3.h +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk104.fuc3.h @@ -314,7 +314,7 @@ uint32_t gk104_grgpc_code[] = { 0x03f01200, 0x0002d000, 0x17f104bd, - 0x10fe0542, + 0x10fe0545, 0x0007f100, 0x0003f007, 0xbd0000d0, @@ -338,184 +338,184 @@ uint32_t gk104_grgpc_code[] = { 0x02d00103, 0xf104bd00, 0xf00c30e7, - 0x24bd50e3, - 0x44bd34bd, -/* 0x0430: init_unk_loop */ - 0xb06821f4, - 0x0bf400f6, - 0x01f7f00f, - 0xfd04f2bb, - 0x30b6054f, -/* 0x0445: init_unk_next */ - 0x0120b601, - 0xb004e0b6, - 0x1bf40126, -/* 0x0451: init_unk_done */ - 0x070380e2, - 0xf1080480, - 0xf0010027, - 0x22cf0223, - 0x9534bd00, - 0x07f10825, - 0x03f0c000, - 0x0005d001, - 0x07f104bd, - 0x03f0c100, - 0x0005d001, - 0x0e9804bd, - 0x010f9800, - 0x015021f5, - 0xbb002fbb, - 0x0e98003f, - 0x020f9801, - 0x015021f5, - 0xfd050e98, - 0x2ebb00ef, - 0x003ebb00, - 0x98020e98, - 0x21f5030f, - 0x0e980150, - 0x00effd07, - 0xbb002ebb, - 0x35b6003e, - 0x0007f102, - 0x0103f0d3, - 0xbd0003d0, - 0x0825b604, - 0xb60635b6, - 0x30b60120, - 0x0824b601, - 0xb90834b6, - 0x21f5022f, - 0x2fbb02d3, - 0x003fbb00, - 0x010007f1, - 0xd00203f0, + 0xe5f050e3, + 0xbd24bd01, +/* 0x0433: init_unk_loop */ + 0xf444bd34, + 0xf6b06821, + 0x0f0bf400, + 0xbb01f7f0, + 0x4ffd04f2, + 0x0130b605, +/* 0x0448: init_unk_next */ + 0xb60120b6, + 0x26b004e0, + 0xe21bf401, +/* 0x0454: init_unk_done */ + 0x80070380, + 0x27f10804, + 0x23f00100, + 0x0022cf02, + 0x259534bd, + 0x0007f108, + 0x0103f0c0, + 0xbd0005d0, + 0x0007f104, + 0x0103f0c1, + 0xbd0005d0, + 0x000e9804, + 0xf5010f98, + 0xbb015021, + 0x3fbb002f, + 0x010e9800, + 0xf5020f98, + 0x98015021, + 0xeffd050e, + 0x002ebb00, + 0x98003ebb, + 0x0f98020e, + 0x5021f503, + 0x070e9801, + 0xbb00effd, + 0x3ebb002e, + 0x0235b600, + 0xd30007f1, + 0xd00103f0, 0x04bd0003, - 0x29f024bd, - 0x0007f11f, - 0x0203f008, - 0xbd0002d0, -/* 0x0505: main */ - 0x0031f404, - 0xf00028f4, - 0x21f424d7, - 0xf401f439, - 0xf404e4b0, - 0x81fe1e18, - 0x0627f001, - 0x12fd20bd, - 0x01e4b604, - 0xfe051efd, - 0x21f50018, - 0x0ef405fa, -/* 0x0535: main_not_ctx_xfer */ - 0x10ef94d3, - 0xf501f5f0, - 0xf4037e21, -/* 0x0542: ih */ - 0x80f9c60e, - 0xf90188fe, - 0xf990f980, - 0xf9b0f9a0, - 0xf9e0f9d0, - 0xf104bdf0, - 0xf00200a7, - 0xaacf00a3, - 0x04abc400, - 0xf02c0bf4, - 0xe7f124d7, - 0xe3f01a00, - 0x00eecf00, - 0x1900f7f1, - 0xcf00f3f0, - 0x21f400ff, - 0x01e7f004, - 0x1d0007f1, - 0xd00003f0, - 0x04bd000e, -/* 0x0590: ih_no_fifo */ - 0x010007f1, - 0xd00003f0, - 0x04bd000a, - 0xe0fcf0fc, - 0xb0fcd0fc, - 0x90fca0fc, - 0x88fe80fc, - 0xf480fc00, - 0x01f80032, -/* 0x05b4: hub_barrier_done */ - 0x9801f7f0, - 0xfebb040e, - 0x02ffb904, - 0x9418e7f1, - 0xf440e3f0, - 0x00f89d21, -/* 0x05cc: ctx_redswitch */ - 0xf120f7f0, + 0xb60825b6, + 0x20b60635, + 0x0130b601, + 0xb60824b6, + 0x2fb90834, + 0xd321f502, + 0x002fbb02, + 0xf1003fbb, + 0xf0010007, + 0x03d00203, + 0xbd04bd00, + 0x1f29f024, + 0x080007f1, + 0xd00203f0, + 0x04bd0002, +/* 0x0508: main */ + 0xf40031f4, + 0xd7f00028, + 0x3921f424, + 0xb0f401f4, + 0x18f404e4, + 0x0181fe1e, + 0xbd0627f0, + 0x0412fd20, + 0xfd01e4b6, + 0x18fe051e, + 0xfd21f500, + 0xd30ef405, +/* 0x0538: main_not_ctx_xfer */ + 0xf010ef94, + 0x21f501f5, + 0x0ef4037e, +/* 0x0545: ih */ + 0xfe80f9c6, + 0x80f90188, + 0xa0f990f9, + 0xd0f9b0f9, + 0xf0f9e0f9, + 0xa7f104bd, + 0xa3f00200, + 0x00aacf00, + 0xf404abc4, + 0xd7f02c0b, + 0x00e7f124, + 0x00e3f01a, + 0xf100eecf, + 0xf01900f7, + 0xffcf00f3, + 0x0421f400, + 0xf101e7f0, + 0xf01d0007, + 0x0ed00003, +/* 0x0593: ih_no_fifo */ + 0xf104bd00, + 0xf0010007, + 0x0ad00003, + 0xfc04bd00, + 0xfce0fcf0, + 0xfcb0fcd0, + 0xfc90fca0, + 0x0088fe80, + 0x32f480fc, +/* 0x05b7: hub_barrier_done */ + 0xf001f800, + 0x0e9801f7, + 0x04febb04, + 0xf102ffb9, + 0xf09418e7, + 0x21f440e3, +/* 0x05cf: ctx_redswitch */ + 0xf000f89d, + 0x07f120f7, + 0x03f08500, + 0x000fd001, + 0xe7f004bd, +/* 0x05e1: ctx_redswitch_delay */ + 0x01e2b608, + 0xf1fd1bf4, + 0xf10800f5, + 0xf10200f5, 0xf0850007, 0x0fd00103, - 0xf004bd00, -/* 0x05de: ctx_redswitch_delay */ - 0xe2b608e7, - 0xfd1bf401, - 0x0800f5f1, - 0x0200f5f1, - 0x850007f1, - 0xd00103f0, - 0x04bd000f, -/* 0x05fa: ctx_xfer */ - 0x07f100f8, - 0x03f08100, - 0x000fd002, - 0x11f404bd, - 0xcc21f507, -/* 0x060d: ctx_xfer_not_load */ - 0x6a21f505, - 0xf124bd02, - 0xf047fc07, - 0x02d00203, - 0xf004bd00, - 0x20b6012c, - 0xfc07f103, - 0x0203f04a, - 0xbd0002d0, - 0x01acf004, - 0xf102a5f0, - 0xf00000b7, - 0x0c9850b3, - 0x0fc4b604, - 0x9800bcbb, - 0x0d98000c, - 0x00e7f001, - 0x016f21f5, - 0xf101acf0, - 0xf04000b7, - 0x0c9850b3, - 0x0fc4b604, - 0x9800bcbb, - 0x0d98010c, - 0x060f9802, - 0x0800e7f1, - 0x016f21f5, + 0xf804bd00, +/* 0x05fd: ctx_xfer */ + 0x0007f100, + 0x0203f081, + 0xbd000fd0, + 0x0711f404, + 0x05cf21f5, +/* 0x0610: ctx_xfer_not_load */ + 0x026a21f5, + 0x07f124bd, + 0x03f047fc, + 0x0002d002, + 0x2cf004bd, + 0x0320b601, + 0x4afc07f1, + 0xd00203f0, + 0x04bd0002, 0xf001acf0, - 0xb7f104a5, - 0xb3f03000, + 0xb7f102a5, + 0xb3f00000, 0x040c9850, 0xbb0fc4b6, 0x0c9800bc, - 0x030d9802, - 0xf1080f98, - 0xf50200e7, - 0xf5016f21, - 0xf4025e21, - 0x12f40601, -/* 0x06a9: ctx_xfer_post */ - 0x7f21f507, -/* 0x06ad: ctx_xfer_done */ - 0xb421f502, - 0x0000f805, - 0x00000000, + 0x010d9800, + 0xf500e7f0, + 0xf0016f21, + 0xb7f101ac, + 0xb3f04000, + 0x040c9850, + 0xbb0fc4b6, + 0x0c9800bc, + 0x020d9801, + 0xf1060f98, + 0xf50800e7, + 0xf0016f21, + 0xa5f001ac, + 0x00b7f104, + 0x50b3f030, + 0xb6040c98, + 0xbcbb0fc4, + 0x020c9800, + 0x98030d98, + 0xe7f1080f, + 0x21f50200, + 0x21f5016f, + 0x01f4025e, + 0x0712f406, +/* 0x06ac: ctx_xfer_post */ + 0x027f21f5, +/* 0x06b0: ctx_xfer_done */ + 0x05b721f5, + 0x000000f8, 0x00000000, 0x00000000, 0x00000000, diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk110.fuc3.h b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk110.fuc3.h index ee8e54db8..1c6e11b05 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk110.fuc3.h +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk110.fuc3.h @@ -314,7 +314,7 @@ uint32_t gk110_grgpc_code[] = { 0x03f01200, 0x0002d000, 0x17f104bd, - 0x10fe0542, + 0x10fe0545, 0x0007f100, 0x0003f007, 0xbd0000d0, @@ -338,184 +338,184 @@ uint32_t gk110_grgpc_code[] = { 0x02d00103, 0xf104bd00, 0xf00c30e7, - 0x24bd50e3, - 0x44bd34bd, -/* 0x0430: init_unk_loop */ - 0xb06821f4, - 0x0bf400f6, - 0x01f7f00f, - 0xfd04f2bb, - 0x30b6054f, -/* 0x0445: init_unk_next */ - 0x0120b601, - 0xb004e0b6, - 0x1bf40226, -/* 0x0451: init_unk_done */ - 0x070380e2, - 0xf1080480, - 0xf0010027, - 0x22cf0223, - 0x9534bd00, - 0x07f10825, - 0x03f0c000, - 0x0005d001, - 0x07f104bd, - 0x03f0c100, - 0x0005d001, - 0x0e9804bd, - 0x010f9800, - 0x015021f5, - 0xbb002fbb, - 0x0e98003f, - 0x020f9801, - 0x015021f5, - 0xfd050e98, - 0x2ebb00ef, - 0x003ebb00, - 0x98020e98, - 0x21f5030f, - 0x0e980150, - 0x00effd07, - 0xbb002ebb, - 0x35b6003e, - 0x0007f102, - 0x0103f0d3, - 0xbd0003d0, - 0x0825b604, - 0xb60635b6, - 0x30b60120, - 0x0824b601, - 0xb90834b6, - 0x21f5022f, - 0x2fbb02d3, - 0x003fbb00, - 0x010007f1, - 0xd00203f0, + 0xe5f050e3, + 0xbd24bd01, +/* 0x0433: init_unk_loop */ + 0xf444bd34, + 0xf6b06821, + 0x0f0bf400, + 0xbb01f7f0, + 0x4ffd04f2, + 0x0130b605, +/* 0x0448: init_unk_next */ + 0xb60120b6, + 0x26b004e0, + 0xe21bf402, +/* 0x0454: init_unk_done */ + 0x80070380, + 0x27f10804, + 0x23f00100, + 0x0022cf02, + 0x259534bd, + 0x0007f108, + 0x0103f0c0, + 0xbd0005d0, + 0x0007f104, + 0x0103f0c1, + 0xbd0005d0, + 0x000e9804, + 0xf5010f98, + 0xbb015021, + 0x3fbb002f, + 0x010e9800, + 0xf5020f98, + 0x98015021, + 0xeffd050e, + 0x002ebb00, + 0x98003ebb, + 0x0f98020e, + 0x5021f503, + 0x070e9801, + 0xbb00effd, + 0x3ebb002e, + 0x0235b600, + 0xd30007f1, + 0xd00103f0, 0x04bd0003, - 0x29f024bd, - 0x0007f11f, - 0x0203f030, - 0xbd0002d0, -/* 0x0505: main */ - 0x0031f404, - 0xf00028f4, - 0x21f424d7, - 0xf401f439, - 0xf404e4b0, - 0x81fe1e18, - 0x0627f001, - 0x12fd20bd, - 0x01e4b604, - 0xfe051efd, - 0x21f50018, - 0x0ef405fa, -/* 0x0535: main_not_ctx_xfer */ - 0x10ef94d3, - 0xf501f5f0, - 0xf4037e21, -/* 0x0542: ih */ - 0x80f9c60e, - 0xf90188fe, - 0xf990f980, - 0xf9b0f9a0, - 0xf9e0f9d0, - 0xf104bdf0, - 0xf00200a7, - 0xaacf00a3, - 0x04abc400, - 0xf02c0bf4, - 0xe7f124d7, - 0xe3f01a00, - 0x00eecf00, - 0x1900f7f1, - 0xcf00f3f0, - 0x21f400ff, - 0x01e7f004, - 0x1d0007f1, - 0xd00003f0, - 0x04bd000e, -/* 0x0590: ih_no_fifo */ - 0x010007f1, - 0xd00003f0, - 0x04bd000a, - 0xe0fcf0fc, - 0xb0fcd0fc, - 0x90fca0fc, - 0x88fe80fc, - 0xf480fc00, - 0x01f80032, -/* 0x05b4: hub_barrier_done */ - 0x9801f7f0, - 0xfebb040e, - 0x02ffb904, - 0x9418e7f1, - 0xf440e3f0, - 0x00f89d21, -/* 0x05cc: ctx_redswitch */ - 0xf120f7f0, + 0xb60825b6, + 0x20b60635, + 0x0130b601, + 0xb60824b6, + 0x2fb90834, + 0xd321f502, + 0x002fbb02, + 0xf1003fbb, + 0xf0010007, + 0x03d00203, + 0xbd04bd00, + 0x1f29f024, + 0x300007f1, + 0xd00203f0, + 0x04bd0002, +/* 0x0508: main */ + 0xf40031f4, + 0xd7f00028, + 0x3921f424, + 0xb0f401f4, + 0x18f404e4, + 0x0181fe1e, + 0xbd0627f0, + 0x0412fd20, + 0xfd01e4b6, + 0x18fe051e, + 0xfd21f500, + 0xd30ef405, +/* 0x0538: main_not_ctx_xfer */ + 0xf010ef94, + 0x21f501f5, + 0x0ef4037e, +/* 0x0545: ih */ + 0xfe80f9c6, + 0x80f90188, + 0xa0f990f9, + 0xd0f9b0f9, + 0xf0f9e0f9, + 0xa7f104bd, + 0xa3f00200, + 0x00aacf00, + 0xf404abc4, + 0xd7f02c0b, + 0x00e7f124, + 0x00e3f01a, + 0xf100eecf, + 0xf01900f7, + 0xffcf00f3, + 0x0421f400, + 0xf101e7f0, + 0xf01d0007, + 0x0ed00003, +/* 0x0593: ih_no_fifo */ + 0xf104bd00, + 0xf0010007, + 0x0ad00003, + 0xfc04bd00, + 0xfce0fcf0, + 0xfcb0fcd0, + 0xfc90fca0, + 0x0088fe80, + 0x32f480fc, +/* 0x05b7: hub_barrier_done */ + 0xf001f800, + 0x0e9801f7, + 0x04febb04, + 0xf102ffb9, + 0xf09418e7, + 0x21f440e3, +/* 0x05cf: ctx_redswitch */ + 0xf000f89d, + 0x07f120f7, + 0x03f08500, + 0x000fd001, + 0xe7f004bd, +/* 0x05e1: ctx_redswitch_delay */ + 0x01e2b608, + 0xf1fd1bf4, + 0xf10800f5, + 0xf10200f5, 0xf0850007, 0x0fd00103, - 0xf004bd00, -/* 0x05de: ctx_redswitch_delay */ - 0xe2b608e7, - 0xfd1bf401, - 0x0800f5f1, - 0x0200f5f1, - 0x850007f1, - 0xd00103f0, - 0x04bd000f, -/* 0x05fa: ctx_xfer */ - 0x07f100f8, - 0x03f08100, - 0x000fd002, - 0x11f404bd, - 0xcc21f507, -/* 0x060d: ctx_xfer_not_load */ - 0x6a21f505, - 0xf124bd02, - 0xf047fc07, - 0x02d00203, - 0xf004bd00, - 0x20b6012c, - 0xfc07f103, - 0x0203f04a, - 0xbd0002d0, - 0x01acf004, - 0xf102a5f0, - 0xf00000b7, - 0x0c9850b3, - 0x0fc4b604, - 0x9800bcbb, - 0x0d98000c, - 0x00e7f001, - 0x016f21f5, - 0xf101acf0, - 0xf04000b7, - 0x0c9850b3, - 0x0fc4b604, - 0x9800bcbb, - 0x0d98010c, - 0x060f9802, - 0x0800e7f1, - 0x016f21f5, + 0xf804bd00, +/* 0x05fd: ctx_xfer */ + 0x0007f100, + 0x0203f081, + 0xbd000fd0, + 0x0711f404, + 0x05cf21f5, +/* 0x0610: ctx_xfer_not_load */ + 0x026a21f5, + 0x07f124bd, + 0x03f047fc, + 0x0002d002, + 0x2cf004bd, + 0x0320b601, + 0x4afc07f1, + 0xd00203f0, + 0x04bd0002, 0xf001acf0, - 0xb7f104a5, - 0xb3f03000, + 0xb7f102a5, + 0xb3f00000, 0x040c9850, 0xbb0fc4b6, 0x0c9800bc, - 0x030d9802, - 0xf1080f98, - 0xf50200e7, - 0xf5016f21, - 0xf4025e21, - 0x12f40601, -/* 0x06a9: ctx_xfer_post */ - 0x7f21f507, -/* 0x06ad: ctx_xfer_done */ - 0xb421f502, - 0x0000f805, - 0x00000000, + 0x010d9800, + 0xf500e7f0, + 0xf0016f21, + 0xb7f101ac, + 0xb3f04000, + 0x040c9850, + 0xbb0fc4b6, + 0x0c9800bc, + 0x020d9801, + 0xf1060f98, + 0xf50800e7, + 0xf0016f21, + 0xa5f001ac, + 0x00b7f104, + 0x50b3f030, + 0xb6040c98, + 0xbcbb0fc4, + 0x020c9800, + 0x98030d98, + 0xe7f1080f, + 0x21f50200, + 0x21f5016f, + 0x01f4025e, + 0x0712f406, +/* 0x06ac: ctx_xfer_post */ + 0x027f21f5, +/* 0x06b0: ctx_xfer_done */ + 0x05b721f5, + 0x000000f8, 0x00000000, 0x00000000, 0x00000000, diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk208.fuc5.h b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk208.fuc5.h index fbcc342f8..84af7ec6a 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk208.fuc5.h +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgk208.fuc5.h @@ -276,7 +276,7 @@ uint32_t gk208_grgpc_code[] = { 0x02020014, 0xf6120040, 0x04bd0002, - 0xfe048141, + 0xfe048441, 0x00400010, 0x0000f607, 0x040204bd, @@ -295,165 +295,165 @@ uint32_t gk208_grgpc_code[] = { 0x01c90080, 0xbd0002f6, 0x0c308e04, - 0xbd24bd50, -/* 0x0383: init_unk_loop */ - 0x7e44bd34, - 0xb0000065, - 0x0bf400f6, - 0xbb010f0e, - 0x4ffd04f2, - 0x0130b605, -/* 0x0398: init_unk_next */ - 0xb60120b6, - 0x26b004e0, - 0xe21bf401, -/* 0x03a4: init_unk_done */ - 0xb50703b5, - 0x00820804, - 0x22cf0201, - 0x9534bd00, - 0x00800825, - 0x05f601c0, - 0x8004bd00, - 0xf601c100, - 0x04bd0005, - 0x98000e98, - 0x207e010f, - 0x2fbb0001, - 0x003fbb00, - 0x98010e98, - 0x207e020f, - 0x0e980001, - 0x00effd05, - 0xbb002ebb, - 0x0e98003e, - 0x030f9802, - 0x0001207e, - 0xfd070e98, - 0x2ebb00ef, - 0x003ebb00, - 0x800235b6, - 0xf601d300, - 0x04bd0003, - 0xb60825b6, - 0x20b60635, - 0x0130b601, - 0xb60824b6, - 0x2fb20834, - 0x0002687e, - 0xbb002fbb, - 0x0080003f, - 0x03f60201, - 0xbd04bd00, - 0x1f29f024, - 0x02300080, - 0xbd0002f6, -/* 0x0445: main */ - 0x0031f404, - 0x0d0028f4, - 0x00377e24, - 0xf401f400, - 0xf404e4b0, - 0x81fe1d18, - 0xbd060201, - 0x0412fd20, - 0xfd01e4b6, - 0x18fe051e, - 0x05187e00, - 0xd40ef400, -/* 0x0474: main_not_ctx_xfer */ - 0xf010ef94, - 0xf87e01f5, - 0x0ef40002, -/* 0x0481: ih */ - 0xfe80f9c7, - 0x80f90188, - 0xa0f990f9, - 0xd0f9b0f9, - 0xf0f9e0f9, - 0x004a04bd, - 0x00aacf02, - 0xf404abc4, - 0x240d1f0b, - 0xcf1a004e, - 0x004f00ee, - 0x00ffcf19, - 0x0000047e, - 0x0040010e, - 0x000ef61d, -/* 0x04be: ih_no_fifo */ - 0x004004bd, - 0x000af601, - 0xf0fc04bd, - 0xd0fce0fc, - 0xa0fcb0fc, - 0x80fc90fc, - 0xfc0088fe, - 0x0032f480, -/* 0x04de: hub_barrier_done */ - 0x010f01f8, - 0xbb040e98, - 0xffb204fe, - 0x4094188e, - 0x00008f7e, -/* 0x04f2: ctx_redswitch */ - 0x200f00f8, + 0x01e5f050, + 0x34bd24bd, +/* 0x0386: init_unk_loop */ + 0x657e44bd, + 0xf6b00000, + 0x0e0bf400, + 0xf2bb010f, + 0x054ffd04, +/* 0x039b: init_unk_next */ + 0xb60130b6, + 0xe0b60120, + 0x0126b004, +/* 0x03a7: init_unk_done */ + 0xb5e21bf4, + 0x04b50703, + 0x01008208, + 0x0022cf02, + 0x259534bd, + 0xc0008008, + 0x0005f601, + 0x008004bd, + 0x05f601c1, + 0x9804bd00, + 0x0f98000e, + 0x01207e01, + 0x002fbb00, + 0x98003fbb, + 0x0f98010e, + 0x01207e02, + 0x050e9800, + 0xbb00effd, + 0x3ebb002e, + 0x020e9800, + 0x7e030f98, + 0x98000120, + 0xeffd070e, + 0x002ebb00, + 0xb6003ebb, + 0x00800235, + 0x03f601d3, + 0xb604bd00, + 0x35b60825, + 0x0120b606, + 0xb60130b6, + 0x34b60824, + 0x7e2fb208, + 0xbb000268, + 0x3fbb002f, + 0x01008000, + 0x0003f602, + 0x24bd04bd, + 0x801f29f0, + 0xf6023000, + 0x04bd0002, +/* 0x0448: main */ + 0xf40031f4, + 0x240d0028, + 0x0000377e, + 0xb0f401f4, + 0x18f404e4, + 0x0181fe1d, + 0x20bd0602, + 0xb60412fd, + 0x1efd01e4, + 0x0018fe05, + 0x00051b7e, +/* 0x0477: main_not_ctx_xfer */ + 0x94d40ef4, + 0xf5f010ef, + 0x02f87e01, + 0xc70ef400, +/* 0x0484: ih */ + 0x88fe80f9, + 0xf980f901, + 0xf9a0f990, + 0xf9d0f9b0, + 0xbdf0f9e0, + 0x02004a04, + 0xc400aacf, + 0x0bf404ab, + 0x4e240d1f, + 0xeecf1a00, + 0x19004f00, + 0x7e00ffcf, + 0x0e000004, + 0x1d004001, + 0xbd000ef6, +/* 0x04c1: ih_no_fifo */ + 0x01004004, + 0xbd000af6, + 0xfcf0fc04, + 0xfcd0fce0, + 0xfca0fcb0, + 0xfe80fc90, + 0x80fc0088, + 0xf80032f4, +/* 0x04e1: hub_barrier_done */ + 0x98010f01, + 0xfebb040e, + 0x8effb204, + 0x7e409418, + 0xf800008f, +/* 0x04f5: ctx_redswitch */ + 0x80200f00, + 0xf6018500, + 0x04bd000f, +/* 0x0502: ctx_redswitch_delay */ + 0xe2b6080e, + 0xfd1bf401, + 0x0800f5f1, + 0x0200f5f1, 0x01850080, 0xbd000ff6, -/* 0x04ff: ctx_redswitch_delay */ - 0xb6080e04, - 0x1bf401e2, - 0x00f5f1fd, - 0x00f5f108, - 0x85008002, - 0x000ff601, - 0x00f804bd, -/* 0x0518: ctx_xfer */ - 0x02810080, - 0xbd000ff6, - 0x0711f404, - 0x0004f27e, -/* 0x0528: ctx_xfer_not_load */ - 0x0002167e, - 0xfc8024bd, - 0x02f60247, - 0xf004bd00, - 0x20b6012c, - 0x4afc8003, +/* 0x051b: ctx_xfer */ + 0x8000f804, + 0xf6028100, + 0x04bd000f, + 0x7e0711f4, +/* 0x052b: ctx_xfer_not_load */ + 0x7e0004f5, + 0xbd000216, + 0x47fc8024, 0x0002f602, - 0xacf004bd, - 0x02a5f001, - 0x5000008b, - 0xb6040c98, - 0xbcbb0fc4, - 0x000c9800, - 0x0e010d98, - 0x013d7e00, - 0x01acf000, - 0x5040008b, - 0xb6040c98, - 0xbcbb0fc4, - 0x010c9800, - 0x98020d98, - 0x004e060f, - 0x013d7e08, - 0x01acf000, - 0x8b04a5f0, - 0x98503000, + 0x2cf004bd, + 0x0320b601, + 0x024afc80, + 0xbd0002f6, + 0x01acf004, + 0x8b02a5f0, + 0x98500000, 0xc4b6040c, 0x00bcbb0f, - 0x98020c98, - 0x0f98030d, - 0x02004e08, + 0x98000c98, + 0x000e010d, 0x00013d7e, - 0x00020a7e, - 0xf40601f4, -/* 0x05b2: ctx_xfer_post */ - 0x277e0712, -/* 0x05b6: ctx_xfer_done */ - 0xde7e0002, - 0x00f80004, - 0x00000000, + 0x8b01acf0, + 0x98504000, + 0xc4b6040c, + 0x00bcbb0f, + 0x98010c98, + 0x0f98020d, + 0x08004e06, + 0x00013d7e, + 0xf001acf0, + 0x008b04a5, + 0x0c985030, + 0x0fc4b604, + 0x9800bcbb, + 0x0d98020c, + 0x080f9803, + 0x7e02004e, + 0x7e00013d, + 0xf400020a, + 0x12f40601, +/* 0x05b5: ctx_xfer_post */ + 0x02277e07, +/* 0x05b9: ctx_xfer_done */ + 0x04e17e00, + 0x0000f800, 0x00000000, 0x00000000, 0x00000000, diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgm107.fuc5.h b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgm107.fuc5.h index 51f5c3c6e..11bf363a6 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgm107.fuc5.h +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/fuc/gpcgm107.fuc5.h @@ -289,7 +289,7 @@ uint32_t gm107_grgpc_code[] = { 0x020014fe, 0x12004002, 0xbd0002f6, - 0x05b04104, + 0x05b34104, 0x400010fe, 0x00f60700, 0x0204bd00, @@ -308,259 +308,259 @@ uint32_t gm107_grgpc_code[] = { 0xc900800f, 0x0002f601, 0x308e04bd, - 0x24bd500c, - 0x44bd34bd, -/* 0x03b0: init_unk_loop */ - 0x0000657e, - 0xf400f6b0, - 0x010f0e0b, - 0xfd04f2bb, - 0x30b6054f, -/* 0x03c5: init_unk_next */ - 0x0120b601, - 0xb004e0b6, - 0x1bf40226, -/* 0x03d1: init_unk_done */ - 0x0703b5e2, - 0x820804b5, - 0xcf020100, - 0x34bd0022, - 0x80082595, - 0xf601c000, + 0xe5f0500c, + 0xbd24bd01, +/* 0x03b3: init_unk_loop */ + 0x7e44bd34, + 0xb0000065, + 0x0bf400f6, + 0xbb010f0e, + 0x4ffd04f2, + 0x0130b605, +/* 0x03c8: init_unk_next */ + 0xb60120b6, + 0x26b004e0, + 0xe21bf402, +/* 0x03d4: init_unk_done */ + 0xb50703b5, + 0x00820804, + 0x22cf0201, + 0x9534bd00, + 0x00800825, + 0x05f601c0, + 0x8004bd00, + 0xf601c100, 0x04bd0005, - 0x01c10080, - 0xbd0005f6, - 0x000e9804, - 0x7e010f98, - 0xbb000120, - 0x3fbb002f, - 0x010e9800, - 0x7e020f98, - 0x98000120, - 0xeffd050e, - 0x002ebb00, - 0x98003ebb, - 0x0f98020e, - 0x01207e03, - 0x070e9800, - 0xbb00effd, - 0x3ebb002e, - 0x0235b600, - 0x01d30080, - 0xbd0003f6, - 0x0825b604, - 0xb60635b6, - 0x30b60120, - 0x0824b601, - 0xb20834b6, - 0x02687e2f, - 0x002fbb00, - 0x0f003fbb, - 0x8effb23f, - 0xf0501d60, - 0x8f7e01e5, - 0x0c0f0000, - 0xa88effb2, - 0xe5f0501d, - 0x008f7e01, - 0x03147e00, - 0xb23f0f00, - 0x1d608eff, - 0x01e5f050, - 0x00008f7e, - 0xffb2000f, - 0x501d9c8e, - 0x7e01e5f0, - 0x0f00008f, - 0x03147e01, - 0x8effb200, + 0x98000e98, + 0x207e010f, + 0x2fbb0001, + 0x003fbb00, + 0x98010e98, + 0x207e020f, + 0x0e980001, + 0x00effd05, + 0xbb002ebb, + 0x0e98003e, + 0x030f9802, + 0x0001207e, + 0xfd070e98, + 0x2ebb00ef, + 0x003ebb00, + 0x800235b6, + 0xf601d300, + 0x04bd0003, + 0xb60825b6, + 0x20b60635, + 0x0130b601, + 0xb60824b6, + 0x2fb20834, + 0x0002687e, + 0xbb002fbb, + 0x3f0f003f, + 0x501d608e, + 0xb201e5f0, + 0x008f7eff, + 0x8e0c0f00, 0xf0501da8, - 0x8f7e01e5, - 0xff0f0000, - 0x988effb2, + 0xffb201e5, + 0x00008f7e, + 0x0003147e, + 0x608e3f0f, 0xe5f0501d, - 0x008f7e01, - 0xb2020f00, - 0x1da88eff, + 0x7effb201, + 0x0f00008f, + 0x1d9c8e00, 0x01e5f050, - 0x00008f7e, + 0x8f7effb2, + 0x010f0000, 0x0003147e, - 0x85050498, - 0x98504000, - 0x64b60406, - 0x0056bb0f, -/* 0x04e0: tpc_strand_init_tpc_loop */ - 0x05705eb8, - 0x00657e00, - 0xbdf6b200, -/* 0x04ed: tpc_strand_init_idx_loop */ - 0x605eb874, - 0x7fb20005, - 0x00008f7e, - 0x05885eb8, - 0x082f9500, - 0x00008f7e, - 0x058c5eb8, - 0x082f9500, + 0x501da88e, + 0xb201e5f0, + 0x008f7eff, + 0x8eff0f00, + 0xf0501d98, + 0xffb201e5, 0x00008f7e, - 0x05905eb8, - 0x00657e00, - 0x06f5b600, - 0xb601f0b6, - 0x2fbb08f4, - 0x003fbb00, - 0xb60170b6, - 0x1bf40162, - 0x0050b7bf, - 0x0142b608, - 0x0fa81bf4, - 0x8effb23f, - 0xf0501d60, - 0x8f7e01e5, - 0x0d0f0000, - 0xa88effb2, + 0xa88e020f, 0xe5f0501d, - 0x008f7e01, - 0x03147e00, - 0x01008000, - 0x0003f602, - 0x24bd04bd, - 0x801f29f0, - 0xf6023000, - 0x04bd0002, -/* 0x0574: main */ - 0xf40031f4, - 0x240d0028, - 0x0000377e, - 0xb0f401f4, - 0x18f404e4, - 0x0181fe1d, - 0x20bd0602, - 0xb60412fd, - 0x1efd01e4, - 0x0018fe05, - 0x0006477e, -/* 0x05a3: main_not_ctx_xfer */ - 0x94d40ef4, - 0xf5f010ef, - 0x02f87e01, - 0xc70ef400, -/* 0x05b0: ih */ - 0x88fe80f9, - 0xf980f901, - 0xf9a0f990, - 0xf9d0f9b0, - 0xbdf0f9e0, - 0x02004a04, - 0xc400aacf, - 0x0bf404ab, - 0x4e240d1f, - 0xeecf1a00, - 0x19004f00, - 0x7e00ffcf, - 0x0e000004, - 0x1d004001, - 0xbd000ef6, -/* 0x05ed: ih_no_fifo */ - 0x01004004, - 0xbd000af6, - 0xfcf0fc04, - 0xfcd0fce0, - 0xfca0fcb0, - 0xfe80fc90, - 0x80fc0088, - 0xf80032f4, -/* 0x060d: hub_barrier_done */ - 0x98010f01, - 0xfebb040e, - 0x8effb204, - 0x7e409418, - 0xf800008f, -/* 0x0621: ctx_redswitch */ - 0x80200f00, + 0x7effb201, + 0x7e00008f, + 0x98000314, + 0x00850504, + 0x06985040, + 0x0f64b604, +/* 0x04e3: tpc_strand_init_tpc_loop */ + 0xb80056bb, + 0x0005705e, + 0x0000657e, + 0x74bdf6b2, +/* 0x04f0: tpc_strand_init_idx_loop */ + 0x05605eb8, + 0x7e7fb200, + 0xb800008f, + 0x0005885e, + 0x7e082f95, + 0xb800008f, + 0x00058c5e, + 0x7e082f95, + 0xb800008f, + 0x0005905e, + 0x0000657e, + 0xb606f5b6, + 0xf4b601f0, + 0x002fbb08, + 0xb6003fbb, + 0x62b60170, + 0xbf1bf401, + 0x080050b7, + 0xf40142b6, + 0x3f0fa81b, + 0x501d608e, + 0xb201e5f0, + 0x008f7eff, + 0x8e0d0f00, + 0xf0501da8, + 0xffb201e5, + 0x00008f7e, + 0x0003147e, + 0x02010080, + 0xbd0003f6, + 0xf024bd04, + 0x00801f29, + 0x02f60230, +/* 0x0577: main */ + 0xf404bd00, + 0x28f40031, + 0x7e240d00, + 0xf4000037, + 0xe4b0f401, + 0x1d18f404, + 0x020181fe, + 0xfd20bd06, + 0xe4b60412, + 0x051efd01, + 0x7e0018fe, + 0xf400064a, +/* 0x05a6: main_not_ctx_xfer */ + 0xef94d40e, + 0x01f5f010, + 0x0002f87e, +/* 0x05b3: ih */ + 0xf9c70ef4, + 0x0188fe80, + 0x90f980f9, + 0xb0f9a0f9, + 0xe0f9d0f9, + 0x04bdf0f9, + 0xcf02004a, + 0xabc400aa, + 0x1f0bf404, + 0x004e240d, + 0x00eecf1a, + 0xcf19004f, + 0x047e00ff, + 0x010e0000, + 0xf61d0040, + 0x04bd000e, +/* 0x05f0: ih_no_fifo */ + 0xf6010040, + 0x04bd000a, + 0xe0fcf0fc, + 0xb0fcd0fc, + 0x90fca0fc, + 0x88fe80fc, + 0xf480fc00, + 0x01f80032, +/* 0x0610: hub_barrier_done */ + 0x0e98010f, + 0x04febb04, + 0x188effb2, + 0x8f7e4094, + 0x00f80000, +/* 0x0624: ctx_redswitch */ + 0x0080200f, + 0x0ff60185, + 0x0e04bd00, +/* 0x0631: ctx_redswitch_delay */ + 0x01e2b608, + 0xf1fd1bf4, + 0xf10800f5, + 0x800200f5, 0xf6018500, 0x04bd000f, -/* 0x062e: ctx_redswitch_delay */ - 0xe2b6080e, - 0xfd1bf401, - 0x0800f5f1, - 0x0200f5f1, - 0x01850080, - 0xbd000ff6, -/* 0x0647: ctx_xfer */ - 0x8000f804, - 0xf6028100, - 0x04bd000f, - 0xc48effb2, - 0xe5f0501d, - 0x008f7e01, - 0x0711f400, - 0x0006217e, -/* 0x0664: ctx_xfer_not_load */ - 0x0002167e, - 0xfc8024bd, - 0x02f60247, - 0xf004bd00, - 0x20b6012c, - 0x4afc8003, +/* 0x064a: ctx_xfer */ + 0x008000f8, + 0x0ff60281, + 0x8e04bd00, + 0xf0501dc4, + 0xffb201e5, + 0x00008f7e, + 0x7e0711f4, +/* 0x0667: ctx_xfer_not_load */ + 0x7e000624, + 0xbd000216, + 0x47fc8024, 0x0002f602, - 0x0c0f04bd, - 0xa88effb2, - 0xe5f0501d, - 0x008f7e01, - 0x03147e00, - 0xb23f0f00, - 0x1d608eff, - 0x01e5f050, + 0x2cf004bd, + 0x0320b601, + 0x024afc80, + 0xbd0002f6, + 0x8e0c0f04, + 0xf0501da8, + 0xffb201e5, 0x00008f7e, - 0xffb2000f, - 0x501d9c8e, - 0x7e01e5f0, + 0x0003147e, + 0x608e3f0f, + 0xe5f0501d, + 0x7effb201, 0x0f00008f, - 0x03147e01, - 0x01fcf000, - 0xb203f0b6, - 0x1da88eff, + 0x1d9c8e00, 0x01e5f050, - 0x00008f7e, - 0xf001acf0, - 0x008b02a5, - 0x0c985000, - 0x0fc4b604, - 0x9800bcbb, - 0x0d98000c, - 0x7e000e01, - 0xf000013d, - 0x008b01ac, - 0x0c985040, - 0x0fc4b604, - 0x9800bcbb, - 0x0d98010c, - 0x060f9802, - 0x7e08004e, - 0xf000013d, + 0x8f7effb2, + 0x010f0000, + 0x0003147e, + 0xb601fcf0, + 0xa88e03f0, + 0xe5f0501d, + 0x7effb201, + 0xf000008f, 0xa5f001ac, - 0x30008b04, + 0x00008b02, 0x040c9850, 0xbb0fc4b6, 0x0c9800bc, - 0x030d9802, - 0x4e080f98, - 0x3d7e0200, - 0x0a7e0001, - 0x147e0002, - 0x01f40003, - 0x1a12f406, -/* 0x073c: ctx_xfer_post */ - 0x0002277e, - 0xffb20d0f, - 0x501da88e, - 0x7e01e5f0, - 0x7e00008f, -/* 0x0753: ctx_xfer_done */ - 0x7e000314, - 0xf800060d, - 0x00000000, + 0x010d9800, + 0x3d7e000e, + 0xacf00001, + 0x40008b01, + 0x040c9850, + 0xbb0fc4b6, + 0x0c9800bc, + 0x020d9801, + 0x4e060f98, + 0x3d7e0800, + 0xacf00001, + 0x04a5f001, + 0x5030008b, + 0xb6040c98, + 0xbcbb0fc4, + 0x020c9800, + 0x98030d98, + 0x004e080f, + 0x013d7e02, + 0x020a7e00, + 0x03147e00, + 0x0601f400, +/* 0x073f: ctx_xfer_post */ + 0x7e1a12f4, + 0x0f000227, + 0x1da88e0d, + 0x01e5f050, + 0x8f7effb2, + 0x147e0000, +/* 0x0756: ctx_xfer_done */ + 0x107e0003, + 0x00f80006, 0x00000000, 0x00000000, 0x00000000, diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/g84.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/g84.c new file mode 100644 index 000000000..ce9133005 --- /dev/null +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/g84.c @@ -0,0 +1,196 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "nv50.h" + +#include <subdev/timer.h> + +static const struct nvkm_bitfield nv50_gr_status[] = { + { 0x00000001, "BUSY" }, /* set when any bit is set */ + { 0x00000002, "DISPATCH" }, + { 0x00000004, "UNK2" }, + { 0x00000008, "UNK3" }, + { 0x00000010, "UNK4" }, + { 0x00000020, "UNK5" }, + { 0x00000040, "M2MF" }, + { 0x00000080, "UNK7" }, + { 0x00000100, "CTXPROG" }, + { 0x00000200, "VFETCH" }, + { 0x00000400, "CCACHE_PREGEOM" }, + { 0x00000800, "STRMOUT_VATTR_POSTGEOM" }, + { 0x00001000, "VCLIP" }, + { 0x00002000, "RATTR_APLANE" }, + { 0x00004000, "TRAST" }, + { 0x00008000, "CLIPID" }, + { 0x00010000, "ZCULL" }, + { 0x00020000, "ENG2D" }, + { 0x00040000, "RMASK" }, + { 0x00080000, "TPC_RAST" }, + { 0x00100000, "TPC_PROP" }, + { 0x00200000, "TPC_TEX" }, + { 0x00400000, "TPC_GEOM" }, + { 0x00800000, "TPC_MP" }, + { 0x01000000, "ROP" }, + {} +}; + +static const struct nvkm_bitfield +nv50_gr_vstatus_0[] = { + { 0x01, "VFETCH" }, + { 0x02, "CCACHE" }, + { 0x04, "PREGEOM" }, + { 0x08, "POSTGEOM" }, + { 0x10, "VATTR" }, + { 0x20, "STRMOUT" }, + { 0x40, "VCLIP" }, + {} +}; + +static const struct nvkm_bitfield +nv50_gr_vstatus_1[] = { + { 0x01, "TPC_RAST" }, + { 0x02, "TPC_PROP" }, + { 0x04, "TPC_TEX" }, + { 0x08, "TPC_GEOM" }, + { 0x10, "TPC_MP" }, + {} +}; + +static const struct nvkm_bitfield +nv50_gr_vstatus_2[] = { + { 0x01, "RATTR" }, + { 0x02, "APLANE" }, + { 0x04, "TRAST" }, + { 0x08, "CLIPID" }, + { 0x10, "ZCULL" }, + { 0x20, "ENG2D" }, + { 0x40, "RMASK" }, + { 0x80, "ROP" }, + {} +}; + +static void +nvkm_gr_vstatus_print(struct nv50_gr *gr, int r, + const struct nvkm_bitfield *units, u32 status) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + u32 stat = status; + u8 mask = 0x00; + char msg[64]; + int i; + + for (i = 0; units[i].name && status; i++) { + if ((status & 7) == 1) + mask |= (1 << i); + status >>= 3; + } + + nvkm_snprintbf(msg, sizeof(msg), units, mask); + nvkm_error(subdev, "PGRAPH_VSTATUS%d: %08x [%s]\n", r, stat, msg); +} + +int +g84_gr_tlb_flush(struct nvkm_gr *base) +{ + struct nv50_gr *gr = nv50_gr(base); + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + struct nvkm_timer *tmr = device->timer; + bool idle, timeout = false; + unsigned long flags; + char status[128]; + u64 start; + u32 tmp; + + spin_lock_irqsave(&gr->lock, flags); + nvkm_mask(device, 0x400500, 0x00000001, 0x00000000); + + start = nvkm_timer_read(tmr); + do { + idle = true; + + for (tmp = nvkm_rd32(device, 0x400380); tmp && idle; tmp >>= 3) { + if ((tmp & 7) == 1) + idle = false; + } + + for (tmp = nvkm_rd32(device, 0x400384); tmp && idle; tmp >>= 3) { + if ((tmp & 7) == 1) + idle = false; + } + + for (tmp = nvkm_rd32(device, 0x400388); tmp && idle; tmp >>= 3) { + if ((tmp & 7) == 1) + idle = false; + } + } while (!idle && + !(timeout = nvkm_timer_read(tmr) - start > 2000000000)); + + if (timeout) { + nvkm_error(subdev, "PGRAPH TLB flush idle timeout fail\n"); + + tmp = nvkm_rd32(device, 0x400700); + nvkm_snprintbf(status, sizeof(status), nv50_gr_status, tmp); + nvkm_error(subdev, "PGRAPH_STATUS %08x [%s]\n", tmp, status); + + nvkm_gr_vstatus_print(gr, 0, nv50_gr_vstatus_0, + nvkm_rd32(device, 0x400380)); + nvkm_gr_vstatus_print(gr, 1, nv50_gr_vstatus_1, + nvkm_rd32(device, 0x400384)); + nvkm_gr_vstatus_print(gr, 2, nv50_gr_vstatus_2, + nvkm_rd32(device, 0x400388)); + } + + + nvkm_wr32(device, 0x100c80, 0x00000001); + nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x100c80) & 0x00000001)) + break; + ); + nvkm_mask(device, 0x400500, 0x00000001, 0x00000001); + spin_unlock_irqrestore(&gr->lock, flags); + return timeout ? -EBUSY : 0; +} + +static const struct nvkm_gr_func +g84_gr = { + .init = nv50_gr_init, + .intr = nv50_gr_intr, + .chan_new = nv50_gr_chan_new, + .tlb_flush = g84_gr_tlb_flush, + .units = nv50_gr_units, + .sclass = { + { -1, -1, 0x0030, &nv50_gr_object }, + { -1, -1, 0x502d, &nv50_gr_object }, + { -1, -1, 0x5039, &nv50_gr_object }, + { -1, -1, 0x50c0, &nv50_gr_object }, + { -1, -1, 0x8297, &nv50_gr_object }, + {} + } +}; + +int +g84_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return nv50_gr_new_(&g84_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.c index 5606c25e5..9f5dfc851 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.c @@ -26,13 +26,12 @@ #include "fuc/os.h" #include <core/client.h> -#include <core/device.h> -#include <core/handle.h> #include <core/option.h> -#include <engine/fifo.h> #include <subdev/fb.h> #include <subdev/mc.h> +#include <subdev/pmu.h> #include <subdev/timer.h> +#include <engine/fifo.h> #include <nvif/class.h> #include <nvif/unpack.h> @@ -42,35 +41,36 @@ ******************************************************************************/ static void -gf100_gr_zbc_clear_color(struct gf100_gr_priv *priv, int zbc) +gf100_gr_zbc_clear_color(struct gf100_gr *gr, int zbc) { - if (priv->zbc_color[zbc].format) { - nv_wr32(priv, 0x405804, priv->zbc_color[zbc].ds[0]); - nv_wr32(priv, 0x405808, priv->zbc_color[zbc].ds[1]); - nv_wr32(priv, 0x40580c, priv->zbc_color[zbc].ds[2]); - nv_wr32(priv, 0x405810, priv->zbc_color[zbc].ds[3]); - } - nv_wr32(priv, 0x405814, priv->zbc_color[zbc].format); - nv_wr32(priv, 0x405820, zbc); - nv_wr32(priv, 0x405824, 0x00000004); /* TRIGGER | WRITE | COLOR */ + struct nvkm_device *device = gr->base.engine.subdev.device; + if (gr->zbc_color[zbc].format) { + nvkm_wr32(device, 0x405804, gr->zbc_color[zbc].ds[0]); + nvkm_wr32(device, 0x405808, gr->zbc_color[zbc].ds[1]); + nvkm_wr32(device, 0x40580c, gr->zbc_color[zbc].ds[2]); + nvkm_wr32(device, 0x405810, gr->zbc_color[zbc].ds[3]); + } + nvkm_wr32(device, 0x405814, gr->zbc_color[zbc].format); + nvkm_wr32(device, 0x405820, zbc); + nvkm_wr32(device, 0x405824, 0x00000004); /* TRIGGER | WRITE | COLOR */ } static int -gf100_gr_zbc_color_get(struct gf100_gr_priv *priv, int format, +gf100_gr_zbc_color_get(struct gf100_gr *gr, int format, const u32 ds[4], const u32 l2[4]) { - struct nvkm_ltc *ltc = nvkm_ltc(priv); + struct nvkm_ltc *ltc = gr->base.engine.subdev.device->ltc; int zbc = -ENOSPC, i; for (i = ltc->zbc_min; i <= ltc->zbc_max; i++) { - if (priv->zbc_color[i].format) { - if (priv->zbc_color[i].format != format) + if (gr->zbc_color[i].format) { + if (gr->zbc_color[i].format != format) continue; - if (memcmp(priv->zbc_color[i].ds, ds, sizeof( - priv->zbc_color[i].ds))) + if (memcmp(gr->zbc_color[i].ds, ds, sizeof( + gr->zbc_color[i].ds))) continue; - if (memcmp(priv->zbc_color[i].l2, l2, sizeof( - priv->zbc_color[i].l2))) { + if (memcmp(gr->zbc_color[i].l2, l2, sizeof( + gr->zbc_color[i].l2))) { WARN_ON(1); return -EINVAL; } @@ -83,38 +83,39 @@ gf100_gr_zbc_color_get(struct gf100_gr_priv *priv, int format, if (zbc < 0) return zbc; - memcpy(priv->zbc_color[zbc].ds, ds, sizeof(priv->zbc_color[zbc].ds)); - memcpy(priv->zbc_color[zbc].l2, l2, sizeof(priv->zbc_color[zbc].l2)); - priv->zbc_color[zbc].format = format; - ltc->zbc_color_get(ltc, zbc, l2); - gf100_gr_zbc_clear_color(priv, zbc); + memcpy(gr->zbc_color[zbc].ds, ds, sizeof(gr->zbc_color[zbc].ds)); + memcpy(gr->zbc_color[zbc].l2, l2, sizeof(gr->zbc_color[zbc].l2)); + gr->zbc_color[zbc].format = format; + nvkm_ltc_zbc_color_get(ltc, zbc, l2); + gf100_gr_zbc_clear_color(gr, zbc); return zbc; } static void -gf100_gr_zbc_clear_depth(struct gf100_gr_priv *priv, int zbc) +gf100_gr_zbc_clear_depth(struct gf100_gr *gr, int zbc) { - if (priv->zbc_depth[zbc].format) - nv_wr32(priv, 0x405818, priv->zbc_depth[zbc].ds); - nv_wr32(priv, 0x40581c, priv->zbc_depth[zbc].format); - nv_wr32(priv, 0x405820, zbc); - nv_wr32(priv, 0x405824, 0x00000005); /* TRIGGER | WRITE | DEPTH */ + struct nvkm_device *device = gr->base.engine.subdev.device; + if (gr->zbc_depth[zbc].format) + nvkm_wr32(device, 0x405818, gr->zbc_depth[zbc].ds); + nvkm_wr32(device, 0x40581c, gr->zbc_depth[zbc].format); + nvkm_wr32(device, 0x405820, zbc); + nvkm_wr32(device, 0x405824, 0x00000005); /* TRIGGER | WRITE | DEPTH */ } static int -gf100_gr_zbc_depth_get(struct gf100_gr_priv *priv, int format, +gf100_gr_zbc_depth_get(struct gf100_gr *gr, int format, const u32 ds, const u32 l2) { - struct nvkm_ltc *ltc = nvkm_ltc(priv); + struct nvkm_ltc *ltc = gr->base.engine.subdev.device->ltc; int zbc = -ENOSPC, i; for (i = ltc->zbc_min; i <= ltc->zbc_max; i++) { - if (priv->zbc_depth[i].format) { - if (priv->zbc_depth[i].format != format) + if (gr->zbc_depth[i].format) { + if (gr->zbc_depth[i].format != format) continue; - if (priv->zbc_depth[i].ds != ds) + if (gr->zbc_depth[i].ds != ds) continue; - if (priv->zbc_depth[i].l2 != l2) { + if (gr->zbc_depth[i].l2 != l2) { WARN_ON(1); return -EINVAL; } @@ -127,11 +128,11 @@ gf100_gr_zbc_depth_get(struct gf100_gr_priv *priv, int format, if (zbc < 0) return zbc; - priv->zbc_depth[zbc].format = format; - priv->zbc_depth[zbc].ds = ds; - priv->zbc_depth[zbc].l2 = l2; - ltc->zbc_depth_get(ltc, zbc, l2); - gf100_gr_zbc_clear_depth(priv, zbc); + gr->zbc_depth[zbc].format = format; + gr->zbc_depth[zbc].ds = ds; + gr->zbc_depth[zbc].l2 = l2; + nvkm_ltc_zbc_depth_get(ltc, zbc, l2); + gf100_gr_zbc_clear_depth(gr, zbc); return zbc; } @@ -142,7 +143,7 @@ gf100_gr_zbc_depth_get(struct gf100_gr_priv *priv, int format, static int gf100_fermi_mthd_zbc_color(struct nvkm_object *object, void *data, u32 size) { - struct gf100_gr_priv *priv = (void *)object->engine; + struct gf100_gr *gr = gf100_gr(nvkm_gr(object->engine)); union { struct fermi_a_zbc_color_v0 v0; } *args = data; @@ -169,7 +170,7 @@ gf100_fermi_mthd_zbc_color(struct nvkm_object *object, void *data, u32 size) case FERMI_A_ZBC_COLOR_V0_FMT_AU8BU8GU8RU8: case FERMI_A_ZBC_COLOR_V0_FMT_A2R10G10B10: case FERMI_A_ZBC_COLOR_V0_FMT_BF10GF11RF11: - ret = gf100_gr_zbc_color_get(priv, args->v0.format, + ret = gf100_gr_zbc_color_get(gr, args->v0.format, args->v0.ds, args->v0.l2); if (ret >= 0) { @@ -188,7 +189,7 @@ gf100_fermi_mthd_zbc_color(struct nvkm_object *object, void *data, u32 size) static int gf100_fermi_mthd_zbc_depth(struct nvkm_object *object, void *data, u32 size) { - struct gf100_gr_priv *priv = (void *)object->engine; + struct gf100_gr *gr = gf100_gr(nvkm_gr(object->engine)); union { struct fermi_a_zbc_depth_v0 v0; } *args = data; @@ -197,7 +198,7 @@ gf100_fermi_mthd_zbc_depth(struct nvkm_object *object, void *data, u32 size) if (nvif_unpack(args->v0, 0, 0, false)) { switch (args->v0.format) { case FERMI_A_ZBC_DEPTH_V0_FMT_FP32: - ret = gf100_gr_zbc_depth_get(priv, args->v0.format, + ret = gf100_gr_zbc_depth_get(gr, args->v0.format, args->v0.ds, args->v0.l2); return (ret >= 0) ? 0 : -ENOSPC; @@ -223,106 +224,176 @@ gf100_fermi_mthd(struct nvkm_object *object, u32 mthd, void *data, u32 size) return -EINVAL; } -struct nvkm_ofuncs -gf100_fermi_ofuncs = { - .ctor = _nvkm_object_ctor, - .dtor = nvkm_object_destroy, - .init = nvkm_object_init, - .fini = nvkm_object_fini, +const struct nvkm_object_func +gf100_fermi = { .mthd = gf100_fermi_mthd, }; -static int -gf100_gr_set_shader_exceptions(struct nvkm_object *object, u32 mthd, - void *pdata, u32 size) +static void +gf100_gr_mthd_set_shader_exceptions(struct nvkm_device *device, u32 data) { - struct gf100_gr_priv *priv = (void *)object->engine; - if (size >= sizeof(u32)) { - u32 data = *(u32 *)pdata ? 0xffffffff : 0x00000000; - nv_wr32(priv, 0x419e44, data); - nv_wr32(priv, 0x419e4c, data); - return 0; + nvkm_wr32(device, 0x419e44, data ? 0xffffffff : 0x00000000); + nvkm_wr32(device, 0x419e4c, data ? 0xffffffff : 0x00000000); +} + +static bool +gf100_gr_mthd_sw(struct nvkm_device *device, u16 class, u32 mthd, u32 data) +{ + switch (class & 0x00ff) { + case 0x97: + case 0xc0: + switch (mthd) { + case 0x1528: + gf100_gr_mthd_set_shader_exceptions(device, data); + return true; + default: + break; + } + break; + default: + break; } - return -EINVAL; + return false; } -struct nvkm_omthds -gf100_gr_9097_omthds[] = { - { 0x1528, 0x1528, gf100_gr_set_shader_exceptions }, - {} -}; +static int +gf100_gr_object_get(struct nvkm_gr *base, int index, struct nvkm_sclass *sclass) +{ + struct gf100_gr *gr = gf100_gr(base); + int c = 0; -struct nvkm_omthds -gf100_gr_90c0_omthds[] = { - { 0x1528, 0x1528, gf100_gr_set_shader_exceptions }, - {} -}; + while (gr->func->sclass[c].oclass) { + if (c++ == index) { + *sclass = gr->func->sclass[index]; + return index; + } + } -struct nvkm_oclass -gf100_gr_sclass[] = { - { FERMI_TWOD_A, &nvkm_object_ofuncs }, - { FERMI_MEMORY_TO_MEMORY_FORMAT_A, &nvkm_object_ofuncs }, - { FERMI_A, &gf100_fermi_ofuncs, gf100_gr_9097_omthds }, - { FERMI_COMPUTE_A, &nvkm_object_ofuncs, gf100_gr_90c0_omthds }, - {} -}; + return c; +} /******************************************************************************* * PGRAPH context ******************************************************************************/ -int -gf100_gr_context_ctor(struct nvkm_object *parent, struct nvkm_object *engine, - struct nvkm_oclass *oclass, void *args, u32 size, - struct nvkm_object **pobject) +static int +gf100_gr_chan_bind(struct nvkm_object *object, struct nvkm_gpuobj *parent, + int align, struct nvkm_gpuobj **pgpuobj) { - struct nvkm_vm *vm = nvkm_client(parent)->vm; - struct gf100_gr_priv *priv = (void *)engine; - struct gf100_gr_data *data = priv->mmio_data; - struct gf100_gr_mmio *mmio = priv->mmio_list; - struct gf100_gr_chan *chan; + struct gf100_gr_chan *chan = gf100_gr_chan(object); + struct gf100_gr *gr = chan->gr; int ret, i; - /* allocate memory for context, and fill with default values */ - ret = nvkm_gr_context_create(parent, engine, oclass, NULL, - priv->size, 0x100, - NVOBJ_FLAG_ZERO_ALLOC, &chan); - *pobject = nv_object(chan); + ret = nvkm_gpuobj_new(gr->base.engine.subdev.device, gr->size, + align, false, parent, pgpuobj); if (ret) return ret; + nvkm_kmap(*pgpuobj); + for (i = 0; i < gr->size; i += 4) + nvkm_wo32(*pgpuobj, i, gr->data[i / 4]); + + if (!gr->firmware) { + nvkm_wo32(*pgpuobj, 0x00, chan->mmio_nr / 2); + nvkm_wo32(*pgpuobj, 0x04, chan->mmio_vma.offset >> 8); + } else { + nvkm_wo32(*pgpuobj, 0xf4, 0); + nvkm_wo32(*pgpuobj, 0xf8, 0); + nvkm_wo32(*pgpuobj, 0x10, chan->mmio_nr / 2); + nvkm_wo32(*pgpuobj, 0x14, lower_32_bits(chan->mmio_vma.offset)); + nvkm_wo32(*pgpuobj, 0x18, upper_32_bits(chan->mmio_vma.offset)); + nvkm_wo32(*pgpuobj, 0x1c, 1); + nvkm_wo32(*pgpuobj, 0x20, 0); + nvkm_wo32(*pgpuobj, 0x28, 0); + nvkm_wo32(*pgpuobj, 0x2c, 0); + } + nvkm_done(*pgpuobj); + return 0; +} + +static void * +gf100_gr_chan_dtor(struct nvkm_object *object) +{ + struct gf100_gr_chan *chan = gf100_gr_chan(object); + int i; + + for (i = 0; i < ARRAY_SIZE(chan->data); i++) { + if (chan->data[i].vma.node) { + nvkm_vm_unmap(&chan->data[i].vma); + nvkm_vm_put(&chan->data[i].vma); + } + nvkm_memory_del(&chan->data[i].mem); + } + + if (chan->mmio_vma.node) { + nvkm_vm_unmap(&chan->mmio_vma); + nvkm_vm_put(&chan->mmio_vma); + } + nvkm_memory_del(&chan->mmio); + return chan; +} + +static const struct nvkm_object_func +gf100_gr_chan = { + .dtor = gf100_gr_chan_dtor, + .bind = gf100_gr_chan_bind, +}; + +static int +gf100_gr_chan_new(struct nvkm_gr *base, struct nvkm_fifo_chan *fifoch, + const struct nvkm_oclass *oclass, + struct nvkm_object **pobject) +{ + struct gf100_gr *gr = gf100_gr(base); + struct gf100_gr_data *data = gr->mmio_data; + struct gf100_gr_mmio *mmio = gr->mmio_list; + struct gf100_gr_chan *chan; + struct nvkm_device *device = gr->base.engine.subdev.device; + int ret, i; + + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&gf100_gr_chan, oclass, &chan->object); + chan->gr = gr; + *pobject = &chan->object; + /* allocate memory for a "mmio list" buffer that's used by the HUB * fuc to modify some per-context register settings on first load * of the context. */ - ret = nvkm_gpuobj_new(nv_object(chan), NULL, 0x1000, 0x100, 0, - &chan->mmio); + ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, 0x1000, 0x100, + false, &chan->mmio); if (ret) return ret; - ret = nvkm_gpuobj_map_vm(nv_gpuobj(chan->mmio), vm, - NV_MEM_ACCESS_RW | NV_MEM_ACCESS_SYS, - &chan->mmio_vma); + ret = nvkm_vm_get(fifoch->vm, 0x1000, 12, NV_MEM_ACCESS_RW | + NV_MEM_ACCESS_SYS, &chan->mmio_vma); if (ret) return ret; + nvkm_memory_map(chan->mmio, &chan->mmio_vma, 0); + /* allocate buffers referenced by mmio list */ - for (i = 0; data->size && i < ARRAY_SIZE(priv->mmio_data); i++) { - ret = nvkm_gpuobj_new(nv_object(chan), NULL, data->size, - data->align, 0, &chan->data[i].mem); + for (i = 0; data->size && i < ARRAY_SIZE(gr->mmio_data); i++) { + ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, + data->size, data->align, false, + &chan->data[i].mem); if (ret) return ret; - ret = nvkm_gpuobj_map_vm(chan->data[i].mem, vm, data->access, - &chan->data[i].vma); + ret = nvkm_vm_get(fifoch->vm, + nvkm_memory_size(chan->data[i].mem), 12, + data->access, &chan->data[i].vma); if (ret) return ret; + nvkm_memory_map(chan->data[i].mem, &chan->data[i].vma, 0); data++; } /* finally, fill in the mmio list and point the context at it */ - for (i = 0; mmio->addr && i < ARRAY_SIZE(priv->mmio_list); i++) { + nvkm_kmap(chan->mmio); + for (i = 0; mmio->addr && i < ARRAY_SIZE(gr->mmio_list); i++) { u32 addr = mmio->addr; u32 data = mmio->data; @@ -331,49 +402,14 @@ gf100_gr_context_ctor(struct nvkm_object *parent, struct nvkm_object *engine, data |= info >> mmio->shift; } - nv_wo32(chan->mmio, chan->mmio_nr++ * 4, addr); - nv_wo32(chan->mmio, chan->mmio_nr++ * 4, data); + nvkm_wo32(chan->mmio, chan->mmio_nr++ * 4, addr); + nvkm_wo32(chan->mmio, chan->mmio_nr++ * 4, data); mmio++; } - - for (i = 0; i < priv->size; i += 4) - nv_wo32(chan, i, priv->data[i / 4]); - - if (!priv->firmware) { - nv_wo32(chan, 0x00, chan->mmio_nr / 2); - nv_wo32(chan, 0x04, chan->mmio_vma.offset >> 8); - } else { - nv_wo32(chan, 0xf4, 0); - nv_wo32(chan, 0xf8, 0); - nv_wo32(chan, 0x10, chan->mmio_nr / 2); - nv_wo32(chan, 0x14, lower_32_bits(chan->mmio_vma.offset)); - nv_wo32(chan, 0x18, upper_32_bits(chan->mmio_vma.offset)); - nv_wo32(chan, 0x1c, 1); - nv_wo32(chan, 0x20, 0); - nv_wo32(chan, 0x28, 0); - nv_wo32(chan, 0x2c, 0); - } - + nvkm_done(chan->mmio); return 0; } -void -gf100_gr_context_dtor(struct nvkm_object *object) -{ - struct gf100_gr_chan *chan = (void *)object; - int i; - - for (i = 0; i < ARRAY_SIZE(chan->data); i++) { - nvkm_gpuobj_unmap(&chan->data[i].vma); - nvkm_gpuobj_ref(NULL, &chan->data[i].mem); - } - - nvkm_gpuobj_unmap(&chan->mmio_vma); - nvkm_gpuobj_ref(NULL, &chan->mmio); - - nvkm_gr_context_destroy(&chan->base); -} - /******************************************************************************* * PGRAPH register lists ******************************************************************************/ @@ -635,7 +671,7 @@ gf100_gr_pack_mmio[] = { ******************************************************************************/ void -gf100_gr_zbc_init(struct gf100_gr_priv *priv) +gf100_gr_zbc_init(struct gf100_gr *gr) { const u32 zero[] = { 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000 }; @@ -645,27 +681,62 @@ gf100_gr_zbc_init(struct gf100_gr_priv *priv) 0x00000000, 0x00000000, 0x00000000, 0x00000000 }; const u32 f32_1[] = { 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000 }; - struct nvkm_ltc *ltc = nvkm_ltc(priv); + struct nvkm_ltc *ltc = gr->base.engine.subdev.device->ltc; int index; - if (!priv->zbc_color[0].format) { - gf100_gr_zbc_color_get(priv, 1, & zero[0], &zero[4]); - gf100_gr_zbc_color_get(priv, 2, & one[0], &one[4]); - gf100_gr_zbc_color_get(priv, 4, &f32_0[0], &f32_0[4]); - gf100_gr_zbc_color_get(priv, 4, &f32_1[0], &f32_1[4]); - gf100_gr_zbc_depth_get(priv, 1, 0x00000000, 0x00000000); - gf100_gr_zbc_depth_get(priv, 1, 0x3f800000, 0x3f800000); + if (!gr->zbc_color[0].format) { + gf100_gr_zbc_color_get(gr, 1, & zero[0], &zero[4]); + gf100_gr_zbc_color_get(gr, 2, & one[0], &one[4]); + gf100_gr_zbc_color_get(gr, 4, &f32_0[0], &f32_0[4]); + gf100_gr_zbc_color_get(gr, 4, &f32_1[0], &f32_1[4]); + gf100_gr_zbc_depth_get(gr, 1, 0x00000000, 0x00000000); + gf100_gr_zbc_depth_get(gr, 1, 0x3f800000, 0x3f800000); } for (index = ltc->zbc_min; index <= ltc->zbc_max; index++) - gf100_gr_zbc_clear_color(priv, index); + gf100_gr_zbc_clear_color(gr, index); for (index = ltc->zbc_min; index <= ltc->zbc_max; index++) - gf100_gr_zbc_clear_depth(priv, index); + gf100_gr_zbc_clear_depth(gr, index); +} + +/** + * Wait until GR goes idle. GR is considered idle if it is disabled by the + * MC (0x200) register, or GR is not busy and a context switch is not in + * progress. + */ +int +gf100_gr_wait_idle(struct gf100_gr *gr) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + unsigned long end_jiffies = jiffies + msecs_to_jiffies(2000); + bool gr_enabled, ctxsw_active, gr_busy; + + do { + /* + * required to make sure FIFO_ENGINE_STATUS (0x2640) is + * up-to-date + */ + nvkm_rd32(device, 0x400700); + + gr_enabled = nvkm_rd32(device, 0x200) & 0x1000; + ctxsw_active = nvkm_rd32(device, 0x2640) & 0x8000; + gr_busy = nvkm_rd32(device, 0x40060c) & 0x1; + + if (!gr_enabled || (!gr_busy && !ctxsw_active)) + return 0; + } while (time_before(jiffies, end_jiffies)); + + nvkm_error(subdev, + "wait for idle timeout (en: %d, ctxsw: %d, busy: %d)\n", + gr_enabled, ctxsw_active, gr_busy); + return -EAGAIN; } void -gf100_gr_mmio(struct gf100_gr_priv *priv, const struct gf100_gr_pack *p) +gf100_gr_mmio(struct gf100_gr *gr, const struct gf100_gr_pack *p) { + struct nvkm_device *device = gr->base.engine.subdev.device; const struct gf100_gr_pack *pack; const struct gf100_gr_init *init; @@ -673,43 +744,54 @@ gf100_gr_mmio(struct gf100_gr_priv *priv, const struct gf100_gr_pack *p) u32 next = init->addr + init->count * init->pitch; u32 addr = init->addr; while (addr < next) { - nv_wr32(priv, addr, init->data); + nvkm_wr32(device, addr, init->data); addr += init->pitch; } } } void -gf100_gr_icmd(struct gf100_gr_priv *priv, const struct gf100_gr_pack *p) +gf100_gr_icmd(struct gf100_gr *gr, const struct gf100_gr_pack *p) { + struct nvkm_device *device = gr->base.engine.subdev.device; const struct gf100_gr_pack *pack; const struct gf100_gr_init *init; u32 data = 0; - nv_wr32(priv, 0x400208, 0x80000000); + nvkm_wr32(device, 0x400208, 0x80000000); pack_for_each_init(init, pack, p) { u32 next = init->addr + init->count * init->pitch; u32 addr = init->addr; if ((pack == p && init == p->init) || data != init->data) { - nv_wr32(priv, 0x400204, init->data); + nvkm_wr32(device, 0x400204, init->data); data = init->data; } while (addr < next) { - nv_wr32(priv, 0x400200, addr); - nv_wait(priv, 0x400700, 0x00000002, 0x00000000); + nvkm_wr32(device, 0x400200, addr); + /** + * Wait for GR to go idle after submitting a + * GO_IDLE bundle + */ + if ((addr & 0xffff) == 0xe100) + gf100_gr_wait_idle(gr); + nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x400700) & 0x00000004)) + break; + ); addr += init->pitch; } } - nv_wr32(priv, 0x400208, 0x00000000); + nvkm_wr32(device, 0x400208, 0x00000000); } void -gf100_gr_mthd(struct gf100_gr_priv *priv, const struct gf100_gr_pack *p) +gf100_gr_mthd(struct gf100_gr *gr, const struct gf100_gr_pack *p) { + struct nvkm_device *device = gr->base.engine.subdev.device; const struct gf100_gr_pack *pack; const struct gf100_gr_init *init; u32 data = 0; @@ -720,79 +802,75 @@ gf100_gr_mthd(struct gf100_gr_priv *priv, const struct gf100_gr_pack *p) u32 addr = init->addr; if ((pack == p && init == p->init) || data != init->data) { - nv_wr32(priv, 0x40448c, init->data); + nvkm_wr32(device, 0x40448c, init->data); data = init->data; } while (addr < next) { - nv_wr32(priv, 0x404488, ctrl | (addr << 14)); + nvkm_wr32(device, 0x404488, ctrl | (addr << 14)); addr += init->pitch; } } } u64 -gf100_gr_units(struct nvkm_gr *gr) +gf100_gr_units(struct nvkm_gr *base) { - struct gf100_gr_priv *priv = (void *)gr; + struct gf100_gr *gr = gf100_gr(base); u64 cfg; - cfg = (u32)priv->gpc_nr; - cfg |= (u32)priv->tpc_total << 8; - cfg |= (u64)priv->rop_nr << 32; + cfg = (u32)gr->gpc_nr; + cfg |= (u32)gr->tpc_total << 8; + cfg |= (u64)gr->rop_nr << 32; return cfg; } -static const struct nvkm_enum gk104_sked_error[] = { - { 7, "CONSTANT_BUFFER_SIZE" }, - { 9, "LOCAL_MEMORY_SIZE_POS" }, - { 10, "LOCAL_MEMORY_SIZE_NEG" }, - { 11, "WARP_CSTACK_SIZE" }, - { 12, "TOTAL_TEMP_SIZE" }, - { 13, "REGISTER_COUNT" }, - { 18, "TOTAL_THREADS" }, - { 20, "PROGRAM_OFFSET" }, - { 21, "SHARED_MEMORY_SIZE" }, - { 25, "SHARED_CONFIG_TOO_SMALL" }, - { 26, "TOTAL_REGISTER_COUNT" }, +static const struct nvkm_bitfield gk104_sked_error[] = { + { 0x00000080, "CONSTANT_BUFFER_SIZE" }, + { 0x00000200, "LOCAL_MEMORY_SIZE_POS" }, + { 0x00000400, "LOCAL_MEMORY_SIZE_NEG" }, + { 0x00000800, "WARP_CSTACK_SIZE" }, + { 0x00001000, "TOTAL_TEMP_SIZE" }, + { 0x00002000, "REGISTER_COUNT" }, + { 0x00040000, "TOTAL_THREADS" }, + { 0x00100000, "PROGRAM_OFFSET" }, + { 0x00200000, "SHARED_MEMORY_SIZE" }, + { 0x02000000, "SHARED_CONFIG_TOO_SMALL" }, + { 0x04000000, "TOTAL_REGISTER_COUNT" }, {} }; -static const struct nvkm_enum gf100_gpc_rop_error[] = { - { 1, "RT_PITCH_OVERRUN" }, - { 4, "RT_WIDTH_OVERRUN" }, - { 5, "RT_HEIGHT_OVERRUN" }, - { 7, "ZETA_STORAGE_TYPE_MISMATCH" }, - { 8, "RT_STORAGE_TYPE_MISMATCH" }, - { 10, "RT_LINEAR_MISMATCH" }, +static const struct nvkm_bitfield gf100_gpc_rop_error[] = { + { 0x00000002, "RT_PITCH_OVERRUN" }, + { 0x00000010, "RT_WIDTH_OVERRUN" }, + { 0x00000020, "RT_HEIGHT_OVERRUN" }, + { 0x00000080, "ZETA_STORAGE_TYPE_MISMATCH" }, + { 0x00000100, "RT_STORAGE_TYPE_MISMATCH" }, + { 0x00000400, "RT_LINEAR_MISMATCH" }, {} }; static void -gf100_gr_trap_gpc_rop(struct gf100_gr_priv *priv, int gpc) +gf100_gr_trap_gpc_rop(struct gf100_gr *gr, int gpc) { + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + char error[128]; u32 trap[4]; - int i; - trap[0] = nv_rd32(priv, GPC_UNIT(gpc, 0x0420)); - trap[1] = nv_rd32(priv, GPC_UNIT(gpc, 0x0434)); - trap[2] = nv_rd32(priv, GPC_UNIT(gpc, 0x0438)); - trap[3] = nv_rd32(priv, GPC_UNIT(gpc, 0x043c)); + trap[0] = nvkm_rd32(device, GPC_UNIT(gpc, 0x0420)) & 0x3fffffff; + trap[1] = nvkm_rd32(device, GPC_UNIT(gpc, 0x0434)); + trap[2] = nvkm_rd32(device, GPC_UNIT(gpc, 0x0438)); + trap[3] = nvkm_rd32(device, GPC_UNIT(gpc, 0x043c)); - nv_error(priv, "GPC%d/PROP trap:", gpc); - for (i = 0; i <= 29; ++i) { - if (!(trap[0] & (1 << i))) - continue; - pr_cont(" "); - nvkm_enum_print(gf100_gpc_rop_error, i); - } - pr_cont("\n"); + nvkm_snprintbf(error, sizeof(error), gf100_gpc_rop_error, trap[0]); - nv_error(priv, "x = %u, y = %u, format = %x, storage type = %x\n", - trap[1] & 0xffff, trap[1] >> 16, (trap[2] >> 8) & 0x3f, - trap[3] & 0xff); - nv_wr32(priv, GPC_UNIT(gpc, 0x0420), 0xc0000000); + nvkm_error(subdev, "GPC%d/PROP trap: %08x [%s] x = %u, y = %u, " + "format = %x, storage type = %x\n", + gpc, trap[0], error, trap[1] & 0xffff, trap[1] >> 16, + (trap[2] >> 8) & 0x3f, trap[3] & 0xff); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0420), 0xc0000000); } static const struct nvkm_enum gf100_mp_warp_error[] = { @@ -804,6 +882,7 @@ static const struct nvkm_enum gf100_mp_warp_error[] = { { 0x0d, "GPR_OUT_OF_BOUNDS" }, { 0x0e, "MEM_OUT_OF_BOUNDS" }, { 0x0f, "UNALIGNED_MEM_ACCESS" }, + { 0x10, "INVALID_ADDR_SPACE" }, { 0x11, "INVALID_PARAM" }, {} }; @@ -815,401 +894,418 @@ static const struct nvkm_bitfield gf100_mp_global_error[] = { }; static void -gf100_gr_trap_mp(struct gf100_gr_priv *priv, int gpc, int tpc) +gf100_gr_trap_mp(struct gf100_gr *gr, int gpc, int tpc) { - u32 werr = nv_rd32(priv, TPC_UNIT(gpc, tpc, 0x648)); - u32 gerr = nv_rd32(priv, TPC_UNIT(gpc, tpc, 0x650)); - - nv_error(priv, "GPC%i/TPC%i/MP trap:", gpc, tpc); - nvkm_bitfield_print(gf100_mp_global_error, gerr); - if (werr) { - pr_cont(" "); - nvkm_enum_print(gf100_mp_warp_error, werr & 0xffff); - } - pr_cont("\n"); - - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x648), 0x00000000); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x650), gerr); + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 werr = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x648)); + u32 gerr = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x650)); + const struct nvkm_enum *warp; + char glob[128]; + + nvkm_snprintbf(glob, sizeof(glob), gf100_mp_global_error, gerr); + warp = nvkm_enum_find(gf100_mp_warp_error, werr & 0xffff); + + nvkm_error(subdev, "GPC%i/TPC%i/MP trap: " + "global %08x [%s] warp %04x [%s]\n", + gpc, tpc, gerr, glob, werr, warp ? warp->name : ""); + + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x648), 0x00000000); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x650), gerr); } static void -gf100_gr_trap_tpc(struct gf100_gr_priv *priv, int gpc, int tpc) +gf100_gr_trap_tpc(struct gf100_gr *gr, int gpc, int tpc) { - u32 stat = nv_rd32(priv, TPC_UNIT(gpc, tpc, 0x0508)); + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 stat = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x0508)); if (stat & 0x00000001) { - u32 trap = nv_rd32(priv, TPC_UNIT(gpc, tpc, 0x0224)); - nv_error(priv, "GPC%d/TPC%d/TEX: 0x%08x\n", gpc, tpc, trap); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x0224), 0xc0000000); + u32 trap = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x0224)); + nvkm_error(subdev, "GPC%d/TPC%d/TEX: %08x\n", gpc, tpc, trap); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x0224), 0xc0000000); stat &= ~0x00000001; } if (stat & 0x00000002) { - gf100_gr_trap_mp(priv, gpc, tpc); + gf100_gr_trap_mp(gr, gpc, tpc); stat &= ~0x00000002; } if (stat & 0x00000004) { - u32 trap = nv_rd32(priv, TPC_UNIT(gpc, tpc, 0x0084)); - nv_error(priv, "GPC%d/TPC%d/POLY: 0x%08x\n", gpc, tpc, trap); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x0084), 0xc0000000); + u32 trap = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x0084)); + nvkm_error(subdev, "GPC%d/TPC%d/POLY: %08x\n", gpc, tpc, trap); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x0084), 0xc0000000); stat &= ~0x00000004; } if (stat & 0x00000008) { - u32 trap = nv_rd32(priv, TPC_UNIT(gpc, tpc, 0x048c)); - nv_error(priv, "GPC%d/TPC%d/L1C: 0x%08x\n", gpc, tpc, trap); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x048c), 0xc0000000); + u32 trap = nvkm_rd32(device, TPC_UNIT(gpc, tpc, 0x048c)); + nvkm_error(subdev, "GPC%d/TPC%d/L1C: %08x\n", gpc, tpc, trap); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x048c), 0xc0000000); stat &= ~0x00000008; } if (stat) { - nv_error(priv, "GPC%d/TPC%d/0x%08x: unknown\n", gpc, tpc, stat); + nvkm_error(subdev, "GPC%d/TPC%d/%08x: unknown\n", gpc, tpc, stat); } } static void -gf100_gr_trap_gpc(struct gf100_gr_priv *priv, int gpc) +gf100_gr_trap_gpc(struct gf100_gr *gr, int gpc) { - u32 stat = nv_rd32(priv, GPC_UNIT(gpc, 0x2c90)); + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 stat = nvkm_rd32(device, GPC_UNIT(gpc, 0x2c90)); int tpc; if (stat & 0x00000001) { - gf100_gr_trap_gpc_rop(priv, gpc); + gf100_gr_trap_gpc_rop(gr, gpc); stat &= ~0x00000001; } if (stat & 0x00000002) { - u32 trap = nv_rd32(priv, GPC_UNIT(gpc, 0x0900)); - nv_error(priv, "GPC%d/ZCULL: 0x%08x\n", gpc, trap); - nv_wr32(priv, GPC_UNIT(gpc, 0x0900), 0xc0000000); + u32 trap = nvkm_rd32(device, GPC_UNIT(gpc, 0x0900)); + nvkm_error(subdev, "GPC%d/ZCULL: %08x\n", gpc, trap); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0900), 0xc0000000); stat &= ~0x00000002; } if (stat & 0x00000004) { - u32 trap = nv_rd32(priv, GPC_UNIT(gpc, 0x1028)); - nv_error(priv, "GPC%d/CCACHE: 0x%08x\n", gpc, trap); - nv_wr32(priv, GPC_UNIT(gpc, 0x1028), 0xc0000000); + u32 trap = nvkm_rd32(device, GPC_UNIT(gpc, 0x1028)); + nvkm_error(subdev, "GPC%d/CCACHE: %08x\n", gpc, trap); + nvkm_wr32(device, GPC_UNIT(gpc, 0x1028), 0xc0000000); stat &= ~0x00000004; } if (stat & 0x00000008) { - u32 trap = nv_rd32(priv, GPC_UNIT(gpc, 0x0824)); - nv_error(priv, "GPC%d/ESETUP: 0x%08x\n", gpc, trap); - nv_wr32(priv, GPC_UNIT(gpc, 0x0824), 0xc0000000); + u32 trap = nvkm_rd32(device, GPC_UNIT(gpc, 0x0824)); + nvkm_error(subdev, "GPC%d/ESETUP: %08x\n", gpc, trap); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0824), 0xc0000000); stat &= ~0x00000009; } - for (tpc = 0; tpc < priv->tpc_nr[gpc]; tpc++) { + for (tpc = 0; tpc < gr->tpc_nr[gpc]; tpc++) { u32 mask = 0x00010000 << tpc; if (stat & mask) { - gf100_gr_trap_tpc(priv, gpc, tpc); - nv_wr32(priv, GPC_UNIT(gpc, 0x2c90), mask); + gf100_gr_trap_tpc(gr, gpc, tpc); + nvkm_wr32(device, GPC_UNIT(gpc, 0x2c90), mask); stat &= ~mask; } } if (stat) { - nv_error(priv, "GPC%d/0x%08x: unknown\n", gpc, stat); + nvkm_error(subdev, "GPC%d/%08x: unknown\n", gpc, stat); } } static void -gf100_gr_trap_intr(struct gf100_gr_priv *priv) +gf100_gr_trap_intr(struct gf100_gr *gr) { - u32 trap = nv_rd32(priv, 0x400108); - int rop, gpc, i; + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 trap = nvkm_rd32(device, 0x400108); + int rop, gpc; if (trap & 0x00000001) { - u32 stat = nv_rd32(priv, 0x404000); - nv_error(priv, "DISPATCH 0x%08x\n", stat); - nv_wr32(priv, 0x404000, 0xc0000000); - nv_wr32(priv, 0x400108, 0x00000001); + u32 stat = nvkm_rd32(device, 0x404000); + nvkm_error(subdev, "DISPATCH %08x\n", stat); + nvkm_wr32(device, 0x404000, 0xc0000000); + nvkm_wr32(device, 0x400108, 0x00000001); trap &= ~0x00000001; } if (trap & 0x00000002) { - u32 stat = nv_rd32(priv, 0x404600); - nv_error(priv, "M2MF 0x%08x\n", stat); - nv_wr32(priv, 0x404600, 0xc0000000); - nv_wr32(priv, 0x400108, 0x00000002); + u32 stat = nvkm_rd32(device, 0x404600); + nvkm_error(subdev, "M2MF %08x\n", stat); + nvkm_wr32(device, 0x404600, 0xc0000000); + nvkm_wr32(device, 0x400108, 0x00000002); trap &= ~0x00000002; } if (trap & 0x00000008) { - u32 stat = nv_rd32(priv, 0x408030); - nv_error(priv, "CCACHE 0x%08x\n", stat); - nv_wr32(priv, 0x408030, 0xc0000000); - nv_wr32(priv, 0x400108, 0x00000008); + u32 stat = nvkm_rd32(device, 0x408030); + nvkm_error(subdev, "CCACHE %08x\n", stat); + nvkm_wr32(device, 0x408030, 0xc0000000); + nvkm_wr32(device, 0x400108, 0x00000008); trap &= ~0x00000008; } if (trap & 0x00000010) { - u32 stat = nv_rd32(priv, 0x405840); - nv_error(priv, "SHADER 0x%08x\n", stat); - nv_wr32(priv, 0x405840, 0xc0000000); - nv_wr32(priv, 0x400108, 0x00000010); + u32 stat = nvkm_rd32(device, 0x405840); + nvkm_error(subdev, "SHADER %08x\n", stat); + nvkm_wr32(device, 0x405840, 0xc0000000); + nvkm_wr32(device, 0x400108, 0x00000010); trap &= ~0x00000010; } if (trap & 0x00000040) { - u32 stat = nv_rd32(priv, 0x40601c); - nv_error(priv, "UNK6 0x%08x\n", stat); - nv_wr32(priv, 0x40601c, 0xc0000000); - nv_wr32(priv, 0x400108, 0x00000040); + u32 stat = nvkm_rd32(device, 0x40601c); + nvkm_error(subdev, "UNK6 %08x\n", stat); + nvkm_wr32(device, 0x40601c, 0xc0000000); + nvkm_wr32(device, 0x400108, 0x00000040); trap &= ~0x00000040; } if (trap & 0x00000080) { - u32 stat = nv_rd32(priv, 0x404490); - nv_error(priv, "MACRO 0x%08x\n", stat); - nv_wr32(priv, 0x404490, 0xc0000000); - nv_wr32(priv, 0x400108, 0x00000080); + u32 stat = nvkm_rd32(device, 0x404490); + nvkm_error(subdev, "MACRO %08x\n", stat); + nvkm_wr32(device, 0x404490, 0xc0000000); + nvkm_wr32(device, 0x400108, 0x00000080); trap &= ~0x00000080; } if (trap & 0x00000100) { - u32 stat = nv_rd32(priv, 0x407020); + u32 stat = nvkm_rd32(device, 0x407020) & 0x3fffffff; + char sked[128]; - nv_error(priv, "SKED:"); - for (i = 0; i <= 29; ++i) { - if (!(stat & (1 << i))) - continue; - pr_cont(" "); - nvkm_enum_print(gk104_sked_error, i); - } - pr_cont("\n"); + nvkm_snprintbf(sked, sizeof(sked), gk104_sked_error, stat); + nvkm_error(subdev, "SKED: %08x [%s]\n", stat, sked); - if (stat & 0x3fffffff) - nv_wr32(priv, 0x407020, 0x40000000); - nv_wr32(priv, 0x400108, 0x00000100); + if (stat) + nvkm_wr32(device, 0x407020, 0x40000000); + nvkm_wr32(device, 0x400108, 0x00000100); trap &= ~0x00000100; } if (trap & 0x01000000) { - u32 stat = nv_rd32(priv, 0x400118); - for (gpc = 0; stat && gpc < priv->gpc_nr; gpc++) { + u32 stat = nvkm_rd32(device, 0x400118); + for (gpc = 0; stat && gpc < gr->gpc_nr; gpc++) { u32 mask = 0x00000001 << gpc; if (stat & mask) { - gf100_gr_trap_gpc(priv, gpc); - nv_wr32(priv, 0x400118, mask); + gf100_gr_trap_gpc(gr, gpc); + nvkm_wr32(device, 0x400118, mask); stat &= ~mask; } } - nv_wr32(priv, 0x400108, 0x01000000); + nvkm_wr32(device, 0x400108, 0x01000000); trap &= ~0x01000000; } if (trap & 0x02000000) { - for (rop = 0; rop < priv->rop_nr; rop++) { - u32 statz = nv_rd32(priv, ROP_UNIT(rop, 0x070)); - u32 statc = nv_rd32(priv, ROP_UNIT(rop, 0x144)); - nv_error(priv, "ROP%d 0x%08x 0x%08x\n", + for (rop = 0; rop < gr->rop_nr; rop++) { + u32 statz = nvkm_rd32(device, ROP_UNIT(rop, 0x070)); + u32 statc = nvkm_rd32(device, ROP_UNIT(rop, 0x144)); + nvkm_error(subdev, "ROP%d %08x %08x\n", rop, statz, statc); - nv_wr32(priv, ROP_UNIT(rop, 0x070), 0xc0000000); - nv_wr32(priv, ROP_UNIT(rop, 0x144), 0xc0000000); + nvkm_wr32(device, ROP_UNIT(rop, 0x070), 0xc0000000); + nvkm_wr32(device, ROP_UNIT(rop, 0x144), 0xc0000000); } - nv_wr32(priv, 0x400108, 0x02000000); + nvkm_wr32(device, 0x400108, 0x02000000); trap &= ~0x02000000; } if (trap) { - nv_error(priv, "TRAP UNHANDLED 0x%08x\n", trap); - nv_wr32(priv, 0x400108, trap); + nvkm_error(subdev, "TRAP UNHANDLED %08x\n", trap); + nvkm_wr32(device, 0x400108, trap); } } static void -gf100_gr_ctxctl_debug_unit(struct gf100_gr_priv *priv, u32 base) +gf100_gr_ctxctl_debug_unit(struct gf100_gr *gr, u32 base) { - nv_error(priv, "%06x - done 0x%08x\n", base, - nv_rd32(priv, base + 0x400)); - nv_error(priv, "%06x - stat 0x%08x 0x%08x 0x%08x 0x%08x\n", base, - nv_rd32(priv, base + 0x800), nv_rd32(priv, base + 0x804), - nv_rd32(priv, base + 0x808), nv_rd32(priv, base + 0x80c)); - nv_error(priv, "%06x - stat 0x%08x 0x%08x 0x%08x 0x%08x\n", base, - nv_rd32(priv, base + 0x810), nv_rd32(priv, base + 0x814), - nv_rd32(priv, base + 0x818), nv_rd32(priv, base + 0x81c)); + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + nvkm_error(subdev, "%06x - done %08x\n", base, + nvkm_rd32(device, base + 0x400)); + nvkm_error(subdev, "%06x - stat %08x %08x %08x %08x\n", base, + nvkm_rd32(device, base + 0x800), + nvkm_rd32(device, base + 0x804), + nvkm_rd32(device, base + 0x808), + nvkm_rd32(device, base + 0x80c)); + nvkm_error(subdev, "%06x - stat %08x %08x %08x %08x\n", base, + nvkm_rd32(device, base + 0x810), + nvkm_rd32(device, base + 0x814), + nvkm_rd32(device, base + 0x818), + nvkm_rd32(device, base + 0x81c)); } void -gf100_gr_ctxctl_debug(struct gf100_gr_priv *priv) +gf100_gr_ctxctl_debug(struct gf100_gr *gr) { - u32 gpcnr = nv_rd32(priv, 0x409604) & 0xffff; + struct nvkm_device *device = gr->base.engine.subdev.device; + u32 gpcnr = nvkm_rd32(device, 0x409604) & 0xffff; u32 gpc; - gf100_gr_ctxctl_debug_unit(priv, 0x409000); + gf100_gr_ctxctl_debug_unit(gr, 0x409000); for (gpc = 0; gpc < gpcnr; gpc++) - gf100_gr_ctxctl_debug_unit(priv, 0x502000 + (gpc * 0x8000)); + gf100_gr_ctxctl_debug_unit(gr, 0x502000 + (gpc * 0x8000)); } static void -gf100_gr_ctxctl_isr(struct gf100_gr_priv *priv) +gf100_gr_ctxctl_isr(struct gf100_gr *gr) { - u32 stat = nv_rd32(priv, 0x409c18); + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 stat = nvkm_rd32(device, 0x409c18); if (stat & 0x00000001) { - u32 code = nv_rd32(priv, 0x409814); + u32 code = nvkm_rd32(device, 0x409814); if (code == E_BAD_FWMTHD) { - u32 class = nv_rd32(priv, 0x409808); - u32 addr = nv_rd32(priv, 0x40980c); + u32 class = nvkm_rd32(device, 0x409808); + u32 addr = nvkm_rd32(device, 0x40980c); u32 subc = (addr & 0x00070000) >> 16; u32 mthd = (addr & 0x00003ffc); - u32 data = nv_rd32(priv, 0x409810); + u32 data = nvkm_rd32(device, 0x409810); - nv_error(priv, "FECS MTHD subc %d class 0x%04x " - "mthd 0x%04x data 0x%08x\n", - subc, class, mthd, data); + nvkm_error(subdev, "FECS MTHD subc %d class %04x " + "mthd %04x data %08x\n", + subc, class, mthd, data); - nv_wr32(priv, 0x409c20, 0x00000001); + nvkm_wr32(device, 0x409c20, 0x00000001); stat &= ~0x00000001; } else { - nv_error(priv, "FECS ucode error %d\n", code); + nvkm_error(subdev, "FECS ucode error %d\n", code); } } if (stat & 0x00080000) { - nv_error(priv, "FECS watchdog timeout\n"); - gf100_gr_ctxctl_debug(priv); - nv_wr32(priv, 0x409c20, 0x00080000); + nvkm_error(subdev, "FECS watchdog timeout\n"); + gf100_gr_ctxctl_debug(gr); + nvkm_wr32(device, 0x409c20, 0x00080000); stat &= ~0x00080000; } if (stat) { - nv_error(priv, "FECS 0x%08x\n", stat); - gf100_gr_ctxctl_debug(priv); - nv_wr32(priv, 0x409c20, stat); + nvkm_error(subdev, "FECS %08x\n", stat); + gf100_gr_ctxctl_debug(gr); + nvkm_wr32(device, 0x409c20, stat); } } static void -gf100_gr_intr(struct nvkm_subdev *subdev) +gf100_gr_intr(struct nvkm_gr *base) { - struct nvkm_fifo *pfifo = nvkm_fifo(subdev); - struct nvkm_engine *engine = nv_engine(subdev); - struct nvkm_object *engctx; - struct nvkm_handle *handle; - struct gf100_gr_priv *priv = (void *)subdev; - u64 inst = nv_rd32(priv, 0x409b00) & 0x0fffffff; - u32 stat = nv_rd32(priv, 0x400100); - u32 addr = nv_rd32(priv, 0x400704); + struct gf100_gr *gr = gf100_gr(base); + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + struct nvkm_fifo_chan *chan; + unsigned long flags; + u64 inst = nvkm_rd32(device, 0x409b00) & 0x0fffffff; + u32 stat = nvkm_rd32(device, 0x400100); + u32 addr = nvkm_rd32(device, 0x400704); u32 mthd = (addr & 0x00003ffc); u32 subc = (addr & 0x00070000) >> 16; - u32 data = nv_rd32(priv, 0x400708); - u32 code = nv_rd32(priv, 0x400110); + u32 data = nvkm_rd32(device, 0x400708); + u32 code = nvkm_rd32(device, 0x400110); u32 class; - int chid; + const char *name = "unknown"; + int chid = -1; - if (nv_device(priv)->card_type < NV_E0 || subc < 4) - class = nv_rd32(priv, 0x404200 + (subc * 4)); + chan = nvkm_fifo_chan_inst(device->fifo, (u64)inst << 12, &flags); + if (chan) { + name = chan->object.client->name; + chid = chan->chid; + } + + if (device->card_type < NV_E0 || subc < 4) + class = nvkm_rd32(device, 0x404200 + (subc * 4)); else class = 0x0000; - engctx = nvkm_engctx_get(engine, inst); - chid = pfifo->chid(pfifo, engctx); - if (stat & 0x00000001) { /* * notifier interrupt, only needed for cyclestats * can be safely ignored */ - nv_wr32(priv, 0x400100, 0x00000001); + nvkm_wr32(device, 0x400100, 0x00000001); stat &= ~0x00000001; } if (stat & 0x00000010) { - handle = nvkm_handle_get_class(engctx, class); - if (!handle || nv_call(handle->object, mthd, data)) { - nv_error(priv, - "ILLEGAL_MTHD ch %d [0x%010llx %s] subc %d class 0x%04x mthd 0x%04x data 0x%08x\n", - chid, inst << 12, nvkm_client_name(engctx), - subc, class, mthd, data); + if (!gf100_gr_mthd_sw(device, class, mthd, data)) { + nvkm_error(subdev, "ILLEGAL_MTHD ch %d [%010llx %s] " + "subc %d class %04x mthd %04x data %08x\n", + chid, inst << 12, name, subc, + class, mthd, data); } - nvkm_handle_put(handle); - nv_wr32(priv, 0x400100, 0x00000010); + nvkm_wr32(device, 0x400100, 0x00000010); stat &= ~0x00000010; } if (stat & 0x00000020) { - nv_error(priv, - "ILLEGAL_CLASS ch %d [0x%010llx %s] subc %d class 0x%04x mthd 0x%04x data 0x%08x\n", - chid, inst << 12, nvkm_client_name(engctx), subc, - class, mthd, data); - nv_wr32(priv, 0x400100, 0x00000020); + nvkm_error(subdev, "ILLEGAL_CLASS ch %d [%010llx %s] " + "subc %d class %04x mthd %04x data %08x\n", + chid, inst << 12, name, subc, class, mthd, data); + nvkm_wr32(device, 0x400100, 0x00000020); stat &= ~0x00000020; } if (stat & 0x00100000) { - nv_error(priv, "DATA_ERROR ["); - nvkm_enum_print(nv50_data_error_names, code); - pr_cont("] ch %d [0x%010llx %s] subc %d class 0x%04x mthd 0x%04x data 0x%08x\n", - chid, inst << 12, nvkm_client_name(engctx), subc, - class, mthd, data); - nv_wr32(priv, 0x400100, 0x00100000); + const struct nvkm_enum *en = + nvkm_enum_find(nv50_data_error_names, code); + nvkm_error(subdev, "DATA_ERROR %08x [%s] ch %d [%010llx %s] " + "subc %d class %04x mthd %04x data %08x\n", + code, en ? en->name : "", chid, inst << 12, + name, subc, class, mthd, data); + nvkm_wr32(device, 0x400100, 0x00100000); stat &= ~0x00100000; } if (stat & 0x00200000) { - nv_error(priv, "TRAP ch %d [0x%010llx %s]\n", chid, inst << 12, - nvkm_client_name(engctx)); - gf100_gr_trap_intr(priv); - nv_wr32(priv, 0x400100, 0x00200000); + nvkm_error(subdev, "TRAP ch %d [%010llx %s]\n", + chid, inst << 12, name); + gf100_gr_trap_intr(gr); + nvkm_wr32(device, 0x400100, 0x00200000); stat &= ~0x00200000; } if (stat & 0x00080000) { - gf100_gr_ctxctl_isr(priv); - nv_wr32(priv, 0x400100, 0x00080000); + gf100_gr_ctxctl_isr(gr); + nvkm_wr32(device, 0x400100, 0x00080000); stat &= ~0x00080000; } if (stat) { - nv_error(priv, "unknown stat 0x%08x\n", stat); - nv_wr32(priv, 0x400100, stat); + nvkm_error(subdev, "intr %08x\n", stat); + nvkm_wr32(device, 0x400100, stat); } - nv_wr32(priv, 0x400500, 0x00010001); - nvkm_engctx_put(engctx); + nvkm_wr32(device, 0x400500, 0x00010001); + nvkm_fifo_chan_put(device->fifo, flags, &chan); } void -gf100_gr_init_fw(struct gf100_gr_priv *priv, u32 fuc_base, +gf100_gr_init_fw(struct gf100_gr *gr, u32 fuc_base, struct gf100_gr_fuc *code, struct gf100_gr_fuc *data) { + struct nvkm_device *device = gr->base.engine.subdev.device; int i; - nv_wr32(priv, fuc_base + 0x01c0, 0x01000000); + nvkm_wr32(device, fuc_base + 0x01c0, 0x01000000); for (i = 0; i < data->size / 4; i++) - nv_wr32(priv, fuc_base + 0x01c4, data->data[i]); + nvkm_wr32(device, fuc_base + 0x01c4, data->data[i]); - nv_wr32(priv, fuc_base + 0x0180, 0x01000000); + nvkm_wr32(device, fuc_base + 0x0180, 0x01000000); for (i = 0; i < code->size / 4; i++) { if ((i & 0x3f) == 0) - nv_wr32(priv, fuc_base + 0x0188, i >> 6); - nv_wr32(priv, fuc_base + 0x0184, code->data[i]); + nvkm_wr32(device, fuc_base + 0x0188, i >> 6); + nvkm_wr32(device, fuc_base + 0x0184, code->data[i]); } /* code must be padded to 0x40 words */ for (; i & 0x3f; i++) - nv_wr32(priv, fuc_base + 0x0184, 0); + nvkm_wr32(device, fuc_base + 0x0184, 0); } static void -gf100_gr_init_csdata(struct gf100_gr_priv *priv, +gf100_gr_init_csdata(struct gf100_gr *gr, const struct gf100_gr_pack *pack, u32 falcon, u32 starstar, u32 base) { + struct nvkm_device *device = gr->base.engine.subdev.device; const struct gf100_gr_pack *iter; const struct gf100_gr_init *init; u32 addr = ~0, prev = ~0, xfer = 0; u32 star, temp; - nv_wr32(priv, falcon + 0x01c0, 0x02000000 + starstar); - star = nv_rd32(priv, falcon + 0x01c4); - temp = nv_rd32(priv, falcon + 0x01c4); + nvkm_wr32(device, falcon + 0x01c0, 0x02000000 + starstar); + star = nvkm_rd32(device, falcon + 0x01c4); + temp = nvkm_rd32(device, falcon + 0x01c4); if (temp > star) star = temp; - nv_wr32(priv, falcon + 0x01c0, 0x01000000 + star); + nvkm_wr32(device, falcon + 0x01c0, 0x01000000 + star); pack_for_each_init(init, iter, pack) { u32 head = init->addr - base; @@ -1218,7 +1314,7 @@ gf100_gr_init_csdata(struct gf100_gr_priv *priv, if (head != prev + 4 || xfer >= 32) { if (xfer) { u32 data = ((--xfer << 26) | addr); - nv_wr32(priv, falcon + 0x01c4, data); + nvkm_wr32(device, falcon + 0x01c4, data); star += 4; } addr = head; @@ -1230,157 +1326,166 @@ gf100_gr_init_csdata(struct gf100_gr_priv *priv, } } - nv_wr32(priv, falcon + 0x01c4, (--xfer << 26) | addr); - nv_wr32(priv, falcon + 0x01c0, 0x01000004 + starstar); - nv_wr32(priv, falcon + 0x01c4, star + 4); + nvkm_wr32(device, falcon + 0x01c4, (--xfer << 26) | addr); + nvkm_wr32(device, falcon + 0x01c0, 0x01000004 + starstar); + nvkm_wr32(device, falcon + 0x01c4, star + 4); } int -gf100_gr_init_ctxctl(struct gf100_gr_priv *priv) +gf100_gr_init_ctxctl(struct gf100_gr *gr) { - struct gf100_gr_oclass *oclass = (void *)nv_object(priv)->oclass; - struct gf100_grctx_oclass *cclass = (void *)nv_engine(priv)->cclass; + const struct gf100_grctx_func *grctx = gr->func->grctx; + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; int i; - if (priv->firmware) { + if (gr->firmware) { /* load fuc microcode */ - nvkm_mc(priv)->unk260(nvkm_mc(priv), 0); - gf100_gr_init_fw(priv, 0x409000, &priv->fuc409c, - &priv->fuc409d); - gf100_gr_init_fw(priv, 0x41a000, &priv->fuc41ac, - &priv->fuc41ad); - nvkm_mc(priv)->unk260(nvkm_mc(priv), 1); + nvkm_mc_unk260(device->mc, 0); + gf100_gr_init_fw(gr, 0x409000, &gr->fuc409c, &gr->fuc409d); + gf100_gr_init_fw(gr, 0x41a000, &gr->fuc41ac, &gr->fuc41ad); + nvkm_mc_unk260(device->mc, 1); /* start both of them running */ - nv_wr32(priv, 0x409840, 0xffffffff); - nv_wr32(priv, 0x41a10c, 0x00000000); - nv_wr32(priv, 0x40910c, 0x00000000); - nv_wr32(priv, 0x41a100, 0x00000002); - nv_wr32(priv, 0x409100, 0x00000002); - if (!nv_wait(priv, 0x409800, 0x00000001, 0x00000001)) - nv_warn(priv, "0x409800 wait failed\n"); - - nv_wr32(priv, 0x409840, 0xffffffff); - nv_wr32(priv, 0x409500, 0x7fffffff); - nv_wr32(priv, 0x409504, 0x00000021); - - nv_wr32(priv, 0x409840, 0xffffffff); - nv_wr32(priv, 0x409500, 0x00000000); - nv_wr32(priv, 0x409504, 0x00000010); - if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) { - nv_error(priv, "fuc09 req 0x10 timeout\n"); + nvkm_wr32(device, 0x409840, 0xffffffff); + nvkm_wr32(device, 0x41a10c, 0x00000000); + nvkm_wr32(device, 0x40910c, 0x00000000); + nvkm_wr32(device, 0x41a100, 0x00000002); + nvkm_wr32(device, 0x409100, 0x00000002); + if (nvkm_msec(device, 2000, + if (nvkm_rd32(device, 0x409800) & 0x00000001) + break; + ) < 0) return -EBUSY; - } - priv->size = nv_rd32(priv, 0x409800); - nv_wr32(priv, 0x409840, 0xffffffff); - nv_wr32(priv, 0x409500, 0x00000000); - nv_wr32(priv, 0x409504, 0x00000016); - if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) { - nv_error(priv, "fuc09 req 0x16 timeout\n"); + nvkm_wr32(device, 0x409840, 0xffffffff); + nvkm_wr32(device, 0x409500, 0x7fffffff); + nvkm_wr32(device, 0x409504, 0x00000021); + + nvkm_wr32(device, 0x409840, 0xffffffff); + nvkm_wr32(device, 0x409500, 0x00000000); + nvkm_wr32(device, 0x409504, 0x00000010); + if (nvkm_msec(device, 2000, + if ((gr->size = nvkm_rd32(device, 0x409800))) + break; + ) < 0) return -EBUSY; - } - nv_wr32(priv, 0x409840, 0xffffffff); - nv_wr32(priv, 0x409500, 0x00000000); - nv_wr32(priv, 0x409504, 0x00000025); - if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) { - nv_error(priv, "fuc09 req 0x25 timeout\n"); + nvkm_wr32(device, 0x409840, 0xffffffff); + nvkm_wr32(device, 0x409500, 0x00000000); + nvkm_wr32(device, 0x409504, 0x00000016); + if (nvkm_msec(device, 2000, + if (nvkm_rd32(device, 0x409800)) + break; + ) < 0) + return -EBUSY; + + nvkm_wr32(device, 0x409840, 0xffffffff); + nvkm_wr32(device, 0x409500, 0x00000000); + nvkm_wr32(device, 0x409504, 0x00000025); + if (nvkm_msec(device, 2000, + if (nvkm_rd32(device, 0x409800)) + break; + ) < 0) return -EBUSY; - } - if (nv_device(priv)->chipset >= 0xe0) { - nv_wr32(priv, 0x409800, 0x00000000); - nv_wr32(priv, 0x409500, 0x00000001); - nv_wr32(priv, 0x409504, 0x00000030); - if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) { - nv_error(priv, "fuc09 req 0x30 timeout\n"); + if (device->chipset >= 0xe0) { + nvkm_wr32(device, 0x409800, 0x00000000); + nvkm_wr32(device, 0x409500, 0x00000001); + nvkm_wr32(device, 0x409504, 0x00000030); + if (nvkm_msec(device, 2000, + if (nvkm_rd32(device, 0x409800)) + break; + ) < 0) return -EBUSY; - } - nv_wr32(priv, 0x409810, 0xb00095c8); - nv_wr32(priv, 0x409800, 0x00000000); - nv_wr32(priv, 0x409500, 0x00000001); - nv_wr32(priv, 0x409504, 0x00000031); - if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) { - nv_error(priv, "fuc09 req 0x31 timeout\n"); + nvkm_wr32(device, 0x409810, 0xb00095c8); + nvkm_wr32(device, 0x409800, 0x00000000); + nvkm_wr32(device, 0x409500, 0x00000001); + nvkm_wr32(device, 0x409504, 0x00000031); + if (nvkm_msec(device, 2000, + if (nvkm_rd32(device, 0x409800)) + break; + ) < 0) return -EBUSY; - } - nv_wr32(priv, 0x409810, 0x00080420); - nv_wr32(priv, 0x409800, 0x00000000); - nv_wr32(priv, 0x409500, 0x00000001); - nv_wr32(priv, 0x409504, 0x00000032); - if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) { - nv_error(priv, "fuc09 req 0x32 timeout\n"); + nvkm_wr32(device, 0x409810, 0x00080420); + nvkm_wr32(device, 0x409800, 0x00000000); + nvkm_wr32(device, 0x409500, 0x00000001); + nvkm_wr32(device, 0x409504, 0x00000032); + if (nvkm_msec(device, 2000, + if (nvkm_rd32(device, 0x409800)) + break; + ) < 0) return -EBUSY; - } - nv_wr32(priv, 0x409614, 0x00000070); - nv_wr32(priv, 0x409614, 0x00000770); - nv_wr32(priv, 0x40802c, 0x00000001); + nvkm_wr32(device, 0x409614, 0x00000070); + nvkm_wr32(device, 0x409614, 0x00000770); + nvkm_wr32(device, 0x40802c, 0x00000001); } - if (priv->data == NULL) { - int ret = gf100_grctx_generate(priv); + if (gr->data == NULL) { + int ret = gf100_grctx_generate(gr); if (ret) { - nv_error(priv, "failed to construct context\n"); + nvkm_error(subdev, "failed to construct context\n"); return ret; } } return 0; } else - if (!oclass->fecs.ucode) { + if (!gr->func->fecs.ucode) { return -ENOSYS; } /* load HUB microcode */ - nvkm_mc(priv)->unk260(nvkm_mc(priv), 0); - nv_wr32(priv, 0x4091c0, 0x01000000); - for (i = 0; i < oclass->fecs.ucode->data.size / 4; i++) - nv_wr32(priv, 0x4091c4, oclass->fecs.ucode->data.data[i]); + nvkm_mc_unk260(device->mc, 0); + nvkm_wr32(device, 0x4091c0, 0x01000000); + for (i = 0; i < gr->func->fecs.ucode->data.size / 4; i++) + nvkm_wr32(device, 0x4091c4, gr->func->fecs.ucode->data.data[i]); - nv_wr32(priv, 0x409180, 0x01000000); - for (i = 0; i < oclass->fecs.ucode->code.size / 4; i++) { + nvkm_wr32(device, 0x409180, 0x01000000); + for (i = 0; i < gr->func->fecs.ucode->code.size / 4; i++) { if ((i & 0x3f) == 0) - nv_wr32(priv, 0x409188, i >> 6); - nv_wr32(priv, 0x409184, oclass->fecs.ucode->code.data[i]); + nvkm_wr32(device, 0x409188, i >> 6); + nvkm_wr32(device, 0x409184, gr->func->fecs.ucode->code.data[i]); } /* load GPC microcode */ - nv_wr32(priv, 0x41a1c0, 0x01000000); - for (i = 0; i < oclass->gpccs.ucode->data.size / 4; i++) - nv_wr32(priv, 0x41a1c4, oclass->gpccs.ucode->data.data[i]); + nvkm_wr32(device, 0x41a1c0, 0x01000000); + for (i = 0; i < gr->func->gpccs.ucode->data.size / 4; i++) + nvkm_wr32(device, 0x41a1c4, gr->func->gpccs.ucode->data.data[i]); - nv_wr32(priv, 0x41a180, 0x01000000); - for (i = 0; i < oclass->gpccs.ucode->code.size / 4; i++) { + nvkm_wr32(device, 0x41a180, 0x01000000); + for (i = 0; i < gr->func->gpccs.ucode->code.size / 4; i++) { if ((i & 0x3f) == 0) - nv_wr32(priv, 0x41a188, i >> 6); - nv_wr32(priv, 0x41a184, oclass->gpccs.ucode->code.data[i]); + nvkm_wr32(device, 0x41a188, i >> 6); + nvkm_wr32(device, 0x41a184, gr->func->gpccs.ucode->code.data[i]); } - nvkm_mc(priv)->unk260(nvkm_mc(priv), 1); + nvkm_mc_unk260(device->mc, 1); /* load register lists */ - gf100_gr_init_csdata(priv, cclass->hub, 0x409000, 0x000, 0x000000); - gf100_gr_init_csdata(priv, cclass->gpc, 0x41a000, 0x000, 0x418000); - gf100_gr_init_csdata(priv, cclass->tpc, 0x41a000, 0x004, 0x419800); - gf100_gr_init_csdata(priv, cclass->ppc, 0x41a000, 0x008, 0x41be00); + gf100_gr_init_csdata(gr, grctx->hub, 0x409000, 0x000, 0x000000); + gf100_gr_init_csdata(gr, grctx->gpc, 0x41a000, 0x000, 0x418000); + gf100_gr_init_csdata(gr, grctx->tpc, 0x41a000, 0x004, 0x419800); + gf100_gr_init_csdata(gr, grctx->ppc, 0x41a000, 0x008, 0x41be00); /* start HUB ucode running, it'll init the GPCs */ - nv_wr32(priv, 0x40910c, 0x00000000); - nv_wr32(priv, 0x409100, 0x00000002); - if (!nv_wait(priv, 0x409800, 0x80000000, 0x80000000)) { - nv_error(priv, "HUB_INIT timed out\n"); - gf100_gr_ctxctl_debug(priv); + nvkm_wr32(device, 0x40910c, 0x00000000); + nvkm_wr32(device, 0x409100, 0x00000002); + if (nvkm_msec(device, 2000, + if (nvkm_rd32(device, 0x409800) & 0x80000000) + break; + ) < 0) { + gf100_gr_ctxctl_debug(gr); return -EBUSY; } - priv->size = nv_rd32(priv, 0x409804); - if (priv->data == NULL) { - int ret = gf100_grctx_generate(priv); + gr->size = nvkm_rd32(device, 0x409804); + if (gr->data == NULL) { + int ret = gf100_grctx_generate(gr); if (ret) { - nv_error(priv, "failed to construct context\n"); + nvkm_error(subdev, "failed to construct context\n"); return ret; } } @@ -1388,143 +1493,162 @@ gf100_gr_init_ctxctl(struct gf100_gr_priv *priv) return 0; } -int -gf100_gr_init(struct nvkm_object *object) +static int +gf100_gr_oneinit(struct nvkm_gr *base) { - struct gf100_gr_oclass *oclass = (void *)object->oclass; - struct gf100_gr_priv *priv = (void *)object; - const u32 magicgpc918 = DIV_ROUND_UP(0x00800000, priv->tpc_total); - u32 data[TPC_MAX / 8] = {}; - u8 tpcnr[GPC_MAX]; - int gpc, tpc, rop; - int ret, i; + struct gf100_gr *gr = gf100_gr(base); + struct nvkm_device *device = gr->base.engine.subdev.device; + int ret, i, j; + + nvkm_pmu_pgob(device->pmu, false); - ret = nvkm_gr_init(&priv->base); + ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, 0x1000, 256, false, + &gr->unk4188b4); if (ret) return ret; - nv_wr32(priv, GPC_BCAST(0x0880), 0x00000000); - nv_wr32(priv, GPC_BCAST(0x08a4), 0x00000000); - nv_wr32(priv, GPC_BCAST(0x0888), 0x00000000); - nv_wr32(priv, GPC_BCAST(0x088c), 0x00000000); - nv_wr32(priv, GPC_BCAST(0x0890), 0x00000000); - nv_wr32(priv, GPC_BCAST(0x0894), 0x00000000); - nv_wr32(priv, GPC_BCAST(0x08b4), priv->unk4188b4->addr >> 8); - nv_wr32(priv, GPC_BCAST(0x08b8), priv->unk4188b8->addr >> 8); - - gf100_gr_mmio(priv, oclass->mmio); - - memcpy(tpcnr, priv->tpc_nr, sizeof(priv->tpc_nr)); - for (i = 0, gpc = -1; i < priv->tpc_total; i++) { - do { - gpc = (gpc + 1) % priv->gpc_nr; - } while (!tpcnr[gpc]); - tpc = priv->tpc_nr[gpc] - tpcnr[gpc]--; - - data[i / 8] |= tpc << ((i % 8) * 4); - } - - nv_wr32(priv, GPC_BCAST(0x0980), data[0]); - nv_wr32(priv, GPC_BCAST(0x0984), data[1]); - nv_wr32(priv, GPC_BCAST(0x0988), data[2]); - nv_wr32(priv, GPC_BCAST(0x098c), data[3]); - - for (gpc = 0; gpc < priv->gpc_nr; gpc++) { - nv_wr32(priv, GPC_UNIT(gpc, 0x0914), - priv->magic_not_rop_nr << 8 | priv->tpc_nr[gpc]); - nv_wr32(priv, GPC_UNIT(gpc, 0x0910), 0x00040000 | - priv->tpc_total); - nv_wr32(priv, GPC_UNIT(gpc, 0x0918), magicgpc918); - } + ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, 0x1000, 256, false, + &gr->unk4188b8); + if (ret) + return ret; - if (nv_device(priv)->chipset != 0xd7) - nv_wr32(priv, GPC_BCAST(0x1bd4), magicgpc918); - else - nv_wr32(priv, GPC_BCAST(0x3fd4), magicgpc918); - - nv_wr32(priv, GPC_BCAST(0x08ac), nv_rd32(priv, 0x100800)); - - nv_wr32(priv, 0x400500, 0x00010001); - - nv_wr32(priv, 0x400100, 0xffffffff); - nv_wr32(priv, 0x40013c, 0xffffffff); - - nv_wr32(priv, 0x409c24, 0x000f0000); - nv_wr32(priv, 0x404000, 0xc0000000); - nv_wr32(priv, 0x404600, 0xc0000000); - nv_wr32(priv, 0x408030, 0xc0000000); - nv_wr32(priv, 0x40601c, 0xc0000000); - nv_wr32(priv, 0x404490, 0xc0000000); - nv_wr32(priv, 0x406018, 0xc0000000); - nv_wr32(priv, 0x405840, 0xc0000000); - nv_wr32(priv, 0x405844, 0x00ffffff); - nv_mask(priv, 0x419cc0, 0x00000008, 0x00000008); - nv_mask(priv, 0x419eb4, 0x00001000, 0x00001000); - - for (gpc = 0; gpc < priv->gpc_nr; gpc++) { - nv_wr32(priv, GPC_UNIT(gpc, 0x0420), 0xc0000000); - nv_wr32(priv, GPC_UNIT(gpc, 0x0900), 0xc0000000); - nv_wr32(priv, GPC_UNIT(gpc, 0x1028), 0xc0000000); - nv_wr32(priv, GPC_UNIT(gpc, 0x0824), 0xc0000000); - for (tpc = 0; tpc < priv->tpc_nr[gpc]; tpc++) { - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x508), 0xffffffff); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x50c), 0xffffffff); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x224), 0xc0000000); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x48c), 0xc0000000); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x084), 0xc0000000); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x644), 0x001ffffe); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x64c), 0x0000000f); + nvkm_kmap(gr->unk4188b4); + for (i = 0; i < 0x1000; i += 4) + nvkm_wo32(gr->unk4188b4, i, 0x00000010); + nvkm_done(gr->unk4188b4); + + nvkm_kmap(gr->unk4188b8); + for (i = 0; i < 0x1000; i += 4) + nvkm_wo32(gr->unk4188b8, i, 0x00000010); + nvkm_done(gr->unk4188b8); + + gr->rop_nr = (nvkm_rd32(device, 0x409604) & 0x001f0000) >> 16; + gr->gpc_nr = nvkm_rd32(device, 0x409604) & 0x0000001f; + for (i = 0; i < gr->gpc_nr; i++) { + gr->tpc_nr[i] = nvkm_rd32(device, GPC_UNIT(i, 0x2608)); + gr->tpc_total += gr->tpc_nr[i]; + gr->ppc_nr[i] = gr->func->ppc_nr; + for (j = 0; j < gr->ppc_nr[i]; j++) { + u8 mask = nvkm_rd32(device, GPC_UNIT(i, 0x0c30 + (j * 4))); + if (mask) + gr->ppc_mask[i] |= (1 << j); + gr->ppc_tpc_nr[i][j] = hweight8(mask); } - nv_wr32(priv, GPC_UNIT(gpc, 0x2c90), 0xffffffff); - nv_wr32(priv, GPC_UNIT(gpc, 0x2c94), 0xffffffff); } - for (rop = 0; rop < priv->rop_nr; rop++) { - nv_wr32(priv, ROP_UNIT(rop, 0x144), 0xc0000000); - nv_wr32(priv, ROP_UNIT(rop, 0x070), 0xc0000000); - nv_wr32(priv, ROP_UNIT(rop, 0x204), 0xffffffff); - nv_wr32(priv, ROP_UNIT(rop, 0x208), 0xffffffff); + /*XXX: these need figuring out... though it might not even matter */ + switch (device->chipset) { + case 0xc0: + if (gr->tpc_total == 11) { /* 465, 3/4/4/0, 4 */ + gr->magic_not_rop_nr = 0x07; + } else + if (gr->tpc_total == 14) { /* 470, 3/3/4/4, 5 */ + gr->magic_not_rop_nr = 0x05; + } else + if (gr->tpc_total == 15) { /* 480, 3/4/4/4, 6 */ + gr->magic_not_rop_nr = 0x06; + } + break; + case 0xc3: /* 450, 4/0/0/0, 2 */ + gr->magic_not_rop_nr = 0x03; + break; + case 0xc4: /* 460, 3/4/0/0, 4 */ + gr->magic_not_rop_nr = 0x01; + break; + case 0xc1: /* 2/0/0/0, 1 */ + gr->magic_not_rop_nr = 0x01; + break; + case 0xc8: /* 4/4/3/4, 5 */ + gr->magic_not_rop_nr = 0x06; + break; + case 0xce: /* 4/4/0/0, 4 */ + gr->magic_not_rop_nr = 0x03; + break; + case 0xcf: /* 4/0/0/0, 3 */ + gr->magic_not_rop_nr = 0x03; + break; + case 0xd7: + case 0xd9: /* 1/0/0/0, 1 */ + case 0xea: /* gk20a */ + case 0x12b: /* gm20b */ + gr->magic_not_rop_nr = 0x01; + break; } - nv_wr32(priv, 0x400108, 0xffffffff); - nv_wr32(priv, 0x400138, 0xffffffff); - nv_wr32(priv, 0x400118, 0xffffffff); - nv_wr32(priv, 0x400130, 0xffffffff); - nv_wr32(priv, 0x40011c, 0xffffffff); - nv_wr32(priv, 0x400134, 0xffffffff); - - nv_wr32(priv, 0x400054, 0x34ce3464); - - gf100_gr_zbc_init(priv); + return 0; +} - return gf100_gr_init_ctxctl(priv); +int +gf100_gr_init_(struct nvkm_gr *base) +{ + struct gf100_gr *gr = gf100_gr(base); + nvkm_pmu_pgob(gr->base.engine.subdev.device->pmu, false); + return gr->func->init(gr); } -static void +void gf100_gr_dtor_fw(struct gf100_gr_fuc *fuc) { kfree(fuc->data); fuc->data = NULL; } +void * +gf100_gr_dtor(struct nvkm_gr *base) +{ + struct gf100_gr *gr = gf100_gr(base); + + if (gr->func->dtor) + gr->func->dtor(gr); + kfree(gr->data); + + gf100_gr_dtor_fw(&gr->fuc409c); + gf100_gr_dtor_fw(&gr->fuc409d); + gf100_gr_dtor_fw(&gr->fuc41ac); + gf100_gr_dtor_fw(&gr->fuc41ad); + + nvkm_memory_del(&gr->unk4188b8); + nvkm_memory_del(&gr->unk4188b4); + return gr; +} + +static const struct nvkm_gr_func +gf100_gr_ = { + .dtor = gf100_gr_dtor, + .oneinit = gf100_gr_oneinit, + .init = gf100_gr_init_, + .intr = gf100_gr_intr, + .units = gf100_gr_units, + .chan_new = gf100_gr_chan_new, + .object_get = gf100_gr_object_get, +}; + int -gf100_gr_ctor_fw(struct gf100_gr_priv *priv, const char *fwname, +gf100_gr_ctor_fw(struct gf100_gr *gr, const char *fwname, struct gf100_gr_fuc *fuc) { - struct nvkm_device *device = nv_device(priv); + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; const struct firmware *fw; - char f[32]; + char f[64]; + char cname[16]; int ret; + int i; - snprintf(f, sizeof(f), "nouveau/nv%02x_%s", device->chipset, fwname); - ret = request_firmware(&fw, f, nv_device_base(device)); + /* Convert device name to lowercase */ + strncpy(cname, device->chip->name, sizeof(cname)); + cname[sizeof(cname) - 1] = '\0'; + i = strlen(cname); + while (i) { + --i; + cname[i] = tolower(cname[i]); + } + + snprintf(f, sizeof(f), "nvidia/%s/%s.bin", cname, fwname); + ret = request_firmware(&fw, f, device->dev); if (ret) { - snprintf(f, sizeof(f), "nouveau/%s", fwname); - ret = request_firmware(&fw, f, nv_device_base(device)); - if (ret) { - nv_error(priv, "failed to load %s\n", fwname); - return ret; - } + nvkm_error(subdev, "failed to load %s\n", fwname); + return ret; } fuc->size = fw->size; @@ -1533,126 +1657,150 @@ gf100_gr_ctor_fw(struct gf100_gr_priv *priv, const char *fwname, return (fuc->data != NULL) ? 0 : -ENOMEM; } -void -gf100_gr_dtor(struct nvkm_object *object) +int +gf100_gr_ctor(const struct gf100_gr_func *func, struct nvkm_device *device, + int index, struct gf100_gr *gr) { - struct gf100_gr_priv *priv = (void *)object; + int ret; - kfree(priv->data); + gr->func = func; + gr->firmware = nvkm_boolopt(device->cfgopt, "NvGrUseFW", + func->fecs.ucode == NULL); - gf100_gr_dtor_fw(&priv->fuc409c); - gf100_gr_dtor_fw(&priv->fuc409d); - gf100_gr_dtor_fw(&priv->fuc41ac); - gf100_gr_dtor_fw(&priv->fuc41ad); + ret = nvkm_gr_ctor(&gf100_gr_, device, index, 0x08001000, + gr->firmware || func->fecs.ucode != NULL, + &gr->base); + if (ret) + return ret; - nvkm_gpuobj_ref(NULL, &priv->unk4188b8); - nvkm_gpuobj_ref(NULL, &priv->unk4188b4); + if (gr->firmware) { + nvkm_info(&gr->base.engine.subdev, "using external firmware\n"); + if (gf100_gr_ctor_fw(gr, "fecs_inst", &gr->fuc409c) || + gf100_gr_ctor_fw(gr, "fecs_data", &gr->fuc409d) || + gf100_gr_ctor_fw(gr, "gpccs_inst", &gr->fuc41ac) || + gf100_gr_ctor_fw(gr, "gpccs_data", &gr->fuc41ad)) + return -ENODEV; + } - nvkm_gr_destroy(&priv->base); + return 0; } int -gf100_gr_ctor(struct nvkm_object *parent, struct nvkm_object *engine, - struct nvkm_oclass *bclass, void *data, u32 size, - struct nvkm_object **pobject) +gf100_gr_new_(const struct gf100_gr_func *func, struct nvkm_device *device, + int index, struct nvkm_gr **pgr) { - struct gf100_gr_oclass *oclass = (void *)bclass; - struct nvkm_device *device = nv_device(parent); - struct gf100_gr_priv *priv; - bool use_ext_fw, enable; - int ret, i, j; + struct gf100_gr *gr; + if (!(gr = kzalloc(sizeof(*gr), GFP_KERNEL))) + return -ENOMEM; + *pgr = &gr->base; + return gf100_gr_ctor(func, device, index, gr); +} - use_ext_fw = nvkm_boolopt(device->cfgopt, "NvGrUseFW", - oclass->fecs.ucode == NULL); - enable = use_ext_fw || oclass->fecs.ucode != NULL; +int +gf100_gr_init(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + const u32 magicgpc918 = DIV_ROUND_UP(0x00800000, gr->tpc_total); + u32 data[TPC_MAX / 8] = {}; + u8 tpcnr[GPC_MAX]; + int gpc, tpc, rop; + int i; - ret = nvkm_gr_create(parent, engine, bclass, enable, &priv); - *pobject = nv_object(priv); - if (ret) - return ret; + nvkm_wr32(device, GPC_BCAST(0x0880), 0x00000000); + nvkm_wr32(device, GPC_BCAST(0x08a4), 0x00000000); + nvkm_wr32(device, GPC_BCAST(0x0888), 0x00000000); + nvkm_wr32(device, GPC_BCAST(0x088c), 0x00000000); + nvkm_wr32(device, GPC_BCAST(0x0890), 0x00000000); + nvkm_wr32(device, GPC_BCAST(0x0894), 0x00000000); + nvkm_wr32(device, GPC_BCAST(0x08b4), nvkm_memory_addr(gr->unk4188b4) >> 8); + nvkm_wr32(device, GPC_BCAST(0x08b8), nvkm_memory_addr(gr->unk4188b8) >> 8); - nv_subdev(priv)->unit = 0x08001000; - nv_subdev(priv)->intr = gf100_gr_intr; + gf100_gr_mmio(gr, gr->func->mmio); - priv->base.units = gf100_gr_units; + memcpy(tpcnr, gr->tpc_nr, sizeof(gr->tpc_nr)); + for (i = 0, gpc = -1; i < gr->tpc_total; i++) { + do { + gpc = (gpc + 1) % gr->gpc_nr; + } while (!tpcnr[gpc]); + tpc = gr->tpc_nr[gpc] - tpcnr[gpc]--; - if (use_ext_fw) { - nv_info(priv, "using external firmware\n"); - if (gf100_gr_ctor_fw(priv, "fuc409c", &priv->fuc409c) || - gf100_gr_ctor_fw(priv, "fuc409d", &priv->fuc409d) || - gf100_gr_ctor_fw(priv, "fuc41ac", &priv->fuc41ac) || - gf100_gr_ctor_fw(priv, "fuc41ad", &priv->fuc41ad)) - return -ENODEV; - priv->firmware = true; + data[i / 8] |= tpc << ((i % 8) * 4); } - ret = nvkm_gpuobj_new(nv_object(priv), NULL, 0x1000, 256, 0, - &priv->unk4188b4); - if (ret) - return ret; - - ret = nvkm_gpuobj_new(nv_object(priv), NULL, 0x1000, 256, 0, - &priv->unk4188b8); - if (ret) - return ret; + nvkm_wr32(device, GPC_BCAST(0x0980), data[0]); + nvkm_wr32(device, GPC_BCAST(0x0984), data[1]); + nvkm_wr32(device, GPC_BCAST(0x0988), data[2]); + nvkm_wr32(device, GPC_BCAST(0x098c), data[3]); - for (i = 0; i < 0x1000; i += 4) { - nv_wo32(priv->unk4188b4, i, 0x00000010); - nv_wo32(priv->unk4188b8, i, 0x00000010); + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + nvkm_wr32(device, GPC_UNIT(gpc, 0x0914), + gr->magic_not_rop_nr << 8 | gr->tpc_nr[gpc]); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0910), 0x00040000 | + gr->tpc_total); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0918), magicgpc918); } - priv->rop_nr = (nv_rd32(priv, 0x409604) & 0x001f0000) >> 16; - priv->gpc_nr = nv_rd32(priv, 0x409604) & 0x0000001f; - for (i = 0; i < priv->gpc_nr; i++) { - priv->tpc_nr[i] = nv_rd32(priv, GPC_UNIT(i, 0x2608)); - priv->tpc_total += priv->tpc_nr[i]; - priv->ppc_nr[i] = oclass->ppc_nr; - for (j = 0; j < priv->ppc_nr[i]; j++) { - u8 mask = nv_rd32(priv, GPC_UNIT(i, 0x0c30 + (j * 4))); - priv->ppc_tpc_nr[i][j] = hweight8(mask); + if (device->chipset != 0xd7) + nvkm_wr32(device, GPC_BCAST(0x1bd4), magicgpc918); + else + nvkm_wr32(device, GPC_BCAST(0x3fd4), magicgpc918); + + nvkm_wr32(device, GPC_BCAST(0x08ac), nvkm_rd32(device, 0x100800)); + + nvkm_wr32(device, 0x400500, 0x00010001); + + nvkm_wr32(device, 0x400100, 0xffffffff); + nvkm_wr32(device, 0x40013c, 0xffffffff); + + nvkm_wr32(device, 0x409c24, 0x000f0000); + nvkm_wr32(device, 0x404000, 0xc0000000); + nvkm_wr32(device, 0x404600, 0xc0000000); + nvkm_wr32(device, 0x408030, 0xc0000000); + nvkm_wr32(device, 0x40601c, 0xc0000000); + nvkm_wr32(device, 0x404490, 0xc0000000); + nvkm_wr32(device, 0x406018, 0xc0000000); + nvkm_wr32(device, 0x405840, 0xc0000000); + nvkm_wr32(device, 0x405844, 0x00ffffff); + nvkm_mask(device, 0x419cc0, 0x00000008, 0x00000008); + nvkm_mask(device, 0x419eb4, 0x00001000, 0x00001000); + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + nvkm_wr32(device, GPC_UNIT(gpc, 0x0420), 0xc0000000); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0900), 0xc0000000); + nvkm_wr32(device, GPC_UNIT(gpc, 0x1028), 0xc0000000); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0824), 0xc0000000); + for (tpc = 0; tpc < gr->tpc_nr[gpc]; tpc++) { + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x508), 0xffffffff); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x50c), 0xffffffff); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x224), 0xc0000000); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x48c), 0xc0000000); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x084), 0xc0000000); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x644), 0x001ffffe); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x64c), 0x0000000f); } + nvkm_wr32(device, GPC_UNIT(gpc, 0x2c90), 0xffffffff); + nvkm_wr32(device, GPC_UNIT(gpc, 0x2c94), 0xffffffff); } - /*XXX: these need figuring out... though it might not even matter */ - switch (nv_device(priv)->chipset) { - case 0xc0: - if (priv->tpc_total == 11) { /* 465, 3/4/4/0, 4 */ - priv->magic_not_rop_nr = 0x07; - } else - if (priv->tpc_total == 14) { /* 470, 3/3/4/4, 5 */ - priv->magic_not_rop_nr = 0x05; - } else - if (priv->tpc_total == 15) { /* 480, 3/4/4/4, 6 */ - priv->magic_not_rop_nr = 0x06; - } - break; - case 0xc3: /* 450, 4/0/0/0, 2 */ - priv->magic_not_rop_nr = 0x03; - break; - case 0xc4: /* 460, 3/4/0/0, 4 */ - priv->magic_not_rop_nr = 0x01; - break; - case 0xc1: /* 2/0/0/0, 1 */ - priv->magic_not_rop_nr = 0x01; - break; - case 0xc8: /* 4/4/3/4, 5 */ - priv->magic_not_rop_nr = 0x06; - break; - case 0xce: /* 4/4/0/0, 4 */ - priv->magic_not_rop_nr = 0x03; - break; - case 0xcf: /* 4/0/0/0, 3 */ - priv->magic_not_rop_nr = 0x03; - break; - case 0xd7: - case 0xd9: /* 1/0/0/0, 1 */ - priv->magic_not_rop_nr = 0x01; - break; + for (rop = 0; rop < gr->rop_nr; rop++) { + nvkm_wr32(device, ROP_UNIT(rop, 0x144), 0xc0000000); + nvkm_wr32(device, ROP_UNIT(rop, 0x070), 0xc0000000); + nvkm_wr32(device, ROP_UNIT(rop, 0x204), 0xffffffff); + nvkm_wr32(device, ROP_UNIT(rop, 0x208), 0xffffffff); } - nv_engine(priv)->cclass = *oclass->cclass; - nv_engine(priv)->sclass = oclass->sclass; - return 0; + nvkm_wr32(device, 0x400108, 0xffffffff); + nvkm_wr32(device, 0x400138, 0xffffffff); + nvkm_wr32(device, 0x400118, 0xffffffff); + nvkm_wr32(device, 0x400130, 0xffffffff); + nvkm_wr32(device, 0x40011c, 0xffffffff); + nvkm_wr32(device, 0x400134, 0xffffffff); + + nvkm_wr32(device, 0x400054, 0x34ce3464); + + gf100_gr_zbc_init(gr); + + return gf100_gr_init_ctxctl(gr); } #include "fuc/hubgf100.fuc3.h" @@ -1675,18 +1823,24 @@ gf100_gr_gpccs_ucode = { .data.size = sizeof(gf100_grgpc_data), }; -struct nvkm_oclass * -gf100_gr_oclass = &(struct gf100_gr_oclass) { - .base.handle = NV_ENGINE(GR, 0xc0), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gf100_gr_ctor, - .dtor = gf100_gr_dtor, - .init = gf100_gr_init, - .fini = _nvkm_gr_fini, - }, - .cclass = &gf100_grctx_oclass, - .sclass = gf100_gr_sclass, +static const struct gf100_gr_func +gf100_gr = { + .init = gf100_gr_init, .mmio = gf100_gr_pack_mmio, .fecs.ucode = &gf100_gr_fecs_ucode, .gpccs.ucode = &gf100_gr_gpccs_ucode, -}.base; + .grctx = &gf100_grctx, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, FERMI_MEMORY_TO_MEMORY_FORMAT_A }, + { -1, -1, FERMI_A, &gf100_fermi }, + { -1, -1, FERMI_COMPUTE_A }, + {} + } +}; + +int +gf100_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(&gf100_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.h b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.h index 8af1a89ed..02e78b8d9 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.h +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.h @@ -21,11 +21,14 @@ * * Authors: Ben Skeggs */ -#ifndef __NVC0_GR_H__ -#define __NVC0_GR_H__ -#include <engine/gr.h> +#ifndef __GF100_GR_H__ +#define __GF100_GR_H__ +#define gf100_gr(p) container_of((p), struct gf100_gr, base) +#include "priv.h" +#include <core/gpuobj.h> #include <subdev/ltc.h> +#include <subdev/mmu.h> #define GPC_MAX 32 #define TPC_MAX (GPC_MAX * 8) @@ -67,7 +70,8 @@ struct gf100_gr_zbc_depth { u32 l2; }; -struct gf100_gr_priv { +struct gf100_gr { + const struct gf100_gr_func *func; struct nvkm_gr base; struct gf100_gr_fuc fuc409c; @@ -76,6 +80,15 @@ struct gf100_gr_priv { struct gf100_gr_fuc fuc41ad; bool firmware; + /* + * Used if the register packs are loaded from NVIDIA fw instead of + * using hardcoded arrays. + */ + struct gf100_gr_pack *fuc_sw_nonctx; + struct gf100_gr_pack *fuc_sw_ctx; + struct gf100_gr_pack *fuc_bundle; + struct gf100_gr_pack *fuc_method; + struct gf100_gr_zbc_color zbc_color[NVKM_LTC_MAX_ZBC_CNT]; struct gf100_gr_zbc_depth zbc_depth[NVKM_LTC_MAX_ZBC_CNT]; @@ -84,10 +97,11 @@ struct gf100_gr_priv { u8 tpc_nr[GPC_MAX]; u8 tpc_total; u8 ppc_nr[GPC_MAX]; + u8 ppc_mask[GPC_MAX]; u8 ppc_tpc_nr[GPC_MAX][4]; - struct nvkm_gpuobj *unk4188b4; - struct nvkm_gpuobj *unk4188b8; + struct nvkm_memory *unk4188b4; + struct nvkm_memory *unk4188b8; struct gf100_gr_data mmio_data[4]; struct gf100_gr_mmio mmio_list[4096/8]; @@ -97,48 +111,65 @@ struct gf100_gr_priv { u8 magic_not_rop_nr; }; +int gf100_gr_ctor(const struct gf100_gr_func *, struct nvkm_device *, + int, struct gf100_gr *); +int gf100_gr_new_(const struct gf100_gr_func *, struct nvkm_device *, + int, struct nvkm_gr **); +void *gf100_gr_dtor(struct nvkm_gr *); + +struct gf100_gr_func { + void (*dtor)(struct gf100_gr *); + int (*init)(struct gf100_gr *); + void (*init_gpc_mmu)(struct gf100_gr *); + void (*set_hww_esr_report_mask)(struct gf100_gr *); + const struct gf100_gr_pack *mmio; + struct { + struct gf100_gr_ucode *ucode; + } fecs; + struct { + struct gf100_gr_ucode *ucode; + } gpccs; + int ppc_nr; + const struct gf100_grctx_func *grctx; + struct nvkm_sclass sclass[]; +}; + +int gf100_gr_init(struct gf100_gr *); + +int gk104_gr_init(struct gf100_gr *); + +int gk20a_gr_new_(const struct gf100_gr_func *, struct nvkm_device *, + int, struct nvkm_gr **); +void gk20a_gr_dtor(struct gf100_gr *); +int gk20a_gr_init(struct gf100_gr *); + +int gm204_gr_init(struct gf100_gr *); + +#define gf100_gr_chan(p) container_of((p), struct gf100_gr_chan, object) + struct gf100_gr_chan { - struct nvkm_gr_chan base; + struct nvkm_object object; + struct gf100_gr *gr; - struct nvkm_gpuobj *mmio; + struct nvkm_memory *mmio; struct nvkm_vma mmio_vma; int mmio_nr; + struct { - struct nvkm_gpuobj *mem; + struct nvkm_memory *mem; struct nvkm_vma vma; } data[4]; }; -int gf100_gr_context_ctor(struct nvkm_object *, struct nvkm_object *, - struct nvkm_oclass *, void *, u32, - struct nvkm_object **); -void gf100_gr_context_dtor(struct nvkm_object *); - -void gf100_gr_ctxctl_debug(struct gf100_gr_priv *); +void gf100_gr_ctxctl_debug(struct gf100_gr *); +void gf100_gr_dtor_fw(struct gf100_gr_fuc *); +int gf100_gr_ctor_fw(struct gf100_gr *, const char *, + struct gf100_gr_fuc *); u64 gf100_gr_units(struct nvkm_gr *); -int gf100_gr_ctor(struct nvkm_object *, struct nvkm_object *, - struct nvkm_oclass *, void *data, u32 size, - struct nvkm_object **); -void gf100_gr_dtor(struct nvkm_object *); -int gf100_gr_init(struct nvkm_object *); -void gf100_gr_zbc_init(struct gf100_gr_priv *); - -int gk104_gr_ctor(struct nvkm_object *, struct nvkm_object *, - struct nvkm_oclass *, void *data, u32 size, - struct nvkm_object **); -int gk104_gr_init(struct nvkm_object *); - -int gm204_gr_init(struct nvkm_object *); +void gf100_gr_zbc_init(struct gf100_gr *); -extern struct nvkm_ofuncs gf100_fermi_ofuncs; - -extern struct nvkm_oclass gf100_gr_sclass[]; -extern struct nvkm_omthds gf100_gr_9097_omthds[]; -extern struct nvkm_omthds gf100_gr_90c0_omthds[]; -extern struct nvkm_oclass gf110_gr_sclass[]; -extern struct nvkm_oclass gk110_gr_sclass[]; -extern struct nvkm_oclass gm204_gr_sclass[]; +extern const struct nvkm_object_func gf100_fermi; struct gf100_gr_init { u32 addr; @@ -167,24 +198,11 @@ extern struct gf100_gr_ucode gf100_gr_gpccs_ucode; extern struct gf100_gr_ucode gk110_gr_fecs_ucode; extern struct gf100_gr_ucode gk110_gr_gpccs_ucode; -struct gf100_gr_oclass { - struct nvkm_oclass base; - struct nvkm_oclass **cclass; - struct nvkm_oclass *sclass; - const struct gf100_gr_pack *mmio; - struct { - struct gf100_gr_ucode *ucode; - } fecs; - struct { - struct gf100_gr_ucode *ucode; - } gpccs; - int ppc_nr; -}; - -void gf100_gr_mmio(struct gf100_gr_priv *, const struct gf100_gr_pack *); -void gf100_gr_icmd(struct gf100_gr_priv *, const struct gf100_gr_pack *); -void gf100_gr_mthd(struct gf100_gr_priv *, const struct gf100_gr_pack *); -int gf100_gr_init_ctxctl(struct gf100_gr_priv *); +int gf100_gr_wait_idle(struct gf100_gr *); +void gf100_gr_mmio(struct gf100_gr *, const struct gf100_gr_pack *); +void gf100_gr_icmd(struct gf100_gr *, const struct gf100_gr_pack *); +void gf100_gr_mthd(struct gf100_gr *, const struct gf100_gr_pack *); +int gf100_gr_init_ctxctl(struct gf100_gr *); /* register init value lists */ @@ -260,7 +278,7 @@ extern const struct gf100_gr_init gm107_gr_init_tex_0[]; extern const struct gf100_gr_init gm107_gr_init_l1c_0[]; extern const struct gf100_gr_init gm107_gr_init_wwdx_0[]; extern const struct gf100_gr_init gm107_gr_init_cbm_0[]; -void gm107_gr_init_bios(struct gf100_gr_priv *); +void gm107_gr_init_bios(struct gf100_gr *); extern const struct gf100_gr_pack gm204_gr_pack_mmio[]; #endif diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf104.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf104.c index 20d3b85db..8f253e0a2 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf104.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf104.c @@ -24,6 +24,8 @@ #include "gf100.h" #include "ctxgf100.h" +#include <nvif/class.h> + /******************************************************************************* * PGRAPH register lists ******************************************************************************/ @@ -110,18 +112,24 @@ gf104_gr_pack_mmio[] = { * PGRAPH engine/subdev functions ******************************************************************************/ -struct nvkm_oclass * -gf104_gr_oclass = &(struct gf100_gr_oclass) { - .base.handle = NV_ENGINE(GR, 0xc3), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gf100_gr_ctor, - .dtor = gf100_gr_dtor, - .init = gf100_gr_init, - .fini = _nvkm_gr_fini, - }, - .cclass = &gf104_grctx_oclass, - .sclass = gf100_gr_sclass, +static const struct gf100_gr_func +gf104_gr = { + .init = gf100_gr_init, .mmio = gf104_gr_pack_mmio, .fecs.ucode = &gf100_gr_fecs_ucode, .gpccs.ucode = &gf100_gr_gpccs_ucode, -}.base; + .grctx = &gf104_grctx, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, FERMI_MEMORY_TO_MEMORY_FORMAT_A }, + { -1, -1, FERMI_A, &gf100_fermi }, + { -1, -1, FERMI_COMPUTE_A }, + {} + } +}; + +int +gf104_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(&gf104_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf108.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf108.c index 8df73421c..815a5aafa 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf108.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf108.c @@ -27,20 +27,6 @@ #include <nvif/class.h> /******************************************************************************* - * Graphics object classes - ******************************************************************************/ - -static struct nvkm_oclass -gf108_gr_sclass[] = { - { FERMI_TWOD_A, &nvkm_object_ofuncs }, - { FERMI_MEMORY_TO_MEMORY_FORMAT_A, &nvkm_object_ofuncs }, - { FERMI_A, &gf100_fermi_ofuncs, gf100_gr_9097_omthds }, - { FERMI_B, &gf100_fermi_ofuncs, gf100_gr_9097_omthds }, - { FERMI_COMPUTE_A, &nvkm_object_ofuncs, gf100_gr_90c0_omthds }, - {} -}; - -/******************************************************************************* * PGRAPH register lists ******************************************************************************/ @@ -117,18 +103,25 @@ gf108_gr_pack_mmio[] = { * PGRAPH engine/subdev functions ******************************************************************************/ -struct nvkm_oclass * -gf108_gr_oclass = &(struct gf100_gr_oclass) { - .base.handle = NV_ENGINE(GR, 0xc1), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gf100_gr_ctor, - .dtor = gf100_gr_dtor, - .init = gf100_gr_init, - .fini = _nvkm_gr_fini, - }, - .cclass = &gf108_grctx_oclass, - .sclass = gf108_gr_sclass, +static const struct gf100_gr_func +gf108_gr = { + .init = gf100_gr_init, .mmio = gf108_gr_pack_mmio, .fecs.ucode = &gf100_gr_fecs_ucode, .gpccs.ucode = &gf100_gr_gpccs_ucode, -}.base; + .grctx = &gf108_grctx, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, FERMI_MEMORY_TO_MEMORY_FORMAT_A }, + { -1, -1, FERMI_A, &gf100_fermi }, + { -1, -1, FERMI_B, &gf100_fermi }, + { -1, -1, FERMI_COMPUTE_A }, + {} + } +}; + +int +gf108_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(&gf108_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf110.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf110.c index ef76e2dd1..d081ee41f 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf110.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf110.c @@ -27,21 +27,6 @@ #include <nvif/class.h> /******************************************************************************* - * Graphics object classes - ******************************************************************************/ - -struct nvkm_oclass -gf110_gr_sclass[] = { - { FERMI_TWOD_A, &nvkm_object_ofuncs }, - { FERMI_MEMORY_TO_MEMORY_FORMAT_A, &nvkm_object_ofuncs }, - { FERMI_A, &gf100_fermi_ofuncs, gf100_gr_9097_omthds }, - { FERMI_B, &gf100_fermi_ofuncs, gf100_gr_9097_omthds }, - { FERMI_C, &gf100_fermi_ofuncs, gf100_gr_9097_omthds }, - { FERMI_COMPUTE_A, &nvkm_object_ofuncs, gf100_gr_90c0_omthds }, - {} -}; - -/******************************************************************************* * PGRAPH register lists ******************************************************************************/ @@ -99,18 +84,27 @@ gf110_gr_pack_mmio[] = { * PGRAPH engine/subdev functions ******************************************************************************/ -struct nvkm_oclass * -gf110_gr_oclass = &(struct gf100_gr_oclass) { - .base.handle = NV_ENGINE(GR, 0xc8), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gf100_gr_ctor, - .dtor = gf100_gr_dtor, - .init = gf100_gr_init, - .fini = _nvkm_gr_fini, - }, - .cclass = &gf110_grctx_oclass, - .sclass = gf110_gr_sclass, +static const struct gf100_gr_func +gf110_gr = { + .init = gf100_gr_init, .mmio = gf110_gr_pack_mmio, .fecs.ucode = &gf100_gr_fecs_ucode, .gpccs.ucode = &gf100_gr_gpccs_ucode, -}.base; + .grctx = &gf110_grctx, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, FERMI_MEMORY_TO_MEMORY_FORMAT_A }, + { -1, -1, FERMI_A, &gf100_fermi }, + { -1, -1, FERMI_B, &gf100_fermi }, + { -1, -1, FERMI_C, &gf100_fermi }, + { -1, -1, FERMI_COMPUTE_A }, + { -1, -1, FERMI_COMPUTE_B }, + {} + } +}; + +int +gf110_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(&gf110_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf117.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf117.c index 871ac5f80..d8e8af4d3 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf117.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf117.c @@ -24,6 +24,8 @@ #include "gf100.h" #include "ctxgf100.h" +#include <nvif/class.h> + /******************************************************************************* * PGRAPH register lists ******************************************************************************/ @@ -118,19 +120,28 @@ gf117_gr_gpccs_ucode = { .data.size = sizeof(gf117_grgpc_data), }; -struct nvkm_oclass * -gf117_gr_oclass = &(struct gf100_gr_oclass) { - .base.handle = NV_ENGINE(GR, 0xd7), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gf100_gr_ctor, - .dtor = gf100_gr_dtor, - .init = gf100_gr_init, - .fini = _nvkm_gr_fini, - }, - .cclass = &gf117_grctx_oclass, - .sclass = gf110_gr_sclass, +static const struct gf100_gr_func +gf117_gr = { + .init = gf100_gr_init, .mmio = gf117_gr_pack_mmio, .fecs.ucode = &gf117_gr_fecs_ucode, .gpccs.ucode = &gf117_gr_gpccs_ucode, .ppc_nr = 1, -}.base; + .grctx = &gf117_grctx, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, FERMI_MEMORY_TO_MEMORY_FORMAT_A }, + { -1, -1, FERMI_A, &gf100_fermi }, + { -1, -1, FERMI_B, &gf100_fermi }, + { -1, -1, FERMI_C, &gf100_fermi }, + { -1, -1, FERMI_COMPUTE_A }, + { -1, -1, FERMI_COMPUTE_B }, + {} + } +}; + +int +gf117_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(&gf117_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf119.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf119.c index e6dd651e2..01faf9a73 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf119.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gf119.c @@ -24,6 +24,8 @@ #include "gf100.h" #include "ctxgf100.h" +#include <nvif/class.h> + /******************************************************************************* * PGRAPH register lists ******************************************************************************/ @@ -173,18 +175,27 @@ gf119_gr_pack_mmio[] = { * PGRAPH engine/subdev functions ******************************************************************************/ -struct nvkm_oclass * -gf119_gr_oclass = &(struct gf100_gr_oclass) { - .base.handle = NV_ENGINE(GR, 0xd9), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gf100_gr_ctor, - .dtor = gf100_gr_dtor, - .init = gf100_gr_init, - .fini = _nvkm_gr_fini, - }, - .cclass = &gf119_grctx_oclass, - .sclass = gf110_gr_sclass, +static const struct gf100_gr_func +gf119_gr = { + .init = gf100_gr_init, .mmio = gf119_gr_pack_mmio, .fecs.ucode = &gf100_gr_fecs_ucode, .gpccs.ucode = &gf100_gr_gpccs_ucode, -}.base; + .grctx = &gf119_grctx, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, FERMI_MEMORY_TO_MEMORY_FORMAT_A }, + { -1, -1, FERMI_A, &gf100_fermi }, + { -1, -1, FERMI_B, &gf100_fermi }, + { -1, -1, FERMI_C, &gf100_fermi }, + { -1, -1, FERMI_COMPUTE_A }, + { -1, -1, FERMI_COMPUTE_B }, + {} + } +}; + +int +gf119_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(&gf119_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk104.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk104.c index 46f7844ec..abf54928a 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk104.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk104.c @@ -24,24 +24,9 @@ #include "gf100.h" #include "ctxgf100.h" -#include <subdev/pmu.h> - #include <nvif/class.h> /******************************************************************************* - * Graphics object classes - ******************************************************************************/ - -static struct nvkm_oclass -gk104_gr_sclass[] = { - { FERMI_TWOD_A, &nvkm_object_ofuncs }, - { KEPLER_INLINE_TO_MEMORY_A, &nvkm_object_ofuncs }, - { KEPLER_A, &gf100_fermi_ofuncs, gf100_gr_9097_omthds }, - { KEPLER_COMPUTE_A, &nvkm_object_ofuncs, gf100_gr_90c0_omthds }, - {} -}; - -/******************************************************************************* * PGRAPH register lists ******************************************************************************/ @@ -193,132 +178,112 @@ gk104_gr_pack_mmio[] = { ******************************************************************************/ int -gk104_gr_init(struct nvkm_object *object) +gk104_gr_init(struct gf100_gr *gr) { - struct gf100_gr_oclass *oclass = (void *)object->oclass; - struct gf100_gr_priv *priv = (void *)object; - struct nvkm_pmu *pmu = nvkm_pmu(priv); - const u32 magicgpc918 = DIV_ROUND_UP(0x00800000, priv->tpc_total); + struct nvkm_device *device = gr->base.engine.subdev.device; + const u32 magicgpc918 = DIV_ROUND_UP(0x00800000, gr->tpc_total); u32 data[TPC_MAX / 8] = {}; u8 tpcnr[GPC_MAX]; int gpc, tpc, rop; - int ret, i; - - if (pmu) - pmu->pgob(pmu, false); + int i; - ret = nvkm_gr_init(&priv->base); - if (ret) - return ret; + nvkm_wr32(device, GPC_BCAST(0x0880), 0x00000000); + nvkm_wr32(device, GPC_BCAST(0x08a4), 0x00000000); + nvkm_wr32(device, GPC_BCAST(0x0888), 0x00000000); + nvkm_wr32(device, GPC_BCAST(0x088c), 0x00000000); + nvkm_wr32(device, GPC_BCAST(0x0890), 0x00000000); + nvkm_wr32(device, GPC_BCAST(0x0894), 0x00000000); + nvkm_wr32(device, GPC_BCAST(0x08b4), nvkm_memory_addr(gr->unk4188b4) >> 8); + nvkm_wr32(device, GPC_BCAST(0x08b8), nvkm_memory_addr(gr->unk4188b8) >> 8); - nv_wr32(priv, GPC_BCAST(0x0880), 0x00000000); - nv_wr32(priv, GPC_BCAST(0x08a4), 0x00000000); - nv_wr32(priv, GPC_BCAST(0x0888), 0x00000000); - nv_wr32(priv, GPC_BCAST(0x088c), 0x00000000); - nv_wr32(priv, GPC_BCAST(0x0890), 0x00000000); - nv_wr32(priv, GPC_BCAST(0x0894), 0x00000000); - nv_wr32(priv, GPC_BCAST(0x08b4), priv->unk4188b4->addr >> 8); - nv_wr32(priv, GPC_BCAST(0x08b8), priv->unk4188b8->addr >> 8); + gf100_gr_mmio(gr, gr->func->mmio); - gf100_gr_mmio(priv, oclass->mmio); - - nv_wr32(priv, GPC_UNIT(0, 0x3018), 0x00000001); + nvkm_wr32(device, GPC_UNIT(0, 0x3018), 0x00000001); memset(data, 0x00, sizeof(data)); - memcpy(tpcnr, priv->tpc_nr, sizeof(priv->tpc_nr)); - for (i = 0, gpc = -1; i < priv->tpc_total; i++) { + memcpy(tpcnr, gr->tpc_nr, sizeof(gr->tpc_nr)); + for (i = 0, gpc = -1; i < gr->tpc_total; i++) { do { - gpc = (gpc + 1) % priv->gpc_nr; + gpc = (gpc + 1) % gr->gpc_nr; } while (!tpcnr[gpc]); - tpc = priv->tpc_nr[gpc] - tpcnr[gpc]--; + tpc = gr->tpc_nr[gpc] - tpcnr[gpc]--; data[i / 8] |= tpc << ((i % 8) * 4); } - nv_wr32(priv, GPC_BCAST(0x0980), data[0]); - nv_wr32(priv, GPC_BCAST(0x0984), data[1]); - nv_wr32(priv, GPC_BCAST(0x0988), data[2]); - nv_wr32(priv, GPC_BCAST(0x098c), data[3]); - - for (gpc = 0; gpc < priv->gpc_nr; gpc++) { - nv_wr32(priv, GPC_UNIT(gpc, 0x0914), - priv->magic_not_rop_nr << 8 | priv->tpc_nr[gpc]); - nv_wr32(priv, GPC_UNIT(gpc, 0x0910), 0x00040000 | - priv->tpc_total); - nv_wr32(priv, GPC_UNIT(gpc, 0x0918), magicgpc918); + nvkm_wr32(device, GPC_BCAST(0x0980), data[0]); + nvkm_wr32(device, GPC_BCAST(0x0984), data[1]); + nvkm_wr32(device, GPC_BCAST(0x0988), data[2]); + nvkm_wr32(device, GPC_BCAST(0x098c), data[3]); + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + nvkm_wr32(device, GPC_UNIT(gpc, 0x0914), + gr->magic_not_rop_nr << 8 | gr->tpc_nr[gpc]); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0910), 0x00040000 | + gr->tpc_total); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0918), magicgpc918); } - nv_wr32(priv, GPC_BCAST(0x3fd4), magicgpc918); - nv_wr32(priv, GPC_BCAST(0x08ac), nv_rd32(priv, 0x100800)); - - nv_wr32(priv, 0x400500, 0x00010001); - - nv_wr32(priv, 0x400100, 0xffffffff); - nv_wr32(priv, 0x40013c, 0xffffffff); - - nv_wr32(priv, 0x409ffc, 0x00000000); - nv_wr32(priv, 0x409c14, 0x00003e3e); - nv_wr32(priv, 0x409c24, 0x000f0001); - nv_wr32(priv, 0x404000, 0xc0000000); - nv_wr32(priv, 0x404600, 0xc0000000); - nv_wr32(priv, 0x408030, 0xc0000000); - nv_wr32(priv, 0x404490, 0xc0000000); - nv_wr32(priv, 0x406018, 0xc0000000); - nv_wr32(priv, 0x407020, 0x40000000); - nv_wr32(priv, 0x405840, 0xc0000000); - nv_wr32(priv, 0x405844, 0x00ffffff); - nv_mask(priv, 0x419cc0, 0x00000008, 0x00000008); - nv_mask(priv, 0x419eb4, 0x00001000, 0x00001000); - - for (gpc = 0; gpc < priv->gpc_nr; gpc++) { - nv_wr32(priv, GPC_UNIT(gpc, 0x3038), 0xc0000000); - nv_wr32(priv, GPC_UNIT(gpc, 0x0420), 0xc0000000); - nv_wr32(priv, GPC_UNIT(gpc, 0x0900), 0xc0000000); - nv_wr32(priv, GPC_UNIT(gpc, 0x1028), 0xc0000000); - nv_wr32(priv, GPC_UNIT(gpc, 0x0824), 0xc0000000); - for (tpc = 0; tpc < priv->tpc_nr[gpc]; tpc++) { - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x508), 0xffffffff); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x50c), 0xffffffff); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x224), 0xc0000000); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x48c), 0xc0000000); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x084), 0xc0000000); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x644), 0x001ffffe); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x64c), 0x0000000f); + nvkm_wr32(device, GPC_BCAST(0x3fd4), magicgpc918); + nvkm_wr32(device, GPC_BCAST(0x08ac), nvkm_rd32(device, 0x100800)); + + nvkm_wr32(device, 0x400500, 0x00010001); + + nvkm_wr32(device, 0x400100, 0xffffffff); + nvkm_wr32(device, 0x40013c, 0xffffffff); + + nvkm_wr32(device, 0x409ffc, 0x00000000); + nvkm_wr32(device, 0x409c14, 0x00003e3e); + nvkm_wr32(device, 0x409c24, 0x000f0001); + nvkm_wr32(device, 0x404000, 0xc0000000); + nvkm_wr32(device, 0x404600, 0xc0000000); + nvkm_wr32(device, 0x408030, 0xc0000000); + nvkm_wr32(device, 0x404490, 0xc0000000); + nvkm_wr32(device, 0x406018, 0xc0000000); + nvkm_wr32(device, 0x407020, 0x40000000); + nvkm_wr32(device, 0x405840, 0xc0000000); + nvkm_wr32(device, 0x405844, 0x00ffffff); + nvkm_mask(device, 0x419cc0, 0x00000008, 0x00000008); + nvkm_mask(device, 0x419eb4, 0x00001000, 0x00001000); + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + nvkm_wr32(device, GPC_UNIT(gpc, 0x3038), 0xc0000000); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0420), 0xc0000000); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0900), 0xc0000000); + nvkm_wr32(device, GPC_UNIT(gpc, 0x1028), 0xc0000000); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0824), 0xc0000000); + for (tpc = 0; tpc < gr->tpc_nr[gpc]; tpc++) { + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x508), 0xffffffff); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x50c), 0xffffffff); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x224), 0xc0000000); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x48c), 0xc0000000); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x084), 0xc0000000); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x644), 0x001ffffe); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x64c), 0x0000000f); } - nv_wr32(priv, GPC_UNIT(gpc, 0x2c90), 0xffffffff); - nv_wr32(priv, GPC_UNIT(gpc, 0x2c94), 0xffffffff); + nvkm_wr32(device, GPC_UNIT(gpc, 0x2c90), 0xffffffff); + nvkm_wr32(device, GPC_UNIT(gpc, 0x2c94), 0xffffffff); } - for (rop = 0; rop < priv->rop_nr; rop++) { - nv_wr32(priv, ROP_UNIT(rop, 0x144), 0xc0000000); - nv_wr32(priv, ROP_UNIT(rop, 0x070), 0xc0000000); - nv_wr32(priv, ROP_UNIT(rop, 0x204), 0xffffffff); - nv_wr32(priv, ROP_UNIT(rop, 0x208), 0xffffffff); + for (rop = 0; rop < gr->rop_nr; rop++) { + nvkm_wr32(device, ROP_UNIT(rop, 0x144), 0xc0000000); + nvkm_wr32(device, ROP_UNIT(rop, 0x070), 0xc0000000); + nvkm_wr32(device, ROP_UNIT(rop, 0x204), 0xffffffff); + nvkm_wr32(device, ROP_UNIT(rop, 0x208), 0xffffffff); } - nv_wr32(priv, 0x400108, 0xffffffff); - nv_wr32(priv, 0x400138, 0xffffffff); - nv_wr32(priv, 0x400118, 0xffffffff); - nv_wr32(priv, 0x400130, 0xffffffff); - nv_wr32(priv, 0x40011c, 0xffffffff); - nv_wr32(priv, 0x400134, 0xffffffff); + nvkm_wr32(device, 0x400108, 0xffffffff); + nvkm_wr32(device, 0x400138, 0xffffffff); + nvkm_wr32(device, 0x400118, 0xffffffff); + nvkm_wr32(device, 0x400130, 0xffffffff); + nvkm_wr32(device, 0x40011c, 0xffffffff); + nvkm_wr32(device, 0x400134, 0xffffffff); - nv_wr32(priv, 0x400054, 0x34ce3464); + nvkm_wr32(device, 0x400054, 0x34ce3464); - gf100_gr_zbc_init(priv); + gf100_gr_zbc_init(gr); - return gf100_gr_init_ctxctl(priv); -} - -int -gk104_gr_ctor(struct nvkm_object *parent, struct nvkm_object *engine, - struct nvkm_oclass *oclass, void *data, u32 size, - struct nvkm_object **pobject) -{ - struct nvkm_pmu *pmu = nvkm_pmu(parent); - if (pmu) - pmu->pgob(pmu, false); - return gf100_gr_ctor(parent, engine, oclass, data, size, pobject); + return gf100_gr_init_ctxctl(gr); } #include "fuc/hubgk104.fuc3.h" @@ -341,19 +306,25 @@ gk104_gr_gpccs_ucode = { .data.size = sizeof(gk104_grgpc_data), }; -struct nvkm_oclass * -gk104_gr_oclass = &(struct gf100_gr_oclass) { - .base.handle = NV_ENGINE(GR, 0xe4), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gk104_gr_ctor, - .dtor = gf100_gr_dtor, - .init = gk104_gr_init, - .fini = _nvkm_gr_fini, - }, - .cclass = &gk104_grctx_oclass, - .sclass = gk104_gr_sclass, +static const struct gf100_gr_func +gk104_gr = { + .init = gk104_gr_init, .mmio = gk104_gr_pack_mmio, .fecs.ucode = &gk104_gr_fecs_ucode, .gpccs.ucode = &gk104_gr_gpccs_ucode, .ppc_nr = 1, -}.base; + .grctx = &gk104_grctx, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, KEPLER_INLINE_TO_MEMORY_A }, + { -1, -1, KEPLER_A, &gf100_fermi }, + { -1, -1, KEPLER_COMPUTE_A }, + {} + } +}; + +int +gk104_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(&gk104_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110.c index f4cd8e554..32aa2946e 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110.c @@ -29,19 +29,6 @@ #include <nvif/class.h> /******************************************************************************* - * Graphics object classes - ******************************************************************************/ - -struct nvkm_oclass -gk110_gr_sclass[] = { - { FERMI_TWOD_A, &nvkm_object_ofuncs }, - { KEPLER_INLINE_TO_MEMORY_B, &nvkm_object_ofuncs }, - { KEPLER_B, &gf100_fermi_ofuncs, gf100_gr_9097_omthds }, - { KEPLER_COMPUTE_B, &nvkm_object_ofuncs, gf100_gr_90c0_omthds }, - {} -}; - -/******************************************************************************* * PGRAPH register lists ******************************************************************************/ @@ -193,19 +180,25 @@ gk110_gr_gpccs_ucode = { .data.size = sizeof(gk110_grgpc_data), }; -struct nvkm_oclass * -gk110_gr_oclass = &(struct gf100_gr_oclass) { - .base.handle = NV_ENGINE(GR, 0xf0), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gk104_gr_ctor, - .dtor = gf100_gr_dtor, - .init = gk104_gr_init, - .fini = _nvkm_gr_fini, - }, - .cclass = &gk110_grctx_oclass, - .sclass = gk110_gr_sclass, +static const struct gf100_gr_func +gk110_gr = { + .init = gk104_gr_init, .mmio = gk110_gr_pack_mmio, .fecs.ucode = &gk110_gr_fecs_ucode, .gpccs.ucode = &gk110_gr_gpccs_ucode, .ppc_nr = 2, -}.base; + .grctx = &gk110_grctx, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, KEPLER_INLINE_TO_MEMORY_B }, + { -1, -1, KEPLER_B, &gf100_fermi }, + { -1, -1, KEPLER_COMPUTE_B }, + {} + } +}; + +int +gk110_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(&gk110_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110b.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110b.c index 9ff9eab0c..22f88afbf 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110b.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk110b.c @@ -24,6 +24,8 @@ #include "gf100.h" #include "ctxgf100.h" +#include <nvif/class.h> + /******************************************************************************* * PGRAPH register lists ******************************************************************************/ @@ -98,19 +100,25 @@ gk110b_gr_pack_mmio[] = { * PGRAPH engine/subdev functions ******************************************************************************/ -struct nvkm_oclass * -gk110b_gr_oclass = &(struct gf100_gr_oclass) { - .base.handle = NV_ENGINE(GR, 0xf1), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gk104_gr_ctor, - .dtor = gf100_gr_dtor, - .init = gk104_gr_init, - .fini = _nvkm_gr_fini, - }, - .cclass = &gk110b_grctx_oclass, - .sclass = gk110_gr_sclass, +static const struct gf100_gr_func +gk110b_gr = { + .init = gk104_gr_init, .mmio = gk110b_gr_pack_mmio, .fecs.ucode = &gk110_gr_fecs_ucode, .gpccs.ucode = &gk110_gr_gpccs_ucode, .ppc_nr = 2, -}.base; + .grctx = &gk110b_grctx, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, KEPLER_INLINE_TO_MEMORY_B }, + { -1, -1, KEPLER_B, &gf100_fermi }, + { -1, -1, KEPLER_COMPUTE_B }, + {} + } +}; + +int +gk110b_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(&gk110b_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk208.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk208.c index 85f44a3d5..ee7554fc8 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk208.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk208.c @@ -29,19 +29,6 @@ #include <nvif/class.h> /******************************************************************************* - * Graphics object classes - ******************************************************************************/ - -static struct nvkm_oclass -gk208_gr_sclass[] = { - { FERMI_TWOD_A, &nvkm_object_ofuncs }, - { KEPLER_INLINE_TO_MEMORY_B, &nvkm_object_ofuncs }, - { KEPLER_B, &gf100_fermi_ofuncs }, - { KEPLER_COMPUTE_B, &nvkm_object_ofuncs }, - {} -}; - -/******************************************************************************* * PGRAPH register lists ******************************************************************************/ @@ -172,19 +159,25 @@ gk208_gr_gpccs_ucode = { .data.size = sizeof(gk208_grgpc_data), }; -struct nvkm_oclass * -gk208_gr_oclass = &(struct gf100_gr_oclass) { - .base.handle = NV_ENGINE(GR, 0x08), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gk104_gr_ctor, - .dtor = gf100_gr_dtor, - .init = gk104_gr_init, - .fini = _nvkm_gr_fini, - }, - .cclass = &gk208_grctx_oclass, - .sclass = gk208_gr_sclass, +static const struct gf100_gr_func +gk208_gr = { + .init = gk104_gr_init, .mmio = gk208_gr_pack_mmio, .fecs.ucode = &gk208_gr_fecs_ucode, .gpccs.ucode = &gk208_gr_gpccs_ucode, .ppc_nr = 1, -}.base; + .grctx = &gk208_grctx, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, KEPLER_INLINE_TO_MEMORY_B }, + { -1, -1, KEPLER_B, &gf100_fermi }, + { -1, -1, KEPLER_COMPUTE_B }, + {} + } +}; + +int +gk208_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(&gk208_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk20a.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk20a.c index 40ff5eb91..b8758d3b8 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk20a.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk20a.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2014, NVIDIA CORPORATION. All rights reserved. + * Copyright (c) 2014-2015, NVIDIA CORPORATION. All rights reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -22,28 +22,335 @@ #include "gf100.h" #include "ctxgf100.h" +#include <subdev/timer.h> + #include <nvif/class.h> -static struct nvkm_oclass -gk20a_gr_sclass[] = { - { FERMI_TWOD_A, &nvkm_object_ofuncs }, - { KEPLER_INLINE_TO_MEMORY_A, &nvkm_object_ofuncs }, - { KEPLER_C, &gf100_fermi_ofuncs, gf100_gr_9097_omthds }, - { KEPLER_COMPUTE_A, &nvkm_object_ofuncs, gf100_gr_90c0_omthds }, - {} +static void +gk20a_gr_init_dtor(struct gf100_gr_pack *pack) +{ + vfree(pack); +} + +struct gk20a_fw_av +{ + u32 addr; + u32 data; +}; + +static struct gf100_gr_pack * +gk20a_gr_av_to_init(struct gf100_gr_fuc *fuc) +{ + struct gf100_gr_init *init; + struct gf100_gr_pack *pack; + const int nent = (fuc->size / sizeof(struct gk20a_fw_av)); + int i; + + pack = vzalloc((sizeof(*pack) * 2) + (sizeof(*init) * (nent + 1))); + if (!pack) + return ERR_PTR(-ENOMEM); + + init = (void *)(pack + 2); + + pack[0].init = init; + + for (i = 0; i < nent; i++) { + struct gf100_gr_init *ent = &init[i]; + struct gk20a_fw_av *av = &((struct gk20a_fw_av *)fuc->data)[i]; + + ent->addr = av->addr; + ent->data = av->data; + ent->count = 1; + ent->pitch = 1; + } + + return pack; +} + +struct gk20a_fw_aiv +{ + u32 addr; + u32 index; + u32 data; }; -struct nvkm_oclass * -gk20a_gr_oclass = &(struct gf100_gr_oclass) { - .base.handle = NV_ENGINE(GR, 0xea), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gf100_gr_ctor, - .dtor = gf100_gr_dtor, - .init = gk104_gr_init, - .fini = _nvkm_gr_fini, - }, - .cclass = &gk20a_grctx_oclass, - .sclass = gk20a_gr_sclass, - .mmio = gk104_gr_pack_mmio, +static struct gf100_gr_pack * +gk20a_gr_aiv_to_init(struct gf100_gr_fuc *fuc) +{ + struct gf100_gr_init *init; + struct gf100_gr_pack *pack; + const int nent = (fuc->size / sizeof(struct gk20a_fw_aiv)); + int i; + + pack = vzalloc((sizeof(*pack) * 2) + (sizeof(*init) * (nent + 1))); + if (!pack) + return ERR_PTR(-ENOMEM); + + init = (void *)(pack + 2); + + pack[0].init = init; + + for (i = 0; i < nent; i++) { + struct gf100_gr_init *ent = &init[i]; + struct gk20a_fw_aiv *av = &((struct gk20a_fw_aiv *)fuc->data)[i]; + + ent->addr = av->addr; + ent->data = av->data; + ent->count = 1; + ent->pitch = 1; + } + + return pack; +} + +static struct gf100_gr_pack * +gk20a_gr_av_to_method(struct gf100_gr_fuc *fuc) +{ + struct gf100_gr_init *init; + struct gf100_gr_pack *pack; + /* We don't suppose we will initialize more than 16 classes here... */ + static const unsigned int max_classes = 16; + const int nent = (fuc->size / sizeof(struct gk20a_fw_av)); + int i, classidx = 0; + u32 prevclass = 0; + + pack = vzalloc((sizeof(*pack) * max_classes) + + (sizeof(*init) * (nent + 1))); + if (!pack) + return ERR_PTR(-ENOMEM); + + init = (void *)(pack + max_classes); + + for (i = 0; i < nent; i++) { + struct gf100_gr_init *ent = &init[i]; + struct gk20a_fw_av *av = &((struct gk20a_fw_av *)fuc->data)[i]; + u32 class = av->addr & 0xffff; + u32 addr = (av->addr & 0xffff0000) >> 14; + + if (prevclass != class) { + pack[classidx].init = ent; + pack[classidx].type = class; + prevclass = class; + if (++classidx >= max_classes) { + vfree(pack); + return ERR_PTR(-ENOSPC); + } + } + + ent->addr = addr; + ent->data = av->data; + ent->count = 1; + ent->pitch = 1; + } + + return pack; +} + +static int +gk20a_gr_wait_mem_scrubbing(struct gf100_gr *gr) +{ + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x40910c) & 0x00000006)) + break; + ) < 0) { + nvkm_error(subdev, "FECS mem scrubbing timeout\n"); + return -ETIMEDOUT; + } + + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x41a10c) & 0x00000006)) + break; + ) < 0) { + nvkm_error(subdev, "GPCCS mem scrubbing timeout\n"); + return -ETIMEDOUT; + } + + return 0; +} + +static void +gk20a_gr_set_hww_esr_report_mask(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_wr32(device, 0x419e44, 0x1ffffe); + nvkm_wr32(device, 0x419e4c, 0x7f); +} + +int +gk20a_gr_init(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + const u32 magicgpc918 = DIV_ROUND_UP(0x00800000, gr->tpc_total); + u32 data[TPC_MAX / 8] = {}; + u8 tpcnr[GPC_MAX]; + int gpc, tpc; + int ret, i; + + /* Clear SCC RAM */ + nvkm_wr32(device, 0x40802c, 0x1); + + gf100_gr_mmio(gr, gr->fuc_sw_nonctx); + + ret = gk20a_gr_wait_mem_scrubbing(gr); + if (ret) + return ret; + + ret = gf100_gr_wait_idle(gr); + if (ret) + return ret; + + /* MMU debug buffer */ + nvkm_wr32(device, 0x100cc8, nvkm_memory_addr(gr->unk4188b4) >> 8); + nvkm_wr32(device, 0x100ccc, nvkm_memory_addr(gr->unk4188b8) >> 8); + + if (gr->func->init_gpc_mmu) + gr->func->init_gpc_mmu(gr); + + /* Set the PE as stream master */ + nvkm_mask(device, 0x503018, 0x1, 0x1); + + /* Zcull init */ + memset(data, 0x00, sizeof(data)); + memcpy(tpcnr, gr->tpc_nr, sizeof(gr->tpc_nr)); + for (i = 0, gpc = -1; i < gr->tpc_total; i++) { + do { + gpc = (gpc + 1) % gr->gpc_nr; + } while (!tpcnr[gpc]); + tpc = gr->tpc_nr[gpc] - tpcnr[gpc]--; + + data[i / 8] |= tpc << ((i % 8) * 4); + } + + nvkm_wr32(device, GPC_BCAST(0x0980), data[0]); + nvkm_wr32(device, GPC_BCAST(0x0984), data[1]); + nvkm_wr32(device, GPC_BCAST(0x0988), data[2]); + nvkm_wr32(device, GPC_BCAST(0x098c), data[3]); + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + nvkm_wr32(device, GPC_UNIT(gpc, 0x0914), + gr->magic_not_rop_nr << 8 | gr->tpc_nr[gpc]); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0910), 0x00040000 | + gr->tpc_total); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0918), magicgpc918); + } + + nvkm_wr32(device, GPC_BCAST(0x3fd4), magicgpc918); + + /* Enable FIFO access */ + nvkm_wr32(device, 0x400500, 0x00010001); + + /* Enable interrupts */ + nvkm_wr32(device, 0x400100, 0xffffffff); + nvkm_wr32(device, 0x40013c, 0xffffffff); + + /* Enable FECS error interrupts */ + nvkm_wr32(device, 0x409c24, 0x000f0000); + + /* Enable hardware warning exceptions */ + nvkm_wr32(device, 0x404000, 0xc0000000); + nvkm_wr32(device, 0x404600, 0xc0000000); + + if (gr->func->set_hww_esr_report_mask) + gr->func->set_hww_esr_report_mask(gr); + + /* Enable TPC exceptions per GPC */ + nvkm_wr32(device, 0x419d0c, 0x2); + nvkm_wr32(device, 0x41ac94, (((1 << gr->tpc_total) - 1) & 0xff) << 16); + + /* Reset and enable all exceptions */ + nvkm_wr32(device, 0x400108, 0xffffffff); + nvkm_wr32(device, 0x400138, 0xffffffff); + nvkm_wr32(device, 0x400118, 0xffffffff); + nvkm_wr32(device, 0x400130, 0xffffffff); + nvkm_wr32(device, 0x40011c, 0xffffffff); + nvkm_wr32(device, 0x400134, 0xffffffff); + + gf100_gr_zbc_init(gr); + + return gf100_gr_init_ctxctl(gr); +} + +void +gk20a_gr_dtor(struct gf100_gr *gr) +{ + gk20a_gr_init_dtor(gr->fuc_method); + gk20a_gr_init_dtor(gr->fuc_bundle); + gk20a_gr_init_dtor(gr->fuc_sw_ctx); + gk20a_gr_init_dtor(gr->fuc_sw_nonctx); +} + +int +gk20a_gr_new_(const struct gf100_gr_func *func, struct nvkm_device *device, + int index, struct nvkm_gr **pgr) +{ + struct gf100_gr_fuc fuc; + struct gf100_gr *gr; + int ret; + + if (!(gr = kzalloc(sizeof(*gr), GFP_KERNEL))) + return -ENOMEM; + *pgr = &gr->base; + + ret = gf100_gr_ctor(func, device, index, gr); + if (ret) + return ret; + + ret = gf100_gr_ctor_fw(gr, "sw_nonctx", &fuc); + if (ret) + return ret; + gr->fuc_sw_nonctx = gk20a_gr_av_to_init(&fuc); + gf100_gr_dtor_fw(&fuc); + if (IS_ERR(gr->fuc_sw_nonctx)) + return PTR_ERR(gr->fuc_sw_nonctx); + + ret = gf100_gr_ctor_fw(gr, "sw_ctx", &fuc); + if (ret) + return ret; + gr->fuc_sw_ctx = gk20a_gr_aiv_to_init(&fuc); + gf100_gr_dtor_fw(&fuc); + if (IS_ERR(gr->fuc_sw_ctx)) + return PTR_ERR(gr->fuc_sw_ctx); + + ret = gf100_gr_ctor_fw(gr, "sw_bundle_init", &fuc); + if (ret) + return ret; + gr->fuc_bundle = gk20a_gr_av_to_init(&fuc); + gf100_gr_dtor_fw(&fuc); + if (IS_ERR(gr->fuc_bundle)) + return PTR_ERR(gr->fuc_bundle); + + ret = gf100_gr_ctor_fw(gr, "sw_method_init", &fuc); + if (ret) + return ret; + gr->fuc_method = gk20a_gr_av_to_method(&fuc); + gf100_gr_dtor_fw(&fuc); + if (IS_ERR(gr->fuc_method)) + return PTR_ERR(gr->fuc_method); + + return 0; +} + +static const struct gf100_gr_func +gk20a_gr = { + .dtor = gk20a_gr_dtor, + .init = gk20a_gr_init, + .set_hww_esr_report_mask = gk20a_gr_set_hww_esr_report_mask, .ppc_nr = 1, -}.base; + .grctx = &gk20a_grctx, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, KEPLER_INLINE_TO_MEMORY_A }, + { -1, -1, KEPLER_C, &gf100_fermi }, + { -1, -1, KEPLER_COMPUTE_A }, + {} + } +}; + +int +gk20a_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return gk20a_gr_new_(&gk20a_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm107.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm107.c index a5ebd459b..56e960212 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm107.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm107.c @@ -30,19 +30,6 @@ #include <nvif/class.h> /******************************************************************************* - * Graphics object classes - ******************************************************************************/ - -static struct nvkm_oclass -gm107_gr_sclass[] = { - { FERMI_TWOD_A, &nvkm_object_ofuncs }, - { KEPLER_INLINE_TO_MEMORY_B, &nvkm_object_ofuncs }, - { MAXWELL_A, &gf100_fermi_ofuncs, gf100_gr_9097_omthds }, - { MAXWELL_COMPUTE_A, &nvkm_object_ofuncs, gf100_gr_90c0_omthds }, - {} -}; - -/******************************************************************************* * PGRAPH register lists ******************************************************************************/ @@ -292,7 +279,7 @@ gm107_gr_pack_mmio[] = { ******************************************************************************/ void -gm107_gr_init_bios(struct gf100_gr_priv *priv) +gm107_gr_init_bios(struct gf100_gr *gr) { static const struct { u32 ctrl; @@ -304,7 +291,8 @@ gm107_gr_init_bios(struct gf100_gr_priv *priv) { 0x419af0, 0x419af4 }, { 0x419af8, 0x419afc }, }; - struct nvkm_bios *bios = nvkm_bios(priv); + struct nvkm_device *device = gr->base.engine.subdev.device; + struct nvkm_bios *bios = device->bios; struct nvbios_P0260E infoE; struct nvbios_P0260X infoX; int E = -1, X; @@ -312,124 +300,119 @@ gm107_gr_init_bios(struct gf100_gr_priv *priv) while (nvbios_P0260Ep(bios, ++E, &ver, &hdr, &infoE)) { if (X = -1, E < ARRAY_SIZE(regs)) { - nv_wr32(priv, regs[E].ctrl, infoE.data); + nvkm_wr32(device, regs[E].ctrl, infoE.data); while (nvbios_P0260Xp(bios, ++X, &ver, &hdr, &infoX)) - nv_wr32(priv, regs[E].data, infoX.data); + nvkm_wr32(device, regs[E].data, infoX.data); } } } int -gm107_gr_init(struct nvkm_object *object) +gm107_gr_init(struct gf100_gr *gr) { - struct gf100_gr_oclass *oclass = (void *)object->oclass; - struct gf100_gr_priv *priv = (void *)object; - const u32 magicgpc918 = DIV_ROUND_UP(0x00800000, priv->tpc_total); + struct nvkm_device *device = gr->base.engine.subdev.device; + const u32 magicgpc918 = DIV_ROUND_UP(0x00800000, gr->tpc_total); u32 data[TPC_MAX / 8] = {}; u8 tpcnr[GPC_MAX]; int gpc, tpc, ppc, rop; - int ret, i; - - ret = nvkm_gr_init(&priv->base); - if (ret) - return ret; + int i; - nv_wr32(priv, GPC_BCAST(0x0880), 0x00000000); - nv_wr32(priv, GPC_BCAST(0x0890), 0x00000000); - nv_wr32(priv, GPC_BCAST(0x0894), 0x00000000); - nv_wr32(priv, GPC_BCAST(0x08b4), priv->unk4188b4->addr >> 8); - nv_wr32(priv, GPC_BCAST(0x08b8), priv->unk4188b8->addr >> 8); + nvkm_wr32(device, GPC_BCAST(0x0880), 0x00000000); + nvkm_wr32(device, GPC_BCAST(0x0890), 0x00000000); + nvkm_wr32(device, GPC_BCAST(0x0894), 0x00000000); + nvkm_wr32(device, GPC_BCAST(0x08b4), nvkm_memory_addr(gr->unk4188b4) >> 8); + nvkm_wr32(device, GPC_BCAST(0x08b8), nvkm_memory_addr(gr->unk4188b8) >> 8); - gf100_gr_mmio(priv, oclass->mmio); + gf100_gr_mmio(gr, gr->func->mmio); - gm107_gr_init_bios(priv); + gm107_gr_init_bios(gr); - nv_wr32(priv, GPC_UNIT(0, 0x3018), 0x00000001); + nvkm_wr32(device, GPC_UNIT(0, 0x3018), 0x00000001); memset(data, 0x00, sizeof(data)); - memcpy(tpcnr, priv->tpc_nr, sizeof(priv->tpc_nr)); - for (i = 0, gpc = -1; i < priv->tpc_total; i++) { + memcpy(tpcnr, gr->tpc_nr, sizeof(gr->tpc_nr)); + for (i = 0, gpc = -1; i < gr->tpc_total; i++) { do { - gpc = (gpc + 1) % priv->gpc_nr; + gpc = (gpc + 1) % gr->gpc_nr; } while (!tpcnr[gpc]); - tpc = priv->tpc_nr[gpc] - tpcnr[gpc]--; + tpc = gr->tpc_nr[gpc] - tpcnr[gpc]--; data[i / 8] |= tpc << ((i % 8) * 4); } - nv_wr32(priv, GPC_BCAST(0x0980), data[0]); - nv_wr32(priv, GPC_BCAST(0x0984), data[1]); - nv_wr32(priv, GPC_BCAST(0x0988), data[2]); - nv_wr32(priv, GPC_BCAST(0x098c), data[3]); - - for (gpc = 0; gpc < priv->gpc_nr; gpc++) { - nv_wr32(priv, GPC_UNIT(gpc, 0x0914), - priv->magic_not_rop_nr << 8 | priv->tpc_nr[gpc]); - nv_wr32(priv, GPC_UNIT(gpc, 0x0910), 0x00040000 | - priv->tpc_total); - nv_wr32(priv, GPC_UNIT(gpc, 0x0918), magicgpc918); + nvkm_wr32(device, GPC_BCAST(0x0980), data[0]); + nvkm_wr32(device, GPC_BCAST(0x0984), data[1]); + nvkm_wr32(device, GPC_BCAST(0x0988), data[2]); + nvkm_wr32(device, GPC_BCAST(0x098c), data[3]); + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + nvkm_wr32(device, GPC_UNIT(gpc, 0x0914), + gr->magic_not_rop_nr << 8 | gr->tpc_nr[gpc]); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0910), 0x00040000 | + gr->tpc_total); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0918), magicgpc918); } - nv_wr32(priv, GPC_BCAST(0x3fd4), magicgpc918); - nv_wr32(priv, GPC_BCAST(0x08ac), nv_rd32(priv, 0x100800)); - - nv_wr32(priv, 0x400500, 0x00010001); - - nv_wr32(priv, 0x400100, 0xffffffff); - nv_wr32(priv, 0x40013c, 0xffffffff); - nv_wr32(priv, 0x400124, 0x00000002); - nv_wr32(priv, 0x409c24, 0x000e0000); - - nv_wr32(priv, 0x404000, 0xc0000000); - nv_wr32(priv, 0x404600, 0xc0000000); - nv_wr32(priv, 0x408030, 0xc0000000); - nv_wr32(priv, 0x404490, 0xc0000000); - nv_wr32(priv, 0x406018, 0xc0000000); - nv_wr32(priv, 0x407020, 0x40000000); - nv_wr32(priv, 0x405840, 0xc0000000); - nv_wr32(priv, 0x405844, 0x00ffffff); - nv_mask(priv, 0x419cc0, 0x00000008, 0x00000008); - - for (gpc = 0; gpc < priv->gpc_nr; gpc++) { - for (ppc = 0; ppc < 2 /* priv->ppc_nr[gpc] */; ppc++) - nv_wr32(priv, PPC_UNIT(gpc, ppc, 0x038), 0xc0000000); - nv_wr32(priv, GPC_UNIT(gpc, 0x0420), 0xc0000000); - nv_wr32(priv, GPC_UNIT(gpc, 0x0900), 0xc0000000); - nv_wr32(priv, GPC_UNIT(gpc, 0x1028), 0xc0000000); - nv_wr32(priv, GPC_UNIT(gpc, 0x0824), 0xc0000000); - for (tpc = 0; tpc < priv->tpc_nr[gpc]; tpc++) { - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x508), 0xffffffff); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x50c), 0xffffffff); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x224), 0xc0000000); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x48c), 0xc0000000); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x084), 0xc0000000); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x430), 0xc0000000); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x644), 0x00dffffe); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x64c), 0x00000005); + nvkm_wr32(device, GPC_BCAST(0x3fd4), magicgpc918); + nvkm_wr32(device, GPC_BCAST(0x08ac), nvkm_rd32(device, 0x100800)); + + nvkm_wr32(device, 0x400500, 0x00010001); + + nvkm_wr32(device, 0x400100, 0xffffffff); + nvkm_wr32(device, 0x40013c, 0xffffffff); + nvkm_wr32(device, 0x400124, 0x00000002); + nvkm_wr32(device, 0x409c24, 0x000e0000); + + nvkm_wr32(device, 0x404000, 0xc0000000); + nvkm_wr32(device, 0x404600, 0xc0000000); + nvkm_wr32(device, 0x408030, 0xc0000000); + nvkm_wr32(device, 0x404490, 0xc0000000); + nvkm_wr32(device, 0x406018, 0xc0000000); + nvkm_wr32(device, 0x407020, 0x40000000); + nvkm_wr32(device, 0x405840, 0xc0000000); + nvkm_wr32(device, 0x405844, 0x00ffffff); + nvkm_mask(device, 0x419cc0, 0x00000008, 0x00000008); + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + for (ppc = 0; ppc < 2 /* gr->ppc_nr[gpc] */; ppc++) + nvkm_wr32(device, PPC_UNIT(gpc, ppc, 0x038), 0xc0000000); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0420), 0xc0000000); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0900), 0xc0000000); + nvkm_wr32(device, GPC_UNIT(gpc, 0x1028), 0xc0000000); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0824), 0xc0000000); + for (tpc = 0; tpc < gr->tpc_nr[gpc]; tpc++) { + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x508), 0xffffffff); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x50c), 0xffffffff); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x224), 0xc0000000); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x48c), 0xc0000000); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x084), 0xc0000000); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x430), 0xc0000000); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x644), 0x00dffffe); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x64c), 0x00000005); } - nv_wr32(priv, GPC_UNIT(gpc, 0x2c90), 0xffffffff); - nv_wr32(priv, GPC_UNIT(gpc, 0x2c94), 0xffffffff); + nvkm_wr32(device, GPC_UNIT(gpc, 0x2c90), 0xffffffff); + nvkm_wr32(device, GPC_UNIT(gpc, 0x2c94), 0xffffffff); } - for (rop = 0; rop < priv->rop_nr; rop++) { - nv_wr32(priv, ROP_UNIT(rop, 0x144), 0x40000000); - nv_wr32(priv, ROP_UNIT(rop, 0x070), 0x40000000); - nv_wr32(priv, ROP_UNIT(rop, 0x204), 0xffffffff); - nv_wr32(priv, ROP_UNIT(rop, 0x208), 0xffffffff); + for (rop = 0; rop < gr->rop_nr; rop++) { + nvkm_wr32(device, ROP_UNIT(rop, 0x144), 0x40000000); + nvkm_wr32(device, ROP_UNIT(rop, 0x070), 0x40000000); + nvkm_wr32(device, ROP_UNIT(rop, 0x204), 0xffffffff); + nvkm_wr32(device, ROP_UNIT(rop, 0x208), 0xffffffff); } - nv_wr32(priv, 0x400108, 0xffffffff); - nv_wr32(priv, 0x400138, 0xffffffff); - nv_wr32(priv, 0x400118, 0xffffffff); - nv_wr32(priv, 0x400130, 0xffffffff); - nv_wr32(priv, 0x40011c, 0xffffffff); - nv_wr32(priv, 0x400134, 0xffffffff); + nvkm_wr32(device, 0x400108, 0xffffffff); + nvkm_wr32(device, 0x400138, 0xffffffff); + nvkm_wr32(device, 0x400118, 0xffffffff); + nvkm_wr32(device, 0x400130, 0xffffffff); + nvkm_wr32(device, 0x40011c, 0xffffffff); + nvkm_wr32(device, 0x400134, 0xffffffff); - nv_wr32(priv, 0x400054, 0x2c350f63); + nvkm_wr32(device, 0x400054, 0x2c350f63); - gf100_gr_zbc_init(priv); + gf100_gr_zbc_init(gr); - return gf100_gr_init_ctxctl(priv); + return gf100_gr_init_ctxctl(gr); } #include "fuc/hubgm107.fuc5.h" @@ -452,19 +435,25 @@ gm107_gr_gpccs_ucode = { .data.size = sizeof(gm107_grgpc_data), }; -struct nvkm_oclass * -gm107_gr_oclass = &(struct gf100_gr_oclass) { - .base.handle = NV_ENGINE(GR, 0x07), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gf100_gr_ctor, - .dtor = gf100_gr_dtor, - .init = gm107_gr_init, - .fini = _nvkm_gr_fini, - }, - .cclass = &gm107_grctx_oclass, - .sclass = gm107_gr_sclass, +static const struct gf100_gr_func +gm107_gr = { + .init = gm107_gr_init, .mmio = gm107_gr_pack_mmio, .fecs.ucode = &gm107_gr_fecs_ucode, .gpccs.ucode = &gm107_gr_gpccs_ucode, .ppc_nr = 2, -}.base; + .grctx = &gm107_grctx, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, KEPLER_INLINE_TO_MEMORY_B }, + { -1, -1, MAXWELL_A, &gf100_fermi }, + { -1, -1, MAXWELL_COMPUTE_A }, + {} + } +}; + +int +gm107_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(&gm107_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm204.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm204.c index fdb1dcf16..90381dde4 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm204.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm204.c @@ -27,19 +27,6 @@ #include <nvif/class.h> /******************************************************************************* - * Graphics object classes - ******************************************************************************/ - -struct nvkm_oclass -gm204_gr_sclass[] = { - { FERMI_TWOD_A, &nvkm_object_ofuncs }, - { KEPLER_INLINE_TO_MEMORY_B, &nvkm_object_ofuncs }, - { MAXWELL_B, &gf100_fermi_ofuncs, gf100_gr_9097_omthds }, - { MAXWELL_COMPUTE_B, &nvkm_object_ofuncs, gf100_gr_90c0_omthds }, - {} -}; - -/******************************************************************************* * PGRAPH register lists ******************************************************************************/ @@ -243,144 +230,144 @@ gm204_gr_data[] = { ******************************************************************************/ static int -gm204_gr_init_ctxctl(struct gf100_gr_priv *priv) +gm204_gr_init_ctxctl(struct gf100_gr *gr) { return 0; } int -gm204_gr_init(struct nvkm_object *object) +gm204_gr_init(struct gf100_gr *gr) { - struct gf100_gr_oclass *oclass = (void *)object->oclass; - struct gf100_gr_priv *priv = (void *)object; - const u32 magicgpc918 = DIV_ROUND_UP(0x00800000, priv->tpc_total); - u32 data[TPC_MAX / 8] = {}; + struct nvkm_device *device = gr->base.engine.subdev.device; + const u32 magicgpc918 = DIV_ROUND_UP(0x00800000, gr->tpc_total); + u32 data[TPC_MAX / 8] = {}, tmp; u8 tpcnr[GPC_MAX]; int gpc, tpc, ppc, rop; - int ret, i; - u32 tmp; - - ret = nvkm_gr_init(&priv->base); - if (ret) - return ret; + int i; - tmp = nv_rd32(priv, 0x100c80); /*XXX: mask? */ - nv_wr32(priv, 0x418880, 0x00001000 | (tmp & 0x00000fff)); - nv_wr32(priv, 0x418890, 0x00000000); - nv_wr32(priv, 0x418894, 0x00000000); - nv_wr32(priv, 0x4188b4, priv->unk4188b4->addr >> 8); - nv_wr32(priv, 0x4188b8, priv->unk4188b8->addr >> 8); - nv_mask(priv, 0x4188b0, 0x00040000, 0x00040000); + tmp = nvkm_rd32(device, 0x100c80); /*XXX: mask? */ + nvkm_wr32(device, 0x418880, 0x00001000 | (tmp & 0x00000fff)); + nvkm_wr32(device, 0x418890, 0x00000000); + nvkm_wr32(device, 0x418894, 0x00000000); + nvkm_wr32(device, 0x4188b4, nvkm_memory_addr(gr->unk4188b4) >> 8); + nvkm_wr32(device, 0x4188b8, nvkm_memory_addr(gr->unk4188b8) >> 8); + nvkm_mask(device, 0x4188b0, 0x00040000, 0x00040000); /*XXX: belongs in fb */ - nv_wr32(priv, 0x100cc8, priv->unk4188b4->addr >> 8); - nv_wr32(priv, 0x100ccc, priv->unk4188b8->addr >> 8); - nv_mask(priv, 0x100cc4, 0x00040000, 0x00040000); + nvkm_wr32(device, 0x100cc8, nvkm_memory_addr(gr->unk4188b4) >> 8); + nvkm_wr32(device, 0x100ccc, nvkm_memory_addr(gr->unk4188b8) >> 8); + nvkm_mask(device, 0x100cc4, 0x00040000, 0x00040000); - gf100_gr_mmio(priv, oclass->mmio); + gf100_gr_mmio(gr, gr->func->mmio); - gm107_gr_init_bios(priv); + gm107_gr_init_bios(gr); - nv_wr32(priv, GPC_UNIT(0, 0x3018), 0x00000001); + nvkm_wr32(device, GPC_UNIT(0, 0x3018), 0x00000001); memset(data, 0x00, sizeof(data)); - memcpy(tpcnr, priv->tpc_nr, sizeof(priv->tpc_nr)); - for (i = 0, gpc = -1; i < priv->tpc_total; i++) { + memcpy(tpcnr, gr->tpc_nr, sizeof(gr->tpc_nr)); + for (i = 0, gpc = -1; i < gr->tpc_total; i++) { do { - gpc = (gpc + 1) % priv->gpc_nr; + gpc = (gpc + 1) % gr->gpc_nr; } while (!tpcnr[gpc]); - tpc = priv->tpc_nr[gpc] - tpcnr[gpc]--; + tpc = gr->tpc_nr[gpc] - tpcnr[gpc]--; data[i / 8] |= tpc << ((i % 8) * 4); } - nv_wr32(priv, GPC_BCAST(0x0980), data[0]); - nv_wr32(priv, GPC_BCAST(0x0984), data[1]); - nv_wr32(priv, GPC_BCAST(0x0988), data[2]); - nv_wr32(priv, GPC_BCAST(0x098c), data[3]); - - for (gpc = 0; gpc < priv->gpc_nr; gpc++) { - nv_wr32(priv, GPC_UNIT(gpc, 0x0914), - priv->magic_not_rop_nr << 8 | priv->tpc_nr[gpc]); - nv_wr32(priv, GPC_UNIT(gpc, 0x0910), 0x00040000 | - priv->tpc_total); - nv_wr32(priv, GPC_UNIT(gpc, 0x0918), magicgpc918); + nvkm_wr32(device, GPC_BCAST(0x0980), data[0]); + nvkm_wr32(device, GPC_BCAST(0x0984), data[1]); + nvkm_wr32(device, GPC_BCAST(0x0988), data[2]); + nvkm_wr32(device, GPC_BCAST(0x098c), data[3]); + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + nvkm_wr32(device, GPC_UNIT(gpc, 0x0914), + gr->magic_not_rop_nr << 8 | gr->tpc_nr[gpc]); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0910), 0x00040000 | + gr->tpc_total); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0918), magicgpc918); } - nv_wr32(priv, GPC_BCAST(0x3fd4), magicgpc918); - nv_wr32(priv, GPC_BCAST(0x08ac), nv_rd32(priv, 0x100800)); - nv_wr32(priv, GPC_BCAST(0x033c), nv_rd32(priv, 0x100804)); - - nv_wr32(priv, 0x400500, 0x00010001); - nv_wr32(priv, 0x400100, 0xffffffff); - nv_wr32(priv, 0x40013c, 0xffffffff); - nv_wr32(priv, 0x400124, 0x00000002); - nv_wr32(priv, 0x409c24, 0x000e0000); - nv_wr32(priv, 0x405848, 0xc0000000); - nv_wr32(priv, 0x40584c, 0x00000001); - nv_wr32(priv, 0x404000, 0xc0000000); - nv_wr32(priv, 0x404600, 0xc0000000); - nv_wr32(priv, 0x408030, 0xc0000000); - nv_wr32(priv, 0x404490, 0xc0000000); - nv_wr32(priv, 0x406018, 0xc0000000); - nv_wr32(priv, 0x407020, 0x40000000); - nv_wr32(priv, 0x405840, 0xc0000000); - nv_wr32(priv, 0x405844, 0x00ffffff); - nv_mask(priv, 0x419cc0, 0x00000008, 0x00000008); - - for (gpc = 0; gpc < priv->gpc_nr; gpc++) { - for (ppc = 0; ppc < priv->ppc_nr[gpc]; ppc++) - nv_wr32(priv, PPC_UNIT(gpc, ppc, 0x038), 0xc0000000); - nv_wr32(priv, GPC_UNIT(gpc, 0x0420), 0xc0000000); - nv_wr32(priv, GPC_UNIT(gpc, 0x0900), 0xc0000000); - nv_wr32(priv, GPC_UNIT(gpc, 0x1028), 0xc0000000); - nv_wr32(priv, GPC_UNIT(gpc, 0x0824), 0xc0000000); - for (tpc = 0; tpc < priv->tpc_nr[gpc]; tpc++) { - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x508), 0xffffffff); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x50c), 0xffffffff); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x224), 0xc0000000); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x48c), 0xc0000000); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x084), 0xc0000000); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x430), 0xc0000000); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x644), 0x00dffffe); - nv_wr32(priv, TPC_UNIT(gpc, tpc, 0x64c), 0x00000005); + nvkm_wr32(device, GPC_BCAST(0x3fd4), magicgpc918); + nvkm_wr32(device, GPC_BCAST(0x08ac), nvkm_rd32(device, 0x100800)); + nvkm_wr32(device, GPC_BCAST(0x033c), nvkm_rd32(device, 0x100804)); + + nvkm_wr32(device, 0x400500, 0x00010001); + nvkm_wr32(device, 0x400100, 0xffffffff); + nvkm_wr32(device, 0x40013c, 0xffffffff); + nvkm_wr32(device, 0x400124, 0x00000002); + nvkm_wr32(device, 0x409c24, 0x000e0000); + nvkm_wr32(device, 0x405848, 0xc0000000); + nvkm_wr32(device, 0x40584c, 0x00000001); + nvkm_wr32(device, 0x404000, 0xc0000000); + nvkm_wr32(device, 0x404600, 0xc0000000); + nvkm_wr32(device, 0x408030, 0xc0000000); + nvkm_wr32(device, 0x404490, 0xc0000000); + nvkm_wr32(device, 0x406018, 0xc0000000); + nvkm_wr32(device, 0x407020, 0x40000000); + nvkm_wr32(device, 0x405840, 0xc0000000); + nvkm_wr32(device, 0x405844, 0x00ffffff); + nvkm_mask(device, 0x419cc0, 0x00000008, 0x00000008); + + for (gpc = 0; gpc < gr->gpc_nr; gpc++) { + for (ppc = 0; ppc < gr->ppc_nr[gpc]; ppc++) + nvkm_wr32(device, PPC_UNIT(gpc, ppc, 0x038), 0xc0000000); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0420), 0xc0000000); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0900), 0xc0000000); + nvkm_wr32(device, GPC_UNIT(gpc, 0x1028), 0xc0000000); + nvkm_wr32(device, GPC_UNIT(gpc, 0x0824), 0xc0000000); + for (tpc = 0; tpc < gr->tpc_nr[gpc]; tpc++) { + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x508), 0xffffffff); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x50c), 0xffffffff); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x224), 0xc0000000); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x48c), 0xc0000000); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x084), 0xc0000000); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x430), 0xc0000000); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x644), 0x00dffffe); + nvkm_wr32(device, TPC_UNIT(gpc, tpc, 0x64c), 0x00000005); } - nv_wr32(priv, GPC_UNIT(gpc, 0x2c90), 0xffffffff); - nv_wr32(priv, GPC_UNIT(gpc, 0x2c94), 0xffffffff); + nvkm_wr32(device, GPC_UNIT(gpc, 0x2c90), 0xffffffff); + nvkm_wr32(device, GPC_UNIT(gpc, 0x2c94), 0xffffffff); } - for (rop = 0; rop < priv->rop_nr; rop++) { - nv_wr32(priv, ROP_UNIT(rop, 0x144), 0x40000000); - nv_wr32(priv, ROP_UNIT(rop, 0x070), 0x40000000); - nv_wr32(priv, ROP_UNIT(rop, 0x204), 0xffffffff); - nv_wr32(priv, ROP_UNIT(rop, 0x208), 0xffffffff); + for (rop = 0; rop < gr->rop_nr; rop++) { + nvkm_wr32(device, ROP_UNIT(rop, 0x144), 0x40000000); + nvkm_wr32(device, ROP_UNIT(rop, 0x070), 0x40000000); + nvkm_wr32(device, ROP_UNIT(rop, 0x204), 0xffffffff); + nvkm_wr32(device, ROP_UNIT(rop, 0x208), 0xffffffff); } - nv_wr32(priv, 0x400108, 0xffffffff); - nv_wr32(priv, 0x400138, 0xffffffff); - nv_wr32(priv, 0x400118, 0xffffffff); - nv_wr32(priv, 0x400130, 0xffffffff); - nv_wr32(priv, 0x40011c, 0xffffffff); - nv_wr32(priv, 0x400134, 0xffffffff); + nvkm_wr32(device, 0x400108, 0xffffffff); + nvkm_wr32(device, 0x400138, 0xffffffff); + nvkm_wr32(device, 0x400118, 0xffffffff); + nvkm_wr32(device, 0x400130, 0xffffffff); + nvkm_wr32(device, 0x40011c, 0xffffffff); + nvkm_wr32(device, 0x400134, 0xffffffff); - nv_wr32(priv, 0x400054, 0x2c350f63); + nvkm_wr32(device, 0x400054, 0x2c350f63); - gf100_gr_zbc_init(priv); + gf100_gr_zbc_init(gr); - return gm204_gr_init_ctxctl(priv); + return gm204_gr_init_ctxctl(gr); } -struct nvkm_oclass * -gm204_gr_oclass = &(struct gf100_gr_oclass) { - .base.handle = NV_ENGINE(GR, 0x24), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gf100_gr_ctor, - .dtor = gf100_gr_dtor, - .init = gm204_gr_init, - .fini = _nvkm_gr_fini, - }, - .cclass = &gm204_grctx_oclass, - .sclass = gm204_gr_sclass, +static const struct gf100_gr_func +gm204_gr = { + .init = gm204_gr_init, .mmio = gm204_gr_pack_mmio, .ppc_nr = 2, -}.base; + .grctx = &gm204_grctx, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, KEPLER_INLINE_TO_MEMORY_B }, + { -1, -1, MAXWELL_B, &gf100_fermi }, + { -1, -1, MAXWELL_COMPUTE_B }, + {} + } +}; + +int +gm204_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(&gm204_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm206.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm206.c index 04b9733d1..341dc560a 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm206.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm206.c @@ -24,17 +24,25 @@ #include "gf100.h" #include "ctxgf100.h" -struct nvkm_oclass * -gm206_gr_oclass = &(struct gf100_gr_oclass) { - .base.handle = NV_ENGINE(GR, 0x26), - .base.ofuncs = &(struct nvkm_ofuncs) { - .ctor = gf100_gr_ctor, - .dtor = gf100_gr_dtor, - .init = gm204_gr_init, - .fini = _nvkm_gr_fini, - }, - .cclass = &gm206_grctx_oclass, - .sclass = gm204_gr_sclass, +#include <nvif/class.h> + +static const struct gf100_gr_func +gm206_gr = { + .init = gm204_gr_init, .mmio = gm204_gr_pack_mmio, .ppc_nr = 2, -}.base; + .grctx = &gm206_grctx, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, KEPLER_INLINE_TO_MEMORY_B }, + { -1, -1, MAXWELL_B, &gf100_fermi }, + { -1, -1, MAXWELL_COMPUTE_B }, + {} + } +}; + +int +gm206_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return gf100_gr_new_(&gm206_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm20b.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm20b.c new file mode 100644 index 000000000..65b6e3d1e --- /dev/null +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gm20b.c @@ -0,0 +1,83 @@ +/* + * Copyright (c) 2015, NVIDIA CORPORATION. All rights reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ +#include "gf100.h" +#include "ctxgf100.h" + +#include <subdev/timer.h> + +#include <nvif/class.h> + +static void +gm20b_gr_init_gpc_mmu(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + u32 val; + + /* TODO this needs to be removed once secure boot works */ + if (1) { + nvkm_wr32(device, 0x100ce4, 0xffffffff); + } + + /* TODO update once secure boot works */ + val = nvkm_rd32(device, 0x100c80); + val &= 0xf000087f; + nvkm_wr32(device, 0x418880, val); + nvkm_wr32(device, 0x418890, 0); + nvkm_wr32(device, 0x418894, 0); + + nvkm_wr32(device, 0x4188b0, nvkm_rd32(device, 0x100cc4)); + nvkm_wr32(device, 0x4188b4, nvkm_rd32(device, 0x100cc8)); + nvkm_wr32(device, 0x4188b8, nvkm_rd32(device, 0x100ccc)); + + nvkm_wr32(device, 0x4188ac, nvkm_rd32(device, 0x100800)); +} + +static void +gm20b_gr_set_hww_esr_report_mask(struct gf100_gr *gr) +{ + struct nvkm_device *device = gr->base.engine.subdev.device; + nvkm_wr32(device, 0x419e44, 0xdffffe); + nvkm_wr32(device, 0x419e4c, 0x5); +} + +static const struct gf100_gr_func +gm20b_gr = { + .dtor = gk20a_gr_dtor, + .init = gk20a_gr_init, + .init_gpc_mmu = gm20b_gr_init_gpc_mmu, + .set_hww_esr_report_mask = gm20b_gr_set_hww_esr_report_mask, + .ppc_nr = 1, + .grctx = &gm20b_grctx, + .sclass = { + { -1, -1, FERMI_TWOD_A }, + { -1, -1, KEPLER_INLINE_TO_MEMORY_B }, + { -1, -1, MAXWELL_B, &gf100_fermi }, + { -1, -1, MAXWELL_COMPUTE_B }, + {} + } +}; + +int +gm20b_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return gk20a_gr_new_(&gm20b_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gt200.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gt200.c new file mode 100644 index 000000000..2e68919f0 --- /dev/null +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gt200.c @@ -0,0 +1,47 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "nv50.h" + +static const struct nvkm_gr_func +gt200_gr = { + .init = nv50_gr_init, + .intr = nv50_gr_intr, + .chan_new = nv50_gr_chan_new, + .tlb_flush = g84_gr_tlb_flush, + .units = nv50_gr_units, + .sclass = { + { -1, -1, 0x0030, &nv50_gr_object }, + { -1, -1, 0x502d, &nv50_gr_object }, + { -1, -1, 0x5039, &nv50_gr_object }, + { -1, -1, 0x50c0, &nv50_gr_object }, + { -1, -1, 0x8397, &nv50_gr_object }, + {} + } +}; + +int +gt200_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return nv50_gr_new_(>200_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gt215.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gt215.c new file mode 100644 index 000000000..2bf7aac36 --- /dev/null +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/gt215.c @@ -0,0 +1,48 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "nv50.h" + +static const struct nvkm_gr_func +gt215_gr = { + .init = nv50_gr_init, + .intr = nv50_gr_intr, + .chan_new = nv50_gr_chan_new, + .tlb_flush = g84_gr_tlb_flush, + .units = nv50_gr_units, + .sclass = { + { -1, -1, 0x0030, &nv50_gr_object }, + { -1, -1, 0x502d, &nv50_gr_object }, + { -1, -1, 0x5039, &nv50_gr_object }, + { -1, -1, 0x50c0, &nv50_gr_object }, + { -1, -1, 0x8597, &nv50_gr_object }, + { -1, -1, 0x85c0, &nv50_gr_object }, + {} + } +}; + +int +gt215_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return nv50_gr_new_(>215_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/mcp79.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/mcp79.c new file mode 100644 index 000000000..95d5219fa --- /dev/null +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/mcp79.c @@ -0,0 +1,46 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "nv50.h" + +static const struct nvkm_gr_func +mcp79_gr = { + .init = nv50_gr_init, + .intr = nv50_gr_intr, + .chan_new = nv50_gr_chan_new, + .units = nv50_gr_units, + .sclass = { + { -1, -1, 0x0030, &nv50_gr_object }, + { -1, -1, 0x502d, &nv50_gr_object }, + { -1, -1, 0x5039, &nv50_gr_object }, + { -1, -1, 0x50c0, &nv50_gr_object }, + { -1, -1, 0x8397, &nv50_gr_object }, + {} + } +}; + +int +mcp79_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return nv50_gr_new_(&mcp79_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/mcp89.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/mcp89.c new file mode 100644 index 000000000..027b58e59 --- /dev/null +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/mcp89.c @@ -0,0 +1,48 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "nv50.h" + +static const struct nvkm_gr_func +mcp89_gr = { + .init = nv50_gr_init, + .intr = nv50_gr_intr, + .chan_new = nv50_gr_chan_new, + .tlb_flush = g84_gr_tlb_flush, + .units = nv50_gr_units, + .sclass = { + { -1, -1, 0x0030, &nv50_gr_object }, + { -1, -1, 0x502d, &nv50_gr_object }, + { -1, -1, 0x5039, &nv50_gr_object }, + { -1, -1, 0x50c0, &nv50_gr_object }, + { -1, -1, 0x85c0, &nv50_gr_object }, + { -1, -1, 0x8697, &nv50_gr_object }, + {} + } +}; + +int +mcp89_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return nv50_gr_new_(&mcp89_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv04.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv04.c index 2614510c2..85c5b7fea 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv04.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv04.c @@ -21,13 +21,13 @@ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER * DEALINGS IN THE SOFTWARE. */ -#include <engine/gr.h> +#include "priv.h" #include "regs.h" #include <core/client.h> -#include <core/device.h> -#include <core/handle.h> +#include <core/gpuobj.h> #include <engine/fifo.h> +#include <engine/fifo/chan.h> #include <subdev/instmem.h> #include <subdev/timer.h> @@ -346,25 +346,23 @@ nv04_gr_ctx_regs[] = { NV04_PGRAPH_DEBUG_3 }; -struct nv04_gr_priv { +#define nv04_gr(p) container_of((p), struct nv04_gr, base) + +struct nv04_gr { struct nvkm_gr base; struct nv04_gr_chan *chan[16]; spinlock_t lock; }; +#define nv04_gr_chan(p) container_of((p), struct nv04_gr_chan, object) + struct nv04_gr_chan { - struct nvkm_object base; + struct nvkm_object object; + struct nv04_gr *gr; int chid; u32 nv04[ARRAY_SIZE(nv04_gr_ctx_regs)]; }; - -static inline struct nv04_gr_priv * -nv04_gr_priv(struct nv04_gr_chan *chan) -{ - return (void *)nv_object(chan)->engine; -} - /******************************************************************************* * Graphics object classes ******************************************************************************/ @@ -444,35 +442,34 @@ nv04_gr_priv(struct nv04_gr_chan *chan) */ static void -nv04_gr_set_ctx1(struct nvkm_object *object, u32 mask, u32 value) +nv04_gr_set_ctx1(struct nvkm_device *device, u32 inst, u32 mask, u32 value) { - struct nv04_gr_priv *priv = (void *)object->engine; - int subc = (nv_rd32(priv, NV04_PGRAPH_TRAPPED_ADDR) >> 13) & 0x7; + int subc = (nvkm_rd32(device, NV04_PGRAPH_TRAPPED_ADDR) >> 13) & 0x7; u32 tmp; - tmp = nv_ro32(object, 0x00); + tmp = nvkm_rd32(device, 0x700000 + inst); tmp &= ~mask; tmp |= value; - nv_wo32(object, 0x00, tmp); + nvkm_wr32(device, 0x700000 + inst, tmp); - nv_wr32(priv, NV04_PGRAPH_CTX_SWITCH1, tmp); - nv_wr32(priv, NV04_PGRAPH_CTX_CACHE1 + (subc<<2), tmp); + nvkm_wr32(device, NV04_PGRAPH_CTX_SWITCH1, tmp); + nvkm_wr32(device, NV04_PGRAPH_CTX_CACHE1 + (subc << 2), tmp); } static void -nv04_gr_set_ctx_val(struct nvkm_object *object, u32 mask, u32 value) +nv04_gr_set_ctx_val(struct nvkm_device *device, u32 inst, u32 mask, u32 value) { int class, op, valid = 1; u32 tmp, ctx1; - ctx1 = nv_ro32(object, 0x00); + ctx1 = nvkm_rd32(device, 0x700000 + inst); class = ctx1 & 0xff; op = (ctx1 >> 15) & 7; - tmp = nv_ro32(object, 0x0c); + tmp = nvkm_rd32(device, 0x70000c + inst); tmp &= ~mask; tmp |= value; - nv_wo32(object, 0x0c, tmp); + nvkm_wr32(device, 0x70000c + inst, tmp); /* check for valid surf2d/surf_dst/surf_color */ if (!(tmp & 0x02000000)) @@ -504,527 +501,567 @@ nv04_gr_set_ctx_val(struct nvkm_object *object, u32 mask, u32 value) break; } - nv04_gr_set_ctx1(object, 0x01000000, valid << 24); + nv04_gr_set_ctx1(device, inst, 0x01000000, valid << 24); } -static int -nv04_gr_mthd_set_operation(struct nvkm_object *object, u32 mthd, - void *args, u32 size) +static bool +nv04_gr_mthd_set_operation(struct nvkm_device *device, u32 inst, u32 data) { - u32 class = nv_ro32(object, 0) & 0xff; - u32 data = *(u32 *)args; + u8 class = nvkm_rd32(device, 0x700000) & 0x000000ff; if (data > 5) - return 1; + return false; /* Old versions of the objects only accept first three operations. */ if (data > 2 && class < 0x40) - return 1; - nv04_gr_set_ctx1(object, 0x00038000, data << 15); + return false; + nv04_gr_set_ctx1(device, inst, 0x00038000, data << 15); /* changing operation changes set of objects needed for validation */ - nv04_gr_set_ctx_val(object, 0, 0); - return 0; + nv04_gr_set_ctx_val(device, inst, 0, 0); + return true; } -static int -nv04_gr_mthd_surf3d_clip_h(struct nvkm_object *object, u32 mthd, - void *args, u32 size) +static bool +nv04_gr_mthd_surf3d_clip_h(struct nvkm_device *device, u32 inst, u32 data) { - struct nv04_gr_priv *priv = (void *)object->engine; - u32 data = *(u32 *)args; u32 min = data & 0xffff, max; u32 w = data >> 16; if (min & 0x8000) /* too large */ - return 1; + return false; if (w & 0x8000) /* yes, it accepts negative for some reason. */ w |= 0xffff0000; max = min + w; max &= 0x3ffff; - nv_wr32(priv, 0x40053c, min); - nv_wr32(priv, 0x400544, max); - return 0; + nvkm_wr32(device, 0x40053c, min); + nvkm_wr32(device, 0x400544, max); + return true; } -static int -nv04_gr_mthd_surf3d_clip_v(struct nvkm_object *object, u32 mthd, - void *args, u32 size) +static bool +nv04_gr_mthd_surf3d_clip_v(struct nvkm_device *device, u32 inst, u32 data) { - struct nv04_gr_priv *priv = (void *)object->engine; - u32 data = *(u32 *)args; u32 min = data & 0xffff, max; u32 w = data >> 16; if (min & 0x8000) /* too large */ - return 1; + return false; if (w & 0x8000) /* yes, it accepts negative for some reason. */ w |= 0xffff0000; max = min + w; max &= 0x3ffff; - nv_wr32(priv, 0x400540, min); - nv_wr32(priv, 0x400548, max); - return 0; + nvkm_wr32(device, 0x400540, min); + nvkm_wr32(device, 0x400548, max); + return true; } -static u16 -nv04_gr_mthd_bind_class(struct nvkm_object *object, u32 *args, u32 size) +static u8 +nv04_gr_mthd_bind_class(struct nvkm_device *device, u32 inst) { - struct nvkm_instmem *imem = nvkm_instmem(object); - u32 inst = *(u32 *)args << 4; - return nv_ro32(imem, inst); + return nvkm_rd32(device, 0x700000 + (inst << 4)); } -static int -nv04_gr_mthd_bind_surf2d(struct nvkm_object *object, u32 mthd, - void *args, u32 size) +static bool +nv04_gr_mthd_bind_surf2d(struct nvkm_device *device, u32 inst, u32 data) { - switch (nv04_gr_mthd_bind_class(object, args, size)) { + switch (nv04_gr_mthd_bind_class(device, data)) { case 0x30: - nv04_gr_set_ctx1(object, 0x00004000, 0); - nv04_gr_set_ctx_val(object, 0x02000000, 0); - return 0; + nv04_gr_set_ctx1(device, inst, 0x00004000, 0); + nv04_gr_set_ctx_val(device, inst, 0x02000000, 0); + return true; case 0x42: - nv04_gr_set_ctx1(object, 0x00004000, 0); - nv04_gr_set_ctx_val(object, 0x02000000, 0x02000000); - return 0; + nv04_gr_set_ctx1(device, inst, 0x00004000, 0); + nv04_gr_set_ctx_val(device, inst, 0x02000000, 0x02000000); + return true; } - return 1; + return false; } -static int -nv04_gr_mthd_bind_surf2d_swzsurf(struct nvkm_object *object, u32 mthd, - void *args, u32 size) +static bool +nv04_gr_mthd_bind_surf2d_swzsurf(struct nvkm_device *device, u32 inst, u32 data) { - switch (nv04_gr_mthd_bind_class(object, args, size)) { + switch (nv04_gr_mthd_bind_class(device, data)) { case 0x30: - nv04_gr_set_ctx1(object, 0x00004000, 0); - nv04_gr_set_ctx_val(object, 0x02000000, 0); - return 0; + nv04_gr_set_ctx1(device, inst, 0x00004000, 0); + nv04_gr_set_ctx_val(device, inst, 0x02000000, 0); + return true; case 0x42: - nv04_gr_set_ctx1(object, 0x00004000, 0); - nv04_gr_set_ctx_val(object, 0x02000000, 0x02000000); - return 0; + nv04_gr_set_ctx1(device, inst, 0x00004000, 0); + nv04_gr_set_ctx_val(device, inst, 0x02000000, 0x02000000); + return true; case 0x52: - nv04_gr_set_ctx1(object, 0x00004000, 0x00004000); - nv04_gr_set_ctx_val(object, 0x02000000, 0x02000000); - return 0; + nv04_gr_set_ctx1(device, inst, 0x00004000, 0x00004000); + nv04_gr_set_ctx_val(device, inst, 0x02000000, 0x02000000); + return true; } - return 1; + return false; } -static int -nv01_gr_mthd_bind_patt(struct nvkm_object *object, u32 mthd, - void *args, u32 size) +static bool +nv01_gr_mthd_bind_patt(struct nvkm_device *device, u32 inst, u32 data) { - switch (nv04_gr_mthd_bind_class(object, args, size)) { + switch (nv04_gr_mthd_bind_class(device, data)) { case 0x30: - nv04_gr_set_ctx_val(object, 0x08000000, 0); - return 0; + nv04_gr_set_ctx_val(device, inst, 0x08000000, 0); + return true; case 0x18: - nv04_gr_set_ctx_val(object, 0x08000000, 0x08000000); - return 0; + nv04_gr_set_ctx_val(device, inst, 0x08000000, 0x08000000); + return true; } - return 1; + return false; } -static int -nv04_gr_mthd_bind_patt(struct nvkm_object *object, u32 mthd, - void *args, u32 size) +static bool +nv04_gr_mthd_bind_patt(struct nvkm_device *device, u32 inst, u32 data) { - switch (nv04_gr_mthd_bind_class(object, args, size)) { + switch (nv04_gr_mthd_bind_class(device, data)) { case 0x30: - nv04_gr_set_ctx_val(object, 0x08000000, 0); - return 0; + nv04_gr_set_ctx_val(device, inst, 0x08000000, 0); + return true; case 0x44: - nv04_gr_set_ctx_val(object, 0x08000000, 0x08000000); - return 0; + nv04_gr_set_ctx_val(device, inst, 0x08000000, 0x08000000); + return true; } - return 1; + return false; } -static int -nv04_gr_mthd_bind_rop(struct nvkm_object *object, u32 mthd, - void *args, u32 size) +static bool +nv04_gr_mthd_bind_rop(struct nvkm_device *device, u32 inst, u32 data) { - switch (nv04_gr_mthd_bind_class(object, args, size)) { + switch (nv04_gr_mthd_bind_class(device, data)) { case 0x30: - nv04_gr_set_ctx_val(object, 0x10000000, 0); - return 0; + nv04_gr_set_ctx_val(device, inst, 0x10000000, 0); + return true; case 0x43: - nv04_gr_set_ctx_val(object, 0x10000000, 0x10000000); - return 0; + nv04_gr_set_ctx_val(device, inst, 0x10000000, 0x10000000); + return true; } - return 1; + return false; } -static int -nv04_gr_mthd_bind_beta1(struct nvkm_object *object, u32 mthd, - void *args, u32 size) +static bool +nv04_gr_mthd_bind_beta1(struct nvkm_device *device, u32 inst, u32 data) { - switch (nv04_gr_mthd_bind_class(object, args, size)) { + switch (nv04_gr_mthd_bind_class(device, data)) { case 0x30: - nv04_gr_set_ctx_val(object, 0x20000000, 0); - return 0; + nv04_gr_set_ctx_val(device, inst, 0x20000000, 0); + return true; case 0x12: - nv04_gr_set_ctx_val(object, 0x20000000, 0x20000000); - return 0; + nv04_gr_set_ctx_val(device, inst, 0x20000000, 0x20000000); + return true; } - return 1; + return false; } -static int -nv04_gr_mthd_bind_beta4(struct nvkm_object *object, u32 mthd, - void *args, u32 size) +static bool +nv04_gr_mthd_bind_beta4(struct nvkm_device *device, u32 inst, u32 data) { - switch (nv04_gr_mthd_bind_class(object, args, size)) { + switch (nv04_gr_mthd_bind_class(device, data)) { case 0x30: - nv04_gr_set_ctx_val(object, 0x40000000, 0); - return 0; + nv04_gr_set_ctx_val(device, inst, 0x40000000, 0); + return true; case 0x72: - nv04_gr_set_ctx_val(object, 0x40000000, 0x40000000); - return 0; + nv04_gr_set_ctx_val(device, inst, 0x40000000, 0x40000000); + return true; } - return 1; + return false; } -static int -nv04_gr_mthd_bind_surf_dst(struct nvkm_object *object, u32 mthd, - void *args, u32 size) +static bool +nv04_gr_mthd_bind_surf_dst(struct nvkm_device *device, u32 inst, u32 data) { - switch (nv04_gr_mthd_bind_class(object, args, size)) { + switch (nv04_gr_mthd_bind_class(device, data)) { case 0x30: - nv04_gr_set_ctx_val(object, 0x02000000, 0); - return 0; + nv04_gr_set_ctx_val(device, inst, 0x02000000, 0); + return true; case 0x58: - nv04_gr_set_ctx_val(object, 0x02000000, 0x02000000); - return 0; + nv04_gr_set_ctx_val(device, inst, 0x02000000, 0x02000000); + return true; } - return 1; + return false; } -static int -nv04_gr_mthd_bind_surf_src(struct nvkm_object *object, u32 mthd, - void *args, u32 size) +static bool +nv04_gr_mthd_bind_surf_src(struct nvkm_device *device, u32 inst, u32 data) { - switch (nv04_gr_mthd_bind_class(object, args, size)) { + switch (nv04_gr_mthd_bind_class(device, data)) { case 0x30: - nv04_gr_set_ctx_val(object, 0x04000000, 0); - return 0; + nv04_gr_set_ctx_val(device, inst, 0x04000000, 0); + return true; case 0x59: - nv04_gr_set_ctx_val(object, 0x04000000, 0x04000000); - return 0; + nv04_gr_set_ctx_val(device, inst, 0x04000000, 0x04000000); + return true; } - return 1; + return false; } -static int -nv04_gr_mthd_bind_surf_color(struct nvkm_object *object, u32 mthd, - void *args, u32 size) +static bool +nv04_gr_mthd_bind_surf_color(struct nvkm_device *device, u32 inst, u32 data) { - switch (nv04_gr_mthd_bind_class(object, args, size)) { + switch (nv04_gr_mthd_bind_class(device, data)) { case 0x30: - nv04_gr_set_ctx_val(object, 0x02000000, 0); - return 0; + nv04_gr_set_ctx_val(device, inst, 0x02000000, 0); + return true; case 0x5a: - nv04_gr_set_ctx_val(object, 0x02000000, 0x02000000); - return 0; + nv04_gr_set_ctx_val(device, inst, 0x02000000, 0x02000000); + return true; } - return 1; + return false; } -static int -nv04_gr_mthd_bind_surf_zeta(struct nvkm_object *object, u32 mthd, - void *args, u32 size) +static bool +nv04_gr_mthd_bind_surf_zeta(struct nvkm_device *device, u32 inst, u32 data) { - switch (nv04_gr_mthd_bind_class(object, args, size)) { + switch (nv04_gr_mthd_bind_class(device, data)) { case 0x30: - nv04_gr_set_ctx_val(object, 0x04000000, 0); - return 0; + nv04_gr_set_ctx_val(device, inst, 0x04000000, 0); + return true; case 0x5b: - nv04_gr_set_ctx_val(object, 0x04000000, 0x04000000); - return 0; + nv04_gr_set_ctx_val(device, inst, 0x04000000, 0x04000000); + return true; } - return 1; + return false; } -static int -nv01_gr_mthd_bind_clip(struct nvkm_object *object, u32 mthd, - void *args, u32 size) +static bool +nv01_gr_mthd_bind_clip(struct nvkm_device *device, u32 inst, u32 data) { - switch (nv04_gr_mthd_bind_class(object, args, size)) { + switch (nv04_gr_mthd_bind_class(device, data)) { case 0x30: - nv04_gr_set_ctx1(object, 0x2000, 0); - return 0; + nv04_gr_set_ctx1(device, inst, 0x2000, 0); + return true; case 0x19: - nv04_gr_set_ctx1(object, 0x2000, 0x2000); - return 0; + nv04_gr_set_ctx1(device, inst, 0x2000, 0x2000); + return true; } - return 1; + return false; } -static int -nv01_gr_mthd_bind_chroma(struct nvkm_object *object, u32 mthd, - void *args, u32 size) +static bool +nv01_gr_mthd_bind_chroma(struct nvkm_device *device, u32 inst, u32 data) { - switch (nv04_gr_mthd_bind_class(object, args, size)) { + switch (nv04_gr_mthd_bind_class(device, data)) { case 0x30: - nv04_gr_set_ctx1(object, 0x1000, 0); - return 0; + nv04_gr_set_ctx1(device, inst, 0x1000, 0); + return true; /* Yes, for some reason even the old versions of objects * accept 0x57 and not 0x17. Consistency be damned. */ case 0x57: - nv04_gr_set_ctx1(object, 0x1000, 0x1000); - return 0; + nv04_gr_set_ctx1(device, inst, 0x1000, 0x1000); + return true; } - return 1; + return false; } -static struct nvkm_omthds -nv03_gr_gdi_omthds[] = { - { 0x0184, 0x0184, nv01_gr_mthd_bind_patt }, - { 0x0188, 0x0188, nv04_gr_mthd_bind_rop }, - { 0x018c, 0x018c, nv04_gr_mthd_bind_beta1 }, - { 0x0190, 0x0190, nv04_gr_mthd_bind_surf_dst }, - { 0x02fc, 0x02fc, nv04_gr_mthd_set_operation }, - {} -}; - -static struct nvkm_omthds -nv04_gr_gdi_omthds[] = { - { 0x0188, 0x0188, nv04_gr_mthd_bind_patt }, - { 0x018c, 0x018c, nv04_gr_mthd_bind_rop }, - { 0x0190, 0x0190, nv04_gr_mthd_bind_beta1 }, - { 0x0194, 0x0194, nv04_gr_mthd_bind_beta4 }, - { 0x0198, 0x0198, nv04_gr_mthd_bind_surf2d }, - { 0x02fc, 0x02fc, nv04_gr_mthd_set_operation }, - {} -}; +static bool +nv03_gr_mthd_gdi(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0184: func = nv01_gr_mthd_bind_patt; break; + case 0x0188: func = nv04_gr_mthd_bind_rop; break; + case 0x018c: func = nv04_gr_mthd_bind_beta1; break; + case 0x0190: func = nv04_gr_mthd_bind_surf_dst; break; + case 0x02fc: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} -static struct nvkm_omthds -nv01_gr_blit_omthds[] = { - { 0x0184, 0x0184, nv01_gr_mthd_bind_chroma }, - { 0x0188, 0x0188, nv01_gr_mthd_bind_clip }, - { 0x018c, 0x018c, nv01_gr_mthd_bind_patt }, - { 0x0190, 0x0190, nv04_gr_mthd_bind_rop }, - { 0x0194, 0x0194, nv04_gr_mthd_bind_beta1 }, - { 0x0198, 0x0198, nv04_gr_mthd_bind_surf_dst }, - { 0x019c, 0x019c, nv04_gr_mthd_bind_surf_src }, - { 0x02fc, 0x02fc, nv04_gr_mthd_set_operation }, - {} -}; +static bool +nv04_gr_mthd_gdi(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0188: func = nv04_gr_mthd_bind_patt; break; + case 0x018c: func = nv04_gr_mthd_bind_rop; break; + case 0x0190: func = nv04_gr_mthd_bind_beta1; break; + case 0x0194: func = nv04_gr_mthd_bind_beta4; break; + case 0x0198: func = nv04_gr_mthd_bind_surf2d; break; + case 0x02fc: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} -static struct nvkm_omthds -nv04_gr_blit_omthds[] = { - { 0x0184, 0x0184, nv01_gr_mthd_bind_chroma }, - { 0x0188, 0x0188, nv01_gr_mthd_bind_clip }, - { 0x018c, 0x018c, nv04_gr_mthd_bind_patt }, - { 0x0190, 0x0190, nv04_gr_mthd_bind_rop }, - { 0x0194, 0x0194, nv04_gr_mthd_bind_beta1 }, - { 0x0198, 0x0198, nv04_gr_mthd_bind_beta4 }, - { 0x019c, 0x019c, nv04_gr_mthd_bind_surf2d }, - { 0x02fc, 0x02fc, nv04_gr_mthd_set_operation }, - {} -}; +static bool +nv01_gr_mthd_blit(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0184: func = nv01_gr_mthd_bind_chroma; break; + case 0x0188: func = nv01_gr_mthd_bind_clip; break; + case 0x018c: func = nv01_gr_mthd_bind_patt; break; + case 0x0190: func = nv04_gr_mthd_bind_rop; break; + case 0x0194: func = nv04_gr_mthd_bind_beta1; break; + case 0x0198: func = nv04_gr_mthd_bind_surf_dst; break; + case 0x019c: func = nv04_gr_mthd_bind_surf_src; break; + case 0x02fc: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} -static struct nvkm_omthds -nv04_gr_iifc_omthds[] = { - { 0x0188, 0x0188, nv01_gr_mthd_bind_chroma }, - { 0x018c, 0x018c, nv01_gr_mthd_bind_clip }, - { 0x0190, 0x0190, nv04_gr_mthd_bind_patt }, - { 0x0194, 0x0194, nv04_gr_mthd_bind_rop }, - { 0x0198, 0x0198, nv04_gr_mthd_bind_beta1 }, - { 0x019c, 0x019c, nv04_gr_mthd_bind_beta4 }, - { 0x01a0, 0x01a0, nv04_gr_mthd_bind_surf2d_swzsurf }, - { 0x03e4, 0x03e4, nv04_gr_mthd_set_operation }, - {} -}; +static bool +nv04_gr_mthd_blit(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0184: func = nv01_gr_mthd_bind_chroma; break; + case 0x0188: func = nv01_gr_mthd_bind_clip; break; + case 0x018c: func = nv04_gr_mthd_bind_patt; break; + case 0x0190: func = nv04_gr_mthd_bind_rop; break; + case 0x0194: func = nv04_gr_mthd_bind_beta1; break; + case 0x0198: func = nv04_gr_mthd_bind_beta4; break; + case 0x019c: func = nv04_gr_mthd_bind_surf2d; break; + case 0x02fc: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} -static struct nvkm_omthds -nv01_gr_ifc_omthds[] = { - { 0x0184, 0x0184, nv01_gr_mthd_bind_chroma }, - { 0x0188, 0x0188, nv01_gr_mthd_bind_clip }, - { 0x018c, 0x018c, nv01_gr_mthd_bind_patt }, - { 0x0190, 0x0190, nv04_gr_mthd_bind_rop }, - { 0x0194, 0x0194, nv04_gr_mthd_bind_beta1 }, - { 0x0198, 0x0198, nv04_gr_mthd_bind_surf_dst }, - { 0x02fc, 0x02fc, nv04_gr_mthd_set_operation }, - {} -}; +static bool +nv04_gr_mthd_iifc(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0188: func = nv01_gr_mthd_bind_chroma; break; + case 0x018c: func = nv01_gr_mthd_bind_clip; break; + case 0x0190: func = nv04_gr_mthd_bind_patt; break; + case 0x0194: func = nv04_gr_mthd_bind_rop; break; + case 0x0198: func = nv04_gr_mthd_bind_beta1; break; + case 0x019c: func = nv04_gr_mthd_bind_beta4; break; + case 0x01a0: func = nv04_gr_mthd_bind_surf2d_swzsurf; break; + case 0x03e4: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} -static struct nvkm_omthds -nv04_gr_ifc_omthds[] = { - { 0x0184, 0x0184, nv01_gr_mthd_bind_chroma }, - { 0x0188, 0x0188, nv01_gr_mthd_bind_clip }, - { 0x018c, 0x018c, nv04_gr_mthd_bind_patt }, - { 0x0190, 0x0190, nv04_gr_mthd_bind_rop }, - { 0x0194, 0x0194, nv04_gr_mthd_bind_beta1 }, - { 0x0198, 0x0198, nv04_gr_mthd_bind_beta4 }, - { 0x019c, 0x019c, nv04_gr_mthd_bind_surf2d }, - { 0x02fc, 0x02fc, nv04_gr_mthd_set_operation }, - {} -}; +static bool +nv01_gr_mthd_ifc(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0184: func = nv01_gr_mthd_bind_chroma; break; + case 0x0188: func = nv01_gr_mthd_bind_clip; break; + case 0x018c: func = nv01_gr_mthd_bind_patt; break; + case 0x0190: func = nv04_gr_mthd_bind_rop; break; + case 0x0194: func = nv04_gr_mthd_bind_beta1; break; + case 0x0198: func = nv04_gr_mthd_bind_surf_dst; break; + case 0x02fc: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} -static struct nvkm_omthds -nv03_gr_sifc_omthds[] = { - { 0x0184, 0x0184, nv01_gr_mthd_bind_chroma }, - { 0x0188, 0x0188, nv01_gr_mthd_bind_patt }, - { 0x018c, 0x018c, nv04_gr_mthd_bind_rop }, - { 0x0190, 0x0190, nv04_gr_mthd_bind_beta1 }, - { 0x0194, 0x0194, nv04_gr_mthd_bind_surf_dst }, - { 0x02fc, 0x02fc, nv04_gr_mthd_set_operation }, - {} -}; +static bool +nv04_gr_mthd_ifc(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0184: func = nv01_gr_mthd_bind_chroma; break; + case 0x0188: func = nv01_gr_mthd_bind_clip; break; + case 0x018c: func = nv04_gr_mthd_bind_patt; break; + case 0x0190: func = nv04_gr_mthd_bind_rop; break; + case 0x0194: func = nv04_gr_mthd_bind_beta1; break; + case 0x0198: func = nv04_gr_mthd_bind_beta4; break; + case 0x019c: func = nv04_gr_mthd_bind_surf2d; break; + case 0x02fc: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} -static struct nvkm_omthds -nv04_gr_sifc_omthds[] = { - { 0x0184, 0x0184, nv01_gr_mthd_bind_chroma }, - { 0x0188, 0x0188, nv04_gr_mthd_bind_patt }, - { 0x018c, 0x018c, nv04_gr_mthd_bind_rop }, - { 0x0190, 0x0190, nv04_gr_mthd_bind_beta1 }, - { 0x0194, 0x0194, nv04_gr_mthd_bind_beta4 }, - { 0x0198, 0x0198, nv04_gr_mthd_bind_surf2d }, - { 0x02fc, 0x02fc, nv04_gr_mthd_set_operation }, - {} -}; +static bool +nv03_gr_mthd_sifc(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0184: func = nv01_gr_mthd_bind_chroma; break; + case 0x0188: func = nv01_gr_mthd_bind_patt; break; + case 0x018c: func = nv04_gr_mthd_bind_rop; break; + case 0x0190: func = nv04_gr_mthd_bind_beta1; break; + case 0x0194: func = nv04_gr_mthd_bind_surf_dst; break; + case 0x02fc: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} -static struct nvkm_omthds -nv03_gr_sifm_omthds[] = { - { 0x0188, 0x0188, nv01_gr_mthd_bind_patt }, - { 0x018c, 0x018c, nv04_gr_mthd_bind_rop }, - { 0x0190, 0x0190, nv04_gr_mthd_bind_beta1 }, - { 0x0194, 0x0194, nv04_gr_mthd_bind_surf_dst }, - { 0x0304, 0x0304, nv04_gr_mthd_set_operation }, - {} -}; +static bool +nv04_gr_mthd_sifc(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0184: func = nv01_gr_mthd_bind_chroma; break; + case 0x0188: func = nv04_gr_mthd_bind_patt; break; + case 0x018c: func = nv04_gr_mthd_bind_rop; break; + case 0x0190: func = nv04_gr_mthd_bind_beta1; break; + case 0x0194: func = nv04_gr_mthd_bind_beta4; break; + case 0x0198: func = nv04_gr_mthd_bind_surf2d; break; + case 0x02fc: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} -static struct nvkm_omthds -nv04_gr_sifm_omthds[] = { - { 0x0188, 0x0188, nv04_gr_mthd_bind_patt }, - { 0x018c, 0x018c, nv04_gr_mthd_bind_rop }, - { 0x0190, 0x0190, nv04_gr_mthd_bind_beta1 }, - { 0x0194, 0x0194, nv04_gr_mthd_bind_beta4 }, - { 0x0198, 0x0198, nv04_gr_mthd_bind_surf2d }, - { 0x0304, 0x0304, nv04_gr_mthd_set_operation }, - {} -}; +static bool +nv03_gr_mthd_sifm(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0188: func = nv01_gr_mthd_bind_patt; break; + case 0x018c: func = nv04_gr_mthd_bind_rop; break; + case 0x0190: func = nv04_gr_mthd_bind_beta1; break; + case 0x0194: func = nv04_gr_mthd_bind_surf_dst; break; + case 0x0304: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} -static struct nvkm_omthds -nv04_gr_surf3d_omthds[] = { - { 0x02f8, 0x02f8, nv04_gr_mthd_surf3d_clip_h }, - { 0x02fc, 0x02fc, nv04_gr_mthd_surf3d_clip_v }, - {} -}; +static bool +nv04_gr_mthd_sifm(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0188: func = nv04_gr_mthd_bind_patt; break; + case 0x018c: func = nv04_gr_mthd_bind_rop; break; + case 0x0190: func = nv04_gr_mthd_bind_beta1; break; + case 0x0194: func = nv04_gr_mthd_bind_beta4; break; + case 0x0198: func = nv04_gr_mthd_bind_surf2d; break; + case 0x0304: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} -static struct nvkm_omthds -nv03_gr_ttri_omthds[] = { - { 0x0188, 0x0188, nv01_gr_mthd_bind_clip }, - { 0x018c, 0x018c, nv04_gr_mthd_bind_surf_color }, - { 0x0190, 0x0190, nv04_gr_mthd_bind_surf_zeta }, - {} -}; +static bool +nv04_gr_mthd_surf3d(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x02f8: func = nv04_gr_mthd_surf3d_clip_h; break; + case 0x02fc: func = nv04_gr_mthd_surf3d_clip_v; break; + default: + return false; + } + return func(device, inst, data); +} -static struct nvkm_omthds -nv01_gr_prim_omthds[] = { - { 0x0184, 0x0184, nv01_gr_mthd_bind_clip }, - { 0x0188, 0x0188, nv01_gr_mthd_bind_patt }, - { 0x018c, 0x018c, nv04_gr_mthd_bind_rop }, - { 0x0190, 0x0190, nv04_gr_mthd_bind_beta1 }, - { 0x0194, 0x0194, nv04_gr_mthd_bind_surf_dst }, - { 0x02fc, 0x02fc, nv04_gr_mthd_set_operation }, - {} -}; +static bool +nv03_gr_mthd_ttri(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0188: func = nv01_gr_mthd_bind_clip; break; + case 0x018c: func = nv04_gr_mthd_bind_surf_color; break; + case 0x0190: func = nv04_gr_mthd_bind_surf_zeta; break; + default: + return false; + } + return func(device, inst, data); +} -static struct nvkm_omthds -nv04_gr_prim_omthds[] = { - { 0x0184, 0x0184, nv01_gr_mthd_bind_clip }, - { 0x0188, 0x0188, nv04_gr_mthd_bind_patt }, - { 0x018c, 0x018c, nv04_gr_mthd_bind_rop }, - { 0x0190, 0x0190, nv04_gr_mthd_bind_beta1 }, - { 0x0194, 0x0194, nv04_gr_mthd_bind_beta4 }, - { 0x0198, 0x0198, nv04_gr_mthd_bind_surf2d }, - { 0x02fc, 0x02fc, nv04_gr_mthd_set_operation }, - {} -}; +static bool +nv01_gr_mthd_prim(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0184: func = nv01_gr_mthd_bind_clip; break; + case 0x0188: func = nv01_gr_mthd_bind_patt; break; + case 0x018c: func = nv04_gr_mthd_bind_rop; break; + case 0x0190: func = nv04_gr_mthd_bind_beta1; break; + case 0x0194: func = nv04_gr_mthd_bind_surf_dst; break; + case 0x02fc: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} -static int -nv04_gr_object_ctor(struct nvkm_object *parent, struct nvkm_object *engine, - struct nvkm_oclass *oclass, void *data, u32 size, - struct nvkm_object **pobject) +static bool +nv04_gr_mthd_prim(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) { - struct nvkm_gpuobj *obj; - int ret; + bool (*func)(struct nvkm_device *, u32, u32); + switch (mthd) { + case 0x0184: func = nv01_gr_mthd_bind_clip; break; + case 0x0188: func = nv04_gr_mthd_bind_patt; break; + case 0x018c: func = nv04_gr_mthd_bind_rop; break; + case 0x0190: func = nv04_gr_mthd_bind_beta1; break; + case 0x0194: func = nv04_gr_mthd_bind_beta4; break; + case 0x0198: func = nv04_gr_mthd_bind_surf2d; break; + case 0x02fc: func = nv04_gr_mthd_set_operation; break; + default: + return false; + } + return func(device, inst, data); +} - ret = nvkm_gpuobj_create(parent, engine, oclass, 0, parent, - 16, 16, 0, &obj); - *pobject = nv_object(obj); - if (ret) - return ret; +static bool +nv04_gr_mthd(struct nvkm_device *device, u32 inst, u32 mthd, u32 data) +{ + bool (*func)(struct nvkm_device *, u32, u32, u32); + switch (nvkm_rd32(device, 0x700000 + inst) & 0x000000ff) { + case 0x1c ... 0x1e: + func = nv01_gr_mthd_prim; break; + case 0x1f: func = nv01_gr_mthd_blit; break; + case 0x21: func = nv01_gr_mthd_ifc; break; + case 0x36: func = nv03_gr_mthd_sifc; break; + case 0x37: func = nv03_gr_mthd_sifm; break; + case 0x48: func = nv03_gr_mthd_ttri; break; + case 0x4a: func = nv04_gr_mthd_gdi; break; + case 0x4b: func = nv03_gr_mthd_gdi; break; + case 0x53: func = nv04_gr_mthd_surf3d; break; + case 0x5c ... 0x5e: + func = nv04_gr_mthd_prim; break; + case 0x5f: func = nv04_gr_mthd_blit; break; + case 0x60: func = nv04_gr_mthd_iifc; break; + case 0x61: func = nv04_gr_mthd_ifc; break; + case 0x76: func = nv04_gr_mthd_sifc; break; + case 0x77: func = nv04_gr_mthd_sifm; break; + default: + return false; + } + return func(device, inst, mthd, data); +} - nv_wo32(obj, 0x00, nv_mclass(obj)); +static int +nv04_gr_object_bind(struct nvkm_object *object, struct nvkm_gpuobj *parent, + int align, struct nvkm_gpuobj **pgpuobj) +{ + int ret = nvkm_gpuobj_new(object->engine->subdev.device, 16, align, + false, parent, pgpuobj); + if (ret == 0) { + nvkm_kmap(*pgpuobj); + nvkm_wo32(*pgpuobj, 0x00, object->oclass); #ifdef __BIG_ENDIAN - nv_mo32(obj, 0x00, 0x00080000, 0x00080000); + nvkm_mo32(*pgpuobj, 0x00, 0x00080000, 0x00080000); #endif - nv_wo32(obj, 0x04, 0x00000000); - nv_wo32(obj, 0x08, 0x00000000); - nv_wo32(obj, 0x0c, 0x00000000); - return 0; + nvkm_wo32(*pgpuobj, 0x04, 0x00000000); + nvkm_wo32(*pgpuobj, 0x08, 0x00000000); + nvkm_wo32(*pgpuobj, 0x0c, 0x00000000); + nvkm_done(*pgpuobj); + } + return ret; } -struct nvkm_ofuncs -nv04_gr_ofuncs = { - .ctor = nv04_gr_object_ctor, - .dtor = _nvkm_gpuobj_dtor, - .init = _nvkm_gpuobj_init, - .fini = _nvkm_gpuobj_fini, - .rd32 = _nvkm_gpuobj_rd32, - .wr32 = _nvkm_gpuobj_wr32, -}; - -static struct nvkm_oclass -nv04_gr_sclass[] = { - { 0x0012, &nv04_gr_ofuncs }, /* beta1 */ - { 0x0017, &nv04_gr_ofuncs }, /* chroma */ - { 0x0018, &nv04_gr_ofuncs }, /* pattern (nv01) */ - { 0x0019, &nv04_gr_ofuncs }, /* clip */ - { 0x001c, &nv04_gr_ofuncs, nv01_gr_prim_omthds }, /* line */ - { 0x001d, &nv04_gr_ofuncs, nv01_gr_prim_omthds }, /* tri */ - { 0x001e, &nv04_gr_ofuncs, nv01_gr_prim_omthds }, /* rect */ - { 0x001f, &nv04_gr_ofuncs, nv01_gr_blit_omthds }, - { 0x0021, &nv04_gr_ofuncs, nv01_gr_ifc_omthds }, - { 0x0030, &nv04_gr_ofuncs }, /* null */ - { 0x0036, &nv04_gr_ofuncs, nv03_gr_sifc_omthds }, - { 0x0037, &nv04_gr_ofuncs, nv03_gr_sifm_omthds }, - { 0x0038, &nv04_gr_ofuncs }, /* dvd subpicture */ - { 0x0039, &nv04_gr_ofuncs }, /* m2mf */ - { 0x0042, &nv04_gr_ofuncs }, /* surf2d */ - { 0x0043, &nv04_gr_ofuncs }, /* rop */ - { 0x0044, &nv04_gr_ofuncs }, /* pattern */ - { 0x0048, &nv04_gr_ofuncs, nv03_gr_ttri_omthds }, - { 0x004a, &nv04_gr_ofuncs, nv04_gr_gdi_omthds }, - { 0x004b, &nv04_gr_ofuncs, nv03_gr_gdi_omthds }, - { 0x0052, &nv04_gr_ofuncs }, /* swzsurf */ - { 0x0053, &nv04_gr_ofuncs, nv04_gr_surf3d_omthds }, - { 0x0054, &nv04_gr_ofuncs }, /* ttri */ - { 0x0055, &nv04_gr_ofuncs }, /* mtri */ - { 0x0057, &nv04_gr_ofuncs }, /* chroma */ - { 0x0058, &nv04_gr_ofuncs }, /* surf_dst */ - { 0x0059, &nv04_gr_ofuncs }, /* surf_src */ - { 0x005a, &nv04_gr_ofuncs }, /* surf_color */ - { 0x005b, &nv04_gr_ofuncs }, /* surf_zeta */ - { 0x005c, &nv04_gr_ofuncs, nv04_gr_prim_omthds }, /* line */ - { 0x005d, &nv04_gr_ofuncs, nv04_gr_prim_omthds }, /* tri */ - { 0x005e, &nv04_gr_ofuncs, nv04_gr_prim_omthds }, /* rect */ - { 0x005f, &nv04_gr_ofuncs, nv04_gr_blit_omthds }, - { 0x0060, &nv04_gr_ofuncs, nv04_gr_iifc_omthds }, - { 0x0061, &nv04_gr_ofuncs, nv04_gr_ifc_omthds }, - { 0x0064, &nv04_gr_ofuncs }, /* iifc (nv05) */ - { 0x0065, &nv04_gr_ofuncs }, /* ifc (nv05) */ - { 0x0066, &nv04_gr_ofuncs }, /* sifc (nv05) */ - { 0x0072, &nv04_gr_ofuncs }, /* beta4 */ - { 0x0076, &nv04_gr_ofuncs, nv04_gr_sifc_omthds }, - { 0x0077, &nv04_gr_ofuncs, nv04_gr_sifm_omthds }, - {}, +const struct nvkm_object_func +nv04_gr_object = { + .bind = nv04_gr_object_bind, }; /******************************************************************************* @@ -1032,13 +1069,14 @@ nv04_gr_sclass[] = { ******************************************************************************/ static struct nv04_gr_chan * -nv04_gr_channel(struct nv04_gr_priv *priv) +nv04_gr_channel(struct nv04_gr *gr) { + struct nvkm_device *device = gr->base.engine.subdev.device; struct nv04_gr_chan *chan = NULL; - if (nv_rd32(priv, NV04_PGRAPH_CTX_CONTROL) & 0x00010000) { - int chid = nv_rd32(priv, NV04_PGRAPH_CTX_USER) >> 24; - if (chid < ARRAY_SIZE(priv->chan)) - chan = priv->chan[chid]; + if (nvkm_rd32(device, NV04_PGRAPH_CTX_CONTROL) & 0x00010000) { + int chid = nvkm_rd32(device, NV04_PGRAPH_CTX_USER) >> 24; + if (chid < ARRAY_SIZE(gr->chan)) + chan = gr->chan[chid]; } return chan; } @@ -1046,55 +1084,52 @@ nv04_gr_channel(struct nv04_gr_priv *priv) static int nv04_gr_load_context(struct nv04_gr_chan *chan, int chid) { - struct nv04_gr_priv *priv = nv04_gr_priv(chan); + struct nvkm_device *device = chan->gr->base.engine.subdev.device; int i; for (i = 0; i < ARRAY_SIZE(nv04_gr_ctx_regs); i++) - nv_wr32(priv, nv04_gr_ctx_regs[i], chan->nv04[i]); + nvkm_wr32(device, nv04_gr_ctx_regs[i], chan->nv04[i]); - nv_wr32(priv, NV04_PGRAPH_CTX_CONTROL, 0x10010100); - nv_mask(priv, NV04_PGRAPH_CTX_USER, 0xff000000, chid << 24); - nv_mask(priv, NV04_PGRAPH_FFINTFC_ST2, 0xfff00000, 0x00000000); + nvkm_wr32(device, NV04_PGRAPH_CTX_CONTROL, 0x10010100); + nvkm_mask(device, NV04_PGRAPH_CTX_USER, 0xff000000, chid << 24); + nvkm_mask(device, NV04_PGRAPH_FFINTFC_ST2, 0xfff00000, 0x00000000); return 0; } static int nv04_gr_unload_context(struct nv04_gr_chan *chan) { - struct nv04_gr_priv *priv = nv04_gr_priv(chan); + struct nvkm_device *device = chan->gr->base.engine.subdev.device; int i; for (i = 0; i < ARRAY_SIZE(nv04_gr_ctx_regs); i++) - chan->nv04[i] = nv_rd32(priv, nv04_gr_ctx_regs[i]); + chan->nv04[i] = nvkm_rd32(device, nv04_gr_ctx_regs[i]); - nv_wr32(priv, NV04_PGRAPH_CTX_CONTROL, 0x10000000); - nv_mask(priv, NV04_PGRAPH_CTX_USER, 0xff000000, 0x0f000000); + nvkm_wr32(device, NV04_PGRAPH_CTX_CONTROL, 0x10000000); + nvkm_mask(device, NV04_PGRAPH_CTX_USER, 0xff000000, 0x0f000000); return 0; } static void -nv04_gr_context_switch(struct nv04_gr_priv *priv) +nv04_gr_context_switch(struct nv04_gr *gr) { + struct nvkm_device *device = gr->base.engine.subdev.device; struct nv04_gr_chan *prev = NULL; struct nv04_gr_chan *next = NULL; - unsigned long flags; int chid; - spin_lock_irqsave(&priv->lock, flags); - nv04_gr_idle(priv); + nv04_gr_idle(&gr->base); /* If previous context is valid, we need to save it */ - prev = nv04_gr_channel(priv); + prev = nv04_gr_channel(gr); if (prev) nv04_gr_unload_context(prev); /* load context for next channel */ - chid = (nv_rd32(priv, NV04_PGRAPH_TRAPPED_ADDR) >> 24) & 0x0f; - next = priv->chan[chid]; + chid = (nvkm_rd32(device, NV04_PGRAPH_TRAPPED_ADDR) >> 24) & 0x0f; + next = gr->chan[chid]; if (next) nv04_gr_load_context(next, chid); - - spin_unlock_irqrestore(&priv->lock, flags); } static u32 *ctx_reg(struct nv04_gr_chan *chan, u32 reg) @@ -1109,98 +1144,85 @@ static u32 *ctx_reg(struct nv04_gr_chan *chan, u32 reg) return NULL; } -static int -nv04_gr_context_ctor(struct nvkm_object *parent, - struct nvkm_object *engine, - struct nvkm_oclass *oclass, void *data, u32 size, - struct nvkm_object **pobject) +static void * +nv04_gr_chan_dtor(struct nvkm_object *object) { - struct nvkm_fifo_chan *fifo = (void *)parent; - struct nv04_gr_priv *priv = (void *)engine; - struct nv04_gr_chan *chan; + struct nv04_gr_chan *chan = nv04_gr_chan(object); + struct nv04_gr *gr = chan->gr; unsigned long flags; - int ret; - - ret = nvkm_object_create(parent, engine, oclass, 0, &chan); - *pobject = nv_object(chan); - if (ret) - return ret; - - spin_lock_irqsave(&priv->lock, flags); - if (priv->chan[fifo->chid]) { - *pobject = nv_object(priv->chan[fifo->chid]); - atomic_inc(&(*pobject)->refcount); - spin_unlock_irqrestore(&priv->lock, flags); - nvkm_object_destroy(&chan->base); - return 1; - } - *ctx_reg(chan, NV04_PGRAPH_DEBUG_3) = 0xfad4ff31; - - priv->chan[fifo->chid] = chan; - chan->chid = fifo->chid; - spin_unlock_irqrestore(&priv->lock, flags); - return 0; + spin_lock_irqsave(&gr->lock, flags); + gr->chan[chan->chid] = NULL; + spin_unlock_irqrestore(&gr->lock, flags); + return chan; } -static void -nv04_gr_context_dtor(struct nvkm_object *object) +static int +nv04_gr_chan_fini(struct nvkm_object *object, bool suspend) { - struct nv04_gr_priv *priv = (void *)object->engine; - struct nv04_gr_chan *chan = (void *)object; + struct nv04_gr_chan *chan = nv04_gr_chan(object); + struct nv04_gr *gr = chan->gr; + struct nvkm_device *device = gr->base.engine.subdev.device; unsigned long flags; - spin_lock_irqsave(&priv->lock, flags); - priv->chan[chan->chid] = NULL; - spin_unlock_irqrestore(&priv->lock, flags); - - nvkm_object_destroy(&chan->base); + spin_lock_irqsave(&gr->lock, flags); + nvkm_mask(device, NV04_PGRAPH_FIFO, 0x00000001, 0x00000000); + if (nv04_gr_channel(gr) == chan) + nv04_gr_unload_context(chan); + nvkm_mask(device, NV04_PGRAPH_FIFO, 0x00000001, 0x00000001); + spin_unlock_irqrestore(&gr->lock, flags); + return 0; } +static const struct nvkm_object_func +nv04_gr_chan = { + .dtor = nv04_gr_chan_dtor, + .fini = nv04_gr_chan_fini, +}; + static int -nv04_gr_context_fini(struct nvkm_object *object, bool suspend) +nv04_gr_chan_new(struct nvkm_gr *base, struct nvkm_fifo_chan *fifoch, + const struct nvkm_oclass *oclass, struct nvkm_object **pobject) { - struct nv04_gr_priv *priv = (void *)object->engine; - struct nv04_gr_chan *chan = (void *)object; + struct nv04_gr *gr = nv04_gr(base); + struct nv04_gr_chan *chan; unsigned long flags; - spin_lock_irqsave(&priv->lock, flags); - nv_mask(priv, NV04_PGRAPH_FIFO, 0x00000001, 0x00000000); - if (nv04_gr_channel(priv) == chan) - nv04_gr_unload_context(chan); - nv_mask(priv, NV04_PGRAPH_FIFO, 0x00000001, 0x00000001); - spin_unlock_irqrestore(&priv->lock, flags); + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&nv04_gr_chan, oclass, &chan->object); + chan->gr = gr; + chan->chid = fifoch->chid; + *pobject = &chan->object; - return nvkm_object_fini(&chan->base, suspend); -} + *ctx_reg(chan, NV04_PGRAPH_DEBUG_3) = 0xfad4ff31; -static struct nvkm_oclass -nv04_gr_cclass = { - .handle = NV_ENGCTX(GR, 0x04), - .ofuncs = &(struct nvkm_ofuncs) { - .ctor = nv04_gr_context_ctor, - .dtor = nv04_gr_context_dtor, - .init = nvkm_object_init, - .fini = nv04_gr_context_fini, - }, -}; + spin_lock_irqsave(&gr->lock, flags); + gr->chan[chan->chid] = chan; + spin_unlock_irqrestore(&gr->lock, flags); + return 0; +} /******************************************************************************* * PGRAPH engine/subdev functions ******************************************************************************/ bool -nv04_gr_idle(void *obj) +nv04_gr_idle(struct nvkm_gr *gr) { - struct nvkm_gr *gr = nvkm_gr(obj); + struct nvkm_subdev *subdev = &gr->engine.subdev; + struct nvkm_device *device = subdev->device; u32 mask = 0xffffffff; - if (nv_device(obj)->card_type == NV_40) + if (device->card_type == NV_40) mask &= ~NV40_PGRAPH_STATUS_SYNC_STALL; - if (!nv_wait(gr, NV04_PGRAPH_STATUS, mask, 0)) { - nv_error(gr, "idle timed out with status 0x%08x\n", - nv_rd32(gr, NV04_PGRAPH_STATUS)); + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, NV04_PGRAPH_STATUS) & mask)) + break; + ) < 0) { + nvkm_error(subdev, "idle timed out with status %08x\n", + nvkm_rd32(device, NV04_PGRAPH_STATUS)); return false; } @@ -1247,136 +1269,159 @@ nv04_gr_nsource[] = { }; static void -nv04_gr_intr(struct nvkm_subdev *subdev) +nv04_gr_intr(struct nvkm_gr *base) { - struct nv04_gr_priv *priv = (void *)subdev; - struct nv04_gr_chan *chan = NULL; - struct nvkm_namedb *namedb = NULL; - struct nvkm_handle *handle = NULL; - u32 stat = nv_rd32(priv, NV03_PGRAPH_INTR); - u32 nsource = nv_rd32(priv, NV03_PGRAPH_NSOURCE); - u32 nstatus = nv_rd32(priv, NV03_PGRAPH_NSTATUS); - u32 addr = nv_rd32(priv, NV04_PGRAPH_TRAPPED_ADDR); + struct nv04_gr *gr = nv04_gr(base); + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 stat = nvkm_rd32(device, NV03_PGRAPH_INTR); + u32 nsource = nvkm_rd32(device, NV03_PGRAPH_NSOURCE); + u32 nstatus = nvkm_rd32(device, NV03_PGRAPH_NSTATUS); + u32 addr = nvkm_rd32(device, NV04_PGRAPH_TRAPPED_ADDR); u32 chid = (addr & 0x0f000000) >> 24; u32 subc = (addr & 0x0000e000) >> 13; u32 mthd = (addr & 0x00001ffc); - u32 data = nv_rd32(priv, NV04_PGRAPH_TRAPPED_DATA); - u32 class = nv_rd32(priv, 0x400180 + subc * 4) & 0xff; - u32 inst = (nv_rd32(priv, 0x40016c) & 0xffff) << 4; + u32 data = nvkm_rd32(device, NV04_PGRAPH_TRAPPED_DATA); + u32 class = nvkm_rd32(device, 0x400180 + subc * 4) & 0xff; + u32 inst = (nvkm_rd32(device, 0x40016c) & 0xffff) << 4; u32 show = stat; + char msg[128], src[128], sta[128]; + struct nv04_gr_chan *chan; unsigned long flags; - spin_lock_irqsave(&priv->lock, flags); - chan = priv->chan[chid]; - if (chan) - namedb = (void *)nv_pclass(nv_object(chan), NV_NAMEDB_CLASS); - spin_unlock_irqrestore(&priv->lock, flags); + spin_lock_irqsave(&gr->lock, flags); + chan = gr->chan[chid]; if (stat & NV_PGRAPH_INTR_NOTIFY) { if (chan && (nsource & NV03_PGRAPH_NSOURCE_ILLEGAL_MTHD)) { - handle = nvkm_namedb_get_vinst(namedb, inst); - if (handle && !nv_call(handle->object, mthd, data)) + if (!nv04_gr_mthd(device, inst, mthd, data)) show &= ~NV_PGRAPH_INTR_NOTIFY; } } if (stat & NV_PGRAPH_INTR_CONTEXT_SWITCH) { - nv_wr32(priv, NV03_PGRAPH_INTR, NV_PGRAPH_INTR_CONTEXT_SWITCH); + nvkm_wr32(device, NV03_PGRAPH_INTR, NV_PGRAPH_INTR_CONTEXT_SWITCH); stat &= ~NV_PGRAPH_INTR_CONTEXT_SWITCH; show &= ~NV_PGRAPH_INTR_CONTEXT_SWITCH; - nv04_gr_context_switch(priv); + nv04_gr_context_switch(gr); } - nv_wr32(priv, NV03_PGRAPH_INTR, stat); - nv_wr32(priv, NV04_PGRAPH_FIFO, 0x00000001); + nvkm_wr32(device, NV03_PGRAPH_INTR, stat); + nvkm_wr32(device, NV04_PGRAPH_FIFO, 0x00000001); if (show) { - nv_error(priv, "%s", ""); - nvkm_bitfield_print(nv04_gr_intr_name, show); - pr_cont(" nsource:"); - nvkm_bitfield_print(nv04_gr_nsource, nsource); - pr_cont(" nstatus:"); - nvkm_bitfield_print(nv04_gr_nstatus, nstatus); - pr_cont("\n"); - nv_error(priv, - "ch %d [%s] subc %d class 0x%04x mthd 0x%04x data 0x%08x\n", - chid, nvkm_client_name(chan), subc, class, mthd, - data); + nvkm_snprintbf(msg, sizeof(msg), nv04_gr_intr_name, show); + nvkm_snprintbf(src, sizeof(src), nv04_gr_nsource, nsource); + nvkm_snprintbf(sta, sizeof(sta), nv04_gr_nstatus, nstatus); + nvkm_error(subdev, "intr %08x [%s] nsource %08x [%s] " + "nstatus %08x [%s] ch %d [%s] subc %d " + "class %04x mthd %04x data %08x\n", + show, msg, nsource, src, nstatus, sta, chid, + chan ? chan->object.client->name : "unknown", + subc, class, mthd, data); } - nvkm_namedb_put(handle); + spin_unlock_irqrestore(&gr->lock, flags); } static int -nv04_gr_ctor(struct nvkm_object *parent, struct nvkm_object *engine, - struct nvkm_oclass *oclass, void *data, u32 size, - struct nvkm_object **pobject) +nv04_gr_init(struct nvkm_gr *base) { - struct nv04_gr_priv *priv; - int ret; - - ret = nvkm_gr_create(parent, engine, oclass, true, &priv); - *pobject = nv_object(priv); - if (ret) - return ret; - - nv_subdev(priv)->unit = 0x00001000; - nv_subdev(priv)->intr = nv04_gr_intr; - nv_engine(priv)->cclass = &nv04_gr_cclass; - nv_engine(priv)->sclass = nv04_gr_sclass; - spin_lock_init(&priv->lock); - return 0; -} - -static int -nv04_gr_init(struct nvkm_object *object) -{ - struct nvkm_engine *engine = nv_engine(object); - struct nv04_gr_priv *priv = (void *)engine; - int ret; - - ret = nvkm_gr_init(&priv->base); - if (ret) - return ret; + struct nv04_gr *gr = nv04_gr(base); + struct nvkm_device *device = gr->base.engine.subdev.device; /* Enable PGRAPH interrupts */ - nv_wr32(priv, NV03_PGRAPH_INTR, 0xFFFFFFFF); - nv_wr32(priv, NV03_PGRAPH_INTR_EN, 0xFFFFFFFF); - - nv_wr32(priv, NV04_PGRAPH_VALID1, 0); - nv_wr32(priv, NV04_PGRAPH_VALID2, 0); - /*nv_wr32(priv, NV04_PGRAPH_DEBUG_0, 0x000001FF); - nv_wr32(priv, NV04_PGRAPH_DEBUG_0, 0x001FFFFF);*/ - nv_wr32(priv, NV04_PGRAPH_DEBUG_0, 0x1231c000); + nvkm_wr32(device, NV03_PGRAPH_INTR, 0xFFFFFFFF); + nvkm_wr32(device, NV03_PGRAPH_INTR_EN, 0xFFFFFFFF); + + nvkm_wr32(device, NV04_PGRAPH_VALID1, 0); + nvkm_wr32(device, NV04_PGRAPH_VALID2, 0); + /*nvkm_wr32(device, NV04_PGRAPH_DEBUG_0, 0x000001FF); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_0, 0x001FFFFF);*/ + nvkm_wr32(device, NV04_PGRAPH_DEBUG_0, 0x1231c000); /*1231C000 blob, 001 haiku*/ /*V_WRITE(NV04_PGRAPH_DEBUG_1, 0xf2d91100);*/ - nv_wr32(priv, NV04_PGRAPH_DEBUG_1, 0x72111100); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_1, 0x72111100); /*0x72111100 blob , 01 haiku*/ - /*nv_wr32(priv, NV04_PGRAPH_DEBUG_2, 0x11d5f870);*/ - nv_wr32(priv, NV04_PGRAPH_DEBUG_2, 0x11d5f071); + /*nvkm_wr32(device, NV04_PGRAPH_DEBUG_2, 0x11d5f870);*/ + nvkm_wr32(device, NV04_PGRAPH_DEBUG_2, 0x11d5f071); /*haiku same*/ - /*nv_wr32(priv, NV04_PGRAPH_DEBUG_3, 0xfad4ff31);*/ - nv_wr32(priv, NV04_PGRAPH_DEBUG_3, 0xf0d4ff31); + /*nvkm_wr32(device, NV04_PGRAPH_DEBUG_3, 0xfad4ff31);*/ + nvkm_wr32(device, NV04_PGRAPH_DEBUG_3, 0xf0d4ff31); /*haiku and blob 10d4*/ - nv_wr32(priv, NV04_PGRAPH_STATE , 0xFFFFFFFF); - nv_wr32(priv, NV04_PGRAPH_CTX_CONTROL , 0x10000100); - nv_mask(priv, NV04_PGRAPH_CTX_USER, 0xff000000, 0x0f000000); + nvkm_wr32(device, NV04_PGRAPH_STATE , 0xFFFFFFFF); + nvkm_wr32(device, NV04_PGRAPH_CTX_CONTROL , 0x10000100); + nvkm_mask(device, NV04_PGRAPH_CTX_USER, 0xff000000, 0x0f000000); /* These don't belong here, they're part of a per-channel context */ - nv_wr32(priv, NV04_PGRAPH_PATTERN_SHAPE, 0x00000000); - nv_wr32(priv, NV04_PGRAPH_BETA_AND , 0xFFFFFFFF); + nvkm_wr32(device, NV04_PGRAPH_PATTERN_SHAPE, 0x00000000); + nvkm_wr32(device, NV04_PGRAPH_BETA_AND , 0xFFFFFFFF); return 0; } -struct nvkm_oclass -nv04_gr_oclass = { - .handle = NV_ENGINE(GR, 0x04), - .ofuncs = &(struct nvkm_ofuncs) { - .ctor = nv04_gr_ctor, - .dtor = _nvkm_gr_dtor, - .init = nv04_gr_init, - .fini = _nvkm_gr_fini, - }, +static const struct nvkm_gr_func +nv04_gr = { + .init = nv04_gr_init, + .intr = nv04_gr_intr, + .chan_new = nv04_gr_chan_new, + .sclass = { + { -1, -1, 0x0012, &nv04_gr_object }, /* beta1 */ + { -1, -1, 0x0017, &nv04_gr_object }, /* chroma */ + { -1, -1, 0x0018, &nv04_gr_object }, /* pattern (nv01) */ + { -1, -1, 0x0019, &nv04_gr_object }, /* clip */ + { -1, -1, 0x001c, &nv04_gr_object }, /* line */ + { -1, -1, 0x001d, &nv04_gr_object }, /* tri */ + { -1, -1, 0x001e, &nv04_gr_object }, /* rect */ + { -1, -1, 0x001f, &nv04_gr_object }, + { -1, -1, 0x0021, &nv04_gr_object }, + { -1, -1, 0x0030, &nv04_gr_object }, /* null */ + { -1, -1, 0x0036, &nv04_gr_object }, + { -1, -1, 0x0037, &nv04_gr_object }, + { -1, -1, 0x0038, &nv04_gr_object }, /* dvd subpicture */ + { -1, -1, 0x0039, &nv04_gr_object }, /* m2mf */ + { -1, -1, 0x0042, &nv04_gr_object }, /* surf2d */ + { -1, -1, 0x0043, &nv04_gr_object }, /* rop */ + { -1, -1, 0x0044, &nv04_gr_object }, /* pattern */ + { -1, -1, 0x0048, &nv04_gr_object }, + { -1, -1, 0x004a, &nv04_gr_object }, + { -1, -1, 0x004b, &nv04_gr_object }, + { -1, -1, 0x0052, &nv04_gr_object }, /* swzsurf */ + { -1, -1, 0x0053, &nv04_gr_object }, + { -1, -1, 0x0054, &nv04_gr_object }, /* ttri */ + { -1, -1, 0x0055, &nv04_gr_object }, /* mtri */ + { -1, -1, 0x0057, &nv04_gr_object }, /* chroma */ + { -1, -1, 0x0058, &nv04_gr_object }, /* surf_dst */ + { -1, -1, 0x0059, &nv04_gr_object }, /* surf_src */ + { -1, -1, 0x005a, &nv04_gr_object }, /* surf_color */ + { -1, -1, 0x005b, &nv04_gr_object }, /* surf_zeta */ + { -1, -1, 0x005c, &nv04_gr_object }, /* line */ + { -1, -1, 0x005d, &nv04_gr_object }, /* tri */ + { -1, -1, 0x005e, &nv04_gr_object }, /* rect */ + { -1, -1, 0x005f, &nv04_gr_object }, + { -1, -1, 0x0060, &nv04_gr_object }, + { -1, -1, 0x0061, &nv04_gr_object }, + { -1, -1, 0x0064, &nv04_gr_object }, /* iifc (nv05) */ + { -1, -1, 0x0065, &nv04_gr_object }, /* ifc (nv05) */ + { -1, -1, 0x0066, &nv04_gr_object }, /* sifc (nv05) */ + { -1, -1, 0x0072, &nv04_gr_object }, /* beta4 */ + { -1, -1, 0x0076, &nv04_gr_object }, + { -1, -1, 0x0077, &nv04_gr_object }, + {} + } }; + +int +nv04_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + struct nv04_gr *gr; + + if (!(gr = kzalloc(sizeof(*gr), GFP_KERNEL))) + return -ENOMEM; + spin_lock_init(&gr->lock); + *pgr = &gr->base; + + return nvkm_gr_ctor(&nv04_gr, device, index, 0x00001000, + true, &gr->base); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv10.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv10.c index 389904eb6..4542867fa 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv10.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv10.c @@ -21,13 +21,13 @@ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER * DEALINGS IN THE SOFTWARE. */ -#include <engine/gr.h> +#include "nv10.h" #include "regs.h" #include <core/client.h> -#include <core/device.h> -#include <core/handle.h> +#include <core/gpuobj.h> #include <engine/fifo.h> +#include <engine/fifo/chan.h> #include <subdev/fb.h> struct pipe_state { @@ -386,14 +386,19 @@ static int nv17_gr_ctx_regs[] = { 0x00400a04, }; -struct nv10_gr_priv { +#define nv10_gr(p) container_of((p), struct nv10_gr, base) + +struct nv10_gr { struct nvkm_gr base; struct nv10_gr_chan *chan[32]; spinlock_t lock; }; +#define nv10_gr_chan(p) container_of((p), struct nv10_gr_chan, object) + struct nv10_gr_chan { - struct nvkm_object base; + struct nvkm_object object; + struct nv10_gr *gr; int chid; int nv10[ARRAY_SIZE(nv10_gr_ctx_regs)]; int nv17[ARRAY_SIZE(nv17_gr_ctx_regs)]; @@ -402,214 +407,151 @@ struct nv10_gr_chan { }; -static inline struct nv10_gr_priv * -nv10_gr_priv(struct nv10_gr_chan *chan) -{ - return (void *)nv_object(chan)->engine; -} - /******************************************************************************* * Graphics object classes ******************************************************************************/ -#define PIPE_SAVE(priv, state, addr) \ +#define PIPE_SAVE(gr, state, addr) \ do { \ int __i; \ - nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, addr); \ + nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, addr); \ for (__i = 0; __i < ARRAY_SIZE(state); __i++) \ - state[__i] = nv_rd32(priv, NV10_PGRAPH_PIPE_DATA); \ + state[__i] = nvkm_rd32(device, NV10_PGRAPH_PIPE_DATA); \ } while (0) -#define PIPE_RESTORE(priv, state, addr) \ +#define PIPE_RESTORE(gr, state, addr) \ do { \ int __i; \ - nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, addr); \ + nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, addr); \ for (__i = 0; __i < ARRAY_SIZE(state); __i++) \ - nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, state[__i]); \ + nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, state[__i]); \ } while (0) -static struct nvkm_oclass -nv10_gr_sclass[] = { - { 0x0012, &nv04_gr_ofuncs }, /* beta1 */ - { 0x0019, &nv04_gr_ofuncs }, /* clip */ - { 0x0030, &nv04_gr_ofuncs }, /* null */ - { 0x0039, &nv04_gr_ofuncs }, /* m2mf */ - { 0x0043, &nv04_gr_ofuncs }, /* rop */ - { 0x0044, &nv04_gr_ofuncs }, /* pattern */ - { 0x004a, &nv04_gr_ofuncs }, /* gdi */ - { 0x0052, &nv04_gr_ofuncs }, /* swzsurf */ - { 0x005f, &nv04_gr_ofuncs }, /* blit */ - { 0x0062, &nv04_gr_ofuncs }, /* surf2d */ - { 0x0072, &nv04_gr_ofuncs }, /* beta4 */ - { 0x0089, &nv04_gr_ofuncs }, /* sifm */ - { 0x008a, &nv04_gr_ofuncs }, /* ifc */ - { 0x009f, &nv04_gr_ofuncs }, /* blit */ - { 0x0093, &nv04_gr_ofuncs }, /* surf3d */ - { 0x0094, &nv04_gr_ofuncs }, /* ttri */ - { 0x0095, &nv04_gr_ofuncs }, /* mtri */ - { 0x0056, &nv04_gr_ofuncs }, /* celcius */ - {}, -}; - -static struct nvkm_oclass -nv15_gr_sclass[] = { - { 0x0012, &nv04_gr_ofuncs }, /* beta1 */ - { 0x0019, &nv04_gr_ofuncs }, /* clip */ - { 0x0030, &nv04_gr_ofuncs }, /* null */ - { 0x0039, &nv04_gr_ofuncs }, /* m2mf */ - { 0x0043, &nv04_gr_ofuncs }, /* rop */ - { 0x0044, &nv04_gr_ofuncs }, /* pattern */ - { 0x004a, &nv04_gr_ofuncs }, /* gdi */ - { 0x0052, &nv04_gr_ofuncs }, /* swzsurf */ - { 0x005f, &nv04_gr_ofuncs }, /* blit */ - { 0x0062, &nv04_gr_ofuncs }, /* surf2d */ - { 0x0072, &nv04_gr_ofuncs }, /* beta4 */ - { 0x0089, &nv04_gr_ofuncs }, /* sifm */ - { 0x008a, &nv04_gr_ofuncs }, /* ifc */ - { 0x009f, &nv04_gr_ofuncs }, /* blit */ - { 0x0093, &nv04_gr_ofuncs }, /* surf3d */ - { 0x0094, &nv04_gr_ofuncs }, /* ttri */ - { 0x0095, &nv04_gr_ofuncs }, /* mtri */ - { 0x0096, &nv04_gr_ofuncs }, /* celcius */ - {}, -}; - -static int -nv17_gr_mthd_lma_window(struct nvkm_object *object, u32 mthd, - void *args, u32 size) +static void +nv17_gr_mthd_lma_window(struct nv10_gr_chan *chan, u32 mthd, u32 data) { - struct nv10_gr_chan *chan = (void *)object->parent; - struct nv10_gr_priv *priv = nv10_gr_priv(chan); + struct nvkm_device *device = chan->object.engine->subdev.device; + struct nvkm_gr *gr = &chan->gr->base; struct pipe_state *pipe = &chan->pipe_state; u32 pipe_0x0040[1], pipe_0x64c0[8], pipe_0x6a80[3], pipe_0x6ab0[3]; u32 xfmode0, xfmode1; - u32 data = *(u32 *)args; int i; chan->lma_window[(mthd - 0x1638) / 4] = data; if (mthd != 0x1644) - return 0; + return; - nv04_gr_idle(priv); + nv04_gr_idle(gr); - PIPE_SAVE(priv, pipe_0x0040, 0x0040); - PIPE_SAVE(priv, pipe->pipe_0x0200, 0x0200); + PIPE_SAVE(device, pipe_0x0040, 0x0040); + PIPE_SAVE(device, pipe->pipe_0x0200, 0x0200); - PIPE_RESTORE(priv, chan->lma_window, 0x6790); + PIPE_RESTORE(device, chan->lma_window, 0x6790); - nv04_gr_idle(priv); + nv04_gr_idle(gr); - xfmode0 = nv_rd32(priv, NV10_PGRAPH_XFMODE0); - xfmode1 = nv_rd32(priv, NV10_PGRAPH_XFMODE1); + xfmode0 = nvkm_rd32(device, NV10_PGRAPH_XFMODE0); + xfmode1 = nvkm_rd32(device, NV10_PGRAPH_XFMODE1); - PIPE_SAVE(priv, pipe->pipe_0x4400, 0x4400); - PIPE_SAVE(priv, pipe_0x64c0, 0x64c0); - PIPE_SAVE(priv, pipe_0x6ab0, 0x6ab0); - PIPE_SAVE(priv, pipe_0x6a80, 0x6a80); + PIPE_SAVE(device, pipe->pipe_0x4400, 0x4400); + PIPE_SAVE(device, pipe_0x64c0, 0x64c0); + PIPE_SAVE(device, pipe_0x6ab0, 0x6ab0); + PIPE_SAVE(device, pipe_0x6a80, 0x6a80); - nv04_gr_idle(priv); + nv04_gr_idle(gr); - nv_wr32(priv, NV10_PGRAPH_XFMODE0, 0x10000000); - nv_wr32(priv, NV10_PGRAPH_XFMODE1, 0x00000000); - nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x000064c0); + nvkm_wr32(device, NV10_PGRAPH_XFMODE0, 0x10000000); + nvkm_wr32(device, NV10_PGRAPH_XFMODE1, 0x00000000); + nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x000064c0); for (i = 0; i < 4; i++) - nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x3f800000); + nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x3f800000); for (i = 0; i < 4; i++) - nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x00000000); + nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x00000000); - nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x00006ab0); + nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x00006ab0); for (i = 0; i < 3; i++) - nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x3f800000); + nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x3f800000); - nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x00006a80); + nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x00006a80); for (i = 0; i < 3; i++) - nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x00000000); - - nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x00000040); - nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x00000008); + nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x00000000); - PIPE_RESTORE(priv, pipe->pipe_0x0200, 0x0200); + nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x00000040); + nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x00000008); - nv04_gr_idle(priv); + PIPE_RESTORE(device, pipe->pipe_0x0200, 0x0200); - PIPE_RESTORE(priv, pipe_0x0040, 0x0040); + nv04_gr_idle(gr); - nv_wr32(priv, NV10_PGRAPH_XFMODE0, xfmode0); - nv_wr32(priv, NV10_PGRAPH_XFMODE1, xfmode1); + PIPE_RESTORE(device, pipe_0x0040, 0x0040); - PIPE_RESTORE(priv, pipe_0x64c0, 0x64c0); - PIPE_RESTORE(priv, pipe_0x6ab0, 0x6ab0); - PIPE_RESTORE(priv, pipe_0x6a80, 0x6a80); - PIPE_RESTORE(priv, pipe->pipe_0x4400, 0x4400); + nvkm_wr32(device, NV10_PGRAPH_XFMODE0, xfmode0); + nvkm_wr32(device, NV10_PGRAPH_XFMODE1, xfmode1); - nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x000000c0); - nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x00000000); + PIPE_RESTORE(device, pipe_0x64c0, 0x64c0); + PIPE_RESTORE(device, pipe_0x6ab0, 0x6ab0); + PIPE_RESTORE(device, pipe_0x6a80, 0x6a80); + PIPE_RESTORE(device, pipe->pipe_0x4400, 0x4400); - nv04_gr_idle(priv); + nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x000000c0); + nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x00000000); - return 0; + nv04_gr_idle(gr); } -static int -nv17_gr_mthd_lma_enable(struct nvkm_object *object, u32 mthd, - void *args, u32 size) +static void +nv17_gr_mthd_lma_enable(struct nv10_gr_chan *chan, u32 mthd, u32 data) { - struct nv10_gr_chan *chan = (void *)object->parent; - struct nv10_gr_priv *priv = nv10_gr_priv(chan); + struct nvkm_device *device = chan->object.engine->subdev.device; + struct nvkm_gr *gr = &chan->gr->base; - nv04_gr_idle(priv); + nv04_gr_idle(gr); - nv_mask(priv, NV10_PGRAPH_DEBUG_4, 0x00000100, 0x00000100); - nv_mask(priv, 0x4006b0, 0x08000000, 0x08000000); - return 0; + nvkm_mask(device, NV10_PGRAPH_DEBUG_4, 0x00000100, 0x00000100); + nvkm_mask(device, 0x4006b0, 0x08000000, 0x08000000); } -static struct nvkm_omthds -nv17_celcius_omthds[] = { - { 0x1638, 0x1638, nv17_gr_mthd_lma_window }, - { 0x163c, 0x163c, nv17_gr_mthd_lma_window }, - { 0x1640, 0x1640, nv17_gr_mthd_lma_window }, - { 0x1644, 0x1644, nv17_gr_mthd_lma_window }, - { 0x1658, 0x1658, nv17_gr_mthd_lma_enable }, - {} -}; +static bool +nv17_gr_mthd_celcius(struct nv10_gr_chan *chan, u32 mthd, u32 data) +{ + void (*func)(struct nv10_gr_chan *, u32, u32); + switch (mthd) { + case 0x1638 ... 0x1644: + func = nv17_gr_mthd_lma_window; break; + case 0x1658: func = nv17_gr_mthd_lma_enable; break; + default: + return false; + } + func(chan, mthd, data); + return true; +} -static struct nvkm_oclass -nv17_gr_sclass[] = { - { 0x0012, &nv04_gr_ofuncs }, /* beta1 */ - { 0x0019, &nv04_gr_ofuncs }, /* clip */ - { 0x0030, &nv04_gr_ofuncs }, /* null */ - { 0x0039, &nv04_gr_ofuncs }, /* m2mf */ - { 0x0043, &nv04_gr_ofuncs }, /* rop */ - { 0x0044, &nv04_gr_ofuncs }, /* pattern */ - { 0x004a, &nv04_gr_ofuncs }, /* gdi */ - { 0x0052, &nv04_gr_ofuncs }, /* swzsurf */ - { 0x005f, &nv04_gr_ofuncs }, /* blit */ - { 0x0062, &nv04_gr_ofuncs }, /* surf2d */ - { 0x0072, &nv04_gr_ofuncs }, /* beta4 */ - { 0x0089, &nv04_gr_ofuncs }, /* sifm */ - { 0x008a, &nv04_gr_ofuncs }, /* ifc */ - { 0x009f, &nv04_gr_ofuncs }, /* blit */ - { 0x0093, &nv04_gr_ofuncs }, /* surf3d */ - { 0x0094, &nv04_gr_ofuncs }, /* ttri */ - { 0x0095, &nv04_gr_ofuncs }, /* mtri */ - { 0x0099, &nv04_gr_ofuncs, nv17_celcius_omthds }, - {}, -}; +static bool +nv10_gr_mthd(struct nv10_gr_chan *chan, u8 class, u32 mthd, u32 data) +{ + bool (*func)(struct nv10_gr_chan *, u32, u32); + switch (class) { + case 0x99: func = nv17_gr_mthd_celcius; break; + default: + return false; + } + return func(chan, mthd, data); +} /******************************************************************************* * PGRAPH context ******************************************************************************/ static struct nv10_gr_chan * -nv10_gr_channel(struct nv10_gr_priv *priv) +nv10_gr_channel(struct nv10_gr *gr) { + struct nvkm_device *device = gr->base.engine.subdev.device; struct nv10_gr_chan *chan = NULL; - if (nv_rd32(priv, 0x400144) & 0x00010000) { - int chid = nv_rd32(priv, 0x400148) >> 24; - if (chid < ARRAY_SIZE(priv->chan)) - chan = priv->chan[chid]; + if (nvkm_rd32(device, 0x400144) & 0x00010000) { + int chid = nvkm_rd32(device, 0x400148) >> 24; + if (chid < ARRAY_SIZE(gr->chan)) + chan = gr->chan[chid]; } return chan; } @@ -617,75 +559,78 @@ nv10_gr_channel(struct nv10_gr_priv *priv) static void nv10_gr_save_pipe(struct nv10_gr_chan *chan) { - struct nv10_gr_priv *priv = nv10_gr_priv(chan); + struct nv10_gr *gr = chan->gr; struct pipe_state *pipe = &chan->pipe_state; - - PIPE_SAVE(priv, pipe->pipe_0x4400, 0x4400); - PIPE_SAVE(priv, pipe->pipe_0x0200, 0x0200); - PIPE_SAVE(priv, pipe->pipe_0x6400, 0x6400); - PIPE_SAVE(priv, pipe->pipe_0x6800, 0x6800); - PIPE_SAVE(priv, pipe->pipe_0x6c00, 0x6c00); - PIPE_SAVE(priv, pipe->pipe_0x7000, 0x7000); - PIPE_SAVE(priv, pipe->pipe_0x7400, 0x7400); - PIPE_SAVE(priv, pipe->pipe_0x7800, 0x7800); - PIPE_SAVE(priv, pipe->pipe_0x0040, 0x0040); - PIPE_SAVE(priv, pipe->pipe_0x0000, 0x0000); + struct nvkm_device *device = gr->base.engine.subdev.device; + + PIPE_SAVE(gr, pipe->pipe_0x4400, 0x4400); + PIPE_SAVE(gr, pipe->pipe_0x0200, 0x0200); + PIPE_SAVE(gr, pipe->pipe_0x6400, 0x6400); + PIPE_SAVE(gr, pipe->pipe_0x6800, 0x6800); + PIPE_SAVE(gr, pipe->pipe_0x6c00, 0x6c00); + PIPE_SAVE(gr, pipe->pipe_0x7000, 0x7000); + PIPE_SAVE(gr, pipe->pipe_0x7400, 0x7400); + PIPE_SAVE(gr, pipe->pipe_0x7800, 0x7800); + PIPE_SAVE(gr, pipe->pipe_0x0040, 0x0040); + PIPE_SAVE(gr, pipe->pipe_0x0000, 0x0000); } static void nv10_gr_load_pipe(struct nv10_gr_chan *chan) { - struct nv10_gr_priv *priv = nv10_gr_priv(chan); + struct nv10_gr *gr = chan->gr; struct pipe_state *pipe = &chan->pipe_state; + struct nvkm_device *device = gr->base.engine.subdev.device; u32 xfmode0, xfmode1; int i; - nv04_gr_idle(priv); + nv04_gr_idle(&gr->base); /* XXX check haiku comments */ - xfmode0 = nv_rd32(priv, NV10_PGRAPH_XFMODE0); - xfmode1 = nv_rd32(priv, NV10_PGRAPH_XFMODE1); - nv_wr32(priv, NV10_PGRAPH_XFMODE0, 0x10000000); - nv_wr32(priv, NV10_PGRAPH_XFMODE1, 0x00000000); - nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x000064c0); + xfmode0 = nvkm_rd32(device, NV10_PGRAPH_XFMODE0); + xfmode1 = nvkm_rd32(device, NV10_PGRAPH_XFMODE1); + nvkm_wr32(device, NV10_PGRAPH_XFMODE0, 0x10000000); + nvkm_wr32(device, NV10_PGRAPH_XFMODE1, 0x00000000); + nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x000064c0); for (i = 0; i < 4; i++) - nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x3f800000); + nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x3f800000); for (i = 0; i < 4; i++) - nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x00000000); + nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x00000000); - nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x00006ab0); + nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x00006ab0); for (i = 0; i < 3; i++) - nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x3f800000); + nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x3f800000); - nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x00006a80); + nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x00006a80); for (i = 0; i < 3; i++) - nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x00000000); + nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x00000000); - nv_wr32(priv, NV10_PGRAPH_PIPE_ADDRESS, 0x00000040); - nv_wr32(priv, NV10_PGRAPH_PIPE_DATA, 0x00000008); + nvkm_wr32(device, NV10_PGRAPH_PIPE_ADDRESS, 0x00000040); + nvkm_wr32(device, NV10_PGRAPH_PIPE_DATA, 0x00000008); - PIPE_RESTORE(priv, pipe->pipe_0x0200, 0x0200); - nv04_gr_idle(priv); + PIPE_RESTORE(gr, pipe->pipe_0x0200, 0x0200); + nv04_gr_idle(&gr->base); /* restore XFMODE */ - nv_wr32(priv, NV10_PGRAPH_XFMODE0, xfmode0); - nv_wr32(priv, NV10_PGRAPH_XFMODE1, xfmode1); - PIPE_RESTORE(priv, pipe->pipe_0x6400, 0x6400); - PIPE_RESTORE(priv, pipe->pipe_0x6800, 0x6800); - PIPE_RESTORE(priv, pipe->pipe_0x6c00, 0x6c00); - PIPE_RESTORE(priv, pipe->pipe_0x7000, 0x7000); - PIPE_RESTORE(priv, pipe->pipe_0x7400, 0x7400); - PIPE_RESTORE(priv, pipe->pipe_0x7800, 0x7800); - PIPE_RESTORE(priv, pipe->pipe_0x4400, 0x4400); - PIPE_RESTORE(priv, pipe->pipe_0x0000, 0x0000); - PIPE_RESTORE(priv, pipe->pipe_0x0040, 0x0040); - nv04_gr_idle(priv); + nvkm_wr32(device, NV10_PGRAPH_XFMODE0, xfmode0); + nvkm_wr32(device, NV10_PGRAPH_XFMODE1, xfmode1); + PIPE_RESTORE(gr, pipe->pipe_0x6400, 0x6400); + PIPE_RESTORE(gr, pipe->pipe_0x6800, 0x6800); + PIPE_RESTORE(gr, pipe->pipe_0x6c00, 0x6c00); + PIPE_RESTORE(gr, pipe->pipe_0x7000, 0x7000); + PIPE_RESTORE(gr, pipe->pipe_0x7400, 0x7400); + PIPE_RESTORE(gr, pipe->pipe_0x7800, 0x7800); + PIPE_RESTORE(gr, pipe->pipe_0x4400, 0x4400); + PIPE_RESTORE(gr, pipe->pipe_0x0000, 0x0000); + PIPE_RESTORE(gr, pipe->pipe_0x0040, 0x0040); + nv04_gr_idle(&gr->base); } static void nv10_gr_create_pipe(struct nv10_gr_chan *chan) { - struct nv10_gr_priv *priv = nv10_gr_priv(chan); + struct nv10_gr *gr = chan->gr; + struct nvkm_subdev *subdev = &gr->base.engine.subdev; struct pipe_state *pipe_state = &chan->pipe_state; u32 *pipe_state_addr; int i; @@ -698,7 +643,7 @@ nv10_gr_create_pipe(struct nv10_gr_chan *chan) u32 *__end_addr = pipe_state->pipe_##addr + \ ARRAY_SIZE(pipe_state->pipe_##addr); \ if (pipe_state_addr != __end_addr) \ - nv_error(priv, "incomplete pipe init for 0x%x : %p/%p\n", \ + nvkm_error(subdev, "incomplete pipe init for 0x%x : %p/%p\n", \ addr, pipe_state_addr, __end_addr); \ } while (0) #define NV_WRITE_PIPE_INIT(value) *(pipe_state_addr++) = value @@ -838,33 +783,36 @@ nv10_gr_create_pipe(struct nv10_gr_chan *chan) } static int -nv10_gr_ctx_regs_find_offset(struct nv10_gr_priv *priv, int reg) +nv10_gr_ctx_regs_find_offset(struct nv10_gr *gr, int reg) { + struct nvkm_subdev *subdev = &gr->base.engine.subdev; int i; for (i = 0; i < ARRAY_SIZE(nv10_gr_ctx_regs); i++) { if (nv10_gr_ctx_regs[i] == reg) return i; } - nv_error(priv, "unknow offset nv10_ctx_regs %d\n", reg); + nvkm_error(subdev, "unknown offset nv10_ctx_regs %d\n", reg); return -1; } static int -nv17_gr_ctx_regs_find_offset(struct nv10_gr_priv *priv, int reg) +nv17_gr_ctx_regs_find_offset(struct nv10_gr *gr, int reg) { + struct nvkm_subdev *subdev = &gr->base.engine.subdev; int i; for (i = 0; i < ARRAY_SIZE(nv17_gr_ctx_regs); i++) { if (nv17_gr_ctx_regs[i] == reg) return i; } - nv_error(priv, "unknow offset nv17_ctx_regs %d\n", reg); + nvkm_error(subdev, "unknown offset nv17_ctx_regs %d\n", reg); return -1; } static void nv10_gr_load_dma_vtxbuf(struct nv10_gr_chan *chan, int chid, u32 inst) { - struct nv10_gr_priv *priv = nv10_gr_priv(chan); + struct nv10_gr *gr = chan->gr; + struct nvkm_device *device = gr->base.engine.subdev.device; u32 st2, st2_dl, st2_dh, fifo_ptr, fifo[0x60/4]; u32 ctx_user, ctx_switch[5]; int i, subchan = -1; @@ -876,7 +824,7 @@ nv10_gr_load_dma_vtxbuf(struct nv10_gr_chan *chan, int chid, u32 inst) /* Look for a celsius object */ for (i = 0; i < 8; i++) { - int class = nv_rd32(priv, NV10_PGRAPH_CTX_CACHE(i, 0)) & 0xfff; + int class = nvkm_rd32(device, NV10_PGRAPH_CTX_CACHE(i, 0)) & 0xfff; if (class == 0x56 || class == 0x96 || class == 0x99) { subchan = i; @@ -888,159 +836,183 @@ nv10_gr_load_dma_vtxbuf(struct nv10_gr_chan *chan, int chid, u32 inst) return; /* Save the current ctx object */ - ctx_user = nv_rd32(priv, NV10_PGRAPH_CTX_USER); + ctx_user = nvkm_rd32(device, NV10_PGRAPH_CTX_USER); for (i = 0; i < 5; i++) - ctx_switch[i] = nv_rd32(priv, NV10_PGRAPH_CTX_SWITCH(i)); + ctx_switch[i] = nvkm_rd32(device, NV10_PGRAPH_CTX_SWITCH(i)); /* Save the FIFO state */ - st2 = nv_rd32(priv, NV10_PGRAPH_FFINTFC_ST2); - st2_dl = nv_rd32(priv, NV10_PGRAPH_FFINTFC_ST2_DL); - st2_dh = nv_rd32(priv, NV10_PGRAPH_FFINTFC_ST2_DH); - fifo_ptr = nv_rd32(priv, NV10_PGRAPH_FFINTFC_FIFO_PTR); + st2 = nvkm_rd32(device, NV10_PGRAPH_FFINTFC_ST2); + st2_dl = nvkm_rd32(device, NV10_PGRAPH_FFINTFC_ST2_DL); + st2_dh = nvkm_rd32(device, NV10_PGRAPH_FFINTFC_ST2_DH); + fifo_ptr = nvkm_rd32(device, NV10_PGRAPH_FFINTFC_FIFO_PTR); for (i = 0; i < ARRAY_SIZE(fifo); i++) - fifo[i] = nv_rd32(priv, 0x4007a0 + 4 * i); + fifo[i] = nvkm_rd32(device, 0x4007a0 + 4 * i); /* Switch to the celsius subchannel */ for (i = 0; i < 5; i++) - nv_wr32(priv, NV10_PGRAPH_CTX_SWITCH(i), - nv_rd32(priv, NV10_PGRAPH_CTX_CACHE(subchan, i))); - nv_mask(priv, NV10_PGRAPH_CTX_USER, 0xe000, subchan << 13); + nvkm_wr32(device, NV10_PGRAPH_CTX_SWITCH(i), + nvkm_rd32(device, NV10_PGRAPH_CTX_CACHE(subchan, i))); + nvkm_mask(device, NV10_PGRAPH_CTX_USER, 0xe000, subchan << 13); /* Inject NV10TCL_DMA_VTXBUF */ - nv_wr32(priv, NV10_PGRAPH_FFINTFC_FIFO_PTR, 0); - nv_wr32(priv, NV10_PGRAPH_FFINTFC_ST2, + nvkm_wr32(device, NV10_PGRAPH_FFINTFC_FIFO_PTR, 0); + nvkm_wr32(device, NV10_PGRAPH_FFINTFC_ST2, 0x2c000000 | chid << 20 | subchan << 16 | 0x18c); - nv_wr32(priv, NV10_PGRAPH_FFINTFC_ST2_DL, inst); - nv_mask(priv, NV10_PGRAPH_CTX_CONTROL, 0, 0x10000); - nv_mask(priv, NV04_PGRAPH_FIFO, 0x00000001, 0x00000001); - nv_mask(priv, NV04_PGRAPH_FIFO, 0x00000001, 0x00000000); + nvkm_wr32(device, NV10_PGRAPH_FFINTFC_ST2_DL, inst); + nvkm_mask(device, NV10_PGRAPH_CTX_CONTROL, 0, 0x10000); + nvkm_mask(device, NV04_PGRAPH_FIFO, 0x00000001, 0x00000001); + nvkm_mask(device, NV04_PGRAPH_FIFO, 0x00000001, 0x00000000); /* Restore the FIFO state */ for (i = 0; i < ARRAY_SIZE(fifo); i++) - nv_wr32(priv, 0x4007a0 + 4 * i, fifo[i]); + nvkm_wr32(device, 0x4007a0 + 4 * i, fifo[i]); - nv_wr32(priv, NV10_PGRAPH_FFINTFC_FIFO_PTR, fifo_ptr); - nv_wr32(priv, NV10_PGRAPH_FFINTFC_ST2, st2); - nv_wr32(priv, NV10_PGRAPH_FFINTFC_ST2_DL, st2_dl); - nv_wr32(priv, NV10_PGRAPH_FFINTFC_ST2_DH, st2_dh); + nvkm_wr32(device, NV10_PGRAPH_FFINTFC_FIFO_PTR, fifo_ptr); + nvkm_wr32(device, NV10_PGRAPH_FFINTFC_ST2, st2); + nvkm_wr32(device, NV10_PGRAPH_FFINTFC_ST2_DL, st2_dl); + nvkm_wr32(device, NV10_PGRAPH_FFINTFC_ST2_DH, st2_dh); /* Restore the current ctx object */ for (i = 0; i < 5; i++) - nv_wr32(priv, NV10_PGRAPH_CTX_SWITCH(i), ctx_switch[i]); - nv_wr32(priv, NV10_PGRAPH_CTX_USER, ctx_user); + nvkm_wr32(device, NV10_PGRAPH_CTX_SWITCH(i), ctx_switch[i]); + nvkm_wr32(device, NV10_PGRAPH_CTX_USER, ctx_user); } static int nv10_gr_load_context(struct nv10_gr_chan *chan, int chid) { - struct nv10_gr_priv *priv = nv10_gr_priv(chan); + struct nv10_gr *gr = chan->gr; + struct nvkm_device *device = gr->base.engine.subdev.device; u32 inst; int i; for (i = 0; i < ARRAY_SIZE(nv10_gr_ctx_regs); i++) - nv_wr32(priv, nv10_gr_ctx_regs[i], chan->nv10[i]); + nvkm_wr32(device, nv10_gr_ctx_regs[i], chan->nv10[i]); - if (nv_device(priv)->card_type >= NV_11 && - nv_device(priv)->chipset >= 0x17) { + if (device->card_type >= NV_11 && device->chipset >= 0x17) { for (i = 0; i < ARRAY_SIZE(nv17_gr_ctx_regs); i++) - nv_wr32(priv, nv17_gr_ctx_regs[i], chan->nv17[i]); + nvkm_wr32(device, nv17_gr_ctx_regs[i], chan->nv17[i]); } nv10_gr_load_pipe(chan); - inst = nv_rd32(priv, NV10_PGRAPH_GLOBALSTATE1) & 0xffff; + inst = nvkm_rd32(device, NV10_PGRAPH_GLOBALSTATE1) & 0xffff; nv10_gr_load_dma_vtxbuf(chan, chid, inst); - nv_wr32(priv, NV10_PGRAPH_CTX_CONTROL, 0x10010100); - nv_mask(priv, NV10_PGRAPH_CTX_USER, 0xff000000, chid << 24); - nv_mask(priv, NV10_PGRAPH_FFINTFC_ST2, 0x30000000, 0x00000000); + nvkm_wr32(device, NV10_PGRAPH_CTX_CONTROL, 0x10010100); + nvkm_mask(device, NV10_PGRAPH_CTX_USER, 0xff000000, chid << 24); + nvkm_mask(device, NV10_PGRAPH_FFINTFC_ST2, 0x30000000, 0x00000000); return 0; } static int nv10_gr_unload_context(struct nv10_gr_chan *chan) { - struct nv10_gr_priv *priv = nv10_gr_priv(chan); + struct nv10_gr *gr = chan->gr; + struct nvkm_device *device = gr->base.engine.subdev.device; int i; for (i = 0; i < ARRAY_SIZE(nv10_gr_ctx_regs); i++) - chan->nv10[i] = nv_rd32(priv, nv10_gr_ctx_regs[i]); + chan->nv10[i] = nvkm_rd32(device, nv10_gr_ctx_regs[i]); - if (nv_device(priv)->card_type >= NV_11 && - nv_device(priv)->chipset >= 0x17) { + if (device->card_type >= NV_11 && device->chipset >= 0x17) { for (i = 0; i < ARRAY_SIZE(nv17_gr_ctx_regs); i++) - chan->nv17[i] = nv_rd32(priv, nv17_gr_ctx_regs[i]); + chan->nv17[i] = nvkm_rd32(device, nv17_gr_ctx_regs[i]); } nv10_gr_save_pipe(chan); - nv_wr32(priv, NV10_PGRAPH_CTX_CONTROL, 0x10000000); - nv_mask(priv, NV10_PGRAPH_CTX_USER, 0xff000000, 0x1f000000); + nvkm_wr32(device, NV10_PGRAPH_CTX_CONTROL, 0x10000000); + nvkm_mask(device, NV10_PGRAPH_CTX_USER, 0xff000000, 0x1f000000); return 0; } static void -nv10_gr_context_switch(struct nv10_gr_priv *priv) +nv10_gr_context_switch(struct nv10_gr *gr) { + struct nvkm_device *device = gr->base.engine.subdev.device; struct nv10_gr_chan *prev = NULL; struct nv10_gr_chan *next = NULL; - unsigned long flags; int chid; - spin_lock_irqsave(&priv->lock, flags); - nv04_gr_idle(priv); + nv04_gr_idle(&gr->base); /* If previous context is valid, we need to save it */ - prev = nv10_gr_channel(priv); + prev = nv10_gr_channel(gr); if (prev) nv10_gr_unload_context(prev); /* load context for next channel */ - chid = (nv_rd32(priv, NV04_PGRAPH_TRAPPED_ADDR) >> 20) & 0x1f; - next = priv->chan[chid]; + chid = (nvkm_rd32(device, NV04_PGRAPH_TRAPPED_ADDR) >> 20) & 0x1f; + next = gr->chan[chid]; if (next) nv10_gr_load_context(next, chid); +} + +static int +nv10_gr_chan_fini(struct nvkm_object *object, bool suspend) +{ + struct nv10_gr_chan *chan = nv10_gr_chan(object); + struct nv10_gr *gr = chan->gr; + struct nvkm_device *device = gr->base.engine.subdev.device; + unsigned long flags; - spin_unlock_irqrestore(&priv->lock, flags); + spin_lock_irqsave(&gr->lock, flags); + nvkm_mask(device, NV04_PGRAPH_FIFO, 0x00000001, 0x00000000); + if (nv10_gr_channel(gr) == chan) + nv10_gr_unload_context(chan); + nvkm_mask(device, NV04_PGRAPH_FIFO, 0x00000001, 0x00000001); + spin_unlock_irqrestore(&gr->lock, flags); + return 0; +} + +static void * +nv10_gr_chan_dtor(struct nvkm_object *object) +{ + struct nv10_gr_chan *chan = nv10_gr_chan(object); + struct nv10_gr *gr = chan->gr; + unsigned long flags; + + spin_lock_irqsave(&gr->lock, flags); + gr->chan[chan->chid] = NULL; + spin_unlock_irqrestore(&gr->lock, flags); + return chan; } +static const struct nvkm_object_func +nv10_gr_chan = { + .dtor = nv10_gr_chan_dtor, + .fini = nv10_gr_chan_fini, +}; + #define NV_WRITE_CTX(reg, val) do { \ - int offset = nv10_gr_ctx_regs_find_offset(priv, reg); \ + int offset = nv10_gr_ctx_regs_find_offset(gr, reg); \ if (offset > 0) \ chan->nv10[offset] = val; \ } while (0) #define NV17_WRITE_CTX(reg, val) do { \ - int offset = nv17_gr_ctx_regs_find_offset(priv, reg); \ + int offset = nv17_gr_ctx_regs_find_offset(gr, reg); \ if (offset > 0) \ chan->nv17[offset] = val; \ } while (0) -static int -nv10_gr_context_ctor(struct nvkm_object *parent, struct nvkm_object *engine, - struct nvkm_oclass *oclass, void *data, u32 size, - struct nvkm_object **pobject) +int +nv10_gr_chan_new(struct nvkm_gr *base, struct nvkm_fifo_chan *fifoch, + const struct nvkm_oclass *oclass, struct nvkm_object **pobject) { - struct nvkm_fifo_chan *fifo = (void *)parent; - struct nv10_gr_priv *priv = (void *)engine; + struct nv10_gr *gr = nv10_gr(base); struct nv10_gr_chan *chan; + struct nvkm_device *device = gr->base.engine.subdev.device; unsigned long flags; - int ret; - - ret = nvkm_object_create(parent, engine, oclass, 0, &chan); - *pobject = nv_object(chan); - if (ret) - return ret; - - spin_lock_irqsave(&priv->lock, flags); - if (priv->chan[fifo->chid]) { - *pobject = nv_object(priv->chan[fifo->chid]); - atomic_inc(&(*pobject)->refcount); - spin_unlock_irqrestore(&priv->lock, flags); - nvkm_object_destroy(&chan->base); - return 1; - } + + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&nv10_gr_chan, oclass, &chan->object); + chan->gr = gr; + chan->chid = fifoch->chid; + *pobject = &chan->object; NV_WRITE_CTX(0x00400e88, 0x08000000); NV_WRITE_CTX(0x00400e9c, 0x4b7fffff); @@ -1049,12 +1021,11 @@ nv10_gr_context_ctor(struct nvkm_object *parent, struct nvkm_object *engine, NV_WRITE_CTX(0x00400e14, 0x00001000); NV_WRITE_CTX(0x00400e30, 0x00080008); NV_WRITE_CTX(0x00400e34, 0x00080008); - if (nv_device(priv)->card_type >= NV_11 && - nv_device(priv)->chipset >= 0x17) { + if (device->card_type >= NV_11 && device->chipset >= 0x17) { /* is it really needed ??? */ NV17_WRITE_CTX(NV10_PGRAPH_DEBUG_4, - nv_rd32(priv, NV10_PGRAPH_DEBUG_4)); - NV17_WRITE_CTX(0x004006b0, nv_rd32(priv, 0x004006b0)); + nvkm_rd32(device, NV10_PGRAPH_DEBUG_4)); + NV17_WRITE_CTX(0x004006b0, nvkm_rd32(device, 0x004006b0)); NV17_WRITE_CTX(0x00400eac, 0x0fff0000); NV17_WRITE_CTX(0x00400eb0, 0x0fff0000); NV17_WRITE_CTX(0x00400ec0, 0x00000080); @@ -1064,74 +1035,32 @@ nv10_gr_context_ctor(struct nvkm_object *parent, struct nvkm_object *engine, nv10_gr_create_pipe(chan); - priv->chan[fifo->chid] = chan; - chan->chid = fifo->chid; - spin_unlock_irqrestore(&priv->lock, flags); + spin_lock_irqsave(&gr->lock, flags); + gr->chan[chan->chid] = chan; + spin_unlock_irqrestore(&gr->lock, flags); return 0; } -static void -nv10_gr_context_dtor(struct nvkm_object *object) -{ - struct nv10_gr_priv *priv = (void *)object->engine; - struct nv10_gr_chan *chan = (void *)object; - unsigned long flags; - - spin_lock_irqsave(&priv->lock, flags); - priv->chan[chan->chid] = NULL; - spin_unlock_irqrestore(&priv->lock, flags); - - nvkm_object_destroy(&chan->base); -} - -static int -nv10_gr_context_fini(struct nvkm_object *object, bool suspend) -{ - struct nv10_gr_priv *priv = (void *)object->engine; - struct nv10_gr_chan *chan = (void *)object; - unsigned long flags; - - spin_lock_irqsave(&priv->lock, flags); - nv_mask(priv, NV04_PGRAPH_FIFO, 0x00000001, 0x00000000); - if (nv10_gr_channel(priv) == chan) - nv10_gr_unload_context(chan); - nv_mask(priv, NV04_PGRAPH_FIFO, 0x00000001, 0x00000001); - spin_unlock_irqrestore(&priv->lock, flags); - - return nvkm_object_fini(&chan->base, suspend); -} - -static struct nvkm_oclass -nv10_gr_cclass = { - .handle = NV_ENGCTX(GR, 0x10), - .ofuncs = &(struct nvkm_ofuncs) { - .ctor = nv10_gr_context_ctor, - .dtor = nv10_gr_context_dtor, - .init = nvkm_object_init, - .fini = nv10_gr_context_fini, - }, -}; - /******************************************************************************* * PGRAPH engine/subdev functions ******************************************************************************/ -static void -nv10_gr_tile_prog(struct nvkm_engine *engine, int i) +void +nv10_gr_tile(struct nvkm_gr *base, int i, struct nvkm_fb_tile *tile) { - struct nvkm_fb_tile *tile = &nvkm_fb(engine)->tile.region[i]; - struct nvkm_fifo *pfifo = nvkm_fifo(engine); - struct nv10_gr_priv *priv = (void *)engine; + struct nv10_gr *gr = nv10_gr(base); + struct nvkm_device *device = gr->base.engine.subdev.device; + struct nvkm_fifo *fifo = device->fifo; unsigned long flags; - pfifo->pause(pfifo, &flags); - nv04_gr_idle(priv); + nvkm_fifo_pause(fifo, &flags); + nv04_gr_idle(&gr->base); - nv_wr32(priv, NV10_PGRAPH_TLIMIT(i), tile->limit); - nv_wr32(priv, NV10_PGRAPH_TSIZE(i), tile->pitch); - nv_wr32(priv, NV10_PGRAPH_TILE(i), tile->addr); + nvkm_wr32(device, NV10_PGRAPH_TLIMIT(i), tile->limit); + nvkm_wr32(device, NV10_PGRAPH_TSIZE(i), tile->pitch); + nvkm_wr32(device, NV10_PGRAPH_TILE(i), tile->addr); - pfifo->start(pfifo, &flags); + nvkm_fifo_start(fifo, &flags); } const struct nvkm_bitfield nv10_gr_intr_name[] = { @@ -1148,168 +1077,145 @@ const struct nvkm_bitfield nv10_gr_nstatus[] = { {} }; -static void -nv10_gr_intr(struct nvkm_subdev *subdev) +void +nv10_gr_intr(struct nvkm_gr *base) { - struct nv10_gr_priv *priv = (void *)subdev; - struct nv10_gr_chan *chan = NULL; - struct nvkm_namedb *namedb = NULL; - struct nvkm_handle *handle = NULL; - u32 stat = nv_rd32(priv, NV03_PGRAPH_INTR); - u32 nsource = nv_rd32(priv, NV03_PGRAPH_NSOURCE); - u32 nstatus = nv_rd32(priv, NV03_PGRAPH_NSTATUS); - u32 addr = nv_rd32(priv, NV04_PGRAPH_TRAPPED_ADDR); + struct nv10_gr *gr = nv10_gr(base); + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 stat = nvkm_rd32(device, NV03_PGRAPH_INTR); + u32 nsource = nvkm_rd32(device, NV03_PGRAPH_NSOURCE); + u32 nstatus = nvkm_rd32(device, NV03_PGRAPH_NSTATUS); + u32 addr = nvkm_rd32(device, NV04_PGRAPH_TRAPPED_ADDR); u32 chid = (addr & 0x01f00000) >> 20; u32 subc = (addr & 0x00070000) >> 16; u32 mthd = (addr & 0x00001ffc); - u32 data = nv_rd32(priv, NV04_PGRAPH_TRAPPED_DATA); - u32 class = nv_rd32(priv, 0x400160 + subc * 4) & 0xfff; + u32 data = nvkm_rd32(device, NV04_PGRAPH_TRAPPED_DATA); + u32 class = nvkm_rd32(device, 0x400160 + subc * 4) & 0xfff; u32 show = stat; + char msg[128], src[128], sta[128]; + struct nv10_gr_chan *chan; unsigned long flags; - spin_lock_irqsave(&priv->lock, flags); - chan = priv->chan[chid]; - if (chan) - namedb = (void *)nv_pclass(nv_object(chan), NV_NAMEDB_CLASS); - spin_unlock_irqrestore(&priv->lock, flags); + spin_lock_irqsave(&gr->lock, flags); + chan = gr->chan[chid]; if (stat & NV_PGRAPH_INTR_ERROR) { if (chan && (nsource & NV03_PGRAPH_NSOURCE_ILLEGAL_MTHD)) { - handle = nvkm_namedb_get_class(namedb, class); - if (handle && !nv_call(handle->object, mthd, data)) + if (!nv10_gr_mthd(chan, class, mthd, data)) show &= ~NV_PGRAPH_INTR_ERROR; } } if (stat & NV_PGRAPH_INTR_CONTEXT_SWITCH) { - nv_wr32(priv, NV03_PGRAPH_INTR, NV_PGRAPH_INTR_CONTEXT_SWITCH); + nvkm_wr32(device, NV03_PGRAPH_INTR, NV_PGRAPH_INTR_CONTEXT_SWITCH); stat &= ~NV_PGRAPH_INTR_CONTEXT_SWITCH; show &= ~NV_PGRAPH_INTR_CONTEXT_SWITCH; - nv10_gr_context_switch(priv); + nv10_gr_context_switch(gr); } - nv_wr32(priv, NV03_PGRAPH_INTR, stat); - nv_wr32(priv, NV04_PGRAPH_FIFO, 0x00000001); + nvkm_wr32(device, NV03_PGRAPH_INTR, stat); + nvkm_wr32(device, NV04_PGRAPH_FIFO, 0x00000001); if (show) { - nv_error(priv, "%s", ""); - nvkm_bitfield_print(nv10_gr_intr_name, show); - pr_cont(" nsource:"); - nvkm_bitfield_print(nv04_gr_nsource, nsource); - pr_cont(" nstatus:"); - nvkm_bitfield_print(nv10_gr_nstatus, nstatus); - pr_cont("\n"); - nv_error(priv, - "ch %d [%s] subc %d class 0x%04x mthd 0x%04x data 0x%08x\n", - chid, nvkm_client_name(chan), subc, class, mthd, - data); + nvkm_snprintbf(msg, sizeof(msg), nv10_gr_intr_name, show); + nvkm_snprintbf(src, sizeof(src), nv04_gr_nsource, nsource); + nvkm_snprintbf(sta, sizeof(sta), nv10_gr_nstatus, nstatus); + nvkm_error(subdev, "intr %08x [%s] nsource %08x [%s] " + "nstatus %08x [%s] ch %d [%s] subc %d " + "class %04x mthd %04x data %08x\n", + show, msg, nsource, src, nstatus, sta, chid, + chan ? chan->object.client->name : "unknown", + subc, class, mthd, data); } - nvkm_namedb_put(handle); + spin_unlock_irqrestore(&gr->lock, flags); } -static int -nv10_gr_ctor(struct nvkm_object *parent, struct nvkm_object *engine, - struct nvkm_oclass *oclass, void *data, u32 size, - struct nvkm_object **pobject) +int +nv10_gr_init(struct nvkm_gr *base) { - struct nv10_gr_priv *priv; - int ret; - - ret = nvkm_gr_create(parent, engine, oclass, true, &priv); - *pobject = nv_object(priv); - if (ret) - return ret; - - nv_subdev(priv)->unit = 0x00001000; - nv_subdev(priv)->intr = nv10_gr_intr; - nv_engine(priv)->cclass = &nv10_gr_cclass; - - if (nv_device(priv)->chipset <= 0x10) - nv_engine(priv)->sclass = nv10_gr_sclass; - else - if (nv_device(priv)->chipset < 0x17 || - nv_device(priv)->card_type < NV_11) - nv_engine(priv)->sclass = nv15_gr_sclass; - else - nv_engine(priv)->sclass = nv17_gr_sclass; - - nv_engine(priv)->tile_prog = nv10_gr_tile_prog; - spin_lock_init(&priv->lock); - return 0; -} - -static void -nv10_gr_dtor(struct nvkm_object *object) -{ - struct nv10_gr_priv *priv = (void *)object; - nvkm_gr_destroy(&priv->base); -} - -static int -nv10_gr_init(struct nvkm_object *object) -{ - struct nvkm_engine *engine = nv_engine(object); - struct nvkm_fb *pfb = nvkm_fb(object); - struct nv10_gr_priv *priv = (void *)engine; - int ret, i; - - ret = nvkm_gr_init(&priv->base); - if (ret) - return ret; - - nv_wr32(priv, NV03_PGRAPH_INTR , 0xFFFFFFFF); - nv_wr32(priv, NV03_PGRAPH_INTR_EN, 0xFFFFFFFF); - - nv_wr32(priv, NV04_PGRAPH_DEBUG_0, 0xFFFFFFFF); - nv_wr32(priv, NV04_PGRAPH_DEBUG_0, 0x00000000); - nv_wr32(priv, NV04_PGRAPH_DEBUG_1, 0x00118700); - /* nv_wr32(priv, NV04_PGRAPH_DEBUG_2, 0x24E00810); */ /* 0x25f92ad9 */ - nv_wr32(priv, NV04_PGRAPH_DEBUG_2, 0x25f92ad9); - nv_wr32(priv, NV04_PGRAPH_DEBUG_3, 0x55DE0830 | (1 << 29) | (1 << 31)); - - if (nv_device(priv)->card_type >= NV_11 && - nv_device(priv)->chipset >= 0x17) { - nv_wr32(priv, NV10_PGRAPH_DEBUG_4, 0x1f000000); - nv_wr32(priv, 0x400a10, 0x03ff3fb6); - nv_wr32(priv, 0x400838, 0x002f8684); - nv_wr32(priv, 0x40083c, 0x00115f3f); - nv_wr32(priv, 0x4006b0, 0x40000020); + struct nv10_gr *gr = nv10_gr(base); + struct nvkm_device *device = gr->base.engine.subdev.device; + + nvkm_wr32(device, NV03_PGRAPH_INTR , 0xFFFFFFFF); + nvkm_wr32(device, NV03_PGRAPH_INTR_EN, 0xFFFFFFFF); + + nvkm_wr32(device, NV04_PGRAPH_DEBUG_0, 0xFFFFFFFF); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_0, 0x00000000); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_1, 0x00118700); + /* nvkm_wr32(device, NV04_PGRAPH_DEBUG_2, 0x24E00810); */ /* 0x25f92ad9 */ + nvkm_wr32(device, NV04_PGRAPH_DEBUG_2, 0x25f92ad9); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_3, 0x55DE0830 | (1 << 29) | (1 << 31)); + + if (device->card_type >= NV_11 && device->chipset >= 0x17) { + nvkm_wr32(device, NV10_PGRAPH_DEBUG_4, 0x1f000000); + nvkm_wr32(device, 0x400a10, 0x03ff3fb6); + nvkm_wr32(device, 0x400838, 0x002f8684); + nvkm_wr32(device, 0x40083c, 0x00115f3f); + nvkm_wr32(device, 0x4006b0, 0x40000020); } else { - nv_wr32(priv, NV10_PGRAPH_DEBUG_4, 0x00000000); + nvkm_wr32(device, NV10_PGRAPH_DEBUG_4, 0x00000000); } - /* Turn all the tiling regions off. */ - for (i = 0; i < pfb->tile.regions; i++) - engine->tile_prog(engine, i); - - nv_wr32(priv, NV10_PGRAPH_CTX_SWITCH(0), 0x00000000); - nv_wr32(priv, NV10_PGRAPH_CTX_SWITCH(1), 0x00000000); - nv_wr32(priv, NV10_PGRAPH_CTX_SWITCH(2), 0x00000000); - nv_wr32(priv, NV10_PGRAPH_CTX_SWITCH(3), 0x00000000); - nv_wr32(priv, NV10_PGRAPH_CTX_SWITCH(4), 0x00000000); - nv_wr32(priv, NV10_PGRAPH_STATE, 0xFFFFFFFF); + nvkm_wr32(device, NV10_PGRAPH_CTX_SWITCH(0), 0x00000000); + nvkm_wr32(device, NV10_PGRAPH_CTX_SWITCH(1), 0x00000000); + nvkm_wr32(device, NV10_PGRAPH_CTX_SWITCH(2), 0x00000000); + nvkm_wr32(device, NV10_PGRAPH_CTX_SWITCH(3), 0x00000000); + nvkm_wr32(device, NV10_PGRAPH_CTX_SWITCH(4), 0x00000000); + nvkm_wr32(device, NV10_PGRAPH_STATE, 0xFFFFFFFF); - nv_mask(priv, NV10_PGRAPH_CTX_USER, 0xff000000, 0x1f000000); - nv_wr32(priv, NV10_PGRAPH_CTX_CONTROL, 0x10000100); - nv_wr32(priv, NV10_PGRAPH_FFINTFC_ST2, 0x08000000); + nvkm_mask(device, NV10_PGRAPH_CTX_USER, 0xff000000, 0x1f000000); + nvkm_wr32(device, NV10_PGRAPH_CTX_CONTROL, 0x10000100); + nvkm_wr32(device, NV10_PGRAPH_FFINTFC_ST2, 0x08000000); return 0; } -static int -nv10_gr_fini(struct nvkm_object *object, bool suspend) +int +nv10_gr_new_(const struct nvkm_gr_func *func, struct nvkm_device *device, + int index, struct nvkm_gr **pgr) { - struct nv10_gr_priv *priv = (void *)object; - return nvkm_gr_fini(&priv->base, suspend); + struct nv10_gr *gr; + + if (!(gr = kzalloc(sizeof(*gr), GFP_KERNEL))) + return -ENOMEM; + spin_lock_init(&gr->lock); + *pgr = &gr->base; + + return nvkm_gr_ctor(func, device, index, 0x00001000, true, &gr->base); } -struct nvkm_oclass -nv10_gr_oclass = { - .handle = NV_ENGINE(GR, 0x10), - .ofuncs = &(struct nvkm_ofuncs) { - .ctor = nv10_gr_ctor, - .dtor = nv10_gr_dtor, - .init = nv10_gr_init, - .fini = nv10_gr_fini, - }, +static const struct nvkm_gr_func +nv10_gr = { + .init = nv10_gr_init, + .intr = nv10_gr_intr, + .tile = nv10_gr_tile, + .chan_new = nv10_gr_chan_new, + .sclass = { + { -1, -1, 0x0012, &nv04_gr_object }, /* beta1 */ + { -1, -1, 0x0019, &nv04_gr_object }, /* clip */ + { -1, -1, 0x0030, &nv04_gr_object }, /* null */ + { -1, -1, 0x0039, &nv04_gr_object }, /* m2mf */ + { -1, -1, 0x0043, &nv04_gr_object }, /* rop */ + { -1, -1, 0x0044, &nv04_gr_object }, /* pattern */ + { -1, -1, 0x004a, &nv04_gr_object }, /* gdi */ + { -1, -1, 0x0052, &nv04_gr_object }, /* swzsurf */ + { -1, -1, 0x005f, &nv04_gr_object }, /* blit */ + { -1, -1, 0x0062, &nv04_gr_object }, /* surf2d */ + { -1, -1, 0x0072, &nv04_gr_object }, /* beta4 */ + { -1, -1, 0x0089, &nv04_gr_object }, /* sifm */ + { -1, -1, 0x008a, &nv04_gr_object }, /* ifc */ + { -1, -1, 0x009f, &nv04_gr_object }, /* blit */ + { -1, -1, 0x0093, &nv04_gr_object }, /* surf3d */ + { -1, -1, 0x0094, &nv04_gr_object }, /* ttri */ + { -1, -1, 0x0095, &nv04_gr_object }, /* mtri */ + { -1, -1, 0x0056, &nv04_gr_object }, /* celcius */ + {} + } }; + +int +nv10_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return nv10_gr_new_(&nv10_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv10.h b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv10.h new file mode 100644 index 000000000..d7c3d86cc --- /dev/null +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv10.h @@ -0,0 +1,13 @@ +#ifndef __NV10_GR_H__ +#define __NV10_GR_H__ +#include "priv.h" + +int nv10_gr_new_(const struct nvkm_gr_func *, struct nvkm_device *, int index, + struct nvkm_gr **); +int nv10_gr_init(struct nvkm_gr *); +void nv10_gr_intr(struct nvkm_gr *); +void nv10_gr_tile(struct nvkm_gr *, int, struct nvkm_fb_tile *); + +int nv10_gr_chan_new(struct nvkm_gr *, struct nvkm_fifo_chan *, + const struct nvkm_oclass *, struct nvkm_object **); +#endif diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv15.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv15.c new file mode 100644 index 000000000..3e2c6856b --- /dev/null +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv15.c @@ -0,0 +1,59 @@ +/* + * Copyright 2007 Matthieu CASTET <castet.matthieu@free.fr> + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragr) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ +#include "nv10.h" + +static const struct nvkm_gr_func +nv15_gr = { + .init = nv10_gr_init, + .intr = nv10_gr_intr, + .tile = nv10_gr_tile, + .chan_new = nv10_gr_chan_new, + .sclass = { + { -1, -1, 0x0012, &nv04_gr_object }, /* beta1 */ + { -1, -1, 0x0019, &nv04_gr_object }, /* clip */ + { -1, -1, 0x0030, &nv04_gr_object }, /* null */ + { -1, -1, 0x0039, &nv04_gr_object }, /* m2mf */ + { -1, -1, 0x0043, &nv04_gr_object }, /* rop */ + { -1, -1, 0x0044, &nv04_gr_object }, /* pattern */ + { -1, -1, 0x004a, &nv04_gr_object }, /* gdi */ + { -1, -1, 0x0052, &nv04_gr_object }, /* swzsurf */ + { -1, -1, 0x005f, &nv04_gr_object }, /* blit */ + { -1, -1, 0x0062, &nv04_gr_object }, /* surf2d */ + { -1, -1, 0x0072, &nv04_gr_object }, /* beta4 */ + { -1, -1, 0x0089, &nv04_gr_object }, /* sifm */ + { -1, -1, 0x008a, &nv04_gr_object }, /* ifc */ + { -1, -1, 0x009f, &nv04_gr_object }, /* blit */ + { -1, -1, 0x0093, &nv04_gr_object }, /* surf3d */ + { -1, -1, 0x0094, &nv04_gr_object }, /* ttri */ + { -1, -1, 0x0095, &nv04_gr_object }, /* mtri */ + { -1, -1, 0x0096, &nv04_gr_object }, /* celcius */ + {} + } +}; + +int +nv15_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return nv10_gr_new_(&nv15_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv17.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv17.c new file mode 100644 index 000000000..12437d085 --- /dev/null +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv17.c @@ -0,0 +1,59 @@ +/* + * Copyright 2007 Matthieu CASTET <castet.matthieu@free.fr> + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragr) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + */ +#include "nv10.h" + +static const struct nvkm_gr_func +nv17_gr = { + .init = nv10_gr_init, + .intr = nv10_gr_intr, + .tile = nv10_gr_tile, + .chan_new = nv10_gr_chan_new, + .sclass = { + { -1, -1, 0x0012, &nv04_gr_object }, /* beta1 */ + { -1, -1, 0x0019, &nv04_gr_object }, /* clip */ + { -1, -1, 0x0030, &nv04_gr_object }, /* null */ + { -1, -1, 0x0039, &nv04_gr_object }, /* m2mf */ + { -1, -1, 0x0043, &nv04_gr_object }, /* rop */ + { -1, -1, 0x0044, &nv04_gr_object }, /* pattern */ + { -1, -1, 0x004a, &nv04_gr_object }, /* gdi */ + { -1, -1, 0x0052, &nv04_gr_object }, /* swzsurf */ + { -1, -1, 0x005f, &nv04_gr_object }, /* blit */ + { -1, -1, 0x0062, &nv04_gr_object }, /* surf2d */ + { -1, -1, 0x0072, &nv04_gr_object }, /* beta4 */ + { -1, -1, 0x0089, &nv04_gr_object }, /* sifm */ + { -1, -1, 0x008a, &nv04_gr_object }, /* ifc */ + { -1, -1, 0x009f, &nv04_gr_object }, /* blit */ + { -1, -1, 0x0093, &nv04_gr_object }, /* surf3d */ + { -1, -1, 0x0094, &nv04_gr_object }, /* ttri */ + { -1, -1, 0x0095, &nv04_gr_object }, /* mtri */ + { -1, -1, 0x0099, &nv04_gr_object }, + {} + } +}; + +int +nv17_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return nv10_gr_new_(&nv17_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.c index 1713ffb66..5caef65d3 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.c @@ -2,375 +2,374 @@ #include "regs.h" #include <core/client.h> -#include <core/device.h> -#include <core/handle.h> +#include <core/gpuobj.h> #include <engine/fifo.h> +#include <engine/fifo/chan.h> #include <subdev/fb.h> #include <subdev/timer.h> /******************************************************************************* - * Graphics object classes + * PGRAPH context ******************************************************************************/ -static struct nvkm_oclass -nv20_gr_sclass[] = { - { 0x0012, &nv04_gr_ofuncs, NULL }, /* beta1 */ - { 0x0019, &nv04_gr_ofuncs, NULL }, /* clip */ - { 0x0030, &nv04_gr_ofuncs, NULL }, /* null */ - { 0x0039, &nv04_gr_ofuncs, NULL }, /* m2mf */ - { 0x0043, &nv04_gr_ofuncs, NULL }, /* rop */ - { 0x0044, &nv04_gr_ofuncs, NULL }, /* patt */ - { 0x004a, &nv04_gr_ofuncs, NULL }, /* gdi */ - { 0x0062, &nv04_gr_ofuncs, NULL }, /* surf2d */ - { 0x0072, &nv04_gr_ofuncs, NULL }, /* beta4 */ - { 0x0089, &nv04_gr_ofuncs, NULL }, /* sifm */ - { 0x008a, &nv04_gr_ofuncs, NULL }, /* ifc */ - { 0x0096, &nv04_gr_ofuncs, NULL }, /* celcius */ - { 0x0097, &nv04_gr_ofuncs, NULL }, /* kelvin */ - { 0x009e, &nv04_gr_ofuncs, NULL }, /* swzsurf */ - { 0x009f, &nv04_gr_ofuncs, NULL }, /* imageblit */ - {}, -}; +int +nv20_gr_chan_init(struct nvkm_object *object) +{ + struct nv20_gr_chan *chan = nv20_gr_chan(object); + struct nv20_gr *gr = chan->gr; + u32 inst = nvkm_memory_addr(chan->inst); -/******************************************************************************* - * PGRAPH context - ******************************************************************************/ + nvkm_kmap(gr->ctxtab); + nvkm_wo32(gr->ctxtab, chan->chid * 4, inst >> 4); + nvkm_done(gr->ctxtab); + return 0; +} + +int +nv20_gr_chan_fini(struct nvkm_object *object, bool suspend) +{ + struct nv20_gr_chan *chan = nv20_gr_chan(object); + struct nv20_gr *gr = chan->gr; + struct nvkm_device *device = gr->base.engine.subdev.device; + u32 inst = nvkm_memory_addr(chan->inst); + int chid = -1; + + nvkm_mask(device, 0x400720, 0x00000001, 0x00000000); + if (nvkm_rd32(device, 0x400144) & 0x00010000) + chid = (nvkm_rd32(device, 0x400148) & 0x1f000000) >> 24; + if (chan->chid == chid) { + nvkm_wr32(device, 0x400784, inst >> 4); + nvkm_wr32(device, 0x400788, 0x00000002); + nvkm_msec(device, 2000, + if (!nvkm_rd32(device, 0x400700)) + break; + ); + nvkm_wr32(device, 0x400144, 0x10000000); + nvkm_mask(device, 0x400148, 0xff000000, 0x1f000000); + } + nvkm_mask(device, 0x400720, 0x00000001, 0x00000001); + + nvkm_kmap(gr->ctxtab); + nvkm_wo32(gr->ctxtab, chan->chid * 4, 0x00000000); + nvkm_done(gr->ctxtab); + return 0; +} + +void * +nv20_gr_chan_dtor(struct nvkm_object *object) +{ + struct nv20_gr_chan *chan = nv20_gr_chan(object); + nvkm_memory_del(&chan->inst); + return chan; +} + +static const struct nvkm_object_func +nv20_gr_chan = { + .dtor = nv20_gr_chan_dtor, + .init = nv20_gr_chan_init, + .fini = nv20_gr_chan_fini, +}; static int -nv20_gr_context_ctor(struct nvkm_object *parent, struct nvkm_object *engine, - struct nvkm_oclass *oclass, void *data, u32 size, - struct nvkm_object **pobject) +nv20_gr_chan_new(struct nvkm_gr *base, struct nvkm_fifo_chan *fifoch, + const struct nvkm_oclass *oclass, struct nvkm_object **pobject) { + struct nv20_gr *gr = nv20_gr(base); struct nv20_gr_chan *chan; int ret, i; - ret = nvkm_gr_context_create(parent, engine, oclass, NULL, 0x37f0, - 16, NVOBJ_FLAG_ZERO_ALLOC, &chan); - *pobject = nv_object(chan); + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&nv20_gr_chan, oclass, &chan->object); + chan->gr = gr; + chan->chid = fifoch->chid; + *pobject = &chan->object; + + ret = nvkm_memory_new(gr->base.engine.subdev.device, + NVKM_MEM_TARGET_INST, 0x37f0, 16, true, + &chan->inst); if (ret) return ret; - chan->chid = nvkm_fifo_chan(parent)->chid; - - nv_wo32(chan, 0x0000, 0x00000001 | (chan->chid << 24)); - nv_wo32(chan, 0x033c, 0xffff0000); - nv_wo32(chan, 0x03a0, 0x0fff0000); - nv_wo32(chan, 0x03a4, 0x0fff0000); - nv_wo32(chan, 0x047c, 0x00000101); - nv_wo32(chan, 0x0490, 0x00000111); - nv_wo32(chan, 0x04a8, 0x44400000); + nvkm_kmap(chan->inst); + nvkm_wo32(chan->inst, 0x0000, 0x00000001 | (chan->chid << 24)); + nvkm_wo32(chan->inst, 0x033c, 0xffff0000); + nvkm_wo32(chan->inst, 0x03a0, 0x0fff0000); + nvkm_wo32(chan->inst, 0x03a4, 0x0fff0000); + nvkm_wo32(chan->inst, 0x047c, 0x00000101); + nvkm_wo32(chan->inst, 0x0490, 0x00000111); + nvkm_wo32(chan->inst, 0x04a8, 0x44400000); for (i = 0x04d4; i <= 0x04e0; i += 4) - nv_wo32(chan, i, 0x00030303); + nvkm_wo32(chan->inst, i, 0x00030303); for (i = 0x04f4; i <= 0x0500; i += 4) - nv_wo32(chan, i, 0x00080000); + nvkm_wo32(chan->inst, i, 0x00080000); for (i = 0x050c; i <= 0x0518; i += 4) - nv_wo32(chan, i, 0x01012000); + nvkm_wo32(chan->inst, i, 0x01012000); for (i = 0x051c; i <= 0x0528; i += 4) - nv_wo32(chan, i, 0x000105b8); + nvkm_wo32(chan->inst, i, 0x000105b8); for (i = 0x052c; i <= 0x0538; i += 4) - nv_wo32(chan, i, 0x00080008); + nvkm_wo32(chan->inst, i, 0x00080008); for (i = 0x055c; i <= 0x0598; i += 4) - nv_wo32(chan, i, 0x07ff0000); - nv_wo32(chan, 0x05a4, 0x4b7fffff); - nv_wo32(chan, 0x05fc, 0x00000001); - nv_wo32(chan, 0x0604, 0x00004000); - nv_wo32(chan, 0x0610, 0x00000001); - nv_wo32(chan, 0x0618, 0x00040000); - nv_wo32(chan, 0x061c, 0x00010000); + nvkm_wo32(chan->inst, i, 0x07ff0000); + nvkm_wo32(chan->inst, 0x05a4, 0x4b7fffff); + nvkm_wo32(chan->inst, 0x05fc, 0x00000001); + nvkm_wo32(chan->inst, 0x0604, 0x00004000); + nvkm_wo32(chan->inst, 0x0610, 0x00000001); + nvkm_wo32(chan->inst, 0x0618, 0x00040000); + nvkm_wo32(chan->inst, 0x061c, 0x00010000); for (i = 0x1c1c; i <= 0x248c; i += 16) { - nv_wo32(chan, (i + 0), 0x10700ff9); - nv_wo32(chan, (i + 4), 0x0436086c); - nv_wo32(chan, (i + 8), 0x000c001b); + nvkm_wo32(chan->inst, (i + 0), 0x10700ff9); + nvkm_wo32(chan->inst, (i + 4), 0x0436086c); + nvkm_wo32(chan->inst, (i + 8), 0x000c001b); } - nv_wo32(chan, 0x281c, 0x3f800000); - nv_wo32(chan, 0x2830, 0x3f800000); - nv_wo32(chan, 0x285c, 0x40000000); - nv_wo32(chan, 0x2860, 0x3f800000); - nv_wo32(chan, 0x2864, 0x3f000000); - nv_wo32(chan, 0x286c, 0x40000000); - nv_wo32(chan, 0x2870, 0x3f800000); - nv_wo32(chan, 0x2878, 0xbf800000); - nv_wo32(chan, 0x2880, 0xbf800000); - nv_wo32(chan, 0x34a4, 0x000fe000); - nv_wo32(chan, 0x3530, 0x000003f8); - nv_wo32(chan, 0x3540, 0x002fe000); + nvkm_wo32(chan->inst, 0x281c, 0x3f800000); + nvkm_wo32(chan->inst, 0x2830, 0x3f800000); + nvkm_wo32(chan->inst, 0x285c, 0x40000000); + nvkm_wo32(chan->inst, 0x2860, 0x3f800000); + nvkm_wo32(chan->inst, 0x2864, 0x3f000000); + nvkm_wo32(chan->inst, 0x286c, 0x40000000); + nvkm_wo32(chan->inst, 0x2870, 0x3f800000); + nvkm_wo32(chan->inst, 0x2878, 0xbf800000); + nvkm_wo32(chan->inst, 0x2880, 0xbf800000); + nvkm_wo32(chan->inst, 0x34a4, 0x000fe000); + nvkm_wo32(chan->inst, 0x3530, 0x000003f8); + nvkm_wo32(chan->inst, 0x3540, 0x002fe000); for (i = 0x355c; i <= 0x3578; i += 4) - nv_wo32(chan, i, 0x001c527c); - return 0; -} - -int -nv20_gr_context_init(struct nvkm_object *object) -{ - struct nv20_gr_priv *priv = (void *)object->engine; - struct nv20_gr_chan *chan = (void *)object; - int ret; - - ret = nvkm_gr_context_init(&chan->base); - if (ret) - return ret; - - nv_wo32(priv->ctxtab, chan->chid * 4, nv_gpuobj(chan)->addr >> 4); + nvkm_wo32(chan->inst, i, 0x001c527c); + nvkm_done(chan->inst); return 0; } -int -nv20_gr_context_fini(struct nvkm_object *object, bool suspend) -{ - struct nv20_gr_priv *priv = (void *)object->engine; - struct nv20_gr_chan *chan = (void *)object; - int chid = -1; - - nv_mask(priv, 0x400720, 0x00000001, 0x00000000); - if (nv_rd32(priv, 0x400144) & 0x00010000) - chid = (nv_rd32(priv, 0x400148) & 0x1f000000) >> 24; - if (chan->chid == chid) { - nv_wr32(priv, 0x400784, nv_gpuobj(chan)->addr >> 4); - nv_wr32(priv, 0x400788, 0x00000002); - nv_wait(priv, 0x400700, 0xffffffff, 0x00000000); - nv_wr32(priv, 0x400144, 0x10000000); - nv_mask(priv, 0x400148, 0xff000000, 0x1f000000); - } - nv_mask(priv, 0x400720, 0x00000001, 0x00000001); - - nv_wo32(priv->ctxtab, chan->chid * 4, 0x00000000); - return nvkm_gr_context_fini(&chan->base, suspend); -} - -static struct nvkm_oclass -nv20_gr_cclass = { - .handle = NV_ENGCTX(GR, 0x20), - .ofuncs = &(struct nvkm_ofuncs) { - .ctor = nv20_gr_context_ctor, - .dtor = _nvkm_gr_context_dtor, - .init = nv20_gr_context_init, - .fini = nv20_gr_context_fini, - .rd32 = _nvkm_gr_context_rd32, - .wr32 = _nvkm_gr_context_wr32, - }, -}; - /******************************************************************************* * PGRAPH engine/subdev functions ******************************************************************************/ void -nv20_gr_tile_prog(struct nvkm_engine *engine, int i) +nv20_gr_tile(struct nvkm_gr *base, int i, struct nvkm_fb_tile *tile) { - struct nvkm_fb_tile *tile = &nvkm_fb(engine)->tile.region[i]; - struct nvkm_fifo *pfifo = nvkm_fifo(engine); - struct nv20_gr_priv *priv = (void *)engine; + struct nv20_gr *gr = nv20_gr(base); + struct nvkm_device *device = gr->base.engine.subdev.device; + struct nvkm_fifo *fifo = device->fifo; unsigned long flags; - pfifo->pause(pfifo, &flags); - nv04_gr_idle(priv); + nvkm_fifo_pause(fifo, &flags); + nv04_gr_idle(&gr->base); - nv_wr32(priv, NV20_PGRAPH_TLIMIT(i), tile->limit); - nv_wr32(priv, NV20_PGRAPH_TSIZE(i), tile->pitch); - nv_wr32(priv, NV20_PGRAPH_TILE(i), tile->addr); + nvkm_wr32(device, NV20_PGRAPH_TLIMIT(i), tile->limit); + nvkm_wr32(device, NV20_PGRAPH_TSIZE(i), tile->pitch); + nvkm_wr32(device, NV20_PGRAPH_TILE(i), tile->addr); - nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00EA0030 + 4 * i); - nv_wr32(priv, NV10_PGRAPH_RDI_DATA, tile->limit); - nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00EA0050 + 4 * i); - nv_wr32(priv, NV10_PGRAPH_RDI_DATA, tile->pitch); - nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00EA0010 + 4 * i); - nv_wr32(priv, NV10_PGRAPH_RDI_DATA, tile->addr); + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00EA0030 + 4 * i); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA, tile->limit); + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00EA0050 + 4 * i); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA, tile->pitch); + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00EA0010 + 4 * i); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA, tile->addr); - if (nv_device(engine)->chipset != 0x34) { - nv_wr32(priv, NV20_PGRAPH_ZCOMP(i), tile->zcomp); - nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00ea0090 + 4 * i); - nv_wr32(priv, NV10_PGRAPH_RDI_DATA, tile->zcomp); + if (device->chipset != 0x34) { + nvkm_wr32(device, NV20_PGRAPH_ZCOMP(i), tile->zcomp); + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00ea0090 + 4 * i); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA, tile->zcomp); } - pfifo->start(pfifo, &flags); + nvkm_fifo_start(fifo, &flags); } void -nv20_gr_intr(struct nvkm_subdev *subdev) +nv20_gr_intr(struct nvkm_gr *base) { - struct nvkm_engine *engine = nv_engine(subdev); - struct nvkm_object *engctx; - struct nvkm_handle *handle; - struct nv20_gr_priv *priv = (void *)subdev; - u32 stat = nv_rd32(priv, NV03_PGRAPH_INTR); - u32 nsource = nv_rd32(priv, NV03_PGRAPH_NSOURCE); - u32 nstatus = nv_rd32(priv, NV03_PGRAPH_NSTATUS); - u32 addr = nv_rd32(priv, NV04_PGRAPH_TRAPPED_ADDR); + struct nv20_gr *gr = nv20_gr(base); + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + struct nvkm_fifo_chan *chan; + u32 stat = nvkm_rd32(device, NV03_PGRAPH_INTR); + u32 nsource = nvkm_rd32(device, NV03_PGRAPH_NSOURCE); + u32 nstatus = nvkm_rd32(device, NV03_PGRAPH_NSTATUS); + u32 addr = nvkm_rd32(device, NV04_PGRAPH_TRAPPED_ADDR); u32 chid = (addr & 0x01f00000) >> 20; u32 subc = (addr & 0x00070000) >> 16; u32 mthd = (addr & 0x00001ffc); - u32 data = nv_rd32(priv, NV04_PGRAPH_TRAPPED_DATA); - u32 class = nv_rd32(priv, 0x400160 + subc * 4) & 0xfff; + u32 data = nvkm_rd32(device, NV04_PGRAPH_TRAPPED_DATA); + u32 class = nvkm_rd32(device, 0x400160 + subc * 4) & 0xfff; u32 show = stat; + char msg[128], src[128], sta[128]; + unsigned long flags; - engctx = nvkm_engctx_get(engine, chid); - if (stat & NV_PGRAPH_INTR_ERROR) { - if (nsource & NV03_PGRAPH_NSOURCE_ILLEGAL_MTHD) { - handle = nvkm_handle_get_class(engctx, class); - if (handle && !nv_call(handle->object, mthd, data)) - show &= ~NV_PGRAPH_INTR_ERROR; - nvkm_handle_put(handle); - } - } + chan = nvkm_fifo_chan_chid(device->fifo, chid, &flags); - nv_wr32(priv, NV03_PGRAPH_INTR, stat); - nv_wr32(priv, NV04_PGRAPH_FIFO, 0x00000001); + nvkm_wr32(device, NV03_PGRAPH_INTR, stat); + nvkm_wr32(device, NV04_PGRAPH_FIFO, 0x00000001); if (show) { - nv_error(priv, "%s", ""); - nvkm_bitfield_print(nv10_gr_intr_name, show); - pr_cont(" nsource:"); - nvkm_bitfield_print(nv04_gr_nsource, nsource); - pr_cont(" nstatus:"); - nvkm_bitfield_print(nv10_gr_nstatus, nstatus); - pr_cont("\n"); - nv_error(priv, - "ch %d [%s] subc %d class 0x%04x mthd 0x%04x data 0x%08x\n", - chid, nvkm_client_name(engctx), subc, class, mthd, - data); + nvkm_snprintbf(msg, sizeof(msg), nv10_gr_intr_name, show); + nvkm_snprintbf(src, sizeof(src), nv04_gr_nsource, nsource); + nvkm_snprintbf(sta, sizeof(sta), nv10_gr_nstatus, nstatus); + nvkm_error(subdev, "intr %08x [%s] nsource %08x [%s] " + "nstatus %08x [%s] ch %d [%s] subc %d " + "class %04x mthd %04x data %08x\n", + show, msg, nsource, src, nstatus, sta, chid, + chan ? chan->object.client->name : "unknown", + subc, class, mthd, data); } - nvkm_engctx_put(engctx); -} - -static int -nv20_gr_ctor(struct nvkm_object *parent, struct nvkm_object *engine, - struct nvkm_oclass *oclass, void *data, u32 size, - struct nvkm_object **pobject) -{ - struct nv20_gr_priv *priv; - int ret; - - ret = nvkm_gr_create(parent, engine, oclass, true, &priv); - *pobject = nv_object(priv); - if (ret) - return ret; - - ret = nvkm_gpuobj_new(nv_object(priv), NULL, 32 * 4, 16, - NVOBJ_FLAG_ZERO_ALLOC, &priv->ctxtab); - if (ret) - return ret; - - nv_subdev(priv)->unit = 0x00001000; - nv_subdev(priv)->intr = nv20_gr_intr; - nv_engine(priv)->cclass = &nv20_gr_cclass; - nv_engine(priv)->sclass = nv20_gr_sclass; - nv_engine(priv)->tile_prog = nv20_gr_tile_prog; - return 0; + nvkm_fifo_chan_put(device->fifo, flags, &chan); } -void -nv20_gr_dtor(struct nvkm_object *object) +int +nv20_gr_oneinit(struct nvkm_gr *base) { - struct nv20_gr_priv *priv = (void *)object; - nvkm_gpuobj_ref(NULL, &priv->ctxtab); - nvkm_gr_destroy(&priv->base); + struct nv20_gr *gr = nv20_gr(base); + return nvkm_memory_new(gr->base.engine.subdev.device, + NVKM_MEM_TARGET_INST, 32 * 4, 16, + true, &gr->ctxtab); } int -nv20_gr_init(struct nvkm_object *object) +nv20_gr_init(struct nvkm_gr *base) { - struct nvkm_engine *engine = nv_engine(object); - struct nv20_gr_priv *priv = (void *)engine; - struct nvkm_fb *pfb = nvkm_fb(object); + struct nv20_gr *gr = nv20_gr(base); + struct nvkm_device *device = gr->base.engine.subdev.device; u32 tmp, vramsz; - int ret, i; - - ret = nvkm_gr_init(&priv->base); - if (ret) - return ret; + int i; - nv_wr32(priv, NV20_PGRAPH_CHANNEL_CTX_TABLE, priv->ctxtab->addr >> 4); + nvkm_wr32(device, NV20_PGRAPH_CHANNEL_CTX_TABLE, + nvkm_memory_addr(gr->ctxtab) >> 4); - if (nv_device(priv)->chipset == 0x20) { - nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x003d0000); + if (device->chipset == 0x20) { + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x003d0000); for (i = 0; i < 15; i++) - nv_wr32(priv, NV10_PGRAPH_RDI_DATA, 0x00000000); - nv_wait(priv, 0x400700, 0xffffffff, 0x00000000); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA, 0x00000000); + nvkm_msec(device, 2000, + if (!nvkm_rd32(device, 0x400700)) + break; + ); } else { - nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x02c80000); + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x02c80000); for (i = 0; i < 32; i++) - nv_wr32(priv, NV10_PGRAPH_RDI_DATA, 0x00000000); - nv_wait(priv, 0x400700, 0xffffffff, 0x00000000); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA, 0x00000000); + nvkm_msec(device, 2000, + if (!nvkm_rd32(device, 0x400700)) + break; + ); } - nv_wr32(priv, NV03_PGRAPH_INTR , 0xFFFFFFFF); - nv_wr32(priv, NV03_PGRAPH_INTR_EN, 0xFFFFFFFF); + nvkm_wr32(device, NV03_PGRAPH_INTR , 0xFFFFFFFF); + nvkm_wr32(device, NV03_PGRAPH_INTR_EN, 0xFFFFFFFF); - nv_wr32(priv, NV04_PGRAPH_DEBUG_0, 0xFFFFFFFF); - nv_wr32(priv, NV04_PGRAPH_DEBUG_0, 0x00000000); - nv_wr32(priv, NV04_PGRAPH_DEBUG_1, 0x00118700); - nv_wr32(priv, NV04_PGRAPH_DEBUG_3, 0xF3CE0475); /* 0x4 = auto ctx switch */ - nv_wr32(priv, NV10_PGRAPH_DEBUG_4, 0x00000000); - nv_wr32(priv, 0x40009C , 0x00000040); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_0, 0xFFFFFFFF); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_0, 0x00000000); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_1, 0x00118700); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_3, 0xF3CE0475); /* 0x4 = auto ctx switch */ + nvkm_wr32(device, NV10_PGRAPH_DEBUG_4, 0x00000000); + nvkm_wr32(device, 0x40009C , 0x00000040); - if (nv_device(priv)->chipset >= 0x25) { - nv_wr32(priv, 0x400890, 0x00a8cfff); - nv_wr32(priv, 0x400610, 0x304B1FB6); - nv_wr32(priv, 0x400B80, 0x1cbd3883); - nv_wr32(priv, 0x400B84, 0x44000000); - nv_wr32(priv, 0x400098, 0x40000080); - nv_wr32(priv, 0x400B88, 0x000000ff); + if (device->chipset >= 0x25) { + nvkm_wr32(device, 0x400890, 0x00a8cfff); + nvkm_wr32(device, 0x400610, 0x304B1FB6); + nvkm_wr32(device, 0x400B80, 0x1cbd3883); + nvkm_wr32(device, 0x400B84, 0x44000000); + nvkm_wr32(device, 0x400098, 0x40000080); + nvkm_wr32(device, 0x400B88, 0x000000ff); } else { - nv_wr32(priv, 0x400880, 0x0008c7df); - nv_wr32(priv, 0x400094, 0x00000005); - nv_wr32(priv, 0x400B80, 0x45eae20e); - nv_wr32(priv, 0x400B84, 0x24000000); - nv_wr32(priv, 0x400098, 0x00000040); - nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00E00038); - nv_wr32(priv, NV10_PGRAPH_RDI_DATA , 0x00000030); - nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00E10038); - nv_wr32(priv, NV10_PGRAPH_RDI_DATA , 0x00000030); + nvkm_wr32(device, 0x400880, 0x0008c7df); + nvkm_wr32(device, 0x400094, 0x00000005); + nvkm_wr32(device, 0x400B80, 0x45eae20e); + nvkm_wr32(device, 0x400B84, 0x24000000); + nvkm_wr32(device, 0x400098, 0x00000040); + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00E00038); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA , 0x00000030); + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00E10038); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA , 0x00000030); } - /* Turn all the tiling regions off. */ - for (i = 0; i < pfb->tile.regions; i++) - engine->tile_prog(engine, i); + nvkm_wr32(device, 0x4009a0, nvkm_rd32(device, 0x100324)); + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00EA000C); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA, nvkm_rd32(device, 0x100324)); - nv_wr32(priv, 0x4009a0, nv_rd32(priv, 0x100324)); - nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00EA000C); - nv_wr32(priv, NV10_PGRAPH_RDI_DATA, nv_rd32(priv, 0x100324)); + nvkm_wr32(device, NV10_PGRAPH_CTX_CONTROL, 0x10000100); + nvkm_wr32(device, NV10_PGRAPH_STATE , 0xFFFFFFFF); - nv_wr32(priv, NV10_PGRAPH_CTX_CONTROL, 0x10000100); - nv_wr32(priv, NV10_PGRAPH_STATE , 0xFFFFFFFF); - - tmp = nv_rd32(priv, NV10_PGRAPH_SURFACE) & 0x0007ff00; - nv_wr32(priv, NV10_PGRAPH_SURFACE, tmp); - tmp = nv_rd32(priv, NV10_PGRAPH_SURFACE) | 0x00020100; - nv_wr32(priv, NV10_PGRAPH_SURFACE, tmp); + tmp = nvkm_rd32(device, NV10_PGRAPH_SURFACE) & 0x0007ff00; + nvkm_wr32(device, NV10_PGRAPH_SURFACE, tmp); + tmp = nvkm_rd32(device, NV10_PGRAPH_SURFACE) | 0x00020100; + nvkm_wr32(device, NV10_PGRAPH_SURFACE, tmp); /* begin RAM config */ - vramsz = nv_device_resource_len(nv_device(priv), 0) - 1; - nv_wr32(priv, 0x4009A4, nv_rd32(priv, 0x100200)); - nv_wr32(priv, 0x4009A8, nv_rd32(priv, 0x100204)); - nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00EA0000); - nv_wr32(priv, NV10_PGRAPH_RDI_DATA , nv_rd32(priv, 0x100200)); - nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00EA0004); - nv_wr32(priv, NV10_PGRAPH_RDI_DATA , nv_rd32(priv, 0x100204)); - nv_wr32(priv, 0x400820, 0); - nv_wr32(priv, 0x400824, 0); - nv_wr32(priv, 0x400864, vramsz - 1); - nv_wr32(priv, 0x400868, vramsz - 1); + vramsz = device->func->resource_size(device, 1) - 1; + nvkm_wr32(device, 0x4009A4, nvkm_rd32(device, 0x100200)); + nvkm_wr32(device, 0x4009A8, nvkm_rd32(device, 0x100204)); + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00EA0000); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA , nvkm_rd32(device, 0x100200)); + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00EA0004); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA , nvkm_rd32(device, 0x100204)); + nvkm_wr32(device, 0x400820, 0); + nvkm_wr32(device, 0x400824, 0); + nvkm_wr32(device, 0x400864, vramsz - 1); + nvkm_wr32(device, 0x400868, vramsz - 1); /* interesting.. the below overwrites some of the tile setup above.. */ - nv_wr32(priv, 0x400B20, 0x00000000); - nv_wr32(priv, 0x400B04, 0xFFFFFFFF); + nvkm_wr32(device, 0x400B20, 0x00000000); + nvkm_wr32(device, 0x400B04, 0xFFFFFFFF); - nv_wr32(priv, NV03_PGRAPH_ABS_UCLIP_XMIN, 0); - nv_wr32(priv, NV03_PGRAPH_ABS_UCLIP_YMIN, 0); - nv_wr32(priv, NV03_PGRAPH_ABS_UCLIP_XMAX, 0x7fff); - nv_wr32(priv, NV03_PGRAPH_ABS_UCLIP_YMAX, 0x7fff); + nvkm_wr32(device, NV03_PGRAPH_ABS_UCLIP_XMIN, 0); + nvkm_wr32(device, NV03_PGRAPH_ABS_UCLIP_YMIN, 0); + nvkm_wr32(device, NV03_PGRAPH_ABS_UCLIP_XMAX, 0x7fff); + nvkm_wr32(device, NV03_PGRAPH_ABS_UCLIP_YMAX, 0x7fff); return 0; } -struct nvkm_oclass -nv20_gr_oclass = { - .handle = NV_ENGINE(GR, 0x20), - .ofuncs = &(struct nvkm_ofuncs) { - .ctor = nv20_gr_ctor, - .dtor = nv20_gr_dtor, - .init = nv20_gr_init, - .fini = _nvkm_gr_fini, - }, +void * +nv20_gr_dtor(struct nvkm_gr *base) +{ + struct nv20_gr *gr = nv20_gr(base); + nvkm_memory_del(&gr->ctxtab); + return gr; +} + +int +nv20_gr_new_(const struct nvkm_gr_func *func, struct nvkm_device *device, + int index, struct nvkm_gr **pgr) +{ + struct nv20_gr *gr; + + if (!(gr = kzalloc(sizeof(*gr), GFP_KERNEL))) + return -ENOMEM; + *pgr = &gr->base; + + return nvkm_gr_ctor(func, device, index, 0x00001000, true, &gr->base); +} + +static const struct nvkm_gr_func +nv20_gr = { + .dtor = nv20_gr_dtor, + .oneinit = nv20_gr_oneinit, + .init = nv20_gr_init, + .intr = nv20_gr_intr, + .tile = nv20_gr_tile, + .chan_new = nv20_gr_chan_new, + .sclass = { + { -1, -1, 0x0012, &nv04_gr_object }, /* beta1 */ + { -1, -1, 0x0019, &nv04_gr_object }, /* clip */ + { -1, -1, 0x0030, &nv04_gr_object }, /* null */ + { -1, -1, 0x0039, &nv04_gr_object }, /* m2mf */ + { -1, -1, 0x0043, &nv04_gr_object }, /* rop */ + { -1, -1, 0x0044, &nv04_gr_object }, /* patt */ + { -1, -1, 0x004a, &nv04_gr_object }, /* gdi */ + { -1, -1, 0x0062, &nv04_gr_object }, /* surf2d */ + { -1, -1, 0x0072, &nv04_gr_object }, /* beta4 */ + { -1, -1, 0x0089, &nv04_gr_object }, /* sifm */ + { -1, -1, 0x008a, &nv04_gr_object }, /* ifc */ + { -1, -1, 0x0096, &nv04_gr_object }, /* celcius */ + { -1, -1, 0x0097, &nv04_gr_object }, /* kelvin */ + { -1, -1, 0x009e, &nv04_gr_object }, /* swzsurf */ + { -1, -1, 0x009f, &nv04_gr_object }, /* imageblit */ + {} + } }; + +int +nv20_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return nv20_gr_new_(&nv20_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.h b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.h index ac4dc048f..cdf4501e3 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.h +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv20.h @@ -1,26 +1,33 @@ #ifndef __NV20_GR_H__ #define __NV20_GR_H__ -#include <engine/gr.h> +#define nv20_gr(p) container_of((p), struct nv20_gr, base) +#include "priv.h" -struct nv20_gr_priv { +struct nv20_gr { struct nvkm_gr base; - struct nvkm_gpuobj *ctxtab; + struct nvkm_memory *ctxtab; }; -struct nv20_gr_chan { - struct nvkm_gr_chan base; - int chid; -}; +int nv20_gr_new_(const struct nvkm_gr_func *, struct nvkm_device *, + int, struct nvkm_gr **); +void *nv20_gr_dtor(struct nvkm_gr *); +int nv20_gr_oneinit(struct nvkm_gr *); +int nv20_gr_init(struct nvkm_gr *); +void nv20_gr_intr(struct nvkm_gr *); +void nv20_gr_tile(struct nvkm_gr *, int, struct nvkm_fb_tile *); -extern struct nvkm_oclass nv25_gr_sclass[]; -int nv20_gr_context_init(struct nvkm_object *); -int nv20_gr_context_fini(struct nvkm_object *, bool); +int nv30_gr_init(struct nvkm_gr *); -void nv20_gr_tile_prog(struct nvkm_engine *, int); -void nv20_gr_intr(struct nvkm_subdev *); +#define nv20_gr_chan(p) container_of((p), struct nv20_gr_chan, object) -void nv20_gr_dtor(struct nvkm_object *); -int nv20_gr_init(struct nvkm_object *); +struct nv20_gr_chan { + struct nvkm_object object; + struct nv20_gr *gr; + int chid; + struct nvkm_memory *inst; +}; -int nv30_gr_init(struct nvkm_object *); +void *nv20_gr_chan_dtor(struct nvkm_object *); +int nv20_gr_chan_init(struct nvkm_object *); +int nv20_gr_chan_fini(struct nvkm_object *, bool); #endif diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv25.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv25.c index bc362519c..6c4a00819 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv25.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv25.c @@ -1,158 +1,134 @@ #include "nv20.h" #include "regs.h" +#include <core/gpuobj.h> #include <engine/fifo.h> +#include <engine/fifo/chan.h> /******************************************************************************* - * Graphics object classes + * PGRAPH context ******************************************************************************/ -struct nvkm_oclass -nv25_gr_sclass[] = { - { 0x0012, &nv04_gr_ofuncs, NULL }, /* beta1 */ - { 0x0019, &nv04_gr_ofuncs, NULL }, /* clip */ - { 0x0030, &nv04_gr_ofuncs, NULL }, /* null */ - { 0x0039, &nv04_gr_ofuncs, NULL }, /* m2mf */ - { 0x0043, &nv04_gr_ofuncs, NULL }, /* rop */ - { 0x0044, &nv04_gr_ofuncs, NULL }, /* patt */ - { 0x004a, &nv04_gr_ofuncs, NULL }, /* gdi */ - { 0x0062, &nv04_gr_ofuncs, NULL }, /* surf2d */ - { 0x0072, &nv04_gr_ofuncs, NULL }, /* beta4 */ - { 0x0089, &nv04_gr_ofuncs, NULL }, /* sifm */ - { 0x008a, &nv04_gr_ofuncs, NULL }, /* ifc */ - { 0x0096, &nv04_gr_ofuncs, NULL }, /* celcius */ - { 0x009e, &nv04_gr_ofuncs, NULL }, /* swzsurf */ - { 0x009f, &nv04_gr_ofuncs, NULL }, /* imageblit */ - { 0x0597, &nv04_gr_ofuncs, NULL }, /* kelvin */ - {}, +static const struct nvkm_object_func +nv25_gr_chan = { + .dtor = nv20_gr_chan_dtor, + .init = nv20_gr_chan_init, + .fini = nv20_gr_chan_fini, }; -/******************************************************************************* - * PGRAPH context - ******************************************************************************/ - static int -nv25_gr_context_ctor(struct nvkm_object *parent, struct nvkm_object *engine, - struct nvkm_oclass *oclass, void *data, u32 size, - struct nvkm_object **pobject) +nv25_gr_chan_new(struct nvkm_gr *base, struct nvkm_fifo_chan *fifoch, + const struct nvkm_oclass *oclass, struct nvkm_object **pobject) { + struct nv20_gr *gr = nv20_gr(base); struct nv20_gr_chan *chan; int ret, i; - ret = nvkm_gr_context_create(parent, engine, oclass, NULL, 0x3724, - 16, NVOBJ_FLAG_ZERO_ALLOC, &chan); - *pobject = nv_object(chan); + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&nv25_gr_chan, oclass, &chan->object); + chan->gr = gr; + chan->chid = fifoch->chid; + *pobject = &chan->object; + + ret = nvkm_memory_new(gr->base.engine.subdev.device, + NVKM_MEM_TARGET_INST, 0x3724, 16, true, + &chan->inst); if (ret) return ret; - chan->chid = nvkm_fifo_chan(parent)->chid; - - nv_wo32(chan, 0x0028, 0x00000001 | (chan->chid << 24)); - nv_wo32(chan, 0x035c, 0xffff0000); - nv_wo32(chan, 0x03c0, 0x0fff0000); - nv_wo32(chan, 0x03c4, 0x0fff0000); - nv_wo32(chan, 0x049c, 0x00000101); - nv_wo32(chan, 0x04b0, 0x00000111); - nv_wo32(chan, 0x04c8, 0x00000080); - nv_wo32(chan, 0x04cc, 0xffff0000); - nv_wo32(chan, 0x04d0, 0x00000001); - nv_wo32(chan, 0x04e4, 0x44400000); - nv_wo32(chan, 0x04fc, 0x4b800000); + nvkm_kmap(chan->inst); + nvkm_wo32(chan->inst, 0x0028, 0x00000001 | (chan->chid << 24)); + nvkm_wo32(chan->inst, 0x035c, 0xffff0000); + nvkm_wo32(chan->inst, 0x03c0, 0x0fff0000); + nvkm_wo32(chan->inst, 0x03c4, 0x0fff0000); + nvkm_wo32(chan->inst, 0x049c, 0x00000101); + nvkm_wo32(chan->inst, 0x04b0, 0x00000111); + nvkm_wo32(chan->inst, 0x04c8, 0x00000080); + nvkm_wo32(chan->inst, 0x04cc, 0xffff0000); + nvkm_wo32(chan->inst, 0x04d0, 0x00000001); + nvkm_wo32(chan->inst, 0x04e4, 0x44400000); + nvkm_wo32(chan->inst, 0x04fc, 0x4b800000); for (i = 0x0510; i <= 0x051c; i += 4) - nv_wo32(chan, i, 0x00030303); + nvkm_wo32(chan->inst, i, 0x00030303); for (i = 0x0530; i <= 0x053c; i += 4) - nv_wo32(chan, i, 0x00080000); + nvkm_wo32(chan->inst, i, 0x00080000); for (i = 0x0548; i <= 0x0554; i += 4) - nv_wo32(chan, i, 0x01012000); + nvkm_wo32(chan->inst, i, 0x01012000); for (i = 0x0558; i <= 0x0564; i += 4) - nv_wo32(chan, i, 0x000105b8); + nvkm_wo32(chan->inst, i, 0x000105b8); for (i = 0x0568; i <= 0x0574; i += 4) - nv_wo32(chan, i, 0x00080008); + nvkm_wo32(chan->inst, i, 0x00080008); for (i = 0x0598; i <= 0x05d4; i += 4) - nv_wo32(chan, i, 0x07ff0000); - nv_wo32(chan, 0x05e0, 0x4b7fffff); - nv_wo32(chan, 0x0620, 0x00000080); - nv_wo32(chan, 0x0624, 0x30201000); - nv_wo32(chan, 0x0628, 0x70605040); - nv_wo32(chan, 0x062c, 0xb0a09080); - nv_wo32(chan, 0x0630, 0xf0e0d0c0); - nv_wo32(chan, 0x0664, 0x00000001); - nv_wo32(chan, 0x066c, 0x00004000); - nv_wo32(chan, 0x0678, 0x00000001); - nv_wo32(chan, 0x0680, 0x00040000); - nv_wo32(chan, 0x0684, 0x00010000); + nvkm_wo32(chan->inst, i, 0x07ff0000); + nvkm_wo32(chan->inst, 0x05e0, 0x4b7fffff); + nvkm_wo32(chan->inst, 0x0620, 0x00000080); + nvkm_wo32(chan->inst, 0x0624, 0x30201000); + nvkm_wo32(chan->inst, 0x0628, 0x70605040); + nvkm_wo32(chan->inst, 0x062c, 0xb0a09080); + nvkm_wo32(chan->inst, 0x0630, 0xf0e0d0c0); + nvkm_wo32(chan->inst, 0x0664, 0x00000001); + nvkm_wo32(chan->inst, 0x066c, 0x00004000); + nvkm_wo32(chan->inst, 0x0678, 0x00000001); + nvkm_wo32(chan->inst, 0x0680, 0x00040000); + nvkm_wo32(chan->inst, 0x0684, 0x00010000); for (i = 0x1b04; i <= 0x2374; i += 16) { - nv_wo32(chan, (i + 0), 0x10700ff9); - nv_wo32(chan, (i + 4), 0x0436086c); - nv_wo32(chan, (i + 8), 0x000c001b); + nvkm_wo32(chan->inst, (i + 0), 0x10700ff9); + nvkm_wo32(chan->inst, (i + 4), 0x0436086c); + nvkm_wo32(chan->inst, (i + 8), 0x000c001b); } - nv_wo32(chan, 0x2704, 0x3f800000); - nv_wo32(chan, 0x2718, 0x3f800000); - nv_wo32(chan, 0x2744, 0x40000000); - nv_wo32(chan, 0x2748, 0x3f800000); - nv_wo32(chan, 0x274c, 0x3f000000); - nv_wo32(chan, 0x2754, 0x40000000); - nv_wo32(chan, 0x2758, 0x3f800000); - nv_wo32(chan, 0x2760, 0xbf800000); - nv_wo32(chan, 0x2768, 0xbf800000); - nv_wo32(chan, 0x308c, 0x000fe000); - nv_wo32(chan, 0x3108, 0x000003f8); - nv_wo32(chan, 0x3468, 0x002fe000); + nvkm_wo32(chan->inst, 0x2704, 0x3f800000); + nvkm_wo32(chan->inst, 0x2718, 0x3f800000); + nvkm_wo32(chan->inst, 0x2744, 0x40000000); + nvkm_wo32(chan->inst, 0x2748, 0x3f800000); + nvkm_wo32(chan->inst, 0x274c, 0x3f000000); + nvkm_wo32(chan->inst, 0x2754, 0x40000000); + nvkm_wo32(chan->inst, 0x2758, 0x3f800000); + nvkm_wo32(chan->inst, 0x2760, 0xbf800000); + nvkm_wo32(chan->inst, 0x2768, 0xbf800000); + nvkm_wo32(chan->inst, 0x308c, 0x000fe000); + nvkm_wo32(chan->inst, 0x3108, 0x000003f8); + nvkm_wo32(chan->inst, 0x3468, 0x002fe000); for (i = 0x3484; i <= 0x34a0; i += 4) - nv_wo32(chan, i, 0x001c527c); + nvkm_wo32(chan->inst, i, 0x001c527c); + nvkm_done(chan->inst); return 0; } -static struct nvkm_oclass -nv25_gr_cclass = { - .handle = NV_ENGCTX(GR, 0x25), - .ofuncs = &(struct nvkm_ofuncs) { - .ctor = nv25_gr_context_ctor, - .dtor = _nvkm_gr_context_dtor, - .init = nv20_gr_context_init, - .fini = nv20_gr_context_fini, - .rd32 = _nvkm_gr_context_rd32, - .wr32 = _nvkm_gr_context_wr32, - }, -}; - /******************************************************************************* * PGRAPH engine/subdev functions ******************************************************************************/ -static int -nv25_gr_ctor(struct nvkm_object *parent, struct nvkm_object *engine, - struct nvkm_oclass *oclass, void *data, u32 size, - struct nvkm_object **pobject) -{ - struct nv20_gr_priv *priv; - int ret; - - ret = nvkm_gr_create(parent, engine, oclass, true, &priv); - *pobject = nv_object(priv); - if (ret) - return ret; - - ret = nvkm_gpuobj_new(nv_object(priv), NULL, 32 * 4, 16, - NVOBJ_FLAG_ZERO_ALLOC, &priv->ctxtab); - if (ret) - return ret; +static const struct nvkm_gr_func +nv25_gr = { + .dtor = nv20_gr_dtor, + .oneinit = nv20_gr_oneinit, + .init = nv20_gr_init, + .intr = nv20_gr_intr, + .tile = nv20_gr_tile, + .chan_new = nv25_gr_chan_new, + .sclass = { + { -1, -1, 0x0012, &nv04_gr_object }, /* beta1 */ + { -1, -1, 0x0019, &nv04_gr_object }, /* clip */ + { -1, -1, 0x0030, &nv04_gr_object }, /* null */ + { -1, -1, 0x0039, &nv04_gr_object }, /* m2mf */ + { -1, -1, 0x0043, &nv04_gr_object }, /* rop */ + { -1, -1, 0x0044, &nv04_gr_object }, /* patt */ + { -1, -1, 0x004a, &nv04_gr_object }, /* gdi */ + { -1, -1, 0x0062, &nv04_gr_object }, /* surf2d */ + { -1, -1, 0x0072, &nv04_gr_object }, /* beta4 */ + { -1, -1, 0x0089, &nv04_gr_object }, /* sifm */ + { -1, -1, 0x008a, &nv04_gr_object }, /* ifc */ + { -1, -1, 0x0096, &nv04_gr_object }, /* celcius */ + { -1, -1, 0x009e, &nv04_gr_object }, /* swzsurf */ + { -1, -1, 0x009f, &nv04_gr_object }, /* imageblit */ + { -1, -1, 0x0597, &nv04_gr_object }, /* kelvin */ + {} + } +}; - nv_subdev(priv)->unit = 0x00001000; - nv_subdev(priv)->intr = nv20_gr_intr; - nv_engine(priv)->cclass = &nv25_gr_cclass; - nv_engine(priv)->sclass = nv25_gr_sclass; - nv_engine(priv)->tile_prog = nv20_gr_tile_prog; - return 0; +int +nv25_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return nv20_gr_new_(&nv25_gr, device, index, pgr); } - -struct nvkm_oclass -nv25_gr_oclass = { - .handle = NV_ENGINE(GR, 0x25), - .ofuncs = &(struct nvkm_ofuncs) { - .ctor = nv25_gr_ctor, - .dtor = nv20_gr_dtor, - .init = nv20_gr_init, - .fini = _nvkm_gr_fini, - }, -}; diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv2a.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv2a.c index 22a5096e2..3cad26dbc 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv2a.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv2a.c @@ -1,125 +1,125 @@ #include "nv20.h" #include "regs.h" +#include <core/gpuobj.h> #include <engine/fifo.h> +#include <engine/fifo/chan.h> /******************************************************************************* * PGRAPH context ******************************************************************************/ +static const struct nvkm_object_func +nv2a_gr_chan = { + .dtor = nv20_gr_chan_dtor, + .init = nv20_gr_chan_init, + .fini = nv20_gr_chan_fini, +}; + static int -nv2a_gr_context_ctor(struct nvkm_object *parent, struct nvkm_object *engine, - struct nvkm_oclass *oclass, void *data, u32 size, - struct nvkm_object **pobject) +nv2a_gr_chan_new(struct nvkm_gr *base, struct nvkm_fifo_chan *fifoch, + const struct nvkm_oclass *oclass, struct nvkm_object **pobject) { + struct nv20_gr *gr = nv20_gr(base); struct nv20_gr_chan *chan; int ret, i; - ret = nvkm_gr_context_create(parent, engine, oclass, NULL, 0x36b0, - 16, NVOBJ_FLAG_ZERO_ALLOC, &chan); - *pobject = nv_object(chan); + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&nv2a_gr_chan, oclass, &chan->object); + chan->gr = gr; + chan->chid = fifoch->chid; + *pobject = &chan->object; + + ret = nvkm_memory_new(gr->base.engine.subdev.device, + NVKM_MEM_TARGET_INST, 0x36b0, 16, true, + &chan->inst); if (ret) return ret; - chan->chid = nvkm_fifo_chan(parent)->chid; - - nv_wo32(chan, 0x0000, 0x00000001 | (chan->chid << 24)); - nv_wo32(chan, 0x033c, 0xffff0000); - nv_wo32(chan, 0x03a0, 0x0fff0000); - nv_wo32(chan, 0x03a4, 0x0fff0000); - nv_wo32(chan, 0x047c, 0x00000101); - nv_wo32(chan, 0x0490, 0x00000111); - nv_wo32(chan, 0x04a8, 0x44400000); + nvkm_kmap(chan->inst); + nvkm_wo32(chan->inst, 0x0000, 0x00000001 | (chan->chid << 24)); + nvkm_wo32(chan->inst, 0x033c, 0xffff0000); + nvkm_wo32(chan->inst, 0x03a0, 0x0fff0000); + nvkm_wo32(chan->inst, 0x03a4, 0x0fff0000); + nvkm_wo32(chan->inst, 0x047c, 0x00000101); + nvkm_wo32(chan->inst, 0x0490, 0x00000111); + nvkm_wo32(chan->inst, 0x04a8, 0x44400000); for (i = 0x04d4; i <= 0x04e0; i += 4) - nv_wo32(chan, i, 0x00030303); + nvkm_wo32(chan->inst, i, 0x00030303); for (i = 0x04f4; i <= 0x0500; i += 4) - nv_wo32(chan, i, 0x00080000); + nvkm_wo32(chan->inst, i, 0x00080000); for (i = 0x050c; i <= 0x0518; i += 4) - nv_wo32(chan, i, 0x01012000); + nvkm_wo32(chan->inst, i, 0x01012000); for (i = 0x051c; i <= 0x0528; i += 4) - nv_wo32(chan, i, 0x000105b8); + nvkm_wo32(chan->inst, i, 0x000105b8); for (i = 0x052c; i <= 0x0538; i += 4) - nv_wo32(chan, i, 0x00080008); + nvkm_wo32(chan->inst, i, 0x00080008); for (i = 0x055c; i <= 0x0598; i += 4) - nv_wo32(chan, i, 0x07ff0000); - nv_wo32(chan, 0x05a4, 0x4b7fffff); - nv_wo32(chan, 0x05fc, 0x00000001); - nv_wo32(chan, 0x0604, 0x00004000); - nv_wo32(chan, 0x0610, 0x00000001); - nv_wo32(chan, 0x0618, 0x00040000); - nv_wo32(chan, 0x061c, 0x00010000); + nvkm_wo32(chan->inst, i, 0x07ff0000); + nvkm_wo32(chan->inst, 0x05a4, 0x4b7fffff); + nvkm_wo32(chan->inst, 0x05fc, 0x00000001); + nvkm_wo32(chan->inst, 0x0604, 0x00004000); + nvkm_wo32(chan->inst, 0x0610, 0x00000001); + nvkm_wo32(chan->inst, 0x0618, 0x00040000); + nvkm_wo32(chan->inst, 0x061c, 0x00010000); for (i = 0x1a9c; i <= 0x22fc; i += 16) { /*XXX: check!! */ - nv_wo32(chan, (i + 0), 0x10700ff9); - nv_wo32(chan, (i + 4), 0x0436086c); - nv_wo32(chan, (i + 8), 0x000c001b); + nvkm_wo32(chan->inst, (i + 0), 0x10700ff9); + nvkm_wo32(chan->inst, (i + 4), 0x0436086c); + nvkm_wo32(chan->inst, (i + 8), 0x000c001b); } - nv_wo32(chan, 0x269c, 0x3f800000); - nv_wo32(chan, 0x26b0, 0x3f800000); - nv_wo32(chan, 0x26dc, 0x40000000); - nv_wo32(chan, 0x26e0, 0x3f800000); - nv_wo32(chan, 0x26e4, 0x3f000000); - nv_wo32(chan, 0x26ec, 0x40000000); - nv_wo32(chan, 0x26f0, 0x3f800000); - nv_wo32(chan, 0x26f8, 0xbf800000); - nv_wo32(chan, 0x2700, 0xbf800000); - nv_wo32(chan, 0x3024, 0x000fe000); - nv_wo32(chan, 0x30a0, 0x000003f8); - nv_wo32(chan, 0x33fc, 0x002fe000); + nvkm_wo32(chan->inst, 0x269c, 0x3f800000); + nvkm_wo32(chan->inst, 0x26b0, 0x3f800000); + nvkm_wo32(chan->inst, 0x26dc, 0x40000000); + nvkm_wo32(chan->inst, 0x26e0, 0x3f800000); + nvkm_wo32(chan->inst, 0x26e4, 0x3f000000); + nvkm_wo32(chan->inst, 0x26ec, 0x40000000); + nvkm_wo32(chan->inst, 0x26f0, 0x3f800000); + nvkm_wo32(chan->inst, 0x26f8, 0xbf800000); + nvkm_wo32(chan->inst, 0x2700, 0xbf800000); + nvkm_wo32(chan->inst, 0x3024, 0x000fe000); + nvkm_wo32(chan->inst, 0x30a0, 0x000003f8); + nvkm_wo32(chan->inst, 0x33fc, 0x002fe000); for (i = 0x341c; i <= 0x3438; i += 4) - nv_wo32(chan, i, 0x001c527c); + nvkm_wo32(chan->inst, i, 0x001c527c); + nvkm_done(chan->inst); return 0; } -static struct nvkm_oclass -nv2a_gr_cclass = { - .handle = NV_ENGCTX(GR, 0x2a), - .ofuncs = &(struct nvkm_ofuncs) { - .ctor = nv2a_gr_context_ctor, - .dtor = _nvkm_gr_context_dtor, - .init = nv20_gr_context_init, - .fini = nv20_gr_context_fini, - .rd32 = _nvkm_gr_context_rd32, - .wr32 = _nvkm_gr_context_wr32, - }, -}; - /******************************************************************************* * PGRAPH engine/subdev functions ******************************************************************************/ -static int -nv2a_gr_ctor(struct nvkm_object *parent, struct nvkm_object *engine, - struct nvkm_oclass *oclass, void *data, u32 size, - struct nvkm_object **pobject) -{ - struct nv20_gr_priv *priv; - int ret; - - ret = nvkm_gr_create(parent, engine, oclass, true, &priv); - *pobject = nv_object(priv); - if (ret) - return ret; - - ret = nvkm_gpuobj_new(nv_object(priv), NULL, 32 * 4, 16, - NVOBJ_FLAG_ZERO_ALLOC, &priv->ctxtab); - if (ret) - return ret; +static const struct nvkm_gr_func +nv2a_gr = { + .dtor = nv20_gr_dtor, + .oneinit = nv20_gr_oneinit, + .init = nv20_gr_init, + .intr = nv20_gr_intr, + .tile = nv20_gr_tile, + .chan_new = nv2a_gr_chan_new, + .sclass = { + { -1, -1, 0x0012, &nv04_gr_object }, /* beta1 */ + { -1, -1, 0x0019, &nv04_gr_object }, /* clip */ + { -1, -1, 0x0030, &nv04_gr_object }, /* null */ + { -1, -1, 0x0039, &nv04_gr_object }, /* m2mf */ + { -1, -1, 0x0043, &nv04_gr_object }, /* rop */ + { -1, -1, 0x0044, &nv04_gr_object }, /* patt */ + { -1, -1, 0x004a, &nv04_gr_object }, /* gdi */ + { -1, -1, 0x0062, &nv04_gr_object }, /* surf2d */ + { -1, -1, 0x0072, &nv04_gr_object }, /* beta4 */ + { -1, -1, 0x0089, &nv04_gr_object }, /* sifm */ + { -1, -1, 0x008a, &nv04_gr_object }, /* ifc */ + { -1, -1, 0x0096, &nv04_gr_object }, /* celcius */ + { -1, -1, 0x009e, &nv04_gr_object }, /* swzsurf */ + { -1, -1, 0x009f, &nv04_gr_object }, /* imageblit */ + { -1, -1, 0x0597, &nv04_gr_object }, /* kelvin */ + {} + } +}; - nv_subdev(priv)->unit = 0x00001000; - nv_subdev(priv)->intr = nv20_gr_intr; - nv_engine(priv)->cclass = &nv2a_gr_cclass; - nv_engine(priv)->sclass = nv25_gr_sclass; - nv_engine(priv)->tile_prog = nv20_gr_tile_prog; - return 0; +int +nv2a_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return nv20_gr_new_(&nv2a_gr, device, index, pgr); } - -struct nvkm_oclass -nv2a_gr_oclass = { - .handle = NV_ENGINE(GR, 0x2a), - .ofuncs = &(struct nvkm_ofuncs) { - .ctor = nv2a_gr_ctor, - .dtor = nv20_gr_dtor, - .init = nv20_gr_init, - .fini = _nvkm_gr_fini, - }, -}; diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv30.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv30.c index dcc84eb54..69de8c625 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv30.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv30.c @@ -1,231 +1,198 @@ #include "nv20.h" #include "regs.h" -#include <core/device.h> +#include <core/gpuobj.h> #include <engine/fifo.h> +#include <engine/fifo/chan.h> #include <subdev/fb.h> /******************************************************************************* - * Graphics object classes + * PGRAPH context ******************************************************************************/ -static struct nvkm_oclass -nv30_gr_sclass[] = { - { 0x0012, &nv04_gr_ofuncs, NULL }, /* beta1 */ - { 0x0019, &nv04_gr_ofuncs, NULL }, /* clip */ - { 0x0030, &nv04_gr_ofuncs, NULL }, /* null */ - { 0x0039, &nv04_gr_ofuncs, NULL }, /* m2mf */ - { 0x0043, &nv04_gr_ofuncs, NULL }, /* rop */ - { 0x0044, &nv04_gr_ofuncs, NULL }, /* patt */ - { 0x004a, &nv04_gr_ofuncs, NULL }, /* gdi */ - { 0x0062, &nv04_gr_ofuncs, NULL }, /* surf2d */ - { 0x0072, &nv04_gr_ofuncs, NULL }, /* beta4 */ - { 0x0089, &nv04_gr_ofuncs, NULL }, /* sifm */ - { 0x008a, &nv04_gr_ofuncs, NULL }, /* ifc */ - { 0x009f, &nv04_gr_ofuncs, NULL }, /* imageblit */ - { 0x0362, &nv04_gr_ofuncs, NULL }, /* surf2d (nv30) */ - { 0x0389, &nv04_gr_ofuncs, NULL }, /* sifm (nv30) */ - { 0x038a, &nv04_gr_ofuncs, NULL }, /* ifc (nv30) */ - { 0x039e, &nv04_gr_ofuncs, NULL }, /* swzsurf (nv30) */ - { 0x0397, &nv04_gr_ofuncs, NULL }, /* rankine */ - {}, +static const struct nvkm_object_func +nv30_gr_chan = { + .dtor = nv20_gr_chan_dtor, + .init = nv20_gr_chan_init, + .fini = nv20_gr_chan_fini, }; -/******************************************************************************* - * PGRAPH context - ******************************************************************************/ - static int -nv30_gr_context_ctor(struct nvkm_object *parent, struct nvkm_object *engine, - struct nvkm_oclass *oclass, void *data, u32 size, - struct nvkm_object **pobject) +nv30_gr_chan_new(struct nvkm_gr *base, struct nvkm_fifo_chan *fifoch, + const struct nvkm_oclass *oclass, struct nvkm_object **pobject) { + struct nv20_gr *gr = nv20_gr(base); struct nv20_gr_chan *chan; int ret, i; - ret = nvkm_gr_context_create(parent, engine, oclass, NULL, 0x5f48, - 16, NVOBJ_FLAG_ZERO_ALLOC, &chan); - *pobject = nv_object(chan); + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&nv30_gr_chan, oclass, &chan->object); + chan->gr = gr; + chan->chid = fifoch->chid; + *pobject = &chan->object; + + ret = nvkm_memory_new(gr->base.engine.subdev.device, + NVKM_MEM_TARGET_INST, 0x5f48, 16, true, + &chan->inst); if (ret) return ret; - chan->chid = nvkm_fifo_chan(parent)->chid; - - nv_wo32(chan, 0x0028, 0x00000001 | (chan->chid << 24)); - nv_wo32(chan, 0x0410, 0x00000101); - nv_wo32(chan, 0x0424, 0x00000111); - nv_wo32(chan, 0x0428, 0x00000060); - nv_wo32(chan, 0x0444, 0x00000080); - nv_wo32(chan, 0x0448, 0xffff0000); - nv_wo32(chan, 0x044c, 0x00000001); - nv_wo32(chan, 0x0460, 0x44400000); - nv_wo32(chan, 0x048c, 0xffff0000); + nvkm_kmap(chan->inst); + nvkm_wo32(chan->inst, 0x0028, 0x00000001 | (chan->chid << 24)); + nvkm_wo32(chan->inst, 0x0410, 0x00000101); + nvkm_wo32(chan->inst, 0x0424, 0x00000111); + nvkm_wo32(chan->inst, 0x0428, 0x00000060); + nvkm_wo32(chan->inst, 0x0444, 0x00000080); + nvkm_wo32(chan->inst, 0x0448, 0xffff0000); + nvkm_wo32(chan->inst, 0x044c, 0x00000001); + nvkm_wo32(chan->inst, 0x0460, 0x44400000); + nvkm_wo32(chan->inst, 0x048c, 0xffff0000); for (i = 0x04e0; i < 0x04e8; i += 4) - nv_wo32(chan, i, 0x0fff0000); - nv_wo32(chan, 0x04ec, 0x00011100); + nvkm_wo32(chan->inst, i, 0x0fff0000); + nvkm_wo32(chan->inst, 0x04ec, 0x00011100); for (i = 0x0508; i < 0x0548; i += 4) - nv_wo32(chan, i, 0x07ff0000); - nv_wo32(chan, 0x0550, 0x4b7fffff); - nv_wo32(chan, 0x058c, 0x00000080); - nv_wo32(chan, 0x0590, 0x30201000); - nv_wo32(chan, 0x0594, 0x70605040); - nv_wo32(chan, 0x0598, 0xb8a89888); - nv_wo32(chan, 0x059c, 0xf8e8d8c8); - nv_wo32(chan, 0x05b0, 0xb0000000); + nvkm_wo32(chan->inst, i, 0x07ff0000); + nvkm_wo32(chan->inst, 0x0550, 0x4b7fffff); + nvkm_wo32(chan->inst, 0x058c, 0x00000080); + nvkm_wo32(chan->inst, 0x0590, 0x30201000); + nvkm_wo32(chan->inst, 0x0594, 0x70605040); + nvkm_wo32(chan->inst, 0x0598, 0xb8a89888); + nvkm_wo32(chan->inst, 0x059c, 0xf8e8d8c8); + nvkm_wo32(chan->inst, 0x05b0, 0xb0000000); for (i = 0x0600; i < 0x0640; i += 4) - nv_wo32(chan, i, 0x00010588); + nvkm_wo32(chan->inst, i, 0x00010588); for (i = 0x0640; i < 0x0680; i += 4) - nv_wo32(chan, i, 0x00030303); + nvkm_wo32(chan->inst, i, 0x00030303); for (i = 0x06c0; i < 0x0700; i += 4) - nv_wo32(chan, i, 0x0008aae4); + nvkm_wo32(chan->inst, i, 0x0008aae4); for (i = 0x0700; i < 0x0740; i += 4) - nv_wo32(chan, i, 0x01012000); + nvkm_wo32(chan->inst, i, 0x01012000); for (i = 0x0740; i < 0x0780; i += 4) - nv_wo32(chan, i, 0x00080008); - nv_wo32(chan, 0x085c, 0x00040000); - nv_wo32(chan, 0x0860, 0x00010000); + nvkm_wo32(chan->inst, i, 0x00080008); + nvkm_wo32(chan->inst, 0x085c, 0x00040000); + nvkm_wo32(chan->inst, 0x0860, 0x00010000); for (i = 0x0864; i < 0x0874; i += 4) - nv_wo32(chan, i, 0x00040004); + nvkm_wo32(chan->inst, i, 0x00040004); for (i = 0x1f18; i <= 0x3088 ; i += 16) { - nv_wo32(chan, i + 0, 0x10700ff9); - nv_wo32(chan, i + 1, 0x0436086c); - nv_wo32(chan, i + 2, 0x000c001b); + nvkm_wo32(chan->inst, i + 0, 0x10700ff9); + nvkm_wo32(chan->inst, i + 1, 0x0436086c); + nvkm_wo32(chan->inst, i + 2, 0x000c001b); } for (i = 0x30b8; i < 0x30c8; i += 4) - nv_wo32(chan, i, 0x0000ffff); - nv_wo32(chan, 0x344c, 0x3f800000); - nv_wo32(chan, 0x3808, 0x3f800000); - nv_wo32(chan, 0x381c, 0x3f800000); - nv_wo32(chan, 0x3848, 0x40000000); - nv_wo32(chan, 0x384c, 0x3f800000); - nv_wo32(chan, 0x3850, 0x3f000000); - nv_wo32(chan, 0x3858, 0x40000000); - nv_wo32(chan, 0x385c, 0x3f800000); - nv_wo32(chan, 0x3864, 0xbf800000); - nv_wo32(chan, 0x386c, 0xbf800000); + nvkm_wo32(chan->inst, i, 0x0000ffff); + nvkm_wo32(chan->inst, 0x344c, 0x3f800000); + nvkm_wo32(chan->inst, 0x3808, 0x3f800000); + nvkm_wo32(chan->inst, 0x381c, 0x3f800000); + nvkm_wo32(chan->inst, 0x3848, 0x40000000); + nvkm_wo32(chan->inst, 0x384c, 0x3f800000); + nvkm_wo32(chan->inst, 0x3850, 0x3f000000); + nvkm_wo32(chan->inst, 0x3858, 0x40000000); + nvkm_wo32(chan->inst, 0x385c, 0x3f800000); + nvkm_wo32(chan->inst, 0x3864, 0xbf800000); + nvkm_wo32(chan->inst, 0x386c, 0xbf800000); + nvkm_done(chan->inst); return 0; } -static struct nvkm_oclass -nv30_gr_cclass = { - .handle = NV_ENGCTX(GR, 0x30), - .ofuncs = &(struct nvkm_ofuncs) { - .ctor = nv30_gr_context_ctor, - .dtor = _nvkm_gr_context_dtor, - .init = nv20_gr_context_init, - .fini = nv20_gr_context_fini, - .rd32 = _nvkm_gr_context_rd32, - .wr32 = _nvkm_gr_context_wr32, - }, -}; - /******************************************************************************* * PGRAPH engine/subdev functions ******************************************************************************/ -static int -nv30_gr_ctor(struct nvkm_object *parent, struct nvkm_object *engine, - struct nvkm_oclass *oclass, void *data, u32 size, - struct nvkm_object **pobject) -{ - struct nv20_gr_priv *priv; - int ret; - - ret = nvkm_gr_create(parent, engine, oclass, true, &priv); - *pobject = nv_object(priv); - if (ret) - return ret; - - ret = nvkm_gpuobj_new(nv_object(priv), NULL, 32 * 4, 16, - NVOBJ_FLAG_ZERO_ALLOC, &priv->ctxtab); - if (ret) - return ret; - - nv_subdev(priv)->unit = 0x00001000; - nv_subdev(priv)->intr = nv20_gr_intr; - nv_engine(priv)->cclass = &nv30_gr_cclass; - nv_engine(priv)->sclass = nv30_gr_sclass; - nv_engine(priv)->tile_prog = nv20_gr_tile_prog; - return 0; -} - int -nv30_gr_init(struct nvkm_object *object) +nv30_gr_init(struct nvkm_gr *base) { - struct nvkm_engine *engine = nv_engine(object); - struct nv20_gr_priv *priv = (void *)engine; - struct nvkm_fb *pfb = nvkm_fb(object); - int ret, i; - - ret = nvkm_gr_init(&priv->base); - if (ret) - return ret; - - nv_wr32(priv, NV20_PGRAPH_CHANNEL_CTX_TABLE, priv->ctxtab->addr >> 4); - - nv_wr32(priv, NV03_PGRAPH_INTR , 0xFFFFFFFF); - nv_wr32(priv, NV03_PGRAPH_INTR_EN, 0xFFFFFFFF); - - nv_wr32(priv, NV04_PGRAPH_DEBUG_0, 0xFFFFFFFF); - nv_wr32(priv, NV04_PGRAPH_DEBUG_0, 0x00000000); - nv_wr32(priv, NV04_PGRAPH_DEBUG_1, 0x401287c0); - nv_wr32(priv, 0x400890, 0x01b463ff); - nv_wr32(priv, NV04_PGRAPH_DEBUG_3, 0xf2de0475); - nv_wr32(priv, NV10_PGRAPH_DEBUG_4, 0x00008000); - nv_wr32(priv, NV04_PGRAPH_LIMIT_VIOL_PIX, 0xf04bdff6); - nv_wr32(priv, 0x400B80, 0x1003d888); - nv_wr32(priv, 0x400B84, 0x0c000000); - nv_wr32(priv, 0x400098, 0x00000000); - nv_wr32(priv, 0x40009C, 0x0005ad00); - nv_wr32(priv, 0x400B88, 0x62ff00ff); /* suspiciously like PGRAPH_DEBUG_2 */ - nv_wr32(priv, 0x4000a0, 0x00000000); - nv_wr32(priv, 0x4000a4, 0x00000008); - nv_wr32(priv, 0x4008a8, 0xb784a400); - nv_wr32(priv, 0x400ba0, 0x002f8685); - nv_wr32(priv, 0x400ba4, 0x00231f3f); - nv_wr32(priv, 0x4008a4, 0x40000020); - - if (nv_device(priv)->chipset == 0x34) { - nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00EA0004); - nv_wr32(priv, NV10_PGRAPH_RDI_DATA , 0x00200201); - nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00EA0008); - nv_wr32(priv, NV10_PGRAPH_RDI_DATA , 0x00000008); - nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00EA0000); - nv_wr32(priv, NV10_PGRAPH_RDI_DATA , 0x00000032); - nv_wr32(priv, NV10_PGRAPH_RDI_INDEX, 0x00E00004); - nv_wr32(priv, NV10_PGRAPH_RDI_DATA , 0x00000002); + struct nv20_gr *gr = nv20_gr(base); + struct nvkm_device *device = gr->base.engine.subdev.device; + + nvkm_wr32(device, NV20_PGRAPH_CHANNEL_CTX_TABLE, + nvkm_memory_addr(gr->ctxtab) >> 4); + + nvkm_wr32(device, NV03_PGRAPH_INTR , 0xFFFFFFFF); + nvkm_wr32(device, NV03_PGRAPH_INTR_EN, 0xFFFFFFFF); + + nvkm_wr32(device, NV04_PGRAPH_DEBUG_0, 0xFFFFFFFF); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_0, 0x00000000); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_1, 0x401287c0); + nvkm_wr32(device, 0x400890, 0x01b463ff); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_3, 0xf2de0475); + nvkm_wr32(device, NV10_PGRAPH_DEBUG_4, 0x00008000); + nvkm_wr32(device, NV04_PGRAPH_LIMIT_VIOL_PIX, 0xf04bdff6); + nvkm_wr32(device, 0x400B80, 0x1003d888); + nvkm_wr32(device, 0x400B84, 0x0c000000); + nvkm_wr32(device, 0x400098, 0x00000000); + nvkm_wr32(device, 0x40009C, 0x0005ad00); + nvkm_wr32(device, 0x400B88, 0x62ff00ff); /* suspiciously like PGRAPH_DEBUG_2 */ + nvkm_wr32(device, 0x4000a0, 0x00000000); + nvkm_wr32(device, 0x4000a4, 0x00000008); + nvkm_wr32(device, 0x4008a8, 0xb784a400); + nvkm_wr32(device, 0x400ba0, 0x002f8685); + nvkm_wr32(device, 0x400ba4, 0x00231f3f); + nvkm_wr32(device, 0x4008a4, 0x40000020); + + if (device->chipset == 0x34) { + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00EA0004); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA , 0x00200201); + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00EA0008); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA , 0x00000008); + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00EA0000); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA , 0x00000032); + nvkm_wr32(device, NV10_PGRAPH_RDI_INDEX, 0x00E00004); + nvkm_wr32(device, NV10_PGRAPH_RDI_DATA , 0x00000002); } - nv_wr32(priv, 0x4000c0, 0x00000016); - - /* Turn all the tiling regions off. */ - for (i = 0; i < pfb->tile.regions; i++) - engine->tile_prog(engine, i); + nvkm_wr32(device, 0x4000c0, 0x00000016); - nv_wr32(priv, NV10_PGRAPH_CTX_CONTROL, 0x10000100); - nv_wr32(priv, NV10_PGRAPH_STATE , 0xFFFFFFFF); - nv_wr32(priv, 0x0040075c , 0x00000001); + nvkm_wr32(device, NV10_PGRAPH_CTX_CONTROL, 0x10000100); + nvkm_wr32(device, NV10_PGRAPH_STATE , 0xFFFFFFFF); + nvkm_wr32(device, 0x0040075c , 0x00000001); /* begin RAM config */ - /* vramsz = pci_resource_len(priv->dev->pdev, 0) - 1; */ - nv_wr32(priv, 0x4009A4, nv_rd32(priv, 0x100200)); - nv_wr32(priv, 0x4009A8, nv_rd32(priv, 0x100204)); - if (nv_device(priv)->chipset != 0x34) { - nv_wr32(priv, 0x400750, 0x00EA0000); - nv_wr32(priv, 0x400754, nv_rd32(priv, 0x100200)); - nv_wr32(priv, 0x400750, 0x00EA0004); - nv_wr32(priv, 0x400754, nv_rd32(priv, 0x100204)); + /* vramsz = pci_resource_len(gr->dev->pdev, 1) - 1; */ + nvkm_wr32(device, 0x4009A4, nvkm_rd32(device, 0x100200)); + nvkm_wr32(device, 0x4009A8, nvkm_rd32(device, 0x100204)); + if (device->chipset != 0x34) { + nvkm_wr32(device, 0x400750, 0x00EA0000); + nvkm_wr32(device, 0x400754, nvkm_rd32(device, 0x100200)); + nvkm_wr32(device, 0x400750, 0x00EA0004); + nvkm_wr32(device, 0x400754, nvkm_rd32(device, 0x100204)); } + return 0; } -struct nvkm_oclass -nv30_gr_oclass = { - .handle = NV_ENGINE(GR, 0x30), - .ofuncs = &(struct nvkm_ofuncs) { - .ctor = nv30_gr_ctor, - .dtor = nv20_gr_dtor, - .init = nv30_gr_init, - .fini = _nvkm_gr_fini, - }, +static const struct nvkm_gr_func +nv30_gr = { + .dtor = nv20_gr_dtor, + .oneinit = nv20_gr_oneinit, + .init = nv30_gr_init, + .intr = nv20_gr_intr, + .tile = nv20_gr_tile, + .chan_new = nv30_gr_chan_new, + .sclass = { + { -1, -1, 0x0012, &nv04_gr_object }, /* beta1 */ + { -1, -1, 0x0019, &nv04_gr_object }, /* clip */ + { -1, -1, 0x0030, &nv04_gr_object }, /* null */ + { -1, -1, 0x0039, &nv04_gr_object }, /* m2mf */ + { -1, -1, 0x0043, &nv04_gr_object }, /* rop */ + { -1, -1, 0x0044, &nv04_gr_object }, /* patt */ + { -1, -1, 0x004a, &nv04_gr_object }, /* gdi */ + { -1, -1, 0x0062, &nv04_gr_object }, /* surf2d */ + { -1, -1, 0x0072, &nv04_gr_object }, /* beta4 */ + { -1, -1, 0x0089, &nv04_gr_object }, /* sifm */ + { -1, -1, 0x008a, &nv04_gr_object }, /* ifc */ + { -1, -1, 0x009f, &nv04_gr_object }, /* imageblit */ + { -1, -1, 0x0362, &nv04_gr_object }, /* surf2d (nv30) */ + { -1, -1, 0x0389, &nv04_gr_object }, /* sifm (nv30) */ + { -1, -1, 0x038a, &nv04_gr_object }, /* ifc (nv30) */ + { -1, -1, 0x039e, &nv04_gr_object }, /* swzsurf (nv30) */ + { -1, -1, 0x0397, &nv04_gr_object }, /* rankine */ + {} + } }; + +int +nv30_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return nv20_gr_new_(&nv30_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv34.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv34.c index 985b7f330..2207dac23 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv34.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv34.c @@ -1,159 +1,135 @@ #include "nv20.h" #include "regs.h" +#include <core/gpuobj.h> #include <engine/fifo.h> +#include <engine/fifo/chan.h> /******************************************************************************* - * Graphics object classes + * PGRAPH context ******************************************************************************/ -static struct nvkm_oclass -nv34_gr_sclass[] = { - { 0x0012, &nv04_gr_ofuncs, NULL }, /* beta1 */ - { 0x0019, &nv04_gr_ofuncs, NULL }, /* clip */ - { 0x0030, &nv04_gr_ofuncs, NULL }, /* null */ - { 0x0039, &nv04_gr_ofuncs, NULL }, /* m2mf */ - { 0x0043, &nv04_gr_ofuncs, NULL }, /* rop */ - { 0x0044, &nv04_gr_ofuncs, NULL }, /* patt */ - { 0x004a, &nv04_gr_ofuncs, NULL }, /* gdi */ - { 0x0062, &nv04_gr_ofuncs, NULL }, /* surf2d */ - { 0x0072, &nv04_gr_ofuncs, NULL }, /* beta4 */ - { 0x0089, &nv04_gr_ofuncs, NULL }, /* sifm */ - { 0x008a, &nv04_gr_ofuncs, NULL }, /* ifc */ - { 0x009f, &nv04_gr_ofuncs, NULL }, /* imageblit */ - { 0x0362, &nv04_gr_ofuncs, NULL }, /* surf2d (nv30) */ - { 0x0389, &nv04_gr_ofuncs, NULL }, /* sifm (nv30) */ - { 0x038a, &nv04_gr_ofuncs, NULL }, /* ifc (nv30) */ - { 0x039e, &nv04_gr_ofuncs, NULL }, /* swzsurf (nv30) */ - { 0x0697, &nv04_gr_ofuncs, NULL }, /* rankine */ - {}, +static const struct nvkm_object_func +nv34_gr_chan = { + .dtor = nv20_gr_chan_dtor, + .init = nv20_gr_chan_init, + .fini = nv20_gr_chan_fini, }; -/******************************************************************************* - * PGRAPH context - ******************************************************************************/ - static int -nv34_gr_context_ctor(struct nvkm_object *parent, struct nvkm_object *engine, - struct nvkm_oclass *oclass, void *data, u32 size, - struct nvkm_object **pobject) +nv34_gr_chan_new(struct nvkm_gr *base, struct nvkm_fifo_chan *fifoch, + const struct nvkm_oclass *oclass, struct nvkm_object **pobject) { + struct nv20_gr *gr = nv20_gr(base); struct nv20_gr_chan *chan; int ret, i; - ret = nvkm_gr_context_create(parent, engine, oclass, NULL, 0x46dc, - 16, NVOBJ_FLAG_ZERO_ALLOC, &chan); - *pobject = nv_object(chan); + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&nv34_gr_chan, oclass, &chan->object); + chan->gr = gr; + chan->chid = fifoch->chid; + *pobject = &chan->object; + + ret = nvkm_memory_new(gr->base.engine.subdev.device, + NVKM_MEM_TARGET_INST, 0x46dc, 16, true, + &chan->inst); if (ret) return ret; - chan->chid = nvkm_fifo_chan(parent)->chid; - - nv_wo32(chan, 0x0028, 0x00000001 | (chan->chid << 24)); - nv_wo32(chan, 0x040c, 0x01000101); - nv_wo32(chan, 0x0420, 0x00000111); - nv_wo32(chan, 0x0424, 0x00000060); - nv_wo32(chan, 0x0440, 0x00000080); - nv_wo32(chan, 0x0444, 0xffff0000); - nv_wo32(chan, 0x0448, 0x00000001); - nv_wo32(chan, 0x045c, 0x44400000); - nv_wo32(chan, 0x0480, 0xffff0000); + nvkm_kmap(chan->inst); + nvkm_wo32(chan->inst, 0x0028, 0x00000001 | (chan->chid << 24)); + nvkm_wo32(chan->inst, 0x040c, 0x01000101); + nvkm_wo32(chan->inst, 0x0420, 0x00000111); + nvkm_wo32(chan->inst, 0x0424, 0x00000060); + nvkm_wo32(chan->inst, 0x0440, 0x00000080); + nvkm_wo32(chan->inst, 0x0444, 0xffff0000); + nvkm_wo32(chan->inst, 0x0448, 0x00000001); + nvkm_wo32(chan->inst, 0x045c, 0x44400000); + nvkm_wo32(chan->inst, 0x0480, 0xffff0000); for (i = 0x04d4; i < 0x04dc; i += 4) - nv_wo32(chan, i, 0x0fff0000); - nv_wo32(chan, 0x04e0, 0x00011100); + nvkm_wo32(chan->inst, i, 0x0fff0000); + nvkm_wo32(chan->inst, 0x04e0, 0x00011100); for (i = 0x04fc; i < 0x053c; i += 4) - nv_wo32(chan, i, 0x07ff0000); - nv_wo32(chan, 0x0544, 0x4b7fffff); - nv_wo32(chan, 0x057c, 0x00000080); - nv_wo32(chan, 0x0580, 0x30201000); - nv_wo32(chan, 0x0584, 0x70605040); - nv_wo32(chan, 0x0588, 0xb8a89888); - nv_wo32(chan, 0x058c, 0xf8e8d8c8); - nv_wo32(chan, 0x05a0, 0xb0000000); + nvkm_wo32(chan->inst, i, 0x07ff0000); + nvkm_wo32(chan->inst, 0x0544, 0x4b7fffff); + nvkm_wo32(chan->inst, 0x057c, 0x00000080); + nvkm_wo32(chan->inst, 0x0580, 0x30201000); + nvkm_wo32(chan->inst, 0x0584, 0x70605040); + nvkm_wo32(chan->inst, 0x0588, 0xb8a89888); + nvkm_wo32(chan->inst, 0x058c, 0xf8e8d8c8); + nvkm_wo32(chan->inst, 0x05a0, 0xb0000000); for (i = 0x05f0; i < 0x0630; i += 4) - nv_wo32(chan, i, 0x00010588); + nvkm_wo32(chan->inst, i, 0x00010588); for (i = 0x0630; i < 0x0670; i += 4) - nv_wo32(chan, i, 0x00030303); + nvkm_wo32(chan->inst, i, 0x00030303); for (i = 0x06b0; i < 0x06f0; i += 4) - nv_wo32(chan, i, 0x0008aae4); + nvkm_wo32(chan->inst, i, 0x0008aae4); for (i = 0x06f0; i < 0x0730; i += 4) - nv_wo32(chan, i, 0x01012000); + nvkm_wo32(chan->inst, i, 0x01012000); for (i = 0x0730; i < 0x0770; i += 4) - nv_wo32(chan, i, 0x00080008); - nv_wo32(chan, 0x0850, 0x00040000); - nv_wo32(chan, 0x0854, 0x00010000); + nvkm_wo32(chan->inst, i, 0x00080008); + nvkm_wo32(chan->inst, 0x0850, 0x00040000); + nvkm_wo32(chan->inst, 0x0854, 0x00010000); for (i = 0x0858; i < 0x0868; i += 4) - nv_wo32(chan, i, 0x00040004); + nvkm_wo32(chan->inst, i, 0x00040004); for (i = 0x15ac; i <= 0x271c ; i += 16) { - nv_wo32(chan, i + 0, 0x10700ff9); - nv_wo32(chan, i + 1, 0x0436086c); - nv_wo32(chan, i + 2, 0x000c001b); + nvkm_wo32(chan->inst, i + 0, 0x10700ff9); + nvkm_wo32(chan->inst, i + 1, 0x0436086c); + nvkm_wo32(chan->inst, i + 2, 0x000c001b); } for (i = 0x274c; i < 0x275c; i += 4) - nv_wo32(chan, i, 0x0000ffff); - nv_wo32(chan, 0x2ae0, 0x3f800000); - nv_wo32(chan, 0x2e9c, 0x3f800000); - nv_wo32(chan, 0x2eb0, 0x3f800000); - nv_wo32(chan, 0x2edc, 0x40000000); - nv_wo32(chan, 0x2ee0, 0x3f800000); - nv_wo32(chan, 0x2ee4, 0x3f000000); - nv_wo32(chan, 0x2eec, 0x40000000); - nv_wo32(chan, 0x2ef0, 0x3f800000); - nv_wo32(chan, 0x2ef8, 0xbf800000); - nv_wo32(chan, 0x2f00, 0xbf800000); + nvkm_wo32(chan->inst, i, 0x0000ffff); + nvkm_wo32(chan->inst, 0x2ae0, 0x3f800000); + nvkm_wo32(chan->inst, 0x2e9c, 0x3f800000); + nvkm_wo32(chan->inst, 0x2eb0, 0x3f800000); + nvkm_wo32(chan->inst, 0x2edc, 0x40000000); + nvkm_wo32(chan->inst, 0x2ee0, 0x3f800000); + nvkm_wo32(chan->inst, 0x2ee4, 0x3f000000); + nvkm_wo32(chan->inst, 0x2eec, 0x40000000); + nvkm_wo32(chan->inst, 0x2ef0, 0x3f800000); + nvkm_wo32(chan->inst, 0x2ef8, 0xbf800000); + nvkm_wo32(chan->inst, 0x2f00, 0xbf800000); + nvkm_done(chan->inst); return 0; } -static struct nvkm_oclass -nv34_gr_cclass = { - .handle = NV_ENGCTX(GR, 0x34), - .ofuncs = &(struct nvkm_ofuncs) { - .ctor = nv34_gr_context_ctor, - .dtor = _nvkm_gr_context_dtor, - .init = nv20_gr_context_init, - .fini = nv20_gr_context_fini, - .rd32 = _nvkm_gr_context_rd32, - .wr32 = _nvkm_gr_context_wr32, - }, -}; - /******************************************************************************* * PGRAPH engine/subdev functions ******************************************************************************/ -static int -nv34_gr_ctor(struct nvkm_object *parent, struct nvkm_object *engine, - struct nvkm_oclass *oclass, void *data, u32 size, - struct nvkm_object **pobject) -{ - struct nv20_gr_priv *priv; - int ret; - - ret = nvkm_gr_create(parent, engine, oclass, true, &priv); - *pobject = nv_object(priv); - if (ret) - return ret; - - ret = nvkm_gpuobj_new(nv_object(priv), NULL, 32 * 4, 16, - NVOBJ_FLAG_ZERO_ALLOC, &priv->ctxtab); - if (ret) - return ret; +static const struct nvkm_gr_func +nv34_gr = { + .dtor = nv20_gr_dtor, + .oneinit = nv20_gr_oneinit, + .init = nv30_gr_init, + .intr = nv20_gr_intr, + .tile = nv20_gr_tile, + .chan_new = nv34_gr_chan_new, + .sclass = { + { -1, -1, 0x0012, &nv04_gr_object }, /* beta1 */ + { -1, -1, 0x0019, &nv04_gr_object }, /* clip */ + { -1, -1, 0x0030, &nv04_gr_object }, /* null */ + { -1, -1, 0x0039, &nv04_gr_object }, /* m2mf */ + { -1, -1, 0x0043, &nv04_gr_object }, /* rop */ + { -1, -1, 0x0044, &nv04_gr_object }, /* patt */ + { -1, -1, 0x004a, &nv04_gr_object }, /* gdi */ + { -1, -1, 0x0062, &nv04_gr_object }, /* surf2d */ + { -1, -1, 0x0072, &nv04_gr_object }, /* beta4 */ + { -1, -1, 0x0089, &nv04_gr_object }, /* sifm */ + { -1, -1, 0x008a, &nv04_gr_object }, /* ifc */ + { -1, -1, 0x009f, &nv04_gr_object }, /* imageblit */ + { -1, -1, 0x0362, &nv04_gr_object }, /* surf2d (nv30) */ + { -1, -1, 0x0389, &nv04_gr_object }, /* sifm (nv30) */ + { -1, -1, 0x038a, &nv04_gr_object }, /* ifc (nv30) */ + { -1, -1, 0x039e, &nv04_gr_object }, /* swzsurf (nv30) */ + { -1, -1, 0x0697, &nv04_gr_object }, /* rankine */ + {} + } +}; - nv_subdev(priv)->unit = 0x00001000; - nv_subdev(priv)->intr = nv20_gr_intr; - nv_engine(priv)->cclass = &nv34_gr_cclass; - nv_engine(priv)->sclass = nv34_gr_sclass; - nv_engine(priv)->tile_prog = nv20_gr_tile_prog; - return 0; +int +nv34_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return nv20_gr_new_(&nv34_gr, device, index, pgr); } - -struct nvkm_oclass -nv34_gr_oclass = { - .handle = NV_ENGINE(GR, 0x34), - .ofuncs = &(struct nvkm_ofuncs) { - .ctor = nv34_gr_ctor, - .dtor = nv20_gr_dtor, - .init = nv30_gr_init, - .fini = _nvkm_gr_fini, - }, -}; diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv35.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv35.c index 707625f19..740df0f52 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv35.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv35.c @@ -1,159 +1,135 @@ #include "nv20.h" #include "regs.h" +#include <core/gpuobj.h> #include <engine/fifo.h> +#include <engine/fifo/chan.h> /******************************************************************************* - * Graphics object classes + * PGRAPH context ******************************************************************************/ -static struct nvkm_oclass -nv35_gr_sclass[] = { - { 0x0012, &nv04_gr_ofuncs, NULL }, /* beta1 */ - { 0x0019, &nv04_gr_ofuncs, NULL }, /* clip */ - { 0x0030, &nv04_gr_ofuncs, NULL }, /* null */ - { 0x0039, &nv04_gr_ofuncs, NULL }, /* m2mf */ - { 0x0043, &nv04_gr_ofuncs, NULL }, /* rop */ - { 0x0044, &nv04_gr_ofuncs, NULL }, /* patt */ - { 0x004a, &nv04_gr_ofuncs, NULL }, /* gdi */ - { 0x0062, &nv04_gr_ofuncs, NULL }, /* surf2d */ - { 0x0072, &nv04_gr_ofuncs, NULL }, /* beta4 */ - { 0x0089, &nv04_gr_ofuncs, NULL }, /* sifm */ - { 0x008a, &nv04_gr_ofuncs, NULL }, /* ifc */ - { 0x009f, &nv04_gr_ofuncs, NULL }, /* imageblit */ - { 0x0362, &nv04_gr_ofuncs, NULL }, /* surf2d (nv30) */ - { 0x0389, &nv04_gr_ofuncs, NULL }, /* sifm (nv30) */ - { 0x038a, &nv04_gr_ofuncs, NULL }, /* ifc (nv30) */ - { 0x039e, &nv04_gr_ofuncs, NULL }, /* swzsurf (nv30) */ - { 0x0497, &nv04_gr_ofuncs, NULL }, /* rankine */ - {}, +static const struct nvkm_object_func +nv35_gr_chan = { + .dtor = nv20_gr_chan_dtor, + .init = nv20_gr_chan_init, + .fini = nv20_gr_chan_fini, }; -/******************************************************************************* - * PGRAPH context - ******************************************************************************/ - static int -nv35_gr_context_ctor(struct nvkm_object *parent, struct nvkm_object *engine, - struct nvkm_oclass *oclass, void *data, u32 size, - struct nvkm_object **pobject) +nv35_gr_chan_new(struct nvkm_gr *base, struct nvkm_fifo_chan *fifoch, + const struct nvkm_oclass *oclass, struct nvkm_object **pobject) { + struct nv20_gr *gr = nv20_gr(base); struct nv20_gr_chan *chan; int ret, i; - ret = nvkm_gr_context_create(parent, engine, oclass, NULL, 0x577c, - 16, NVOBJ_FLAG_ZERO_ALLOC, &chan); - *pobject = nv_object(chan); + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&nv35_gr_chan, oclass, &chan->object); + chan->gr = gr; + chan->chid = fifoch->chid; + *pobject = &chan->object; + + ret = nvkm_memory_new(gr->base.engine.subdev.device, + NVKM_MEM_TARGET_INST, 0x577c, 16, true, + &chan->inst); if (ret) return ret; - chan->chid = nvkm_fifo_chan(parent)->chid; - - nv_wo32(chan, 0x0028, 0x00000001 | (chan->chid << 24)); - nv_wo32(chan, 0x040c, 0x00000101); - nv_wo32(chan, 0x0420, 0x00000111); - nv_wo32(chan, 0x0424, 0x00000060); - nv_wo32(chan, 0x0440, 0x00000080); - nv_wo32(chan, 0x0444, 0xffff0000); - nv_wo32(chan, 0x0448, 0x00000001); - nv_wo32(chan, 0x045c, 0x44400000); - nv_wo32(chan, 0x0488, 0xffff0000); + nvkm_kmap(chan->inst); + nvkm_wo32(chan->inst, 0x0028, 0x00000001 | (chan->chid << 24)); + nvkm_wo32(chan->inst, 0x040c, 0x00000101); + nvkm_wo32(chan->inst, 0x0420, 0x00000111); + nvkm_wo32(chan->inst, 0x0424, 0x00000060); + nvkm_wo32(chan->inst, 0x0440, 0x00000080); + nvkm_wo32(chan->inst, 0x0444, 0xffff0000); + nvkm_wo32(chan->inst, 0x0448, 0x00000001); + nvkm_wo32(chan->inst, 0x045c, 0x44400000); + nvkm_wo32(chan->inst, 0x0488, 0xffff0000); for (i = 0x04dc; i < 0x04e4; i += 4) - nv_wo32(chan, i, 0x0fff0000); - nv_wo32(chan, 0x04e8, 0x00011100); + nvkm_wo32(chan->inst, i, 0x0fff0000); + nvkm_wo32(chan->inst, 0x04e8, 0x00011100); for (i = 0x0504; i < 0x0544; i += 4) - nv_wo32(chan, i, 0x07ff0000); - nv_wo32(chan, 0x054c, 0x4b7fffff); - nv_wo32(chan, 0x0588, 0x00000080); - nv_wo32(chan, 0x058c, 0x30201000); - nv_wo32(chan, 0x0590, 0x70605040); - nv_wo32(chan, 0x0594, 0xb8a89888); - nv_wo32(chan, 0x0598, 0xf8e8d8c8); - nv_wo32(chan, 0x05ac, 0xb0000000); + nvkm_wo32(chan->inst, i, 0x07ff0000); + nvkm_wo32(chan->inst, 0x054c, 0x4b7fffff); + nvkm_wo32(chan->inst, 0x0588, 0x00000080); + nvkm_wo32(chan->inst, 0x058c, 0x30201000); + nvkm_wo32(chan->inst, 0x0590, 0x70605040); + nvkm_wo32(chan->inst, 0x0594, 0xb8a89888); + nvkm_wo32(chan->inst, 0x0598, 0xf8e8d8c8); + nvkm_wo32(chan->inst, 0x05ac, 0xb0000000); for (i = 0x0604; i < 0x0644; i += 4) - nv_wo32(chan, i, 0x00010588); + nvkm_wo32(chan->inst, i, 0x00010588); for (i = 0x0644; i < 0x0684; i += 4) - nv_wo32(chan, i, 0x00030303); + nvkm_wo32(chan->inst, i, 0x00030303); for (i = 0x06c4; i < 0x0704; i += 4) - nv_wo32(chan, i, 0x0008aae4); + nvkm_wo32(chan->inst, i, 0x0008aae4); for (i = 0x0704; i < 0x0744; i += 4) - nv_wo32(chan, i, 0x01012000); + nvkm_wo32(chan->inst, i, 0x01012000); for (i = 0x0744; i < 0x0784; i += 4) - nv_wo32(chan, i, 0x00080008); - nv_wo32(chan, 0x0860, 0x00040000); - nv_wo32(chan, 0x0864, 0x00010000); + nvkm_wo32(chan->inst, i, 0x00080008); + nvkm_wo32(chan->inst, 0x0860, 0x00040000); + nvkm_wo32(chan->inst, 0x0864, 0x00010000); for (i = 0x0868; i < 0x0878; i += 4) - nv_wo32(chan, i, 0x00040004); + nvkm_wo32(chan->inst, i, 0x00040004); for (i = 0x1f1c; i <= 0x308c ; i += 16) { - nv_wo32(chan, i + 0, 0x10700ff9); - nv_wo32(chan, i + 4, 0x0436086c); - nv_wo32(chan, i + 8, 0x000c001b); + nvkm_wo32(chan->inst, i + 0, 0x10700ff9); + nvkm_wo32(chan->inst, i + 4, 0x0436086c); + nvkm_wo32(chan->inst, i + 8, 0x000c001b); } for (i = 0x30bc; i < 0x30cc; i += 4) - nv_wo32(chan, i, 0x0000ffff); - nv_wo32(chan, 0x3450, 0x3f800000); - nv_wo32(chan, 0x380c, 0x3f800000); - nv_wo32(chan, 0x3820, 0x3f800000); - nv_wo32(chan, 0x384c, 0x40000000); - nv_wo32(chan, 0x3850, 0x3f800000); - nv_wo32(chan, 0x3854, 0x3f000000); - nv_wo32(chan, 0x385c, 0x40000000); - nv_wo32(chan, 0x3860, 0x3f800000); - nv_wo32(chan, 0x3868, 0xbf800000); - nv_wo32(chan, 0x3870, 0xbf800000); + nvkm_wo32(chan->inst, i, 0x0000ffff); + nvkm_wo32(chan->inst, 0x3450, 0x3f800000); + nvkm_wo32(chan->inst, 0x380c, 0x3f800000); + nvkm_wo32(chan->inst, 0x3820, 0x3f800000); + nvkm_wo32(chan->inst, 0x384c, 0x40000000); + nvkm_wo32(chan->inst, 0x3850, 0x3f800000); + nvkm_wo32(chan->inst, 0x3854, 0x3f000000); + nvkm_wo32(chan->inst, 0x385c, 0x40000000); + nvkm_wo32(chan->inst, 0x3860, 0x3f800000); + nvkm_wo32(chan->inst, 0x3868, 0xbf800000); + nvkm_wo32(chan->inst, 0x3870, 0xbf800000); + nvkm_done(chan->inst); return 0; } -static struct nvkm_oclass -nv35_gr_cclass = { - .handle = NV_ENGCTX(GR, 0x35), - .ofuncs = &(struct nvkm_ofuncs) { - .ctor = nv35_gr_context_ctor, - .dtor = _nvkm_gr_context_dtor, - .init = nv20_gr_context_init, - .fini = nv20_gr_context_fini, - .rd32 = _nvkm_gr_context_rd32, - .wr32 = _nvkm_gr_context_wr32, - }, -}; - /******************************************************************************* * PGRAPH engine/subdev functions ******************************************************************************/ -static int -nv35_gr_ctor(struct nvkm_object *parent, struct nvkm_object *engine, - struct nvkm_oclass *oclass, void *data, u32 size, - struct nvkm_object **pobject) -{ - struct nv20_gr_priv *priv; - int ret; - - ret = nvkm_gr_create(parent, engine, oclass, true, &priv); - *pobject = nv_object(priv); - if (ret) - return ret; - - ret = nvkm_gpuobj_new(nv_object(priv), NULL, 32 * 4, 16, - NVOBJ_FLAG_ZERO_ALLOC, &priv->ctxtab); - if (ret) - return ret; +static const struct nvkm_gr_func +nv35_gr = { + .dtor = nv20_gr_dtor, + .oneinit = nv20_gr_oneinit, + .init = nv30_gr_init, + .intr = nv20_gr_intr, + .tile = nv20_gr_tile, + .chan_new = nv35_gr_chan_new, + .sclass = { + { -1, -1, 0x0012, &nv04_gr_object }, /* beta1 */ + { -1, -1, 0x0019, &nv04_gr_object }, /* clip */ + { -1, -1, 0x0030, &nv04_gr_object }, /* null */ + { -1, -1, 0x0039, &nv04_gr_object }, /* m2mf */ + { -1, -1, 0x0043, &nv04_gr_object }, /* rop */ + { -1, -1, 0x0044, &nv04_gr_object }, /* patt */ + { -1, -1, 0x004a, &nv04_gr_object }, /* gdi */ + { -1, -1, 0x0062, &nv04_gr_object }, /* surf2d */ + { -1, -1, 0x0072, &nv04_gr_object }, /* beta4 */ + { -1, -1, 0x0089, &nv04_gr_object }, /* sifm */ + { -1, -1, 0x008a, &nv04_gr_object }, /* ifc */ + { -1, -1, 0x009f, &nv04_gr_object }, /* imageblit */ + { -1, -1, 0x0362, &nv04_gr_object }, /* surf2d (nv30) */ + { -1, -1, 0x0389, &nv04_gr_object }, /* sifm (nv30) */ + { -1, -1, 0x038a, &nv04_gr_object }, /* ifc (nv30) */ + { -1, -1, 0x039e, &nv04_gr_object }, /* swzsurf (nv30) */ + { -1, -1, 0x0497, &nv04_gr_object }, /* rankine */ + {} + } +}; - nv_subdev(priv)->unit = 0x00001000; - nv_subdev(priv)->intr = nv20_gr_intr; - nv_engine(priv)->cclass = &nv35_gr_cclass; - nv_engine(priv)->sclass = nv35_gr_sclass; - nv_engine(priv)->tile_prog = nv20_gr_tile_prog; - return 0; +int +nv35_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return nv20_gr_new_(&nv35_gr, device, index, pgr); } - -struct nvkm_oclass -nv35_gr_oclass = { - .handle = NV_ENGINE(GR, 0x35), - .ofuncs = &(struct nvkm_ofuncs) { - .ctor = nv35_gr_ctor, - .dtor = nv20_gr_dtor, - .init = nv30_gr_init, - .fini = _nvkm_gr_fini, - }, -}; diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv40.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv40.c index 7e1937980..05a895496 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv40.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv40.c @@ -25,26 +25,15 @@ #include "regs.h" #include <core/client.h> -#include <core/handle.h> +#include <core/gpuobj.h> #include <subdev/fb.h> #include <subdev/timer.h> #include <engine/fifo.h> -struct nv40_gr_priv { - struct nvkm_gr base; - u32 size; -}; - -struct nv40_gr_chan { - struct nvkm_gr_chan base; -}; - -static u64 +u64 nv40_gr_units(struct nvkm_gr *gr) { - struct nv40_gr_priv *priv = (void *)gr; - - return nv_rd32(priv, 0x1540); + return nvkm_rd32(gr->engine.subdev.device, 0x1540); } /******************************************************************************* @@ -52,80 +41,29 @@ nv40_gr_units(struct nvkm_gr *gr) ******************************************************************************/ static int -nv40_gr_object_ctor(struct nvkm_object *parent, struct nvkm_object *engine, - struct nvkm_oclass *oclass, void *data, u32 size, - struct nvkm_object **pobject) +nv40_gr_object_bind(struct nvkm_object *object, struct nvkm_gpuobj *parent, + int align, struct nvkm_gpuobj **pgpuobj) { - struct nvkm_gpuobj *obj; - int ret; - - ret = nvkm_gpuobj_create(parent, engine, oclass, 0, parent, - 20, 16, 0, &obj); - *pobject = nv_object(obj); - if (ret) - return ret; - - nv_wo32(obj, 0x00, nv_mclass(obj)); - nv_wo32(obj, 0x04, 0x00000000); - nv_wo32(obj, 0x08, 0x00000000); + int ret = nvkm_gpuobj_new(object->engine->subdev.device, 20, align, + false, parent, pgpuobj); + if (ret == 0) { + nvkm_kmap(*pgpuobj); + nvkm_wo32(*pgpuobj, 0x00, object->oclass); + nvkm_wo32(*pgpuobj, 0x04, 0x00000000); + nvkm_wo32(*pgpuobj, 0x08, 0x00000000); #ifdef __BIG_ENDIAN - nv_mo32(obj, 0x08, 0x01000000, 0x01000000); + nvkm_mo32(*pgpuobj, 0x08, 0x01000000, 0x01000000); #endif - nv_wo32(obj, 0x0c, 0x00000000); - nv_wo32(obj, 0x10, 0x00000000); - return 0; + nvkm_wo32(*pgpuobj, 0x0c, 0x00000000); + nvkm_wo32(*pgpuobj, 0x10, 0x00000000); + nvkm_done(*pgpuobj); + } + return ret; } -static struct nvkm_ofuncs -nv40_gr_ofuncs = { - .ctor = nv40_gr_object_ctor, - .dtor = _nvkm_gpuobj_dtor, - .init = _nvkm_gpuobj_init, - .fini = _nvkm_gpuobj_fini, - .rd32 = _nvkm_gpuobj_rd32, - .wr32 = _nvkm_gpuobj_wr32, -}; - -static struct nvkm_oclass -nv40_gr_sclass[] = { - { 0x0012, &nv40_gr_ofuncs, NULL }, /* beta1 */ - { 0x0019, &nv40_gr_ofuncs, NULL }, /* clip */ - { 0x0030, &nv40_gr_ofuncs, NULL }, /* null */ - { 0x0039, &nv40_gr_ofuncs, NULL }, /* m2mf */ - { 0x0043, &nv40_gr_ofuncs, NULL }, /* rop */ - { 0x0044, &nv40_gr_ofuncs, NULL }, /* patt */ - { 0x004a, &nv40_gr_ofuncs, NULL }, /* gdi */ - { 0x0062, &nv40_gr_ofuncs, NULL }, /* surf2d */ - { 0x0072, &nv40_gr_ofuncs, NULL }, /* beta4 */ - { 0x0089, &nv40_gr_ofuncs, NULL }, /* sifm */ - { 0x008a, &nv40_gr_ofuncs, NULL }, /* ifc */ - { 0x009f, &nv40_gr_ofuncs, NULL }, /* imageblit */ - { 0x3062, &nv40_gr_ofuncs, NULL }, /* surf2d (nv40) */ - { 0x3089, &nv40_gr_ofuncs, NULL }, /* sifm (nv40) */ - { 0x309e, &nv40_gr_ofuncs, NULL }, /* swzsurf (nv40) */ - { 0x4097, &nv40_gr_ofuncs, NULL }, /* curie */ - {}, -}; - -static struct nvkm_oclass -nv44_gr_sclass[] = { - { 0x0012, &nv40_gr_ofuncs, NULL }, /* beta1 */ - { 0x0019, &nv40_gr_ofuncs, NULL }, /* clip */ - { 0x0030, &nv40_gr_ofuncs, NULL }, /* null */ - { 0x0039, &nv40_gr_ofuncs, NULL }, /* m2mf */ - { 0x0043, &nv40_gr_ofuncs, NULL }, /* rop */ - { 0x0044, &nv40_gr_ofuncs, NULL }, /* patt */ - { 0x004a, &nv40_gr_ofuncs, NULL }, /* gdi */ - { 0x0062, &nv40_gr_ofuncs, NULL }, /* surf2d */ - { 0x0072, &nv40_gr_ofuncs, NULL }, /* beta4 */ - { 0x0089, &nv40_gr_ofuncs, NULL }, /* sifm */ - { 0x008a, &nv40_gr_ofuncs, NULL }, /* ifc */ - { 0x009f, &nv40_gr_ofuncs, NULL }, /* imageblit */ - { 0x3062, &nv40_gr_ofuncs, NULL }, /* surf2d (nv40) */ - { 0x3089, &nv40_gr_ofuncs, NULL }, /* sifm (nv40) */ - { 0x309e, &nv40_gr_ofuncs, NULL }, /* swzsurf (nv40) */ - { 0x4497, &nv40_gr_ofuncs, NULL }, /* curie */ - {}, +const struct nvkm_object_func +nv40_gr_object = { + .bind = nv40_gr_object_bind, }; /******************************************************************************* @@ -133,361 +71,335 @@ nv44_gr_sclass[] = { ******************************************************************************/ static int -nv40_gr_context_ctor(struct nvkm_object *parent, struct nvkm_object *engine, - struct nvkm_oclass *oclass, void *data, u32 size, - struct nvkm_object **pobject) +nv40_gr_chan_bind(struct nvkm_object *object, struct nvkm_gpuobj *parent, + int align, struct nvkm_gpuobj **pgpuobj) { - struct nv40_gr_priv *priv = (void *)engine; - struct nv40_gr_chan *chan; - int ret; - - ret = nvkm_gr_context_create(parent, engine, oclass, NULL, priv->size, - 16, NVOBJ_FLAG_ZERO_ALLOC, &chan); - *pobject = nv_object(chan); - if (ret) - return ret; - - nv40_grctx_fill(nv_device(priv), nv_gpuobj(chan)); - nv_wo32(chan, 0x00000, nv_gpuobj(chan)->addr >> 4); - return 0; + struct nv40_gr_chan *chan = nv40_gr_chan(object); + struct nv40_gr *gr = chan->gr; + int ret = nvkm_gpuobj_new(gr->base.engine.subdev.device, gr->size, + align, true, parent, pgpuobj); + if (ret == 0) { + chan->inst = (*pgpuobj)->addr; + nvkm_kmap(*pgpuobj); + nv40_grctx_fill(gr->base.engine.subdev.device, *pgpuobj); + nvkm_wo32(*pgpuobj, 0x00000, chan->inst >> 4); + nvkm_done(*pgpuobj); + } + return ret; } static int -nv40_gr_context_fini(struct nvkm_object *object, bool suspend) +nv40_gr_chan_fini(struct nvkm_object *object, bool suspend) { - struct nv40_gr_priv *priv = (void *)object->engine; - struct nv40_gr_chan *chan = (void *)object; - u32 inst = 0x01000000 | nv_gpuobj(chan)->addr >> 4; + struct nv40_gr_chan *chan = nv40_gr_chan(object); + struct nv40_gr *gr = chan->gr; + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 inst = 0x01000000 | chan->inst >> 4; int ret = 0; - nv_mask(priv, 0x400720, 0x00000001, 0x00000000); + nvkm_mask(device, 0x400720, 0x00000001, 0x00000000); - if (nv_rd32(priv, 0x40032c) == inst) { + if (nvkm_rd32(device, 0x40032c) == inst) { if (suspend) { - nv_wr32(priv, 0x400720, 0x00000000); - nv_wr32(priv, 0x400784, inst); - nv_mask(priv, 0x400310, 0x00000020, 0x00000020); - nv_mask(priv, 0x400304, 0x00000001, 0x00000001); - if (!nv_wait(priv, 0x400300, 0x00000001, 0x00000000)) { - u32 insn = nv_rd32(priv, 0x400308); - nv_warn(priv, "ctxprog timeout 0x%08x\n", insn); + nvkm_wr32(device, 0x400720, 0x00000000); + nvkm_wr32(device, 0x400784, inst); + nvkm_mask(device, 0x400310, 0x00000020, 0x00000020); + nvkm_mask(device, 0x400304, 0x00000001, 0x00000001); + if (nvkm_msec(device, 2000, + if (!(nvkm_rd32(device, 0x400300) & 0x00000001)) + break; + ) < 0) { + u32 insn = nvkm_rd32(device, 0x400308); + nvkm_warn(subdev, "ctxprog timeout %08x\n", insn); ret = -EBUSY; } } - nv_mask(priv, 0x40032c, 0x01000000, 0x00000000); + nvkm_mask(device, 0x40032c, 0x01000000, 0x00000000); } - if (nv_rd32(priv, 0x400330) == inst) - nv_mask(priv, 0x400330, 0x01000000, 0x00000000); + if (nvkm_rd32(device, 0x400330) == inst) + nvkm_mask(device, 0x400330, 0x01000000, 0x00000000); - nv_mask(priv, 0x400720, 0x00000001, 0x00000001); + nvkm_mask(device, 0x400720, 0x00000001, 0x00000001); return ret; } -static struct nvkm_oclass -nv40_gr_cclass = { - .handle = NV_ENGCTX(GR, 0x40), - .ofuncs = &(struct nvkm_ofuncs) { - .ctor = nv40_gr_context_ctor, - .dtor = _nvkm_gr_context_dtor, - .init = _nvkm_gr_context_init, - .fini = nv40_gr_context_fini, - .rd32 = _nvkm_gr_context_rd32, - .wr32 = _nvkm_gr_context_wr32, - }, +static void * +nv40_gr_chan_dtor(struct nvkm_object *object) +{ + struct nv40_gr_chan *chan = nv40_gr_chan(object); + unsigned long flags; + spin_lock_irqsave(&chan->gr->base.engine.lock, flags); + list_del(&chan->head); + spin_unlock_irqrestore(&chan->gr->base.engine.lock, flags); + return chan; +} + +static const struct nvkm_object_func +nv40_gr_chan = { + .dtor = nv40_gr_chan_dtor, + .fini = nv40_gr_chan_fini, + .bind = nv40_gr_chan_bind, }; +int +nv40_gr_chan_new(struct nvkm_gr *base, struct nvkm_fifo_chan *fifoch, + const struct nvkm_oclass *oclass, struct nvkm_object **pobject) +{ + struct nv40_gr *gr = nv40_gr(base); + struct nv40_gr_chan *chan; + unsigned long flags; + + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&nv40_gr_chan, oclass, &chan->object); + chan->gr = gr; + chan->fifo = fifoch; + *pobject = &chan->object; + + spin_lock_irqsave(&chan->gr->base.engine.lock, flags); + list_add(&chan->head, &gr->chan); + spin_unlock_irqrestore(&chan->gr->base.engine.lock, flags); + return 0; +} + /******************************************************************************* * PGRAPH engine/subdev functions ******************************************************************************/ static void -nv40_gr_tile_prog(struct nvkm_engine *engine, int i) +nv40_gr_tile(struct nvkm_gr *base, int i, struct nvkm_fb_tile *tile) { - struct nvkm_fb_tile *tile = &nvkm_fb(engine)->tile.region[i]; - struct nvkm_fifo *pfifo = nvkm_fifo(engine); - struct nv40_gr_priv *priv = (void *)engine; + struct nv40_gr *gr = nv40_gr(base); + struct nvkm_device *device = gr->base.engine.subdev.device; + struct nvkm_fifo *fifo = device->fifo; unsigned long flags; - pfifo->pause(pfifo, &flags); - nv04_gr_idle(priv); + nvkm_fifo_pause(fifo, &flags); + nv04_gr_idle(&gr->base); - switch (nv_device(priv)->chipset) { + switch (device->chipset) { case 0x40: case 0x41: case 0x42: case 0x43: case 0x45: - case 0x4e: - nv_wr32(priv, NV20_PGRAPH_TSIZE(i), tile->pitch); - nv_wr32(priv, NV20_PGRAPH_TLIMIT(i), tile->limit); - nv_wr32(priv, NV20_PGRAPH_TILE(i), tile->addr); - nv_wr32(priv, NV40_PGRAPH_TSIZE1(i), tile->pitch); - nv_wr32(priv, NV40_PGRAPH_TLIMIT1(i), tile->limit); - nv_wr32(priv, NV40_PGRAPH_TILE1(i), tile->addr); - switch (nv_device(priv)->chipset) { + nvkm_wr32(device, NV20_PGRAPH_TSIZE(i), tile->pitch); + nvkm_wr32(device, NV20_PGRAPH_TLIMIT(i), tile->limit); + nvkm_wr32(device, NV20_PGRAPH_TILE(i), tile->addr); + nvkm_wr32(device, NV40_PGRAPH_TSIZE1(i), tile->pitch); + nvkm_wr32(device, NV40_PGRAPH_TLIMIT1(i), tile->limit); + nvkm_wr32(device, NV40_PGRAPH_TILE1(i), tile->addr); + switch (device->chipset) { case 0x40: case 0x45: - nv_wr32(priv, NV20_PGRAPH_ZCOMP(i), tile->zcomp); - nv_wr32(priv, NV40_PGRAPH_ZCOMP1(i), tile->zcomp); + nvkm_wr32(device, NV20_PGRAPH_ZCOMP(i), tile->zcomp); + nvkm_wr32(device, NV40_PGRAPH_ZCOMP1(i), tile->zcomp); break; case 0x41: case 0x42: case 0x43: - nv_wr32(priv, NV41_PGRAPH_ZCOMP0(i), tile->zcomp); - nv_wr32(priv, NV41_PGRAPH_ZCOMP1(i), tile->zcomp); + nvkm_wr32(device, NV41_PGRAPH_ZCOMP0(i), tile->zcomp); + nvkm_wr32(device, NV41_PGRAPH_ZCOMP1(i), tile->zcomp); break; default: break; } break; - case 0x44: - case 0x4a: - nv_wr32(priv, NV20_PGRAPH_TSIZE(i), tile->pitch); - nv_wr32(priv, NV20_PGRAPH_TLIMIT(i), tile->limit); - nv_wr32(priv, NV20_PGRAPH_TILE(i), tile->addr); - break; - case 0x46: - case 0x4c: case 0x47: case 0x49: case 0x4b: - case 0x63: - case 0x67: - case 0x68: - nv_wr32(priv, NV47_PGRAPH_TSIZE(i), tile->pitch); - nv_wr32(priv, NV47_PGRAPH_TLIMIT(i), tile->limit); - nv_wr32(priv, NV47_PGRAPH_TILE(i), tile->addr); - nv_wr32(priv, NV40_PGRAPH_TSIZE1(i), tile->pitch); - nv_wr32(priv, NV40_PGRAPH_TLIMIT1(i), tile->limit); - nv_wr32(priv, NV40_PGRAPH_TILE1(i), tile->addr); - switch (nv_device(priv)->chipset) { - case 0x47: - case 0x49: - case 0x4b: - nv_wr32(priv, NV47_PGRAPH_ZCOMP0(i), tile->zcomp); - nv_wr32(priv, NV47_PGRAPH_ZCOMP1(i), tile->zcomp); - break; - default: - break; - } + nvkm_wr32(device, NV47_PGRAPH_TSIZE(i), tile->pitch); + nvkm_wr32(device, NV47_PGRAPH_TLIMIT(i), tile->limit); + nvkm_wr32(device, NV47_PGRAPH_TILE(i), tile->addr); + nvkm_wr32(device, NV40_PGRAPH_TSIZE1(i), tile->pitch); + nvkm_wr32(device, NV40_PGRAPH_TLIMIT1(i), tile->limit); + nvkm_wr32(device, NV40_PGRAPH_TILE1(i), tile->addr); + nvkm_wr32(device, NV47_PGRAPH_ZCOMP0(i), tile->zcomp); + nvkm_wr32(device, NV47_PGRAPH_ZCOMP1(i), tile->zcomp); break; default: + WARN_ON(1); break; } - pfifo->start(pfifo, &flags); + nvkm_fifo_start(fifo, &flags); } -static void -nv40_gr_intr(struct nvkm_subdev *subdev) +void +nv40_gr_intr(struct nvkm_gr *base) { - struct nvkm_fifo *pfifo = nvkm_fifo(subdev); - struct nvkm_engine *engine = nv_engine(subdev); - struct nvkm_object *engctx; - struct nvkm_handle *handle = NULL; - struct nv40_gr_priv *priv = (void *)subdev; - u32 stat = nv_rd32(priv, NV03_PGRAPH_INTR); - u32 nsource = nv_rd32(priv, NV03_PGRAPH_NSOURCE); - u32 nstatus = nv_rd32(priv, NV03_PGRAPH_NSTATUS); - u32 inst = nv_rd32(priv, 0x40032c) & 0x000fffff; - u32 addr = nv_rd32(priv, NV04_PGRAPH_TRAPPED_ADDR); + struct nv40_gr *gr = nv40_gr(base); + struct nv40_gr_chan *temp, *chan = NULL; + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 stat = nvkm_rd32(device, NV03_PGRAPH_INTR); + u32 nsource = nvkm_rd32(device, NV03_PGRAPH_NSOURCE); + u32 nstatus = nvkm_rd32(device, NV03_PGRAPH_NSTATUS); + u32 inst = nvkm_rd32(device, 0x40032c) & 0x000fffff; + u32 addr = nvkm_rd32(device, NV04_PGRAPH_TRAPPED_ADDR); u32 subc = (addr & 0x00070000) >> 16; u32 mthd = (addr & 0x00001ffc); - u32 data = nv_rd32(priv, NV04_PGRAPH_TRAPPED_DATA); - u32 class = nv_rd32(priv, 0x400160 + subc * 4) & 0xffff; + u32 data = nvkm_rd32(device, NV04_PGRAPH_TRAPPED_DATA); + u32 class = nvkm_rd32(device, 0x400160 + subc * 4) & 0xffff; u32 show = stat; - int chid; - - engctx = nvkm_engctx_get(engine, inst); - chid = pfifo->chid(pfifo, engctx); + char msg[128], src[128], sta[128]; + unsigned long flags; - if (stat & NV_PGRAPH_INTR_ERROR) { - if (nsource & NV03_PGRAPH_NSOURCE_ILLEGAL_MTHD) { - handle = nvkm_handle_get_class(engctx, class); - if (handle && !nv_call(handle->object, mthd, data)) - show &= ~NV_PGRAPH_INTR_ERROR; - nvkm_handle_put(handle); + spin_lock_irqsave(&gr->base.engine.lock, flags); + list_for_each_entry(temp, &gr->chan, head) { + if (temp->inst >> 4 == inst) { + chan = temp; + list_del(&chan->head); + list_add(&chan->head, &gr->chan); + break; } + } + if (stat & NV_PGRAPH_INTR_ERROR) { if (nsource & NV03_PGRAPH_NSOURCE_DMA_VTX_PROTECTION) { - nv_mask(priv, 0x402000, 0, 0); + nvkm_mask(device, 0x402000, 0, 0); } } - nv_wr32(priv, NV03_PGRAPH_INTR, stat); - nv_wr32(priv, NV04_PGRAPH_FIFO, 0x00000001); + nvkm_wr32(device, NV03_PGRAPH_INTR, stat); + nvkm_wr32(device, NV04_PGRAPH_FIFO, 0x00000001); if (show) { - nv_error(priv, "%s", ""); - nvkm_bitfield_print(nv10_gr_intr_name, show); - pr_cont(" nsource:"); - nvkm_bitfield_print(nv04_gr_nsource, nsource); - pr_cont(" nstatus:"); - nvkm_bitfield_print(nv10_gr_nstatus, nstatus); - pr_cont("\n"); - nv_error(priv, - "ch %d [0x%08x %s] subc %d class 0x%04x mthd 0x%04x data 0x%08x\n", - chid, inst << 4, nvkm_client_name(engctx), subc, - class, mthd, data); + nvkm_snprintbf(msg, sizeof(msg), nv10_gr_intr_name, show); + nvkm_snprintbf(src, sizeof(src), nv04_gr_nsource, nsource); + nvkm_snprintbf(sta, sizeof(sta), nv10_gr_nstatus, nstatus); + nvkm_error(subdev, "intr %08x [%s] nsource %08x [%s] " + "nstatus %08x [%s] ch %d [%08x %s] subc %d " + "class %04x mthd %04x data %08x\n", + show, msg, nsource, src, nstatus, sta, + chan ? chan->fifo->chid : -1, inst << 4, + chan ? chan->fifo->object.client->name : "unknown", + subc, class, mthd, data); } - nvkm_engctx_put(engctx); + spin_unlock_irqrestore(&gr->base.engine.lock, flags); } -static int -nv40_gr_ctor(struct nvkm_object *parent, struct nvkm_object *engine, - struct nvkm_oclass *oclass, void *data, u32 size, - struct nvkm_object **pobject) +int +nv40_gr_init(struct nvkm_gr *base) { - struct nv40_gr_priv *priv; - int ret; - - ret = nvkm_gr_create(parent, engine, oclass, true, &priv); - *pobject = nv_object(priv); - if (ret) - return ret; - - nv_subdev(priv)->unit = 0x00001000; - nv_subdev(priv)->intr = nv40_gr_intr; - nv_engine(priv)->cclass = &nv40_gr_cclass; - if (nv44_gr_class(priv)) - nv_engine(priv)->sclass = nv44_gr_sclass; - else - nv_engine(priv)->sclass = nv40_gr_sclass; - nv_engine(priv)->tile_prog = nv40_gr_tile_prog; - - priv->base.units = nv40_gr_units; - return 0; -} - -static int -nv40_gr_init(struct nvkm_object *object) -{ - struct nvkm_engine *engine = nv_engine(object); - struct nvkm_fb *pfb = nvkm_fb(object); - struct nv40_gr_priv *priv = (void *)engine; + struct nv40_gr *gr = nv40_gr(base); + struct nvkm_device *device = gr->base.engine.subdev.device; int ret, i, j; u32 vramsz; - ret = nvkm_gr_init(&priv->base); - if (ret) - return ret; - /* generate and upload context program */ - ret = nv40_grctx_init(nv_device(priv), &priv->size); + ret = nv40_grctx_init(device, &gr->size); if (ret) return ret; /* No context present currently */ - nv_wr32(priv, NV40_PGRAPH_CTXCTL_CUR, 0x00000000); + nvkm_wr32(device, NV40_PGRAPH_CTXCTL_CUR, 0x00000000); - nv_wr32(priv, NV03_PGRAPH_INTR , 0xFFFFFFFF); - nv_wr32(priv, NV40_PGRAPH_INTR_EN, 0xFFFFFFFF); + nvkm_wr32(device, NV03_PGRAPH_INTR , 0xFFFFFFFF); + nvkm_wr32(device, NV40_PGRAPH_INTR_EN, 0xFFFFFFFF); - nv_wr32(priv, NV04_PGRAPH_DEBUG_0, 0xFFFFFFFF); - nv_wr32(priv, NV04_PGRAPH_DEBUG_0, 0x00000000); - nv_wr32(priv, NV04_PGRAPH_DEBUG_1, 0x401287c0); - nv_wr32(priv, NV04_PGRAPH_DEBUG_3, 0xe0de8055); - nv_wr32(priv, NV10_PGRAPH_DEBUG_4, 0x00008000); - nv_wr32(priv, NV04_PGRAPH_LIMIT_VIOL_PIX, 0x00be3c5f); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_0, 0xFFFFFFFF); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_0, 0x00000000); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_1, 0x401287c0); + nvkm_wr32(device, NV04_PGRAPH_DEBUG_3, 0xe0de8055); + nvkm_wr32(device, NV10_PGRAPH_DEBUG_4, 0x00008000); + nvkm_wr32(device, NV04_PGRAPH_LIMIT_VIOL_PIX, 0x00be3c5f); - nv_wr32(priv, NV10_PGRAPH_CTX_CONTROL, 0x10010100); - nv_wr32(priv, NV10_PGRAPH_STATE , 0xFFFFFFFF); + nvkm_wr32(device, NV10_PGRAPH_CTX_CONTROL, 0x10010100); + nvkm_wr32(device, NV10_PGRAPH_STATE , 0xFFFFFFFF); - j = nv_rd32(priv, 0x1540) & 0xff; + j = nvkm_rd32(device, 0x1540) & 0xff; if (j) { for (i = 0; !(j & 1); j >>= 1, i++) ; - nv_wr32(priv, 0x405000, i); + nvkm_wr32(device, 0x405000, i); } - if (nv_device(priv)->chipset == 0x40) { - nv_wr32(priv, 0x4009b0, 0x83280fff); - nv_wr32(priv, 0x4009b4, 0x000000a0); + if (device->chipset == 0x40) { + nvkm_wr32(device, 0x4009b0, 0x83280fff); + nvkm_wr32(device, 0x4009b4, 0x000000a0); } else { - nv_wr32(priv, 0x400820, 0x83280eff); - nv_wr32(priv, 0x400824, 0x000000a0); + nvkm_wr32(device, 0x400820, 0x83280eff); + nvkm_wr32(device, 0x400824, 0x000000a0); } - switch (nv_device(priv)->chipset) { + switch (device->chipset) { case 0x40: case 0x45: - nv_wr32(priv, 0x4009b8, 0x0078e366); - nv_wr32(priv, 0x4009bc, 0x0000014c); + nvkm_wr32(device, 0x4009b8, 0x0078e366); + nvkm_wr32(device, 0x4009bc, 0x0000014c); break; case 0x41: case 0x42: /* pciid also 0x00Cx */ /* case 0x0120: XXX (pciid) */ - nv_wr32(priv, 0x400828, 0x007596ff); - nv_wr32(priv, 0x40082c, 0x00000108); + nvkm_wr32(device, 0x400828, 0x007596ff); + nvkm_wr32(device, 0x40082c, 0x00000108); break; case 0x43: - nv_wr32(priv, 0x400828, 0x0072cb77); - nv_wr32(priv, 0x40082c, 0x00000108); + nvkm_wr32(device, 0x400828, 0x0072cb77); + nvkm_wr32(device, 0x40082c, 0x00000108); break; case 0x44: case 0x46: /* G72 */ case 0x4a: case 0x4c: /* G7x-based C51 */ case 0x4e: - nv_wr32(priv, 0x400860, 0); - nv_wr32(priv, 0x400864, 0); + nvkm_wr32(device, 0x400860, 0); + nvkm_wr32(device, 0x400864, 0); break; case 0x47: /* G70 */ case 0x49: /* G71 */ case 0x4b: /* G73 */ - nv_wr32(priv, 0x400828, 0x07830610); - nv_wr32(priv, 0x40082c, 0x0000016A); + nvkm_wr32(device, 0x400828, 0x07830610); + nvkm_wr32(device, 0x40082c, 0x0000016A); break; default: break; } - nv_wr32(priv, 0x400b38, 0x2ffff800); - nv_wr32(priv, 0x400b3c, 0x00006000); + nvkm_wr32(device, 0x400b38, 0x2ffff800); + nvkm_wr32(device, 0x400b3c, 0x00006000); /* Tiling related stuff. */ - switch (nv_device(priv)->chipset) { + switch (device->chipset) { case 0x44: case 0x4a: - nv_wr32(priv, 0x400bc4, 0x1003d888); - nv_wr32(priv, 0x400bbc, 0xb7a7b500); + nvkm_wr32(device, 0x400bc4, 0x1003d888); + nvkm_wr32(device, 0x400bbc, 0xb7a7b500); break; case 0x46: - nv_wr32(priv, 0x400bc4, 0x0000e024); - nv_wr32(priv, 0x400bbc, 0xb7a7b520); + nvkm_wr32(device, 0x400bc4, 0x0000e024); + nvkm_wr32(device, 0x400bbc, 0xb7a7b520); break; case 0x4c: case 0x4e: case 0x67: - nv_wr32(priv, 0x400bc4, 0x1003d888); - nv_wr32(priv, 0x400bbc, 0xb7a7b540); + nvkm_wr32(device, 0x400bc4, 0x1003d888); + nvkm_wr32(device, 0x400bbc, 0xb7a7b540); break; default: break; } - /* Turn all the tiling regions off. */ - for (i = 0; i < pfb->tile.regions; i++) - engine->tile_prog(engine, i); - /* begin RAM config */ - vramsz = nv_device_resource_len(nv_device(priv), 0) - 1; - switch (nv_device(priv)->chipset) { + vramsz = device->func->resource_size(device, 1) - 1; + switch (device->chipset) { case 0x40: - nv_wr32(priv, 0x4009A4, nv_rd32(priv, 0x100200)); - nv_wr32(priv, 0x4009A8, nv_rd32(priv, 0x100204)); - nv_wr32(priv, 0x4069A4, nv_rd32(priv, 0x100200)); - nv_wr32(priv, 0x4069A8, nv_rd32(priv, 0x100204)); - nv_wr32(priv, 0x400820, 0); - nv_wr32(priv, 0x400824, 0); - nv_wr32(priv, 0x400864, vramsz); - nv_wr32(priv, 0x400868, vramsz); + nvkm_wr32(device, 0x4009A4, nvkm_rd32(device, 0x100200)); + nvkm_wr32(device, 0x4009A8, nvkm_rd32(device, 0x100204)); + nvkm_wr32(device, 0x4069A4, nvkm_rd32(device, 0x100200)); + nvkm_wr32(device, 0x4069A8, nvkm_rd32(device, 0x100204)); + nvkm_wr32(device, 0x400820, 0); + nvkm_wr32(device, 0x400824, 0); + nvkm_wr32(device, 0x400864, vramsz); + nvkm_wr32(device, 0x400868, vramsz); break; default: - switch (nv_device(priv)->chipset) { + switch (device->chipset) { case 0x41: case 0x42: case 0x43: @@ -495,33 +407,70 @@ nv40_gr_init(struct nvkm_object *object) case 0x4e: case 0x44: case 0x4a: - nv_wr32(priv, 0x4009F0, nv_rd32(priv, 0x100200)); - nv_wr32(priv, 0x4009F4, nv_rd32(priv, 0x100204)); + nvkm_wr32(device, 0x4009F0, nvkm_rd32(device, 0x100200)); + nvkm_wr32(device, 0x4009F4, nvkm_rd32(device, 0x100204)); break; default: - nv_wr32(priv, 0x400DF0, nv_rd32(priv, 0x100200)); - nv_wr32(priv, 0x400DF4, nv_rd32(priv, 0x100204)); + nvkm_wr32(device, 0x400DF0, nvkm_rd32(device, 0x100200)); + nvkm_wr32(device, 0x400DF4, nvkm_rd32(device, 0x100204)); break; } - nv_wr32(priv, 0x4069F0, nv_rd32(priv, 0x100200)); - nv_wr32(priv, 0x4069F4, nv_rd32(priv, 0x100204)); - nv_wr32(priv, 0x400840, 0); - nv_wr32(priv, 0x400844, 0); - nv_wr32(priv, 0x4008A0, vramsz); - nv_wr32(priv, 0x4008A4, vramsz); + nvkm_wr32(device, 0x4069F0, nvkm_rd32(device, 0x100200)); + nvkm_wr32(device, 0x4069F4, nvkm_rd32(device, 0x100204)); + nvkm_wr32(device, 0x400840, 0); + nvkm_wr32(device, 0x400844, 0); + nvkm_wr32(device, 0x4008A0, vramsz); + nvkm_wr32(device, 0x4008A4, vramsz); break; } return 0; } -struct nvkm_oclass -nv40_gr_oclass = { - .handle = NV_ENGINE(GR, 0x40), - .ofuncs = &(struct nvkm_ofuncs) { - .ctor = nv40_gr_ctor, - .dtor = _nvkm_gr_dtor, - .init = nv40_gr_init, - .fini = _nvkm_gr_fini, - }, +int +nv40_gr_new_(const struct nvkm_gr_func *func, struct nvkm_device *device, + int index, struct nvkm_gr **pgr) +{ + struct nv40_gr *gr; + + if (!(gr = kzalloc(sizeof(*gr), GFP_KERNEL))) + return -ENOMEM; + *pgr = &gr->base; + INIT_LIST_HEAD(&gr->chan); + + return nvkm_gr_ctor(func, device, index, 0x00001000, true, &gr->base); +} + +static const struct nvkm_gr_func +nv40_gr = { + .init = nv40_gr_init, + .intr = nv40_gr_intr, + .tile = nv40_gr_tile, + .units = nv40_gr_units, + .chan_new = nv40_gr_chan_new, + .sclass = { + { -1, -1, 0x0012, &nv40_gr_object }, /* beta1 */ + { -1, -1, 0x0019, &nv40_gr_object }, /* clip */ + { -1, -1, 0x0030, &nv40_gr_object }, /* null */ + { -1, -1, 0x0039, &nv40_gr_object }, /* m2mf */ + { -1, -1, 0x0043, &nv40_gr_object }, /* rop */ + { -1, -1, 0x0044, &nv40_gr_object }, /* patt */ + { -1, -1, 0x004a, &nv40_gr_object }, /* gdi */ + { -1, -1, 0x0062, &nv40_gr_object }, /* surf2d */ + { -1, -1, 0x0072, &nv40_gr_object }, /* beta4 */ + { -1, -1, 0x0089, &nv40_gr_object }, /* sifm */ + { -1, -1, 0x008a, &nv40_gr_object }, /* ifc */ + { -1, -1, 0x009f, &nv40_gr_object }, /* imageblit */ + { -1, -1, 0x3062, &nv40_gr_object }, /* surf2d (nv40) */ + { -1, -1, 0x3089, &nv40_gr_object }, /* sifm (nv40) */ + { -1, -1, 0x309e, &nv40_gr_object }, /* swzsurf (nv40) */ + { -1, -1, 0x4097, &nv40_gr_object }, /* curie */ + {} + } }; + +int +nv40_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return nv40_gr_new_(&nv40_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv40.h b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv40.h index d852bd6de..2812ed11f 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv40.h +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv40.h @@ -1,22 +1,45 @@ #ifndef __NV40_GR_H__ #define __NV40_GR_H__ -#include <engine/gr.h> +#define nv40_gr(p) container_of((p), struct nv40_gr, base) +#include "priv.h" -#include <core/device.h> -struct nvkm_gpuobj; +struct nv40_gr { + struct nvkm_gr base; + u32 size; + struct list_head chan; +}; + +int nv40_gr_new_(const struct nvkm_gr_func *, struct nvkm_device *, int index, + struct nvkm_gr **); +int nv40_gr_init(struct nvkm_gr *); +void nv40_gr_intr(struct nvkm_gr *); +u64 nv40_gr_units(struct nvkm_gr *); + +#define nv40_gr_chan(p) container_of((p), struct nv40_gr_chan, object) + +struct nv40_gr_chan { + struct nvkm_object object; + struct nv40_gr *gr; + struct nvkm_fifo_chan *fifo; + u32 inst; + struct list_head head; +}; + +int nv40_gr_chan_new(struct nvkm_gr *, struct nvkm_fifo_chan *, + const struct nvkm_oclass *, struct nvkm_object **); + +extern const struct nvkm_object_func nv40_gr_object; /* returns 1 if device is one of the nv4x using the 0x4497 object class, * helpful to determine a number of other hardware features */ static inline int -nv44_gr_class(void *priv) +nv44_gr_class(struct nvkm_device *device) { - struct nvkm_device *device = nv_device(priv); - if ((device->chipset & 0xf0) == 0x60) return 1; - return !(0x0baf & (1 << (device->chipset & 0x0f))); + return !(0x0aaf & (1 << (device->chipset & 0x0f))); } int nv40_grctx_init(struct nvkm_device *, u32 *size); diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv44.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv44.c new file mode 100644 index 000000000..45ff80254 --- /dev/null +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv44.c @@ -0,0 +1,108 @@ +/* + * Copyright 2012 Red Hat Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Ben Skeggs + */ +#include "nv40.h" +#include "regs.h" + +#include <subdev/fb.h> +#include <engine/fifo.h> + +static void +nv44_gr_tile(struct nvkm_gr *base, int i, struct nvkm_fb_tile *tile) +{ + struct nv40_gr *gr = nv40_gr(base); + struct nvkm_device *device = gr->base.engine.subdev.device; + struct nvkm_fifo *fifo = device->fifo; + unsigned long flags; + + nvkm_fifo_pause(fifo, &flags); + nv04_gr_idle(&gr->base); + + switch (device->chipset) { + case 0x44: + case 0x4a: + nvkm_wr32(device, NV20_PGRAPH_TSIZE(i), tile->pitch); + nvkm_wr32(device, NV20_PGRAPH_TLIMIT(i), tile->limit); + nvkm_wr32(device, NV20_PGRAPH_TILE(i), tile->addr); + break; + case 0x46: + case 0x4c: + case 0x63: + case 0x67: + case 0x68: + nvkm_wr32(device, NV47_PGRAPH_TSIZE(i), tile->pitch); + nvkm_wr32(device, NV47_PGRAPH_TLIMIT(i), tile->limit); + nvkm_wr32(device, NV47_PGRAPH_TILE(i), tile->addr); + nvkm_wr32(device, NV40_PGRAPH_TSIZE1(i), tile->pitch); + nvkm_wr32(device, NV40_PGRAPH_TLIMIT1(i), tile->limit); + nvkm_wr32(device, NV40_PGRAPH_TILE1(i), tile->addr); + break; + case 0x4e: + nvkm_wr32(device, NV20_PGRAPH_TSIZE(i), tile->pitch); + nvkm_wr32(device, NV20_PGRAPH_TLIMIT(i), tile->limit); + nvkm_wr32(device, NV20_PGRAPH_TILE(i), tile->addr); + nvkm_wr32(device, NV40_PGRAPH_TSIZE1(i), tile->pitch); + nvkm_wr32(device, NV40_PGRAPH_TLIMIT1(i), tile->limit); + nvkm_wr32(device, NV40_PGRAPH_TILE1(i), tile->addr); + break; + default: + WARN_ON(1); + break; + } + + nvkm_fifo_start(fifo, &flags); +} + +static const struct nvkm_gr_func +nv44_gr = { + .init = nv40_gr_init, + .intr = nv40_gr_intr, + .tile = nv44_gr_tile, + .units = nv40_gr_units, + .chan_new = nv40_gr_chan_new, + .sclass = { + { -1, -1, 0x0012, &nv40_gr_object }, /* beta1 */ + { -1, -1, 0x0019, &nv40_gr_object }, /* clip */ + { -1, -1, 0x0030, &nv40_gr_object }, /* null */ + { -1, -1, 0x0039, &nv40_gr_object }, /* m2mf */ + { -1, -1, 0x0043, &nv40_gr_object }, /* rop */ + { -1, -1, 0x0044, &nv40_gr_object }, /* patt */ + { -1, -1, 0x004a, &nv40_gr_object }, /* gdi */ + { -1, -1, 0x0062, &nv40_gr_object }, /* surf2d */ + { -1, -1, 0x0072, &nv40_gr_object }, /* beta4 */ + { -1, -1, 0x0089, &nv40_gr_object }, /* sifm */ + { -1, -1, 0x008a, &nv40_gr_object }, /* ifc */ + { -1, -1, 0x009f, &nv40_gr_object }, /* imageblit */ + { -1, -1, 0x3062, &nv40_gr_object }, /* surf2d (nv40) */ + { -1, -1, 0x3089, &nv40_gr_object }, /* sifm (nv40) */ + { -1, -1, 0x309e, &nv40_gr_object }, /* swzsurf (nv40) */ + { -1, -1, 0x4497, &nv40_gr_object }, /* curie */ + {} + } +}; + +int +nv44_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return nv40_gr_new_(&nv44_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.c b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.c index 270d7cd63..b19b912d5 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.c +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.c @@ -24,27 +24,13 @@ #include "nv50.h" #include <core/client.h> -#include <core/device.h> -#include <core/handle.h> +#include <core/gpuobj.h> #include <engine/fifo.h> -#include <subdev/timer.h> -struct nv50_gr_priv { - struct nvkm_gr base; - spinlock_t lock; - u32 size; -}; - -struct nv50_gr_chan { - struct nvkm_gr_chan base; -}; - -static u64 +u64 nv50_gr_units(struct nvkm_gr *gr) { - struct nv50_gr_priv *priv = (void *)gr; - - return nv_rd32(priv, 0x1540); + return nvkm_rd32(gr->engine.subdev.device, 0x1540); } /******************************************************************************* @@ -52,86 +38,25 @@ nv50_gr_units(struct nvkm_gr *gr) ******************************************************************************/ static int -nv50_gr_object_ctor(struct nvkm_object *parent, struct nvkm_object *engine, - struct nvkm_oclass *oclass, void *data, u32 size, - struct nvkm_object **pobject) +nv50_gr_object_bind(struct nvkm_object *object, struct nvkm_gpuobj *parent, + int align, struct nvkm_gpuobj **pgpuobj) { - struct nvkm_gpuobj *obj; - int ret; - - ret = nvkm_gpuobj_create(parent, engine, oclass, 0, parent, - 16, 16, 0, &obj); - *pobject = nv_object(obj); - if (ret) - return ret; - - nv_wo32(obj, 0x00, nv_mclass(obj)); - nv_wo32(obj, 0x04, 0x00000000); - nv_wo32(obj, 0x08, 0x00000000); - nv_wo32(obj, 0x0c, 0x00000000); - return 0; + int ret = nvkm_gpuobj_new(object->engine->subdev.device, 16, + align, false, parent, pgpuobj); + if (ret == 0) { + nvkm_kmap(*pgpuobj); + nvkm_wo32(*pgpuobj, 0x00, object->oclass); + nvkm_wo32(*pgpuobj, 0x04, 0x00000000); + nvkm_wo32(*pgpuobj, 0x08, 0x00000000); + nvkm_wo32(*pgpuobj, 0x0c, 0x00000000); + nvkm_done(*pgpuobj); + } + return ret; } -static struct nvkm_ofuncs -nv50_gr_ofuncs = { - .ctor = nv50_gr_object_ctor, - .dtor = _nvkm_gpuobj_dtor, - .init = _nvkm_gpuobj_init, - .fini = _nvkm_gpuobj_fini, - .rd32 = _nvkm_gpuobj_rd32, - .wr32 = _nvkm_gpuobj_wr32, -}; - -static struct nvkm_oclass -nv50_gr_sclass[] = { - { 0x0030, &nv50_gr_ofuncs }, - { 0x502d, &nv50_gr_ofuncs }, - { 0x5039, &nv50_gr_ofuncs }, - { 0x5097, &nv50_gr_ofuncs }, - { 0x50c0, &nv50_gr_ofuncs }, - {} -}; - -static struct nvkm_oclass -g84_gr_sclass[] = { - { 0x0030, &nv50_gr_ofuncs }, - { 0x502d, &nv50_gr_ofuncs }, - { 0x5039, &nv50_gr_ofuncs }, - { 0x50c0, &nv50_gr_ofuncs }, - { 0x8297, &nv50_gr_ofuncs }, - {} -}; - -static struct nvkm_oclass -gt200_gr_sclass[] = { - { 0x0030, &nv50_gr_ofuncs }, - { 0x502d, &nv50_gr_ofuncs }, - { 0x5039, &nv50_gr_ofuncs }, - { 0x50c0, &nv50_gr_ofuncs }, - { 0x8397, &nv50_gr_ofuncs }, - {} -}; - -static struct nvkm_oclass -gt215_gr_sclass[] = { - { 0x0030, &nv50_gr_ofuncs }, - { 0x502d, &nv50_gr_ofuncs }, - { 0x5039, &nv50_gr_ofuncs }, - { 0x50c0, &nv50_gr_ofuncs }, - { 0x8597, &nv50_gr_ofuncs }, - { 0x85c0, &nv50_gr_ofuncs }, - {} -}; - -static struct nvkm_oclass -mcp89_gr_sclass[] = { - { 0x0030, &nv50_gr_ofuncs }, - { 0x502d, &nv50_gr_ofuncs }, - { 0x5039, &nv50_gr_ofuncs }, - { 0x50c0, &nv50_gr_ofuncs }, - { 0x85c0, &nv50_gr_ofuncs }, - { 0x8697, &nv50_gr_ofuncs }, - {} +const struct nvkm_object_func +nv50_gr_object = { + .bind = nv50_gr_object_bind, }; /******************************************************************************* @@ -139,160 +64,43 @@ mcp89_gr_sclass[] = { ******************************************************************************/ static int -nv50_gr_context_ctor(struct nvkm_object *parent, struct nvkm_object *engine, - struct nvkm_oclass *oclass, void *data, u32 size, - struct nvkm_object **pobject) +nv50_gr_chan_bind(struct nvkm_object *object, struct nvkm_gpuobj *parent, + int align, struct nvkm_gpuobj **pgpuobj) { - struct nv50_gr_priv *priv = (void *)engine; - struct nv50_gr_chan *chan; - int ret; - - ret = nvkm_gr_context_create(parent, engine, oclass, NULL, priv->size, - 0, NVOBJ_FLAG_ZERO_ALLOC, &chan); - *pobject = nv_object(chan); - if (ret) - return ret; - - nv50_grctx_fill(nv_device(priv), nv_gpuobj(chan)); - return 0; + struct nv50_gr *gr = nv50_gr_chan(object)->gr; + int ret = nvkm_gpuobj_new(gr->base.engine.subdev.device, gr->size, + align, true, parent, pgpuobj); + if (ret == 0) { + nvkm_kmap(*pgpuobj); + nv50_grctx_fill(gr->base.engine.subdev.device, *pgpuobj); + nvkm_done(*pgpuobj); + } + return ret; } -static struct nvkm_oclass -nv50_gr_cclass = { - .handle = NV_ENGCTX(GR, 0x50), - .ofuncs = &(struct nvkm_ofuncs) { - .ctor = nv50_gr_context_ctor, - .dtor = _nvkm_gr_context_dtor, - .init = _nvkm_gr_context_init, - .fini = _nvkm_gr_context_fini, - .rd32 = _nvkm_gr_context_rd32, - .wr32 = _nvkm_gr_context_wr32, - }, -}; - -/******************************************************************************* - * PGRAPH engine/subdev functions - ******************************************************************************/ - -static const struct nvkm_bitfield nv50_pgr_status[] = { - { 0x00000001, "BUSY" }, /* set when any bit is set */ - { 0x00000002, "DISPATCH" }, - { 0x00000004, "UNK2" }, - { 0x00000008, "UNK3" }, - { 0x00000010, "UNK4" }, - { 0x00000020, "UNK5" }, - { 0x00000040, "M2MF" }, - { 0x00000080, "UNK7" }, - { 0x00000100, "CTXPROG" }, - { 0x00000200, "VFETCH" }, - { 0x00000400, "CCACHE_PREGEOM" }, - { 0x00000800, "STRMOUT_VATTR_POSTGEOM" }, - { 0x00001000, "VCLIP" }, - { 0x00002000, "RATTR_APLANE" }, - { 0x00004000, "TRAST" }, - { 0x00008000, "CLIPID" }, - { 0x00010000, "ZCULL" }, - { 0x00020000, "ENG2D" }, - { 0x00040000, "RMASK" }, - { 0x00080000, "TPC_RAST" }, - { 0x00100000, "TPC_PROP" }, - { 0x00200000, "TPC_TEX" }, - { 0x00400000, "TPC_GEOM" }, - { 0x00800000, "TPC_MP" }, - { 0x01000000, "ROP" }, - {} -}; - -static const char *const nv50_pgr_vstatus_0[] = { - "VFETCH", "CCACHE", "PREGEOM", "POSTGEOM", "VATTR", "STRMOUT", "VCLIP", - NULL -}; - -static const char *const nv50_pgr_vstatus_1[] = { - "TPC_RAST", "TPC_PROP", "TPC_TEX", "TPC_GEOM", "TPC_MP", NULL -}; - -static const char *const nv50_pgr_vstatus_2[] = { - "RATTR", "APLANE", "TRAST", "CLIPID", "ZCULL", "ENG2D", "RMASK", - "ROP", NULL +static const struct nvkm_object_func +nv50_gr_chan = { + .bind = nv50_gr_chan_bind, }; -static void -nvkm_pgr_vstatus_print(struct nv50_gr_priv *priv, int r, - const char *const units[], u32 status) +int +nv50_gr_chan_new(struct nvkm_gr *base, struct nvkm_fifo_chan *fifoch, + const struct nvkm_oclass *oclass, struct nvkm_object **pobject) { - int i; - - nv_error(priv, "PGRAPH_VSTATUS%d: 0x%08x", r, status); + struct nv50_gr *gr = nv50_gr(base); + struct nv50_gr_chan *chan; - for (i = 0; units[i] && status; i++) { - if ((status & 7) == 1) - pr_cont(" %s", units[i]); - status >>= 3; - } - if (status) - pr_cont(" (invalid: 0x%x)", status); - pr_cont("\n"); + if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL))) + return -ENOMEM; + nvkm_object_ctor(&nv50_gr_chan, oclass, &chan->object); + chan->gr = gr; + *pobject = &chan->object; + return 0; } -static int -g84_gr_tlb_flush(struct nvkm_engine *engine) -{ - struct nvkm_timer *ptimer = nvkm_timer(engine); - struct nv50_gr_priv *priv = (void *)engine; - bool idle, timeout = false; - unsigned long flags; - u64 start; - u32 tmp; - - spin_lock_irqsave(&priv->lock, flags); - nv_mask(priv, 0x400500, 0x00000001, 0x00000000); - - start = ptimer->read(ptimer); - do { - idle = true; - - for (tmp = nv_rd32(priv, 0x400380); tmp && idle; tmp >>= 3) { - if ((tmp & 7) == 1) - idle = false; - } - - for (tmp = nv_rd32(priv, 0x400384); tmp && idle; tmp >>= 3) { - if ((tmp & 7) == 1) - idle = false; - } - - for (tmp = nv_rd32(priv, 0x400388); tmp && idle; tmp >>= 3) { - if ((tmp & 7) == 1) - idle = false; - } - } while (!idle && - !(timeout = ptimer->read(ptimer) - start > 2000000000)); - - if (timeout) { - nv_error(priv, "PGRAPH TLB flush idle timeout fail\n"); - - tmp = nv_rd32(priv, 0x400700); - nv_error(priv, "PGRAPH_STATUS : 0x%08x", tmp); - nvkm_bitfield_print(nv50_pgr_status, tmp); - pr_cont("\n"); - - nvkm_pgr_vstatus_print(priv, 0, nv50_pgr_vstatus_0, - nv_rd32(priv, 0x400380)); - nvkm_pgr_vstatus_print(priv, 1, nv50_pgr_vstatus_1, - nv_rd32(priv, 0x400384)); - nvkm_pgr_vstatus_print(priv, 2, nv50_pgr_vstatus_2, - nv_rd32(priv, 0x400388)); - } - - - nv_wr32(priv, 0x100c80, 0x00000001); - if (!nv_wait(priv, 0x100c80, 0x00000001, 0x00000000)) - nv_error(priv, "vm flush timeout\n"); - nv_mask(priv, 0x400500, 0x00000001, 0x00000001); - spin_unlock_irqrestore(&priv->lock, flags); - return timeout ? -EBUSY : 0; -} +/******************************************************************************* + * PGRAPH engine/subdev functions + ******************************************************************************/ static const struct nvkm_bitfield nv50_mp_exec_errors[] = { { 0x01, "STACK_UNDERFLOW" }, @@ -427,157 +235,172 @@ static const struct nvkm_bitfield nv50_gr_trap_prop[] = { }; static void -nv50_priv_prop_trap(struct nv50_gr_priv *priv, - u32 ustatus_addr, u32 ustatus, u32 tp) +nv50_gr_prop_trap(struct nv50_gr *gr, u32 ustatus_addr, u32 ustatus, u32 tp) { - u32 e0c = nv_rd32(priv, ustatus_addr + 0x04); - u32 e10 = nv_rd32(priv, ustatus_addr + 0x08); - u32 e14 = nv_rd32(priv, ustatus_addr + 0x0c); - u32 e18 = nv_rd32(priv, ustatus_addr + 0x10); - u32 e1c = nv_rd32(priv, ustatus_addr + 0x14); - u32 e20 = nv_rd32(priv, ustatus_addr + 0x18); - u32 e24 = nv_rd32(priv, ustatus_addr + 0x1c); + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 e0c = nvkm_rd32(device, ustatus_addr + 0x04); + u32 e10 = nvkm_rd32(device, ustatus_addr + 0x08); + u32 e14 = nvkm_rd32(device, ustatus_addr + 0x0c); + u32 e18 = nvkm_rd32(device, ustatus_addr + 0x10); + u32 e1c = nvkm_rd32(device, ustatus_addr + 0x14); + u32 e20 = nvkm_rd32(device, ustatus_addr + 0x18); + u32 e24 = nvkm_rd32(device, ustatus_addr + 0x1c); + char msg[128]; /* CUDA memory: l[], g[] or stack. */ if (ustatus & 0x00000080) { if (e18 & 0x80000000) { /* g[] read fault? */ - nv_error(priv, "TRAP_PROP - TP %d - CUDA_FAULT - Global read fault at address %02x%08x\n", + nvkm_error(subdev, "TRAP_PROP - TP %d - CUDA_FAULT - Global read fault at address %02x%08x\n", tp, e14, e10 | ((e18 >> 24) & 0x1f)); e18 &= ~0x1f000000; } else if (e18 & 0xc) { /* g[] write fault? */ - nv_error(priv, "TRAP_PROP - TP %d - CUDA_FAULT - Global write fault at address %02x%08x\n", + nvkm_error(subdev, "TRAP_PROP - TP %d - CUDA_FAULT - Global write fault at address %02x%08x\n", tp, e14, e10 | ((e18 >> 7) & 0x1f)); e18 &= ~0x00000f80; } else { - nv_error(priv, "TRAP_PROP - TP %d - Unknown CUDA fault at address %02x%08x\n", + nvkm_error(subdev, "TRAP_PROP - TP %d - Unknown CUDA fault at address %02x%08x\n", tp, e14, e10); } ustatus &= ~0x00000080; } if (ustatus) { - nv_error(priv, "TRAP_PROP - TP %d -", tp); - nvkm_bitfield_print(nv50_gr_trap_prop, ustatus); - pr_cont(" - Address %02x%08x\n", e14, e10); + nvkm_snprintbf(msg, sizeof(msg), nv50_gr_trap_prop, ustatus); + nvkm_error(subdev, "TRAP_PROP - TP %d - %08x [%s] - " + "Address %02x%08x\n", + tp, ustatus, msg, e14, e10); } - nv_error(priv, "TRAP_PROP - TP %d - e0c: %08x, e18: %08x, e1c: %08x, e20: %08x, e24: %08x\n", + nvkm_error(subdev, "TRAP_PROP - TP %d - e0c: %08x, e18: %08x, e1c: %08x, e20: %08x, e24: %08x\n", tp, e0c, e18, e1c, e20, e24); } static void -nv50_priv_mp_trap(struct nv50_gr_priv *priv, int tpid, int display) +nv50_gr_mp_trap(struct nv50_gr *gr, int tpid, int display) { - u32 units = nv_rd32(priv, 0x1540); + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 units = nvkm_rd32(device, 0x1540); u32 addr, mp10, status, pc, oplow, ophigh; + char msg[128]; int i; int mps = 0; for (i = 0; i < 4; i++) { if (!(units & 1 << (i+24))) continue; - if (nv_device(priv)->chipset < 0xa0) + if (device->chipset < 0xa0) addr = 0x408200 + (tpid << 12) + (i << 7); else addr = 0x408100 + (tpid << 11) + (i << 7); - mp10 = nv_rd32(priv, addr + 0x10); - status = nv_rd32(priv, addr + 0x14); + mp10 = nvkm_rd32(device, addr + 0x10); + status = nvkm_rd32(device, addr + 0x14); if (!status) continue; if (display) { - nv_rd32(priv, addr + 0x20); - pc = nv_rd32(priv, addr + 0x24); - oplow = nv_rd32(priv, addr + 0x70); - ophigh = nv_rd32(priv, addr + 0x74); - nv_error(priv, "TRAP_MP_EXEC - " - "TP %d MP %d:", tpid, i); - nvkm_bitfield_print(nv50_mp_exec_errors, status); - pr_cont(" at %06x warp %d, opcode %08x %08x\n", - pc&0xffffff, pc >> 24, - oplow, ophigh); + nvkm_rd32(device, addr + 0x20); + pc = nvkm_rd32(device, addr + 0x24); + oplow = nvkm_rd32(device, addr + 0x70); + ophigh = nvkm_rd32(device, addr + 0x74); + nvkm_snprintbf(msg, sizeof(msg), + nv50_mp_exec_errors, status); + nvkm_error(subdev, "TRAP_MP_EXEC - TP %d MP %d: " + "%08x [%s] at %06x warp %d, " + "opcode %08x %08x\n", + tpid, i, status, msg, pc & 0xffffff, + pc >> 24, oplow, ophigh); } - nv_wr32(priv, addr + 0x10, mp10); - nv_wr32(priv, addr + 0x14, 0); + nvkm_wr32(device, addr + 0x10, mp10); + nvkm_wr32(device, addr + 0x14, 0); mps++; } if (!mps && display) - nv_error(priv, "TRAP_MP_EXEC - TP %d: " + nvkm_error(subdev, "TRAP_MP_EXEC - TP %d: " "No MPs claiming errors?\n", tpid); } static void -nv50_priv_tp_trap(struct nv50_gr_priv *priv, int type, u32 ustatus_old, +nv50_gr_tp_trap(struct nv50_gr *gr, int type, u32 ustatus_old, u32 ustatus_new, int display, const char *name) { + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 units = nvkm_rd32(device, 0x1540); int tps = 0; - u32 units = nv_rd32(priv, 0x1540); int i, r; + char msg[128]; u32 ustatus_addr, ustatus; for (i = 0; i < 16; i++) { if (!(units & (1 << i))) continue; - if (nv_device(priv)->chipset < 0xa0) + if (device->chipset < 0xa0) ustatus_addr = ustatus_old + (i << 12); else ustatus_addr = ustatus_new + (i << 11); - ustatus = nv_rd32(priv, ustatus_addr) & 0x7fffffff; + ustatus = nvkm_rd32(device, ustatus_addr) & 0x7fffffff; if (!ustatus) continue; tps++; switch (type) { case 6: /* texture error... unknown for now */ if (display) { - nv_error(priv, "magic set %d:\n", i); + nvkm_error(subdev, "magic set %d:\n", i); for (r = ustatus_addr + 4; r <= ustatus_addr + 0x10; r += 4) - nv_error(priv, "\t0x%08x: 0x%08x\n", r, - nv_rd32(priv, r)); + nvkm_error(subdev, "\t%08x: %08x\n", r, + nvkm_rd32(device, r)); if (ustatus) { - nv_error(priv, "%s - TP%d:", name, i); - nvkm_bitfield_print(nv50_tex_traps, - ustatus); - pr_cont("\n"); + nvkm_snprintbf(msg, sizeof(msg), + nv50_tex_traps, ustatus); + nvkm_error(subdev, + "%s - TP%d: %08x [%s]\n", + name, i, ustatus, msg); ustatus = 0; } } break; case 7: /* MP error */ if (ustatus & 0x04030000) { - nv50_priv_mp_trap(priv, i, display); + nv50_gr_mp_trap(gr, i, display); ustatus &= ~0x04030000; } if (ustatus && display) { - nv_error(priv, "%s - TP%d:", name, i); - nvkm_bitfield_print(nv50_mpc_traps, ustatus); - pr_cont("\n"); + nvkm_snprintbf(msg, sizeof(msg), + nv50_mpc_traps, ustatus); + nvkm_error(subdev, "%s - TP%d: %08x [%s]\n", + name, i, ustatus, msg); ustatus = 0; } break; case 8: /* PROP error */ if (display) - nv50_priv_prop_trap( - priv, ustatus_addr, ustatus, i); + nv50_gr_prop_trap( + gr, ustatus_addr, ustatus, i); ustatus = 0; break; } if (ustatus) { if (display) - nv_error(priv, "%s - TP%d: Unhandled ustatus 0x%08x\n", name, i, ustatus); + nvkm_error(subdev, "%s - TP%d: Unhandled ustatus %08x\n", name, i, ustatus); } - nv_wr32(priv, ustatus_addr, 0xc0000000); + nvkm_wr32(device, ustatus_addr, 0xc0000000); } if (!tps && display) - nv_warn(priv, "%s - No TPs claiming errors?\n", name); + nvkm_warn(subdev, "%s - No TPs claiming errors?\n", name); } static int -nv50_gr_trap_handler(struct nv50_gr_priv *priv, u32 display, - int chid, u64 inst, struct nvkm_object *engctx) +nv50_gr_trap_handler(struct nv50_gr *gr, u32 display, + int chid, u64 inst, const char *name) { - u32 status = nv_rd32(priv, 0x400108); + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + u32 status = nvkm_rd32(device, 0x400108); u32 ustatus; + char msg[128]; if (!status && display) { - nv_error(priv, "TRAP: no units reporting traps?\n"); + nvkm_error(subdev, "TRAP: no units reporting traps?\n"); return 1; } @@ -585,71 +408,72 @@ nv50_gr_trap_handler(struct nv50_gr_priv *priv, u32 display, * COND, QUERY. If you get a trap from it, the command is still stuck * in DISPATCH and you need to do something about it. */ if (status & 0x001) { - ustatus = nv_rd32(priv, 0x400804) & 0x7fffffff; + ustatus = nvkm_rd32(device, 0x400804) & 0x7fffffff; if (!ustatus && display) { - nv_error(priv, "TRAP_DISPATCH - no ustatus?\n"); + nvkm_error(subdev, "TRAP_DISPATCH - no ustatus?\n"); } - nv_wr32(priv, 0x400500, 0x00000000); + nvkm_wr32(device, 0x400500, 0x00000000); /* Known to be triggered by screwed up NOTIFY and COND... */ if (ustatus & 0x00000001) { - u32 addr = nv_rd32(priv, 0x400808); + u32 addr = nvkm_rd32(device, 0x400808); u32 subc = (addr & 0x00070000) >> 16; u32 mthd = (addr & 0x00001ffc); - u32 datal = nv_rd32(priv, 0x40080c); - u32 datah = nv_rd32(priv, 0x400810); - u32 class = nv_rd32(priv, 0x400814); - u32 r848 = nv_rd32(priv, 0x400848); + u32 datal = nvkm_rd32(device, 0x40080c); + u32 datah = nvkm_rd32(device, 0x400810); + u32 class = nvkm_rd32(device, 0x400814); + u32 r848 = nvkm_rd32(device, 0x400848); - nv_error(priv, "TRAP DISPATCH_FAULT\n"); + nvkm_error(subdev, "TRAP DISPATCH_FAULT\n"); if (display && (addr & 0x80000000)) { - nv_error(priv, - "ch %d [0x%010llx %s] subc %d class 0x%04x mthd 0x%04x data 0x%08x%08x 400808 0x%08x 400848 0x%08x\n", - chid, inst, - nvkm_client_name(engctx), subc, - class, mthd, datah, datal, addr, r848); + nvkm_error(subdev, + "ch %d [%010llx %s] subc %d " + "class %04x mthd %04x data %08x%08x " + "400808 %08x 400848 %08x\n", + chid, inst, name, subc, class, mthd, + datah, datal, addr, r848); } else if (display) { - nv_error(priv, "no stuck command?\n"); + nvkm_error(subdev, "no stuck command?\n"); } - nv_wr32(priv, 0x400808, 0); - nv_wr32(priv, 0x4008e8, nv_rd32(priv, 0x4008e8) & 3); - nv_wr32(priv, 0x400848, 0); + nvkm_wr32(device, 0x400808, 0); + nvkm_wr32(device, 0x4008e8, nvkm_rd32(device, 0x4008e8) & 3); + nvkm_wr32(device, 0x400848, 0); ustatus &= ~0x00000001; } if (ustatus & 0x00000002) { - u32 addr = nv_rd32(priv, 0x40084c); + u32 addr = nvkm_rd32(device, 0x40084c); u32 subc = (addr & 0x00070000) >> 16; u32 mthd = (addr & 0x00001ffc); - u32 data = nv_rd32(priv, 0x40085c); - u32 class = nv_rd32(priv, 0x400814); + u32 data = nvkm_rd32(device, 0x40085c); + u32 class = nvkm_rd32(device, 0x400814); - nv_error(priv, "TRAP DISPATCH_QUERY\n"); + nvkm_error(subdev, "TRAP DISPATCH_QUERY\n"); if (display && (addr & 0x80000000)) { - nv_error(priv, - "ch %d [0x%010llx %s] subc %d class 0x%04x mthd 0x%04x data 0x%08x 40084c 0x%08x\n", - chid, inst, - nvkm_client_name(engctx), subc, - class, mthd, data, addr); + nvkm_error(subdev, + "ch %d [%010llx %s] subc %d " + "class %04x mthd %04x data %08x " + "40084c %08x\n", chid, inst, name, + subc, class, mthd, data, addr); } else if (display) { - nv_error(priv, "no stuck command?\n"); + nvkm_error(subdev, "no stuck command?\n"); } - nv_wr32(priv, 0x40084c, 0); + nvkm_wr32(device, 0x40084c, 0); ustatus &= ~0x00000002; } if (ustatus && display) { - nv_error(priv, "TRAP_DISPATCH (unknown " - "0x%08x)\n", ustatus); + nvkm_error(subdev, "TRAP_DISPATCH " + "(unknown %08x)\n", ustatus); } - nv_wr32(priv, 0x400804, 0xc0000000); - nv_wr32(priv, 0x400108, 0x001); + nvkm_wr32(device, 0x400804, 0xc0000000); + nvkm_wr32(device, 0x400108, 0x001); status &= ~0x001; if (!status) return 0; @@ -657,81 +481,91 @@ nv50_gr_trap_handler(struct nv50_gr_priv *priv, u32 display, /* M2MF: Memory to memory copy engine. */ if (status & 0x002) { - u32 ustatus = nv_rd32(priv, 0x406800) & 0x7fffffff; + u32 ustatus = nvkm_rd32(device, 0x406800) & 0x7fffffff; if (display) { - nv_error(priv, "TRAP_M2MF"); - nvkm_bitfield_print(nv50_gr_trap_m2mf, ustatus); - pr_cont("\n"); - nv_error(priv, "TRAP_M2MF %08x %08x %08x %08x\n", - nv_rd32(priv, 0x406804), nv_rd32(priv, 0x406808), - nv_rd32(priv, 0x40680c), nv_rd32(priv, 0x406810)); - + nvkm_snprintbf(msg, sizeof(msg), + nv50_gr_trap_m2mf, ustatus); + nvkm_error(subdev, "TRAP_M2MF %08x [%s]\n", + ustatus, msg); + nvkm_error(subdev, "TRAP_M2MF %08x %08x %08x %08x\n", + nvkm_rd32(device, 0x406804), + nvkm_rd32(device, 0x406808), + nvkm_rd32(device, 0x40680c), + nvkm_rd32(device, 0x406810)); } /* No sane way found yet -- just reset the bugger. */ - nv_wr32(priv, 0x400040, 2); - nv_wr32(priv, 0x400040, 0); - nv_wr32(priv, 0x406800, 0xc0000000); - nv_wr32(priv, 0x400108, 0x002); + nvkm_wr32(device, 0x400040, 2); + nvkm_wr32(device, 0x400040, 0); + nvkm_wr32(device, 0x406800, 0xc0000000); + nvkm_wr32(device, 0x400108, 0x002); status &= ~0x002; } /* VFETCH: Fetches data from vertex buffers. */ if (status & 0x004) { - u32 ustatus = nv_rd32(priv, 0x400c04) & 0x7fffffff; + u32 ustatus = nvkm_rd32(device, 0x400c04) & 0x7fffffff; if (display) { - nv_error(priv, "TRAP_VFETCH"); - nvkm_bitfield_print(nv50_gr_trap_vfetch, ustatus); - pr_cont("\n"); - nv_error(priv, "TRAP_VFETCH %08x %08x %08x %08x\n", - nv_rd32(priv, 0x400c00), nv_rd32(priv, 0x400c08), - nv_rd32(priv, 0x400c0c), nv_rd32(priv, 0x400c10)); + nvkm_snprintbf(msg, sizeof(msg), + nv50_gr_trap_vfetch, ustatus); + nvkm_error(subdev, "TRAP_VFETCH %08x [%s]\n", + ustatus, msg); + nvkm_error(subdev, "TRAP_VFETCH %08x %08x %08x %08x\n", + nvkm_rd32(device, 0x400c00), + nvkm_rd32(device, 0x400c08), + nvkm_rd32(device, 0x400c0c), + nvkm_rd32(device, 0x400c10)); } - nv_wr32(priv, 0x400c04, 0xc0000000); - nv_wr32(priv, 0x400108, 0x004); + nvkm_wr32(device, 0x400c04, 0xc0000000); + nvkm_wr32(device, 0x400108, 0x004); status &= ~0x004; } /* STRMOUT: DirectX streamout / OpenGL transform feedback. */ if (status & 0x008) { - ustatus = nv_rd32(priv, 0x401800) & 0x7fffffff; + ustatus = nvkm_rd32(device, 0x401800) & 0x7fffffff; if (display) { - nv_error(priv, "TRAP_STRMOUT"); - nvkm_bitfield_print(nv50_gr_trap_strmout, ustatus); - pr_cont("\n"); - nv_error(priv, "TRAP_STRMOUT %08x %08x %08x %08x\n", - nv_rd32(priv, 0x401804), nv_rd32(priv, 0x401808), - nv_rd32(priv, 0x40180c), nv_rd32(priv, 0x401810)); - + nvkm_snprintbf(msg, sizeof(msg), + nv50_gr_trap_strmout, ustatus); + nvkm_error(subdev, "TRAP_STRMOUT %08x [%s]\n", + ustatus, msg); + nvkm_error(subdev, "TRAP_STRMOUT %08x %08x %08x %08x\n", + nvkm_rd32(device, 0x401804), + nvkm_rd32(device, 0x401808), + nvkm_rd32(device, 0x40180c), + nvkm_rd32(device, 0x401810)); } /* No sane way found yet -- just reset the bugger. */ - nv_wr32(priv, 0x400040, 0x80); - nv_wr32(priv, 0x400040, 0); - nv_wr32(priv, 0x401800, 0xc0000000); - nv_wr32(priv, 0x400108, 0x008); + nvkm_wr32(device, 0x400040, 0x80); + nvkm_wr32(device, 0x400040, 0); + nvkm_wr32(device, 0x401800, 0xc0000000); + nvkm_wr32(device, 0x400108, 0x008); status &= ~0x008; } /* CCACHE: Handles code and c[] caches and fills them. */ if (status & 0x010) { - ustatus = nv_rd32(priv, 0x405018) & 0x7fffffff; + ustatus = nvkm_rd32(device, 0x405018) & 0x7fffffff; if (display) { - nv_error(priv, "TRAP_CCACHE"); - nvkm_bitfield_print(nv50_gr_trap_ccache, ustatus); - pr_cont("\n"); - nv_error(priv, "TRAP_CCACHE %08x %08x %08x %08x" - " %08x %08x %08x\n", - nv_rd32(priv, 0x405000), nv_rd32(priv, 0x405004), - nv_rd32(priv, 0x405008), nv_rd32(priv, 0x40500c), - nv_rd32(priv, 0x405010), nv_rd32(priv, 0x405014), - nv_rd32(priv, 0x40501c)); - + nvkm_snprintbf(msg, sizeof(msg), + nv50_gr_trap_ccache, ustatus); + nvkm_error(subdev, "TRAP_CCACHE %08x [%s]\n", + ustatus, msg); + nvkm_error(subdev, "TRAP_CCACHE %08x %08x %08x %08x " + "%08x %08x %08x\n", + nvkm_rd32(device, 0x405000), + nvkm_rd32(device, 0x405004), + nvkm_rd32(device, 0x405008), + nvkm_rd32(device, 0x40500c), + nvkm_rd32(device, 0x405010), + nvkm_rd32(device, 0x405014), + nvkm_rd32(device, 0x40501c)); } - nv_wr32(priv, 0x405018, 0xc0000000); - nv_wr32(priv, 0x400108, 0x010); + nvkm_wr32(device, 0x405018, 0xc0000000); + nvkm_wr32(device, 0x400108, 0x010); status &= ~0x010; } @@ -739,239 +573,174 @@ nv50_gr_trap_handler(struct nv50_gr_priv *priv, u32 display, * remaining, so try to handle it anyway. Perhaps related to that * unknown DMA slot on tesla? */ if (status & 0x20) { - ustatus = nv_rd32(priv, 0x402000) & 0x7fffffff; + ustatus = nvkm_rd32(device, 0x402000) & 0x7fffffff; if (display) - nv_error(priv, "TRAP_UNKC04 0x%08x\n", ustatus); - nv_wr32(priv, 0x402000, 0xc0000000); + nvkm_error(subdev, "TRAP_UNKC04 %08x\n", ustatus); + nvkm_wr32(device, 0x402000, 0xc0000000); /* no status modifiction on purpose */ } /* TEXTURE: CUDA texturing units */ if (status & 0x040) { - nv50_priv_tp_trap(priv, 6, 0x408900, 0x408600, display, + nv50_gr_tp_trap(gr, 6, 0x408900, 0x408600, display, "TRAP_TEXTURE"); - nv_wr32(priv, 0x400108, 0x040); + nvkm_wr32(device, 0x400108, 0x040); status &= ~0x040; } /* MP: CUDA execution engines. */ if (status & 0x080) { - nv50_priv_tp_trap(priv, 7, 0x408314, 0x40831c, display, + nv50_gr_tp_trap(gr, 7, 0x408314, 0x40831c, display, "TRAP_MP"); - nv_wr32(priv, 0x400108, 0x080); + nvkm_wr32(device, 0x400108, 0x080); status &= ~0x080; } /* PROP: Handles TP-initiated uncached memory accesses: * l[], g[], stack, 2d surfaces, render targets. */ if (status & 0x100) { - nv50_priv_tp_trap(priv, 8, 0x408e08, 0x408708, display, + nv50_gr_tp_trap(gr, 8, 0x408e08, 0x408708, display, "TRAP_PROP"); - nv_wr32(priv, 0x400108, 0x100); + nvkm_wr32(device, 0x400108, 0x100); status &= ~0x100; } if (status) { if (display) - nv_error(priv, "TRAP: unknown 0x%08x\n", status); - nv_wr32(priv, 0x400108, status); + nvkm_error(subdev, "TRAP: unknown %08x\n", status); + nvkm_wr32(device, 0x400108, status); } return 1; } -static void -nv50_gr_intr(struct nvkm_subdev *subdev) +void +nv50_gr_intr(struct nvkm_gr *base) { - struct nvkm_fifo *pfifo = nvkm_fifo(subdev); - struct nvkm_engine *engine = nv_engine(subdev); - struct nvkm_object *engctx; - struct nvkm_handle *handle = NULL; - struct nv50_gr_priv *priv = (void *)subdev; - u32 stat = nv_rd32(priv, 0x400100); - u32 inst = nv_rd32(priv, 0x40032c) & 0x0fffffff; - u32 addr = nv_rd32(priv, 0x400704); + struct nv50_gr *gr = nv50_gr(base); + struct nvkm_subdev *subdev = &gr->base.engine.subdev; + struct nvkm_device *device = subdev->device; + struct nvkm_fifo_chan *chan; + u32 stat = nvkm_rd32(device, 0x400100); + u32 inst = nvkm_rd32(device, 0x40032c) & 0x0fffffff; + u32 addr = nvkm_rd32(device, 0x400704); u32 subc = (addr & 0x00070000) >> 16; u32 mthd = (addr & 0x00001ffc); - u32 data = nv_rd32(priv, 0x400708); - u32 class = nv_rd32(priv, 0x400814); + u32 data = nvkm_rd32(device, 0x400708); + u32 class = nvkm_rd32(device, 0x400814); u32 show = stat, show_bitfield = stat; - int chid; - - engctx = nvkm_engctx_get(engine, inst); - chid = pfifo->chid(pfifo, engctx); - - if (stat & 0x00000010) { - handle = nvkm_handle_get_class(engctx, class); - if (handle && !nv_call(handle->object, mthd, data)) - show &= ~0x00000010; - nvkm_handle_put(handle); + const struct nvkm_enum *en; + unsigned long flags; + const char *name = "unknown"; + char msg[128]; + int chid = -1; + + chan = nvkm_fifo_chan_inst(device->fifo, (u64)inst << 12, &flags); + if (chan) { + name = chan->object.client->name; + chid = chan->chid; } if (show & 0x00100000) { - u32 ecode = nv_rd32(priv, 0x400110); - nv_error(priv, "DATA_ERROR "); - nvkm_enum_print(nv50_data_error_names, ecode); - pr_cont("\n"); + u32 ecode = nvkm_rd32(device, 0x400110); + en = nvkm_enum_find(nv50_data_error_names, ecode); + nvkm_error(subdev, "DATA_ERROR %08x [%s]\n", + ecode, en ? en->name : ""); show_bitfield &= ~0x00100000; } if (stat & 0x00200000) { - if (!nv50_gr_trap_handler(priv, show, chid, (u64)inst << 12, - engctx)) + if (!nv50_gr_trap_handler(gr, show, chid, (u64)inst << 12, name)) show &= ~0x00200000; show_bitfield &= ~0x00200000; } - nv_wr32(priv, 0x400100, stat); - nv_wr32(priv, 0x400500, 0x00010001); + nvkm_wr32(device, 0x400100, stat); + nvkm_wr32(device, 0x400500, 0x00010001); if (show) { show &= show_bitfield; - if (show) { - nv_error(priv, "%s", ""); - nvkm_bitfield_print(nv50_gr_intr_name, show); - pr_cont("\n"); - } - nv_error(priv, - "ch %d [0x%010llx %s] subc %d class 0x%04x mthd 0x%04x data 0x%08x\n", - chid, (u64)inst << 12, nvkm_client_name(engctx), - subc, class, mthd, data); + nvkm_snprintbf(msg, sizeof(msg), nv50_gr_intr_name, show); + nvkm_error(subdev, "%08x [%s] ch %d [%010llx %s] subc %d " + "class %04x mthd %04x data %08x\n", + stat, msg, chid, (u64)inst << 12, name, + subc, class, mthd, data); } - if (nv_rd32(priv, 0x400824) & (1 << 31)) - nv_wr32(priv, 0x400824, nv_rd32(priv, 0x400824) & ~(1 << 31)); + if (nvkm_rd32(device, 0x400824) & (1 << 31)) + nvkm_wr32(device, 0x400824, nvkm_rd32(device, 0x400824) & ~(1 << 31)); - nvkm_engctx_put(engctx); + nvkm_fifo_chan_put(device->fifo, flags, &chan); } -static int -nv50_gr_ctor(struct nvkm_object *parent, struct nvkm_object *engine, - struct nvkm_oclass *oclass, void *data, u32 size, - struct nvkm_object **pobject) +int +nv50_gr_init(struct nvkm_gr *base) { - struct nv50_gr_priv *priv; - int ret; - - ret = nvkm_gr_create(parent, engine, oclass, true, &priv); - *pobject = nv_object(priv); - if (ret) - return ret; - - nv_subdev(priv)->unit = 0x00201000; - nv_subdev(priv)->intr = nv50_gr_intr; - nv_engine(priv)->cclass = &nv50_gr_cclass; - - priv->base.units = nv50_gr_units; - - switch (nv_device(priv)->chipset) { - case 0x50: - nv_engine(priv)->sclass = nv50_gr_sclass; - break; - case 0x84: - case 0x86: - case 0x92: - case 0x94: - case 0x96: - case 0x98: - nv_engine(priv)->sclass = g84_gr_sclass; - break; - case 0xa0: - case 0xaa: - case 0xac: - nv_engine(priv)->sclass = gt200_gr_sclass; - break; - case 0xa3: - case 0xa5: - case 0xa8: - nv_engine(priv)->sclass = gt215_gr_sclass; - break; - case 0xaf: - nv_engine(priv)->sclass = mcp89_gr_sclass; - break; - - } - - /* unfortunate hw bug workaround... */ - if (nv_device(priv)->chipset != 0x50 && - nv_device(priv)->chipset != 0xac) - nv_engine(priv)->tlb_flush = g84_gr_tlb_flush; - - spin_lock_init(&priv->lock); - return 0; -} - -static int -nv50_gr_init(struct nvkm_object *object) -{ - struct nv50_gr_priv *priv = (void *)object; + struct nv50_gr *gr = nv50_gr(base); + struct nvkm_device *device = gr->base.engine.subdev.device; int ret, units, i; - ret = nvkm_gr_init(&priv->base); - if (ret) - return ret; - /* NV_PGRAPH_DEBUG_3_HW_CTX_SWITCH_ENABLED */ - nv_wr32(priv, 0x40008c, 0x00000004); + nvkm_wr32(device, 0x40008c, 0x00000004); /* reset/enable traps and interrupts */ - nv_wr32(priv, 0x400804, 0xc0000000); - nv_wr32(priv, 0x406800, 0xc0000000); - nv_wr32(priv, 0x400c04, 0xc0000000); - nv_wr32(priv, 0x401800, 0xc0000000); - nv_wr32(priv, 0x405018, 0xc0000000); - nv_wr32(priv, 0x402000, 0xc0000000); - - units = nv_rd32(priv, 0x001540); + nvkm_wr32(device, 0x400804, 0xc0000000); + nvkm_wr32(device, 0x406800, 0xc0000000); + nvkm_wr32(device, 0x400c04, 0xc0000000); + nvkm_wr32(device, 0x401800, 0xc0000000); + nvkm_wr32(device, 0x405018, 0xc0000000); + nvkm_wr32(device, 0x402000, 0xc0000000); + + units = nvkm_rd32(device, 0x001540); for (i = 0; i < 16; i++) { if (!(units & (1 << i))) continue; - if (nv_device(priv)->chipset < 0xa0) { - nv_wr32(priv, 0x408900 + (i << 12), 0xc0000000); - nv_wr32(priv, 0x408e08 + (i << 12), 0xc0000000); - nv_wr32(priv, 0x408314 + (i << 12), 0xc0000000); + if (device->chipset < 0xa0) { + nvkm_wr32(device, 0x408900 + (i << 12), 0xc0000000); + nvkm_wr32(device, 0x408e08 + (i << 12), 0xc0000000); + nvkm_wr32(device, 0x408314 + (i << 12), 0xc0000000); } else { - nv_wr32(priv, 0x408600 + (i << 11), 0xc0000000); - nv_wr32(priv, 0x408708 + (i << 11), 0xc0000000); - nv_wr32(priv, 0x40831c + (i << 11), 0xc0000000); + nvkm_wr32(device, 0x408600 + (i << 11), 0xc0000000); + nvkm_wr32(device, 0x408708 + (i << 11), 0xc0000000); + nvkm_wr32(device, 0x40831c + (i << 11), 0xc0000000); } } - nv_wr32(priv, 0x400108, 0xffffffff); - nv_wr32(priv, 0x400138, 0xffffffff); - nv_wr32(priv, 0x400100, 0xffffffff); - nv_wr32(priv, 0x40013c, 0xffffffff); - nv_wr32(priv, 0x400500, 0x00010001); + nvkm_wr32(device, 0x400108, 0xffffffff); + nvkm_wr32(device, 0x400138, 0xffffffff); + nvkm_wr32(device, 0x400100, 0xffffffff); + nvkm_wr32(device, 0x40013c, 0xffffffff); + nvkm_wr32(device, 0x400500, 0x00010001); /* upload context program, initialise ctxctl defaults */ - ret = nv50_grctx_init(nv_device(priv), &priv->size); + ret = nv50_grctx_init(device, &gr->size); if (ret) return ret; - nv_wr32(priv, 0x400824, 0x00000000); - nv_wr32(priv, 0x400828, 0x00000000); - nv_wr32(priv, 0x40082c, 0x00000000); - nv_wr32(priv, 0x400830, 0x00000000); - nv_wr32(priv, 0x40032c, 0x00000000); - nv_wr32(priv, 0x400330, 0x00000000); + nvkm_wr32(device, 0x400824, 0x00000000); + nvkm_wr32(device, 0x400828, 0x00000000); + nvkm_wr32(device, 0x40082c, 0x00000000); + nvkm_wr32(device, 0x400830, 0x00000000); + nvkm_wr32(device, 0x40032c, 0x00000000); + nvkm_wr32(device, 0x400330, 0x00000000); /* some unknown zcull magic */ - switch (nv_device(priv)->chipset & 0xf0) { + switch (device->chipset & 0xf0) { case 0x50: case 0x80: case 0x90: - nv_wr32(priv, 0x402ca8, 0x00000800); + nvkm_wr32(device, 0x402ca8, 0x00000800); break; case 0xa0: default: - if (nv_device(priv)->chipset == 0xa0 || - nv_device(priv)->chipset == 0xaa || - nv_device(priv)->chipset == 0xac) { - nv_wr32(priv, 0x402ca8, 0x00000802); + if (device->chipset == 0xa0 || + device->chipset == 0xaa || + device->chipset == 0xac) { + nvkm_wr32(device, 0x402ca8, 0x00000802); } else { - nv_wr32(priv, 0x402cc0, 0x00000000); - nv_wr32(priv, 0x402ca8, 0x00000002); + nvkm_wr32(device, 0x402cc0, 0x00000000); + nvkm_wr32(device, 0x402ca8, 0x00000002); } break; @@ -979,21 +748,47 @@ nv50_gr_init(struct nvkm_object *object) /* zero out zcull regions */ for (i = 0; i < 8; i++) { - nv_wr32(priv, 0x402c20 + (i * 0x10), 0x00000000); - nv_wr32(priv, 0x402c24 + (i * 0x10), 0x00000000); - nv_wr32(priv, 0x402c28 + (i * 0x10), 0x00000000); - nv_wr32(priv, 0x402c2c + (i * 0x10), 0x00000000); + nvkm_wr32(device, 0x402c20 + (i * 0x10), 0x00000000); + nvkm_wr32(device, 0x402c24 + (i * 0x10), 0x00000000); + nvkm_wr32(device, 0x402c28 + (i * 0x10), 0x00000000); + nvkm_wr32(device, 0x402c2c + (i * 0x10), 0x00000000); } + return 0; } -struct nvkm_oclass -nv50_gr_oclass = { - .handle = NV_ENGINE(GR, 0x50), - .ofuncs = &(struct nvkm_ofuncs) { - .ctor = nv50_gr_ctor, - .dtor = _nvkm_gr_dtor, - .init = nv50_gr_init, - .fini = _nvkm_gr_fini, - }, +int +nv50_gr_new_(const struct nvkm_gr_func *func, struct nvkm_device *device, + int index, struct nvkm_gr **pgr) +{ + struct nv50_gr *gr; + + if (!(gr = kzalloc(sizeof(*gr), GFP_KERNEL))) + return -ENOMEM; + spin_lock_init(&gr->lock); + *pgr = &gr->base; + + return nvkm_gr_ctor(func, device, index, 0x00201000, true, &gr->base); +} + +static const struct nvkm_gr_func +nv50_gr = { + .init = nv50_gr_init, + .intr = nv50_gr_intr, + .chan_new = nv50_gr_chan_new, + .units = nv50_gr_units, + .sclass = { + { -1, -1, 0x0030, &nv50_gr_object }, + { -1, -1, 0x502d, &nv50_gr_object }, + { -1, -1, 0x5039, &nv50_gr_object }, + { -1, -1, 0x5097, &nv50_gr_object }, + { -1, -1, 0x50c0, &nv50_gr_object }, + {} + } }; + +int +nv50_gr_new(struct nvkm_device *device, int index, struct nvkm_gr **pgr) +{ + return nv50_gr_new_(&nv50_gr, device, index, pgr); +} diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.h b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.h index bcf786f6b..45eec83a5 100644 --- a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.h +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/nv50.h @@ -1,8 +1,34 @@ #ifndef __NV50_GR_H__ #define __NV50_GR_H__ -#include <engine/gr.h> -struct nvkm_device; -struct nvkm_gpuobj; +#define nv50_gr(p) container_of((p), struct nv50_gr, base) +#include "priv.h" + +struct nv50_gr { + struct nvkm_gr base; + const struct nv50_gr_func *func; + spinlock_t lock; + u32 size; +}; + +int nv50_gr_new_(const struct nvkm_gr_func *, struct nvkm_device *, int index, + struct nvkm_gr **); +int nv50_gr_init(struct nvkm_gr *); +void nv50_gr_intr(struct nvkm_gr *); +u64 nv50_gr_units(struct nvkm_gr *); + +int g84_gr_tlb_flush(struct nvkm_gr *); + +#define nv50_gr_chan(p) container_of((p), struct nv50_gr_chan, object) + +struct nv50_gr_chan { + struct nvkm_object object; + struct nv50_gr *gr; +}; + +int nv50_gr_chan_new(struct nvkm_gr *, struct nvkm_fifo_chan *, + const struct nvkm_oclass *, struct nvkm_object **); + +extern const struct nvkm_object_func nv50_gr_object; int nv50_grctx_init(struct nvkm_device *, u32 *size); void nv50_grctx_fill(struct nvkm_device *, struct nvkm_gpuobj *); diff --git a/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/priv.h b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/priv.h new file mode 100644 index 000000000..a234590be --- /dev/null +++ b/kernel/drivers/gpu/drm/nouveau/nvkm/engine/gr/priv.h @@ -0,0 +1,38 @@ +#ifndef __NVKM_GR_PRIV_H__ +#define __NVKM_GR_PRIV_H__ +#define nvkm_gr(p) container_of((p), struct nvkm_gr, engine) +#include <engine/gr.h> +#include <core/enum.h> +struct nvkm_fb_tile; +struct nvkm_fifo_chan; + +int nvkm_gr_ctor(const struct nvkm_gr_func *, struct nvkm_device *, + int index, u32 pmc_enable, bool enable, + struct nvkm_gr *); + +bool nv04_gr_idle(struct nvkm_gr *); + +struct nvkm_gr_func { + void *(*dtor)(struct nvkm_gr *); + int (*oneinit)(struct nvkm_gr *); + int (*init)(struct nvkm_gr *); + void (*intr)(struct nvkm_gr *); + void (*tile)(struct nvkm_gr *, int region, struct nvkm_fb_tile *); + int (*tlb_flush)(struct nvkm_gr *); + int (*chan_new)(struct nvkm_gr *, struct nvkm_fifo_chan *, + const struct nvkm_oclass *, struct nvkm_object **); + int (*object_get)(struct nvkm_gr *, int, struct nvkm_sclass *); + /* Returns chipset-specific counts of units packed into an u64. + */ + u64 (*units)(struct nvkm_gr *); + struct nvkm_sclass sclass[]; +}; + +extern const struct nvkm_bitfield nv04_gr_nsource[]; +extern const struct nvkm_object_func nv04_gr_object; + +extern const struct nvkm_bitfield nv10_gr_intr_name[]; +extern const struct nvkm_bitfield nv10_gr_nstatus[]; + +extern const struct nvkm_enum nv50_data_error_names[]; +#endif |