diff options
Diffstat (limited to 'kernel/net/sunrpc/xprtrdma')
| -rw-r--r-- | kernel/net/sunrpc/xprtrdma/Makefile | 15 | ||||
| -rw-r--r-- | kernel/net/sunrpc/xprtrdma/backchannel.c | 394 | ||||
| -rw-r--r-- | kernel/net/sunrpc/xprtrdma/fmr_ops.c | 120 | ||||
| -rw-r--r-- | kernel/net/sunrpc/xprtrdma/frwr_ops.c | 343 | ||||
| -rw-r--r-- | kernel/net/sunrpc/xprtrdma/module.c | 46 | ||||
| -rw-r--r-- | kernel/net/sunrpc/xprtrdma/physical_ops.c | 31 | ||||
| -rw-r--r-- | kernel/net/sunrpc/xprtrdma/rpc_rdma.c | 353 | ||||
| -rw-r--r-- | kernel/net/sunrpc/xprtrdma/svc_rdma.c | 14 | ||||
| -rw-r--r-- | kernel/net/sunrpc/xprtrdma/svc_rdma_marshal.c | 140 | ||||
| -rw-r--r-- | kernel/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c | 141 | ||||
| -rw-r--r-- | kernel/net/sunrpc/xprtrdma/svc_rdma_sendto.c | 117 | ||||
| -rw-r--r-- | kernel/net/sunrpc/xprtrdma/svc_rdma_transport.c | 256 | ||||
| -rw-r--r-- | kernel/net/sunrpc/xprtrdma/transport.c | 149 | ||||
| -rw-r--r-- | kernel/net/sunrpc/xprtrdma/verbs.c | 1029 | ||||
| -rw-r--r-- | kernel/net/sunrpc/xprtrdma/xprt_rdma.h | 129 |
15 files changed, 1869 insertions, 1408 deletions
diff --git a/kernel/net/sunrpc/xprtrdma/Makefile b/kernel/net/sunrpc/xprtrdma/Makefile index 579f72bbc..33f99d300 100644 --- a/kernel/net/sunrpc/xprtrdma/Makefile +++ b/kernel/net/sunrpc/xprtrdma/Makefile @@ -1,9 +1,8 @@ -obj-$(CONFIG_SUNRPC_XPRT_RDMA_CLIENT) += xprtrdma.o +obj-$(CONFIG_SUNRPC_XPRT_RDMA) += rpcrdma.o -xprtrdma-y := transport.o rpc_rdma.o verbs.o \ - fmr_ops.o frwr_ops.o physical_ops.o - -obj-$(CONFIG_SUNRPC_XPRT_RDMA_SERVER) += svcrdma.o - -svcrdma-y := svc_rdma.o svc_rdma_transport.o \ - svc_rdma_marshal.o svc_rdma_sendto.o svc_rdma_recvfrom.o +rpcrdma-y := transport.o rpc_rdma.o verbs.o \ + fmr_ops.o frwr_ops.o physical_ops.o \ + svc_rdma.o svc_rdma_transport.o \ + svc_rdma_marshal.o svc_rdma_sendto.o svc_rdma_recvfrom.o \ + module.o +rpcrdma-$(CONFIG_SUNRPC_BACKCHANNEL) += backchannel.o diff --git a/kernel/net/sunrpc/xprtrdma/backchannel.c b/kernel/net/sunrpc/xprtrdma/backchannel.c new file mode 100644 index 000000000..2dcb44f69 --- /dev/null +++ b/kernel/net/sunrpc/xprtrdma/backchannel.c @@ -0,0 +1,394 @@ +/* + * Copyright (c) 2015 Oracle. All rights reserved. + * + * Support for backward direction RPCs on RPC/RDMA. + */ + +#include <linux/module.h> +#include <linux/sunrpc/xprt.h> +#include <linux/sunrpc/svc.h> +#include <linux/sunrpc/svc_xprt.h> + +#include "xprt_rdma.h" + +#if IS_ENABLED(CONFIG_SUNRPC_DEBUG) +# define RPCDBG_FACILITY RPCDBG_TRANS +#endif + +#define RPCRDMA_BACKCHANNEL_DEBUG + +static void rpcrdma_bc_free_rqst(struct rpcrdma_xprt *r_xprt, + struct rpc_rqst *rqst) +{ + struct rpcrdma_buffer *buf = &r_xprt->rx_buf; + struct rpcrdma_req *req = rpcr_to_rdmar(rqst); + + spin_lock(&buf->rb_reqslock); + list_del(&req->rl_all); + spin_unlock(&buf->rb_reqslock); + + rpcrdma_destroy_req(&r_xprt->rx_ia, req); + + kfree(rqst); +} + +static int rpcrdma_bc_setup_rqst(struct rpcrdma_xprt *r_xprt, + struct rpc_rqst *rqst) +{ + struct rpcrdma_ia *ia = &r_xprt->rx_ia; + struct rpcrdma_regbuf *rb; + struct rpcrdma_req *req; + struct xdr_buf *buf; + size_t size; + + req = rpcrdma_create_req(r_xprt); + if (!req) + return -ENOMEM; + req->rl_backchannel = true; + + size = RPCRDMA_INLINE_WRITE_THRESHOLD(rqst); + rb = rpcrdma_alloc_regbuf(ia, size, GFP_KERNEL); + if (IS_ERR(rb)) + goto out_fail; + req->rl_rdmabuf = rb; + + size += RPCRDMA_INLINE_READ_THRESHOLD(rqst); + rb = rpcrdma_alloc_regbuf(ia, size, GFP_KERNEL); + if (IS_ERR(rb)) + goto out_fail; + rb->rg_owner = req; + req->rl_sendbuf = rb; + /* so that rpcr_to_rdmar works when receiving a request */ + rqst->rq_buffer = (void *)req->rl_sendbuf->rg_base; + + buf = &rqst->rq_snd_buf; + buf->head[0].iov_base = rqst->rq_buffer; + buf->head[0].iov_len = 0; + buf->tail[0].iov_base = NULL; + buf->tail[0].iov_len = 0; + buf->page_len = 0; + buf->len = 0; + buf->buflen = size; + + return 0; + +out_fail: + rpcrdma_bc_free_rqst(r_xprt, rqst); + return -ENOMEM; +} + +/* Allocate and add receive buffers to the rpcrdma_buffer's + * existing list of rep's. These are released when the + * transport is destroyed. + */ +static int rpcrdma_bc_setup_reps(struct rpcrdma_xprt *r_xprt, + unsigned int count) +{ + struct rpcrdma_buffer *buffers = &r_xprt->rx_buf; + struct rpcrdma_rep *rep; + unsigned long flags; + int rc = 0; + + while (count--) { + rep = rpcrdma_create_rep(r_xprt); + if (IS_ERR(rep)) { + pr_err("RPC: %s: reply buffer alloc failed\n", + __func__); + rc = PTR_ERR(rep); + break; + } + + spin_lock_irqsave(&buffers->rb_lock, flags); + list_add(&rep->rr_list, &buffers->rb_recv_bufs); + spin_unlock_irqrestore(&buffers->rb_lock, flags); + } + + return rc; +} + +/** + * xprt_rdma_bc_setup - Pre-allocate resources for handling backchannel requests + * @xprt: transport associated with these backchannel resources + * @reqs: number of concurrent incoming requests to expect + * + * Returns 0 on success; otherwise a negative errno + */ +int xprt_rdma_bc_setup(struct rpc_xprt *xprt, unsigned int reqs) +{ + struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt); + struct rpcrdma_buffer *buffer = &r_xprt->rx_buf; + struct rpc_rqst *rqst; + unsigned int i; + int rc; + + /* The backchannel reply path returns each rpc_rqst to the + * bc_pa_list _after_ the reply is sent. If the server is + * faster than the client, it can send another backward + * direction request before the rpc_rqst is returned to the + * list. The client rejects the request in this case. + * + * Twice as many rpc_rqsts are prepared to ensure there is + * always an rpc_rqst available as soon as a reply is sent. + */ + if (reqs > RPCRDMA_BACKWARD_WRS >> 1) + goto out_err; + + for (i = 0; i < (reqs << 1); i++) { + rqst = kzalloc(sizeof(*rqst), GFP_KERNEL); + if (!rqst) { + pr_err("RPC: %s: Failed to create bc rpc_rqst\n", + __func__); + goto out_free; + } + + rqst->rq_xprt = &r_xprt->rx_xprt; + INIT_LIST_HEAD(&rqst->rq_list); + INIT_LIST_HEAD(&rqst->rq_bc_list); + + if (rpcrdma_bc_setup_rqst(r_xprt, rqst)) + goto out_free; + + spin_lock_bh(&xprt->bc_pa_lock); + list_add(&rqst->rq_bc_pa_list, &xprt->bc_pa_list); + spin_unlock_bh(&xprt->bc_pa_lock); + } + + rc = rpcrdma_bc_setup_reps(r_xprt, reqs); + if (rc) + goto out_free; + + rc = rpcrdma_ep_post_extra_recv(r_xprt, reqs); + if (rc) + goto out_free; + + buffer->rb_bc_srv_max_requests = reqs; + request_module("svcrdma"); + + return 0; + +out_free: + xprt_rdma_bc_destroy(xprt, reqs); + +out_err: + pr_err("RPC: %s: setup backchannel transport failed\n", __func__); + return -ENOMEM; +} + +/** + * xprt_rdma_bc_up - Create transport endpoint for backchannel service + * @serv: server endpoint + * @net: network namespace + * + * The "xprt" is an implied argument: it supplies the name of the + * backchannel transport class. + * + * Returns zero on success, negative errno on failure + */ +int xprt_rdma_bc_up(struct svc_serv *serv, struct net *net) +{ + int ret; + + ret = svc_create_xprt(serv, "rdma-bc", net, PF_INET, 0, 0); + if (ret < 0) + return ret; + return 0; +} + +/** + * rpcrdma_bc_marshal_reply - Send backwards direction reply + * @rqst: buffer containing RPC reply data + * + * Returns zero on success. + */ +int rpcrdma_bc_marshal_reply(struct rpc_rqst *rqst) +{ + struct rpc_xprt *xprt = rqst->rq_xprt; + struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt); + struct rpcrdma_req *req = rpcr_to_rdmar(rqst); + struct rpcrdma_msg *headerp; + size_t rpclen; + + headerp = rdmab_to_msg(req->rl_rdmabuf); + headerp->rm_xid = rqst->rq_xid; + headerp->rm_vers = rpcrdma_version; + headerp->rm_credit = + cpu_to_be32(r_xprt->rx_buf.rb_bc_srv_max_requests); + headerp->rm_type = rdma_msg; + headerp->rm_body.rm_chunks[0] = xdr_zero; + headerp->rm_body.rm_chunks[1] = xdr_zero; + headerp->rm_body.rm_chunks[2] = xdr_zero; + + rpclen = rqst->rq_svec[0].iov_len; + + pr_info("RPC: %s: rpclen %zd headerp 0x%p lkey 0x%x\n", + __func__, rpclen, headerp, rdmab_lkey(req->rl_rdmabuf)); + pr_info("RPC: %s: RPC/RDMA: %*ph\n", + __func__, (int)RPCRDMA_HDRLEN_MIN, headerp); + pr_info("RPC: %s: RPC: %*ph\n", + __func__, (int)rpclen, rqst->rq_svec[0].iov_base); + + req->rl_send_iov[0].addr = rdmab_addr(req->rl_rdmabuf); + req->rl_send_iov[0].length = RPCRDMA_HDRLEN_MIN; + req->rl_send_iov[0].lkey = rdmab_lkey(req->rl_rdmabuf); + + req->rl_send_iov[1].addr = rdmab_addr(req->rl_sendbuf); + req->rl_send_iov[1].length = rpclen; + req->rl_send_iov[1].lkey = rdmab_lkey(req->rl_sendbuf); + + req->rl_niovs = 2; + return 0; +} + +/** + * xprt_rdma_bc_destroy - Release resources for handling backchannel requests + * @xprt: transport associated with these backchannel resources + * @reqs: number of incoming requests to destroy; ignored + */ +void xprt_rdma_bc_destroy(struct rpc_xprt *xprt, unsigned int reqs) +{ + struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt); + struct rpc_rqst *rqst, *tmp; + + spin_lock_bh(&xprt->bc_pa_lock); + list_for_each_entry_safe(rqst, tmp, &xprt->bc_pa_list, rq_bc_pa_list) { + list_del(&rqst->rq_bc_pa_list); + spin_unlock_bh(&xprt->bc_pa_lock); + + rpcrdma_bc_free_rqst(r_xprt, rqst); + + spin_lock_bh(&xprt->bc_pa_lock); + } + spin_unlock_bh(&xprt->bc_pa_lock); +} + +/** + * xprt_rdma_bc_free_rqst - Release a backchannel rqst + * @rqst: request to release + */ +void xprt_rdma_bc_free_rqst(struct rpc_rqst *rqst) +{ + struct rpc_xprt *xprt = rqst->rq_xprt; + + smp_mb__before_atomic(); + WARN_ON_ONCE(!test_bit(RPC_BC_PA_IN_USE, &rqst->rq_bc_pa_state)); + clear_bit(RPC_BC_PA_IN_USE, &rqst->rq_bc_pa_state); + smp_mb__after_atomic(); + + spin_lock_bh(&xprt->bc_pa_lock); + list_add_tail(&rqst->rq_bc_pa_list, &xprt->bc_pa_list); + spin_unlock_bh(&xprt->bc_pa_lock); +} + +/** + * rpcrdma_bc_receive_call - Handle a backward direction call + * @xprt: transport receiving the call + * @rep: receive buffer containing the call + * + * Called in the RPC reply handler, which runs in a tasklet. + * Be quick about it. + * + * Operational assumptions: + * o Backchannel credits are ignored, just as the NFS server + * forechannel currently does + * o The ULP manages a replay cache (eg, NFSv4.1 sessions). + * No replay detection is done at the transport level + */ +void rpcrdma_bc_receive_call(struct rpcrdma_xprt *r_xprt, + struct rpcrdma_rep *rep) +{ + struct rpc_xprt *xprt = &r_xprt->rx_xprt; + struct rpcrdma_msg *headerp; + struct svc_serv *bc_serv; + struct rpcrdma_req *req; + struct rpc_rqst *rqst; + struct xdr_buf *buf; + size_t size; + __be32 *p; + + headerp = rdmab_to_msg(rep->rr_rdmabuf); +#ifdef RPCRDMA_BACKCHANNEL_DEBUG + pr_info("RPC: %s: callback XID %08x, length=%u\n", + __func__, be32_to_cpu(headerp->rm_xid), rep->rr_len); + pr_info("RPC: %s: %*ph\n", __func__, rep->rr_len, headerp); +#endif + + /* Sanity check: + * Need at least enough bytes for RPC/RDMA header, as code + * here references the header fields by array offset. Also, + * backward calls are always inline, so ensure there + * are some bytes beyond the RPC/RDMA header. + */ + if (rep->rr_len < RPCRDMA_HDRLEN_MIN + 24) + goto out_short; + p = (__be32 *)((unsigned char *)headerp + RPCRDMA_HDRLEN_MIN); + size = rep->rr_len - RPCRDMA_HDRLEN_MIN; + + /* Grab a free bc rqst */ + spin_lock(&xprt->bc_pa_lock); + if (list_empty(&xprt->bc_pa_list)) { + spin_unlock(&xprt->bc_pa_lock); + goto out_overflow; + } + rqst = list_first_entry(&xprt->bc_pa_list, + struct rpc_rqst, rq_bc_pa_list); + list_del(&rqst->rq_bc_pa_list); + spin_unlock(&xprt->bc_pa_lock); +#ifdef RPCRDMA_BACKCHANNEL_DEBUG + pr_info("RPC: %s: using rqst %p\n", __func__, rqst); +#endif + + /* Prepare rqst */ + rqst->rq_reply_bytes_recvd = 0; + rqst->rq_bytes_sent = 0; + rqst->rq_xid = headerp->rm_xid; + set_bit(RPC_BC_PA_IN_USE, &rqst->rq_bc_pa_state); + + buf = &rqst->rq_rcv_buf; + memset(buf, 0, sizeof(*buf)); + buf->head[0].iov_base = p; + buf->head[0].iov_len = size; + buf->len = size; + + /* The receive buffer has to be hooked to the rpcrdma_req + * so that it can be reposted after the server is done + * parsing it but just before sending the backward + * direction reply. + */ + req = rpcr_to_rdmar(rqst); +#ifdef RPCRDMA_BACKCHANNEL_DEBUG + pr_info("RPC: %s: attaching rep %p to req %p\n", + __func__, rep, req); +#endif + req->rl_reply = rep; + + /* Defeat the retransmit detection logic in send_request */ + req->rl_connect_cookie = 0; + + /* Queue rqst for ULP's callback service */ + bc_serv = xprt->bc_serv; + spin_lock(&bc_serv->sv_cb_lock); + list_add(&rqst->rq_bc_list, &bc_serv->sv_cb_list); + spin_unlock(&bc_serv->sv_cb_lock); + + wake_up(&bc_serv->sv_cb_waitq); + + r_xprt->rx_stats.bcall_count++; + return; + +out_overflow: + pr_warn("RPC/RDMA backchannel overflow\n"); + xprt_disconnect_done(xprt); + /* This receive buffer gets reposted automatically + * when the connection is re-established. + */ + return; + +out_short: + pr_warn("RPC/RDMA short backward direction call\n"); + + if (rpcrdma_ep_post_recv(&r_xprt->rx_ia, &r_xprt->rx_ep, rep)) + xprt_disconnect_done(xprt); + else + pr_warn("RPC: %s: reposting rep %p\n", + __func__, rep); +} diff --git a/kernel/net/sunrpc/xprtrdma/fmr_ops.c b/kernel/net/sunrpc/xprtrdma/fmr_ops.c index 302d4ebf6..f1e8dafbd 100644 --- a/kernel/net/sunrpc/xprtrdma/fmr_ops.c +++ b/kernel/net/sunrpc/xprtrdma/fmr_ops.c @@ -11,6 +11,21 @@ * can take tens of usecs to complete. */ +/* Normal operation + * + * A Memory Region is prepared for RDMA READ or WRITE using the + * ib_map_phys_fmr verb (fmr_op_map). When the RDMA operation is + * finished, the Memory Region is unmapped using the ib_unmap_fmr + * verb (fmr_op_unmap). + */ + +/* Transport recovery + * + * After a transport reconnect, fmr_op_map re-uses the MR already + * allocated for the RPC, but generates a fresh rkey then maps the + * MR again. This process is synchronous. + */ + #include "xprt_rdma.h" #if IS_ENABLED(CONFIG_SUNRPC_DEBUG) @@ -50,19 +65,28 @@ fmr_op_init(struct rpcrdma_xprt *r_xprt) struct rpcrdma_mw *r; int i, rc; + spin_lock_init(&buf->rb_mwlock); INIT_LIST_HEAD(&buf->rb_mws); INIT_LIST_HEAD(&buf->rb_all); - i = (buf->rb_max_requests + 1) * RPCRDMA_MAX_SEGS; - dprintk("RPC: %s: initializing %d FMRs\n", __func__, i); + i = max_t(int, RPCRDMA_MAX_DATA_SEGS / RPCRDMA_MAX_FMR_SGES, 1); + i += 2; /* head + tail */ + i *= buf->rb_max_requests; /* one set for each RPC slot */ + dprintk("RPC: %s: initalizing %d FMRs\n", __func__, i); + rc = -ENOMEM; while (i--) { r = kzalloc(sizeof(*r), GFP_KERNEL); if (!r) - return -ENOMEM; + goto out; - r->r.fmr = ib_alloc_fmr(pd, mr_access_flags, &fmr_attr); - if (IS_ERR(r->r.fmr)) + r->r.fmr.physaddrs = kmalloc(RPCRDMA_MAX_FMR_SGES * + sizeof(u64), GFP_KERNEL); + if (!r->r.fmr.physaddrs) + goto out_free; + + r->r.fmr.fmr = ib_alloc_fmr(pd, mr_access_flags, &fmr_attr); + if (IS_ERR(r->r.fmr.fmr)) goto out_fmr_err; list_add(&r->mw_list, &buf->rb_mws); @@ -71,12 +95,24 @@ fmr_op_init(struct rpcrdma_xprt *r_xprt) return 0; out_fmr_err: - rc = PTR_ERR(r->r.fmr); + rc = PTR_ERR(r->r.fmr.fmr); dprintk("RPC: %s: ib_alloc_fmr status %i\n", __func__, rc); + kfree(r->r.fmr.physaddrs); +out_free: kfree(r); +out: return rc; } +static int +__fmr_unmap(struct rpcrdma_mw *r) +{ + LIST_HEAD(l); + + list_add(&r->r.fmr.fmr->list, &l); + return ib_unmap_fmr(&l); +} + /* Use the ib_map_phys_fmr() verb to register a memory region * for remote access via RDMA READ or RDMA WRITE. */ @@ -85,12 +121,24 @@ fmr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg, int nsegs, bool writing) { struct rpcrdma_ia *ia = &r_xprt->rx_ia; - struct ib_device *device = ia->ri_id->device; + struct ib_device *device = ia->ri_device; enum dma_data_direction direction = rpcrdma_data_dir(writing); struct rpcrdma_mr_seg *seg1 = seg; - struct rpcrdma_mw *mw = seg1->rl_mw; - u64 physaddrs[RPCRDMA_MAX_DATA_SEGS]; int len, pageoff, i, rc; + struct rpcrdma_mw *mw; + + mw = seg1->rl_mw; + seg1->rl_mw = NULL; + if (!mw) { + mw = rpcrdma_get_mw(r_xprt); + if (!mw) + return -ENOMEM; + } else { + /* this is a retransmit; generate a fresh rkey */ + rc = __fmr_unmap(mw); + if (rc) + return rc; + } pageoff = offset_in_page(seg1->mr_offset); seg1->mr_offset -= pageoff; /* start of page */ @@ -100,7 +148,7 @@ fmr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg, nsegs = RPCRDMA_MAX_FMR_SGES; for (i = 0; i < nsegs;) { rpcrdma_map_one(device, seg, direction); - physaddrs[i] = seg->mr_dma; + mw->r.fmr.physaddrs[i] = seg->mr_dma; len += seg->mr_len; ++seg; ++i; @@ -110,11 +158,13 @@ fmr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg, break; } - rc = ib_map_phys_fmr(mw->r.fmr, physaddrs, i, seg1->mr_dma); + rc = ib_map_phys_fmr(mw->r.fmr.fmr, mw->r.fmr.physaddrs, + i, seg1->mr_dma); if (rc) goto out_maperr; - seg1->mr_rkey = mw->r.fmr->rkey; + seg1->rl_mw = mw; + seg1->mr_rkey = mw->r.fmr.fmr->rkey; seg1->mr_base = seg1->mr_dma + pageoff; seg1->mr_nsegs = i; seg1->mr_len = len; @@ -137,48 +187,28 @@ fmr_op_unmap(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg) { struct rpcrdma_ia *ia = &r_xprt->rx_ia; struct rpcrdma_mr_seg *seg1 = seg; - struct ib_device *device; + struct rpcrdma_mw *mw = seg1->rl_mw; int rc, nsegs = seg->mr_nsegs; - LIST_HEAD(l); - list_add(&seg1->rl_mw->r.fmr->list, &l); - rc = ib_unmap_fmr(&l); - read_lock(&ia->ri_qplock); - device = ia->ri_id->device; + dprintk("RPC: %s: FMR %p\n", __func__, mw); + + seg1->rl_mw = NULL; while (seg1->mr_nsegs--) - rpcrdma_unmap_one(device, seg++); - read_unlock(&ia->ri_qplock); + rpcrdma_unmap_one(ia->ri_device, seg++); + rc = __fmr_unmap(mw); if (rc) goto out_err; + rpcrdma_put_mw(r_xprt, mw); return nsegs; out_err: + /* The FMR is abandoned, but remains in rb_all. fmr_op_destroy + * will attempt to release it when the transport is destroyed. + */ dprintk("RPC: %s: ib_unmap_fmr status %i\n", __func__, rc); return nsegs; } -/* After a disconnect, unmap all FMRs. - * - * This is invoked only in the transport connect worker in order - * to serialize with rpcrdma_register_fmr_external(). - */ -static void -fmr_op_reset(struct rpcrdma_xprt *r_xprt) -{ - struct rpcrdma_buffer *buf = &r_xprt->rx_buf; - struct rpcrdma_mw *r; - LIST_HEAD(list); - int rc; - - list_for_each_entry(r, &buf->rb_all, mw_all) - list_add(&r->r.fmr->list, &list); - - rc = ib_unmap_fmr(&list); - if (rc) - dprintk("RPC: %s: ib_unmap_fmr failed %i\n", - __func__, rc); -} - static void fmr_op_destroy(struct rpcrdma_buffer *buf) { @@ -188,10 +218,13 @@ fmr_op_destroy(struct rpcrdma_buffer *buf) while (!list_empty(&buf->rb_all)) { r = list_entry(buf->rb_all.next, struct rpcrdma_mw, mw_all); list_del(&r->mw_all); - rc = ib_dealloc_fmr(r->r.fmr); + kfree(r->r.fmr.physaddrs); + + rc = ib_dealloc_fmr(r->r.fmr.fmr); if (rc) dprintk("RPC: %s: ib_dealloc_fmr failed %i\n", __func__, rc); + kfree(r); } } @@ -202,7 +235,6 @@ const struct rpcrdma_memreg_ops rpcrdma_fmr_memreg_ops = { .ro_open = fmr_op_open, .ro_maxpages = fmr_op_maxpages, .ro_init = fmr_op_init, - .ro_reset = fmr_op_reset, .ro_destroy = fmr_op_destroy, .ro_displayname = "fmr", }; diff --git a/kernel/net/sunrpc/xprtrdma/frwr_ops.c b/kernel/net/sunrpc/xprtrdma/frwr_ops.c index dff0481db..88cf9e726 100644 --- a/kernel/net/sunrpc/xprtrdma/frwr_ops.c +++ b/kernel/net/sunrpc/xprtrdma/frwr_ops.c @@ -11,12 +11,136 @@ * but most complex memory registration mode. */ +/* Normal operation + * + * A Memory Region is prepared for RDMA READ or WRITE using a FAST_REG + * Work Request (frmr_op_map). When the RDMA operation is finished, this + * Memory Region is invalidated using a LOCAL_INV Work Request + * (frmr_op_unmap). + * + * Typically these Work Requests are not signaled, and neither are RDMA + * SEND Work Requests (with the exception of signaling occasionally to + * prevent provider work queue overflows). This greatly reduces HCA + * interrupt workload. + * + * As an optimization, frwr_op_unmap marks MRs INVALID before the + * LOCAL_INV WR is posted. If posting succeeds, the MR is placed on + * rb_mws immediately so that no work (like managing a linked list + * under a spinlock) is needed in the completion upcall. + * + * But this means that frwr_op_map() can occasionally encounter an MR + * that is INVALID but the LOCAL_INV WR has not completed. Work Queue + * ordering prevents a subsequent FAST_REG WR from executing against + * that MR while it is still being invalidated. + */ + +/* Transport recovery + * + * ->op_map and the transport connect worker cannot run at the same + * time, but ->op_unmap can fire while the transport connect worker + * is running. Thus MR recovery is handled in ->op_map, to guarantee + * that recovered MRs are owned by a sending RPC, and not one where + * ->op_unmap could fire at the same time transport reconnect is + * being done. + * + * When the underlying transport disconnects, MRs are left in one of + * three states: + * + * INVALID: The MR was not in use before the QP entered ERROR state. + * (Or, the LOCAL_INV WR has not completed or flushed yet). + * + * STALE: The MR was being registered or unregistered when the QP + * entered ERROR state, and the pending WR was flushed. + * + * VALID: The MR was registered before the QP entered ERROR state. + * + * When frwr_op_map encounters STALE and VALID MRs, they are recovered + * with ib_dereg_mr and then are re-initialized. Beause MR recovery + * allocates fresh resources, it is deferred to a workqueue, and the + * recovered MRs are placed back on the rb_mws list when recovery is + * complete. frwr_op_map allocates another MR for the current RPC while + * the broken MR is reset. + * + * To ensure that frwr_op_map doesn't encounter an MR that is marked + * INVALID but that is about to be flushed due to a previous transport + * disconnect, the transport connect worker attempts to drain all + * pending send queue WRs before the transport is reconnected. + */ + #include "xprt_rdma.h" #if IS_ENABLED(CONFIG_SUNRPC_DEBUG) # define RPCDBG_FACILITY RPCDBG_TRANS #endif +static struct workqueue_struct *frwr_recovery_wq; + +#define FRWR_RECOVERY_WQ_FLAGS (WQ_UNBOUND | WQ_MEM_RECLAIM) + +int +frwr_alloc_recovery_wq(void) +{ + frwr_recovery_wq = alloc_workqueue("frwr_recovery", + FRWR_RECOVERY_WQ_FLAGS, 0); + return !frwr_recovery_wq ? -ENOMEM : 0; +} + +void +frwr_destroy_recovery_wq(void) +{ + struct workqueue_struct *wq; + + if (!frwr_recovery_wq) + return; + + wq = frwr_recovery_wq; + frwr_recovery_wq = NULL; + destroy_workqueue(wq); +} + +/* Deferred reset of a single FRMR. Generate a fresh rkey by + * replacing the MR. + * + * There's no recovery if this fails. The FRMR is abandoned, but + * remains in rb_all. It will be cleaned up when the transport is + * destroyed. + */ +static void +__frwr_recovery_worker(struct work_struct *work) +{ + struct rpcrdma_mw *r = container_of(work, struct rpcrdma_mw, + r.frmr.fr_work); + struct rpcrdma_xprt *r_xprt = r->r.frmr.fr_xprt; + unsigned int depth = r_xprt->rx_ia.ri_max_frmr_depth; + struct ib_pd *pd = r_xprt->rx_ia.ri_pd; + + if (ib_dereg_mr(r->r.frmr.fr_mr)) + goto out_fail; + + r->r.frmr.fr_mr = ib_alloc_mr(pd, IB_MR_TYPE_MEM_REG, depth); + if (IS_ERR(r->r.frmr.fr_mr)) + goto out_fail; + + dprintk("RPC: %s: recovered FRMR %p\n", __func__, r); + r->r.frmr.fr_state = FRMR_IS_INVALID; + rpcrdma_put_mw(r_xprt, r); + return; + +out_fail: + pr_warn("RPC: %s: FRMR %p unrecovered\n", + __func__, r); +} + +/* A broken MR was discovered in a context that can't sleep. + * Defer recovery to the recovery worker. + */ +static void +__frwr_queue_recovery(struct rpcrdma_mw *r) +{ + INIT_WORK(&r->r.frmr.fr_work, __frwr_recovery_worker); + queue_work(frwr_recovery_wq, &r->r.frmr.fr_work); +} + static int __frwr_init(struct rpcrdma_mw *r, struct ib_pd *pd, struct ib_device *device, unsigned int depth) @@ -24,24 +148,28 @@ __frwr_init(struct rpcrdma_mw *r, struct ib_pd *pd, struct ib_device *device, struct rpcrdma_frmr *f = &r->r.frmr; int rc; - f->fr_mr = ib_alloc_fast_reg_mr(pd, depth); + f->fr_mr = ib_alloc_mr(pd, IB_MR_TYPE_MEM_REG, depth); if (IS_ERR(f->fr_mr)) goto out_mr_err; - f->fr_pgl = ib_alloc_fast_reg_page_list(device, depth); - if (IS_ERR(f->fr_pgl)) + + f->sg = kcalloc(depth, sizeof(*f->sg), GFP_KERNEL); + if (!f->sg) goto out_list_err; + + sg_init_table(f->sg, depth); + return 0; out_mr_err: rc = PTR_ERR(f->fr_mr); - dprintk("RPC: %s: ib_alloc_fast_reg_mr status %i\n", + dprintk("RPC: %s: ib_alloc_mr status %i\n", __func__, rc); return rc; out_list_err: - rc = PTR_ERR(f->fr_pgl); - dprintk("RPC: %s: ib_alloc_fast_reg_page_list status %i\n", - __func__, rc); + rc = -ENOMEM; + dprintk("RPC: %s: sg allocation failure\n", + __func__); ib_dereg_mr(f->fr_mr); return rc; } @@ -55,7 +183,7 @@ __frwr_release(struct rpcrdma_mw *r) if (rc) dprintk("RPC: %s: ib_dereg_mr status %i\n", __func__, rc); - ib_free_fast_reg_page_list(r->r.frmr.fr_pgl); + kfree(r->r.frmr.sg); } static int @@ -128,8 +256,11 @@ frwr_sendcompletion(struct ib_wc *wc) /* WARNING: Only wr_id and status are reliable at this point */ r = (struct rpcrdma_mw *)(unsigned long)wc->wr_id; - dprintk("RPC: %s: frmr %p (stale), status %d\n", - __func__, r, wc->status); + if (wc->status == IB_WC_WR_FLUSH_ERR) + dprintk("RPC: %s: frmr %p flushed\n", __func__, r); + else + pr_warn("RPC: %s: frmr %p error, status %s (%d)\n", + __func__, r, ib_wc_status_msg(wc->status), wc->status); r->r.frmr.fr_state = FRMR_IS_STALE; } @@ -137,16 +268,19 @@ static int frwr_op_init(struct rpcrdma_xprt *r_xprt) { struct rpcrdma_buffer *buf = &r_xprt->rx_buf; - struct ib_device *device = r_xprt->rx_ia.ri_id->device; + struct ib_device *device = r_xprt->rx_ia.ri_device; unsigned int depth = r_xprt->rx_ia.ri_max_frmr_depth; struct ib_pd *pd = r_xprt->rx_ia.ri_pd; int i; + spin_lock_init(&buf->rb_mwlock); INIT_LIST_HEAD(&buf->rb_mws); INIT_LIST_HEAD(&buf->rb_all); - i = (buf->rb_max_requests + 1) * RPCRDMA_MAX_SEGS; - dprintk("RPC: %s: initializing %d FRMRs\n", __func__, i); + i = max_t(int, RPCRDMA_MAX_DATA_SEGS / depth, 1); + i += 2; /* head + tail */ + i *= buf->rb_max_requests; /* one set for each RPC slot */ + dprintk("RPC: %s: initalizing %d FRMRs\n", __func__, i); while (i--) { struct rpcrdma_mw *r; @@ -165,6 +299,7 @@ frwr_op_init(struct rpcrdma_xprt *r_xprt) list_add(&r->mw_list, &buf->rb_mws); list_add(&r->mw_all, &buf->rb_all); r->mw_sendcompletion = frwr_sendcompletion; + r->r.frmr.fr_xprt = r_xprt; } return 0; @@ -178,78 +313,103 @@ frwr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg, int nsegs, bool writing) { struct rpcrdma_ia *ia = &r_xprt->rx_ia; - struct ib_device *device = ia->ri_id->device; + struct ib_device *device = ia->ri_device; enum dma_data_direction direction = rpcrdma_data_dir(writing); struct rpcrdma_mr_seg *seg1 = seg; - struct rpcrdma_mw *mw = seg1->rl_mw; - struct rpcrdma_frmr *frmr = &mw->r.frmr; - struct ib_mr *mr = frmr->fr_mr; - struct ib_send_wr fastreg_wr, *bad_wr; + struct rpcrdma_mw *mw; + struct rpcrdma_frmr *frmr; + struct ib_mr *mr; + struct ib_reg_wr reg_wr; + struct ib_send_wr *bad_wr; + int rc, i, n, dma_nents; u8 key; - int len, pageoff; - int i, rc; - int seg_len; - u64 pa; - int page_no; - - pageoff = offset_in_page(seg1->mr_offset); - seg1->mr_offset -= pageoff; /* start of page */ - seg1->mr_len += pageoff; - len = -pageoff; + + mw = seg1->rl_mw; + seg1->rl_mw = NULL; + do { + if (mw) + __frwr_queue_recovery(mw); + mw = rpcrdma_get_mw(r_xprt); + if (!mw) + return -ENOMEM; + } while (mw->r.frmr.fr_state != FRMR_IS_INVALID); + frmr = &mw->r.frmr; + frmr->fr_state = FRMR_IS_VALID; + mr = frmr->fr_mr; + if (nsegs > ia->ri_max_frmr_depth) nsegs = ia->ri_max_frmr_depth; - for (page_no = i = 0; i < nsegs;) { - rpcrdma_map_one(device, seg, direction); - pa = seg->mr_dma; - for (seg_len = seg->mr_len; seg_len > 0; seg_len -= PAGE_SIZE) { - frmr->fr_pgl->page_list[page_no++] = pa; - pa += PAGE_SIZE; - } - len += seg->mr_len; + + for (i = 0; i < nsegs;) { + if (seg->mr_page) + sg_set_page(&frmr->sg[i], + seg->mr_page, + seg->mr_len, + offset_in_page(seg->mr_offset)); + else + sg_set_buf(&frmr->sg[i], seg->mr_offset, + seg->mr_len); + ++seg; ++i; + /* Check for holes */ if ((i < nsegs && offset_in_page(seg->mr_offset)) || offset_in_page((seg-1)->mr_offset + (seg-1)->mr_len)) break; } - dprintk("RPC: %s: Using frmr %p to map %d segments (%d bytes)\n", - __func__, mw, i, len); + frmr->sg_nents = i; - frmr->fr_state = FRMR_IS_VALID; + dma_nents = ib_dma_map_sg(device, frmr->sg, frmr->sg_nents, direction); + if (!dma_nents) { + pr_err("RPC: %s: failed to dma map sg %p sg_nents %u\n", + __func__, frmr->sg, frmr->sg_nents); + return -ENOMEM; + } + + n = ib_map_mr_sg(mr, frmr->sg, frmr->sg_nents, PAGE_SIZE); + if (unlikely(n != frmr->sg_nents)) { + pr_err("RPC: %s: failed to map mr %p (%u/%u)\n", + __func__, frmr->fr_mr, n, frmr->sg_nents); + rc = n < 0 ? n : -EINVAL; + goto out_senderr; + } + + dprintk("RPC: %s: Using frmr %p to map %u segments (%u bytes)\n", + __func__, mw, frmr->sg_nents, mr->length); - memset(&fastreg_wr, 0, sizeof(fastreg_wr)); - fastreg_wr.wr_id = (unsigned long)(void *)mw; - fastreg_wr.opcode = IB_WR_FAST_REG_MR; - fastreg_wr.wr.fast_reg.iova_start = seg1->mr_dma + pageoff; - fastreg_wr.wr.fast_reg.page_list = frmr->fr_pgl; - fastreg_wr.wr.fast_reg.page_shift = PAGE_SHIFT; - fastreg_wr.wr.fast_reg.page_list_len = page_no; - fastreg_wr.wr.fast_reg.length = len; - fastreg_wr.wr.fast_reg.access_flags = writing ? - IB_ACCESS_REMOTE_WRITE | IB_ACCESS_LOCAL_WRITE : - IB_ACCESS_REMOTE_READ; key = (u8)(mr->rkey & 0x000000FF); ib_update_fast_reg_key(mr, ++key); - fastreg_wr.wr.fast_reg.rkey = mr->rkey; + + reg_wr.wr.next = NULL; + reg_wr.wr.opcode = IB_WR_REG_MR; + reg_wr.wr.wr_id = (uintptr_t)mw; + reg_wr.wr.num_sge = 0; + reg_wr.wr.send_flags = 0; + reg_wr.mr = mr; + reg_wr.key = mr->rkey; + reg_wr.access = writing ? + IB_ACCESS_REMOTE_WRITE | IB_ACCESS_LOCAL_WRITE : + IB_ACCESS_REMOTE_READ; DECR_CQCOUNT(&r_xprt->rx_ep); - rc = ib_post_send(ia->ri_id->qp, &fastreg_wr, &bad_wr); + rc = ib_post_send(ia->ri_id->qp, ®_wr.wr, &bad_wr); if (rc) goto out_senderr; + seg1->mr_dir = direction; + seg1->rl_mw = mw; seg1->mr_rkey = mr->rkey; - seg1->mr_base = seg1->mr_dma + pageoff; - seg1->mr_nsegs = i; - seg1->mr_len = len; - return i; + seg1->mr_base = mr->iova; + seg1->mr_nsegs = frmr->sg_nents; + seg1->mr_len = mr->length; + + return frmr->sg_nents; out_senderr: dprintk("RPC: %s: ib_post_send status %i\n", __func__, rc); - ib_update_fast_reg_key(mr, --key); - frmr->fr_state = FRMR_IS_INVALID; - while (i--) - rpcrdma_unmap_one(device, --seg); + ib_dma_unmap_sg(device, frmr->sg, dma_nents, direction); + __frwr_queue_recovery(mw); return rc; } @@ -261,78 +421,46 @@ frwr_op_unmap(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg) { struct rpcrdma_mr_seg *seg1 = seg; struct rpcrdma_ia *ia = &r_xprt->rx_ia; + struct rpcrdma_mw *mw = seg1->rl_mw; + struct rpcrdma_frmr *frmr = &mw->r.frmr; struct ib_send_wr invalidate_wr, *bad_wr; int rc, nsegs = seg->mr_nsegs; - struct ib_device *device; - seg1->rl_mw->r.frmr.fr_state = FRMR_IS_INVALID; + dprintk("RPC: %s: FRMR %p\n", __func__, mw); + + seg1->rl_mw = NULL; + frmr->fr_state = FRMR_IS_INVALID; memset(&invalidate_wr, 0, sizeof(invalidate_wr)); - invalidate_wr.wr_id = (unsigned long)(void *)seg1->rl_mw; + invalidate_wr.wr_id = (unsigned long)(void *)mw; invalidate_wr.opcode = IB_WR_LOCAL_INV; - invalidate_wr.ex.invalidate_rkey = seg1->rl_mw->r.frmr.fr_mr->rkey; + invalidate_wr.ex.invalidate_rkey = frmr->fr_mr->rkey; DECR_CQCOUNT(&r_xprt->rx_ep); + ib_dma_unmap_sg(ia->ri_device, frmr->sg, frmr->sg_nents, seg1->mr_dir); read_lock(&ia->ri_qplock); - device = ia->ri_id->device; - while (seg1->mr_nsegs--) - rpcrdma_unmap_one(device, seg++); rc = ib_post_send(ia->ri_id->qp, &invalidate_wr, &bad_wr); read_unlock(&ia->ri_qplock); if (rc) goto out_err; + + rpcrdma_put_mw(r_xprt, mw); return nsegs; out_err: - /* Force rpcrdma_buffer_get() to retry */ - seg1->rl_mw->r.frmr.fr_state = FRMR_IS_STALE; dprintk("RPC: %s: ib_post_send status %i\n", __func__, rc); + __frwr_queue_recovery(mw); return nsegs; } -/* After a disconnect, a flushed FAST_REG_MR can leave an FRMR in - * an unusable state. Find FRMRs in this state and dereg / reg - * each. FRMRs that are VALID and attached to an rpcrdma_req are - * also torn down. - * - * This gives all in-use FRMRs a fresh rkey and leaves them INVALID. - * - * This is invoked only in the transport connect worker in order - * to serialize with rpcrdma_register_frmr_external(). - */ -static void -frwr_op_reset(struct rpcrdma_xprt *r_xprt) -{ - struct rpcrdma_buffer *buf = &r_xprt->rx_buf; - struct ib_device *device = r_xprt->rx_ia.ri_id->device; - unsigned int depth = r_xprt->rx_ia.ri_max_frmr_depth; - struct ib_pd *pd = r_xprt->rx_ia.ri_pd; - struct rpcrdma_mw *r; - int rc; - - list_for_each_entry(r, &buf->rb_all, mw_all) { - if (r->r.frmr.fr_state == FRMR_IS_INVALID) - continue; - - __frwr_release(r); - rc = __frwr_init(r, pd, device, depth); - if (rc) { - dprintk("RPC: %s: mw %p left %s\n", - __func__, r, - (r->r.frmr.fr_state == FRMR_IS_STALE ? - "stale" : "valid")); - continue; - } - - r->r.frmr.fr_state = FRMR_IS_INVALID; - } -} - static void frwr_op_destroy(struct rpcrdma_buffer *buf) { struct rpcrdma_mw *r; + /* Ensure stale MWs for "buf" are no longer in flight */ + flush_workqueue(frwr_recovery_wq); + while (!list_empty(&buf->rb_all)) { r = list_entry(buf->rb_all.next, struct rpcrdma_mw, mw_all); list_del(&r->mw_all); @@ -347,7 +475,6 @@ const struct rpcrdma_memreg_ops rpcrdma_frwr_memreg_ops = { .ro_open = frwr_op_open, .ro_maxpages = frwr_op_maxpages, .ro_init = frwr_op_init, - .ro_reset = frwr_op_reset, .ro_destroy = frwr_op_destroy, .ro_displayname = "frwr", }; diff --git a/kernel/net/sunrpc/xprtrdma/module.c b/kernel/net/sunrpc/xprtrdma/module.c new file mode 100644 index 000000000..560712bd9 --- /dev/null +++ b/kernel/net/sunrpc/xprtrdma/module.c @@ -0,0 +1,46 @@ +/* + * Copyright (c) 2015 Oracle. All rights reserved. + */ + +/* rpcrdma.ko module initialization + */ + +#include <linux/module.h> +#include <linux/init.h> +#include <linux/sunrpc/svc_rdma.h> +#include "xprt_rdma.h" + +#if IS_ENABLED(CONFIG_SUNRPC_DEBUG) +# define RPCDBG_FACILITY RPCDBG_TRANS +#endif + +MODULE_AUTHOR("Open Grid Computing and Network Appliance, Inc."); +MODULE_DESCRIPTION("RPC/RDMA Transport"); +MODULE_LICENSE("Dual BSD/GPL"); +MODULE_ALIAS("svcrdma"); +MODULE_ALIAS("xprtrdma"); + +static void __exit rpc_rdma_cleanup(void) +{ + xprt_rdma_cleanup(); + svc_rdma_cleanup(); +} + +static int __init rpc_rdma_init(void) +{ + int rc; + + rc = svc_rdma_init(); + if (rc) + goto out; + + rc = xprt_rdma_init(); + if (rc) + svc_rdma_cleanup(); + +out: + return rc; +} + +module_init(rpc_rdma_init); +module_exit(rpc_rdma_cleanup); diff --git a/kernel/net/sunrpc/xprtrdma/physical_ops.c b/kernel/net/sunrpc/xprtrdma/physical_ops.c index ba518af16..617b76f22 100644 --- a/kernel/net/sunrpc/xprtrdma/physical_ops.c +++ b/kernel/net/sunrpc/xprtrdma/physical_ops.c @@ -23,6 +23,21 @@ static int physical_op_open(struct rpcrdma_ia *ia, struct rpcrdma_ep *ep, struct rpcrdma_create_data_internal *cdata) { + struct ib_mr *mr; + + /* Obtain an rkey to use for RPC data payloads. + */ + mr = ib_get_dma_mr(ia->ri_pd, + IB_ACCESS_LOCAL_WRITE | + IB_ACCESS_REMOTE_WRITE | + IB_ACCESS_REMOTE_READ); + if (IS_ERR(mr)) { + pr_err("%s: ib_get_dma_mr for failed with %lX\n", + __func__, PTR_ERR(mr)); + return -ENOMEM; + } + + ia->ri_dma_mr = mr; return 0; } @@ -50,9 +65,8 @@ physical_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg, { struct rpcrdma_ia *ia = &r_xprt->rx_ia; - rpcrdma_map_one(ia->ri_id->device, seg, - rpcrdma_data_dir(writing)); - seg->mr_rkey = ia->ri_bind_mem->rkey; + rpcrdma_map_one(ia->ri_device, seg, rpcrdma_data_dir(writing)); + seg->mr_rkey = ia->ri_dma_mr->rkey; seg->mr_base = seg->mr_dma; seg->mr_nsegs = 1; return 1; @@ -65,19 +79,11 @@ physical_op_unmap(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg) { struct rpcrdma_ia *ia = &r_xprt->rx_ia; - read_lock(&ia->ri_qplock); - rpcrdma_unmap_one(ia->ri_id->device, seg); - read_unlock(&ia->ri_qplock); - + rpcrdma_unmap_one(ia->ri_device, seg); return 1; } static void -physical_op_reset(struct rpcrdma_xprt *r_xprt) -{ -} - -static void physical_op_destroy(struct rpcrdma_buffer *buf) { } @@ -88,7 +94,6 @@ const struct rpcrdma_memreg_ops rpcrdma_physical_memreg_ops = { .ro_open = physical_op_open, .ro_maxpages = physical_op_maxpages, .ro_init = physical_op_init, - .ro_reset = physical_op_reset, .ro_destroy = physical_op_destroy, .ro_displayname = "physical", }; diff --git a/kernel/net/sunrpc/xprtrdma/rpc_rdma.c b/kernel/net/sunrpc/xprtrdma/rpc_rdma.c index 2c53ea9e1..c10d96994 100644 --- a/kernel/net/sunrpc/xprtrdma/rpc_rdma.c +++ b/kernel/net/sunrpc/xprtrdma/rpc_rdma.c @@ -71,6 +71,67 @@ static const char transfertypes[][12] = { }; #endif +/* The client can send a request inline as long as the RPCRDMA header + * plus the RPC call fit under the transport's inline limit. If the + * combined call message size exceeds that limit, the client must use + * the read chunk list for this operation. + */ +static bool rpcrdma_args_inline(struct rpc_rqst *rqst) +{ + unsigned int callsize = RPCRDMA_HDRLEN_MIN + rqst->rq_snd_buf.len; + + return callsize <= RPCRDMA_INLINE_WRITE_THRESHOLD(rqst); +} + +/* The client can't know how large the actual reply will be. Thus it + * plans for the largest possible reply for that particular ULP + * operation. If the maximum combined reply message size exceeds that + * limit, the client must provide a write list or a reply chunk for + * this request. + */ +static bool rpcrdma_results_inline(struct rpc_rqst *rqst) +{ + unsigned int repsize = RPCRDMA_HDRLEN_MIN + rqst->rq_rcv_buf.buflen; + + return repsize <= RPCRDMA_INLINE_READ_THRESHOLD(rqst); +} + +static int +rpcrdma_tail_pullup(struct xdr_buf *buf) +{ + size_t tlen = buf->tail[0].iov_len; + size_t skip = tlen & 3; + + /* Do not include the tail if it is only an XDR pad */ + if (tlen < 4) + return 0; + + /* xdr_write_pages() adds a pad at the beginning of the tail + * if the content in "buf->pages" is unaligned. Force the + * tail's actual content to land at the next XDR position + * after the head instead. + */ + if (skip) { + unsigned char *src, *dst; + unsigned int count; + + src = buf->tail[0].iov_base; + dst = buf->head[0].iov_base; + dst += buf->head[0].iov_len; + + src += skip; + tlen -= skip; + + dprintk("RPC: %s: skip=%zu, memmove(%p, %p, %zu)\n", + __func__, skip, dst, src, tlen); + + for (count = tlen; count; count--) + *dst++ = *src++; + } + + return tlen; +} + /* * Chunk assembly from upper layer xdr_buf. * @@ -122,6 +183,10 @@ rpcrdma_convert_iovs(struct xdr_buf *xdrbuf, unsigned int pos, if (len && n == nsegs) return -EIO; + /* When encoding the read list, the tail is always sent inline */ + if (type == rpcrdma_readch) + return n; + if (xdrbuf->tail[0].iov_len) { /* the rpcrdma protocol allows us to omit any trailing * xdr pad bytes, saving the server an RDMA operation. */ @@ -284,9 +349,6 @@ rpcrdma_create_chunks(struct rpc_rqst *rqst, struct xdr_buf *target, return (unsigned char *)iptr - (unsigned char *)headerp; out: - if (r_xprt->rx_ia.ri_memreg_strategy == RPCRDMA_FRMR) - return n; - for (pos = 0; nchunks--;) pos += r_xprt->rx_ia.ri_ops->ro_unmap(r_xprt, &req->rl_segments[pos]); @@ -300,8 +362,7 @@ out: * pre-registered memory buffer for this request. For small amounts * of data, this is efficient. The cutoff value is tunable. */ -static int -rpcrdma_inline_pullup(struct rpc_rqst *rqst, int pad) +static void rpcrdma_inline_pullup(struct rpc_rqst *rqst) { int i, npages, curlen; int copy_len; @@ -313,16 +374,9 @@ rpcrdma_inline_pullup(struct rpc_rqst *rqst, int pad) destp = rqst->rq_svec[0].iov_base; curlen = rqst->rq_svec[0].iov_len; destp += curlen; - /* - * Do optional padding where it makes sense. Alignment of write - * payload can help the server, if our setting is accurate. - */ - pad -= (curlen + 36/*sizeof(struct rpcrdma_msg_padded)*/); - if (pad < 0 || rqst->rq_slen - curlen < RPCRDMA_INLINE_PAD_THRESH) - pad = 0; /* don't pad this request */ - dprintk("RPC: %s: pad %d destp 0x%p len %d hdrlen %d\n", - __func__, pad, destp, rqst->rq_slen, curlen); + dprintk("RPC: %s: destp 0x%p len %d hdrlen %d\n", + __func__, destp, rqst->rq_slen, curlen); copy_len = rqst->rq_snd_buf.page_len; @@ -358,7 +412,6 @@ rpcrdma_inline_pullup(struct rpc_rqst *rqst, int pad) page_base = 0; } /* header now contains entire send message */ - return pad; } /* @@ -383,11 +436,16 @@ rpcrdma_marshal_req(struct rpc_rqst *rqst) struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt); struct rpcrdma_req *req = rpcr_to_rdmar(rqst); char *base; - size_t rpclen, padlen; + size_t rpclen; ssize_t hdrlen; enum rpcrdma_chunktype rtype, wtype; struct rpcrdma_msg *headerp; +#if defined(CONFIG_SUNRPC_BACKCHANNEL) + if (test_bit(RPC_BC_PA_IN_USE, &rqst->rq_bc_pa_state)) + return rpcrdma_bc_marshal_reply(rqst); +#endif + /* * rpclen gets amount of data in first buffer, which is the * pre-registered buffer. @@ -405,28 +463,15 @@ rpcrdma_marshal_req(struct rpc_rqst *rqst) /* * Chunks needed for results? * + * o Read ops return data as write chunk(s), header as inline. * o If the expected result is under the inline threshold, all ops - * return as inline (but see later). + * return as inline. * o Large non-read ops return as a single reply chunk. - * o Large read ops return data as write chunk(s), header as inline. - * - * Note: the NFS code sending down multiple result segments implies - * the op is one of read, readdir[plus], readlink or NFSv4 getacl. - */ - - /* - * This code can handle read chunks, write chunks OR reply - * chunks -- only one type. If the request is too big to fit - * inline, then we will choose read chunks. If the request is - * a READ, then use write chunks to separate the file data - * into pages; otherwise use reply chunks. */ - if (rqst->rq_rcv_buf.buflen <= RPCRDMA_INLINE_READ_THRESHOLD(rqst)) - wtype = rpcrdma_noch; - else if (rqst->rq_rcv_buf.page_len == 0) - wtype = rpcrdma_replych; - else if (rqst->rq_rcv_buf.flags & XDRBUF_READ) + if (rqst->rq_rcv_buf.flags & XDRBUF_READ) wtype = rpcrdma_writech; + else if (rpcrdma_results_inline(rqst)) + wtype = rpcrdma_noch; else wtype = rpcrdma_replych; @@ -435,21 +480,25 @@ rpcrdma_marshal_req(struct rpc_rqst *rqst) * * o If the total request is under the inline threshold, all ops * are sent as inline. - * o Large non-write ops are sent with the entire message as a - * single read chunk (protocol 0-position special case). * o Large write ops transmit data as read chunk(s), header as * inline. + * o Large non-write ops are sent with the entire message as a + * single read chunk (protocol 0-position special case). * - * Note: the NFS code sending down multiple argument segments - * implies the op is a write. - * TBD check NFSv4 setacl + * This assumes that the upper layer does not present a request + * that both has a data payload, and whose non-data arguments + * by themselves are larger than the inline threshold. */ - if (rqst->rq_snd_buf.len <= RPCRDMA_INLINE_WRITE_THRESHOLD(rqst)) + if (rpcrdma_args_inline(rqst)) { rtype = rpcrdma_noch; - else if (rqst->rq_snd_buf.page_len == 0) - rtype = rpcrdma_areadch; - else + } else if (rqst->rq_snd_buf.flags & XDRBUF_WRITE) { rtype = rpcrdma_readch; + } else { + r_xprt->rx_stats.nomsg_call_count++; + headerp->rm_type = htonl(RDMA_NOMSG); + rtype = rpcrdma_areadch; + rpclen = 0; + } /* The following simplification is not true forever */ if (rtype != rpcrdma_noch && wtype == rpcrdma_replych) @@ -461,7 +510,6 @@ rpcrdma_marshal_req(struct rpc_rqst *rqst) } hdrlen = RPCRDMA_HDRLEN_MIN; - padlen = 0; /* * Pull up any extra send data into the preregistered buffer. @@ -470,45 +518,15 @@ rpcrdma_marshal_req(struct rpc_rqst *rqst) */ if (rtype == rpcrdma_noch) { - padlen = rpcrdma_inline_pullup(rqst, - RPCRDMA_INLINE_PAD_VALUE(rqst)); - - if (padlen) { - headerp->rm_type = rdma_msgp; - headerp->rm_body.rm_padded.rm_align = - cpu_to_be32(RPCRDMA_INLINE_PAD_VALUE(rqst)); - headerp->rm_body.rm_padded.rm_thresh = - cpu_to_be32(RPCRDMA_INLINE_PAD_THRESH); - headerp->rm_body.rm_padded.rm_pempty[0] = xdr_zero; - headerp->rm_body.rm_padded.rm_pempty[1] = xdr_zero; - headerp->rm_body.rm_padded.rm_pempty[2] = xdr_zero; - hdrlen += 2 * sizeof(u32); /* extra words in padhdr */ - if (wtype != rpcrdma_noch) { - dprintk("RPC: %s: invalid chunk list\n", - __func__); - return -EIO; - } - } else { - headerp->rm_body.rm_nochunks.rm_empty[0] = xdr_zero; - headerp->rm_body.rm_nochunks.rm_empty[1] = xdr_zero; - headerp->rm_body.rm_nochunks.rm_empty[2] = xdr_zero; - /* new length after pullup */ - rpclen = rqst->rq_svec[0].iov_len; - /* - * Currently we try to not actually use read inline. - * Reply chunks have the desirable property that - * they land, packed, directly in the target buffers - * without headers, so they require no fixup. The - * additional RDMA Write op sends the same amount - * of data, streams on-the-wire and adds no overhead - * on receive. Therefore, we request a reply chunk - * for non-writes wherever feasible and efficient. - */ - if (wtype == rpcrdma_noch) - wtype = rpcrdma_replych; - } - } + rpcrdma_inline_pullup(rqst); + headerp->rm_body.rm_nochunks.rm_empty[0] = xdr_zero; + headerp->rm_body.rm_nochunks.rm_empty[1] = xdr_zero; + headerp->rm_body.rm_nochunks.rm_empty[2] = xdr_zero; + /* new length after pullup */ + rpclen = rqst->rq_svec[0].iov_len; + } else if (rtype == rpcrdma_readch) + rpclen += rpcrdma_tail_pullup(&rqst->rq_snd_buf); if (rtype != rpcrdma_noch) { hdrlen = rpcrdma_create_chunks(rqst, &rqst->rq_snd_buf, headerp, rtype); @@ -521,9 +539,9 @@ rpcrdma_marshal_req(struct rpc_rqst *rqst) if (hdrlen < 0) return hdrlen; - dprintk("RPC: %s: %s: hdrlen %zd rpclen %zd padlen %zd" + dprintk("RPC: %s: %s: hdrlen %zd rpclen %zd" " headerp 0x%p base 0x%p lkey 0x%x\n", - __func__, transfertypes[wtype], hdrlen, rpclen, padlen, + __func__, transfertypes[wtype], hdrlen, rpclen, headerp, base, rdmab_lkey(req->rl_rdmabuf)); /* @@ -537,26 +555,15 @@ rpcrdma_marshal_req(struct rpc_rqst *rqst) req->rl_send_iov[0].length = hdrlen; req->rl_send_iov[0].lkey = rdmab_lkey(req->rl_rdmabuf); + req->rl_niovs = 1; + if (rtype == rpcrdma_areadch) + return 0; + req->rl_send_iov[1].addr = rdmab_addr(req->rl_sendbuf); req->rl_send_iov[1].length = rpclen; req->rl_send_iov[1].lkey = rdmab_lkey(req->rl_sendbuf); req->rl_niovs = 2; - - if (padlen) { - struct rpcrdma_ep *ep = &r_xprt->rx_ep; - - req->rl_send_iov[2].addr = rdmab_addr(ep->rep_padbuf); - req->rl_send_iov[2].length = padlen; - req->rl_send_iov[2].lkey = rdmab_lkey(ep->rep_padbuf); - - req->rl_send_iov[3].addr = req->rl_send_iov[1].addr + rpclen; - req->rl_send_iov[3].length = rqst->rq_slen - rpclen; - req->rl_send_iov[3].lkey = rdmab_lkey(req->rl_sendbuf); - - req->rl_niovs = 4; - } - return 0; } @@ -709,6 +716,37 @@ rpcrdma_connect_worker(struct work_struct *work) spin_unlock_bh(&xprt->transport_lock); } +#if defined(CONFIG_SUNRPC_BACKCHANNEL) +/* By convention, backchannel calls arrive via rdma_msg type + * messages, and never populate the chunk lists. This makes + * the RPC/RDMA header small and fixed in size, so it is + * straightforward to check the RPC header's direction field. + */ +static bool +rpcrdma_is_bcall(struct rpcrdma_msg *headerp) +{ + __be32 *p = (__be32 *)headerp; + + if (headerp->rm_type != rdma_msg) + return false; + if (headerp->rm_body.rm_chunks[0] != xdr_zero) + return false; + if (headerp->rm_body.rm_chunks[1] != xdr_zero) + return false; + if (headerp->rm_body.rm_chunks[2] != xdr_zero) + return false; + + /* sanity */ + if (p[7] != headerp->rm_xid) + return false; + /* call direction */ + if (p[8] != cpu_to_be32(RPC_CALL)) + return false; + + return true; +} +#endif /* CONFIG_SUNRPC_BACKCHANNEL */ + /* * This function is called when an async event is posted to * the connection which changes the connection state. All it @@ -721,8 +759,8 @@ rpcrdma_conn_func(struct rpcrdma_ep *ep) schedule_delayed_work(&ep->rep_connect_worker, 0); } -/* - * Called as a tasklet to do req/reply match and complete a request +/* Process received RPC/RDMA messages. + * * Errors must result in the RPC task either being awakened, or * allowed to timeout, to discover the errors at that time. */ @@ -732,60 +770,39 @@ rpcrdma_reply_handler(struct rpcrdma_rep *rep) struct rpcrdma_msg *headerp; struct rpcrdma_req *req; struct rpc_rqst *rqst; - struct rpc_xprt *xprt = rep->rr_xprt; - struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt); + struct rpcrdma_xprt *r_xprt = rep->rr_rxprt; + struct rpc_xprt *xprt = &r_xprt->rx_xprt; __be32 *iptr; int rdmalen, status; unsigned long cwnd; u32 credits; - /* Check status. If bad, signal disconnect and return rep to pool */ - if (rep->rr_len == ~0U) { - rpcrdma_recv_buffer_put(rep); - if (r_xprt->rx_ep.rep_connected == 1) { - r_xprt->rx_ep.rep_connected = -EIO; - rpcrdma_conn_func(&r_xprt->rx_ep); - } - return; - } - if (rep->rr_len < RPCRDMA_HDRLEN_MIN) { - dprintk("RPC: %s: short/invalid reply\n", __func__); - goto repost; - } + dprintk("RPC: %s: incoming rep %p\n", __func__, rep); + + if (rep->rr_len == RPCRDMA_BAD_LEN) + goto out_badstatus; + if (rep->rr_len < RPCRDMA_HDRLEN_MIN) + goto out_shortreply; + headerp = rdmab_to_msg(rep->rr_rdmabuf); - if (headerp->rm_vers != rpcrdma_version) { - dprintk("RPC: %s: invalid version %d\n", - __func__, be32_to_cpu(headerp->rm_vers)); - goto repost; - } + if (headerp->rm_vers != rpcrdma_version) + goto out_badversion; +#if defined(CONFIG_SUNRPC_BACKCHANNEL) + if (rpcrdma_is_bcall(headerp)) + goto out_bcall; +#endif - /* Get XID and try for a match. */ - spin_lock(&xprt->transport_lock); + /* Match incoming rpcrdma_rep to an rpcrdma_req to + * get context for handling any incoming chunks. + */ + spin_lock_bh(&xprt->transport_lock); rqst = xprt_lookup_rqst(xprt, headerp->rm_xid); - if (rqst == NULL) { - spin_unlock(&xprt->transport_lock); - dprintk("RPC: %s: reply 0x%p failed " - "to match any request xid 0x%08x len %d\n", - __func__, rep, be32_to_cpu(headerp->rm_xid), - rep->rr_len); -repost: - r_xprt->rx_stats.bad_reply_count++; - rep->rr_func = rpcrdma_reply_handler; - if (rpcrdma_ep_post_recv(&r_xprt->rx_ia, &r_xprt->rx_ep, rep)) - rpcrdma_recv_buffer_put(rep); - - return; - } + if (!rqst) + goto out_nomatch; - /* get request object */ req = rpcr_to_rdmar(rqst); - if (req->rl_reply) { - spin_unlock(&xprt->transport_lock); - dprintk("RPC: %s: duplicate reply 0x%p to RPC " - "request 0x%p: xid 0x%08x\n", __func__, rep, req, - be32_to_cpu(headerp->rm_xid)); - goto repost; - } + if (req->rl_reply) + goto out_duplicate; dprintk("RPC: %s: reply 0x%p completes request 0x%p\n" " RPC request 0x%p xid 0x%08x\n", @@ -882,8 +899,50 @@ badheader: if (xprt->cwnd > cwnd) xprt_release_rqst_cong(rqst->rq_task); + xprt_complete_rqst(rqst->rq_task, status); + spin_unlock_bh(&xprt->transport_lock); dprintk("RPC: %s: xprt_complete_rqst(0x%p, 0x%p, %d)\n", __func__, xprt, rqst, status); - xprt_complete_rqst(rqst->rq_task, status); - spin_unlock(&xprt->transport_lock); + return; + +out_badstatus: + rpcrdma_recv_buffer_put(rep); + if (r_xprt->rx_ep.rep_connected == 1) { + r_xprt->rx_ep.rep_connected = -EIO; + rpcrdma_conn_func(&r_xprt->rx_ep); + } + return; + +#if defined(CONFIG_SUNRPC_BACKCHANNEL) +out_bcall: + rpcrdma_bc_receive_call(r_xprt, rep); + return; +#endif + +out_shortreply: + dprintk("RPC: %s: short/invalid reply\n", __func__); + goto repost; + +out_badversion: + dprintk("RPC: %s: invalid version %d\n", + __func__, be32_to_cpu(headerp->rm_vers)); + goto repost; + +out_nomatch: + spin_unlock_bh(&xprt->transport_lock); + dprintk("RPC: %s: no match for incoming xid 0x%08x len %d\n", + __func__, be32_to_cpu(headerp->rm_xid), + rep->rr_len); + goto repost; + +out_duplicate: + spin_unlock_bh(&xprt->transport_lock); + dprintk("RPC: %s: " + "duplicate reply %p to RPC request %p: xid 0x%08x\n", + __func__, rep, req, be32_to_cpu(headerp->rm_xid)); + +repost: + r_xprt->rx_stats.bad_reply_count++; + if (rpcrdma_ep_post_recv(&r_xprt->rx_ia, &r_xprt->rx_ep, rep)) + rpcrdma_recv_buffer_put(rep); } diff --git a/kernel/net/sunrpc/xprtrdma/svc_rdma.c b/kernel/net/sunrpc/xprtrdma/svc_rdma.c index c1b627026..1b7051bdb 100644 --- a/kernel/net/sunrpc/xprtrdma/svc_rdma.c +++ b/kernel/net/sunrpc/xprtrdma/svc_rdma.c @@ -38,8 +38,7 @@ * * Author: Tom Tucker <tom@opengridcomputing.com> */ -#include <linux/module.h> -#include <linux/init.h> + #include <linux/slab.h> #include <linux/fs.h> #include <linux/sysctl.h> @@ -240,6 +239,9 @@ void svc_rdma_cleanup(void) unregister_sysctl_table(svcrdma_table_header); svcrdma_table_header = NULL; } +#if defined(CONFIG_SUNRPC_BACKCHANNEL) + svc_unreg_xprt_class(&svc_rdma_bc_class); +#endif svc_unreg_xprt_class(&svc_rdma_class); kmem_cache_destroy(svc_rdma_map_cachep); kmem_cache_destroy(svc_rdma_ctxt_cachep); @@ -287,6 +289,9 @@ int svc_rdma_init(void) /* Register RDMA with the SVC transport switch */ svc_reg_xprt_class(&svc_rdma_class); +#if defined(CONFIG_SUNRPC_BACKCHANNEL) + svc_reg_xprt_class(&svc_rdma_bc_class); +#endif return 0; err1: kmem_cache_destroy(svc_rdma_map_cachep); @@ -295,8 +300,3 @@ int svc_rdma_init(void) destroy_workqueue(svc_rdma_wq); return -ENOMEM; } -MODULE_AUTHOR("Tom Tucker <tom@opengridcomputing.com>"); -MODULE_DESCRIPTION("SVC RDMA Transport"); -MODULE_LICENSE("Dual BSD/GPL"); -module_init(svc_rdma_init); -module_exit(svc_rdma_cleanup); diff --git a/kernel/net/sunrpc/xprtrdma/svc_rdma_marshal.c b/kernel/net/sunrpc/xprtrdma/svc_rdma_marshal.c index b681855cf..e2fca7617 100644 --- a/kernel/net/sunrpc/xprtrdma/svc_rdma_marshal.c +++ b/kernel/net/sunrpc/xprtrdma/svc_rdma_marshal.c @@ -50,12 +50,12 @@ /* * Decodes a read chunk list. The expected format is as follows: * descrim : xdr_one - * position : u32 offset into XDR stream - * handle : u32 RKEY + * position : __be32 offset into XDR stream + * handle : __be32 RKEY * . . . * end-of-list: xdr_zero */ -static u32 *decode_read_list(u32 *va, u32 *vaend) +static __be32 *decode_read_list(__be32 *va, __be32 *vaend) { struct rpcrdma_read_chunk *ch = (struct rpcrdma_read_chunk *)va; @@ -67,20 +67,20 @@ static u32 *decode_read_list(u32 *va, u32 *vaend) } ch++; } - return (u32 *)&ch->rc_position; + return &ch->rc_position; } /* * Decodes a write chunk list. The expected format is as follows: * descrim : xdr_one * nchunks : <count> - * handle : u32 RKEY ---+ - * length : u32 <len of segment> | + * handle : __be32 RKEY ---+ + * length : __be32 <len of segment> | * offset : remove va + <count> * . . . | * ---+ */ -static u32 *decode_write_list(u32 *va, u32 *vaend) +static __be32 *decode_write_list(__be32 *va, __be32 *vaend) { unsigned long start, end; int nchunks; @@ -90,14 +90,14 @@ static u32 *decode_write_list(u32 *va, u32 *vaend) /* Check for not write-array */ if (ary->wc_discrim == xdr_zero) - return (u32 *)&ary->wc_nchunks; + return &ary->wc_nchunks; if ((unsigned long)ary + sizeof(struct rpcrdma_write_array) > (unsigned long)vaend) { dprintk("svcrdma: ary=%p, vaend=%p\n", ary, vaend); return NULL; } - nchunks = ntohl(ary->wc_nchunks); + nchunks = be32_to_cpu(ary->wc_nchunks); start = (unsigned long)&ary->wc_array[0]; end = (unsigned long)vaend; @@ -112,10 +112,10 @@ static u32 *decode_write_list(u32 *va, u32 *vaend) * rs_length is the 2nd 4B field in wc_target and taking its * address skips the list terminator */ - return (u32 *)&ary->wc_array[nchunks].wc_target.rs_length; + return &ary->wc_array[nchunks].wc_target.rs_length; } -static u32 *decode_reply_array(u32 *va, u32 *vaend) +static __be32 *decode_reply_array(__be32 *va, __be32 *vaend) { unsigned long start, end; int nchunks; @@ -124,14 +124,14 @@ static u32 *decode_reply_array(u32 *va, u32 *vaend) /* Check for no reply-array */ if (ary->wc_discrim == xdr_zero) - return (u32 *)&ary->wc_nchunks; + return &ary->wc_nchunks; if ((unsigned long)ary + sizeof(struct rpcrdma_write_array) > (unsigned long)vaend) { dprintk("svcrdma: ary=%p, vaend=%p\n", ary, vaend); return NULL; } - nchunks = ntohl(ary->wc_nchunks); + nchunks = be32_to_cpu(ary->wc_nchunks); start = (unsigned long)&ary->wc_array[0]; end = (unsigned long)vaend; @@ -142,15 +142,14 @@ static u32 *decode_reply_array(u32 *va, u32 *vaend) ary, nchunks, vaend); return NULL; } - return (u32 *)&ary->wc_array[nchunks]; + return (__be32 *)&ary->wc_array[nchunks]; } int svc_rdma_xdr_decode_req(struct rpcrdma_msg **rdma_req, struct svc_rqst *rqstp) { struct rpcrdma_msg *rmsgp = NULL; - u32 *va; - u32 *vaend; + __be32 *va, *vaend; u32 hdr_len; rmsgp = (struct rpcrdma_msg *)rqstp->rq_arg.head[0].iov_base; @@ -162,22 +161,17 @@ int svc_rdma_xdr_decode_req(struct rpcrdma_msg **rdma_req, return -EINVAL; } - /* Decode the header */ - rmsgp->rm_xid = ntohl(rmsgp->rm_xid); - rmsgp->rm_vers = ntohl(rmsgp->rm_vers); - rmsgp->rm_credit = ntohl(rmsgp->rm_credit); - rmsgp->rm_type = ntohl(rmsgp->rm_type); - - if (rmsgp->rm_vers != RPCRDMA_VERSION) + if (rmsgp->rm_vers != rpcrdma_version) return -ENOSYS; /* Pull in the extra for the padded case and bump our pointer */ - if (rmsgp->rm_type == RDMA_MSGP) { + if (rmsgp->rm_type == rdma_msgp) { int hdrlen; + rmsgp->rm_body.rm_padded.rm_align = - ntohl(rmsgp->rm_body.rm_padded.rm_align); + be32_to_cpu(rmsgp->rm_body.rm_padded.rm_align); rmsgp->rm_body.rm_padded.rm_thresh = - ntohl(rmsgp->rm_body.rm_padded.rm_thresh); + be32_to_cpu(rmsgp->rm_body.rm_padded.rm_thresh); va = &rmsgp->rm_body.rm_padded.rm_pempty[4]; rqstp->rq_arg.head[0].iov_base = va; @@ -192,7 +186,7 @@ int svc_rdma_xdr_decode_req(struct rpcrdma_msg **rdma_req, * chunk list and a reply chunk list. */ va = &rmsgp->rm_body.rm_chunks[0]; - vaend = (u32 *)((unsigned long)rmsgp + rqstp->rq_arg.len); + vaend = (__be32 *)((unsigned long)rmsgp + rqstp->rq_arg.len); va = decode_read_list(va, vaend); if (!va) return -EINVAL; @@ -211,76 +205,20 @@ int svc_rdma_xdr_decode_req(struct rpcrdma_msg **rdma_req, return hdr_len; } -int svc_rdma_xdr_decode_deferred_req(struct svc_rqst *rqstp) -{ - struct rpcrdma_msg *rmsgp = NULL; - struct rpcrdma_read_chunk *ch; - struct rpcrdma_write_array *ary; - u32 *va; - u32 hdrlen; - - dprintk("svcrdma: processing deferred RDMA header on rqstp=%p\n", - rqstp); - rmsgp = (struct rpcrdma_msg *)rqstp->rq_arg.head[0].iov_base; - - /* Pull in the extra for the padded case and bump our pointer */ - if (rmsgp->rm_type == RDMA_MSGP) { - va = &rmsgp->rm_body.rm_padded.rm_pempty[4]; - rqstp->rq_arg.head[0].iov_base = va; - hdrlen = (u32)((unsigned long)va - (unsigned long)rmsgp); - rqstp->rq_arg.head[0].iov_len -= hdrlen; - return hdrlen; - } - - /* - * Skip all chunks to find RPC msg. These were previously processed - */ - va = &rmsgp->rm_body.rm_chunks[0]; - - /* Skip read-list */ - for (ch = (struct rpcrdma_read_chunk *)va; - ch->rc_discrim != xdr_zero; ch++); - va = (u32 *)&ch->rc_position; - - /* Skip write-list */ - ary = (struct rpcrdma_write_array *)va; - if (ary->wc_discrim == xdr_zero) - va = (u32 *)&ary->wc_nchunks; - else - /* - * rs_length is the 2nd 4B field in wc_target and taking its - * address skips the list terminator - */ - va = (u32 *)&ary->wc_array[ary->wc_nchunks].wc_target.rs_length; - - /* Skip reply-array */ - ary = (struct rpcrdma_write_array *)va; - if (ary->wc_discrim == xdr_zero) - va = (u32 *)&ary->wc_nchunks; - else - va = (u32 *)&ary->wc_array[ary->wc_nchunks]; - - rqstp->rq_arg.head[0].iov_base = va; - hdrlen = (unsigned long)va - (unsigned long)rmsgp; - rqstp->rq_arg.head[0].iov_len -= hdrlen; - - return hdrlen; -} - int svc_rdma_xdr_encode_error(struct svcxprt_rdma *xprt, struct rpcrdma_msg *rmsgp, - enum rpcrdma_errcode err, u32 *va) + enum rpcrdma_errcode err, __be32 *va) { - u32 *startp = va; + __be32 *startp = va; - *va++ = htonl(rmsgp->rm_xid); - *va++ = htonl(rmsgp->rm_vers); - *va++ = htonl(xprt->sc_max_requests); - *va++ = htonl(RDMA_ERROR); - *va++ = htonl(err); + *va++ = rmsgp->rm_xid; + *va++ = rmsgp->rm_vers; + *va++ = cpu_to_be32(xprt->sc_max_requests); + *va++ = rdma_error; + *va++ = cpu_to_be32(err); if (err == ERR_VERS) { - *va++ = htonl(RPCRDMA_VERSION); - *va++ = htonl(RPCRDMA_VERSION); + *va++ = rpcrdma_version; + *va++ = rpcrdma_version; } return (int)((unsigned long)va - (unsigned long)startp); @@ -297,7 +235,7 @@ int svc_rdma_xdr_get_reply_hdr_len(struct rpcrdma_msg *rmsgp) &rmsgp->rm_body.rm_chunks[1]; if (wr_ary->wc_discrim) wr_ary = (struct rpcrdma_write_array *) - &wr_ary->wc_array[ntohl(wr_ary->wc_nchunks)]. + &wr_ary->wc_array[be32_to_cpu(wr_ary->wc_nchunks)]. wc_target.rs_length; else wr_ary = (struct rpcrdma_write_array *) @@ -306,7 +244,7 @@ int svc_rdma_xdr_get_reply_hdr_len(struct rpcrdma_msg *rmsgp) /* skip reply array */ if (wr_ary->wc_discrim) wr_ary = (struct rpcrdma_write_array *) - &wr_ary->wc_array[ntohl(wr_ary->wc_nchunks)]; + &wr_ary->wc_array[be32_to_cpu(wr_ary->wc_nchunks)]; else wr_ary = (struct rpcrdma_write_array *) &wr_ary->wc_nchunks; @@ -325,7 +263,7 @@ void svc_rdma_xdr_encode_write_list(struct rpcrdma_msg *rmsgp, int chunks) ary = (struct rpcrdma_write_array *) &rmsgp->rm_body.rm_chunks[1]; ary->wc_discrim = xdr_one; - ary->wc_nchunks = htonl(chunks); + ary->wc_nchunks = cpu_to_be32(chunks); /* write-list terminator */ ary->wc_array[chunks].wc_target.rs_handle = xdr_zero; @@ -338,7 +276,7 @@ void svc_rdma_xdr_encode_reply_array(struct rpcrdma_write_array *ary, int chunks) { ary->wc_discrim = xdr_one; - ary->wc_nchunks = htonl(chunks); + ary->wc_nchunks = cpu_to_be32(chunks); } void svc_rdma_xdr_encode_array_chunk(struct rpcrdma_write_array *ary, @@ -350,7 +288,7 @@ void svc_rdma_xdr_encode_array_chunk(struct rpcrdma_write_array *ary, struct rpcrdma_segment *seg = &ary->wc_array[chunk_no].wc_target; seg->rs_handle = rs_handle; seg->rs_offset = rs_offset; - seg->rs_length = htonl(write_len); + seg->rs_length = cpu_to_be32(write_len); } void svc_rdma_xdr_encode_reply_header(struct svcxprt_rdma *xprt, @@ -358,10 +296,10 @@ void svc_rdma_xdr_encode_reply_header(struct svcxprt_rdma *xprt, struct rpcrdma_msg *rdma_resp, enum rpcrdma_proc rdma_type) { - rdma_resp->rm_xid = htonl(rdma_argp->rm_xid); - rdma_resp->rm_vers = htonl(rdma_argp->rm_vers); - rdma_resp->rm_credit = htonl(xprt->sc_max_requests); - rdma_resp->rm_type = htonl(rdma_type); + rdma_resp->rm_xid = rdma_argp->rm_xid; + rdma_resp->rm_vers = rdma_argp->rm_vers; + rdma_resp->rm_credit = cpu_to_be32(xprt->sc_max_requests); + rdma_resp->rm_type = cpu_to_be32(rdma_type); /* Encode <nul> chunks lists */ rdma_resp->rm_body.rm_chunks[0] = xdr_zero; diff --git a/kernel/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c b/kernel/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c index f9f13a32d..ff4f01e52 100644 --- a/kernel/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c +++ b/kernel/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c @@ -85,7 +85,7 @@ static void rdma_build_arg_xdr(struct svc_rqst *rqstp, /* RDMA_NOMSG: RDMA READ data should land just after RDMA RECV data */ rmsgp = (struct rpcrdma_msg *)rqstp->rq_arg.head[0].iov_base; - if (be32_to_cpu(rmsgp->rm_type) == RDMA_NOMSG) + if (rmsgp->rm_type == rdma_nomsg) rqstp->rq_arg.pages = &rqstp->rq_pages[0]; else rqstp->rq_arg.pages = &rqstp->rq_pages[1]; @@ -115,15 +115,6 @@ static void rdma_build_arg_xdr(struct svc_rqst *rqstp, rqstp->rq_arg.tail[0].iov_len = 0; } -static int rdma_read_max_sge(struct svcxprt_rdma *xprt, int sge_count) -{ - if (rdma_node_get_transport(xprt->sc_cm_id->device->node_type) == - RDMA_TRANSPORT_IWARP) - return 1; - else - return min_t(int, sge_count, xprt->sc_max_sge); -} - /* Issue an RDMA_READ using the local lkey to map the data sink */ int rdma_read_chunk_lcl(struct svcxprt_rdma *xprt, struct svc_rqst *rqstp, @@ -135,7 +126,7 @@ int rdma_read_chunk_lcl(struct svcxprt_rdma *xprt, u64 rs_offset, bool last) { - struct ib_send_wr read_wr; + struct ib_rdma_wr read_wr; int pages_needed = PAGE_ALIGN(*page_offset + rs_length) >> PAGE_SHIFT; struct svc_rdma_op_ctxt *ctxt = svc_rdma_get_context(xprt); int ret, read, pno; @@ -144,9 +135,9 @@ int rdma_read_chunk_lcl(struct svcxprt_rdma *xprt, ctxt->direction = DMA_FROM_DEVICE; ctxt->read_hdr = head; - pages_needed = - min_t(int, pages_needed, rdma_read_max_sge(xprt, pages_needed)); - read = min_t(int, pages_needed << PAGE_SHIFT, rs_length); + pages_needed = min_t(int, pages_needed, xprt->sc_max_sge_rd); + read = min_t(int, (pages_needed << PAGE_SHIFT) - *page_offset, + rs_length); for (pno = 0; pno < pages_needed; pno++) { int len = min_t(int, rs_length, PAGE_SIZE - pg_off); @@ -189,16 +180,16 @@ int rdma_read_chunk_lcl(struct svcxprt_rdma *xprt, clear_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags); memset(&read_wr, 0, sizeof(read_wr)); - read_wr.wr_id = (unsigned long)ctxt; - read_wr.opcode = IB_WR_RDMA_READ; - ctxt->wr_op = read_wr.opcode; - read_wr.send_flags = IB_SEND_SIGNALED; - read_wr.wr.rdma.rkey = rs_handle; - read_wr.wr.rdma.remote_addr = rs_offset; - read_wr.sg_list = ctxt->sge; - read_wr.num_sge = pages_needed; - - ret = svc_rdma_send(xprt, &read_wr); + read_wr.wr.wr_id = (unsigned long)ctxt; + read_wr.wr.opcode = IB_WR_RDMA_READ; + ctxt->wr_op = read_wr.wr.opcode; + read_wr.wr.send_flags = IB_SEND_SIGNALED; + read_wr.rkey = rs_handle; + read_wr.remote_addr = rs_offset; + read_wr.wr.sg_list = ctxt->sge; + read_wr.wr.num_sge = pages_needed; + + ret = svc_rdma_send(xprt, &read_wr.wr); if (ret) { pr_err("svcrdma: Error %d posting RDMA_READ\n", ret); set_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags); @@ -228,14 +219,14 @@ int rdma_read_chunk_frmr(struct svcxprt_rdma *xprt, u64 rs_offset, bool last) { - struct ib_send_wr read_wr; + struct ib_rdma_wr read_wr; struct ib_send_wr inv_wr; - struct ib_send_wr fastreg_wr; + struct ib_reg_wr reg_wr; u8 key; - int pages_needed = PAGE_ALIGN(*page_offset + rs_length) >> PAGE_SHIFT; + int nents = PAGE_ALIGN(*page_offset + rs_length) >> PAGE_SHIFT; struct svc_rdma_op_ctxt *ctxt = svc_rdma_get_context(xprt); struct svc_rdma_fastreg_mr *frmr = svc_rdma_get_frmr(xprt); - int ret, read, pno; + int ret, read, pno, dma_nents, n; u32 pg_off = *page_offset; u32 pg_no = *page_no; @@ -244,16 +235,14 @@ int rdma_read_chunk_frmr(struct svcxprt_rdma *xprt, ctxt->direction = DMA_FROM_DEVICE; ctxt->frmr = frmr; - pages_needed = min_t(int, pages_needed, xprt->sc_frmr_pg_list_len); - read = min_t(int, pages_needed << PAGE_SHIFT, rs_length); + nents = min_t(unsigned int, nents, xprt->sc_frmr_pg_list_len); + read = min_t(int, (nents << PAGE_SHIFT) - *page_offset, rs_length); - frmr->kva = page_address(rqstp->rq_arg.pages[pg_no]); frmr->direction = DMA_FROM_DEVICE; frmr->access_flags = (IB_ACCESS_LOCAL_WRITE|IB_ACCESS_REMOTE_WRITE); - frmr->map_len = pages_needed << PAGE_SHIFT; - frmr->page_list_len = pages_needed; + frmr->sg_nents = nents; - for (pno = 0; pno < pages_needed; pno++) { + for (pno = 0; pno < nents; pno++) { int len = min_t(int, rs_length, PAGE_SIZE - pg_off); head->arg.pages[pg_no] = rqstp->rq_arg.pages[pg_no]; @@ -261,17 +250,12 @@ int rdma_read_chunk_frmr(struct svcxprt_rdma *xprt, head->arg.len += len; if (!pg_off) head->count++; + + sg_set_page(&frmr->sg[pno], rqstp->rq_arg.pages[pg_no], + len, pg_off); + rqstp->rq_respages = &rqstp->rq_arg.pages[pg_no+1]; rqstp->rq_next_page = rqstp->rq_respages + 1; - frmr->page_list->page_list[pno] = - ib_dma_map_page(xprt->sc_cm_id->device, - head->arg.pages[pg_no], 0, - PAGE_SIZE, DMA_FROM_DEVICE); - ret = ib_dma_mapping_error(xprt->sc_cm_id->device, - frmr->page_list->page_list[pno]); - if (ret) - goto err; - atomic_inc(&xprt->sc_dma_used); /* adjust offset and wrap to next page if needed */ pg_off += len; @@ -287,43 +271,57 @@ int rdma_read_chunk_frmr(struct svcxprt_rdma *xprt, else clear_bit(RDMACTXT_F_LAST_CTXT, &ctxt->flags); + dma_nents = ib_dma_map_sg(xprt->sc_cm_id->device, + frmr->sg, frmr->sg_nents, + frmr->direction); + if (!dma_nents) { + pr_err("svcrdma: failed to dma map sg %p\n", + frmr->sg); + return -ENOMEM; + } + atomic_inc(&xprt->sc_dma_used); + + n = ib_map_mr_sg(frmr->mr, frmr->sg, frmr->sg_nents, PAGE_SIZE); + if (unlikely(n != frmr->sg_nents)) { + pr_err("svcrdma: failed to map mr %p (%d/%d elements)\n", + frmr->mr, n, frmr->sg_nents); + return n < 0 ? n : -EINVAL; + } + /* Bump the key */ key = (u8)(frmr->mr->lkey & 0x000000FF); ib_update_fast_reg_key(frmr->mr, ++key); - ctxt->sge[0].addr = (unsigned long)frmr->kva + *page_offset; + ctxt->sge[0].addr = frmr->mr->iova; ctxt->sge[0].lkey = frmr->mr->lkey; - ctxt->sge[0].length = read; + ctxt->sge[0].length = frmr->mr->length; ctxt->count = 1; ctxt->read_hdr = head; - /* Prepare FASTREG WR */ - memset(&fastreg_wr, 0, sizeof(fastreg_wr)); - fastreg_wr.opcode = IB_WR_FAST_REG_MR; - fastreg_wr.send_flags = IB_SEND_SIGNALED; - fastreg_wr.wr.fast_reg.iova_start = (unsigned long)frmr->kva; - fastreg_wr.wr.fast_reg.page_list = frmr->page_list; - fastreg_wr.wr.fast_reg.page_list_len = frmr->page_list_len; - fastreg_wr.wr.fast_reg.page_shift = PAGE_SHIFT; - fastreg_wr.wr.fast_reg.length = frmr->map_len; - fastreg_wr.wr.fast_reg.access_flags = frmr->access_flags; - fastreg_wr.wr.fast_reg.rkey = frmr->mr->lkey; - fastreg_wr.next = &read_wr; + /* Prepare REG WR */ + reg_wr.wr.opcode = IB_WR_REG_MR; + reg_wr.wr.wr_id = 0; + reg_wr.wr.send_flags = IB_SEND_SIGNALED; + reg_wr.wr.num_sge = 0; + reg_wr.mr = frmr->mr; + reg_wr.key = frmr->mr->lkey; + reg_wr.access = frmr->access_flags; + reg_wr.wr.next = &read_wr.wr; /* Prepare RDMA_READ */ memset(&read_wr, 0, sizeof(read_wr)); - read_wr.send_flags = IB_SEND_SIGNALED; - read_wr.wr.rdma.rkey = rs_handle; - read_wr.wr.rdma.remote_addr = rs_offset; - read_wr.sg_list = ctxt->sge; - read_wr.num_sge = 1; + read_wr.wr.send_flags = IB_SEND_SIGNALED; + read_wr.rkey = rs_handle; + read_wr.remote_addr = rs_offset; + read_wr.wr.sg_list = ctxt->sge; + read_wr.wr.num_sge = 1; if (xprt->sc_dev_caps & SVCRDMA_DEVCAP_READ_W_INV) { - read_wr.opcode = IB_WR_RDMA_READ_WITH_INV; - read_wr.wr_id = (unsigned long)ctxt; - read_wr.ex.invalidate_rkey = ctxt->frmr->mr->lkey; + read_wr.wr.opcode = IB_WR_RDMA_READ_WITH_INV; + read_wr.wr.wr_id = (unsigned long)ctxt; + read_wr.wr.ex.invalidate_rkey = ctxt->frmr->mr->lkey; } else { - read_wr.opcode = IB_WR_RDMA_READ; - read_wr.next = &inv_wr; + read_wr.wr.opcode = IB_WR_RDMA_READ; + read_wr.wr.next = &inv_wr; /* Prepare invalidate */ memset(&inv_wr, 0, sizeof(inv_wr)); inv_wr.wr_id = (unsigned long)ctxt; @@ -331,10 +329,10 @@ int rdma_read_chunk_frmr(struct svcxprt_rdma *xprt, inv_wr.send_flags = IB_SEND_SIGNALED | IB_SEND_FENCE; inv_wr.ex.invalidate_rkey = frmr->mr->lkey; } - ctxt->wr_op = read_wr.opcode; + ctxt->wr_op = read_wr.wr.opcode; /* Post the chain */ - ret = svc_rdma_send(xprt, &fastreg_wr); + ret = svc_rdma_send(xprt, ®_wr.wr); if (ret) { pr_err("svcrdma: Error %d posting RDMA_READ\n", ret); set_bit(XPT_CLOSE, &xprt->sc_xprt.xpt_flags); @@ -348,7 +346,8 @@ int rdma_read_chunk_frmr(struct svcxprt_rdma *xprt, atomic_inc(&rdma_stat_read); return ret; err: - svc_rdma_unmap_dma(ctxt); + ib_dma_unmap_sg(xprt->sc_cm_id->device, + frmr->sg, frmr->sg_nents, frmr->direction); svc_rdma_put_context(ctxt, 0); svc_rdma_put_frmr(xprt, frmr); return ret; @@ -541,7 +540,7 @@ static int rdma_read_complete(struct svc_rqst *rqstp, rqstp->rq_arg.page_base = head->arg.page_base; /* rq_respages starts after the last arg page */ - rqstp->rq_respages = &rqstp->rq_arg.pages[page_no]; + rqstp->rq_respages = &rqstp->rq_pages[page_no]; rqstp->rq_next_page = rqstp->rq_respages + 1; /* Rebuild rq_arg head and tail. */ diff --git a/kernel/net/sunrpc/xprtrdma/svc_rdma_sendto.c b/kernel/net/sunrpc/xprtrdma/svc_rdma_sendto.c index 7de33d1af..969a1ab75 100644 --- a/kernel/net/sunrpc/xprtrdma/svc_rdma_sendto.c +++ b/kernel/net/sunrpc/xprtrdma/svc_rdma_sendto.c @@ -136,6 +136,79 @@ static dma_addr_t dma_map_xdr(struct svcxprt_rdma *xprt, return dma_addr; } +/* Returns the address of the first read chunk or <nul> if no read chunk + * is present + */ +struct rpcrdma_read_chunk * +svc_rdma_get_read_chunk(struct rpcrdma_msg *rmsgp) +{ + struct rpcrdma_read_chunk *ch = + (struct rpcrdma_read_chunk *)&rmsgp->rm_body.rm_chunks[0]; + + if (ch->rc_discrim == xdr_zero) + return NULL; + return ch; +} + +/* Returns the address of the first read write array element or <nul> + * if no write array list is present + */ +static struct rpcrdma_write_array * +svc_rdma_get_write_array(struct rpcrdma_msg *rmsgp) +{ + if (rmsgp->rm_body.rm_chunks[0] != xdr_zero || + rmsgp->rm_body.rm_chunks[1] == xdr_zero) + return NULL; + return (struct rpcrdma_write_array *)&rmsgp->rm_body.rm_chunks[1]; +} + +/* Returns the address of the first reply array element or <nul> if no + * reply array is present + */ +static struct rpcrdma_write_array * +svc_rdma_get_reply_array(struct rpcrdma_msg *rmsgp) +{ + struct rpcrdma_read_chunk *rch; + struct rpcrdma_write_array *wr_ary; + struct rpcrdma_write_array *rp_ary; + + /* XXX: Need to fix when reply chunk may occur with read list + * and/or write list. + */ + if (rmsgp->rm_body.rm_chunks[0] != xdr_zero || + rmsgp->rm_body.rm_chunks[1] != xdr_zero) + return NULL; + + rch = svc_rdma_get_read_chunk(rmsgp); + if (rch) { + while (rch->rc_discrim != xdr_zero) + rch++; + + /* The reply chunk follows an empty write array located + * at 'rc_position' here. The reply array is at rc_target. + */ + rp_ary = (struct rpcrdma_write_array *)&rch->rc_target; + goto found_it; + } + + wr_ary = svc_rdma_get_write_array(rmsgp); + if (wr_ary) { + int chunk = be32_to_cpu(wr_ary->wc_nchunks); + + rp_ary = (struct rpcrdma_write_array *) + &wr_ary->wc_array[chunk].wc_target.rs_length; + goto found_it; + } + + /* No read list, no write list */ + rp_ary = (struct rpcrdma_write_array *)&rmsgp->rm_body.rm_chunks[2]; + + found_it: + if (rp_ary->wc_discrim == xdr_zero) + return NULL; + return rp_ary; +} + /* Assumptions: * - The specified write_len can be represented in sc_max_sge * PAGE_SIZE */ @@ -144,7 +217,7 @@ static int send_write(struct svcxprt_rdma *xprt, struct svc_rqst *rqstp, u32 xdr_off, int write_len, struct svc_rdma_req_map *vec) { - struct ib_send_wr write_wr; + struct ib_rdma_wr write_wr; struct ib_sge *sge; int xdr_sge_no; int sge_no; @@ -209,17 +282,17 @@ static int send_write(struct svcxprt_rdma *xprt, struct svc_rqst *rqstp, /* Prepare WRITE WR */ memset(&write_wr, 0, sizeof write_wr); ctxt->wr_op = IB_WR_RDMA_WRITE; - write_wr.wr_id = (unsigned long)ctxt; - write_wr.sg_list = &sge[0]; - write_wr.num_sge = sge_no; - write_wr.opcode = IB_WR_RDMA_WRITE; - write_wr.send_flags = IB_SEND_SIGNALED; - write_wr.wr.rdma.rkey = rmr; - write_wr.wr.rdma.remote_addr = to; + write_wr.wr.wr_id = (unsigned long)ctxt; + write_wr.wr.sg_list = &sge[0]; + write_wr.wr.num_sge = sge_no; + write_wr.wr.opcode = IB_WR_RDMA_WRITE; + write_wr.wr.send_flags = IB_SEND_SIGNALED; + write_wr.rkey = rmr; + write_wr.remote_addr = to; /* Post It */ atomic_inc(&rdma_stat_write); - if (svc_rdma_send(xprt, &write_wr)) + if (svc_rdma_send(xprt, &write_wr.wr)) goto err; return write_len - bc; err: @@ -240,6 +313,7 @@ static int send_write_chunks(struct svcxprt_rdma *xprt, u32 xdr_off; int chunk_off; int chunk_no; + int nchunks; struct rpcrdma_write_array *arg_ary; struct rpcrdma_write_array *res_ary; int ret; @@ -251,14 +325,15 @@ static int send_write_chunks(struct svcxprt_rdma *xprt, &rdma_resp->rm_body.rm_chunks[1]; /* Write chunks start at the pagelist */ + nchunks = be32_to_cpu(arg_ary->wc_nchunks); for (xdr_off = rqstp->rq_res.head[0].iov_len, chunk_no = 0; - xfer_len && chunk_no < arg_ary->wc_nchunks; + xfer_len && chunk_no < nchunks; chunk_no++) { struct rpcrdma_segment *arg_ch; u64 rs_offset; arg_ch = &arg_ary->wc_array[chunk_no].wc_target; - write_len = min(xfer_len, ntohl(arg_ch->rs_length)); + write_len = min(xfer_len, be32_to_cpu(arg_ch->rs_length)); /* Prepare the response chunk given the length actually * written */ @@ -270,7 +345,7 @@ static int send_write_chunks(struct svcxprt_rdma *xprt, chunk_off = 0; while (write_len) { ret = send_write(xprt, rqstp, - ntohl(arg_ch->rs_handle), + be32_to_cpu(arg_ch->rs_handle), rs_offset + chunk_off, xdr_off, write_len, @@ -318,13 +393,13 @@ static int send_reply_chunks(struct svcxprt_rdma *xprt, &rdma_resp->rm_body.rm_chunks[2]; /* xdr offset starts at RPC message */ - nchunks = ntohl(arg_ary->wc_nchunks); + nchunks = be32_to_cpu(arg_ary->wc_nchunks); for (xdr_off = 0, chunk_no = 0; xfer_len && chunk_no < nchunks; chunk_no++) { u64 rs_offset; ch = &arg_ary->wc_array[chunk_no].wc_target; - write_len = min(xfer_len, htonl(ch->rs_length)); + write_len = min(xfer_len, be32_to_cpu(ch->rs_length)); /* Prepare the reply chunk given the length actually * written */ @@ -335,7 +410,7 @@ static int send_reply_chunks(struct svcxprt_rdma *xprt, chunk_off = 0; while (write_len) { ret = send_write(xprt, rqstp, - ntohl(ch->rs_handle), + be32_to_cpu(ch->rs_handle), rs_offset + chunk_off, xdr_off, write_len, @@ -382,6 +457,7 @@ static int send_reply(struct svcxprt_rdma *rdma, int byte_count) { struct ib_send_wr send_wr; + u32 xdr_off; int sge_no; int sge_bytes; int page_no; @@ -416,8 +492,8 @@ static int send_reply(struct svcxprt_rdma *rdma, ctxt->direction = DMA_TO_DEVICE; /* Map the payload indicated by 'byte_count' */ + xdr_off = 0; for (sge_no = 1; byte_count && sge_no < vec->count; sge_no++) { - int xdr_off = 0; sge_bytes = min_t(size_t, vec->sge[sge_no].iov_len, byte_count); byte_count -= sge_bytes; ctxt->sge[sge_no].addr = @@ -455,6 +531,13 @@ static int send_reply(struct svcxprt_rdma *rdma, } rqstp->rq_next_page = rqstp->rq_respages + 1; + /* The loop above bumps sc_dma_used for each sge. The + * xdr_buf.tail gets a separate sge, but resides in the + * same page as xdr_buf.head. Don't count it twice. + */ + if (sge_no > ctxt->count) + atomic_dec(&rdma->sc_dma_used); + if (sge_no > rdma->sc_max_sge) { pr_err("svcrdma: Too many sges (%d)\n", sge_no); goto err; @@ -515,7 +598,7 @@ int svc_rdma_sendto(struct svc_rqst *rqstp) inline_bytes = rqstp->rq_res.len; /* Create the RDMA response header */ - res_page = svc_rdma_get_page(); + res_page = alloc_page(GFP_KERNEL | __GFP_NOFAIL); rdma_resp = page_address(res_page); reply_ary = svc_rdma_get_reply_array(rdma_argp); if (reply_ary) diff --git a/kernel/net/sunrpc/xprtrdma/svc_rdma_transport.c b/kernel/net/sunrpc/xprtrdma/svc_rdma_transport.c index f609c1c2d..b348b4ade 100644 --- a/kernel/net/sunrpc/xprtrdma/svc_rdma_transport.c +++ b/kernel/net/sunrpc/xprtrdma/svc_rdma_transport.c @@ -56,6 +56,7 @@ #define RPCDBG_FACILITY RPCDBG_SVCXPRT +static struct svcxprt_rdma *rdma_create_xprt(struct svc_serv *, int); static struct svc_xprt *svc_rdma_create(struct svc_serv *serv, struct net *net, struct sockaddr *sa, int salen, @@ -95,16 +96,69 @@ struct svc_xprt_class svc_rdma_class = { .xcl_ident = XPRT_TRANSPORT_RDMA, }; +#if defined(CONFIG_SUNRPC_BACKCHANNEL) +static struct svc_xprt *svc_rdma_bc_create(struct svc_serv *, struct net *, + struct sockaddr *, int, int); +static void svc_rdma_bc_detach(struct svc_xprt *); +static void svc_rdma_bc_free(struct svc_xprt *); + +static struct svc_xprt_ops svc_rdma_bc_ops = { + .xpo_create = svc_rdma_bc_create, + .xpo_detach = svc_rdma_bc_detach, + .xpo_free = svc_rdma_bc_free, + .xpo_prep_reply_hdr = svc_rdma_prep_reply_hdr, + .xpo_secure_port = svc_rdma_secure_port, +}; + +struct svc_xprt_class svc_rdma_bc_class = { + .xcl_name = "rdma-bc", + .xcl_owner = THIS_MODULE, + .xcl_ops = &svc_rdma_bc_ops, + .xcl_max_payload = (1024 - RPCRDMA_HDRLEN_MIN) +}; + +static struct svc_xprt *svc_rdma_bc_create(struct svc_serv *serv, + struct net *net, + struct sockaddr *sa, int salen, + int flags) +{ + struct svcxprt_rdma *cma_xprt; + struct svc_xprt *xprt; + + cma_xprt = rdma_create_xprt(serv, 0); + if (!cma_xprt) + return ERR_PTR(-ENOMEM); + xprt = &cma_xprt->sc_xprt; + + svc_xprt_init(net, &svc_rdma_bc_class, xprt, serv); + serv->sv_bc_xprt = xprt; + + dprintk("svcrdma: %s(%p)\n", __func__, xprt); + return xprt; +} + +static void svc_rdma_bc_detach(struct svc_xprt *xprt) +{ + dprintk("svcrdma: %s(%p)\n", __func__, xprt); +} + +static void svc_rdma_bc_free(struct svc_xprt *xprt) +{ + struct svcxprt_rdma *rdma = + container_of(xprt, struct svcxprt_rdma, sc_xprt); + + dprintk("svcrdma: %s(%p)\n", __func__, xprt); + if (xprt) + kfree(rdma); +} +#endif /* CONFIG_SUNRPC_BACKCHANNEL */ + struct svc_rdma_op_ctxt *svc_rdma_get_context(struct svcxprt_rdma *xprt) { struct svc_rdma_op_ctxt *ctxt; - while (1) { - ctxt = kmem_cache_alloc(svc_rdma_ctxt_cachep, GFP_KERNEL); - if (ctxt) - break; - schedule_timeout_uninterruptible(msecs_to_jiffies(500)); - } + ctxt = kmem_cache_alloc(svc_rdma_ctxt_cachep, + GFP_KERNEL | __GFP_NOFAIL); ctxt->xprt = xprt; INIT_LIST_HEAD(&ctxt->dto_q); ctxt->count = 0; @@ -156,12 +210,8 @@ void svc_rdma_put_context(struct svc_rdma_op_ctxt *ctxt, int free_pages) struct svc_rdma_req_map *svc_rdma_get_req_map(void) { struct svc_rdma_req_map *map; - while (1) { - map = kmem_cache_alloc(svc_rdma_map_cachep, GFP_KERNEL); - if (map) - break; - schedule_timeout_uninterruptible(msecs_to_jiffies(500)); - } + map = kmem_cache_alloc(svc_rdma_map_cachep, + GFP_KERNEL | __GFP_NOFAIL); map->count = 0; return map; } @@ -175,8 +225,8 @@ void svc_rdma_put_req_map(struct svc_rdma_req_map *map) static void cq_event_handler(struct ib_event *event, void *context) { struct svc_xprt *xprt = context; - dprintk("svcrdma: received CQ event id=%d, context=%p\n", - event->event, context); + dprintk("svcrdma: received CQ event %s (%d), context=%p\n", + ib_event_msg(event->event), event->event, context); set_bit(XPT_CLOSE, &xprt->xpt_flags); } @@ -191,8 +241,9 @@ static void qp_event_handler(struct ib_event *event, void *context) case IB_EVENT_COMM_EST: case IB_EVENT_SQ_DRAINED: case IB_EVENT_QP_LAST_WQE_REACHED: - dprintk("svcrdma: QP event %d received for QP=%p\n", - event->event, event->element.qp); + dprintk("svcrdma: QP event %s (%d) received for QP=%p\n", + ib_event_msg(event->event), event->event, + event->element.qp); break; /* These are considered fatal events */ case IB_EVENT_PATH_MIG_ERR: @@ -201,9 +252,10 @@ static void qp_event_handler(struct ib_event *event, void *context) case IB_EVENT_QP_ACCESS_ERR: case IB_EVENT_DEVICE_FATAL: default: - dprintk("svcrdma: QP ERROR event %d received for QP=%p, " + dprintk("svcrdma: QP ERROR event %s (%d) received for QP=%p, " "closing transport\n", - event->event, event->element.qp); + ib_event_msg(event->event), event->event, + event->element.qp); set_bit(XPT_CLOSE, &xprt->xpt_flags); break; } @@ -402,7 +454,8 @@ static void sq_cq_reap(struct svcxprt_rdma *xprt) for (i = 0; i < ret; i++) { wc = &wc_a[i]; if (wc->status != IB_WC_SUCCESS) { - dprintk("svcrdma: sq wc err status %d\n", + dprintk("svcrdma: sq wc err status %s (%d)\n", + ib_wc_status_msg(wc->status), wc->status); /* Close the transport */ @@ -490,18 +543,6 @@ static struct svcxprt_rdma *rdma_create_xprt(struct svc_serv *serv, return cma_xprt; } -struct page *svc_rdma_get_page(void) -{ - struct page *page; - - while ((page = alloc_page(GFP_KERNEL)) == NULL) { - /* If we can't get memory, wait a bit and try again */ - printk(KERN_INFO "svcrdma: out of memory...retrying in 1s\n"); - schedule_timeout_uninterruptible(msecs_to_jiffies(1000)); - } - return page; -} - int svc_rdma_post_recv(struct svcxprt_rdma *xprt) { struct ib_recv_wr recv_wr, *bad_recv_wr; @@ -520,7 +561,7 @@ int svc_rdma_post_recv(struct svcxprt_rdma *xprt) pr_err("svcrdma: Too many sges (%d)\n", sge_no); goto err_put_ctxt; } - page = svc_rdma_get_page(); + page = alloc_page(GFP_KERNEL | __GFP_NOFAIL); ctxt->pages[sge_no] = page; pa = ib_dma_map_page(xprt->sc_cm_id->device, page, 0, PAGE_SIZE, @@ -616,7 +657,8 @@ static int rdma_listen_handler(struct rdma_cm_id *cma_id, switch (event->event) { case RDMA_CM_EVENT_CONNECT_REQUEST: dprintk("svcrdma: Connect request on cma_id=%p, xprt = %p, " - "event=%d\n", cma_id, cma_id->context, event->event); + "event = %s (%d)\n", cma_id, cma_id->context, + rdma_event_msg(event->event), event->event); handle_connect_req(cma_id, event->param.conn.initiator_depth); break; @@ -636,7 +678,8 @@ static int rdma_listen_handler(struct rdma_cm_id *cma_id, default: dprintk("svcrdma: Unexpected event on listening endpoint %p, " - "event=%d\n", cma_id, event->event); + "event = %s (%d)\n", cma_id, + rdma_event_msg(event->event), event->event); break; } @@ -669,15 +712,18 @@ static int rdma_cma_handler(struct rdma_cm_id *cma_id, break; case RDMA_CM_EVENT_DEVICE_REMOVAL: dprintk("svcrdma: Device removal cma_id=%p, xprt = %p, " - "event=%d\n", cma_id, xprt, event->event); + "event = %s (%d)\n", cma_id, xprt, + rdma_event_msg(event->event), event->event); if (xprt) { set_bit(XPT_CLOSE, &xprt->xpt_flags); svc_xprt_enqueue(xprt); + svc_xprt_put(xprt); } break; default: dprintk("svcrdma: Unexpected event on DTO endpoint %p, " - "event=%d\n", cma_id, event->event); + "event = %s (%d)\n", cma_id, + rdma_event_msg(event->event), event->event); break; } return 0; @@ -704,8 +750,8 @@ static struct svc_xprt *svc_rdma_create(struct svc_serv *serv, if (!cma_xprt) return ERR_PTR(-ENOMEM); - listen_id = rdma_create_id(rdma_listen_handler, cma_xprt, RDMA_PS_TCP, - IB_QPT_RC); + listen_id = rdma_create_id(&init_net, rdma_listen_handler, cma_xprt, + RDMA_PS_TCP, IB_QPT_RC); if (IS_ERR(listen_id)) { ret = PTR_ERR(listen_id); dprintk("svcrdma: rdma_create_id failed = %d\n", ret); @@ -744,24 +790,27 @@ static struct svc_xprt *svc_rdma_create(struct svc_serv *serv, static struct svc_rdma_fastreg_mr *rdma_alloc_frmr(struct svcxprt_rdma *xprt) { struct ib_mr *mr; - struct ib_fast_reg_page_list *pl; + struct scatterlist *sg; struct svc_rdma_fastreg_mr *frmr; + u32 num_sg; frmr = kmalloc(sizeof(*frmr), GFP_KERNEL); if (!frmr) goto err; - mr = ib_alloc_fast_reg_mr(xprt->sc_pd, RPCSVC_MAXPAGES); + num_sg = min_t(u32, RPCSVC_MAXPAGES, xprt->sc_frmr_pg_list_len); + mr = ib_alloc_mr(xprt->sc_pd, IB_MR_TYPE_MEM_REG, num_sg); if (IS_ERR(mr)) goto err_free_frmr; - pl = ib_alloc_fast_reg_page_list(xprt->sc_cm_id->device, - RPCSVC_MAXPAGES); - if (IS_ERR(pl)) + sg = kcalloc(RPCSVC_MAXPAGES, sizeof(*sg), GFP_KERNEL); + if (!sg) goto err_free_mr; + sg_init_table(sg, RPCSVC_MAXPAGES); + frmr->mr = mr; - frmr->page_list = pl; + frmr->sg = sg; INIT_LIST_HEAD(&frmr->frmr_list); return frmr; @@ -781,8 +830,8 @@ static void rdma_dealloc_frmr_q(struct svcxprt_rdma *xprt) frmr = list_entry(xprt->sc_frmr_q.next, struct svc_rdma_fastreg_mr, frmr_list); list_del_init(&frmr->frmr_list); + kfree(frmr->sg); ib_dereg_mr(frmr->mr); - ib_free_fast_reg_page_list(frmr->page_list); kfree(frmr); } } @@ -796,8 +845,7 @@ struct svc_rdma_fastreg_mr *svc_rdma_get_frmr(struct svcxprt_rdma *rdma) frmr = list_entry(rdma->sc_frmr_q.next, struct svc_rdma_fastreg_mr, frmr_list); list_del_init(&frmr->frmr_list); - frmr->map_len = 0; - frmr->page_list_len = 0; + frmr->sg_nents = 0; } spin_unlock_bh(&rdma->sc_frmr_q_lock); if (frmr) @@ -806,25 +854,13 @@ struct svc_rdma_fastreg_mr *svc_rdma_get_frmr(struct svcxprt_rdma *rdma) return rdma_alloc_frmr(rdma); } -static void frmr_unmap_dma(struct svcxprt_rdma *xprt, - struct svc_rdma_fastreg_mr *frmr) -{ - int page_no; - for (page_no = 0; page_no < frmr->page_list_len; page_no++) { - dma_addr_t addr = frmr->page_list->page_list[page_no]; - if (ib_dma_mapping_error(frmr->mr->device, addr)) - continue; - atomic_dec(&xprt->sc_dma_used); - ib_dma_unmap_page(frmr->mr->device, addr, PAGE_SIZE, - frmr->direction); - } -} - void svc_rdma_put_frmr(struct svcxprt_rdma *rdma, struct svc_rdma_fastreg_mr *frmr) { if (frmr) { - frmr_unmap_dma(rdma, frmr); + ib_dma_unmap_sg(rdma->sc_cm_id->device, + frmr->sg, frmr->sg_nents, frmr->direction); + atomic_dec(&rdma->sc_dma_used); spin_lock_bh(&rdma->sc_frmr_q_lock); WARN_ON_ONCE(!list_empty(&frmr->frmr_list)); list_add(&frmr->frmr_list, &rdma->sc_frmr_q); @@ -848,10 +884,11 @@ static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt) struct svcxprt_rdma *listen_rdma; struct svcxprt_rdma *newxprt = NULL; struct rdma_conn_param conn_param; + struct ib_cq_init_attr cq_attr = {}; struct ib_qp_init_attr qp_attr; struct ib_device_attr devattr; int uninitialized_var(dma_mr_acc); - int need_dma_mr; + int need_dma_mr = 0; int ret; int i; @@ -884,6 +921,8 @@ static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt) * capabilities of this particular device */ newxprt->sc_max_sge = min((size_t)devattr.max_sge, (size_t)RPCSVC_MAXPAGES); + newxprt->sc_max_sge_rd = min_t(size_t, devattr.max_sge_rd, + RPCSVC_MAXPAGES); newxprt->sc_max_requests = min((size_t)devattr.max_qp_wr, (size_t)svcrdma_max_requests); newxprt->sc_sq_depth = RPCRDMA_SQ_DEPTH_MULT * newxprt->sc_max_requests; @@ -900,22 +939,22 @@ static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt) dprintk("svcrdma: error creating PD for connect request\n"); goto errout; } + cq_attr.cqe = newxprt->sc_sq_depth; newxprt->sc_sq_cq = ib_create_cq(newxprt->sc_cm_id->device, sq_comp_handler, cq_event_handler, newxprt, - newxprt->sc_sq_depth, - 0); + &cq_attr); if (IS_ERR(newxprt->sc_sq_cq)) { dprintk("svcrdma: error creating SQ CQ for connect request\n"); goto errout; } + cq_attr.cqe = newxprt->sc_max_requests; newxprt->sc_rq_cq = ib_create_cq(newxprt->sc_cm_id->device, rq_comp_handler, cq_event_handler, newxprt, - newxprt->sc_max_requests, - 0); + &cq_attr); if (IS_ERR(newxprt->sc_rq_cq)) { dprintk("svcrdma: error creating RQ CQ for connect request\n"); goto errout; @@ -985,35 +1024,26 @@ static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt) /* * Determine if a DMA MR is required and if so, what privs are required */ - switch (rdma_node_get_transport(newxprt->sc_cm_id->device->node_type)) { - case RDMA_TRANSPORT_IWARP: - newxprt->sc_dev_caps |= SVCRDMA_DEVCAP_READ_W_INV; - if (!(newxprt->sc_dev_caps & SVCRDMA_DEVCAP_FAST_REG)) { - need_dma_mr = 1; - dma_mr_acc = - (IB_ACCESS_LOCAL_WRITE | - IB_ACCESS_REMOTE_WRITE); - } else if (!(devattr.device_cap_flags & IB_DEVICE_LOCAL_DMA_LKEY)) { - need_dma_mr = 1; - dma_mr_acc = IB_ACCESS_LOCAL_WRITE; - } else - need_dma_mr = 0; - break; - case RDMA_TRANSPORT_IB: - if (!(newxprt->sc_dev_caps & SVCRDMA_DEVCAP_FAST_REG)) { - need_dma_mr = 1; - dma_mr_acc = IB_ACCESS_LOCAL_WRITE; - } else if (!(devattr.device_cap_flags & - IB_DEVICE_LOCAL_DMA_LKEY)) { - need_dma_mr = 1; - dma_mr_acc = IB_ACCESS_LOCAL_WRITE; - } else - need_dma_mr = 0; - break; - default: + if (!rdma_protocol_iwarp(newxprt->sc_cm_id->device, + newxprt->sc_cm_id->port_num) && + !rdma_ib_or_roce(newxprt->sc_cm_id->device, + newxprt->sc_cm_id->port_num)) goto errout; + + if (!(newxprt->sc_dev_caps & SVCRDMA_DEVCAP_FAST_REG) || + !(devattr.device_cap_flags & IB_DEVICE_LOCAL_DMA_LKEY)) { + need_dma_mr = 1; + dma_mr_acc = IB_ACCESS_LOCAL_WRITE; + if (rdma_protocol_iwarp(newxprt->sc_cm_id->device, + newxprt->sc_cm_id->port_num) && + !(newxprt->sc_dev_caps & SVCRDMA_DEVCAP_FAST_REG)) + dma_mr_acc |= IB_ACCESS_REMOTE_WRITE; } + if (rdma_protocol_iwarp(newxprt->sc_cm_id->device, + newxprt->sc_cm_id->port_num)) + newxprt->sc_dev_caps |= SVCRDMA_DEVCAP_READ_W_INV; + /* Create the DMA MR if needed, otherwise, use the DMA LKEY */ if (need_dma_mr) { /* Register all of physical memory */ @@ -1067,6 +1097,7 @@ static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt) " remote_ip : %pI4\n" " remote_port : %d\n" " max_sge : %d\n" + " max_sge_rd : %d\n" " sq_depth : %d\n" " max_requests : %d\n" " ord : %d\n", @@ -1080,6 +1111,7 @@ static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt) ntohs(((struct sockaddr_in *)&newxprt->sc_cm_id-> route.addr.dst_addr)->sin_port), newxprt->sc_max_sge, + newxprt->sc_max_sge_rd, newxprt->sc_sq_depth, newxprt->sc_max_requests, newxprt->sc_ord); @@ -1222,40 +1254,6 @@ static int svc_rdma_secure_port(struct svc_rqst *rqstp) return 1; } -/* - * Attempt to register the kvec representing the RPC memory with the - * device. - * - * Returns: - * NULL : The device does not support fastreg or there were no more - * fastreg mr. - * frmr : The kvec register request was successfully posted. - * <0 : An error was encountered attempting to register the kvec. - */ -int svc_rdma_fastreg(struct svcxprt_rdma *xprt, - struct svc_rdma_fastreg_mr *frmr) -{ - struct ib_send_wr fastreg_wr; - u8 key; - - /* Bump the key */ - key = (u8)(frmr->mr->lkey & 0x000000FF); - ib_update_fast_reg_key(frmr->mr, ++key); - - /* Prepare FASTREG WR */ - memset(&fastreg_wr, 0, sizeof fastreg_wr); - fastreg_wr.opcode = IB_WR_FAST_REG_MR; - fastreg_wr.send_flags = IB_SEND_SIGNALED; - fastreg_wr.wr.fast_reg.iova_start = (unsigned long)frmr->kva; - fastreg_wr.wr.fast_reg.page_list = frmr->page_list; - fastreg_wr.wr.fast_reg.page_list_len = frmr->page_list_len; - fastreg_wr.wr.fast_reg.page_shift = PAGE_SHIFT; - fastreg_wr.wr.fast_reg.length = frmr->map_len; - fastreg_wr.wr.fast_reg.access_flags = frmr->access_flags; - fastreg_wr.wr.fast_reg.rkey = frmr->mr->lkey; - return svc_rdma_send(xprt, &fastreg_wr); -} - int svc_rdma_send(struct svcxprt_rdma *xprt, struct ib_send_wr *wr) { struct ib_send_wr *bad_wr, *n_wr; @@ -1319,11 +1317,11 @@ void svc_rdma_send_error(struct svcxprt_rdma *xprt, struct rpcrdma_msg *rmsgp, struct ib_send_wr err_wr; struct page *p; struct svc_rdma_op_ctxt *ctxt; - u32 *va; + __be32 *va; int length; int ret; - p = svc_rdma_get_page(); + p = alloc_page(GFP_KERNEL | __GFP_NOFAIL); va = page_address(p); /* XDR encode error */ diff --git a/kernel/net/sunrpc/xprtrdma/transport.c b/kernel/net/sunrpc/xprtrdma/transport.c index 54f23b1be..8c545f7d7 100644 --- a/kernel/net/sunrpc/xprtrdma/transport.c +++ b/kernel/net/sunrpc/xprtrdma/transport.c @@ -48,7 +48,6 @@ */ #include <linux/module.h> -#include <linux/init.h> #include <linux/slab.h> #include <linux/seq_file.h> #include <linux/sunrpc/addr.h> @@ -59,11 +58,6 @@ # define RPCDBG_FACILITY RPCDBG_TRANS #endif -MODULE_LICENSE("Dual BSD/GPL"); - -MODULE_DESCRIPTION("RPC/RDMA Transport for Linux kernel NFS"); -MODULE_AUTHOR("Network Appliance, Inc."); - /* * tunables */ @@ -181,10 +175,8 @@ xprt_rdma_format_addresses6(struct rpc_xprt *xprt, struct sockaddr *sap) } static void -xprt_rdma_format_addresses(struct rpc_xprt *xprt) +xprt_rdma_format_addresses(struct rpc_xprt *xprt, struct sockaddr *sap) { - struct sockaddr *sap = (struct sockaddr *) - &rpcx_to_rdmad(xprt).addr; char buf[128]; switch (sap->sa_family) { @@ -246,6 +238,16 @@ xprt_rdma_connect_worker(struct work_struct *work) xprt_clear_connecting(xprt); } +static void +xprt_rdma_inject_disconnect(struct rpc_xprt *xprt) +{ + struct rpcrdma_xprt *r_xprt = container_of(xprt, struct rpcrdma_xprt, + rx_xprt); + + pr_info("rpcrdma: injecting transport disconnect on xprt=%p\n", xprt); + rdma_disconnect(r_xprt->rx_ia.ri_id); +} + /* * xprt_rdma_destroy * @@ -268,8 +270,8 @@ xprt_rdma_destroy(struct rpc_xprt *xprt) xprt_clear_connected(xprt); - rpcrdma_buffer_destroy(&r_xprt->rx_buf); rpcrdma_ep_destroy(&r_xprt->rx_ep, &r_xprt->rx_ia); + rpcrdma_buffer_destroy(&r_xprt->rx_buf); rpcrdma_ia_close(&r_xprt->rx_ia); xprt_rdma_free_addresses(xprt); @@ -298,7 +300,7 @@ xprt_setup_rdma(struct xprt_create *args) struct rpc_xprt *xprt; struct rpcrdma_xprt *new_xprt; struct rpcrdma_ep *new_ep; - struct sockaddr_in *sin; + struct sockaddr *sap; int rc; if (args->addrlen > sizeof(xprt->addr)) { @@ -329,26 +331,20 @@ xprt_setup_rdma(struct xprt_create *args) * Set up RDMA-specific connect data. */ - /* Put server RDMA address in local cdata */ - memcpy(&cdata.addr, args->dstaddr, args->addrlen); + sap = (struct sockaddr *)&cdata.addr; + memcpy(sap, args->dstaddr, args->addrlen); /* Ensure xprt->addr holds valid server TCP (not RDMA) * address, for any side protocols which peek at it */ xprt->prot = IPPROTO_TCP; xprt->addrlen = args->addrlen; - memcpy(&xprt->addr, &cdata.addr, xprt->addrlen); + memcpy(&xprt->addr, sap, xprt->addrlen); - sin = (struct sockaddr_in *)&cdata.addr; - if (ntohs(sin->sin_port) != 0) + if (rpc_get_port(sap)) xprt_set_bound(xprt); - dprintk("RPC: %s: %pI4:%u\n", - __func__, &sin->sin_addr.s_addr, ntohs(sin->sin_port)); - - /* Set max requests */ cdata.max_requests = xprt->max_reqs; - /* Set some length limits */ cdata.rsize = RPCRDMA_MAX_SEGS * PAGE_SIZE; /* RDMA write max */ cdata.wsize = RPCRDMA_MAX_SEGS * PAGE_SIZE; /* RDMA read max */ @@ -371,8 +367,7 @@ xprt_setup_rdma(struct xprt_create *args) new_xprt = rpcx_to_rdmax(xprt); - rc = rpcrdma_ia_open(new_xprt, (struct sockaddr *) &cdata.addr, - xprt_rdma_memreg_strategy); + rc = rpcrdma_ia_open(new_xprt, sap, xprt_rdma_memreg_strategy); if (rc) goto out1; @@ -405,7 +400,7 @@ xprt_setup_rdma(struct xprt_create *args) INIT_DELAYED_WORK(&new_xprt->rx_connect_worker, xprt_rdma_connect_worker); - xprt_rdma_format_addresses(xprt); + xprt_rdma_format_addresses(xprt, sap); xprt->max_payload = new_xprt->rx_ia.ri_ops->ro_maxpages(new_xprt); if (xprt->max_payload == 0) goto out4; @@ -416,6 +411,9 @@ xprt_setup_rdma(struct xprt_create *args) if (!try_module_get(THIS_MODULE)) goto out4; + dprintk("RPC: %s: %s:%s\n", __func__, + xprt->address_strings[RPC_DISPLAY_ADDR], + xprt->address_strings[RPC_DISPLAY_PORT]); return xprt; out4: @@ -618,12 +616,6 @@ xprt_rdma_send_request(struct rpc_task *task) if (req->rl_reply == NULL) /* e.g. reconnection */ rpcrdma_recv_buffer_get(req); - if (req->rl_reply) { - req->rl_reply->rr_func = rpcrdma_reply_handler; - /* this need only be done once, but... */ - req->rl_reply->rr_xprt = xprt; - } - /* Must suppress retransmit to maintain credits */ if (req->rl_connect_cookie == xprt->connect_cookie) goto drop_connection; @@ -655,31 +647,41 @@ static void xprt_rdma_print_stats(struct rpc_xprt *xprt, struct seq_file *seq) if (xprt_connected(xprt)) idle_time = (long)(jiffies - xprt->last_used) / HZ; - seq_printf(seq, - "\txprt:\trdma %u %lu %lu %lu %ld %lu %lu %lu %Lu %Lu " - "%lu %lu %lu %Lu %Lu %Lu %Lu %lu %lu %lu\n", - - 0, /* need a local port? */ - xprt->stat.bind_count, - xprt->stat.connect_count, - xprt->stat.connect_time, - idle_time, - xprt->stat.sends, - xprt->stat.recvs, - xprt->stat.bad_xids, - xprt->stat.req_u, - xprt->stat.bklog_u, - - r_xprt->rx_stats.read_chunk_count, - r_xprt->rx_stats.write_chunk_count, - r_xprt->rx_stats.reply_chunk_count, - r_xprt->rx_stats.total_rdma_request, - r_xprt->rx_stats.total_rdma_reply, - r_xprt->rx_stats.pullup_copy_count, - r_xprt->rx_stats.fixup_copy_count, - r_xprt->rx_stats.hardway_register_count, - r_xprt->rx_stats.failed_marshal_count, - r_xprt->rx_stats.bad_reply_count); + seq_puts(seq, "\txprt:\trdma "); + seq_printf(seq, "%u %lu %lu %lu %ld %lu %lu %lu %llu %llu ", + 0, /* need a local port? */ + xprt->stat.bind_count, + xprt->stat.connect_count, + xprt->stat.connect_time, + idle_time, + xprt->stat.sends, + xprt->stat.recvs, + xprt->stat.bad_xids, + xprt->stat.req_u, + xprt->stat.bklog_u); + seq_printf(seq, "%lu %lu %lu %llu %llu %llu %llu %lu %lu %lu %lu\n", + r_xprt->rx_stats.read_chunk_count, + r_xprt->rx_stats.write_chunk_count, + r_xprt->rx_stats.reply_chunk_count, + r_xprt->rx_stats.total_rdma_request, + r_xprt->rx_stats.total_rdma_reply, + r_xprt->rx_stats.pullup_copy_count, + r_xprt->rx_stats.fixup_copy_count, + r_xprt->rx_stats.hardway_register_count, + r_xprt->rx_stats.failed_marshal_count, + r_xprt->rx_stats.bad_reply_count, + r_xprt->rx_stats.nomsg_call_count); +} + +static int +xprt_rdma_enable_swap(struct rpc_xprt *xprt) +{ + return 0; +} + +static void +xprt_rdma_disable_swap(struct rpc_xprt *xprt) +{ } /* @@ -700,7 +702,16 @@ static struct rpc_xprt_ops xprt_rdma_procs = { .send_request = xprt_rdma_send_request, .close = xprt_rdma_close, .destroy = xprt_rdma_destroy, - .print_stats = xprt_rdma_print_stats + .print_stats = xprt_rdma_print_stats, + .enable_swap = xprt_rdma_enable_swap, + .disable_swap = xprt_rdma_disable_swap, + .inject_disconnect = xprt_rdma_inject_disconnect, +#if defined(CONFIG_SUNRPC_BACKCHANNEL) + .bc_setup = xprt_rdma_bc_setup, + .bc_up = xprt_rdma_bc_up, + .bc_free_rqst = xprt_rdma_bc_free_rqst, + .bc_destroy = xprt_rdma_bc_destroy, +#endif }; static struct xprt_class xprt_rdma = { @@ -711,7 +722,7 @@ static struct xprt_class xprt_rdma = { .setup = xprt_setup_rdma, }; -static void __exit xprt_rdma_cleanup(void) +void xprt_rdma_cleanup(void) { int rc; @@ -726,17 +737,32 @@ static void __exit xprt_rdma_cleanup(void) if (rc) dprintk("RPC: %s: xprt_unregister returned %i\n", __func__, rc); + + rpcrdma_destroy_wq(); + frwr_destroy_recovery_wq(); } -static int __init xprt_rdma_init(void) +int xprt_rdma_init(void) { int rc; - rc = xprt_register_transport(&xprt_rdma); - + rc = frwr_alloc_recovery_wq(); if (rc) return rc; + rc = rpcrdma_alloc_wq(); + if (rc) { + frwr_destroy_recovery_wq(); + return rc; + } + + rc = xprt_register_transport(&xprt_rdma); + if (rc) { + rpcrdma_destroy_wq(); + frwr_destroy_recovery_wq(); + return rc; + } + dprintk("RPCRDMA Module Init, register RPC RDMA transport\n"); dprintk("Defaults:\n"); @@ -753,6 +779,3 @@ static int __init xprt_rdma_init(void) #endif return 0; } - -module_init(xprt_rdma_init); -module_exit(xprt_rdma_cleanup); diff --git a/kernel/net/sunrpc/xprtrdma/verbs.c b/kernel/net/sunrpc/xprtrdma/verbs.c index 4870d272e..eadd16551 100644 --- a/kernel/net/sunrpc/xprtrdma/verbs.c +++ b/kernel/net/sunrpc/xprtrdma/verbs.c @@ -52,6 +52,7 @@ #include <linux/prefetch.h> #include <linux/sunrpc/addr.h> #include <asm/bitops.h> +#include <linux/module.h> /* try_module_get()/module_put() */ #include "xprt_rdma.h" @@ -67,79 +68,33 @@ * internal functions */ -/* - * handle replies in tasklet context, using a single, global list - * rdma tasklet function -- just turn around and call the func - * for all replies on the list - */ - -static DEFINE_SPINLOCK(rpcrdma_tk_lock_g); -static LIST_HEAD(rpcrdma_tasklets_g); +static struct workqueue_struct *rpcrdma_receive_wq; -static void -rpcrdma_run_tasklet(unsigned long data) +int +rpcrdma_alloc_wq(void) { - struct rpcrdma_rep *rep; - void (*func)(struct rpcrdma_rep *); - unsigned long flags; + struct workqueue_struct *recv_wq; - data = data; - spin_lock_irqsave(&rpcrdma_tk_lock_g, flags); - while (!list_empty(&rpcrdma_tasklets_g)) { - rep = list_entry(rpcrdma_tasklets_g.next, - struct rpcrdma_rep, rr_list); - list_del(&rep->rr_list); - func = rep->rr_func; - rep->rr_func = NULL; - spin_unlock_irqrestore(&rpcrdma_tk_lock_g, flags); - - if (func) - func(rep); - else - rpcrdma_recv_buffer_put(rep); - - spin_lock_irqsave(&rpcrdma_tk_lock_g, flags); - } - spin_unlock_irqrestore(&rpcrdma_tk_lock_g, flags); -} + recv_wq = alloc_workqueue("xprtrdma_receive", + WQ_MEM_RECLAIM | WQ_UNBOUND | WQ_HIGHPRI, + 0); + if (!recv_wq) + return -ENOMEM; -static DECLARE_TASKLET(rpcrdma_tasklet_g, rpcrdma_run_tasklet, 0UL); - -static const char * const async_event[] = { - "CQ error", - "QP fatal error", - "QP request error", - "QP access error", - "communication established", - "send queue drained", - "path migration successful", - "path mig error", - "device fatal error", - "port active", - "port error", - "LID change", - "P_key change", - "SM change", - "SRQ error", - "SRQ limit reached", - "last WQE reached", - "client reregister", - "GID change", -}; - -#define ASYNC_MSG(status) \ - ((status) < ARRAY_SIZE(async_event) ? \ - async_event[(status)] : "unknown async error") + rpcrdma_receive_wq = recv_wq; + return 0; +} -static void -rpcrdma_schedule_tasklet(struct list_head *sched_list) +void +rpcrdma_destroy_wq(void) { - unsigned long flags; + struct workqueue_struct *wq; - spin_lock_irqsave(&rpcrdma_tk_lock_g, flags); - list_splice_tail(sched_list, &rpcrdma_tasklets_g); - spin_unlock_irqrestore(&rpcrdma_tk_lock_g, flags); - tasklet_schedule(&rpcrdma_tasklet_g); + if (rpcrdma_receive_wq) { + wq = rpcrdma_receive_wq; + rpcrdma_receive_wq = NULL; + destroy_workqueue(wq); + } } static void @@ -148,7 +103,7 @@ rpcrdma_qp_async_error_upcall(struct ib_event *event, void *context) struct rpcrdma_ep *ep = context; pr_err("RPC: %s: %s on device %s ep %p\n", - __func__, ASYNC_MSG(event->event), + __func__, ib_event_msg(event->event), event->device->name, context); if (ep->rep_connected == 1) { ep->rep_connected = -EIO; @@ -163,7 +118,7 @@ rpcrdma_cq_async_error_upcall(struct ib_event *event, void *context) struct rpcrdma_ep *ep = context; pr_err("RPC: %s: %s on device %s ep %p\n", - __func__, ASYNC_MSG(event->event), + __func__, ib_event_msg(event->event), event->device->name, context); if (ep->rep_connected == 1) { ep->rep_connected = -EIO; @@ -172,35 +127,6 @@ rpcrdma_cq_async_error_upcall(struct ib_event *event, void *context) } } -static const char * const wc_status[] = { - "success", - "local length error", - "local QP operation error", - "local EE context operation error", - "local protection error", - "WR flushed", - "memory management operation error", - "bad response error", - "local access error", - "remote invalid request error", - "remote access error", - "remote operation error", - "transport retry counter exceeded", - "RNR retry counter exceeded", - "local RDD violation error", - "remove invalid RD request", - "operation aborted", - "invalid EE context number", - "invalid EE context state", - "fatal error", - "response timeout error", - "general error", -}; - -#define COMPLETION_MSG(status) \ - ((status) < ARRAY_SIZE(wc_status) ? \ - wc_status[(status)] : "unexpected completion error") - static void rpcrdma_sendcq_process_wc(struct ib_wc *wc) { @@ -209,7 +135,7 @@ rpcrdma_sendcq_process_wc(struct ib_wc *wc) if (wc->status != IB_WC_SUCCESS && wc->status != IB_WC_WR_FLUSH_ERR) pr_err("RPC: %s: SEND: %s\n", - __func__, COMPLETION_MSG(wc->status)); + __func__, ib_wc_status_msg(wc->status)); } else { struct rpcrdma_mw *r; @@ -218,63 +144,54 @@ rpcrdma_sendcq_process_wc(struct ib_wc *wc) } } -static int -rpcrdma_sendcq_poll(struct ib_cq *cq, struct rpcrdma_ep *ep) +/* The common case is a single send completion is waiting. By + * passing two WC entries to ib_poll_cq, a return code of 1 + * means there is exactly one WC waiting and no more. We don't + * have to invoke ib_poll_cq again to know that the CQ has been + * properly drained. + */ +static void +rpcrdma_sendcq_poll(struct ib_cq *cq) { - struct ib_wc *wcs; - int budget, count, rc; + struct ib_wc *pos, wcs[2]; + int count, rc; - budget = RPCRDMA_WC_BUDGET / RPCRDMA_POLLSIZE; do { - wcs = ep->rep_send_wcs; + pos = wcs; - rc = ib_poll_cq(cq, RPCRDMA_POLLSIZE, wcs); - if (rc <= 0) - return rc; + rc = ib_poll_cq(cq, ARRAY_SIZE(wcs), pos); + if (rc < 0) + break; count = rc; while (count-- > 0) - rpcrdma_sendcq_process_wc(wcs++); - } while (rc == RPCRDMA_POLLSIZE && --budget); - return 0; + rpcrdma_sendcq_process_wc(pos++); + } while (rc == ARRAY_SIZE(wcs)); + return; } -/* - * Handle send, fast_reg_mr, and local_inv completions. - * - * Send events are typically suppressed and thus do not result - * in an upcall. Occasionally one is signaled, however. This - * prevents the provider's completion queue from wrapping and - * losing a completion. +/* Handle provider send completion upcalls. */ static void rpcrdma_sendcq_upcall(struct ib_cq *cq, void *cq_context) { - struct rpcrdma_ep *ep = (struct rpcrdma_ep *)cq_context; - int rc; - - rc = rpcrdma_sendcq_poll(cq, ep); - if (rc) { - dprintk("RPC: %s: ib_poll_cq failed: %i\n", - __func__, rc); - return; - } + do { + rpcrdma_sendcq_poll(cq); + } while (ib_req_notify_cq(cq, IB_CQ_NEXT_COMP | + IB_CQ_REPORT_MISSED_EVENTS) > 0); +} - rc = ib_req_notify_cq(cq, - IB_CQ_NEXT_COMP | IB_CQ_REPORT_MISSED_EVENTS); - if (rc == 0) - return; - if (rc < 0) { - dprintk("RPC: %s: ib_req_notify_cq failed: %i\n", - __func__, rc); - return; - } +static void +rpcrdma_receive_worker(struct work_struct *work) +{ + struct rpcrdma_rep *rep = + container_of(work, struct rpcrdma_rep, rr_work); - rpcrdma_sendcq_poll(cq, ep); + rpcrdma_reply_handler(rep); } static void -rpcrdma_recvcq_process_wc(struct ib_wc *wc, struct list_head *sched_list) +rpcrdma_recvcq_process_wc(struct ib_wc *wc) { struct rpcrdma_rep *rep = (struct rpcrdma_rep *)(unsigned long)wc->wr_id; @@ -291,126 +208,70 @@ rpcrdma_recvcq_process_wc(struct ib_wc *wc, struct list_head *sched_list) __func__, rep, wc->byte_len); rep->rr_len = wc->byte_len; - ib_dma_sync_single_for_cpu(rdmab_to_ia(rep->rr_buffer)->ri_id->device, + ib_dma_sync_single_for_cpu(rep->rr_device, rdmab_addr(rep->rr_rdmabuf), rep->rr_len, DMA_FROM_DEVICE); prefetch(rdmab_to_msg(rep->rr_rdmabuf)); out_schedule: - list_add_tail(&rep->rr_list, sched_list); + queue_work(rpcrdma_receive_wq, &rep->rr_work); return; + out_fail: if (wc->status != IB_WC_WR_FLUSH_ERR) pr_err("RPC: %s: rep %p: %s\n", - __func__, rep, COMPLETION_MSG(wc->status)); - rep->rr_len = ~0U; + __func__, rep, ib_wc_status_msg(wc->status)); + rep->rr_len = RPCRDMA_BAD_LEN; goto out_schedule; } -static int -rpcrdma_recvcq_poll(struct ib_cq *cq, struct rpcrdma_ep *ep) +/* The wc array is on stack: automatic memory is always CPU-local. + * + * struct ib_wc is 64 bytes, making the poll array potentially + * large. But this is at the bottom of the call chain. Further + * substantial work is done in another thread. + */ +static void +rpcrdma_recvcq_poll(struct ib_cq *cq) { - struct list_head sched_list; - struct ib_wc *wcs; - int budget, count, rc; + struct ib_wc *pos, wcs[4]; + int count, rc; - INIT_LIST_HEAD(&sched_list); - budget = RPCRDMA_WC_BUDGET / RPCRDMA_POLLSIZE; do { - wcs = ep->rep_recv_wcs; + pos = wcs; - rc = ib_poll_cq(cq, RPCRDMA_POLLSIZE, wcs); - if (rc <= 0) - goto out_schedule; + rc = ib_poll_cq(cq, ARRAY_SIZE(wcs), pos); + if (rc < 0) + break; count = rc; while (count-- > 0) - rpcrdma_recvcq_process_wc(wcs++, &sched_list); - } while (rc == RPCRDMA_POLLSIZE && --budget); - rc = 0; - -out_schedule: - rpcrdma_schedule_tasklet(&sched_list); - return rc; + rpcrdma_recvcq_process_wc(pos++); + } while (rc == ARRAY_SIZE(wcs)); } -/* - * Handle receive completions. - * - * It is reentrant but processes single events in order to maintain - * ordering of receives to keep server credits. - * - * It is the responsibility of the scheduled tasklet to return - * recv buffers to the pool. NOTE: this affects synchronization of - * connection shutdown. That is, the structures required for - * the completion of the reply handler must remain intact until - * all memory has been reclaimed. +/* Handle provider receive completion upcalls. */ static void rpcrdma_recvcq_upcall(struct ib_cq *cq, void *cq_context) { - struct rpcrdma_ep *ep = (struct rpcrdma_ep *)cq_context; - int rc; - - rc = rpcrdma_recvcq_poll(cq, ep); - if (rc) { - dprintk("RPC: %s: ib_poll_cq failed: %i\n", - __func__, rc); - return; - } - - rc = ib_req_notify_cq(cq, - IB_CQ_NEXT_COMP | IB_CQ_REPORT_MISSED_EVENTS); - if (rc == 0) - return; - if (rc < 0) { - dprintk("RPC: %s: ib_req_notify_cq failed: %i\n", - __func__, rc); - return; - } - - rpcrdma_recvcq_poll(cq, ep); + do { + rpcrdma_recvcq_poll(cq); + } while (ib_req_notify_cq(cq, IB_CQ_NEXT_COMP | + IB_CQ_REPORT_MISSED_EVENTS) > 0); } static void rpcrdma_flush_cqs(struct rpcrdma_ep *ep) { struct ib_wc wc; - LIST_HEAD(sched_list); while (ib_poll_cq(ep->rep_attr.recv_cq, 1, &wc) > 0) - rpcrdma_recvcq_process_wc(&wc, &sched_list); - if (!list_empty(&sched_list)) - rpcrdma_schedule_tasklet(&sched_list); + rpcrdma_recvcq_process_wc(&wc); while (ib_poll_cq(ep->rep_attr.send_cq, 1, &wc) > 0) rpcrdma_sendcq_process_wc(&wc); } -#if IS_ENABLED(CONFIG_SUNRPC_DEBUG) -static const char * const conn[] = { - "address resolved", - "address error", - "route resolved", - "route error", - "connect request", - "connect response", - "connect error", - "unreachable", - "rejected", - "established", - "disconnected", - "device removal", - "multicast join", - "multicast error", - "address change", - "timewait exit", -}; - -#define CONNECTION_MSG(status) \ - ((status) < ARRAY_SIZE(conn) ? \ - conn[(status)] : "unrecognized connection error") -#endif - static int rpcrdma_conn_upcall(struct rdma_cm_id *id, struct rdma_cm_event *event) { @@ -476,7 +337,7 @@ connected: default: dprintk("RPC: %s: %pIS:%u (ep 0x%p): %s\n", __func__, sap, rpc_get_port(sap), ep, - CONNECTION_MSG(event->event)); + rdma_event_msg(event->event)); break; } @@ -487,7 +348,7 @@ connected: pr_info("rpcrdma: connection to %pIS:%u on %s, memreg '%s', %d credits, %d responders%s\n", sap, rpc_get_port(sap), - ia->ri_id->device->name, + ia->ri_device->name, ia->ri_ops->ro_displayname, xprt->rx_buf.rb_max_requests, ird, ird < 4 && ird < tird / 2 ? " (low!)" : ""); @@ -500,6 +361,14 @@ connected: return 0; } +static void rpcrdma_destroy_id(struct rdma_cm_id *id) +{ + if (id) { + module_put(id->device->owner); + rdma_destroy_id(id); + } +} + static struct rdma_cm_id * rpcrdma_create_id(struct rpcrdma_xprt *xprt, struct rpcrdma_ia *ia, struct sockaddr *addr) @@ -509,7 +378,8 @@ rpcrdma_create_id(struct rpcrdma_xprt *xprt, init_completion(&ia->ri_done); - id = rdma_create_id(rpcrdma_conn_upcall, xprt, RDMA_PS_TCP, IB_QPT_RC); + id = rdma_create_id(&init_net, rpcrdma_conn_upcall, xprt, RDMA_PS_TCP, + IB_QPT_RC); if (IS_ERR(id)) { rc = PTR_ERR(id); dprintk("RPC: %s: rdma_create_id() failed %i\n", @@ -526,6 +396,17 @@ rpcrdma_create_id(struct rpcrdma_xprt *xprt, } wait_for_completion_interruptible_timeout(&ia->ri_done, msecs_to_jiffies(RDMA_RESOLVE_TIMEOUT) + 1); + + /* FIXME: + * Until xprtrdma supports DEVICE_REMOVAL, the provider must + * be pinned while there are active NFS/RDMA mounts to prevent + * hangs and crashes at umount time. + */ + if (!ia->ri_async_rc && !try_module_get(id->device->owner)) { + dprintk("RPC: %s: Failed to get device module\n", + __func__); + ia->ri_async_rc = -ENODEV; + } rc = ia->ri_async_rc; if (rc) goto out; @@ -535,16 +416,17 @@ rpcrdma_create_id(struct rpcrdma_xprt *xprt, if (rc) { dprintk("RPC: %s: rdma_resolve_route() failed %i\n", __func__, rc); - goto out; + goto put; } wait_for_completion_interruptible_timeout(&ia->ri_done, msecs_to_jiffies(RDMA_RESOLVE_TIMEOUT) + 1); rc = ia->ri_async_rc; if (rc) - goto out; + goto put; return id; - +put: + module_put(id->device->owner); out: rdma_destroy_id(id); return ERR_PTR(rc); @@ -579,17 +461,20 @@ rpcrdma_clean_cq(struct ib_cq *cq) int rpcrdma_ia_open(struct rpcrdma_xprt *xprt, struct sockaddr *addr, int memreg) { - int rc, mem_priv; struct rpcrdma_ia *ia = &xprt->rx_ia; struct ib_device_attr *devattr = &ia->ri_devattr; + int rc; + + ia->ri_dma_mr = NULL; ia->ri_id = rpcrdma_create_id(xprt, ia, addr); if (IS_ERR(ia->ri_id)) { rc = PTR_ERR(ia->ri_id); goto out1; } + ia->ri_device = ia->ri_id->device; - ia->ri_pd = ib_alloc_pd(ia->ri_id->device); + ia->ri_pd = ib_alloc_pd(ia->ri_device); if (IS_ERR(ia->ri_pd)) { rc = PTR_ERR(ia->ri_pd); dprintk("RPC: %s: ib_alloc_pd() failed %i\n", @@ -597,69 +482,39 @@ rpcrdma_ia_open(struct rpcrdma_xprt *xprt, struct sockaddr *addr, int memreg) goto out2; } - rc = ib_query_device(ia->ri_id->device, devattr); + rc = ib_query_device(ia->ri_device, devattr); if (rc) { dprintk("RPC: %s: ib_query_device failed %d\n", __func__, rc); goto out3; } - if (devattr->device_cap_flags & IB_DEVICE_LOCAL_DMA_LKEY) { - ia->ri_have_dma_lkey = 1; - ia->ri_dma_lkey = ia->ri_id->device->local_dma_lkey; - } - if (memreg == RPCRDMA_FRMR) { - /* Requires both frmr reg and local dma lkey */ - if (((devattr->device_cap_flags & - (IB_DEVICE_MEM_MGT_EXTENSIONS|IB_DEVICE_LOCAL_DMA_LKEY)) != - (IB_DEVICE_MEM_MGT_EXTENSIONS|IB_DEVICE_LOCAL_DMA_LKEY)) || - (devattr->max_fast_reg_page_list_len == 0)) { + if (!(devattr->device_cap_flags & IB_DEVICE_MEM_MGT_EXTENSIONS) || + (devattr->max_fast_reg_page_list_len == 0)) { dprintk("RPC: %s: FRMR registration " "not supported by HCA\n", __func__); memreg = RPCRDMA_MTHCAFMR; } } if (memreg == RPCRDMA_MTHCAFMR) { - if (!ia->ri_id->device->alloc_fmr) { + if (!ia->ri_device->alloc_fmr) { dprintk("RPC: %s: MTHCAFMR registration " "not supported by HCA\n", __func__); - memreg = RPCRDMA_ALLPHYSICAL; + rc = -EINVAL; + goto out3; } } - /* - * Optionally obtain an underlying physical identity mapping in - * order to do a memory window-based bind. This base registration - * is protected from remote access - that is enabled only by binding - * for the specific bytes targeted during each RPC operation, and - * revoked after the corresponding completion similar to a storage - * adapter. - */ switch (memreg) { case RPCRDMA_FRMR: ia->ri_ops = &rpcrdma_frwr_memreg_ops; break; case RPCRDMA_ALLPHYSICAL: ia->ri_ops = &rpcrdma_physical_memreg_ops; - mem_priv = IB_ACCESS_LOCAL_WRITE | - IB_ACCESS_REMOTE_WRITE | - IB_ACCESS_REMOTE_READ; - goto register_setup; + break; case RPCRDMA_MTHCAFMR: ia->ri_ops = &rpcrdma_fmr_memreg_ops; - if (ia->ri_have_dma_lkey) - break; - mem_priv = IB_ACCESS_LOCAL_WRITE; - register_setup: - ia->ri_bind_mem = ib_get_dma_mr(ia->ri_pd, mem_priv); - if (IS_ERR(ia->ri_bind_mem)) { - printk(KERN_ALERT "%s: ib_get_dma_mr for " - "phys register failed with %lX\n", - __func__, PTR_ERR(ia->ri_bind_mem)); - rc = -ENOMEM; - goto out3; - } break; default: printk(KERN_ERR "RPC: Unsupported memory " @@ -670,9 +525,6 @@ rpcrdma_ia_open(struct rpcrdma_xprt *xprt, struct sockaddr *addr, int memreg) dprintk("RPC: %s: memory registration strategy is '%s'\n", __func__, ia->ri_ops->ro_displayname); - /* Else will do memory reg/dereg for each chunk */ - ia->ri_memreg_strategy = memreg; - rwlock_init(&ia->ri_qplock); return 0; @@ -680,7 +532,7 @@ out3: ib_dealloc_pd(ia->ri_pd); ia->ri_pd = NULL; out2: - rdma_destroy_id(ia->ri_id); + rpcrdma_destroy_id(ia->ri_id); ia->ri_id = NULL; out1: return rc; @@ -694,25 +546,17 @@ out1: void rpcrdma_ia_close(struct rpcrdma_ia *ia) { - int rc; - dprintk("RPC: %s: entering\n", __func__); - if (ia->ri_bind_mem != NULL) { - rc = ib_dereg_mr(ia->ri_bind_mem); - dprintk("RPC: %s: ib_dereg_mr returned %i\n", - __func__, rc); - } if (ia->ri_id != NULL && !IS_ERR(ia->ri_id)) { if (ia->ri_id->qp) rdma_destroy_qp(ia->ri_id); - rdma_destroy_id(ia->ri_id); + rpcrdma_destroy_id(ia->ri_id); ia->ri_id = NULL; } - if (ia->ri_pd != NULL && !IS_ERR(ia->ri_pd)) { - rc = ib_dealloc_pd(ia->ri_pd); - dprintk("RPC: %s: ib_dealloc_pd returned %i\n", - __func__, rc); - } + + /* If the pd is still busy, xprtrdma missed freeing a resource */ + if (ia->ri_pd && !IS_ERR(ia->ri_pd)) + ib_dealloc_pd(ia->ri_pd); } /* @@ -724,35 +568,44 @@ rpcrdma_ep_create(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia, { struct ib_device_attr *devattr = &ia->ri_devattr; struct ib_cq *sendcq, *recvcq; + struct ib_cq_init_attr cq_attr = {}; + unsigned int max_qp_wr; int rc, err; + if (devattr->max_sge < RPCRDMA_MAX_IOVS) { + dprintk("RPC: %s: insufficient sge's available\n", + __func__); + return -ENOMEM; + } + + if (devattr->max_qp_wr <= RPCRDMA_BACKWARD_WRS) { + dprintk("RPC: %s: insufficient wqe's available\n", + __func__); + return -ENOMEM; + } + max_qp_wr = devattr->max_qp_wr - RPCRDMA_BACKWARD_WRS; + /* check provider's send/recv wr limits */ - if (cdata->max_requests > devattr->max_qp_wr) - cdata->max_requests = devattr->max_qp_wr; + if (cdata->max_requests > max_qp_wr) + cdata->max_requests = max_qp_wr; ep->rep_attr.event_handler = rpcrdma_qp_async_error_upcall; ep->rep_attr.qp_context = ep; ep->rep_attr.srq = NULL; ep->rep_attr.cap.max_send_wr = cdata->max_requests; + ep->rep_attr.cap.max_send_wr += RPCRDMA_BACKWARD_WRS; rc = ia->ri_ops->ro_open(ia, ep, cdata); if (rc) return rc; ep->rep_attr.cap.max_recv_wr = cdata->max_requests; - ep->rep_attr.cap.max_send_sge = (cdata->padding ? 4 : 2); + ep->rep_attr.cap.max_recv_wr += RPCRDMA_BACKWARD_WRS; + ep->rep_attr.cap.max_send_sge = RPCRDMA_MAX_IOVS; ep->rep_attr.cap.max_recv_sge = 1; ep->rep_attr.cap.max_inline_data = 0; ep->rep_attr.sq_sig_type = IB_SIGNAL_REQ_WR; ep->rep_attr.qp_type = IB_QPT_RC; ep->rep_attr.port_num = ~0; - if (cdata->padding) { - ep->rep_padbuf = rpcrdma_alloc_regbuf(ia, cdata->padding, - GFP_KERNEL); - if (IS_ERR(ep->rep_padbuf)) - return PTR_ERR(ep->rep_padbuf); - } else - ep->rep_padbuf = NULL; - dprintk("RPC: %s: requested max: dtos: send %d recv %d; " "iovs: send %d recv %d\n", __func__, @@ -771,9 +624,9 @@ rpcrdma_ep_create(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia, init_waitqueue_head(&ep->rep_connect_wait); INIT_DELAYED_WORK(&ep->rep_connect_worker, rpcrdma_connect_worker); - sendcq = ib_create_cq(ia->ri_id->device, rpcrdma_sendcq_upcall, - rpcrdma_cq_async_error_upcall, ep, - ep->rep_attr.cap.max_send_wr + 1, 0); + cq_attr.cqe = ep->rep_attr.cap.max_send_wr + 1; + sendcq = ib_create_cq(ia->ri_device, rpcrdma_sendcq_upcall, + rpcrdma_cq_async_error_upcall, NULL, &cq_attr); if (IS_ERR(sendcq)) { rc = PTR_ERR(sendcq); dprintk("RPC: %s: failed to create send CQ: %i\n", @@ -788,9 +641,9 @@ rpcrdma_ep_create(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia, goto out2; } - recvcq = ib_create_cq(ia->ri_id->device, rpcrdma_recvcq_upcall, - rpcrdma_cq_async_error_upcall, ep, - ep->rep_attr.cap.max_recv_wr + 1, 0); + cq_attr.cqe = ep->rep_attr.cap.max_recv_wr + 1; + recvcq = ib_create_cq(ia->ri_device, rpcrdma_recvcq_upcall, + rpcrdma_cq_async_error_upcall, NULL, &cq_attr); if (IS_ERR(recvcq)) { rc = PTR_ERR(recvcq); dprintk("RPC: %s: failed to create recv CQ: %i\n", @@ -835,7 +688,8 @@ out2: dprintk("RPC: %s: ib_destroy_cq returned %i\n", __func__, err); out1: - rpcrdma_free_regbuf(ia, ep->rep_padbuf); + if (ia->ri_dma_mr) + ib_dereg_mr(ia->ri_dma_mr); return rc; } @@ -856,25 +710,32 @@ rpcrdma_ep_destroy(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia) cancel_delayed_work_sync(&ep->rep_connect_worker); - if (ia->ri_id->qp) { + if (ia->ri_id->qp) rpcrdma_ep_disconnect(ep, ia); + + rpcrdma_clean_cq(ep->rep_attr.recv_cq); + rpcrdma_clean_cq(ep->rep_attr.send_cq); + + if (ia->ri_id->qp) { rdma_destroy_qp(ia->ri_id); ia->ri_id->qp = NULL; } - rpcrdma_free_regbuf(ia, ep->rep_padbuf); - - rpcrdma_clean_cq(ep->rep_attr.recv_cq); rc = ib_destroy_cq(ep->rep_attr.recv_cq); if (rc) dprintk("RPC: %s: ib_destroy_cq returned %i\n", __func__, rc); - rpcrdma_clean_cq(ep->rep_attr.send_cq); rc = ib_destroy_cq(ep->rep_attr.send_cq); if (rc) dprintk("RPC: %s: ib_destroy_cq returned %i\n", __func__, rc); + + if (ia->ri_dma_mr) { + rc = ib_dereg_mr(ia->ri_dma_mr); + dprintk("RPC: %s: ib_dereg_mr returned %i\n", + __func__, rc); + } } /* @@ -896,8 +757,6 @@ retry: rpcrdma_flush_cqs(ep); xprt = container_of(ia, struct rpcrdma_xprt, rx_ia); - ia->ri_ops->ro_reset(xprt); - id = rpcrdma_create_id(xprt, ia, (struct sockaddr *)&xprt->rx_data.addr); if (IS_ERR(id)) { @@ -911,10 +770,10 @@ retry: * More stuff I haven't thought of! * Rrrgh! */ - if (ia->ri_id->device != id->device) { + if (ia->ri_device != id->device) { printk("RPC: %s: can't reconnect on " "different device!\n", __func__); - rdma_destroy_id(id); + rpcrdma_destroy_id(id); rc = -ENETUNREACH; goto out; } @@ -923,7 +782,7 @@ retry: if (rc) { dprintk("RPC: %s: rdma_create_qp failed %i\n", __func__, rc); - rdma_destroy_id(id); + rpcrdma_destroy_id(id); rc = -ENETUNREACH; goto out; } @@ -934,7 +793,7 @@ retry: write_unlock(&ia->ri_qplock); rdma_destroy_qp(old); - rdma_destroy_id(old); + rpcrdma_destroy_id(old); } else { dprintk("RPC: %s: connecting...\n", __func__); rc = rdma_create_qp(ia->ri_id, ia->ri_pd, &ep->rep_attr); @@ -983,7 +842,21 @@ retry: } rc = ep->rep_connected; } else { + struct rpcrdma_xprt *r_xprt; + unsigned int extras; + dprintk("RPC: %s: connected\n", __func__); + + r_xprt = container_of(ia, struct rpcrdma_xprt, rx_ia); + extras = r_xprt->rx_buf.rb_bc_srv_max_requests; + + if (extras) { + rc = rpcrdma_ep_post_extra_recv(r_xprt, extras); + if (rc) + pr_warn("%s: rpcrdma_ep_post_extra_recv: %i\n", + __func__, rc); + rc = 0; + } } out: @@ -1020,20 +893,25 @@ rpcrdma_ep_disconnect(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia) } } -static struct rpcrdma_req * +struct rpcrdma_req * rpcrdma_create_req(struct rpcrdma_xprt *r_xprt) { + struct rpcrdma_buffer *buffer = &r_xprt->rx_buf; struct rpcrdma_req *req; req = kzalloc(sizeof(*req), GFP_KERNEL); if (req == NULL) return ERR_PTR(-ENOMEM); + INIT_LIST_HEAD(&req->rl_free); + spin_lock(&buffer->rb_reqslock); + list_add(&req->rl_all, &buffer->rb_allreqs); + spin_unlock(&buffer->rb_reqslock); req->rl_buffer = &r_xprt->rx_buf; return req; } -static struct rpcrdma_rep * +struct rpcrdma_rep * rpcrdma_create_rep(struct rpcrdma_xprt *r_xprt) { struct rpcrdma_create_data_internal *cdata = &r_xprt->rx_data; @@ -1053,7 +931,9 @@ rpcrdma_create_rep(struct rpcrdma_xprt *r_xprt) goto out_free; } - rep->rr_buffer = &r_xprt->rx_buf; + rep->rr_device = ia->ri_device; + rep->rr_rxprt = r_xprt; + INIT_WORK(&rep->rr_work, rpcrdma_receive_worker); return rep; out_free: @@ -1067,44 +947,21 @@ rpcrdma_buffer_create(struct rpcrdma_xprt *r_xprt) { struct rpcrdma_buffer *buf = &r_xprt->rx_buf; struct rpcrdma_ia *ia = &r_xprt->rx_ia; - struct rpcrdma_create_data_internal *cdata = &r_xprt->rx_data; - char *p; - size_t len; int i, rc; - buf->rb_max_requests = cdata->max_requests; + buf->rb_max_requests = r_xprt->rx_data.max_requests; + buf->rb_bc_srv_max_requests = 0; spin_lock_init(&buf->rb_lock); - /* Need to allocate: - * 1. arrays for send and recv pointers - * 2. arrays of struct rpcrdma_req to fill in pointers - * 3. array of struct rpcrdma_rep for replies - * Send/recv buffers in req/rep need to be registered - */ - len = buf->rb_max_requests * - (sizeof(struct rpcrdma_req *) + sizeof(struct rpcrdma_rep *)); - - p = kzalloc(len, GFP_KERNEL); - if (p == NULL) { - dprintk("RPC: %s: req_t/rep_t/pad kzalloc(%zd) failed\n", - __func__, len); - rc = -ENOMEM; - goto out; - } - buf->rb_pool = p; /* for freeing it later */ - - buf->rb_send_bufs = (struct rpcrdma_req **) p; - p = (char *) &buf->rb_send_bufs[buf->rb_max_requests]; - buf->rb_recv_bufs = (struct rpcrdma_rep **) p; - p = (char *) &buf->rb_recv_bufs[buf->rb_max_requests]; - rc = ia->ri_ops->ro_init(r_xprt); if (rc) goto out; + INIT_LIST_HEAD(&buf->rb_send_bufs); + INIT_LIST_HEAD(&buf->rb_allreqs); + spin_lock_init(&buf->rb_reqslock); for (i = 0; i < buf->rb_max_requests; i++) { struct rpcrdma_req *req; - struct rpcrdma_rep *rep; req = rpcrdma_create_req(r_xprt); if (IS_ERR(req)) { @@ -1113,7 +970,13 @@ rpcrdma_buffer_create(struct rpcrdma_xprt *r_xprt) rc = PTR_ERR(req); goto out; } - buf->rb_send_bufs[i] = req; + req->rl_backchannel = false; + list_add(&req->rl_free, &buf->rb_send_bufs); + } + + INIT_LIST_HEAD(&buf->rb_recv_bufs); + for (i = 0; i < buf->rb_max_requests + 2; i++) { + struct rpcrdma_rep *rep; rep = rpcrdma_create_rep(r_xprt); if (IS_ERR(rep)) { @@ -1122,7 +985,7 @@ rpcrdma_buffer_create(struct rpcrdma_xprt *r_xprt) rc = PTR_ERR(rep); goto out; } - buf->rb_recv_bufs[i] = rep; + list_add(&rep->rr_list, &buf->rb_recv_bufs); } return 0; @@ -1131,22 +994,38 @@ out: return rc; } +static struct rpcrdma_req * +rpcrdma_buffer_get_req_locked(struct rpcrdma_buffer *buf) +{ + struct rpcrdma_req *req; + + req = list_first_entry(&buf->rb_send_bufs, + struct rpcrdma_req, rl_free); + list_del(&req->rl_free); + return req; +} + +static struct rpcrdma_rep * +rpcrdma_buffer_get_rep_locked(struct rpcrdma_buffer *buf) +{ + struct rpcrdma_rep *rep; + + rep = list_first_entry(&buf->rb_recv_bufs, + struct rpcrdma_rep, rr_list); + list_del(&rep->rr_list); + return rep; +} + static void rpcrdma_destroy_rep(struct rpcrdma_ia *ia, struct rpcrdma_rep *rep) { - if (!rep) - return; - rpcrdma_free_regbuf(ia, rep->rr_rdmabuf); kfree(rep); } -static void +void rpcrdma_destroy_req(struct rpcrdma_ia *ia, struct rpcrdma_req *req) { - if (!req) - return; - rpcrdma_free_regbuf(ia, req->rl_sendbuf); rpcrdma_free_regbuf(ia, req->rl_rdmabuf); kfree(req); @@ -1156,220 +1035,88 @@ void rpcrdma_buffer_destroy(struct rpcrdma_buffer *buf) { struct rpcrdma_ia *ia = rdmab_to_ia(buf); - int i; - - /* clean up in reverse order from create - * 1. recv mr memory (mr free, then kfree) - * 2. send mr memory (mr free, then kfree) - * 3. MWs - */ - dprintk("RPC: %s: entering\n", __func__); - - for (i = 0; i < buf->rb_max_requests; i++) { - if (buf->rb_recv_bufs) - rpcrdma_destroy_rep(ia, buf->rb_recv_bufs[i]); - if (buf->rb_send_bufs) - rpcrdma_destroy_req(ia, buf->rb_send_bufs[i]); - } - - ia->ri_ops->ro_destroy(buf); - kfree(buf->rb_pool); -} + while (!list_empty(&buf->rb_recv_bufs)) { + struct rpcrdma_rep *rep; -/* "*mw" can be NULL when rpcrdma_buffer_get_mrs() fails, leaving - * some req segments uninitialized. - */ -static void -rpcrdma_buffer_put_mr(struct rpcrdma_mw **mw, struct rpcrdma_buffer *buf) -{ - if (*mw) { - list_add_tail(&(*mw)->mw_list, &buf->rb_mws); - *mw = NULL; + rep = rpcrdma_buffer_get_rep_locked(buf); + rpcrdma_destroy_rep(ia, rep); } -} -/* Cycle mw's back in reverse order, and "spin" them. - * This delays and scrambles reuse as much as possible. - */ -static void -rpcrdma_buffer_put_mrs(struct rpcrdma_req *req, struct rpcrdma_buffer *buf) -{ - struct rpcrdma_mr_seg *seg = req->rl_segments; - struct rpcrdma_mr_seg *seg1 = seg; - int i; + spin_lock(&buf->rb_reqslock); + while (!list_empty(&buf->rb_allreqs)) { + struct rpcrdma_req *req; - for (i = 1, seg++; i < RPCRDMA_MAX_SEGS; seg++, i++) - rpcrdma_buffer_put_mr(&seg->rl_mw, buf); - rpcrdma_buffer_put_mr(&seg1->rl_mw, buf); -} + req = list_first_entry(&buf->rb_allreqs, + struct rpcrdma_req, rl_all); + list_del(&req->rl_all); -static void -rpcrdma_buffer_put_sendbuf(struct rpcrdma_req *req, struct rpcrdma_buffer *buf) -{ - buf->rb_send_bufs[--buf->rb_send_index] = req; - req->rl_niovs = 0; - if (req->rl_reply) { - buf->rb_recv_bufs[--buf->rb_recv_index] = req->rl_reply; - req->rl_reply->rr_func = NULL; - req->rl_reply = NULL; + spin_unlock(&buf->rb_reqslock); + rpcrdma_destroy_req(ia, req); + spin_lock(&buf->rb_reqslock); } -} - -/* rpcrdma_unmap_one() was already done during deregistration. - * Redo only the ib_post_send(). - */ -static void -rpcrdma_retry_local_inv(struct rpcrdma_mw *r, struct rpcrdma_ia *ia) -{ - struct rpcrdma_xprt *r_xprt = - container_of(ia, struct rpcrdma_xprt, rx_ia); - struct ib_send_wr invalidate_wr, *bad_wr; - int rc; - - dprintk("RPC: %s: FRMR %p is stale\n", __func__, r); - - /* When this FRMR is re-inserted into rb_mws, it is no longer stale */ - r->r.frmr.fr_state = FRMR_IS_INVALID; - - memset(&invalidate_wr, 0, sizeof(invalidate_wr)); - invalidate_wr.wr_id = (unsigned long)(void *)r; - invalidate_wr.opcode = IB_WR_LOCAL_INV; - invalidate_wr.ex.invalidate_rkey = r->r.frmr.fr_mr->rkey; - DECR_CQCOUNT(&r_xprt->rx_ep); - - dprintk("RPC: %s: frmr %p invalidating rkey %08x\n", - __func__, r, r->r.frmr.fr_mr->rkey); + spin_unlock(&buf->rb_reqslock); - read_lock(&ia->ri_qplock); - rc = ib_post_send(ia->ri_id->qp, &invalidate_wr, &bad_wr); - read_unlock(&ia->ri_qplock); - if (rc) { - /* Force rpcrdma_buffer_get() to retry */ - r->r.frmr.fr_state = FRMR_IS_STALE; - dprintk("RPC: %s: ib_post_send failed, %i\n", - __func__, rc); - } + ia->ri_ops->ro_destroy(buf); } -static void -rpcrdma_retry_flushed_linv(struct list_head *stale, - struct rpcrdma_buffer *buf) +struct rpcrdma_mw * +rpcrdma_get_mw(struct rpcrdma_xprt *r_xprt) { - struct rpcrdma_ia *ia = rdmab_to_ia(buf); - struct list_head *pos; - struct rpcrdma_mw *r; - unsigned long flags; - - list_for_each(pos, stale) { - r = list_entry(pos, struct rpcrdma_mw, mw_list); - rpcrdma_retry_local_inv(r, ia); - } - - spin_lock_irqsave(&buf->rb_lock, flags); - list_splice_tail(stale, &buf->rb_mws); - spin_unlock_irqrestore(&buf->rb_lock, flags); -} + struct rpcrdma_buffer *buf = &r_xprt->rx_buf; + struct rpcrdma_mw *mw = NULL; -static struct rpcrdma_req * -rpcrdma_buffer_get_frmrs(struct rpcrdma_req *req, struct rpcrdma_buffer *buf, - struct list_head *stale) -{ - struct rpcrdma_mw *r; - int i; - - i = RPCRDMA_MAX_SEGS - 1; - while (!list_empty(&buf->rb_mws)) { - r = list_entry(buf->rb_mws.next, - struct rpcrdma_mw, mw_list); - list_del(&r->mw_list); - if (r->r.frmr.fr_state == FRMR_IS_STALE) { - list_add(&r->mw_list, stale); - continue; - } - req->rl_segments[i].rl_mw = r; - if (unlikely(i-- == 0)) - return req; /* Success */ + spin_lock(&buf->rb_mwlock); + if (!list_empty(&buf->rb_mws)) { + mw = list_first_entry(&buf->rb_mws, + struct rpcrdma_mw, mw_list); + list_del_init(&mw->mw_list); } + spin_unlock(&buf->rb_mwlock); - /* Not enough entries on rb_mws for this req */ - rpcrdma_buffer_put_sendbuf(req, buf); - rpcrdma_buffer_put_mrs(req, buf); - return NULL; + if (!mw) + pr_err("RPC: %s: no MWs available\n", __func__); + return mw; } -static struct rpcrdma_req * -rpcrdma_buffer_get_fmrs(struct rpcrdma_req *req, struct rpcrdma_buffer *buf) +void +rpcrdma_put_mw(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mw *mw) { - struct rpcrdma_mw *r; - int i; - - i = RPCRDMA_MAX_SEGS - 1; - while (!list_empty(&buf->rb_mws)) { - r = list_entry(buf->rb_mws.next, - struct rpcrdma_mw, mw_list); - list_del(&r->mw_list); - req->rl_segments[i].rl_mw = r; - if (unlikely(i-- == 0)) - return req; /* Success */ - } + struct rpcrdma_buffer *buf = &r_xprt->rx_buf; - /* Not enough entries on rb_mws for this req */ - rpcrdma_buffer_put_sendbuf(req, buf); - rpcrdma_buffer_put_mrs(req, buf); - return NULL; + spin_lock(&buf->rb_mwlock); + list_add_tail(&mw->mw_list, &buf->rb_mws); + spin_unlock(&buf->rb_mwlock); } /* * Get a set of request/reply buffers. * - * Reply buffer (if needed) is attached to send buffer upon return. - * Rule: - * rb_send_index and rb_recv_index MUST always be pointing to the - * *next* available buffer (non-NULL). They are incremented after - * removing buffers, and decremented *before* returning them. + * Reply buffer (if available) is attached to send buffer upon return. */ struct rpcrdma_req * rpcrdma_buffer_get(struct rpcrdma_buffer *buffers) { - struct rpcrdma_ia *ia = rdmab_to_ia(buffers); - struct list_head stale; struct rpcrdma_req *req; - unsigned long flags; - - spin_lock_irqsave(&buffers->rb_lock, flags); - if (buffers->rb_send_index == buffers->rb_max_requests) { - spin_unlock_irqrestore(&buffers->rb_lock, flags); - dprintk("RPC: %s: out of request buffers\n", __func__); - return ((struct rpcrdma_req *)NULL); - } - req = buffers->rb_send_bufs[buffers->rb_send_index]; - if (buffers->rb_send_index < buffers->rb_recv_index) { - dprintk("RPC: %s: %d extra receives outstanding (ok)\n", - __func__, - buffers->rb_recv_index - buffers->rb_send_index); - req->rl_reply = NULL; - } else { - req->rl_reply = buffers->rb_recv_bufs[buffers->rb_recv_index]; - buffers->rb_recv_bufs[buffers->rb_recv_index++] = NULL; - } - buffers->rb_send_bufs[buffers->rb_send_index++] = NULL; + spin_lock(&buffers->rb_lock); + if (list_empty(&buffers->rb_send_bufs)) + goto out_reqbuf; + req = rpcrdma_buffer_get_req_locked(buffers); + if (list_empty(&buffers->rb_recv_bufs)) + goto out_repbuf; + req->rl_reply = rpcrdma_buffer_get_rep_locked(buffers); + spin_unlock(&buffers->rb_lock); + return req; - INIT_LIST_HEAD(&stale); - switch (ia->ri_memreg_strategy) { - case RPCRDMA_FRMR: - req = rpcrdma_buffer_get_frmrs(req, buffers, &stale); - break; - case RPCRDMA_MTHCAFMR: - req = rpcrdma_buffer_get_fmrs(req, buffers); - break; - default: - break; - } - spin_unlock_irqrestore(&buffers->rb_lock, flags); - if (!list_empty(&stale)) - rpcrdma_retry_flushed_linv(&stale, buffers); +out_reqbuf: + spin_unlock(&buffers->rb_lock); + pr_warn("RPC: %s: out of request buffers\n", __func__); + return NULL; +out_repbuf: + spin_unlock(&buffers->rb_lock); + pr_warn("RPC: %s: out of reply buffers\n", __func__); + req->rl_reply = NULL; return req; } @@ -1381,39 +1128,31 @@ void rpcrdma_buffer_put(struct rpcrdma_req *req) { struct rpcrdma_buffer *buffers = req->rl_buffer; - struct rpcrdma_ia *ia = rdmab_to_ia(buffers); - unsigned long flags; + struct rpcrdma_rep *rep = req->rl_reply; - spin_lock_irqsave(&buffers->rb_lock, flags); - rpcrdma_buffer_put_sendbuf(req, buffers); - switch (ia->ri_memreg_strategy) { - case RPCRDMA_FRMR: - case RPCRDMA_MTHCAFMR: - rpcrdma_buffer_put_mrs(req, buffers); - break; - default: - break; - } - spin_unlock_irqrestore(&buffers->rb_lock, flags); + req->rl_niovs = 0; + req->rl_reply = NULL; + + spin_lock(&buffers->rb_lock); + list_add_tail(&req->rl_free, &buffers->rb_send_bufs); + if (rep) + list_add_tail(&rep->rr_list, &buffers->rb_recv_bufs); + spin_unlock(&buffers->rb_lock); } /* * Recover reply buffers from pool. - * This happens when recovering from error conditions. - * Post-increment counter/array index. + * This happens when recovering from disconnect. */ void rpcrdma_recv_buffer_get(struct rpcrdma_req *req) { struct rpcrdma_buffer *buffers = req->rl_buffer; - unsigned long flags; - spin_lock_irqsave(&buffers->rb_lock, flags); - if (buffers->rb_recv_index < buffers->rb_max_requests) { - req->rl_reply = buffers->rb_recv_bufs[buffers->rb_recv_index]; - buffers->rb_recv_bufs[buffers->rb_recv_index++] = NULL; - } - spin_unlock_irqrestore(&buffers->rb_lock, flags); + spin_lock(&buffers->rb_lock); + if (!list_empty(&buffers->rb_recv_bufs)) + req->rl_reply = rpcrdma_buffer_get_rep_locked(buffers); + spin_unlock(&buffers->rb_lock); } /* @@ -1423,13 +1162,11 @@ rpcrdma_recv_buffer_get(struct rpcrdma_req *req) void rpcrdma_recv_buffer_put(struct rpcrdma_rep *rep) { - struct rpcrdma_buffer *buffers = rep->rr_buffer; - unsigned long flags; + struct rpcrdma_buffer *buffers = &rep->rr_rxprt->rx_buf; - rep->rr_func = NULL; - spin_lock_irqsave(&buffers->rb_lock, flags); - buffers->rb_recv_bufs[--buffers->rb_recv_index] = rep; - spin_unlock_irqrestore(&buffers->rb_lock, flags); + spin_lock(&buffers->rb_lock); + list_add_tail(&rep->rr_list, &buffers->rb_recv_bufs); + spin_unlock(&buffers->rb_lock); } /* @@ -1444,75 +1181,6 @@ rpcrdma_mapping_error(struct rpcrdma_mr_seg *seg) (unsigned long long)seg->mr_dma, seg->mr_dmalen); } -static int -rpcrdma_register_internal(struct rpcrdma_ia *ia, void *va, int len, - struct ib_mr **mrp, struct ib_sge *iov) -{ - struct ib_phys_buf ipb; - struct ib_mr *mr; - int rc; - - /* - * All memory passed here was kmalloc'ed, therefore phys-contiguous. - */ - iov->addr = ib_dma_map_single(ia->ri_id->device, - va, len, DMA_BIDIRECTIONAL); - if (ib_dma_mapping_error(ia->ri_id->device, iov->addr)) - return -ENOMEM; - - iov->length = len; - - if (ia->ri_have_dma_lkey) { - *mrp = NULL; - iov->lkey = ia->ri_dma_lkey; - return 0; - } else if (ia->ri_bind_mem != NULL) { - *mrp = NULL; - iov->lkey = ia->ri_bind_mem->lkey; - return 0; - } - - ipb.addr = iov->addr; - ipb.size = iov->length; - mr = ib_reg_phys_mr(ia->ri_pd, &ipb, 1, - IB_ACCESS_LOCAL_WRITE, &iov->addr); - - dprintk("RPC: %s: phys convert: 0x%llx " - "registered 0x%llx length %d\n", - __func__, (unsigned long long)ipb.addr, - (unsigned long long)iov->addr, len); - - if (IS_ERR(mr)) { - *mrp = NULL; - rc = PTR_ERR(mr); - dprintk("RPC: %s: failed with %i\n", __func__, rc); - } else { - *mrp = mr; - iov->lkey = mr->lkey; - rc = 0; - } - - return rc; -} - -static int -rpcrdma_deregister_internal(struct rpcrdma_ia *ia, - struct ib_mr *mr, struct ib_sge *iov) -{ - int rc; - - ib_dma_unmap_single(ia->ri_id->device, - iov->addr, iov->length, DMA_BIDIRECTIONAL); - - if (NULL == mr) - return 0; - - rc = ib_dereg_mr(mr); - if (rc) - dprintk("RPC: %s: ib_dereg_mr failed %i\n", __func__, rc); - return rc; -} - /** * rpcrdma_alloc_regbuf - kmalloc and register memory for SEND/RECV buffers * @ia: controlling rpcrdma_ia @@ -1532,26 +1200,29 @@ struct rpcrdma_regbuf * rpcrdma_alloc_regbuf(struct rpcrdma_ia *ia, size_t size, gfp_t flags) { struct rpcrdma_regbuf *rb; - int rc; + struct ib_sge *iov; - rc = -ENOMEM; rb = kmalloc(sizeof(*rb) + size, flags); if (rb == NULL) goto out; - rb->rg_size = size; - rb->rg_owner = NULL; - rc = rpcrdma_register_internal(ia, rb->rg_base, size, - &rb->rg_mr, &rb->rg_iov); - if (rc) + iov = &rb->rg_iov; + iov->addr = ib_dma_map_single(ia->ri_device, + (void *)rb->rg_base, size, + DMA_BIDIRECTIONAL); + if (ib_dma_mapping_error(ia->ri_device, iov->addr)) goto out_free; + iov->length = size; + iov->lkey = ia->ri_pd->local_dma_lkey; + rb->rg_size = size; + rb->rg_owner = NULL; return rb; out_free: kfree(rb); out: - return ERR_PTR(rc); + return ERR_PTR(-ENOMEM); } /** @@ -1562,10 +1233,15 @@ out: void rpcrdma_free_regbuf(struct rpcrdma_ia *ia, struct rpcrdma_regbuf *rb) { - if (rb) { - rpcrdma_deregister_internal(ia, rb->rg_mr, &rb->rg_iov); - kfree(rb); - } + struct ib_sge *iov; + + if (!rb) + return; + + iov = &rb->rg_iov; + ib_dma_unmap_single(ia->ri_device, + iov->addr, iov->length, DMA_BIDIRECTIONAL); + kfree(rb); } /* @@ -1578,9 +1254,11 @@ rpcrdma_ep_post(struct rpcrdma_ia *ia, struct rpcrdma_ep *ep, struct rpcrdma_req *req) { + struct ib_device *device = ia->ri_device; struct ib_send_wr send_wr, *send_wr_fail; struct rpcrdma_rep *rep = req->rl_reply; - int rc; + struct ib_sge *iov = req->rl_send_iov; + int i, rc; if (rep) { rc = rpcrdma_ep_post_recv(ia, ep, rep); @@ -1591,19 +1269,15 @@ rpcrdma_ep_post(struct rpcrdma_ia *ia, send_wr.next = NULL; send_wr.wr_id = RPCRDMA_IGNORE_COMPLETION; - send_wr.sg_list = req->rl_send_iov; + send_wr.sg_list = iov; send_wr.num_sge = req->rl_niovs; send_wr.opcode = IB_WR_SEND; - if (send_wr.num_sge == 4) /* no need to sync any pad (constant) */ - ib_dma_sync_single_for_device(ia->ri_id->device, - req->rl_send_iov[3].addr, req->rl_send_iov[3].length, - DMA_TO_DEVICE); - ib_dma_sync_single_for_device(ia->ri_id->device, - req->rl_send_iov[1].addr, req->rl_send_iov[1].length, - DMA_TO_DEVICE); - ib_dma_sync_single_for_device(ia->ri_id->device, - req->rl_send_iov[0].addr, req->rl_send_iov[0].length, - DMA_TO_DEVICE); + + for (i = 0; i < send_wr.num_sge; i++) + ib_dma_sync_single_for_device(device, iov[i].addr, + iov[i].length, DMA_TO_DEVICE); + dprintk("RPC: %s: posting %d s/g entries\n", + __func__, send_wr.num_sge); if (DECR_CQCOUNT(ep) > 0) send_wr.send_flags = 0; @@ -1636,7 +1310,7 @@ rpcrdma_ep_post_recv(struct rpcrdma_ia *ia, recv_wr.sg_list = &rep->rr_rdmabuf->rg_iov; recv_wr.num_sge = 1; - ib_dma_sync_single_for_cpu(ia->ri_id->device, + ib_dma_sync_single_for_cpu(ia->ri_device, rdmab_addr(rep->rr_rdmabuf), rdmab_length(rep->rr_rdmabuf), DMA_BIDIRECTIONAL); @@ -1649,6 +1323,47 @@ rpcrdma_ep_post_recv(struct rpcrdma_ia *ia, return rc; } +/** + * rpcrdma_ep_post_extra_recv - Post buffers for incoming backchannel requests + * @r_xprt: transport associated with these backchannel resources + * @min_reqs: minimum number of incoming requests expected + * + * Returns zero if all requested buffers were posted, or a negative errno. + */ +int +rpcrdma_ep_post_extra_recv(struct rpcrdma_xprt *r_xprt, unsigned int count) +{ + struct rpcrdma_buffer *buffers = &r_xprt->rx_buf; + struct rpcrdma_ia *ia = &r_xprt->rx_ia; + struct rpcrdma_ep *ep = &r_xprt->rx_ep; + struct rpcrdma_rep *rep; + unsigned long flags; + int rc; + + while (count--) { + spin_lock_irqsave(&buffers->rb_lock, flags); + if (list_empty(&buffers->rb_recv_bufs)) + goto out_reqbuf; + rep = rpcrdma_buffer_get_rep_locked(buffers); + spin_unlock_irqrestore(&buffers->rb_lock, flags); + + rc = rpcrdma_ep_post_recv(ia, ep, rep); + if (rc) + goto out_rc; + } + + return 0; + +out_reqbuf: + spin_unlock_irqrestore(&buffers->rb_lock, flags); + pr_warn("%s: no extra receive buffers\n", __func__); + return -ENOMEM; + +out_rc: + rpcrdma_recv_buffer_put(rep); + return rc; +} + /* How many chunk list items fit within our inline buffers? */ unsigned int diff --git a/kernel/net/sunrpc/xprtrdma/xprt_rdma.h b/kernel/net/sunrpc/xprtrdma/xprt_rdma.h index 78e0b8bea..ac7f8d4f6 100644 --- a/kernel/net/sunrpc/xprtrdma/xprt_rdma.h +++ b/kernel/net/sunrpc/xprtrdma/xprt_rdma.h @@ -51,7 +51,6 @@ #include <linux/sunrpc/clnt.h> /* rpc_xprt */ #include <linux/sunrpc/rpc_rdma.h> /* RPC/RDMA protocol */ #include <linux/sunrpc/xprtrdma.h> /* xprt parameters */ -#include <linux/sunrpc/svc.h> /* RPCSVC_MAXPAYLOAD */ #define RDMA_RESOLVE_TIMEOUT (5000) /* 5 seconds */ #define RDMA_CONNECT_RETRY_MAX (2) /* retries if no listener backlog */ @@ -62,14 +61,12 @@ struct rpcrdma_ia { const struct rpcrdma_memreg_ops *ri_ops; rwlock_t ri_qplock; + struct ib_device *ri_device; struct rdma_cm_id *ri_id; struct ib_pd *ri_pd; - struct ib_mr *ri_bind_mem; - u32 ri_dma_lkey; - int ri_have_dma_lkey; + struct ib_mr *ri_dma_mr; struct completion ri_done; int ri_async_rc; - enum rpcrdma_memreg ri_memreg_strategy; unsigned int ri_max_frmr_depth; struct ib_device_attr ri_devattr; struct ib_qp_attr ri_qp_attr; @@ -80,21 +77,15 @@ struct rpcrdma_ia { * RDMA Endpoint -- one per transport instance */ -#define RPCRDMA_WC_BUDGET (128) -#define RPCRDMA_POLLSIZE (16) - struct rpcrdma_ep { atomic_t rep_cqcount; int rep_cqinit; int rep_connected; struct ib_qp_init_attr rep_attr; wait_queue_head_t rep_connect_wait; - struct rpcrdma_regbuf *rep_padbuf; struct rdma_conn_param rep_remote_cma; struct sockaddr_storage rep_remote_addr; struct delayed_work rep_connect_worker; - struct ib_wc rep_send_wcs[RPCRDMA_POLLSIZE]; - struct ib_wc rep_recv_wcs[RPCRDMA_POLLSIZE]; }; /* @@ -110,6 +101,16 @@ struct rpcrdma_ep { */ #define RPCRDMA_IGNORE_COMPLETION (0ULL) +/* Pre-allocate extra Work Requests for handling backward receives + * and sends. This is a fixed value because the Work Queues are + * allocated when the forward channel is set up. + */ +#if defined(CONFIG_SUNRPC_BACKCHANNEL) +#define RPCRDMA_BACKWARD_WRS (8) +#else +#define RPCRDMA_BACKWARD_WRS (0) +#endif + /* Registered buffer -- registered kmalloc'd memory for RDMA SEND/RECV * * The below structure appears at the front of a large region of kmalloc'd @@ -119,7 +120,6 @@ struct rpcrdma_ep { struct rpcrdma_regbuf { size_t rg_size; struct rpcrdma_req *rg_owner; - struct ib_mr *rg_mr; struct ib_sge rg_iov; __be32 rg_base[0] __attribute__ ((aligned(256))); }; @@ -165,21 +165,22 @@ rdmab_to_msg(struct rpcrdma_regbuf *rb) * struct rpcrdma_buffer. N is the max number of outstanding requests. */ -/* temporary static scatter/gather max */ -#define RPCRDMA_MAX_DATA_SEGS (64) /* max scatter/gather */ +#define RPCRDMA_MAX_DATA_SEGS ((1 * 1024 * 1024) / PAGE_SIZE) #define RPCRDMA_MAX_SEGS (RPCRDMA_MAX_DATA_SEGS + 2) /* head+tail = 2 */ struct rpcrdma_buffer; struct rpcrdma_rep { unsigned int rr_len; - struct rpcrdma_buffer *rr_buffer; - struct rpc_xprt *rr_xprt; - void (*rr_func)(struct rpcrdma_rep *); + struct ib_device *rr_device; + struct rpcrdma_xprt *rr_rxprt; + struct work_struct rr_work; struct list_head rr_list; struct rpcrdma_regbuf *rr_rdmabuf; }; +#define RPCRDMA_BAD_LEN (~0U) + /* * struct rpcrdma_mw - external memory region metadata * @@ -200,14 +201,22 @@ enum rpcrdma_frmr_state { }; struct rpcrdma_frmr { - struct ib_fast_reg_page_list *fr_pgl; + struct scatterlist *sg; + int sg_nents; struct ib_mr *fr_mr; enum rpcrdma_frmr_state fr_state; + struct work_struct fr_work; + struct rpcrdma_xprt *fr_xprt; +}; + +struct rpcrdma_fmr { + struct ib_fmr *fmr; + u64 *physaddrs; }; struct rpcrdma_mw { union { - struct ib_fmr *fmr; + struct rpcrdma_fmr fmr; struct rpcrdma_frmr frmr; } r; void (*mw_sendcompletion)(struct ib_wc *); @@ -252,16 +261,22 @@ struct rpcrdma_mr_seg { /* chunk descriptors */ char *mr_offset; /* kva if no page, else offset */ }; +#define RPCRDMA_MAX_IOVS (2) + struct rpcrdma_req { - unsigned int rl_niovs; /* 0, 2 or 4 */ - unsigned int rl_nchunks; /* non-zero if chunks */ - unsigned int rl_connect_cookie; /* retry detection */ - struct rpcrdma_buffer *rl_buffer; /* home base for this structure */ + struct list_head rl_free; + unsigned int rl_niovs; + unsigned int rl_nchunks; + unsigned int rl_connect_cookie; + struct rpcrdma_buffer *rl_buffer; struct rpcrdma_rep *rl_reply;/* holder for reply buffer */ - struct ib_sge rl_send_iov[4]; /* for active requests */ - struct rpcrdma_regbuf *rl_rdmabuf; - struct rpcrdma_regbuf *rl_sendbuf; - struct rpcrdma_mr_seg rl_segments[RPCRDMA_MAX_SEGS]; + struct ib_sge rl_send_iov[RPCRDMA_MAX_IOVS]; + struct rpcrdma_regbuf *rl_rdmabuf; + struct rpcrdma_regbuf *rl_sendbuf; + struct rpcrdma_mr_seg rl_segments[RPCRDMA_MAX_SEGS]; + + struct list_head rl_all; + bool rl_backchannel; }; static inline struct rpcrdma_req * @@ -281,15 +296,19 @@ rpcr_to_rdmar(struct rpc_rqst *rqst) * One of these is associated with a transport instance */ struct rpcrdma_buffer { - spinlock_t rb_lock; /* protects indexes */ - u32 rb_max_requests;/* client max requests */ - struct list_head rb_mws; /* optional memory windows/fmrs/frmrs */ - struct list_head rb_all; - int rb_send_index; - struct rpcrdma_req **rb_send_bufs; - int rb_recv_index; - struct rpcrdma_rep **rb_recv_bufs; - char *rb_pool; + spinlock_t rb_mwlock; /* protect rb_mws list */ + struct list_head rb_mws; + struct list_head rb_all; + char *rb_pool; + + spinlock_t rb_lock; /* protect buf lists */ + struct list_head rb_send_bufs; + struct list_head rb_recv_bufs; + u32 rb_max_requests; + + u32 rb_bc_srv_max_requests; + spinlock_t rb_reqslock; /* protect rb_allreqs */ + struct list_head rb_allreqs; }; #define rdmab_to_ia(b) (&container_of((b), struct rpcrdma_xprt, rx_buf)->rx_ia) @@ -334,6 +353,8 @@ struct rpcrdma_stats { unsigned long hardway_register_count; unsigned long failed_marshal_count; unsigned long bad_reply_count; + unsigned long nomsg_call_count; + unsigned long bcall_count; }; /* @@ -350,7 +371,6 @@ struct rpcrdma_memreg_ops { struct rpcrdma_create_data_internal *); size_t (*ro_maxpages)(struct rpcrdma_xprt *); int (*ro_init)(struct rpcrdma_xprt *); - void (*ro_reset)(struct rpcrdma_xprt *); void (*ro_destroy)(struct rpcrdma_buffer *); const char *ro_displayname; }; @@ -410,9 +430,14 @@ int rpcrdma_ep_post_recv(struct rpcrdma_ia *, struct rpcrdma_ep *, /* * Buffer calls - xprtrdma/verbs.c */ +struct rpcrdma_req *rpcrdma_create_req(struct rpcrdma_xprt *); +struct rpcrdma_rep *rpcrdma_create_rep(struct rpcrdma_xprt *); +void rpcrdma_destroy_req(struct rpcrdma_ia *, struct rpcrdma_req *); int rpcrdma_buffer_create(struct rpcrdma_xprt *); void rpcrdma_buffer_destroy(struct rpcrdma_buffer *); +struct rpcrdma_mw *rpcrdma_get_mw(struct rpcrdma_xprt *); +void rpcrdma_put_mw(struct rpcrdma_xprt *, struct rpcrdma_mw *); struct rpcrdma_req *rpcrdma_buffer_get(struct rpcrdma_buffer *); void rpcrdma_buffer_put(struct rpcrdma_req *); void rpcrdma_recv_buffer_get(struct rpcrdma_req *); @@ -424,6 +449,13 @@ void rpcrdma_free_regbuf(struct rpcrdma_ia *, struct rpcrdma_regbuf *); unsigned int rpcrdma_max_segments(struct rpcrdma_xprt *); +int rpcrdma_ep_post_extra_recv(struct rpcrdma_xprt *, unsigned int); + +int frwr_alloc_recovery_wq(void); +void frwr_destroy_recovery_wq(void); + +int rpcrdma_alloc_wq(void); +void rpcrdma_destroy_wq(void); /* * Wrappers for chunk registration, shared by read/write chunk code. @@ -480,6 +512,23 @@ void rpcrdma_reply_handler(struct rpcrdma_rep *); */ int rpcrdma_marshal_req(struct rpc_rqst *); +/* RPC/RDMA module init - xprtrdma/transport.c + */ +int xprt_rdma_init(void); +void xprt_rdma_cleanup(void); + +/* Backchannel calls - xprtrdma/backchannel.c + */ +#if defined(CONFIG_SUNRPC_BACKCHANNEL) +int xprt_rdma_bc_setup(struct rpc_xprt *, unsigned int); +int xprt_rdma_bc_up(struct svc_serv *, struct net *); +int rpcrdma_bc_post_recv(struct rpcrdma_xprt *, unsigned int); +void rpcrdma_bc_receive_call(struct rpcrdma_xprt *, struct rpcrdma_rep *); +int rpcrdma_bc_marshal_reply(struct rpc_rqst *); +void xprt_rdma_bc_free_rqst(struct rpc_rqst *); +void xprt_rdma_bc_destroy(struct rpc_xprt *, unsigned int); +#endif /* CONFIG_SUNRPC_BACKCHANNEL */ + /* Temporary NFS request map cache. Created in svc_rdma.c */ extern struct kmem_cache *svc_rdma_map_cachep; /* WR context cache. Created in svc_rdma.c */ @@ -487,10 +536,4 @@ extern struct kmem_cache *svc_rdma_ctxt_cachep; /* Workqueue created in svc_rdma.c */ extern struct workqueue_struct *svc_rdma_wq; -#if RPCSVC_MAXPAYLOAD < (RPCRDMA_MAX_DATA_SEGS << PAGE_SHIFT) -#define RPCSVC_MAXPAYLOAD_RDMA RPCSVC_MAXPAYLOAD -#else -#define RPCSVC_MAXPAYLOAD_RDMA (RPCRDMA_MAX_DATA_SEGS << PAGE_SHIFT) -#endif - #endif /* _LINUX_SUNRPC_XPRT_RDMA_H */ |