summaryrefslogtreecommitdiffstats
path: root/qemu/roms/ipxe/src/net/infiniband.c
diff options
context:
space:
mode:
Diffstat (limited to 'qemu/roms/ipxe/src/net/infiniband.c')
-rw-r--r--qemu/roms/ipxe/src/net/infiniband.c999
1 files changed, 999 insertions, 0 deletions
diff --git a/qemu/roms/ipxe/src/net/infiniband.c b/qemu/roms/ipxe/src/net/infiniband.c
new file mode 100644
index 000000000..12d1d83ce
--- /dev/null
+++ b/qemu/roms/ipxe/src/net/infiniband.c
@@ -0,0 +1,999 @@
+/*
+ * Copyright (C) 2007 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER );
+
+#include <stdint.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <unistd.h>
+#include <byteswap.h>
+#include <errno.h>
+#include <assert.h>
+#include <ipxe/list.h>
+#include <ipxe/errortab.h>
+#include <ipxe/if_arp.h>
+#include <ipxe/netdevice.h>
+#include <ipxe/iobuf.h>
+#include <ipxe/process.h>
+#include <ipxe/infiniband.h>
+#include <ipxe/ib_mi.h>
+#include <ipxe/ib_sma.h>
+
+/** @file
+ *
+ * Infiniband protocol
+ *
+ */
+
+/** List of Infiniband devices */
+struct list_head ib_devices = LIST_HEAD_INIT ( ib_devices );
+
+/** List of open Infiniband devices, in reverse order of opening */
+static struct list_head open_ib_devices = LIST_HEAD_INIT ( open_ib_devices );
+
+/* Disambiguate the various possible EINPROGRESSes */
+#define EINPROGRESS_INIT __einfo_error ( EINFO_EINPROGRESS_INIT )
+#define EINFO_EINPROGRESS_INIT __einfo_uniqify \
+ ( EINFO_EINPROGRESS, 0x01, "Initialising" )
+#define EINPROGRESS_ARMED __einfo_error ( EINFO_EINPROGRESS_ARMED )
+#define EINFO_EINPROGRESS_ARMED __einfo_uniqify \
+ ( EINFO_EINPROGRESS, 0x02, "Armed" )
+
+/** Human-readable message for the link statuses */
+struct errortab infiniband_errors[] __errortab = {
+ __einfo_errortab ( EINFO_EINPROGRESS_INIT ),
+ __einfo_errortab ( EINFO_EINPROGRESS_ARMED ),
+};
+
+/***************************************************************************
+ *
+ * Completion queues
+ *
+ ***************************************************************************
+ */
+
+/**
+ * Create completion queue
+ *
+ * @v ibdev Infiniband device
+ * @v num_cqes Number of completion queue entries
+ * @v op Completion queue operations
+ * @ret cq New completion queue
+ */
+struct ib_completion_queue *
+ib_create_cq ( struct ib_device *ibdev, unsigned int num_cqes,
+ struct ib_completion_queue_operations *op ) {
+ struct ib_completion_queue *cq;
+ int rc;
+
+ DBGC ( ibdev, "IBDEV %p creating completion queue\n", ibdev );
+
+ /* Allocate and initialise data structure */
+ cq = zalloc ( sizeof ( *cq ) );
+ if ( ! cq )
+ goto err_alloc_cq;
+ cq->ibdev = ibdev;
+ list_add ( &cq->list, &ibdev->cqs );
+ cq->num_cqes = num_cqes;
+ INIT_LIST_HEAD ( &cq->work_queues );
+ cq->op = op;
+
+ /* Perform device-specific initialisation and get CQN */
+ if ( ( rc = ibdev->op->create_cq ( ibdev, cq ) ) != 0 ) {
+ DBGC ( ibdev, "IBDEV %p could not initialise completion "
+ "queue: %s\n", ibdev, strerror ( rc ) );
+ goto err_dev_create_cq;
+ }
+
+ DBGC ( ibdev, "IBDEV %p created %d-entry completion queue %p (%p) "
+ "with CQN %#lx\n", ibdev, num_cqes, cq,
+ ib_cq_get_drvdata ( cq ), cq->cqn );
+ return cq;
+
+ ibdev->op->destroy_cq ( ibdev, cq );
+ err_dev_create_cq:
+ list_del ( &cq->list );
+ free ( cq );
+ err_alloc_cq:
+ return NULL;
+}
+
+/**
+ * Destroy completion queue
+ *
+ * @v ibdev Infiniband device
+ * @v cq Completion queue
+ */
+void ib_destroy_cq ( struct ib_device *ibdev,
+ struct ib_completion_queue *cq ) {
+ DBGC ( ibdev, "IBDEV %p destroying completion queue %#lx\n",
+ ibdev, cq->cqn );
+ assert ( list_empty ( &cq->work_queues ) );
+ ibdev->op->destroy_cq ( ibdev, cq );
+ list_del ( &cq->list );
+ free ( cq );
+}
+
+/**
+ * Poll completion queue
+ *
+ * @v ibdev Infiniband device
+ * @v cq Completion queue
+ */
+void ib_poll_cq ( struct ib_device *ibdev,
+ struct ib_completion_queue *cq ) {
+ struct ib_work_queue *wq;
+
+ /* Poll completion queue */
+ ibdev->op->poll_cq ( ibdev, cq );
+
+ /* Refill receive work queues */
+ list_for_each_entry ( wq, &cq->work_queues, list ) {
+ if ( ! wq->is_send )
+ ib_refill_recv ( ibdev, wq->qp );
+ }
+}
+
+/***************************************************************************
+ *
+ * Work queues
+ *
+ ***************************************************************************
+ */
+
+/**
+ * Create queue pair
+ *
+ * @v ibdev Infiniband device
+ * @v type Queue pair type
+ * @v num_send_wqes Number of send work queue entries
+ * @v send_cq Send completion queue
+ * @v num_recv_wqes Number of receive work queue entries
+ * @v recv_cq Receive completion queue
+ * @v op Queue pair operations
+ * @ret qp Queue pair
+ *
+ * The queue pair will be left in the INIT state; you must call
+ * ib_modify_qp() before it is ready to use for sending and receiving.
+ */
+struct ib_queue_pair * ib_create_qp ( struct ib_device *ibdev,
+ enum ib_queue_pair_type type,
+ unsigned int num_send_wqes,
+ struct ib_completion_queue *send_cq,
+ unsigned int num_recv_wqes,
+ struct ib_completion_queue *recv_cq,
+ struct ib_queue_pair_operations *op ) {
+ struct ib_queue_pair *qp;
+ size_t total_size;
+ int rc;
+
+ DBGC ( ibdev, "IBDEV %p creating queue pair\n", ibdev );
+
+ /* Allocate and initialise data structure */
+ total_size = ( sizeof ( *qp ) +
+ ( num_send_wqes * sizeof ( qp->send.iobufs[0] ) ) +
+ ( num_recv_wqes * sizeof ( qp->recv.iobufs[0] ) ) );
+ qp = zalloc ( total_size );
+ if ( ! qp )
+ goto err_alloc_qp;
+ qp->ibdev = ibdev;
+ list_add ( &qp->list, &ibdev->qps );
+ qp->type = type;
+ qp->send.qp = qp;
+ qp->send.is_send = 1;
+ qp->send.cq = send_cq;
+ list_add ( &qp->send.list, &send_cq->work_queues );
+ qp->send.psn = ( random() & 0xffffffUL );
+ qp->send.num_wqes = num_send_wqes;
+ qp->send.iobufs = ( ( ( void * ) qp ) + sizeof ( *qp ) );
+ qp->recv.qp = qp;
+ qp->recv.cq = recv_cq;
+ list_add ( &qp->recv.list, &recv_cq->work_queues );
+ qp->recv.psn = ( random() & 0xffffffUL );
+ qp->recv.num_wqes = num_recv_wqes;
+ qp->recv.iobufs = ( ( ( void * ) qp ) + sizeof ( *qp ) +
+ ( num_send_wqes * sizeof ( qp->send.iobufs[0] ) ));
+ INIT_LIST_HEAD ( &qp->mgids );
+ qp->op = op;
+
+ /* Perform device-specific initialisation and get QPN */
+ if ( ( rc = ibdev->op->create_qp ( ibdev, qp ) ) != 0 ) {
+ DBGC ( ibdev, "IBDEV %p could not initialise queue pair: "
+ "%s\n", ibdev, strerror ( rc ) );
+ goto err_dev_create_qp;
+ }
+ DBGC ( ibdev, "IBDEV %p created queue pair %p (%p) with QPN %#lx\n",
+ ibdev, qp, ib_qp_get_drvdata ( qp ), qp->qpn );
+ DBGC ( ibdev, "IBDEV %p QPN %#lx has %d send entries at [%p,%p)\n",
+ ibdev, qp->qpn, num_send_wqes, qp->send.iobufs,
+ qp->recv.iobufs );
+ DBGC ( ibdev, "IBDEV %p QPN %#lx has %d receive entries at [%p,%p)\n",
+ ibdev, qp->qpn, num_recv_wqes, qp->recv.iobufs,
+ ( ( ( void * ) qp ) + total_size ) );
+
+ /* Calculate externally-visible QPN */
+ switch ( type ) {
+ case IB_QPT_SMI:
+ qp->ext_qpn = IB_QPN_SMI;
+ break;
+ case IB_QPT_GSI:
+ qp->ext_qpn = IB_QPN_GSI;
+ break;
+ default:
+ qp->ext_qpn = qp->qpn;
+ break;
+ }
+ if ( qp->ext_qpn != qp->qpn ) {
+ DBGC ( ibdev, "IBDEV %p QPN %#lx has external QPN %#lx\n",
+ ibdev, qp->qpn, qp->ext_qpn );
+ }
+
+ return qp;
+
+ ibdev->op->destroy_qp ( ibdev, qp );
+ err_dev_create_qp:
+ list_del ( &qp->send.list );
+ list_del ( &qp->recv.list );
+ list_del ( &qp->list );
+ free ( qp );
+ err_alloc_qp:
+ return NULL;
+}
+
+/**
+ * Modify queue pair
+ *
+ * @v ibdev Infiniband device
+ * @v qp Queue pair
+ * @ret rc Return status code
+ */
+int ib_modify_qp ( struct ib_device *ibdev, struct ib_queue_pair *qp ) {
+ int rc;
+
+ DBGC ( ibdev, "IBDEV %p modifying QPN %#lx\n", ibdev, qp->qpn );
+
+ if ( ( rc = ibdev->op->modify_qp ( ibdev, qp ) ) != 0 ) {
+ DBGC ( ibdev, "IBDEV %p could not modify QPN %#lx: %s\n",
+ ibdev, qp->qpn, strerror ( rc ) );
+ return rc;
+ }
+
+ return 0;
+}
+
+/**
+ * Destroy queue pair
+ *
+ * @v ibdev Infiniband device
+ * @v qp Queue pair
+ */
+void ib_destroy_qp ( struct ib_device *ibdev, struct ib_queue_pair *qp ) {
+ struct io_buffer *iobuf;
+ unsigned int i;
+
+ DBGC ( ibdev, "IBDEV %p destroying QPN %#lx\n",
+ ibdev, qp->qpn );
+
+ assert ( list_empty ( &qp->mgids ) );
+
+ /* Perform device-specific destruction */
+ ibdev->op->destroy_qp ( ibdev, qp );
+
+ /* Complete any remaining I/O buffers with errors */
+ for ( i = 0 ; i < qp->send.num_wqes ; i++ ) {
+ if ( ( iobuf = qp->send.iobufs[i] ) != NULL )
+ ib_complete_send ( ibdev, qp, iobuf, -ECANCELED );
+ }
+ for ( i = 0 ; i < qp->recv.num_wqes ; i++ ) {
+ if ( ( iobuf = qp->recv.iobufs[i] ) != NULL ) {
+ ib_complete_recv ( ibdev, qp, NULL, NULL, iobuf,
+ -ECANCELED );
+ }
+ }
+
+ /* Remove work queues from completion queue */
+ list_del ( &qp->send.list );
+ list_del ( &qp->recv.list );
+
+ /* Free QP */
+ list_del ( &qp->list );
+ free ( qp );
+}
+
+/**
+ * Find queue pair by QPN
+ *
+ * @v ibdev Infiniband device
+ * @v qpn Queue pair number
+ * @ret qp Queue pair, or NULL
+ */
+struct ib_queue_pair * ib_find_qp_qpn ( struct ib_device *ibdev,
+ unsigned long qpn ) {
+ struct ib_queue_pair *qp;
+
+ list_for_each_entry ( qp, &ibdev->qps, list ) {
+ if ( ( qpn == qp->qpn ) || ( qpn == qp->ext_qpn ) )
+ return qp;
+ }
+ return NULL;
+}
+
+/**
+ * Find queue pair by multicast GID
+ *
+ * @v ibdev Infiniband device
+ * @v gid Multicast GID
+ * @ret qp Queue pair, or NULL
+ */
+struct ib_queue_pair * ib_find_qp_mgid ( struct ib_device *ibdev,
+ union ib_gid *gid ) {
+ struct ib_queue_pair *qp;
+ struct ib_multicast_gid *mgid;
+
+ list_for_each_entry ( qp, &ibdev->qps, list ) {
+ list_for_each_entry ( mgid, &qp->mgids, list ) {
+ if ( memcmp ( &mgid->gid, gid,
+ sizeof ( mgid->gid ) ) == 0 ) {
+ return qp;
+ }
+ }
+ }
+ return NULL;
+}
+
+/**
+ * Find work queue belonging to completion queue
+ *
+ * @v cq Completion queue
+ * @v qpn Queue pair number
+ * @v is_send Find send work queue (rather than receive)
+ * @ret wq Work queue, or NULL if not found
+ */
+struct ib_work_queue * ib_find_wq ( struct ib_completion_queue *cq,
+ unsigned long qpn, int is_send ) {
+ struct ib_work_queue *wq;
+
+ list_for_each_entry ( wq, &cq->work_queues, list ) {
+ if ( ( wq->qp->qpn == qpn ) && ( wq->is_send == is_send ) )
+ return wq;
+ }
+ return NULL;
+}
+
+/**
+ * Post send work queue entry
+ *
+ * @v ibdev Infiniband device
+ * @v qp Queue pair
+ * @v dest Destination address vector
+ * @v iobuf I/O buffer
+ * @ret rc Return status code
+ */
+int ib_post_send ( struct ib_device *ibdev, struct ib_queue_pair *qp,
+ struct ib_address_vector *dest,
+ struct io_buffer *iobuf ) {
+ struct ib_address_vector dest_copy;
+ int rc;
+
+ /* Check queue fill level */
+ if ( qp->send.fill >= qp->send.num_wqes ) {
+ DBGC ( ibdev, "IBDEV %p QPN %#lx send queue full\n",
+ ibdev, qp->qpn );
+ return -ENOBUFS;
+ }
+
+ /* Use default address vector if none specified */
+ if ( ! dest )
+ dest = &qp->av;
+
+ /* Make modifiable copy of address vector */
+ memcpy ( &dest_copy, dest, sizeof ( dest_copy ) );
+ dest = &dest_copy;
+
+ /* Fill in optional parameters in address vector */
+ if ( ! dest->qkey )
+ dest->qkey = qp->qkey;
+ if ( ! dest->rate )
+ dest->rate = IB_RATE_2_5;
+
+ /* Post to hardware */
+ if ( ( rc = ibdev->op->post_send ( ibdev, qp, dest, iobuf ) ) != 0 ) {
+ DBGC ( ibdev, "IBDEV %p QPN %#lx could not post send WQE: "
+ "%s\n", ibdev, qp->qpn, strerror ( rc ) );
+ return rc;
+ }
+
+ qp->send.fill++;
+ return 0;
+}
+
+/**
+ * Post receive work queue entry
+ *
+ * @v ibdev Infiniband device
+ * @v qp Queue pair
+ * @v iobuf I/O buffer
+ * @ret rc Return status code
+ */
+int ib_post_recv ( struct ib_device *ibdev, struct ib_queue_pair *qp,
+ struct io_buffer *iobuf ) {
+ int rc;
+
+ /* Check packet length */
+ if ( iob_tailroom ( iobuf ) < IB_MAX_PAYLOAD_SIZE ) {
+ DBGC ( ibdev, "IBDEV %p QPN %#lx wrong RX buffer size (%zd)\n",
+ ibdev, qp->qpn, iob_tailroom ( iobuf ) );
+ return -EINVAL;
+ }
+
+ /* Check queue fill level */
+ if ( qp->recv.fill >= qp->recv.num_wqes ) {
+ DBGC ( ibdev, "IBDEV %p QPN %#lx receive queue full\n",
+ ibdev, qp->qpn );
+ return -ENOBUFS;
+ }
+
+ /* Post to hardware */
+ if ( ( rc = ibdev->op->post_recv ( ibdev, qp, iobuf ) ) != 0 ) {
+ DBGC ( ibdev, "IBDEV %p QPN %#lx could not post receive WQE: "
+ "%s\n", ibdev, qp->qpn, strerror ( rc ) );
+ return rc;
+ }
+
+ qp->recv.fill++;
+ return 0;
+}
+
+/**
+ * Complete send work queue entry
+ *
+ * @v ibdev Infiniband device
+ * @v qp Queue pair
+ * @v iobuf I/O buffer
+ * @v rc Completion status code
+ */
+void ib_complete_send ( struct ib_device *ibdev, struct ib_queue_pair *qp,
+ struct io_buffer *iobuf, int rc ) {
+
+ if ( qp->send.cq->op->complete_send ) {
+ qp->send.cq->op->complete_send ( ibdev, qp, iobuf, rc );
+ } else {
+ free_iob ( iobuf );
+ }
+ qp->send.fill--;
+}
+
+/**
+ * Complete receive work queue entry
+ *
+ * @v ibdev Infiniband device
+ * @v qp Queue pair
+ * @v dest Destination address vector, or NULL
+ * @v source Source address vector, or NULL
+ * @v iobuf I/O buffer
+ * @v rc Completion status code
+ */
+void ib_complete_recv ( struct ib_device *ibdev, struct ib_queue_pair *qp,
+ struct ib_address_vector *dest,
+ struct ib_address_vector *source,
+ struct io_buffer *iobuf, int rc ) {
+
+ if ( qp->recv.cq->op->complete_recv ) {
+ qp->recv.cq->op->complete_recv ( ibdev, qp, dest, source,
+ iobuf, rc );
+ } else {
+ free_iob ( iobuf );
+ }
+ qp->recv.fill--;
+}
+
+/**
+ * Refill receive work queue
+ *
+ * @v ibdev Infiniband device
+ * @v qp Queue pair
+ */
+void ib_refill_recv ( struct ib_device *ibdev, struct ib_queue_pair *qp ) {
+ struct io_buffer *iobuf;
+ int rc;
+
+ /* Keep filling while unfilled entries remain */
+ while ( qp->recv.fill < qp->recv.num_wqes ) {
+
+ /* Allocate I/O buffer */
+ iobuf = qp->op->alloc_iob ( IB_MAX_PAYLOAD_SIZE );
+ if ( ! iobuf ) {
+ /* Non-fatal; we will refill on next attempt */
+ return;
+ }
+
+ /* Post I/O buffer */
+ if ( ( rc = ib_post_recv ( ibdev, qp, iobuf ) ) != 0 ) {
+ DBGC ( ibdev, "IBDEV %p could not refill: %s\n",
+ ibdev, strerror ( rc ) );
+ free_iob ( iobuf );
+ /* Give up */
+ return;
+ }
+ }
+}
+
+/***************************************************************************
+ *
+ * Link control
+ *
+ ***************************************************************************
+ */
+
+/**
+ * Get link state
+ *
+ * @v ibdev Infiniband device
+ * @ret rc Link status code
+ */
+int ib_link_rc ( struct ib_device *ibdev ) {
+ switch ( ibdev->port_state ) {
+ case IB_PORT_STATE_DOWN: return -ENOTCONN;
+ case IB_PORT_STATE_INIT: return -EINPROGRESS_INIT;
+ case IB_PORT_STATE_ARMED: return -EINPROGRESS_ARMED;
+ case IB_PORT_STATE_ACTIVE: return 0;
+ default: return -EINVAL;
+ }
+}
+
+/**
+ * Textual representation of Infiniband link state
+ *
+ * @v ibdev Infiniband device
+ * @ret link_text Link state text
+ */
+static const char * ib_link_state_text ( struct ib_device *ibdev ) {
+ switch ( ibdev->port_state ) {
+ case IB_PORT_STATE_DOWN: return "DOWN";
+ case IB_PORT_STATE_INIT: return "INIT";
+ case IB_PORT_STATE_ARMED: return "ARMED";
+ case IB_PORT_STATE_ACTIVE: return "ACTIVE";
+ default: return "UNKNOWN";
+ }
+}
+
+/**
+ * Notify drivers of Infiniband device or link state change
+ *
+ * @v ibdev Infiniband device
+ */
+static void ib_notify ( struct ib_device *ibdev ) {
+ struct ib_driver *driver;
+
+ for_each_table_entry ( driver, IB_DRIVERS )
+ driver->notify ( ibdev );
+}
+
+/**
+ * Notify of Infiniband link state change
+ *
+ * @v ibdev Infiniband device
+ */
+void ib_link_state_changed ( struct ib_device *ibdev ) {
+
+ DBGC ( ibdev, "IBDEV %p link state is %s\n",
+ ibdev, ib_link_state_text ( ibdev ) );
+
+ /* Notify drivers of link state change */
+ ib_notify ( ibdev );
+}
+
+/**
+ * Open port
+ *
+ * @v ibdev Infiniband device
+ * @ret rc Return status code
+ */
+int ib_open ( struct ib_device *ibdev ) {
+ int rc;
+
+ /* Increment device open request counter */
+ if ( ibdev->open_count++ > 0 ) {
+ /* Device was already open; do nothing */
+ return 0;
+ }
+
+ /* Open device */
+ if ( ( rc = ibdev->op->open ( ibdev ) ) != 0 ) {
+ DBGC ( ibdev, "IBDEV %p could not open: %s\n",
+ ibdev, strerror ( rc ) );
+ goto err_open;
+ }
+
+ /* Create subnet management interface */
+ ibdev->smi = ib_create_mi ( ibdev, IB_QPT_SMI );
+ if ( ! ibdev->smi ) {
+ DBGC ( ibdev, "IBDEV %p could not create SMI\n", ibdev );
+ rc = -ENOMEM;
+ goto err_create_smi;
+ }
+
+ /* Create subnet management agent */
+ if ( ( rc = ib_create_sma ( ibdev, ibdev->smi ) ) != 0 ) {
+ DBGC ( ibdev, "IBDEV %p could not create SMA: %s\n",
+ ibdev, strerror ( rc ) );
+ goto err_create_sma;
+ }
+
+ /* Create general services interface */
+ ibdev->gsi = ib_create_mi ( ibdev, IB_QPT_GSI );
+ if ( ! ibdev->gsi ) {
+ DBGC ( ibdev, "IBDEV %p could not create GSI\n", ibdev );
+ rc = -ENOMEM;
+ goto err_create_gsi;
+ }
+
+ /* Add to head of open devices list */
+ list_add ( &ibdev->open_list, &open_ib_devices );
+
+ /* Notify drivers of device state change */
+ ib_notify ( ibdev );
+
+ assert ( ibdev->open_count == 1 );
+ return 0;
+
+ ib_destroy_mi ( ibdev, ibdev->gsi );
+ err_create_gsi:
+ ib_destroy_sma ( ibdev, ibdev->smi );
+ err_create_sma:
+ ib_destroy_mi ( ibdev, ibdev->smi );
+ err_create_smi:
+ ibdev->op->close ( ibdev );
+ err_open:
+ assert ( ibdev->open_count == 1 );
+ ibdev->open_count = 0;
+ return rc;
+}
+
+/**
+ * Close port
+ *
+ * @v ibdev Infiniband device
+ */
+void ib_close ( struct ib_device *ibdev ) {
+
+ /* Decrement device open request counter */
+ ibdev->open_count--;
+
+ /* Close device if this was the last remaining requested opening */
+ if ( ibdev->open_count == 0 ) {
+ ib_notify ( ibdev );
+ list_del ( &ibdev->open_list );
+ ib_destroy_mi ( ibdev, ibdev->gsi );
+ ib_destroy_sma ( ibdev, ibdev->smi );
+ ib_destroy_mi ( ibdev, ibdev->smi );
+ ibdev->op->close ( ibdev );
+ ibdev->port_state = IB_PORT_STATE_DOWN;
+ }
+}
+
+/***************************************************************************
+ *
+ * Multicast
+ *
+ ***************************************************************************
+ */
+
+/**
+ * Attach to multicast group
+ *
+ * @v ibdev Infiniband device
+ * @v qp Queue pair
+ * @v gid Multicast GID
+ * @ret rc Return status code
+ *
+ * Note that this function handles only the local device's attachment
+ * to the multicast GID; it does not issue the relevant MADs to join
+ * the multicast group on the subnet.
+ */
+int ib_mcast_attach ( struct ib_device *ibdev, struct ib_queue_pair *qp,
+ union ib_gid *gid ) {
+ struct ib_multicast_gid *mgid;
+ int rc;
+
+ /* Add to software multicast GID list */
+ mgid = zalloc ( sizeof ( *mgid ) );
+ if ( ! mgid ) {
+ rc = -ENOMEM;
+ goto err_alloc_mgid;
+ }
+ memcpy ( &mgid->gid, gid, sizeof ( mgid->gid ) );
+ list_add ( &mgid->list, &qp->mgids );
+
+ /* Add to hardware multicast GID list */
+ if ( ( rc = ibdev->op->mcast_attach ( ibdev, qp, gid ) ) != 0 )
+ goto err_dev_mcast_attach;
+
+ return 0;
+
+ err_dev_mcast_attach:
+ list_del ( &mgid->list );
+ free ( mgid );
+ err_alloc_mgid:
+ return rc;
+}
+
+/**
+ * Detach from multicast group
+ *
+ * @v ibdev Infiniband device
+ * @v qp Queue pair
+ * @v gid Multicast GID
+ */
+void ib_mcast_detach ( struct ib_device *ibdev, struct ib_queue_pair *qp,
+ union ib_gid *gid ) {
+ struct ib_multicast_gid *mgid;
+
+ /* Remove from hardware multicast GID list */
+ ibdev->op->mcast_detach ( ibdev, qp, gid );
+
+ /* Remove from software multicast GID list */
+ list_for_each_entry ( mgid, &qp->mgids, list ) {
+ if ( memcmp ( &mgid->gid, gid, sizeof ( mgid->gid ) ) == 0 ) {
+ list_del ( &mgid->list );
+ free ( mgid );
+ break;
+ }
+ }
+}
+
+/***************************************************************************
+ *
+ * Miscellaneous
+ *
+ ***************************************************************************
+ */
+
+/**
+ * Count Infiniband HCA ports
+ *
+ * @v ibdev Infiniband device
+ * @ret num_ports Number of ports
+ */
+int ib_count_ports ( struct ib_device *ibdev ) {
+ struct ib_device *tmp;
+ int num_ports = 0;
+
+ /* Search for IB devices with the same physical device to
+ * identify port count.
+ */
+ for_each_ibdev ( tmp ) {
+ if ( tmp->dev == ibdev->dev )
+ num_ports++;
+ }
+ return num_ports;
+}
+
+/**
+ * Set port information
+ *
+ * @v ibdev Infiniband device
+ * @v mad Set port information MAD
+ */
+int ib_set_port_info ( struct ib_device *ibdev, union ib_mad *mad ) {
+ int rc;
+
+ /* Adapters with embedded SMAs do not need to support this method */
+ if ( ! ibdev->op->set_port_info ) {
+ DBGC ( ibdev, "IBDEV %p does not support setting port "
+ "information\n", ibdev );
+ return -ENOTSUP;
+ }
+
+ if ( ( rc = ibdev->op->set_port_info ( ibdev, mad ) ) != 0 ) {
+ DBGC ( ibdev, "IBDEV %p could not set port information: %s\n",
+ ibdev, strerror ( rc ) );
+ return rc;
+ }
+
+ return 0;
+};
+
+/**
+ * Set partition key table
+ *
+ * @v ibdev Infiniband device
+ * @v mad Set partition key table MAD
+ */
+int ib_set_pkey_table ( struct ib_device *ibdev, union ib_mad *mad ) {
+ int rc;
+
+ /* Adapters with embedded SMAs do not need to support this method */
+ if ( ! ibdev->op->set_pkey_table ) {
+ DBGC ( ibdev, "IBDEV %p does not support setting partition "
+ "key table\n", ibdev );
+ return -ENOTSUP;
+ }
+
+ if ( ( rc = ibdev->op->set_pkey_table ( ibdev, mad ) ) != 0 ) {
+ DBGC ( ibdev, "IBDEV %p could not set partition key table: "
+ "%s\n", ibdev, strerror ( rc ) );
+ return rc;
+ }
+
+ return 0;
+};
+
+/***************************************************************************
+ *
+ * Event queues
+ *
+ ***************************************************************************
+ */
+
+/**
+ * Poll event queue
+ *
+ * @v ibdev Infiniband device
+ */
+void ib_poll_eq ( struct ib_device *ibdev ) {
+ struct ib_completion_queue *cq;
+
+ /* Poll device's event queue */
+ ibdev->op->poll_eq ( ibdev );
+
+ /* Poll all completion queues */
+ list_for_each_entry ( cq, &ibdev->cqs, list )
+ ib_poll_cq ( ibdev, cq );
+}
+
+/**
+ * Single-step the Infiniband event queue
+ *
+ * @v process Infiniband event queue process
+ */
+static void ib_step ( struct process *process __unused ) {
+ struct ib_device *ibdev;
+
+ list_for_each_entry ( ibdev, &open_ib_devices, open_list )
+ ib_poll_eq ( ibdev );
+}
+
+/** Infiniband event queue process */
+PERMANENT_PROCESS ( ib_process, ib_step );
+
+/***************************************************************************
+ *
+ * Infiniband device creation/destruction
+ *
+ ***************************************************************************
+ */
+
+/**
+ * Allocate Infiniband device
+ *
+ * @v priv_size Size of driver private data area
+ * @ret ibdev Infiniband device, or NULL
+ */
+struct ib_device * alloc_ibdev ( size_t priv_size ) {
+ struct ib_device *ibdev;
+ void *drv_priv;
+ size_t total_len;
+
+ total_len = ( sizeof ( *ibdev ) + priv_size );
+ ibdev = zalloc ( total_len );
+ if ( ibdev ) {
+ drv_priv = ( ( ( void * ) ibdev ) + sizeof ( *ibdev ) );
+ ib_set_drvdata ( ibdev, drv_priv );
+ INIT_LIST_HEAD ( &ibdev->list );
+ INIT_LIST_HEAD ( &ibdev->open_list );
+ INIT_LIST_HEAD ( &ibdev->cqs );
+ INIT_LIST_HEAD ( &ibdev->qps );
+ ibdev->port_state = IB_PORT_STATE_DOWN;
+ ibdev->lid = IB_LID_NONE;
+ ibdev->pkey = IB_PKEY_DEFAULT;
+ }
+ return ibdev;
+}
+
+/**
+ * Register Infiniband device
+ *
+ * @v ibdev Infiniband device
+ * @ret rc Return status code
+ */
+int register_ibdev ( struct ib_device *ibdev ) {
+ struct ib_driver *driver;
+ int rc;
+
+ /* Add to device list */
+ ibdev_get ( ibdev );
+ list_add_tail ( &ibdev->list, &ib_devices );
+ DBGC ( ibdev, "IBDEV %p registered (phys %s)\n", ibdev,
+ ibdev->dev->name );
+
+ /* Probe device */
+ for_each_table_entry ( driver, IB_DRIVERS ) {
+ if ( ( rc = driver->probe ( ibdev ) ) != 0 ) {
+ DBGC ( ibdev, "IBDEV %p could not add %s device: %s\n",
+ ibdev, driver->name, strerror ( rc ) );
+ goto err_probe;
+ }
+ }
+
+ return 0;
+
+ err_probe:
+ for_each_table_entry_continue_reverse ( driver, IB_DRIVERS )
+ driver->remove ( ibdev );
+ list_del ( &ibdev->list );
+ ibdev_put ( ibdev );
+ return rc;
+}
+
+/**
+ * Unregister Infiniband device
+ *
+ * @v ibdev Infiniband device
+ */
+void unregister_ibdev ( struct ib_device *ibdev ) {
+ struct ib_driver *driver;
+
+ /* Remove device */
+ for_each_table_entry_reverse ( driver, IB_DRIVERS )
+ driver->remove ( ibdev );
+
+ /* Remove from device list */
+ list_del ( &ibdev->list );
+ ibdev_put ( ibdev );
+ DBGC ( ibdev, "IBDEV %p unregistered\n", ibdev );
+}
+
+/**
+ * Find Infiniband device by GID
+ *
+ * @v gid GID
+ * @ret ibdev Infiniband device, or NULL
+ */
+struct ib_device * find_ibdev ( union ib_gid *gid ) {
+ struct ib_device *ibdev;
+
+ for_each_ibdev ( ibdev ) {
+ if ( memcmp ( gid, &ibdev->gid, sizeof ( *gid ) ) == 0 )
+ return ibdev;
+ }
+ return NULL;
+}
+
+/**
+ * Get most recently opened Infiniband device
+ *
+ * @ret ibdev Most recently opened Infiniband device, or NULL
+ */
+struct ib_device * last_opened_ibdev ( void ) {
+ struct ib_device *ibdev;
+
+ ibdev = list_first_entry ( &open_ib_devices, struct ib_device,
+ open_list );
+ if ( ! ibdev )
+ return NULL;
+
+ assert ( ibdev->open_count != 0 );
+ return ibdev;
+}
+
+/* Drag in IPoIB */
+REQUIRE_OBJECT ( ipoib );