diff options
Diffstat (limited to 'drivers/infiniband')
23 files changed, 769 insertions, 270 deletions
diff --git a/drivers/infiniband/core/cache.c b/drivers/infiniband/core/cache.c index 558c9a0fc8b..e85f7013de5 100644 --- a/drivers/infiniband/core/cache.c +++ b/drivers/infiniband/core/cache.c @@ -38,6 +38,7 @@ #include <linux/module.h> #include <linux/errno.h> #include <linux/slab.h> +#include <linux/workqueue.h> #include <rdma/ib_cache.h> diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c index eff591deeb4..40c004a2697 100644 --- a/drivers/infiniband/core/cm.c +++ b/drivers/infiniband/core/cm.c @@ -306,7 +306,9 @@ static int cm_alloc_id(struct cm_id_private *cm_id_priv) do { spin_lock_irqsave(&cm.lock, flags); ret = idr_get_new_above(&cm.local_id_table, cm_id_priv, - next_id++, &id); + next_id, &id); + if (!ret) + next_id = ((unsigned) id + 1) & MAX_ID_MASK; spin_unlock_irqrestore(&cm.lock, flags); } while( (ret == -EAGAIN) && idr_pre_get(&cm.local_id_table, GFP_KERNEL) ); @@ -1295,26 +1297,29 @@ static struct cm_id_private * cm_match_req(struct cm_work *work, req_msg = (struct cm_req_msg *)work->mad_recv_wc->recv_buf.mad; - /* Check for duplicate REQ and stale connections. */ + /* Check for possible duplicate REQ. */ spin_lock_irqsave(&cm.lock, flags); timewait_info = cm_insert_remote_id(cm_id_priv->timewait_info); - if (!timewait_info) - timewait_info = cm_insert_remote_qpn(cm_id_priv->timewait_info); - if (timewait_info) { cur_cm_id_priv = cm_get_id(timewait_info->work.local_id, timewait_info->work.remote_id); - cm_cleanup_timewait(cm_id_priv->timewait_info); spin_unlock_irqrestore(&cm.lock, flags); if (cur_cm_id_priv) { cm_dup_req_handler(work, cur_cm_id_priv); cm_deref_id(cur_cm_id_priv); - } else - cm_issue_rej(work->port, work->mad_recv_wc, - IB_CM_REJ_STALE_CONN, CM_MSG_RESPONSE_REQ, - NULL, 0); - listen_cm_id_priv = NULL; - goto out; + } + return NULL; + } + + /* Check for stale connections. */ + timewait_info = cm_insert_remote_qpn(cm_id_priv->timewait_info); + if (timewait_info) { + cm_cleanup_timewait(cm_id_priv->timewait_info); + spin_unlock_irqrestore(&cm.lock, flags); + cm_issue_rej(work->port, work->mad_recv_wc, + IB_CM_REJ_STALE_CONN, CM_MSG_RESPONSE_REQ, + NULL, 0); + return NULL; } /* Find matching listen request. */ diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c index 592c90aa318..3ada17c0f23 100644 --- a/drivers/infiniband/core/device.c +++ b/drivers/infiniband/core/device.c @@ -40,6 +40,7 @@ #include <linux/slab.h> #include <linux/init.h> #include <linux/mutex.h> +#include <linux/workqueue.h> #include "core_priv.h" @@ -149,6 +150,18 @@ static int alloc_name(char *name) return 0; } +static int start_port(struct ib_device *device) +{ + return (device->node_type == RDMA_NODE_IB_SWITCH) ? 0 : 1; +} + + +static int end_port(struct ib_device *device) +{ + return (device->node_type == RDMA_NODE_IB_SWITCH) ? + 0 : device->phys_port_cnt; +} + /** * ib_alloc_device - allocate an IB device struct * @size:size of structure to allocate @@ -208,6 +221,45 @@ static int add_client_context(struct ib_device *device, struct ib_client *client return 0; } +static int read_port_table_lengths(struct ib_device *device) +{ + struct ib_port_attr *tprops = NULL; + int num_ports, ret = -ENOMEM; + u8 port_index; + + tprops = kmalloc(sizeof *tprops, GFP_KERNEL); + if (!tprops) + goto out; + + num_ports = end_port(device) - start_port(device) + 1; + + device->pkey_tbl_len = kmalloc(sizeof *device->pkey_tbl_len * num_ports, + GFP_KERNEL); + device->gid_tbl_len = kmalloc(sizeof *device->gid_tbl_len * num_ports, + GFP_KERNEL); + if (!device->pkey_tbl_len || !device->gid_tbl_len) + goto err; + + for (port_index = 0; port_index < num_ports; ++port_index) { + ret = ib_query_port(device, port_index + start_port(device), + tprops); + if (ret) + goto err; + device->pkey_tbl_len[port_index] = tprops->pkey_tbl_len; + device->gid_tbl_len[port_index] = tprops->gid_tbl_len; + } + + ret = 0; + goto out; + +err: + kfree(device->gid_tbl_len); + kfree(device->pkey_tbl_len); +out: + kfree(tprops); + return ret; +} + /** * ib_register_device - Register an IB device with IB core * @device:Device to register @@ -239,10 +291,19 @@ int ib_register_device(struct ib_device *device) spin_lock_init(&device->event_handler_lock); spin_lock_init(&device->client_data_lock); + ret = read_port_table_lengths(device); + if (ret) { + printk(KERN_WARNING "Couldn't create table lengths cache for device %s\n", + device->name); + goto out; + } + ret = ib_device_register_sysfs(device); if (ret) { printk(KERN_WARNING "Couldn't register device %s with driver model\n", device->name); + kfree(device->gid_tbl_len); + kfree(device->pkey_tbl_len); goto out; } @@ -284,6 +345,9 @@ void ib_unregister_device(struct ib_device *device) list_del(&device->core_list); + kfree(device->gid_tbl_len); + kfree(device->pkey_tbl_len); + mutex_unlock(&device_mutex); spin_lock_irqsave(&device->client_data_lock, flags); @@ -506,10 +570,7 @@ int ib_query_port(struct ib_device *device, u8 port_num, struct ib_port_attr *port_attr) { - if (device->node_type == RDMA_NODE_IB_SWITCH) { - if (port_num) - return -EINVAL; - } else if (port_num < 1 || port_num > device->phys_port_cnt) + if (port_num < start_port(device) || port_num > end_port(device)) return -EINVAL; return device->query_port(device, port_num, port_attr); @@ -581,10 +642,7 @@ int ib_modify_port(struct ib_device *device, u8 port_num, int port_modify_mask, struct ib_port_modify *port_modify) { - if (device->node_type == RDMA_NODE_IB_SWITCH) { - if (port_num) - return -EINVAL; - } else if (port_num < 1 || port_num > device->phys_port_cnt) + if (port_num < start_port(device) || port_num > end_port(device)) return -EINVAL; return device->modify_port(device, port_num, port_modify_mask, @@ -592,6 +650,68 @@ int ib_modify_port(struct ib_device *device, } EXPORT_SYMBOL(ib_modify_port); +/** + * ib_find_gid - Returns the port number and GID table index where + * a specified GID value occurs. + * @device: The device to query. + * @gid: The GID value to search for. + * @port_num: The port number of the device where the GID value was found. + * @index: The index into the GID table where the GID was found. This + * parameter may be NULL. + */ +int ib_find_gid(struct ib_device *device, union ib_gid *gid, + u8 *port_num, u16 *index) +{ + union ib_gid tmp_gid; + int ret, port, i; + + for (port = start_port(device); port <= end_port(device); ++port) { + for (i = 0; i < device->gid_tbl_len[port - start_port(device)]; ++i) { + ret = ib_query_gid(device, port, i, &tmp_gid); + if (ret) + return ret; + if (!memcmp(&tmp_gid, gid, sizeof *gid)) { + *port_num = port; + if (index) + *index = i; + return 0; + } + } + } + + return -ENOENT; +} +EXPORT_SYMBOL(ib_find_gid); + +/** + * ib_find_pkey - Returns the PKey table index where a specified + * PKey value occurs. + * @device: The device to query. + * @port_num: The port number of the device to search for the PKey. + * @pkey: The PKey value to search for. + * @index: The index into the PKey table where the PKey was found. + */ +int ib_find_pkey(struct ib_device *device, + u8 port_num, u16 pkey, u16 *index) +{ + int ret, i; + u16 tmp_pkey; + + for (i = 0; i < device->pkey_tbl_len[port_num - start_port(device)]; ++i) { + ret = ib_query_pkey(device, port_num, i, &tmp_pkey); + if (ret) + return ret; + + if (pkey == tmp_pkey) { + *index = i; + return 0; + } + } + + return -ENOENT; +} +EXPORT_SYMBOL(ib_find_pkey); + static int __init ib_core_init(void) { int ret; diff --git a/drivers/infiniband/core/umem.c b/drivers/infiniband/core/umem.c index f32ca5fbb26..b4aec5103c9 100644 --- a/drivers/infiniband/core/umem.c +++ b/drivers/infiniband/core/umem.c @@ -36,6 +36,7 @@ #include <linux/mm.h> #include <linux/dma-mapping.h> +#include <linux/sched.h> #include "uverbs.h" @@ -209,8 +210,10 @@ void ib_umem_release(struct ib_umem *umem) __ib_umem_release(umem->context->device, umem, 1); mm = get_task_mm(current); - if (!mm) + if (!mm) { + kfree(umem); return; + } diff = PAGE_ALIGN(umem->length + umem->offset) >> PAGE_SHIFT; diff --git a/drivers/infiniband/hw/ehca/ehca_mrmw.c b/drivers/infiniband/hw/ehca/ehca_mrmw.c index 84c5bb49856..add79bd44e3 100644 --- a/drivers/infiniband/hw/ehca/ehca_mrmw.c +++ b/drivers/infiniband/hw/ehca/ehca_mrmw.c @@ -2050,13 +2050,10 @@ int ehca_mrmw_map_hrc_alloc(const u64 hipz_rc) switch (hipz_rc) { case H_SUCCESS: /* successful completion */ return 0; - case H_ADAPTER_PARM: /* invalid adapter handle */ - case H_RT_PARM: /* invalid resource type */ case H_NOT_ENOUGH_RESOURCES: /* insufficient resources */ - case H_MLENGTH_PARM: /* invalid memory length */ - case H_MEM_ACCESS_PARM: /* invalid access controls */ case H_CONSTRAINED: /* resource constraint */ - return -EINVAL; + case H_NO_MEM: + return -ENOMEM; case H_BUSY: /* long busy */ return -EBUSY; default: diff --git a/drivers/infiniband/hw/ehca/hcp_if.c b/drivers/infiniband/hw/ehca/hcp_if.c index 7f0beec74f7..5766ae3a202 100644 --- a/drivers/infiniband/hw/ehca/hcp_if.c +++ b/drivers/infiniband/hw/ehca/hcp_if.c @@ -331,7 +331,7 @@ u64 hipz_h_alloc_resource_qp(const struct ipz_adapter_handle adapter_handle, 0); qp->ipz_qp_handle.handle = outs[0]; qp->real_qp_num = (u32)outs[1]; - parms->act_nr_send_sges = + parms->act_nr_send_wqes = (u16)EHCA_BMASK_GET(H_ALL_RES_QP_ACT_OUTST_SEND_WR, outs[2]); parms->act_nr_recv_wqes = (u16)EHCA_BMASK_GET(H_ALL_RES_QP_ACT_OUTST_RECV_WR, outs[2]); diff --git a/drivers/infiniband/hw/ipath/ipath_verbs_mcast.c b/drivers/infiniband/hw/ipath/ipath_verbs_mcast.c index 085e28b939e..dd691cfa507 100644 --- a/drivers/infiniband/hw/ipath/ipath_verbs_mcast.c +++ b/drivers/infiniband/hw/ipath/ipath_verbs_mcast.c @@ -165,10 +165,9 @@ static int ipath_mcast_add(struct ipath_ibdev *dev, { struct rb_node **n = &mcast_tree.rb_node; struct rb_node *pn = NULL; - unsigned long flags; int ret; - spin_lock_irqsave(&mcast_lock, flags); + spin_lock_irq(&mcast_lock); while (*n) { struct ipath_mcast *tmcast; @@ -228,7 +227,7 @@ static int ipath_mcast_add(struct ipath_ibdev *dev, ret = 0; bail: - spin_unlock_irqrestore(&mcast_lock, flags); + spin_unlock_irq(&mcast_lock); return ret; } @@ -289,17 +288,16 @@ int ipath_multicast_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid) struct ipath_mcast *mcast = NULL; struct ipath_mcast_qp *p, *tmp; struct rb_node *n; - unsigned long flags; int last = 0; int ret; - spin_lock_irqsave(&mcast_lock, flags); + spin_lock_irq(&mcast_lock); /* Find the GID in the mcast table. */ n = mcast_tree.rb_node; while (1) { if (n == NULL) { - spin_unlock_irqrestore(&mcast_lock, flags); + spin_unlock_irq(&mcast_lock); ret = -EINVAL; goto bail; } @@ -334,7 +332,7 @@ int ipath_multicast_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid) break; } - spin_unlock_irqrestore(&mcast_lock, flags); + spin_unlock_irq(&mcast_lock); if (p) { /* @@ -348,9 +346,9 @@ int ipath_multicast_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid) atomic_dec(&mcast->refcount); wait_event(mcast->wait, !atomic_read(&mcast->refcount)); ipath_mcast_free(mcast); - spin_lock(&dev->n_mcast_grps_lock); + spin_lock_irq(&dev->n_mcast_grps_lock); dev->n_mcast_grps_allocated--; - spin_unlock(&dev->n_mcast_grps_lock); + spin_unlock_irq(&dev->n_mcast_grps_lock); } ret = 0; diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c index 5cd70690845..dc137dec230 100644 --- a/drivers/infiniband/hw/mlx4/qp.c +++ b/drivers/infiniband/hw/mlx4/qp.c @@ -188,14 +188,32 @@ static int send_wqe_overhead(enum ib_qp_type type) } } -static int set_qp_size(struct mlx4_ib_dev *dev, struct ib_qp_cap *cap, - enum ib_qp_type type, struct mlx4_ib_qp *qp) +static int set_rq_size(struct mlx4_ib_dev *dev, struct ib_qp_cap *cap, + struct mlx4_ib_qp *qp) { - /* Sanity check QP size before proceeding */ + /* Sanity check RQ size before proceeding */ + if (cap->max_recv_wr > dev->dev->caps.max_wqes || + cap->max_recv_sge > dev->dev->caps.max_rq_sg) + return -EINVAL; + + qp->rq.max = cap->max_recv_wr ? roundup_pow_of_two(cap->max_recv_wr) : 0; + + qp->rq.wqe_shift = ilog2(roundup_pow_of_two(cap->max_recv_sge * + sizeof (struct mlx4_wqe_data_seg))); + qp->rq.max_gs = (1 << qp->rq.wqe_shift) / sizeof (struct mlx4_wqe_data_seg); + + cap->max_recv_wr = qp->rq.max; + cap->max_recv_sge = qp->rq.max_gs; + + return 0; +} + +static int set_kernel_sq_size(struct mlx4_ib_dev *dev, struct ib_qp_cap *cap, + enum ib_qp_type type, struct mlx4_ib_qp *qp) +{ + /* Sanity check SQ size before proceeding */ if (cap->max_send_wr > dev->dev->caps.max_wqes || - cap->max_recv_wr > dev->dev->caps.max_wqes || cap->max_send_sge > dev->dev->caps.max_sq_sg || - cap->max_recv_sge > dev->dev->caps.max_rq_sg || cap->max_inline_data + send_wqe_overhead(type) + sizeof (struct mlx4_wqe_inline_seg) > dev->dev->caps.max_sq_desc_sz) return -EINVAL; @@ -208,12 +226,7 @@ static int set_qp_size(struct mlx4_ib_dev *dev, struct ib_qp_cap *cap, cap->max_send_sge + 2 > dev->dev->caps.max_sq_sg) return -EINVAL; - qp->rq.max = cap->max_recv_wr ? roundup_pow_of_two(cap->max_recv_wr) : 0; - qp->sq.max = cap->max_send_wr ? roundup_pow_of_two(cap->max_send_wr) : 0; - - qp->rq.wqe_shift = ilog2(roundup_pow_of_two(cap->max_recv_sge * - sizeof (struct mlx4_wqe_data_seg))); - qp->rq.max_gs = (1 << qp->rq.wqe_shift) / sizeof (struct mlx4_wqe_data_seg); + qp->sq.max = cap->max_send_wr ? roundup_pow_of_two(cap->max_send_wr) : 1; qp->sq.wqe_shift = ilog2(roundup_pow_of_two(max(cap->max_send_sge * sizeof (struct mlx4_wqe_data_seg), @@ -233,23 +246,31 @@ static int set_qp_size(struct mlx4_ib_dev *dev, struct ib_qp_cap *cap, qp->sq.offset = 0; } - cap->max_send_wr = qp->sq.max; - cap->max_recv_wr = qp->rq.max; - cap->max_send_sge = qp->sq.max_gs; - cap->max_recv_sge = qp->rq.max_gs; + cap->max_send_wr = qp->sq.max; + cap->max_send_sge = qp->sq.max_gs; cap->max_inline_data = (1 << qp->sq.wqe_shift) - send_wqe_overhead(type) - sizeof (struct mlx4_wqe_inline_seg); return 0; } +static int set_user_sq_size(struct mlx4_ib_qp *qp, + struct mlx4_ib_create_qp *ucmd) +{ + qp->sq.max = 1 << ucmd->log_sq_bb_count; + qp->sq.wqe_shift = ucmd->log_sq_stride; + + qp->buf_size = (qp->rq.max << qp->rq.wqe_shift) + + (qp->sq.max << qp->sq.wqe_shift); + + return 0; +} + static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd, struct ib_qp_init_attr *init_attr, struct ib_udata *udata, int sqpn, struct mlx4_ib_qp *qp) { - struct mlx4_wqe_ctrl_seg *ctrl; int err; - int i; mutex_init(&qp->mutex); spin_lock_init(&qp->sq.lock); @@ -264,7 +285,7 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd, qp->sq.head = 0; qp->sq.tail = 0; - err = set_qp_size(dev, &init_attr->cap, init_attr->qp_type, qp); + err = set_rq_size(dev, &init_attr->cap, qp); if (err) goto err; @@ -276,6 +297,10 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd, goto err; } + err = set_user_sq_size(qp, &ucmd); + if (err) + goto err; + qp->umem = ib_umem_get(pd->uobject->context, ucmd.buf_addr, qp->buf_size, 0); if (IS_ERR(qp->umem)) { @@ -292,16 +317,24 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd, if (err) goto err_mtt; - err = mlx4_ib_db_map_user(to_mucontext(pd->uobject->context), - ucmd.db_addr, &qp->db); - if (err) - goto err_mtt; + if (!init_attr->srq) { + err = mlx4_ib_db_map_user(to_mucontext(pd->uobject->context), + ucmd.db_addr, &qp->db); + if (err) + goto err_mtt; + } } else { - err = mlx4_ib_db_alloc(dev, &qp->db, 0); + err = set_kernel_sq_size(dev, &init_attr->cap, init_attr->qp_type, qp); if (err) goto err; - *qp->db.db = 0; + if (!init_attr->srq) { + err = mlx4_ib_db_alloc(dev, &qp->db, 0); + if (err) + goto err; + + *qp->db.db = 0; + } if (mlx4_buf_alloc(dev->dev, qp->buf_size, PAGE_SIZE * 2, &qp->buf)) { err = -ENOMEM; @@ -317,11 +350,6 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd, if (err) goto err_mtt; - for (i = 0; i < qp->sq.max; ++i) { - ctrl = get_send_wqe(qp, i); - ctrl->owner_opcode = cpu_to_be32(1 << 31); - } - qp->sq.wrid = kmalloc(qp->sq.max * sizeof (u64), GFP_KERNEL); qp->rq.wrid = kmalloc(qp->rq.max * sizeof (u64), GFP_KERNEL); @@ -355,7 +383,7 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd, return 0; err_wrid: - if (pd->uobject) + if (pd->uobject && !init_attr->srq) mlx4_ib_db_unmap_user(to_mucontext(pd->uobject->context), &qp->db); else { kfree(qp->sq.wrid); @@ -372,7 +400,7 @@ err_buf: mlx4_buf_free(dev->dev, qp->buf_size, &qp->buf); err_db: - if (!pd->uobject) + if (!pd->uobject && !init_attr->srq) mlx4_ib_db_free(dev, &qp->db); err: @@ -450,14 +478,16 @@ static void destroy_qp_common(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *qp, mlx4_mtt_cleanup(dev->dev, &qp->mtt); if (is_user) { - mlx4_ib_db_unmap_user(to_mucontext(qp->ibqp.uobject->context), - &qp->db); + if (!qp->ibqp.srq) + mlx4_ib_db_unmap_user(to_mucontext(qp->ibqp.uobject->context), + &qp->db); ib_umem_release(qp->umem); } else { kfree(qp->sq.wrid); kfree(qp->rq.wrid); mlx4_buf_free(dev->dev, qp->buf_size, &qp->buf); - mlx4_ib_db_free(dev, &qp->db); + if (!qp->ibqp.srq) + mlx4_ib_db_free(dev, &qp->db); } } @@ -573,7 +603,7 @@ static int to_mlx4_st(enum ib_qp_type type) } } -static __be32 to_mlx4_access_flags(struct mlx4_ib_qp *qp, struct ib_qp_attr *attr, +static __be32 to_mlx4_access_flags(struct mlx4_ib_qp *qp, const struct ib_qp_attr *attr, int attr_mask) { u8 dest_rd_atomic; @@ -603,7 +633,7 @@ static __be32 to_mlx4_access_flags(struct mlx4_ib_qp *qp, struct ib_qp_attr *att return cpu_to_be32(hw_access_flags); } -static void store_sqp_attrs(struct mlx4_ib_sqp *sqp, struct ib_qp_attr *attr, +static void store_sqp_attrs(struct mlx4_ib_sqp *sqp, const struct ib_qp_attr *attr, int attr_mask) { if (attr_mask & IB_QP_PKEY_INDEX) @@ -619,7 +649,7 @@ static void mlx4_set_sched(struct mlx4_qp_path *path, u8 port) path->sched_queue = (path->sched_queue & 0xbf) | ((port - 1) << 6); } -static int mlx4_set_path(struct mlx4_ib_dev *dev, struct ib_ah_attr *ah, +static int mlx4_set_path(struct mlx4_ib_dev *dev, const struct ib_ah_attr *ah, struct mlx4_qp_path *path, u8 port) { path->grh_mylmc = ah->src_path_bits & 0x7f; @@ -655,14 +685,14 @@ static int mlx4_set_path(struct mlx4_ib_dev *dev, struct ib_ah_attr *ah, return 0; } -int mlx4_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, - int attr_mask, struct ib_udata *udata) +static int __mlx4_ib_modify_qp(struct ib_qp *ibqp, + const struct ib_qp_attr *attr, int attr_mask, + enum ib_qp_state cur_state, enum ib_qp_state new_state) { struct mlx4_ib_dev *dev = to_mdev(ibqp->device); struct mlx4_ib_qp *qp = to_mqp(ibqp); struct mlx4_qp_context *context; enum mlx4_qp_optpar optpar = 0; - enum ib_qp_state cur_state, new_state; int sqd_event; int err = -EINVAL; @@ -670,34 +700,6 @@ int mlx4_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, if (!context) return -ENOMEM; - mutex_lock(&qp->mutex); - - cur_state = attr_mask & IB_QP_CUR_STATE ? attr->cur_qp_state : qp->state; - new_state = attr_mask & IB_QP_STATE ? attr->qp_state : cur_state; - - if (!ib_modify_qp_is_ok(cur_state, new_state, ibqp->qp_type, attr_mask)) - goto out; - - if ((attr_mask & IB_QP_PKEY_INDEX) && - attr->pkey_index >= dev->dev->caps.pkey_table_len) { - goto out; - } - - if ((attr_mask & IB_QP_PORT) && - (attr->port_num == 0 || attr->port_num > dev->dev->caps.num_ports)) { - goto out; - } - - if (attr_mask & IB_QP_MAX_QP_RD_ATOMIC && - attr->max_rd_atomic > dev->dev->caps.max_qp_init_rdma) { - goto out; - } - - if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC && - attr->max_dest_rd_atomic > 1 << dev->dev->caps.max_qp_dest_rdma) { - goto out; - } - context->flags = cpu_to_be32((to_mlx4_state(new_state) << 28) | (to_mlx4_st(ibqp->qp_type) << 16)); context->flags |= cpu_to_be32(1 << 8); /* DE? */ @@ -849,7 +851,7 @@ int mlx4_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, if (ibqp->srq) context->srqn = cpu_to_be32(1 << 24 | to_msrq(ibqp->srq)->msrq.srqn); - if (cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT) + if (!ibqp->srq && cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT) context->db_rec_addr = cpu_to_be64(qp->db.dma); if (cur_state == IB_QPS_INIT && @@ -869,6 +871,21 @@ int mlx4_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, else sqd_event = 0; + /* + * Before passing a kernel QP to the HW, make sure that the + * ownership bits of the send queue are set so that the + * hardware doesn't start processing stale work requests. + */ + if (!ibqp->uobject && cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT) { + struct mlx4_wqe_ctrl_seg *ctrl; + int i; + + for (i = 0; i < qp->sq.max; ++i) { + ctrl = get_send_wqe(qp, i); + ctrl->owner_opcode = cpu_to_be32(1 << 31); + } + } + err = mlx4_qp_modify(dev->dev, &qp->mtt, to_mlx4_state(cur_state), to_mlx4_state(new_state), context, optpar, sqd_event, &qp->mqp); @@ -916,15 +933,89 @@ int mlx4_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, qp->rq.tail = 0; qp->sq.head = 0; qp->sq.tail = 0; - *qp->db.db = 0; + if (!ibqp->srq) + *qp->db.db = 0; } out: - mutex_unlock(&qp->mutex); kfree(context); return err; } +static const struct ib_qp_attr mlx4_ib_qp_attr = { .port_num = 1 }; +static const int mlx4_ib_qp_attr_mask_table[IB_QPT_UD + 1] = { + [IB_QPT_UD] = (IB_QP_PKEY_INDEX | + IB_QP_PORT | + IB_QP_QKEY), + [IB_QPT_UC] = (IB_QP_PKEY_INDEX | + IB_QP_PORT | + IB_QP_ACCESS_FLAGS), + [IB_QPT_RC] = (IB_QP_PKEY_INDEX | + IB_QP_PORT | + IB_QP_ACCESS_FLAGS), + [IB_QPT_SMI] = (IB_QP_PKEY_INDEX | + IB_QP_QKEY), + [IB_QPT_GSI] = (IB_QP_PKEY_INDEX | + IB_QP_QKEY), +}; + +int mlx4_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, + int attr_mask, struct ib_udata *udata) +{ + struct mlx4_ib_dev *dev = to_mdev(ibqp->device); + struct mlx4_ib_qp *qp = to_mqp(ibqp); + enum ib_qp_state cur_state, new_state; + int err = -EINVAL; + + mutex_lock(&qp->mutex); + + cur_state = attr_mask & IB_QP_CUR_STATE ? attr->cur_qp_state : qp->state; + new_state = attr_mask & IB_QP_STATE ? attr->qp_state : cur_state; + + if (!ib_modify_qp_is_ok(cur_state, new_state, ibqp->qp_type, attr_mask)) + goto out; + + if ((attr_mask & IB_QP_PKEY_INDEX) && + attr->pkey_index >= dev->dev->caps.pkey_table_len) { + goto out; + } + + if ((attr_mask & IB_QP_PORT) && + (attr->port_num == 0 || attr->port_num > dev->dev->caps.num_ports)) { + goto out; + } + + if (attr_mask & IB_QP_MAX_QP_RD_ATOMIC && + attr->max_rd_atomic > dev->dev->caps.max_qp_init_rdma) { + goto out; + } + + if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC && + attr->max_dest_rd_atomic > dev->dev->caps.max_qp_dest_rdma) { + goto out; + } + + if (cur_state == new_state && cur_state == IB_QPS_RESET) { + err = 0; + goto out; + } + + if (cur_state == IB_QPS_RESET && new_state == IB_QPS_ERR) { + err = __mlx4_ib_modify_qp(ibqp, &mlx4_ib_qp_attr, + mlx4_ib_qp_attr_mask_table[ibqp->qp_type], + IB_QPS_RESET, IB_QPS_INIT); + if (err) + goto out; + cur_state = IB_QPS_INIT; + } + + err = __mlx4_ib_modify_qp(ibqp, attr, attr_mask, cur_state, new_state); + +out: + mutex_unlock(&qp->mutex); + return err; +} + static int build_mlx_header(struct mlx4_ib_sqp *sqp, struct ib_send_wr *wr, void *wqe) { @@ -952,6 +1043,7 @@ static int build_mlx_header(struct mlx4_ib_sqp *sqp, struct ib_send_wr *wr, (be32_to_cpu(ah->av.sl_tclass_flowlabel) >> 20) & 0xff; sqp->ud_header.grh.flow_label = ah->av.sl_tclass_flowlabel & cpu_to_be32(0xfffff); + sqp->ud_header.grh.hop_limit = ah->av.hop_limit; ib_get_cached_gid(ib_dev, be32_to_cpu(ah->av.port_pd) >> 24, ah->av.gid_index, &sqp->ud_header.grh.source_gid); memcpy(sqp->ud_header.grh.destination_gid.raw, @@ -1192,7 +1284,7 @@ int mlx4_ib_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr, */ wmb(); - if (wr->opcode < 0 || wr->opcode > ARRAY_SIZE(mlx4_ib_opcode)) { + if (wr->opcode < 0 || wr->opcode >= ARRAY_SIZE(mlx4_ib_opcode)) { err = -EINVAL; goto out; } diff --git a/drivers/infiniband/hw/mlx4/srq.c b/drivers/infiniband/hw/mlx4/srq.c index 42ab4a801d6..12fac1c8989 100644 --- a/drivers/infiniband/hw/mlx4/srq.c +++ b/drivers/infiniband/hw/mlx4/srq.c @@ -297,6 +297,12 @@ int mlx4_ib_post_srq_recv(struct ib_srq *ibsrq, struct ib_recv_wr *wr, break; } + if (unlikely(srq->head == srq->tail)) { + err = -ENOMEM; + *bad_wr = wr; + break; + } + srq->wrid[srq->head] = wr->wr_id; next = get_wqe(srq, srq->head); diff --git a/drivers/infiniband/hw/mlx4/user.h b/drivers/infiniband/hw/mlx4/user.h index 5b8eddc9fa8..88c72d56368 100644 --- a/drivers/infiniband/hw/mlx4/user.h +++ b/drivers/infiniband/hw/mlx4/user.h @@ -39,7 +39,7 @@ * Increment this value if any changes that break userspace ABI * compatibility are made. */ -#define MLX4_IB_UVERBS_ABI_VERSION 1 +#define MLX4_IB_UVERBS_ABI_VERSION 2 /* * Make sure that all structs defined in this file remain laid out so @@ -87,6 +87,9 @@ struct mlx4_ib_create_srq_resp { struct mlx4_ib_create_qp { __u64 buf_addr; __u64 db_addr; + __u8 log_sq_bb_count; + __u8 log_sq_stride; + __u8 reserved[6]; }; #endif /* MLX4_IB_USER_H */ diff --git a/drivers/infiniband/hw/mthca/mthca_av.c b/drivers/infiniband/hw/mthca/mthca_av.c index 27caf3b0648..4b111a852ff 100644 --- a/drivers/infiniband/hw/mthca/mthca_av.c +++ b/drivers/infiniband/hw/mthca/mthca_av.c @@ -279,6 +279,7 @@ int mthca_read_ah(struct mthca_dev *dev, struct mthca_ah *ah, (be32_to_cpu(ah->av->sl_tclass_flowlabel) >> 20) & 0xff; header->grh.flow_label = ah->av->sl_tclass_flowlabel & cpu_to_be32(0xfffff); + header->grh.hop_limit = ah->av->hop_limit; ib_get_cached_gid(&dev->ib_dev, be32_to_cpu(ah->av->port_pd) >> 24, ah->av->gid_index % dev->limits.gid_table_len, diff --git a/drivers/infiniband/hw/mthca/mthca_cmd.c b/drivers/infiniband/hw/mthca/mthca_cmd.c index 71314460b11..38102520ffb 100644 --- a/drivers/infiniband/hw/mthca/mthca_cmd.c +++ b/drivers/infiniband/hw/mthca/mthca_cmd.c @@ -37,6 +37,7 @@ #include <linux/completion.h> #include <linux/pci.h> #include <linux/errno.h> +#include <linux/sched.h> #include <asm/io.h> #include <rdma/ib_mad.h> diff --git a/drivers/infiniband/hw/mthca/mthca_cq.c b/drivers/infiniband/hw/mthca/mthca_cq.c index ca224d018af..be6e1e03bda 100644 |