summaryrefslogtreecommitdiffstats
path: root/drivers/infiniband
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/infiniband')
-rw-r--r--drivers/infiniband/core/cm.c2
-rw-r--r--drivers/infiniband/core/cq.c14
-rw-r--r--drivers/infiniband/core/rdma_core.c6
-rw-r--r--drivers/infiniband/core/ucma.c4
-rw-r--r--drivers/infiniband/hw/hns/hns_roce_hw_v2.c34
-rw-r--r--drivers/infiniband/hw/hns/hns_roce_mr.c2
-rw-r--r--drivers/infiniband/hw/mlx5/odp.c27
-rw-r--r--drivers/infiniband/hw/mlx5/qp.c29
-rw-r--r--drivers/infiniband/hw/mlx5/srq_cmd.c4
-rw-r--r--drivers/infiniband/sw/rdmavt/qp.c33
-rw-r--r--drivers/infiniband/sw/rdmavt/rc.c4
11 files changed, 87 insertions, 72 deletions
diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c
index 0d1377232933..dc0558b23158 100644
--- a/drivers/infiniband/core/cm.c
+++ b/drivers/infiniband/core/cm.c
@@ -3676,10 +3676,12 @@ static int cm_send_sidr_rep_locked(struct cm_id_private *cm_id_priv,
return ret;
}
cm_id_priv->id.state = IB_CM_IDLE;
+ spin_lock_irq(&cm.lock);
if (!RB_EMPTY_NODE(&cm_id_priv->sidr_id_node)) {
rb_erase(&cm_id_priv->sidr_id_node, &cm.remote_sidr_table);
RB_CLEAR_NODE(&cm_id_priv->sidr_id_node);
}
+ spin_unlock_irq(&cm.lock);
return 0;
}
diff --git a/drivers/infiniband/core/cq.c b/drivers/infiniband/core/cq.c
index 655795bfa0ee..513825e424bf 100644
--- a/drivers/infiniband/core/cq.c
+++ b/drivers/infiniband/core/cq.c
@@ -72,6 +72,15 @@ static void rdma_dim_init(struct ib_cq *cq)
INIT_WORK(&dim->work, ib_cq_rdma_dim_work);
}
+static void rdma_dim_destroy(struct ib_cq *cq)
+{
+ if (!cq->dim)
+ return;
+
+ cancel_work_sync(&cq->dim->work);
+ kfree(cq->dim);
+}
+
static int __poll_cq(struct ib_cq *cq, int num_entries, struct ib_wc *wc)
{
int rc;
@@ -266,6 +275,7 @@ struct ib_cq *__ib_alloc_cq_user(struct ib_device *dev, void *private,
return cq;
out_destroy_cq:
+ rdma_dim_destroy(cq);
rdma_restrack_del(&cq->res);
cq->device->ops.destroy_cq(cq, udata);
out_free_wc:
@@ -331,12 +341,10 @@ void ib_free_cq_user(struct ib_cq *cq, struct ib_udata *udata)
WARN_ON_ONCE(1);
}
+ rdma_dim_destroy(cq);
trace_cq_free(cq);
rdma_restrack_del(&cq->res);
cq->device->ops.destroy_cq(cq, udata);
- if (cq->dim)
- cancel_work_sync(&cq->dim->work);
- kfree(cq->dim);
kfree(cq->wc);
kfree(cq);
}
diff --git a/drivers/infiniband/core/rdma_core.c b/drivers/infiniband/core/rdma_core.c
index 3027cd2fb247..6d3ed7c6e19e 100644
--- a/drivers/infiniband/core/rdma_core.c
+++ b/drivers/infiniband/core/rdma_core.c
@@ -649,9 +649,6 @@ void rdma_alloc_commit_uobject(struct ib_uobject *uobj,
{
struct ib_uverbs_file *ufile = attrs->ufile;
- /* alloc_commit consumes the uobj kref */
- uobj->uapi_object->type_class->alloc_commit(uobj);
-
/* kref is held so long as the uobj is on the uobj list. */
uverbs_uobject_get(uobj);
spin_lock_irq(&ufile->uobjects_lock);
@@ -661,6 +658,9 @@ void rdma_alloc_commit_uobject(struct ib_uobject *uobj,
/* matches atomic_set(-1) in alloc_uobj */
atomic_set(&uobj->usecnt, 0);
+ /* alloc_commit consumes the uobj kref */
+ uobj->uapi_object->type_class->alloc_commit(uobj);
+
/* Matches the down_read in rdma_alloc_begin_uobject */
up_read(&ufile->hw_destroy_rwsem);
}
diff --git a/drivers/infiniband/core/ucma.c b/drivers/infiniband/core/ucma.c
index 5b87eee8ccc8..d03dacaef788 100644
--- a/drivers/infiniband/core/ucma.c
+++ b/drivers/infiniband/core/ucma.c
@@ -1084,6 +1084,8 @@ static ssize_t ucma_connect(struct ucma_file *file, const char __user *inbuf,
size_t in_size;
int ret;
+ if (in_len < offsetofend(typeof(cmd), reserved))
+ return -EINVAL;
in_size = min_t(size_t, in_len, sizeof(cmd));
if (copy_from_user(&cmd, inbuf, in_size))
return -EFAULT;
@@ -1141,6 +1143,8 @@ static ssize_t ucma_accept(struct ucma_file *file, const char __user *inbuf,
size_t in_size;
int ret;
+ if (in_len < offsetofend(typeof(cmd), reserved))
+ return -EINVAL;
in_size = min_t(size_t, in_len, sizeof(cmd));
if (copy_from_user(&cmd, inbuf, in_size))
return -EFAULT;
diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
index dd01a51816cc..0618ced45bf8 100644
--- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
+++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
@@ -3954,6 +3954,15 @@ static int config_qp_sq_buf(struct hns_roce_dev *hr_dev,
return 0;
}
+static inline enum ib_mtu get_mtu(struct ib_qp *ibqp,
+ const struct ib_qp_attr *attr)
+{
+ if (ibqp->qp_type == IB_QPT_GSI || ibqp->qp_type == IB_QPT_UD)
+ return IB_MTU_4096;
+
+ return attr->path_mtu;
+}
+
static int modify_qp_init_to_rtr(struct ib_qp *ibqp,
const struct ib_qp_attr *attr, int attr_mask,
struct hns_roce_v2_qp_context *context,
@@ -3965,6 +3974,7 @@ static int modify_qp_init_to_rtr(struct ib_qp *ibqp,
struct ib_device *ibdev = &hr_dev->ib_dev;
dma_addr_t trrl_ba;
dma_addr_t irrl_ba;
+ enum ib_mtu mtu;
u8 port_num;
u64 *mtts;
u8 *dmac;
@@ -4062,23 +4072,23 @@ static int modify_qp_init_to_rtr(struct ib_qp *ibqp,
roce_set_field(qpc_mask->byte_52_udpspn_dmac, V2_QPC_BYTE_52_DMAC_M,
V2_QPC_BYTE_52_DMAC_S, 0);
- /* mtu*(2^LP_PKTN_INI) should not bigger than 1 message length 64kb */
+ mtu = get_mtu(ibqp, attr);
+
+ if (attr_mask & IB_QP_PATH_MTU) {
+ roce_set_field(context->byte_24_mtu_tc, V2_QPC_BYTE_24_MTU_M,
+ V2_QPC_BYTE_24_MTU_S, mtu);
+ roce_set_field(qpc_mask->byte_24_mtu_tc, V2_QPC_BYTE_24_MTU_M,
+ V2_QPC_BYTE_24_MTU_S, 0);
+ }
+
+#define MAX_LP_MSG_LEN 65536
+ /* MTU*(2^LP_PKTN_INI) shouldn't be bigger than 64kb */
roce_set_field(context->byte_56_dqpn_err, V2_QPC_BYTE_56_LP_PKTN_INI_M,
V2_QPC_BYTE_56_LP_PKTN_INI_S,
- ilog2(hr_dev->caps.max_sq_inline / IB_MTU_4096));
+ ilog2(MAX_LP_MSG_LEN / ib_mtu_enum_to_int(mtu)));
roce_set_field(qpc_mask->byte_56_dqpn_err, V2_QPC_BYTE_56_LP_PKTN_INI_M,
V2_QPC_BYTE_56_LP_PKTN_INI_S, 0);
- if (ibqp->qp_type == IB_QPT_GSI || ibqp->qp_type == IB_QPT_UD)
- roce_set_field(context->byte_24_mtu_tc, V2_QPC_BYTE_24_MTU_M,
- V2_QPC_BYTE_24_MTU_S, IB_MTU_4096);
- else if (attr_mask & IB_QP_PATH_MTU)
- roce_set_field(context->byte_24_mtu_tc, V2_QPC_BYTE_24_MTU_M,
- V2_QPC_BYTE_24_MTU_S, attr->path_mtu);
-
- roce_set_field(qpc_mask->byte_24_mtu_tc, V2_QPC_BYTE_24_MTU_M,
- V2_QPC_BYTE_24_MTU_S, 0);
-
roce_set_bit(qpc_mask->byte_108_rx_reqepsn,
V2_QPC_BYTE_108_RX_REQ_PSN_ERR_S, 0);
roce_set_field(qpc_mask->byte_96_rx_reqmsn, V2_QPC_BYTE_96_RX_REQ_MSN_M,
diff --git a/drivers/infiniband/hw/hns/hns_roce_mr.c b/drivers/infiniband/hw/hns/hns_roce_mr.c
index 0e71ebee9e52..6b226a5eb7db 100644
--- a/drivers/infiniband/hw/hns/hns_roce_mr.c
+++ b/drivers/infiniband/hw/hns/hns_roce_mr.c
@@ -120,7 +120,7 @@ static int alloc_mr_pbl(struct hns_roce_dev *hr_dev, struct hns_roce_mr *mr,
mr->pbl_hop_num = is_fast ? 1 : hr_dev->caps.pbl_hop_num;
buf_attr.page_shift = is_fast ? PAGE_SHIFT :
- hr_dev->caps.pbl_buf_pg_sz + HNS_HW_PAGE_SHIFT;
+ hr_dev->caps.pbl_buf_pg_sz + PAGE_SHIFT;
buf_attr.region[0].size = length;
buf_attr.region[0].hopnum = mr->pbl_hop_num;
buf_attr.region_count = 1;
diff --git a/drivers/infiniband/hw/mlx5/odp.c b/drivers/infiniband/hw/mlx5/odp.c
index 7d2ec9ee5097..77dca1e05bba 100644
--- a/drivers/infiniband/hw/mlx5/odp.c
+++ b/drivers/infiniband/hw/mlx5/odp.c
@@ -601,6 +601,23 @@ void mlx5_ib_free_implicit_mr(struct mlx5_ib_mr *imr)
*/
synchronize_srcu(&dev->odp_srcu);
+ /*
+ * All work on the prefetch list must be completed, xa_erase() prevented
+ * new work from being created.
+ */
+ wait_event(imr->q_deferred_work, !atomic_read(&imr->num_deferred_work));
+
+ /*
+ * At this point it is forbidden for any other thread to enter
+ * pagefault_mr() on this imr. It is already forbidden to call
+ * pagefault_mr() on an implicit child. Due to this additions to
+ * implicit_children are prevented.
+ */
+
+ /*
+ * Block destroy_unused_implicit_child_mr() from incrementing
+ * num_deferred_work.
+ */
xa_lock(&imr->implicit_children);
xa_for_each (&imr->implicit_children, idx, mtt) {
__xa_erase(&imr->implicit_children, idx);
@@ -609,9 +626,8 @@ void mlx5_ib_free_implicit_mr(struct mlx5_ib_mr *imr)
xa_unlock(&imr->implicit_children);
/*
- * num_deferred_work can only be incremented inside the odp_srcu, or
- * under xa_lock while the child is in the xarray. Thus at this point
- * it is only decreasing, and all work holding it is now on the wq.
+ * Wait for any concurrent destroy_unused_implicit_child_mr() to
+ * complete.
*/
wait_event(imr->q_deferred_work, !atomic_read(&imr->num_deferred_work));
@@ -1781,9 +1797,7 @@ static bool init_prefetch_work(struct ib_pd *pd,
work->frags[i].mr =
get_prefetchable_mr(pd, advice, sg_list[i].lkey);
if (!work->frags[i].mr) {
- work->num_sge = i - 1;
- if (i)
- destroy_prefetch_work(work);
+ work->num_sge = i;
return false;
}
@@ -1849,6 +1863,7 @@ int mlx5_ib_advise_mr_prefetch(struct ib_pd *pd,
srcu_key = srcu_read_lock(&dev->odp_srcu);
if (!init_prefetch_work(pd, advice, pf_flags, work, sg_list, num_sge)) {
srcu_read_unlock(&dev->odp_srcu, srcu_key);
+ destroy_prefetch_work(work);
return -EINVAL;
}
queue_work(system_unbound_wq, &work->work);
diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c
index e050eade97a1..1225b8d77510 100644
--- a/drivers/infiniband/hw/mlx5/qp.c
+++ b/drivers/infiniband/hw/mlx5/qp.c
@@ -1766,15 +1766,14 @@ err:
}
static void configure_requester_scat_cqe(struct mlx5_ib_dev *dev,
+ struct mlx5_ib_qp *qp,
struct ib_qp_init_attr *init_attr,
- struct mlx5_ib_create_qp *ucmd,
void *qpc)
{
int scqe_sz;
bool allow_scat_cqe = false;
- if (ucmd)
- allow_scat_cqe = ucmd->flags & MLX5_QP_FLAG_ALLOW_SCATTER_CQE;
+ allow_scat_cqe = qp->flags_en & MLX5_QP_FLAG_ALLOW_SCATTER_CQE;
if (!allow_scat_cqe && init_attr->sq_sig_type != IB_SIGNAL_ALL_WR)
return;
@@ -1853,8 +1852,6 @@ static int create_xrc_tgt_qp(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
u32 *in;
int err;
- mutex_init(&qp->mutex);
-
if (attr->sq_sig_type == IB_SIGNAL_ALL_WR)
qp->sq_signal_bits = MLX5_WQE_CTRL_CQ_UPDATE;
@@ -1938,7 +1935,6 @@ static int create_user_qp(struct mlx5_ib_dev *dev, struct ib_pd *pd,
u32 *in;
int err;
- mutex_init(&qp->mutex);
spin_lock_init(&qp->sq.lock);
spin_lock_init(&qp->rq.lock);
@@ -2012,7 +2008,7 @@ static int create_user_qp(struct mlx5_ib_dev *dev, struct ib_pd *pd,
}
if ((qp->flags_en & MLX5_QP_FLAG_SCATTER_CQE) &&
(qp->type == MLX5_IB_QPT_DCI || qp->type == IB_QPT_RC))
- configure_requester_scat_cqe(dev, init_attr, ucmd, qpc);
+ configure_requester_scat_cqe(dev, qp, init_attr, qpc);
if (qp->rq.wqe_cnt) {
MLX5_SET(qpc, qpc, log_rq_stride, qp->rq.wqe_shift - 4);
@@ -2129,7 +2125,6 @@ static int create_kernel_qp(struct mlx5_ib_dev *dev, struct ib_pd *pd,
u32 *in;
int err;
- mutex_init(&qp->mutex);
spin_lock_init(&qp->sq.lock);
spin_lock_init(&qp->rq.lock);
@@ -2543,13 +2538,18 @@ static void process_vendor_flag(struct mlx5_ib_dev *dev, int *flags, int flag,
return;
}
- if (flag == MLX5_QP_FLAG_SCATTER_CQE) {
+ switch (flag) {
+ case MLX5_QP_FLAG_SCATTER_CQE:
+ case MLX5_QP_FLAG_ALLOW_SCATTER_CQE:
/*
- * We don't return error if this flag was provided,
- * and mlx5 doesn't have right capability.
- */
- *flags &= ~MLX5_QP_FLAG_SCATTER_CQE;
+ * We don't return error if these flags were provided,
+ * and mlx5 doesn't have right capability.
+ */
+ *flags &= ~(MLX5_QP_FLAG_SCATTER_CQE |
+ MLX5_QP_FLAG_ALLOW_SCATTER_CQE);
return;
+ default:
+ break;
}
mlx5_ib_dbg(dev, "Vendor create QP flag 0x%X is not supported\n", flag);
}
@@ -2589,6 +2589,8 @@ static int process_vendor_flags(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
process_vendor_flag(dev, &flags, MLX5_QP_FLAG_SIGNATURE, true, qp);
process_vendor_flag(dev, &flags, MLX5_QP_FLAG_SCATTER_CQE,
MLX5_CAP_GEN(mdev, sctr_data_cqe), qp);
+ process_vendor_flag(dev, &flags, MLX5_QP_FLAG_ALLOW_SCATTER_CQE,
+ MLX5_CAP_GEN(mdev, sctr_data_cqe), qp);
if (qp->type == IB_QPT_RAW_PACKET) {
cond = MLX5_CAP_ETH(mdev, tunnel_stateless_vxlan) ||
@@ -2963,6 +2965,7 @@ struct ib_qp *mlx5_ib_create_qp(struct ib_pd *pd, struct ib_qp_init_attr *attr,
goto free_ucmd;
}
+ mutex_init(&qp->mutex);
qp->type = type;
if (udata) {
err = process_vendor_flags(dev, qp, params.ucmd, attr);
diff --git a/drivers/infiniband/hw/mlx5/srq_cmd.c b/drivers/infiniband/hw/mlx5/srq_cmd.c
index 6f5eadc4d183..37aaacebd3f2 100644
--- a/drivers/infiniband/hw/mlx5/srq_cmd.c
+++ b/drivers/infiniband/hw/mlx5/srq_cmd.c
@@ -83,11 +83,11 @@ struct mlx5_core_srq *mlx5_cmd_get_srq(struct mlx5_ib_dev *dev, u32 srqn)
struct mlx5_srq_table *table = &dev->srq_table;
struct mlx5_core_srq *srq;
- xa_lock(&table->array);
+ xa_lock_irq(&table->array);
srq = xa_load(&table->array, srqn);
if (srq)
refcount_inc(&srq->common.refcount);
- xa_unlock(&table->array);
+ xa_unlock_irq(&table->array);
return srq;
}
diff --git a/drivers/infiniband/sw/rdmavt/qp.c b/drivers/infiniband/sw/rdmavt/qp.c
index 7db35dd6ad74..332a8ba94b81 100644
--- a/drivers/infiniband/sw/rdmavt/qp.c
+++ b/drivers/infiniband/sw/rdmavt/qp.c
@@ -901,8 +901,6 @@ static void rvt_init_qp(struct rvt_dev_info *rdi, struct rvt_qp *qp,
qp->s_tail_ack_queue = 0;
qp->s_acked_ack_queue = 0;
qp->s_num_rd_atomic = 0;
- if (qp->r_rq.kwq)
- qp->r_rq.kwq->count = qp->r_rq.size;
qp->r_sge.num_sge = 0;
atomic_set(&qp->s_reserved_used, 0);
}
@@ -2367,31 +2365,6 @@ bad_lkey:
}
/**
- * get_count - count numbers of request work queue entries
- * in circular buffer
- * @rq: data structure for request queue entry
- * @tail: tail indices of the circular buffer
- * @head: head indices of the circular buffer
- *
- * Return - total number of entries in the circular buffer
- */
-static u32 get_count(struct rvt_rq *rq, u32 tail, u32 head)
-{
- u32 count;
-
- count = head;
-
- if (count >= rq->size)
- count = 0;
- if (count < tail)
- count += rq->size - tail;
- else
- count -= tail;
-
- return count;
-}
-
-/**
* get_rvt_head - get head indices of the circular buffer
* @rq: data structure for request queue entry
* @ip: the QP
@@ -2465,7 +2438,7 @@ int rvt_get_rwqe(struct rvt_qp *qp, bool wr_id_only)
if (kwq->count < RVT_RWQ_COUNT_THRESHOLD) {
head = get_rvt_head(rq, ip);
- kwq->count = get_count(rq, tail, head);
+ kwq->count = rvt_get_rq_count(rq, head, tail);
}
if (unlikely(kwq->count == 0)) {
ret = 0;
@@ -2500,7 +2473,9 @@ int rvt_get_rwqe(struct rvt_qp *qp, bool wr_id_only)
* the number of remaining WQEs.
*/
if (kwq->count < srq->limit) {
- kwq->count = get_count(rq, tail, get_rvt_head(rq, ip));
+ kwq->count =
+ rvt_get_rq_count(rq,
+ get_rvt_head(rq, ip), tail);
if (kwq->count < srq->limit) {
struct ib_event ev;
diff --git a/drivers/infiniband/sw/rdmavt/rc.c b/drivers/infiniband/sw/rdmavt/rc.c
index 977906cc0d11..c58735f4c94a 100644
--- a/drivers/infiniband/sw/rdmavt/rc.c
+++ b/drivers/infiniband/sw/rdmavt/rc.c
@@ -127,9 +127,7 @@ __be32 rvt_compute_aeth(struct rvt_qp *qp)
* not atomic, which is OK, since the fuzziness is
* resolved as further ACKs go out.
*/
- credits = head - tail;
- if ((int)credits < 0)
- credits += qp->r_rq.size;
+ credits = rvt_get_rq_count(&qp->r_rq, head, tail);
}
/*
* Binary search the credit table to find the code to
generated by cgit v1.2.3 (git 2.25.1) at 2024-08-08 23:25:00 +0000