summaryrefslogtreecommitdiff
path: root/drivers/infiniband
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/infiniband')
-rw-r--r--drivers/infiniband/core/addr.c40
-rw-r--r--drivers/infiniband/core/cma.c15
-rw-r--r--drivers/infiniband/core/cq.c21
-rw-r--r--drivers/infiniband/core/device.c9
-rw-r--r--drivers/infiniband/core/sa_query.c7
-rw-r--r--drivers/infiniband/core/ucma.c73
-rw-r--r--drivers/infiniband/hw/bnxt_re/bnxt_re.h4
-rw-r--r--drivers/infiniband/hw/bnxt_re/ib_verbs.c28
-rw-r--r--drivers/infiniband/hw/bnxt_re/ib_verbs.h3
-rw-r--r--drivers/infiniband/hw/bnxt_re/main.c12
-rw-r--r--drivers/infiniband/hw/bnxt_re/qplib_fp.c113
-rw-r--r--drivers/infiniband/hw/bnxt_re/qplib_fp.h12
-rw-r--r--drivers/infiniband/hw/bnxt_re/qplib_rcfw.c9
-rw-r--r--drivers/infiniband/hw/bnxt_re/qplib_rcfw.h1
-rw-r--r--drivers/infiniband/hw/bnxt_re/qplib_sp.c3
-rw-r--r--drivers/infiniband/hw/bnxt_re/roce_hsi.h25
-rw-r--r--drivers/infiniband/hw/hns/hns_roce_hw_v2.c2
-rw-r--r--drivers/infiniband/hw/mlx4/cq.c4
-rw-r--r--drivers/infiniband/hw/mlx4/main.c11
-rw-r--r--drivers/infiniband/hw/mlx5/cq.c10
-rw-r--r--drivers/infiniband/hw/mlx5/main.c54
-rw-r--r--drivers/infiniband/hw/mlx5/mlx5_ib.h3
-rw-r--r--drivers/infiniband/hw/mlx5/mr.c17
-rw-r--r--drivers/infiniband/hw/mlx5/qp.c34
-rw-r--r--drivers/infiniband/hw/mlx5/srq.c15
-rw-r--r--drivers/infiniband/hw/qedr/main.c3
-rw-r--r--drivers/infiniband/hw/qedr/qedr_iw_cm.c19
-rw-r--r--drivers/infiniband/hw/qedr/verbs.c58
-rw-r--r--drivers/infiniband/sw/rdmavt/mr.c10
29 files changed, 377 insertions, 238 deletions
diff --git a/drivers/infiniband/core/addr.c b/drivers/infiniband/core/addr.c
index a5b4cf030c11..cb1d2ab13c66 100644
--- a/drivers/infiniband/core/addr.c
+++ b/drivers/infiniband/core/addr.c
@@ -207,6 +207,22 @@ int rdma_addr_size(struct sockaddr *addr)
}
EXPORT_SYMBOL(rdma_addr_size);
+int rdma_addr_size_in6(struct sockaddr_in6 *addr)
+{
+ int ret = rdma_addr_size((struct sockaddr *) addr);
+
+ return ret <= sizeof(*addr) ? ret : 0;
+}
+EXPORT_SYMBOL(rdma_addr_size_in6);
+
+int rdma_addr_size_kss(struct __kernel_sockaddr_storage *addr)
+{
+ int ret = rdma_addr_size((struct sockaddr *) addr);
+
+ return ret <= sizeof(*addr) ? ret : 0;
+}
+EXPORT_SYMBOL(rdma_addr_size_kss);
+
static struct rdma_addr_client self;
void rdma_addr_register_client(struct rdma_addr_client *client)
@@ -550,18 +566,13 @@ static int addr_resolve(struct sockaddr *src_in,
dst_release(dst);
}
- if (ndev->flags & IFF_LOOPBACK) {
- ret = rdma_translate_ip(dst_in, addr);
- /*
- * Put the loopback device and get the translated
- * device instead.
- */
+ if (ndev) {
+ if (ndev->flags & IFF_LOOPBACK)
+ ret = rdma_translate_ip(dst_in, addr);
+ else
+ addr->bound_dev_if = ndev->ifindex;
dev_put(ndev);
- ndev = dev_get_by_index(addr->net, addr->bound_dev_if);
- } else {
- addr->bound_dev_if = ndev->ifindex;
}
- dev_put(ndev);
return ret;
}
@@ -591,6 +602,15 @@ static void process_one_req(struct work_struct *_work)
list_del(&req->list);
mutex_unlock(&lock);
+ /*
+ * Although the work will normally have been canceled by the
+ * workqueue, it can still be requeued as long as it is on the
+ * req_list, so it could have been requeued before we grabbed &lock.
+ * We need to cancel it after it is removed from req_list to really be
+ * sure it is safe to free.
+ */
+ cancel_delayed_work(&req->work);
+
req->callback(req->status, (struct sockaddr *)&req->src_addr,
req->addr, req->context);
put_client(req->client);
diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c
index e66963ca58bd..a5367c5efbe7 100644
--- a/drivers/infiniband/core/cma.c
+++ b/drivers/infiniband/core/cma.c
@@ -3069,7 +3069,8 @@ static int cma_port_is_unique(struct rdma_bind_list *bind_list,
continue;
/* different dest port -> unique */
- if (!cma_any_port(cur_daddr) &&
+ if (!cma_any_port(daddr) &&
+ !cma_any_port(cur_daddr) &&
(dport != cur_dport))
continue;
@@ -3080,7 +3081,8 @@ static int cma_port_is_unique(struct rdma_bind_list *bind_list,
continue;
/* different dst address -> unique */
- if (!cma_any_addr(cur_daddr) &&
+ if (!cma_any_addr(daddr) &&
+ !cma_any_addr(cur_daddr) &&
cma_addr_cmp(daddr, cur_daddr))
continue;
@@ -3378,13 +3380,13 @@ int rdma_bind_addr(struct rdma_cm_id *id, struct sockaddr *addr)
}
#endif
}
+ daddr = cma_dst_addr(id_priv);
+ daddr->sa_family = addr->sa_family;
+
ret = cma_get_port(id_priv);
if (ret)
goto err2;
- daddr = cma_dst_addr(id_priv);
- daddr->sa_family = addr->sa_family;
-
return 0;
err2:
if (id_priv->cma_dev)
@@ -4173,6 +4175,9 @@ int rdma_join_multicast(struct rdma_cm_id *id, struct sockaddr *addr,
struct cma_multicast *mc;
int ret;
+ if (!id->device)
+ return -EINVAL;
+
id_priv = container_of(id, struct rdma_id_private, id);
if (!cma_comp(id_priv, RDMA_CM_ADDR_BOUND) &&
!cma_comp(id_priv, RDMA_CM_ADDR_RESOLVED))
diff --git a/drivers/infiniband/core/cq.c b/drivers/infiniband/core/cq.c
index bc79ca8215d7..af5ad6a56ae4 100644
--- a/drivers/infiniband/core/cq.c
+++ b/drivers/infiniband/core/cq.c
@@ -17,6 +17,7 @@
/* # of WCs to poll for with a single call to ib_poll_cq */
#define IB_POLL_BATCH 16
+#define IB_POLL_BATCH_DIRECT 8
/* # of WCs to iterate over before yielding */
#define IB_POLL_BUDGET_IRQ 256
@@ -25,18 +26,18 @@
#define IB_POLL_FLAGS \
(IB_CQ_NEXT_COMP | IB_CQ_REPORT_MISSED_EVENTS)
-static int __ib_process_cq(struct ib_cq *cq, int budget, struct ib_wc *poll_wc)
+static int __ib_process_cq(struct ib_cq *cq, int budget, struct ib_wc *wcs,
+ int batch)
{
int i, n, completed = 0;
- struct ib_wc *wcs = poll_wc ? : cq->wc;
/*
* budget might be (-1) if the caller does not
* want to bound this call, thus we need unsigned
* minimum here.
*/
- while ((n = ib_poll_cq(cq, min_t(u32, IB_POLL_BATCH,
- budget - completed), wcs)) > 0) {
+ while ((n = ib_poll_cq(cq, min_t(u32, batch,
+ budget - completed), wcs)) > 0) {
for (i = 0; i < n; i++) {
struct ib_wc *wc = &wcs[i];
@@ -48,8 +49,7 @@ static int __ib_process_cq(struct ib_cq *cq, int budget, struct ib_wc *poll_wc)
completed += n;
- if (n != IB_POLL_BATCH ||
- (budget != -1 && completed >= budget))
+ if (n != batch || (budget != -1 && completed >= budget))
break;
}
@@ -72,9 +72,9 @@ static int __ib_process_cq(struct ib_cq *cq, int budget, struct ib_wc *poll_wc)
*/
int ib_process_cq_direct(struct ib_cq *cq, int budget)
{
- struct ib_wc wcs[IB_POLL_BATCH];
+ struct ib_wc wcs[IB_POLL_BATCH_DIRECT];
- return __ib_process_cq(cq, budget, wcs);
+ return __ib_process_cq(cq, budget, wcs, IB_POLL_BATCH_DIRECT);
}
EXPORT_SYMBOL(ib_process_cq_direct);
@@ -88,7 +88,7 @@ static int ib_poll_handler(struct irq_poll *iop, int budget)
struct ib_cq *cq = container_of(iop, struct ib_cq, iop);
int completed;
- completed = __ib_process_cq(cq, budget, NULL);
+ completed = __ib_process_cq(cq, budget, cq->wc, IB_POLL_BATCH);
if (completed < budget) {
irq_poll_complete(&cq->iop);
if (ib_req_notify_cq(cq, IB_POLL_FLAGS) > 0)
@@ -108,7 +108,8 @@ static void ib_cq_poll_work(struct work_struct *work)
struct ib_cq *cq = container_of(work, struct ib_cq, work);
int completed;
- completed = __ib_process_cq(cq, IB_POLL_BUDGET_WORKQUEUE, NULL);
+ completed = __ib_process_cq(cq, IB_POLL_BUDGET_WORKQUEUE, cq->wc,
+ IB_POLL_BATCH);
if (completed >= IB_POLL_BUDGET_WORKQUEUE ||
ib_req_notify_cq(cq, IB_POLL_FLAGS) > 0)
queue_work(ib_comp_wq, &cq->work);
diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c
index e8010e73a1cf..b7459cf524e4 100644
--- a/drivers/infiniband/core/device.c
+++ b/drivers/infiniband/core/device.c
@@ -290,6 +290,7 @@ void ib_dealloc_device(struct ib_device *device)
{
WARN_ON(device->reg_state != IB_DEV_UNREGISTERED &&
device->reg_state != IB_DEV_UNINITIALIZED);
+ rdma_restrack_clean(&device->res);
put_device(&device->dev);
}
EXPORT_SYMBOL(ib_dealloc_device);
@@ -536,14 +537,14 @@ int ib_register_device(struct ib_device *device,
ret = device->query_device(device, &device->attrs, &uhw);
if (ret) {
pr_warn("Couldn't query the device attributes\n");
- goto cache_cleanup;
+ goto cg_cleanup;
}
ret = ib_device_register_sysfs(device, port_callback);
if (ret) {
pr_warn("Couldn't register device %s with driver model\n",
device->name);
- goto cache_cleanup;
+ goto cg_cleanup;
}
device->reg_state = IB_DEV_REGISTERED;
@@ -559,6 +560,8 @@ int ib_register_device(struct ib_device *device,
mutex_unlock(&device_mutex);
return 0;
+cg_cleanup:
+ ib_device_unregister_rdmacg(device);
cache_cleanup:
ib_cache_cleanup_one(device);
ib_cache_release_one(device);
@@ -598,8 +601,6 @@ void ib_unregister_device(struct ib_device *device)
}
up_read(&lists_rwsem);
- rdma_restrack_clean(&device->res);
-
ib_device_unregister_rdmacg(device);
ib_device_unregister_sysfs(device);
diff --git a/drivers/infiniband/core/sa_query.c b/drivers/infiniband/core/sa_query.c
index 8cf15d4a8ac4..9f029a1ca5ea 100644
--- a/drivers/infiniband/core/sa_query.c
+++ b/drivers/infiniband/core/sa_query.c
@@ -1291,10 +1291,9 @@ int ib_init_ah_attr_from_path(struct ib_device *device, u8 port_num,
resolved_dev = dev_get_by_index(dev_addr.net,
dev_addr.bound_dev_if);
- if (resolved_dev->flags & IFF_LOOPBACK) {
- dev_put(resolved_dev);
- resolved_dev = idev;
- dev_hold(resolved_dev);
+ if (!resolved_dev) {
+ dev_put(idev);
+ return -ENODEV;
}
ndev = ib_get_ndev_from_path(rec);
rcu_read_lock();
diff --git a/drivers/infiniband/core/ucma.c b/drivers/infiniband/core/ucma.c
index f015f1bf88c9..d933336d7e01 100644
--- a/drivers/infiniband/core/ucma.c
+++ b/drivers/infiniband/core/ucma.c
@@ -132,7 +132,7 @@ static inline struct ucma_context *_ucma_find_context(int id,
ctx = idr_find(&ctx_idr, id);
if (!ctx)
ctx = ERR_PTR(-ENOENT);
- else if (ctx->file != file)
+ else if (ctx->file != file || !ctx->cm_id)
ctx = ERR_PTR(-EINVAL);
return ctx;
}
@@ -456,6 +456,7 @@ static ssize_t ucma_create_id(struct ucma_file *file, const char __user *inbuf,
struct rdma_ucm_create_id cmd;
struct rdma_ucm_create_id_resp resp;
struct ucma_context *ctx;
+ struct rdma_cm_id *cm_id;
enum ib_qp_type qp_type;
int ret;
@@ -476,10 +477,10 @@ static ssize_t ucma_create_id(struct ucma_file *file, const char __user *inbuf,
return -ENOMEM;
ctx->uid = cmd.uid;
- ctx->cm_id = rdma_create_id(current->nsproxy->net_ns,
- ucma_event_handler, ctx, cmd.ps, qp_type);
- if (IS_ERR(ctx->cm_id)) {
- ret = PTR_ERR(ctx->cm_id);
+ cm_id = rdma_create_id(current->nsproxy->net_ns,
+ ucma_event_handler, ctx, cmd.ps, qp_type);
+ if (IS_ERR(cm_id)) {
+ ret = PTR_ERR(cm_id);
goto err1;
}
@@ -489,14 +490,19 @@ static ssize_t ucma_create_id(struct ucma_file *file, const char __user *inbuf,
ret = -EFAULT;
goto err2;
}
+
+ ctx->cm_id = cm_id;
return 0;
err2:
- rdma_destroy_id(ctx->cm_id);
+ rdma_destroy_id(cm_id);
err1:
mutex_lock(&mut);
idr_remove(&ctx_idr, ctx->id);
mutex_unlock(&mut);
+ mutex_lock(&file->mut);
+ list_del(&ctx->list);
+ mutex_unlock(&file->mut);
kfree(ctx);
return ret;
}
@@ -626,6 +632,9 @@ static ssize_t ucma_bind_ip(struct ucma_file *file, const char __user *inbuf,
if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
return -EFAULT;
+ if (!rdma_addr_size_in6(&cmd.addr))
+ return -EINVAL;
+
ctx = ucma_get_ctx(file, cmd.id);
if (IS_ERR(ctx))
return PTR_ERR(ctx);
@@ -639,22 +648,21 @@ static ssize_t ucma_bind(struct ucma_file *file, const char __user *inbuf,
int in_len, int out_len)
{
struct rdma_ucm_bind cmd;
- struct sockaddr *addr;
struct ucma_context *ctx;
int ret;
if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
return -EFAULT;
- addr = (struct sockaddr *) &cmd.addr;
- if (cmd.reserved || !cmd.addr_size || (cmd.addr_size != rdma_addr_size(addr)))
+ if (cmd.reserved || !cmd.addr_size ||
+ cmd.addr_size != rdma_addr_size_kss(&cmd.addr))
return -EINVAL;
ctx = ucma_get_ctx(file, cmd.id);
if (IS_ERR(ctx))
return PTR_ERR(ctx);
- ret = rdma_bind_addr(ctx->cm_id, addr);
+ ret = rdma_bind_addr(ctx->cm_id, (struct sockaddr *) &cmd.addr);
ucma_put_ctx(ctx);
return ret;
}
@@ -670,13 +678,16 @@ static ssize_t ucma_resolve_ip(struct ucma_file *file,
if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
return -EFAULT;
+ if (!rdma_addr_size_in6(&cmd.src_addr) ||
+ !rdma_addr_size_in6(&cmd.dst_addr))
+ return -EINVAL;
+
ctx = ucma_get_ctx(file, cmd.id);
if (IS_ERR(ctx))
return PTR_ERR(ctx);
ret = rdma_resolve_addr(ctx->cm_id, (struct sockaddr *) &cmd.src_addr,
- (struct sockaddr *) &cmd.dst_addr,
- cmd.timeout_ms);
+ (struct sockaddr *) &cmd.dst_addr, cmd.timeout_ms);
ucma_put_ctx(ctx);
return ret;
}
@@ -686,24 +697,23 @@ static ssize_t ucma_resolve_addr(struct ucma_file *file,
int in_len, int out_len)
{
struct rdma_ucm_resolve_addr cmd;
- struct sockaddr *src, *dst;
struct ucma_context *ctx;
int ret;
if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
return -EFAULT;
- src = (struct sockaddr *) &cmd.src_addr;
- dst = (struct sockaddr *) &cmd.dst_addr;
- if (cmd.reserved || (cmd.src_size && (cmd.src_size != rdma_addr_size(src))) ||
- !cmd.dst_size || (cmd.dst_size != rdma_addr_size(dst)))
+ if (cmd.reserved ||
+ (cmd.src_size && (cmd.src_size != rdma_addr_size_kss(&cmd.src_addr))) ||
+ !cmd.dst_size || (cmd.dst_size != rdma_addr_size_kss(&cmd.dst_addr)))
return -EINVAL;
ctx = ucma_get_ctx(file, cmd.id);
if (IS_ERR(ctx))
return PTR_ERR(ctx);
- ret = rdma_resolve_addr(ctx->cm_id, src, dst, cmd.timeout_ms);
+ ret = rdma_resolve_addr(ctx->cm_id, (struct sockaddr *) &cmd.src_addr,
+ (struct sockaddr *) &cmd.dst_addr, cmd.timeout_ms);
ucma_put_ctx(ctx);
return ret;
}
@@ -1149,10 +1159,18 @@ static ssize_t ucma_init_qp_attr(struct ucma_file *file,
if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
return -EFAULT;
+ if (cmd.qp_state > IB_QPS_ERR)
+ return -EINVAL;
+
ctx = ucma_get_ctx(file, cmd.id);
if (IS_ERR(ctx))
return PTR_ERR(ctx);
+ if (!ctx->cm_id->device) {
+ ret = -EINVAL;
+ goto out;
+ }
+
resp.qp_attr_mask = 0;
memset(&qp_attr, 0, sizeof qp_attr);
qp_attr.qp_state = cmd.qp_state;
@@ -1294,6 +1312,9 @@ static ssize_t ucma_set_option(struct ucma_file *file, const char __user *inbuf,
if (IS_ERR(ctx))
return PTR_ERR(ctx);
+ if (unlikely(cmd.optlen > KMALLOC_MAX_SIZE))
+ return -EINVAL;
+
optval = memdup_user((void __user *) (unsigned long) cmd.optval,
cmd.optlen);
if (IS_ERR(optval)) {
@@ -1315,7 +1336,7 @@ static ssize_t ucma_notify(struct ucma_file *file, const char __user *inbuf,
{
struct rdma_ucm_notify cmd;
struct ucma_context *ctx;
- int ret;
+ int ret = -EINVAL;
if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
return -EFAULT;
@@ -1324,7 +1345,9 @@ static ssize_t ucma_notify(struct ucma_file *file, const char __user *inbuf,
if (IS_ERR(ctx))
return PTR_ERR(ctx);
- ret = rdma_notify(ctx->cm_id, (enum ib_event_type) cmd.event);
+ if (ctx->cm_id->device)
+ ret = rdma_notify(ctx->cm_id, (enum ib_event_type)cmd.event);
+
ucma_put_ctx(ctx);
return ret;
}
@@ -1343,7 +1366,7 @@ static ssize_t ucma_process_join(struct ucma_file *file,
return -ENOSPC;
addr = (struct sockaddr *) &cmd->addr;
- if (!cmd->addr_size || (cmd->addr_size != rdma_addr_size(addr)))
+ if (cmd->addr_size != rdma_addr_size(addr))
return -EINVAL;
if (cmd->join_flags == RDMA_MC_JOIN_FLAG_FULLMEMBER)
@@ -1410,7 +1433,10 @@ static ssize_t ucma_join_ip_multicast(struct ucma_file *file,
join_cmd.response = cmd.response;
join_cmd.uid = cmd.uid;
join_cmd.id = cmd.id;
- join_cmd.addr_size = rdma_addr_size((struct sockaddr *) &cmd.addr);
+ join_cmd.addr_size = rdma_addr_size_in6(&cmd.addr);
+ if (!join_cmd.addr_size)
+ return -EINVAL;
+
join_cmd.join_flags = RDMA_MC_JOIN_FLAG_FULLMEMBER;
memcpy(&join_cmd.addr, &cmd.addr, join_cmd.addr_size);
@@ -1426,6 +1452,9 @@ static ssize_t ucma_join_multicast(struct ucma_file *file,
if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
return -EFAULT;
+ if (!rdma_addr_size_kss(&cmd.addr))
+ return -EINVAL;
+
return ucma_process_join(file, &cmd, out_len);
}
diff --git a/drivers/infiniband/hw/bnxt_re/bnxt_re.h b/drivers/infiniband/hw/bnxt_re/bnxt_re.h
index 3eb7a8387116..96f76896488d 100644
--- a/drivers/infiniband/hw/bnxt_re/bnxt_re.h
+++ b/drivers/infiniband/hw/bnxt_re/bnxt_re.h
@@ -57,8 +57,8 @@
#define BNXT_RE_PAGE_SIZE_8M BIT(BNXT_RE_PAGE_SHIFT_8M)
#define BNXT_RE_PAGE_SIZE_1G BIT(BNXT_RE_PAGE_SHIFT_1G)
-#define BNXT_RE_MAX_MR_SIZE_LOW BIT(BNXT_RE_PAGE_SHIFT_1G)
-#define BNXT_RE_MAX_MR_SIZE_HIGH BIT(39)
+#define BNXT_RE_MAX_MR_SIZE_LOW BIT_ULL(BNXT_RE_PAGE_SHIFT_1G)
+#define BNXT_RE_MAX_MR_SIZE_HIGH BIT_ULL(39)
#define BNXT_RE_MAX_MR_SIZE BNXT_RE_MAX_MR_SIZE_HIGH
#define BNXT_RE_MAX_QPC_COUNT (64 * 1024)
diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.c b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
index 643174d949a8..8301d7e5fa8c 100644
--- a/drivers/infiniband/hw/bnxt_re/ib_verbs.c
+++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
@@ -785,7 +785,7 @@ int bnxt_re_query_ah(struct ib_ah *ib_ah, struct rdma_ah_attr *ah_attr)
return 0;
}
-static unsigned long bnxt_re_lock_cqs(struct bnxt_re_qp *qp)
+unsigned long bnxt_re_lock_cqs(struct bnxt_re_qp *qp)
__acquires(&qp->scq->cq_lock) __acquires(&qp->rcq->cq_lock)
{
unsigned long flags;
@@ -799,8 +799,8 @@ static unsigned long bnxt_re_lock_cqs(struct bnxt_re_qp *qp)
return flags;
}
-static void bnxt_re_unlock_cqs(struct bnxt_re_qp *qp,
- unsigned long flags)
+void bnxt_re_unlock_cqs(struct bnxt_re_qp *qp,
+ unsigned long flags)
__releases(&qp->scq->cq_lock) __releases(&qp->rcq->cq_lock)
{
if (qp->rcq != qp->scq)
@@ -1606,6 +1606,7 @@ int bnxt_re_modify_qp(struct ib_qp *ib_qp, struct ib_qp_attr *qp_attr,
int status;
union ib_gid sgid;
struct ib_gid_attr sgid_attr;
+ unsigned int flags;
u8 nw_type;
qp->qplib_qp.modify_flags = 0;
@@ -1634,14 +1635,18 @@ int bnxt_re_modify_qp(struct ib_qp *ib_qp, struct ib_qp_attr *qp_attr,
dev_dbg(rdev_to_dev(rdev),
"Move QP = %p to flush list\n",
qp);
+ flags = bnxt_re_lock_cqs(qp);
bnxt_qplib_add_flush_qp(&qp->qplib_qp);
+ bnxt_re_unlock_cqs(qp, flags);
}
if (!qp->sumem &&
qp->qplib_qp.state == CMDQ_MODIFY_QP_NEW_STATE_RESET) {
dev_dbg(rdev_to_dev(rdev),
"Move QP = %p out of flush list\n",
qp);
+ flags = bnxt_re_lock_cqs(qp);
bnxt_qplib_clean_qp(&qp->qplib_qp);
+ bnxt_re_unlock_cqs(qp, flags);
}
}
if (qp_attr_mask & IB_QP_EN_SQD_ASYNC_NOTIFY) {
@@ -2227,10 +2232,13 @@ static int bnxt_re_build_inv_wqe(struct ib_send_wr *wr,
wqe->type = BNXT_QPLIB_SWQE_TYPE_LOCAL_INV;
wqe->local_inv.inv_l_key = wr->ex.invalidate_rkey;
+ /* Need unconditional fence for local invalidate
+ * opcode to work as expected.
+ */
+ wqe->flags |= BNXT_QPLIB_SWQE_FLAGS_UC_FENCE;
+
if (wr->send_flags & IB_SEND_SIGNALED)
wqe->flags |= BNXT_QPLIB_SWQE_FLAGS_SIGNAL_COMP;
- if (wr->send_flags & IB_SEND_FENCE)
- wqe->flags |= BNXT_QPLIB_SWQE_FLAGS_UC_FENCE;
if (wr->send_flags & IB_SEND_SOLICITED)
wqe->flags |= BNXT_QPLIB_SWQE_FLAGS_SOLICIT_EVENT;
@@ -2251,8 +2259,12 @@ static int bnxt_re_build_reg_wqe(struct ib_reg_wr *wr,
wqe->frmr.levels = qplib_frpl->hwq.level + 1;
wqe->type = BNXT_QPLIB_SWQE_TYPE_REG_MR;
- if (wr->wr.send_flags & IB_SEND_FENCE)
- wqe->flags |= BNXT_QPLIB_SWQE_FLAGS_UC_FENCE;
+ /* Need unconditional fence for reg_mr
+ * opcode to function as expected.
+ */
+
+ wqe->flags |= BNXT_QPLIB_SWQE_FLAGS_UC_FENCE;
+
if (wr->wr.send_flags & IB_SEND_SIGNALED)
wqe->flags |= BNXT_QPLIB_SWQE_FLAGS_SIGNAL_COMP;
@@ -3586,7 +3598,7 @@ struct ib_mr *bnxt_re_reg_user_mr(struct ib_pd *ib_pd, u64 start, u64 length,
int umem_pgs, page_shift, rc;
if (length > BNXT_RE_MAX_MR_SIZE) {
- dev_err(rdev_to_dev(rdev), "MR Size: %lld > Max supported:%ld\n",
+ dev_err(rdev_to_dev(rdev), "MR Size: %lld > Max supported:%lld\n",
length, BNXT_RE_MAX_MR_SIZE);
return ERR_PTR(-ENOMEM);
}
diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.h b/drivers/infiniband/hw/bnxt_re/ib_verbs.h
index b88a48d43a9d..e62b7c2c7da6 100644
--- a/drivers/infiniband/hw/bnxt_re/ib_verbs.h
+++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.h
@@ -222,4 +222,7 @@ struct ib_ucontext *bnxt_re_alloc_ucontext(struct ib_device *ibdev,
struct ib_udata *udata);
int bnxt_re_dealloc_ucontext(struct ib_ucontext *context);
int bnxt_re_mmap(struct ib_ucontext *context, struct vm_area_struct *vma);
+
+unsigned long bnxt_re_lock_cqs(struct bnxt_re_qp *qp);
+void bnxt_re_unlock_cqs(struct bnxt_re_qp *qp, unsigned long flags);
#endif /* __BNXT_RE_IB_VERBS_H__ */
diff --git a/drivers/infiniband/hw/bnxt_re/main.c b/drivers/infiniband/hw/bnxt_re/main.c
index 33a448036c2e..f6e361750466 100644
--- a/drivers/infiniband/hw/bnxt_re/main.c
+++ b/drivers/infiniband/hw/bnxt_re/main.c
@@ -730,6 +730,13 @@ static int bnxt_re_handle_qp_async_event(struct creq_qp_event *qp_event,
struct bnxt_re_qp *qp)
{
struct ib_event event;
+ unsigned int flags;
+
+ if (qp->qplib_qp.state == CMDQ_MODIFY_QP_NEW_STATE_ERR) {
+ flags = bnxt_re_lock_cqs(qp);
+ bnxt_qplib_add_flush_qp(&qp->qplib_qp);
+ bnxt_re_unlock_cqs(qp, flags);
+ }
memset(&event, 0, sizeof(event));
if (qp->qplib_qp.srq) {
@@ -1416,9 +1423,12 @@ static void bnxt_re_task(struct work_struct *work)
switch (re_work->event) {
case NETDEV_REGISTER:
rc = bnxt_re_ib_reg(rdev);
- if (rc)
+ if (rc) {
dev_err(rdev_to_dev(rdev),
"Failed to register with IB: %#x", rc);
+ bnxt_re_remove_one(rdev);
+ bnxt_re_dev_unreg(rdev);
+ }
break;
case NETDEV_UP:
bnxt_re_dispatch_event(&rdev->ibdev, NULL, 1,
diff --git a/drivers/infiniband/hw/bnxt_re/qplib_fp.c b/drivers/infiniband/hw/bnxt_re/qplib_fp.c
index 3ea5b9624f6b..3a78faba8d91 100644
--- a/drivers/infiniband/hw/bnxt_re/qplib_fp.c
+++ b/drivers/infiniband/hw/bnxt_re/qplib_fp.c
@@ -88,75 +88,35 @@ static void __bnxt_qplib_add_flush_qp(struct bnxt_qplib_qp *qp)
}
}
-void bnxt_qplib_acquire_cq_locks(struct bnxt_qplib_qp *qp,
- unsigned long *flags)
- __acquires(&qp->scq->hwq.lock) __acquires(&qp->rcq->hwq.lock)
+static void bnxt_qplib_acquire_cq_flush_locks(struct bnxt_qplib_qp *qp,
+ unsigned long *flags)
+ __acquires(&qp->scq->flush_lock) __acquires(&qp->rcq->flush_lock)
{
- spin_lock_irqsave(&qp->scq->hwq.lock, *flags);
+ spin_lock_irqsave(&qp->scq->flush_lock, *flags);
if (qp->scq == qp->rcq)
- __acquire(&qp->rcq->hwq.lock);
+ __acquire(&qp->rcq->flush_lock);
else
- spin_lock(&qp->rcq->hwq.lock);
+ spin_lock(&qp->rcq->flush_lock);
}
-void bnxt_qplib_release_cq_locks(struct bnxt_qplib_qp *qp,
- unsigned long *flags)
- __releases(&qp->scq->hwq.lock) __releases(&qp->rcq->hwq.lock)
+static void bnxt_qplib_release_cq_flush_locks(struct bnxt_qplib_qp *qp,
+ unsigned long *flags)
+ __releases(&qp->scq->flush_lock) __releases(&qp->rcq->flush_lock)
{
if (qp->scq == qp->rcq)
- __release(&qp->rcq->hwq.lock);
+ __release(&qp->rcq->flush_lock);
else
- spin_unlock(&qp->rcq->hwq.lock);
- spin_unlock_irqrestore(&qp->scq->hwq.lock, *flags);
-}
-
-static struct bnxt_qplib_cq *bnxt_qplib_find_buddy_cq(struct bnxt_qplib_qp *qp,
- struct bnxt_qplib_cq *cq)
-{
- struct bnxt_qplib_cq *buddy_cq = NULL;
-
- if (qp->scq == qp->rcq)
- buddy_cq = NULL;
- else if (qp->scq == cq)
- buddy_cq = qp->rcq;
- else
- buddy_cq = qp->scq;
- return buddy_cq;
-}
-
-static void bnxt_qplib_lock_buddy_cq(struct bnxt_qplib_qp *qp,
- struct bnxt_qplib_cq *cq)
- __acquires(&buddy_cq->hwq.lock)
-{
- struct bnxt_qplib_cq *buddy_cq = NULL;
-
- buddy_cq = bnxt_qplib_find_buddy_cq(qp, cq);
- if (!buddy_cq)
- __acquire(&cq->hwq.lock);
- else
- spin_lock(&buddy_cq->hwq.lock);
-}
-
-static void bnxt_qplib_unlock_buddy_cq(struct bnxt_qplib_qp *qp,
- struct bnxt_qplib_cq *cq)
- __releases(&buddy_cq->hwq.lock)
-{
- struct bnxt_qplib_cq *buddy_cq = NULL;
-
- buddy_cq = bnxt_qplib_find_buddy_cq(qp, cq);
- if (!buddy_cq)
- __release(&cq->hwq.lock);
- else
- spin_unlock(&buddy_cq->hwq.lock);
+ spin_unlock(&qp->rcq->flush_lock);
+ spin_unlock_irqrestore(&qp->scq->flush_lock, *flags);
}
void bnxt_qplib_add_flush_qp(struct bnxt_qplib_qp *qp)
{
unsigned long flags;
- bnxt_qplib_acquire_cq_locks(qp, &flags);
+ bnxt_qplib_acquire_cq_flush_locks(qp, &flags);
__bnxt_qplib_add_flush_qp(qp);
- bnxt_qplib_release_cq_locks(qp, &flags);
+ bnxt_qplib_release_cq_flush_locks(qp, &flags);
}
static void __bnxt_qplib_del_flush_qp(struct bnxt_qplib_qp *qp)
@@ -177,7 +137,7 @@ void bnxt_qplib_clean_qp(struct bnxt_qplib_qp *qp)
{
unsigned long flags;
- bnxt_qplib_acquire_cq_locks(qp, &flags);
+ bnxt_qplib_acquire_cq_flush_locks(qp, &flags);
__clean_cq(qp->scq, (u64)(unsigned long)qp);
qp->sq.hwq.prod = 0;
qp->sq.hwq.cons = 0;
@@ -186,7 +146,7 @@ void bnxt_qplib_clean_qp(struct bnxt_qplib_qp *qp)
qp->rq.hwq.cons = 0;
__bnxt_qplib_del_flush_qp(qp);
- bnxt_qplib_release_cq_locks(qp, &flags);
+ bnxt_qplib_release_cq_flush_locks(qp, &flags);
}
static void bnxt_qpn_cqn_sched_task(struct work_struct *work)
@@ -283,7 +243,7 @@ static void bnxt_qplib_service_nq(unsigned long data)
u32 sw_cons, raw_cons;
u16 type;
int budget = nq->budget;
- u64 q_handle;
+ uintptr_t q_handle;
/* Service the NQ until empty */
raw_cons = hwq->cons;
@@ -566,7 +526,7 @@ int bnxt_qplib_create_srq(struct bnxt_qplib_res *res,
/* Configure the request */
req.dpi = cpu_to_le32(srq->dpi->dpi);
- req.srq_handle = cpu_to_le64(srq);
+ req.srq_handle = cpu_to_le64((uintptr_t)srq);
req.srq_size = cpu_to_le16((u16)srq->hwq.max_elements);
pbl = &srq->hwq.pbl[PBL_LVL_0];
@@ -2107,9 +2067,6 @@ void bnxt_qplib_mark_qp_error(void *qp_handle)
/* Must block new posting of SQ and RQ */
qp->state = CMDQ_MODIFY_QP_NEW_STATE_ERR;
bnxt_qplib_cancel_phantom_processing(qp);
-
- /* Add qp to flush list of the CQ */
- __bnxt_qplib_add_flush_qp(qp);
}
/* Note: SQE is valid from sw_sq_cons up to cqe_sq_cons (exclusive)
@@ -2285,9 +2242,9 @@ static int bnxt_qplib_cq_process_req(struct bnxt_qplib_cq *cq,
sw_sq_cons, cqe->wr_id, cqe->status);
cqe++;
(*budget)--;
- bnxt_qplib_lock_buddy_cq(qp, cq);
bnxt_qplib_mark_qp_error(qp);
- bnxt_qplib_unlock_buddy_cq(qp, cq);
+ /* Add qp to flush list of the CQ */
+ bnxt_qplib_add_flush_qp(qp);
} else {
if (swq->flags & SQ_SEND_FLAGS_SIGNAL_COMP) {
/* Before we complete, do WA 9060 */
@@ -2403,9 +2360,7 @@ static int bnxt_qplib_cq_process_res_rc(struct bnxt_qplib_cq *cq,
if (hwcqe->status != CQ_RES_RC_STATUS_OK) {
qp->state = CMDQ_MODIFY_QP_NEW_STATE_ERR;
/* Add qp to flush list of the CQ */
- bnxt_qplib_lock_buddy_cq(qp, cq);
- __bnxt_qplib_add_flush_qp(qp);
- bnxt_qplib_unlock_buddy_cq(qp, cq);
+ bnxt_qplib_add_flush_qp(qp);
}
}
@@ -2489,9 +2444,7 @@ static int bnxt_qplib_cq_process_res_ud(struct bnxt_qplib_cq *cq,
if (hwcqe->status != CQ_RES_RC_STATUS_OK) {
qp->state = CMDQ_MODIFY_QP_NEW_STATE_ERR;
/* Add qp to flush list of the CQ */
- bnxt_qplib_lock_buddy_cq(qp, cq);
- __bnxt_qplib_add_flush_qp(qp);
- bnxt_qplib_unlock_buddy_cq(qp, cq);
+ bnxt_qplib_add_flush_qp(qp);
}
}
done:
@@ -2501,11 +2454,9 @@ done:
bool bnxt_qplib_is_cq_empty(struct bnxt_qplib_cq *cq)
{
struct cq_base *hw_cqe, **hw_cqe_ptr;
- unsigned long flags;
u32 sw_cons, raw_cons;
bool rc = true;
- spin_lock_irqsave(&cq->hwq.lock, flags);
raw_cons = cq->hwq.cons;
sw_cons = HWQ_CMP(raw_cons, &cq->hwq);
hw_cqe_ptr = (struct cq_base **)cq->hwq.pbl_ptr;
@@ -2513,7 +2464,6 @@ bool bnxt_qplib_is_cq_empty(struct bnxt_qplib_cq *cq)
/* Check for Valid bit. If the CQE is valid, return false */
rc = !CQE_CMP_VALID(hw_cqe, raw_cons, cq->hwq.max_elements);
- spin_unlock_irqrestore(&cq->hwq.lock, flags);
return rc;
}
@@ -2602,9 +2552,7 @@ static int bnxt_qplib_cq_process_res_raweth_qp1(struct bnxt_qplib_cq *cq,
if (hwcqe->status != CQ_RES_RC_STATUS_OK) {
qp->state = CMDQ_MODIFY_QP_NEW_STATE_ERR;
/* Add qp to flush list of the CQ */
- bnxt_qplib_lock_buddy_cq(qp, cq);
- __bnxt_qplib_add_flush_qp(qp);
- bnxt_qplib_unlock_buddy_cq(qp, cq);
+ bnxt_qplib_add_flush_qp(qp);
}
}
@@ -2719,9 +2667,7 @@ do_rq:
*/
/* Add qp to flush list of the CQ */
- bnxt_qplib_lock_buddy_cq(qp, cq);
- __bnxt_qplib_add_flush_qp(qp);
- bnxt_qplib_unlock_buddy_cq(qp, cq);
+ bnxt_qplib_add_flush_qp(qp);
done:
return rc;
}
@@ -2750,7 +2696,7 @@ int bnxt_qplib_process_flush_list(struct bnxt_qplib_cq *cq,
u32 budget = num_cqes;
unsigned long flags;
- spin_lock_irqsave(&cq->hwq.lock, flags);
+ spin_lock_irqsave(&cq->flush_lock, flags);
list_for_each_entry(qp, &cq->sqf_head, sq_flush) {
dev_dbg(&cq->hwq.pdev->dev,
"QPLIB: FP: Flushing SQ QP= %p",
@@ -2764,7 +2710,7 @@ int bnxt_qplib_process_flush_list(struct bnxt_qplib_cq *cq,
qp);
__flush_rq(&qp->rq, qp, &cqe, &budget);
}
- spin_unlock_irqrestore(&cq->hwq.lock, flags);
+ spin_unlock_irqrestore(&cq->flush_lock, flags);
return num_cqes - budget;
}
@@ -2773,11 +2719,9 @@ int bnxt_qplib_poll_cq(struct bnxt_qplib_cq *cq, struct bnxt_qplib_cqe *cqe,
int num_cqes, struct bnxt_qplib_qp **lib_qp)
{
struct cq_base *hw_cqe, **hw_cqe_ptr;
- unsigned long flags;
u32 sw_cons, raw_cons;
int budget, rc = 0;
- spin_lock_irqsave(&cq->hwq.lock, flags);
raw_cons = cq->hwq.cons;
budget = num_cqes;
@@ -2853,20 +2797,15 @@ int bnxt_qplib_poll_cq(struct bnxt_qplib_cq *cq, struct bnxt_qplib_cqe *cqe,
bnxt_qplib_arm_cq(cq, DBR_DBR_TYPE_CQ);
}
exit:
- spin_unlock_irqrestore(&cq->hwq.lock, flags);
return num_cqes - budget;
}
void bnxt_qplib_req_notify_cq(struct bnxt_qplib_cq *cq, u32 arm_type)
{
- unsigned long flags;
-
- spin_lock_irqsave(&cq->hwq.lock, flags);
if (arm_type)
bnxt_qplib_arm_cq(cq, arm_type);
/* Using cq->arm_state variable to track whether to issue cq handler */
atomic_set(&cq->arm_state, 1);
- spin_unlock_irqrestore(&cq->hwq.lock, flags);
}
void bnxt_qplib_flush_cqn_wq(struct bnxt_qplib_qp *qp)
diff --git a/drivers/infiniband/hw/bnxt_re/qplib_fp.h b/drivers/infiniband/hw/bnxt_re/qplib_fp.h
index ca0a2ffa3509..ade9f13c0fd1 100644
--- a/drivers/infiniband/hw/bnxt_re/qplib_fp.h
+++ b/drivers/infiniband/hw/bnxt_re/qplib_fp.h
@@ -389,6 +389,18 @@ struct bnxt_qplib_cq {
struct list_head sqf_head, rqf_head;
atomic_t arm_state;
spinlock_t compl_lock; /* synch CQ handlers */
+/* Locking Notes:
+ * QP can move to error state from modify_qp, async error event or error
+ * CQE as part of poll_cq. When QP is moved to error state, it gets added
+ * to two flush lists, one each for SQ and RQ.
+ * Each flush list is protected by qplib_cq->flush_lock. Both scq and rcq
+ * flush_locks should be acquired when QP is moved to error. The control path
+ * operations(modify_qp and async error events) are synchronized with poll_cq
+ * using upper level CQ locks (bnxt_re_cq->cq_lock) of both SCQ and RCQ.
+ * The qplib_cq->flush_lock is required to synchronize two instances of poll_cq
+ * of the same QP while manipulating the flush list.
+ */
+ spinlock_t flush_lock; /* QP flush management */
};
#define BNXT_QPLIB_MAX_IRRQE_ENTRY_SIZE sizeof(struct xrrq_irrq)
diff --git a/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c b/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c
index 8329ec6a7946..80027a494730 100644
--- a/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c
+++ b/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c
@@ -305,9 +305,8 @@ static int bnxt_qplib_process_qp_event(struct bnxt_qplib_rcfw *rcfw,
err_event->res_err_state_reason);
if (!qp)
break;
- bnxt_qplib_acquire_cq_locks(qp, &flags);
bnxt_qplib_mark_qp_error(qp);
- bnxt_qplib_release_cq_locks(qp, &flags);
+ rcfw->aeq_handler(rcfw, qp_event, qp);
break;
default:
/* Command Response */
@@ -460,7 +459,11 @@ int bnxt_qplib_init_rcfw(struct bnxt_qplib_rcfw *rcfw,
int rc;
RCFW_CMD_PREP(req, INITIALIZE_FW, cmd_flags);
-
+ /* Supply (log-base-2-of-host-page-size - base-page-shift)
+ * to bono to adjust the doorbell page sizes.
+ */
+ req.log2_dbr_pg_size = cpu_to_le16(PAGE_SHIFT -
+ RCFW_DBR_BASE_PAGE_SHIFT);
/*
* VFs need not setup the HW context area, PF
* shall setup this area for VF. Skipping the
diff --git a/drivers/infiniband/hw/bnxt_re/qplib_rcfw.h b/drivers/infiniband/hw/bnxt_re/qplib_rcfw.h
index 6bee6e3636ea..c7cce2e4185e 100644
--- a/drivers/infiniband/hw/bnxt_re/qplib_rcfw.h
+++ b/drivers/infiniband/hw/bnxt_re/qplib_rcfw.h
@@ -49,6 +49,7 @@
#define RCFW_COMM_SIZE 0x104
#define RCFW_DBR_PCI_BAR_REGION 2
+#define RCFW_DBR_BASE_PAGE_SHIFT 12
#define RCFW_CMD_PREP(req, CMD, cmd_flags) \
do { \
diff --git a/drivers/infiniband/hw/bnxt_re/qplib_sp.c b/drivers/infiniband/hw/bnxt_re/qplib_sp.c
index 03057983341f..ee98e5efef84 100644
--- a/drivers/infiniband/hw/bnxt_re/qplib_sp.c
+++ b/drivers/infiniband/hw/bnxt_re/qplib_sp.c
@@ -139,7 +139,8 @@ int bnxt_qplib_get_dev_attr(struct bnxt_qplib_rcfw *rcfw,
attr->max_pkey = le32_to_cpu(sb->max_pkeys);
attr->max_inline_data = le32_to_cpu(sb->max_inline_data);
- attr->l2_db_size = (sb->l2_db_space_size + 1) * PAGE_SIZE;
+ attr->l2_db_size = (sb->l2_db_space_size + 1) *
+ (0x01 << RCFW_DBR_BASE_PAGE_SHIFT);
attr->max_sgid = le32_to_cpu(sb->max_gid);
bnxt_qplib_query_version(rcfw, attr->fw_ver);
diff --git a/drivers/infiniband/hw/bnxt_re/roce_hsi.h b/drivers/infiniband/hw/bnxt_re/roce_hsi.h
index 2d7ea096a247..3e5a4f760d0e 100644
--- a/drivers/infiniband/hw/bnxt_re/roce_hsi.h
+++ b/drivers/infiniband/hw/bnxt_re/roce_hsi.h
@@ -1761,7 +1761,30 @@ struct cmdq_initialize_fw {
#define CMDQ_INITIALIZE_FW_TIM_PG_SIZE_PG_2M (0x3UL << 4)
#define CMDQ_INITIALIZE_FW_TIM_PG_SIZE_PG_8M (0x4UL << 4)
#define CMDQ_INITIALIZE_FW_TIM_PG_SIZE_PG_1G (0x5UL << 4)
- __le16 reserved16;
+ /* This value is (log-base-2-of-DBR-page-size - 12).
+ * 0 for 4KB. HW supported values are enumerated below.
+ */
+ __le16 log2_dbr_pg_size;
+ #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_MASK 0xfUL
+ #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_SFT 0
+ #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_4K 0x0UL
+ #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_8K 0x1UL
+ #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_16K 0x2UL
+ #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_32K 0x3UL
+ #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_64K 0x4UL
+ #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_128K 0x5UL
+ #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_256K 0x6UL
+ #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_512K 0x7UL
+ #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_1M 0x8UL
+ #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_2M 0x9UL
+ #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_4M 0xaUL
+ #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_8M 0xbUL
+ #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_16M 0xcUL
+ #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_32M 0xdUL
+ #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_64M 0xeUL
+ #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_128M 0xfUL
+ #define CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_LAST \
+ CMDQ_INITIALIZE_FW_LOG2_DBR_PG_SIZE_PG_128M
__le64 qpc_page_dir;
__le64 mrw_page_dir;
__le64 srq_page_dir;
diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
index db2ff352d75f..ec638778661c 100644
--- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
+++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
@@ -4383,7 +4383,7 @@ err_dma_alloc_buf:
eq->l0_dma = 0;
if (mhop_num == 1)
- for (i -= i; i >= 0; i--)
+ for (i -= 1; i >= 0; i--)
dma_free_coherent(dev, buf_chk_sz, eq->buf[i],
eq->buf_dma[i]);
else if (mhop_num == 2) {
diff --git a/drivers/infiniband/hw/mlx4/cq.c b/drivers/infiniband/hw/mlx4/cq.c
index 9a566ee3ceff..82adc0d1d30e 100644
--- a/drivers/infiniband/hw/mlx4/cq.c
+++ b/drivers/infiniband/hw/mlx4/cq.c
@@ -601,6 +601,7 @@ static void use_tunnel_data(struct mlx4_ib_qp *qp, struct mlx4_ib_cq *cq, struct
wc->dlid_path_bits = 0;
if (is_eth) {
+ wc->slid = 0;
wc->vlan_id = be16_to_cpu(hdr->tun.sl_vid);
memcpy(&(wc->smac[0]), (char *)&hdr->tun.mac_31_0, 4);
memcpy(&(wc->smac[4]), (char *)&hdr->tun.slid_mac_47_32, 2);
@@ -851,7 +852,6 @@ repoll:
}
}
- wc->slid = be16_to_cpu(cqe->rlid);
g_mlpath_rqpn = be32_to_cpu(cqe->g_mlpath_rqpn);
wc->src_qp = g_mlpath_rqpn & 0xffffff;
wc->dlid_path_bits = (g_mlpath_rqpn >> 24) & 0x7f;
@@ -860,6 +860,7 @@ repoll:
wc->wc_flags |= mlx4_ib_ipoib_csum_ok(cqe->status,
cqe->checksum) ? IB_WC_IP_CSUM_OK : 0;
if (is_eth) {
+ wc->slid = 0;
wc->sl = be16_to_cpu(cqe->sl_vid) >> 13;
if (be32_to_cpu(cqe->vlan_my_qpn) &
MLX4_CQE_CVLAN_PRESENT_MASK) {
@@ -871,6 +872,7 @@ repoll:
memcpy(wc->smac, cqe->smac, ETH_ALEN);
wc->wc_flags |= (IB_WC_WITH_VLAN | IB_WC_WITH_SMAC);
} else {
+ wc->slid = be16_to_cpu(cqe->rlid);
wc->sl = be16_to_cpu(cqe->sl_vid) >> 12;
wc->vlan_id = 0xffff;
}
diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c
index 8d2ee9322f2e..5a0e4fc4785a 100644
--- a/drivers/infiniband/hw/mlx4/main.c
+++ b/drivers/infiniband/hw/mlx4/main.c
@@ -219,8 +219,6 @@ static int mlx4_ib_update_gids_v1_v2(struct gid_entry *gids,
gid_tbl[i].version = 2;
if (!ipv6_addr_v4mapped((struct in6_addr *)&gids[i].gid))
gid_tbl[i].type = 1;
- else
- memset(&gid_tbl[i].gid, 0, 12);
}
}
@@ -366,8 +364,13 @@ static int mlx4_ib_del_gid(struct ib_device *device,
if (!gids) {
ret = -ENOMEM;
} else {
- for (i = 0; i < MLX4_MAX_PORT_GIDS; i++)
- memcpy(&gids[i].gid, &port_gid_table->gids[i].gid, sizeof(union ib_gid));
+ for (i = 0; i < MLX4_MAX_PORT_GIDS; i++) {
+ memcpy(&gids[i].gid,
+ &port_gid_table->gids[i].gid,
+ sizeof(union ib_gid));
+ gids[i].gid_type =
+ port_gid_table->gids[i].gid_type;
+ }
}
}
spin_unlock_bh(&iboe->lock);
diff --git a/drivers/infiniband/hw/mlx5/cq.c b/drivers/infiniband/hw/mlx5/cq.c
index 5b974fb97611..15457c9569a7 100644
--- a/drivers/infiniband/hw/mlx5/cq.c
+++ b/drivers/infiniband/hw/mlx5/cq.c
@@ -226,7 +226,6 @@ static void handle_responder(struct ib_wc *wc, struct mlx5_cqe64 *cqe,
wc->ex.invalidate_rkey = be32_to_cpu(cqe->imm_inval_pkey);
break;
}
- wc->slid = be16_to_cpu(cqe->slid);
wc->src_qp = be32_to_cpu(cqe->flags_rqpn) & 0xffffff;
wc->dlid_path_bits = cqe->ml_path;
g = (be32_to_cpu(cqe->flags_rqpn) >> 28) & 3;
@@ -241,10 +240,12 @@ static void handle_responder(struct ib_wc *wc, struct mlx5_cqe64 *cqe,
}
if (ll != IB_LINK_LAYER_ETHERNET) {
+ wc->slid = be16_to_cpu(cqe->slid);
wc->sl = (be32_to_cpu(cqe->flags_rqpn) >> 24) & 0xf;
return;
}
+ wc->slid = 0;
vlan_present = cqe->l4_l3_hdr_type & 0x1;
roce_packet_type = (be32_to_cpu(cqe->flags_rqpn) >> 24) & 0x3;
if (vlan_present) {
@@ -1177,7 +1178,12 @@ static int resize_user(struct mlx5_ib_dev *dev, struct mlx5_ib_cq *cq,
if (ucmd.reserved0 || ucmd.reserved1)
return -EINVAL;
- umem = ib_umem_get(context, ucmd.buf_addr, entries * ucmd.cqe_size,
+ /* check multiplication overflow */
+ if (ucmd.cqe_size && SIZE_MAX / ucmd.cqe_size <= entries - 1)
+ return -EINVAL;
+
+ umem = ib_umem_get(context, ucmd.buf_addr,
+ (size_t)ucmd.cqe_size * entries,
IB_ACCESS_LOCAL_WRITE, 1);
if (IS_ERR(umem)) {
err = PTR_ERR(umem);
diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index 4236c8086820..7f8bda3a2005 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -245,12 +245,16 @@ struct mlx5_core_dev *mlx5_ib_get_native_port_mdev(struct mlx5_ib_dev *ibdev,
struct mlx5_ib_multiport_info *mpi;
struct mlx5_ib_port *port;
+ if (!mlx5_core_mp_enabled(ibdev->mdev) ||
+ ll != IB_LINK_LAYER_ETHERNET) {
+ if (native_port_num)
+ *native_port_num = ib_port_num;
+ return ibdev->mdev;
+ }
+
if (native_port_num)
*native_port_num = 1;
- if (!mlx5_core_mp_enabled(ibdev->mdev) || ll != IB_LINK_LAYER_ETHERNET)
- return ibdev->mdev;
-
port = &ibdev->port[ib_port_num - 1];
if (!port)
return NULL;
@@ -3263,7 +3267,7 @@ static void mlx5_ib_handle_event(struct work_struct *_work)
struct mlx5_ib_dev *ibdev;
struct ib_event ibev;
bool fatal = false;
- u8 port = 0;
+ u8 port = (u8)work->param;
if (mlx5_core_is_mp_slave(work->dev)) {
ibdev = mlx5_ib_get_ibdev_from_mpi(work->context);
@@ -3283,8 +3287,6 @@ static void mlx5_ib_handle_event(struct work_struct *_work)
case MLX5_DEV_EVENT_PORT_UP:
case MLX5_DEV_EVENT_PORT_DOWN:
case MLX5_DEV_EVENT_PORT_INITIALIZED:
- port = (u8)work->param;
-
/* In RoCE, port up/down events are handled in
* mlx5_netdev_event().
*/
@@ -3298,24 +3300,19 @@ static void mlx5_ib_handle_event(struct work_struct *_work)
case MLX5_DEV_EVENT_LID_CHANGE:
ibev.event = IB_EVENT_LID_CHANGE;
- port = (u8)work->param;
break;
case MLX5_DEV_EVENT_PKEY_CHANGE:
ibev.event = IB_EVENT_PKEY_CHANGE;
- port = (u8)work->param;
-
schedule_work(&ibdev->devr.ports[port - 1].pkey_change_work);
break;
case MLX5_DEV_EVENT_GUID_CHANGE:
ibev.event = IB_EVENT_GID_CHANGE;
- port = (u8)work->param;
break;
case MLX5_DEV_EVENT_CLIENT_REREG:
ibev.event = IB_EVENT_CLIENT_REREGISTER;
- port = (u8)work->param;
break;
case MLX5_DEV_EVENT_DELAY_DROP_TIMEOUT:
schedule_work(&ibdev->delay_drop.delay_drop_work);
@@ -3327,7 +3324,7 @@ static void mlx5_ib_handle_event(struct work_struct *_work)
ibev.device = &ibdev->ib_dev;
ibev.element.port_num = port;
- if (port < 1 || port > ibdev->num_ports) {
+ if (!rdma_is_port_valid(&ibdev->ib_dev, port)) {
mlx5_ib_warn(ibdev, "warning: event on port %d\n", port);
goto out;
}
@@ -3451,9 +3448,12 @@ static void destroy_umrc_res(struct mlx5_ib_dev *dev)
if (err)
mlx5_ib_warn(dev, "mr cache cleanup failed\n");
- mlx5_ib_destroy_qp(dev->umrc.qp);
- ib_free_cq(dev->umrc.cq);
- ib_dealloc_pd(dev->umrc.pd);
+ if (dev->umrc.qp)
+ mlx5_ib_destroy_qp(dev->umrc.qp);
+ if (dev->umrc.cq)
+ ib_free_cq(dev->umrc.cq);
+ if (dev->umrc.pd)
+ ib_dealloc_pd(dev->umrc.pd);
}
enum {
@@ -3555,12 +3555,15 @@ static int create_umr_res(struct mlx5_ib_dev *dev)
error_4:
mlx5_ib_destroy_qp(qp);
+ dev->umrc.qp = NULL;
error_3:
ib_free_cq(cq);
+ dev->umrc.cq = NULL;
error_2:
ib_dealloc_pd(pd);
+ dev->umrc.pd = NULL;
error_0:
kfree(attr);
@@ -4863,19 +4866,19 @@ static int mlx5_ib_stage_ib_reg_init(struct mlx5_ib_dev *dev)
return ib_register_device(&dev->ib_dev, NULL);
}
-static void mlx5_ib_stage_ib_reg_cleanup(struct mlx5_ib_dev *dev)
+static void mlx5_ib_stage_pre_ib_reg_umr_cleanup(struct mlx5_ib_dev *dev)
{
- ib_unregister_device(&dev->ib_dev);
+ destroy_umrc_res(dev);
}
-static int mlx5_ib_stage_umr_res_init(struct mlx5_ib_dev *dev)
+static void mlx5_ib_stage_ib_reg_cleanup(struct mlx5_ib_dev *dev)
{
- return create_umr_res(dev);
+ ib_unregister_device(&dev->ib_dev);
}
-static void mlx5_ib_stage_umr_res_cleanup(struct mlx5_ib_dev *dev)
+static int mlx5_ib_stage_post_ib_reg_umr_init(struct mlx5_ib_dev *dev)
{
- destroy_umrc_res(dev);
+ return create_umr_res(dev);
}
static int mlx5_ib_stage_delay_drop_init(struct mlx5_ib_dev *dev)
@@ -4985,12 +4988,15 @@ static const struct mlx5_ib_profile pf_profile = {
STAGE_CREATE(MLX5_IB_STAGE_BFREG,
mlx5_ib_stage_bfrag_init,
mlx5_ib_stage_bfrag_cleanup),
+ STAGE_CREATE(MLX5_IB_STAGE_PRE_IB_REG_UMR,
+ NULL,
+ mlx5_ib_stage_pre_ib_reg_umr_cleanup),
STAGE_CREATE(MLX5_IB_STAGE_IB_REG,
mlx5_ib_stage_ib_reg_init,
mlx5_ib_stage_ib_reg_cleanup),
- STAGE_CREATE(MLX5_IB_STAGE_UMR_RESOURCES,
- mlx5_ib_stage_umr_res_init,
- mlx5_ib_stage_umr_res_cleanup),
+ STAGE_CREATE(MLX5_IB_STAGE_POST_IB_REG_UMR,
+ mlx5_ib_stage_post_ib_reg_umr_init,
+ NULL),
STAGE_CREATE(MLX5_IB_STAGE_DELAY_DROP,
mlx5_ib_stage_delay_drop_init,
mlx5_ib_stage_delay_drop_cleanup),
diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index 139385129973..a5272499b600 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -739,8 +739,9 @@ enum mlx5_ib_stages {
MLX5_IB_STAGE_CONG_DEBUGFS,
MLX5_IB_STAGE_UAR,
MLX5_IB_STAGE_BFREG,
+ MLX5_IB_STAGE_PRE_IB_REG_UMR,
MLX5_IB_STAGE_IB_REG,
- MLX5_IB_STAGE_UMR_RESOURCES,
+ MLX5_IB_STAGE_POST_IB_REG_UMR,
MLX5_IB_STAGE_DELAY_DROP,
MLX5_IB_STAGE_CLASS_ATTR,
MLX5_IB_STAGE_MAX,
diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c
index 556e015678de..3e0b3f0238d6 100644
--- a/drivers/infiniband/hw/mlx5/mr.c
+++ b/drivers/infiniband/hw/mlx5/mr.c
@@ -739,6 +739,9 @@ int mlx5_mr_cache_cleanup(struct mlx5_ib_dev *dev)
{
int i;
+ if (!dev->cache.wq)
+ return 0;
+
dev->cache.stopped = 1;
flush_workqueue(dev->cache.wq);
@@ -838,7 +841,8 @@ static int mr_umem_get(struct ib_pd *pd, u64 start, u64 length,
*umem = ib_umem_get(pd->uobject->context, start, length,
access_flags, 0);
err = PTR_ERR_OR_ZERO(*umem);
- if (err < 0) {
+ if (err) {
+ *umem = NULL;
mlx5_ib_err(dev, "umem get failed (%d)\n", err);
return err;
}
@@ -1415,6 +1419,7 @@ int mlx5_ib_rereg_user_mr(struct ib_mr *ib_mr, int flags, u64 start,
if (err) {
mlx5_ib_warn(dev, "Failed to rereg UMR\n");
ib_umem_release(mr->umem);
+ mr->umem = NULL;
clean_mr(dev, mr);
return err;
}
@@ -1498,14 +1503,11 @@ static int clean_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
u32 key = mr->mmkey.key;
err = destroy_mkey(dev, mr);
- kfree(mr);
if (err) {
mlx5_ib_warn(dev, "failed to destroy mkey 0x%x (%d)\n",
key, err);
return err;
}
- } else {
- mlx5_mr_cache_free(dev, mr);
}
return 0;
@@ -1548,6 +1550,11 @@ static int dereg_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
atomic_sub(npages, &dev->mdev->priv.reg_pages);
}
+ if (!mr->allocated_from_cache)
+ kfree(mr);
+ else
+ mlx5_mr_cache_free(dev, mr);
+
return 0;
}
@@ -1816,7 +1823,6 @@ mlx5_ib_sg_to_klms(struct mlx5_ib_mr *mr,
mr->ibmr.iova = sg_dma_address(sg) + sg_offset;
mr->ibmr.length = 0;
- mr->ndescs = sg_nents;
for_each_sg(sgl, sg, sg_nents, i) {
if (unlikely(i >= mr->max_descs))
@@ -1828,6 +1834,7 @@ mlx5_ib_sg_to_klms(struct mlx5_ib_mr *mr,
sg_offset = 0;
}
+ mr->ndescs = i;
if (sg_offset_p)
*sg_offset_p = sg_offset;
diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c
index 39d24bf694a8..a2e1aa86e133 100644
--- a/drivers/infiniband/hw/mlx5/qp.c
+++ b/drivers/infiniband/hw/mlx5/qp.c
@@ -1161,7 +1161,7 @@ static void destroy_raw_packet_qp_sq(struct mlx5_ib_dev *dev,
ib_umem_release(sq->ubuffer.umem);
}
-static int get_rq_pas_size(void *qpc)
+static size_t get_rq_pas_size(void *qpc)
{
u32 log_page_size = MLX5_GET(qpc, qpc, log_page_size) + 12;
u32 log_rq_stride = MLX5_GET(qpc, qpc, log_rq_stride);
@@ -1177,7 +1177,8 @@ static int get_rq_pas_size(void *qpc)
}
static int create_raw_packet_qp_rq(struct mlx5_ib_dev *dev,
- struct mlx5_ib_rq *rq, void *qpin)
+ struct mlx5_ib_rq *rq, void *qpin,
+ size_t qpinlen)
{
struct mlx5_ib_qp *mqp = rq->base.container_mibqp;
__be64 *pas;
@@ -1186,9 +1187,12 @@ static int create_raw_packet_qp_rq(struct mlx5_ib_dev *dev,
void *rqc;
void *wq;
void *qpc = MLX5_ADDR_OF(create_qp_in, qpin, qpc);
- int inlen;
+ size_t rq_pas_size = get_rq_pas_size(qpc);
+ size_t inlen;
int err;
- u32 rq_pas_size = get_rq_pas_size(qpc);
+
+ if (qpinlen < rq_pas_size + MLX5_BYTE_OFF(create_qp_in, pas))
+ return -EINVAL;
inlen = MLX5_ST_SZ_BYTES(create_rq_in) + rq_pas_size;
in = kvzalloc(inlen, GFP_KERNEL);
@@ -1277,7 +1281,7 @@ static void destroy_raw_packet_qp_tir(struct mlx5_ib_dev *dev,
}
static int create_raw_packet_qp(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
- u32 *in,
+ u32 *in, size_t inlen,
struct ib_pd *pd)
{
struct mlx5_ib_raw_packet_qp *raw_packet_qp = &qp->raw_packet_qp;
@@ -1309,7 +1313,7 @@ static int create_raw_packet_qp(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
rq->flags |= MLX5_IB_RQ_CVLAN_STRIPPING;
if (qp->flags & MLX5_IB_QP_PCI_WRITE_END_PADDING)
rq->flags |= MLX5_IB_RQ_PCI_WRITE_END_PADDING;
- err = create_raw_packet_qp_rq(dev, rq, in);
+ err = create_raw_packet_qp_rq(dev, rq, in, inlen);
if (err)
goto err_destroy_sq;
@@ -1584,6 +1588,7 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd,
u32 uidx = MLX5_IB_DEFAULT_UIDX;
struct mlx5_ib_create_qp ucmd;
struct mlx5_ib_qp_base *base;
+ int mlx5_st;
void *qpc;
u32 *in;
int err;
@@ -1592,6 +1597,10 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd,
spin_lock_init(&qp->sq.lock);
spin_lock_init(&qp->rq.lock);
+ mlx5_st = to_mlx5_st(init_attr->qp_type);
+ if (mlx5_st < 0)
+ return -EINVAL;
+
if (init_attr->rwq_ind_tbl) {
if (!udata)
return -ENOSYS;
@@ -1753,7 +1762,7 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd,
qpc = MLX5_ADDR_OF(create_qp_in, in, qpc);
- MLX5_SET(qpc, qpc, st, to_mlx5_st(init_attr->qp_type));
+ MLX5_SET(qpc, qpc, st, mlx5_st);
MLX5_SET(qpc, qpc, pm_state, MLX5_QP_PM_MIGRATED);
if (init_attr->qp_type != MLX5_IB_QPT_REG_UMR)
@@ -1867,11 +1876,16 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd,
}
}
+ if (inlen < 0) {
+ err = -EINVAL;
+ goto err;
+ }
+
if (init_attr->qp_type == IB_QPT_RAW_PACKET ||
qp->flags & MLX5_IB_QP_UNDERLAY) {
qp->raw_packet_qp.sq.ubuffer.buf_addr = ucmd.sq_buf_addr;
raw_packet_qp_copy_info(qp, &qp->raw_packet_qp);
- err = create_raw_packet_qp(dev, qp, in, pd);
+ err = create_raw_packet_qp(dev, qp, in, inlen, pd);
} else {
err = mlx5_core_create_qp(dev->mdev, &base->mqp, in, inlen);
}
@@ -3095,8 +3109,10 @@ static int __mlx5_ib_modify_qp(struct ib_qp *ibqp,
goto out;
if (mlx5_cur >= MLX5_QP_NUM_STATE || mlx5_new >= MLX5_QP_NUM_STATE ||
- !optab[mlx5_cur][mlx5_new])
+ !optab[mlx5_cur][mlx5_new]) {
+ err = -EINVAL;
goto out;
+ }
op = optab[mlx5_cur][mlx5_new];
optpar = ib_mask_to_mlx5_opt(attr_mask);
diff --git a/drivers/infiniband/hw/mlx5/srq.c b/drivers/infiniband/hw/mlx5/srq.c
index 6d5fadad9090..3c7522d025f2 100644
--- a/drivers/infiniband/hw/mlx5/srq.c
+++ b/drivers/infiniband/hw/mlx5/srq.c
@@ -241,8 +241,8 @@ struct ib_srq *mlx5_ib_create_srq(struct ib_pd *pd,
{
struct mlx5_ib_dev *dev = to_mdev(pd->device);
struct mlx5_ib_srq *srq;
- int desc_size;
- int buf_size;
+ size_t desc_size;
+ size_t buf_size;
int err;
struct mlx5_srq_attr in = {0};
__u32 max_srq_wqes = 1 << MLX5_CAP_GEN(dev->mdev, log_max_srq_sz);
@@ -266,15 +266,18 @@ struct ib_srq *mlx5_ib_create_srq(struct ib_pd *pd,
desc_size = sizeof(struct mlx5_wqe_srq_next_seg) +
srq->msrq.max_gs * sizeof(struct mlx5_wqe_data_seg);
+ if (desc_size == 0 || srq->msrq.max_gs > desc_size)
+ return ERR_PTR(-EINVAL);
desc_size = roundup_pow_of_two(desc_size);
- desc_size = max_t(int, 32, desc_size);
+ desc_size = max_t(size_t, 32, desc_size);
+ if (desc_size < sizeof(struct mlx5_wqe_srq_next_seg))
+ return ERR_PTR(-EINVAL);
srq->msrq.max_avail_gather = (desc_size - sizeof(struct mlx5_wqe_srq_next_seg)) /
sizeof(struct mlx5_wqe_data_seg);
srq->msrq.wqe_shift = ilog2(desc_size);
buf_size = srq->msrq.max * desc_size;
- mlx5_ib_dbg(dev, "desc_size 0x%x, req wr 0x%x, srq size 0x%x, max_gs 0x%x, max_avail_gather 0x%x\n",
- desc_size, init_attr->attr.max_wr, srq->msrq.max, srq->msrq.max_gs,
- srq->msrq.max_avail_gather);
+ if (buf_size < desc_size)
+ return ERR_PTR(-EINVAL);
in.type = init_attr->srq_type;
if (pd->uobject)
diff --git a/drivers/infiniband/hw/qedr/main.c b/drivers/infiniband/hw/qedr/main.c
index db4bf97c0e15..0ffb9b93e22d 100644
--- a/drivers/infiniband/hw/qedr/main.c
+++ b/drivers/infiniband/hw/qedr/main.c
@@ -833,7 +833,8 @@ static struct qedr_dev *qedr_add(struct qed_dev *cdev, struct pci_dev *pdev,
dev->num_cnq = dev->ops->rdma_get_min_cnq_msix(cdev);
if (!dev->num_cnq) {
- DP_ERR(dev, "not enough CNQ resources.\n");
+ DP_ERR(dev, "Failed. At least one CNQ is required.\n");
+ rc = -ENOMEM;
goto init_err;
}
diff --git a/drivers/infiniband/hw/qedr/qedr_iw_cm.c b/drivers/infiniband/hw/qedr/qedr_iw_cm.c
index 478b7317b80a..26dc374787f7 100644
--- a/drivers/infiniband/hw/qedr/qedr_iw_cm.c
+++ b/drivers/infiniband/hw/qedr/qedr_iw_cm.c
@@ -458,8 +458,7 @@ qedr_addr6_resolve(struct qedr_dev *dev,
}
return -EINVAL;
}
- neigh = dst_neigh_lookup(dst, &dst_in);
-
+ neigh = dst_neigh_lookup(dst, &fl6.daddr);
if (neigh) {
rcu_read_lock();
if (neigh->nud_state & NUD_VALID) {
@@ -494,10 +493,14 @@ int qedr_iw_connect(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
qp = idr_find(&dev->qpidr, conn_param->qpn);
- laddr = (struct sockaddr_in *)&cm_id->local_addr;
- raddr = (struct sockaddr_in *)&cm_id->remote_addr;
- laddr6 = (struct sockaddr_in6 *)&cm_id->local_addr;
- raddr6 = (struct sockaddr_in6 *)&cm_id->remote_addr;
+ laddr = (struct sockaddr_in *)&cm_id->m_local_addr;
+ raddr = (struct sockaddr_in *)&cm_id->m_remote_addr;
+ laddr6 = (struct sockaddr_in6 *)&cm_id->m_local_addr;
+ raddr6 = (struct sockaddr_in6 *)&cm_id->m_remote_addr;
+
+ DP_DEBUG(dev, QEDR_MSG_IWARP, "MAPPED %d %d\n",
+ ntohs(((struct sockaddr_in *)&cm_id->remote_addr)->sin_port),
+ ntohs(raddr->sin_port));
DP_DEBUG(dev, QEDR_MSG_IWARP,
"Connect source address: %pISpc, remote address: %pISpc\n",
@@ -599,8 +602,8 @@ int qedr_iw_create_listen(struct iw_cm_id *cm_id, int backlog)
int rc;
int i;
- laddr = (struct sockaddr_in *)&cm_id->local_addr;
- laddr6 = (struct sockaddr_in6 *)&cm_id->local_addr;
+ laddr = (struct sockaddr_in *)&cm_id->m_local_addr;
+ laddr6 = (struct sockaddr_in6 *)&cm_id->m_local_addr;
DP_DEBUG(dev, QEDR_MSG_IWARP,
"Create Listener address: %pISpc\n", &cm_id->local_addr);
diff --git a/drivers/infiniband/hw/qedr/verbs.c b/drivers/infiniband/hw/qedr/verbs.c
index 53f00dbf313f..419a158e8fca 100644
--- a/drivers/infiniband/hw/qedr/verbs.c
+++ b/drivers/infiniband/hw/qedr/verbs.c
@@ -1841,14 +1841,15 @@ static void qedr_reset_qp_hwq_info(struct qedr_qp_hwq_info *qph)
static int qedr_update_qp_state(struct qedr_dev *dev,
struct qedr_qp *qp,
+ enum qed_roce_qp_state cur_state,
enum qed_roce_qp_state new_state)
{
int status = 0;
- if (new_state == qp->state)
+ if (new_state == cur_state)
return 0;
- switch (qp->state) {
+ switch (cur_state) {
case QED_ROCE_QP_STATE_RESET:
switch (new_state) {
case QED_ROCE_QP_STATE_INIT:
@@ -1955,6 +1956,7 @@ int qedr_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
struct qedr_dev *dev = get_qedr_dev(&qp->dev->ibdev);
const struct ib_global_route *grh = rdma_ah_read_grh(&attr->ah_attr);
enum ib_qp_state old_qp_state, new_qp_state;
+ enum qed_roce_qp_state cur_state;
int rc = 0;
DP_DEBUG(dev, QEDR_MSG_QP,
@@ -2086,18 +2088,23 @@ int qedr_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
SET_FIELD(qp_params.modify_flags,
QED_ROCE_MODIFY_QP_VALID_ACK_TIMEOUT, 1);
- qp_params.ack_timeout = attr->timeout;
- if (attr->timeout) {
- u32 temp;
-
- temp = 4096 * (1UL << attr->timeout) / 1000 / 1000;
- /* FW requires [msec] */
- qp_params.ack_timeout = temp;
- } else {
- /* Infinite */
+ /* The received timeout value is an exponent used like this:
+ * "12.7.34 LOCAL ACK TIMEOUT
+ * Value representing the transport (ACK) timeout for use by
+ * the remote, expressed as: 4.096 * 2^timeout [usec]"
+ * The FW expects timeout in msec so we need to divide the usec
+ * result by 1000. We'll approximate 1000~2^10, and 4.096 ~ 2^2,
+ * so we get: 2^2 * 2^timeout / 2^10 = 2^(timeout - 8).
+ * The value of zero means infinite so we use a 'max_t' to make
+ * sure that sub 1 msec values will be configured as 1 msec.
+ */
+ if (attr->timeout)
+ qp_params.ack_timeout =
+ 1 << max_t(int, attr->timeout - 8, 0);
+ else
qp_params.ack_timeout = 0;
- }
}
+
if (attr_mask & IB_QP_RETRY_CNT) {
SET_FIELD(qp_params.modify_flags,
QED_ROCE_MODIFY_QP_VALID_RETRY_CNT, 1);
@@ -2170,13 +2177,25 @@ int qedr_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
qp->dest_qp_num = attr->dest_qp_num;
}
+ cur_state = qp->state;
+
+ /* Update the QP state before the actual ramrod to prevent a race with
+ * fast path. Modifying the QP state to error will cause the device to
+ * flush the CQEs and while polling the flushed CQEs will considered as
+ * a potential issue if the QP isn't in error state.
+ */
+ if ((attr_mask & IB_QP_STATE) && qp->qp_type != IB_QPT_GSI &&
+ !udata && qp_params.new_state == QED_ROCE_QP_STATE_ERR)
+ qp->state = QED_ROCE_QP_STATE_ERR;
+
if (qp->qp_type != IB_QPT_GSI)
rc = dev->ops->rdma_modify_qp(dev->rdma_ctx,
qp->qed_qp, &qp_params);
if (attr_mask & IB_QP_STATE) {
if ((qp->qp_type != IB_QPT_GSI) && (!udata))
- rc = qedr_update_qp_state(dev, qp, qp_params.new_state);
+ rc = qedr_update_qp_state(dev, qp, cur_state,
+ qp_params.new_state);
qp->state = qp_params.new_state;
}
@@ -3034,6 +3053,11 @@ static int __qedr_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
switch (wr->opcode) {
case IB_WR_SEND_WITH_IMM:
+ if (unlikely(rdma_protocol_iwarp(&dev->ibdev, 1))) {
+ rc = -EINVAL;
+ *bad_wr = wr;
+ break;
+ }
wqe->req_type = RDMA_SQ_REQ_TYPE_SEND_WITH_IMM;
swqe = (struct rdma_sq_send_wqe_1st *)wqe;
swqe->wqe_size = 2;
@@ -3075,6 +3099,11 @@ static int __qedr_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
break;
case IB_WR_RDMA_WRITE_WITH_IMM:
+ if (unlikely(rdma_protocol_iwarp(&dev->ibdev, 1))) {
+ rc = -EINVAL;
+ *bad_wr = wr;
+ break;
+ }
wqe->req_type = RDMA_SQ_REQ_TYPE_RDMA_WR_WITH_IMM;
rwqe = (struct rdma_sq_rdma_wqe_1st *)wqe;
@@ -3724,7 +3753,7 @@ int qedr_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *wc)
{
struct qedr_dev *dev = get_qedr_dev(ibcq->device);
struct qedr_cq *cq = get_qedr_cq(ibcq);
- union rdma_cqe *cqe = cq->latest_cqe;
+ union rdma_cqe *cqe;
u32 old_cons, new_cons;
unsigned long flags;
int update = 0;
@@ -3741,6 +3770,7 @@ int qedr_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *wc)
return qedr_gsi_poll_cq(ibcq, num_entries, wc);
spin_lock_irqsave(&cq->cq_lock, flags);
+ cqe = cq->latest_cqe;
old_cons = qed_chain_get_cons_idx_u32(&cq->pbl);
while (num_entries && is_valid_cqe(cq, cqe)) {
struct qedr_qp *qp;
diff --git a/drivers/infiniband/sw/rdmavt/mr.c b/drivers/infiniband/sw/rdmavt/mr.c
index 1b2e5362a3ff..cc429b567d0a 100644
--- a/drivers/infiniband/sw/rdmavt/mr.c
+++ b/drivers/infiniband/sw/rdmavt/mr.c
@@ -489,11 +489,13 @@ static int rvt_check_refs(struct rvt_mregion *mr, const char *t)
unsigned long timeout;
struct rvt_dev_info *rdi = ib_to_rvt(mr->pd->device);
- if (percpu_ref_is_zero(&mr->refcount))
- return 0;
- /* avoid dma mr */
- if (mr->lkey)
+ if (mr->lkey) {
+ /* avoid dma mr */
rvt_dereg_clean_qps(mr);
+ /* @mr was indexed on rcu protected @lkey_table */
+ synchronize_rcu();
+ }
+
timeout = wait_for_completion_timeout(&mr->comp, 5 * HZ);
if (!timeout) {
rvt_pr_err(rdi,