提交 7dad6b2e 编写于 作者: B Bart Van Assche 提交者: Christoph Hellwig

IB/srp: Fix a race condition triggered by destroying a queue pair

At least LID reassignment can trigger a race condition in the SRP
initiator driver, namely the receive completion handler trying to
post a request on a QP during or after QP destruction and before
the CQ's have been destroyed. Avoid this race by modifying a QP
into the error state and by waiting until all receive completions
have been processed before destroying a QP.
Reported-by: NMax Gurtuvoy <maxg@mellanox.com>
Signed-off-by: NBart Van Assche <bvanassche@acm.org>
Reviewed-by: NSagi Grimberg <sagig@mellanox.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>
上级 d92c0da7
...@@ -453,6 +453,41 @@ static struct srp_fr_pool *srp_alloc_fr_pool(struct srp_target_port *target) ...@@ -453,6 +453,41 @@ static struct srp_fr_pool *srp_alloc_fr_pool(struct srp_target_port *target)
dev->max_pages_per_mr); dev->max_pages_per_mr);
} }
/**
* srp_destroy_qp() - destroy an RDMA queue pair
* @ch: SRP RDMA channel.
*
* Change a queue pair into the error state and wait until all receive
* completions have been processed before destroying it. This avoids that
* the receive completion handler can access the queue pair while it is
* being destroyed.
*/
static void srp_destroy_qp(struct srp_rdma_ch *ch)
{
struct srp_target_port *target = ch->target;
static struct ib_qp_attr attr = { .qp_state = IB_QPS_ERR };
static struct ib_recv_wr wr = { .wr_id = SRP_LAST_WR_ID };
struct ib_recv_wr *bad_wr;
int ret;
/* Destroying a QP and reusing ch->done is only safe if not connected */
WARN_ON_ONCE(target->connected);
ret = ib_modify_qp(ch->qp, &attr, IB_QP_STATE);
WARN_ONCE(ret, "ib_cm_init_qp_attr() returned %d\n", ret);
if (ret)
goto out;
init_completion(&ch->done);
ret = ib_post_recv(ch->qp, &wr, &bad_wr);
WARN_ONCE(ret, "ib_post_recv() returned %d\n", ret);
if (ret == 0)
wait_for_completion(&ch->done);
out:
ib_destroy_qp(ch->qp);
}
static int srp_create_ch_ib(struct srp_rdma_ch *ch) static int srp_create_ch_ib(struct srp_rdma_ch *ch)
{ {
struct srp_target_port *target = ch->target; struct srp_target_port *target = ch->target;
...@@ -469,8 +504,9 @@ static int srp_create_ch_ib(struct srp_rdma_ch *ch) ...@@ -469,8 +504,9 @@ static int srp_create_ch_ib(struct srp_rdma_ch *ch)
if (!init_attr) if (!init_attr)
return -ENOMEM; return -ENOMEM;
/* + 1 for SRP_LAST_WR_ID */
recv_cq = ib_create_cq(dev->dev, srp_recv_completion, NULL, ch, recv_cq = ib_create_cq(dev->dev, srp_recv_completion, NULL, ch,
target->queue_size, ch->comp_vector); target->queue_size + 1, ch->comp_vector);
if (IS_ERR(recv_cq)) { if (IS_ERR(recv_cq)) {
ret = PTR_ERR(recv_cq); ret = PTR_ERR(recv_cq);
goto err; goto err;
...@@ -487,7 +523,7 @@ static int srp_create_ch_ib(struct srp_rdma_ch *ch) ...@@ -487,7 +523,7 @@ static int srp_create_ch_ib(struct srp_rdma_ch *ch)
init_attr->event_handler = srp_qp_event; init_attr->event_handler = srp_qp_event;
init_attr->cap.max_send_wr = m * target->queue_size; init_attr->cap.max_send_wr = m * target->queue_size;
init_attr->cap.max_recv_wr = target->queue_size; init_attr->cap.max_recv_wr = target->queue_size + 1;
init_attr->cap.max_recv_sge = 1; init_attr->cap.max_recv_sge = 1;
init_attr->cap.max_send_sge = 1; init_attr->cap.max_send_sge = 1;
init_attr->sq_sig_type = IB_SIGNAL_REQ_WR; init_attr->sq_sig_type = IB_SIGNAL_REQ_WR;
...@@ -530,7 +566,7 @@ static int srp_create_ch_ib(struct srp_rdma_ch *ch) ...@@ -530,7 +566,7 @@ static int srp_create_ch_ib(struct srp_rdma_ch *ch)
} }
if (ch->qp) if (ch->qp)
ib_destroy_qp(ch->qp); srp_destroy_qp(ch);
if (ch->recv_cq) if (ch->recv_cq)
ib_destroy_cq(ch->recv_cq); ib_destroy_cq(ch->recv_cq);
if (ch->send_cq) if (ch->send_cq)
...@@ -586,7 +622,7 @@ static void srp_free_ch_ib(struct srp_target_port *target, ...@@ -586,7 +622,7 @@ static void srp_free_ch_ib(struct srp_target_port *target,
if (ch->fmr_pool) if (ch->fmr_pool)
ib_destroy_fmr_pool(ch->fmr_pool); ib_destroy_fmr_pool(ch->fmr_pool);
} }
ib_destroy_qp(ch->qp); srp_destroy_qp(ch);
ib_destroy_cq(ch->send_cq); ib_destroy_cq(ch->send_cq);
ib_destroy_cq(ch->recv_cq); ib_destroy_cq(ch->recv_cq);
...@@ -1883,8 +1919,15 @@ static void srp_tl_err_work(struct work_struct *work) ...@@ -1883,8 +1919,15 @@ static void srp_tl_err_work(struct work_struct *work)
} }
static void srp_handle_qp_err(u64 wr_id, enum ib_wc_status wc_status, static void srp_handle_qp_err(u64 wr_id, enum ib_wc_status wc_status,
bool send_err, struct srp_target_port *target) bool send_err, struct srp_rdma_ch *ch)
{ {
struct srp_target_port *target = ch->target;
if (wr_id == SRP_LAST_WR_ID) {
complete(&ch->done);
return;
}
if (target->connected && !target->qp_in_error) { if (target->connected && !target->qp_in_error) {
if (wr_id & LOCAL_INV_WR_ID_MASK) { if (wr_id & LOCAL_INV_WR_ID_MASK) {
shost_printk(KERN_ERR, target->scsi_host, PFX shost_printk(KERN_ERR, target->scsi_host, PFX
...@@ -1915,8 +1958,7 @@ static void srp_recv_completion(struct ib_cq *cq, void *ch_ptr) ...@@ -1915,8 +1958,7 @@ static void srp_recv_completion(struct ib_cq *cq, void *ch_ptr)
if (likely(wc.status == IB_WC_SUCCESS)) { if (likely(wc.status == IB_WC_SUCCESS)) {
srp_handle_recv(ch, &wc); srp_handle_recv(ch, &wc);
} else { } else {
srp_handle_qp_err(wc.wr_id, wc.status, false, srp_handle_qp_err(wc.wr_id, wc.status, false, ch);
ch->target);
} }
} }
} }
...@@ -1932,8 +1974,7 @@ static void srp_send_completion(struct ib_cq *cq, void *ch_ptr) ...@@ -1932,8 +1974,7 @@ static void srp_send_completion(struct ib_cq *cq, void *ch_ptr)
iu = (struct srp_iu *) (uintptr_t) wc.wr_id; iu = (struct srp_iu *) (uintptr_t) wc.wr_id;
list_add(&iu->list, &ch->free_tx); list_add(&iu->list, &ch->free_tx);
} else { } else {
srp_handle_qp_err(wc.wr_id, wc.status, true, srp_handle_qp_err(wc.wr_id, wc.status, true, ch);
ch->target);
} }
} }
} }
......
...@@ -70,6 +70,8 @@ enum { ...@@ -70,6 +70,8 @@ enum {
LOCAL_INV_WR_ID_MASK = 1, LOCAL_INV_WR_ID_MASK = 1,
FAST_REG_WR_ID_MASK = 2, FAST_REG_WR_ID_MASK = 2,
SRP_LAST_WR_ID = 0xfffffffcU,
}; };
enum srp_target_state { enum srp_target_state {
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册