A recent patch replaced a tasklet execution of cq->comp_handler
by a direct call. While this made sense it let changes to cq->notify
state be unprotected and assumed that the cq completion machinery
and the ulp done callbacks were reentrant. The result is that in
some cases completion events can be lost. This patch moves the
cq->comp_handler call inside of the spinlock in rxe_cq_post which
solves both issues. This is compatible with the matching code in
the request notify verb.
Fixes: 78b26a335310 ("RDMA/rxe: Remove tasklet call from rxe_cq.c")
Signed-off-by: Bob Pearson <rpearsonhpe@xxxxxxxxx>
---
drivers/infiniband/sw/rxe/rxe_cq.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/infiniband/sw/rxe/rxe_cq.c b/drivers/infiniband/sw/rxe/rxe_cq.c
index 20ff0c0c4605..6ca2a05b6a2a 100644
--- a/drivers/infiniband/sw/rxe/rxe_cq.c
+++ b/drivers/infiniband/sw/rxe/rxe_cq.c
@@ -113,8 +113,6 @@ int rxe_cq_post(struct rxe_cq *cq, struct rxe_cqe *cqe, int solicited)
queue_advance_producer(cq->queue, QUEUE_TYPE_TO_CLIENT);
- spin_unlock_irqrestore(&cq->cq_lock, flags);
-
if ((cq->notify == IB_CQ_NEXT_COMP) ||
(cq->notify == IB_CQ_SOLICITED && solicited)) {
cq->notify = 0;
@@ -122,6 +120,8 @@ int rxe_cq_post(struct rxe_cq *cq, struct rxe_cqe *cqe, int solicited)
cq->ibcq.comp_handler(&cq->ibcq, cq->ibcq.cq_context);
}
+ spin_unlock_irqrestore(&cq->cq_lock, flags);
+
return 0;
}
base-commit: 830f93f47068b1632cc127871fbf27e918efdf46
--
2.39.2
