[PATCH] rxe: Fix a sleep-in-atomic bug in post_one_send
From: Jia-Ju Bai
Date: Wed May 31 2017 - 21:47:33 EST
The driver may sleep under a spin lock, and the function call path is:
post_one_send (acquire the lock by spin_lock_irqsave)
init_send_wqe
copy_from_user --> may sleep
To fix it, the lock is released before copy_from_user, and the lock is
acquired again after this function.
Signed-off-by: Jia-Ju Bai <baijiaju1990@xxxxxxx>
---
drivers/infiniband/sw/rxe/rxe_verbs.c | 9 ++++++---
1 file changed, 6 insertions(+), 3 deletions(-)
diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.c b/drivers/infiniband/sw/rxe/rxe_verbs.c
index 83d709e..6fb7e1a 100644
--- a/drivers/infiniband/sw/rxe/rxe_verbs.c
+++ b/drivers/infiniband/sw/rxe/rxe_verbs.c
@@ -725,7 +725,7 @@ static int init_send_wqe(struct rxe_qp *qp, struct ib_send_wr *ibwr,
{
int num_sge = ibwr->num_sge;
struct ib_sge *sge;
- int i;
+ int i, err;
u8 *p;
init_send_wr(qp, &wqe->wr, ibwr);
@@ -740,8 +740,11 @@ static int init_send_wqe(struct rxe_qp *qp, struct ib_send_wr *ibwr,
sge = ibwr->sg_list;
for (i = 0; i < num_sge; i++, sge++) {
- if (qp->is_user && copy_from_user(p, (__user void *)
- (uintptr_t)sge->addr, sge->length))
+ spin_unlock_irq(&qp->sq.sq_lock);
+ err = copy_from_user(p, (__user void *)
+ (uintptr_t)sge->addr, sge->length);
+ spin_lock_irq(&qp->sq.sq_lock);
+ if (qp->is_user && err)
return -EFAULT;
else if (!qp->is_user)
--
1.7.9.5