On 5/17/23 08:39, Jason Gunthorpe wrote:
On Tue, May 16, 2023 at 07:32:35PM -0500, Bob Pearson wrote:
On 5/16/23 19:11, Jason Gunthorpe wrote:
On Wed, May 10, 2023 at 11:50:56AM +0800, Guoqing Jiang wrote:
We need to call spin_lock_irqsave/spin_unlock_irqrestore for state_lock
in rxe, otherwsie the callchain
ib_post_send_mad
-> spin_lock_irqsave
-> ib_post_send -> rxe_post_send
-> spin_lock_bh
-> spin_unlock_bh
-> spin_unlock_irqrestore
caused below traces during run block nvmeof-mp/001 test.
..
Fixes: f605f26ea196 ("RDMA/rxe: Protect QP state with qp->state_lock")
Signed-off-by: Guoqing Jiang <guoqing.jiang@xxxxxxxxx>
Reviewed-by: Bob Pearson <rpearsonhpe@xxxxxxxxx>
---
drivers/infiniband/sw/rxe/rxe_comp.c | 26 +++++++++++--------
drivers/infiniband/sw/rxe/rxe_net.c | 7 +++---
drivers/infiniband/sw/rxe/rxe_qp.c | 36 +++++++++++++++++----------
drivers/infiniband/sw/rxe/rxe_recv.c | 9 ++++---
drivers/infiniband/sw/rxe/rxe_req.c | 30 ++++++++++++----------
drivers/infiniband/sw/rxe/rxe_resp.c | 14 ++++++-----
drivers/infiniband/sw/rxe/rxe_verbs.c | 25 ++++++++++---------
7 files changed, 86 insertions(+), 61 deletions(-)
Applied to for-rc, thanks
Jason
You didn't mention it but this shouldn't have applied/compiled without
fixing the overlap of these two patches. ??
oh, I fixed it, it is trivial
Thanks for the fixing and review.
Guoqing
