diff options
| author |   Michal Kalderon <michal.kalderon@marvell.com> | 2021-05-06 10:08:19 +0300 |
|---|---|---|
| committer |   Christoph Hellwig <hch@lst.de> | 2021-05-11 18:30:45 +0200 |
| commit | 8cc365f9559b86802afc0208389f5c8d46b4ad61 (patch) | |
| tree | c85d4c2f2ea98e272db3ce654acd65abf360605f | |
| parent | nvmet: fix inline bio check for passthru (diff) | |
| download | linux-dev-8cc365f9559b86802afc0208389f5c8d46b4ad61.tar.xz linux-dev-8cc365f9559b86802afc0208389f5c8d46b4ad61.zip | |
nvmet-rdma: Fix NULL deref when SEND is completed with error
When running some traffic and taking down the link on peer, a
retry counter exceeded error is received. This leads to
nvmet_rdma_error_comp which tried accessing the cq_context to
obtain the queue. The cq_context is no longer valid after the
fix to use shared CQ mechanism and should be obtained similar
to how it is obtained in other functions from the wc->qp.
[ 905.786331] nvmet_rdma: SEND for CQE 0x00000000e3337f90 failed with status transport retry counter exceeded (12).
[ 905.832048] BUG: unable to handle kernel NULL pointer dereference at 0000000000000048
[ 905.839919] PGD 0 P4D 0
[ 905.842464] Oops: 0000 1 SMP NOPTI
[ 905.846144] CPU: 13 PID: 1557 Comm: kworker/13:1H Kdump: loaded Tainted: G OE --------- - - 4.18.0-304.el8.x86_64 #1
[ 905.872135] RIP: 0010:nvmet_rdma_error_comp+0x5/0x1b [nvmet_rdma]
[ 905.878259] Code: 19 4f c0 e8 89 b3 a5 f6 e9 5b e0 ff ff 0f b7 75 14 4c 89 ea 48 c7 c7 08 1a 4f c0 e8 71 b3 a5 f6 e9 4b e0 ff ff 0f 1f 44 00 00 <48> 8b 47 48 48 85 c0 74 08 48 89 c7 e9 98 bf 49 00 e9 c3 e3 ff ff
[ 905.897135] RSP: 0018:ffffab601c45fe28 EFLAGS: 00010246
[ 905.902387] RAX: 0000000000000065 RBX: ffff9e729ea2f800 RCX: 0000000000000000
[ 905.909558] RDX: 0000000000000000 RSI: ffff9e72df9567c8 RDI: 0000000000000000
[ 905.916731] RBP: ffff9e729ea2b400 R08: 000000000000074d R09: 0000000000000074
[ 905.923903] R10: 0000000000000000 R11: ffffab601c45fcc0 R12: 0000000000000010
[ 905.931074] R13: 0000000000000000 R14: 0000000000000010 R15: ffff9e729ea2f400
[ 905.938247] FS: 0000000000000000(0000) GS:ffff9e72df940000(0000) knlGS:0000000000000000
[ 905.938249] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 905.950067] nvmet_rdma: SEND for CQE 0x00000000c7356cca failed with status transport retry counter exceeded (12).
[ 905.961855] CR2: 0000000000000048 CR3: 000000678d010004 CR4: 00000000007706e0
[ 905.961855] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 905.961856] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[ 905.961857] PKRU: 55555554
[ 906.010315] Call Trace:
[ 906.012778] __ib_process_cq+0x89/0x170 [ib_core]
[ 906.017509] ib_cq_poll_work+0x26/0x80 [ib_core]
[ 906.022152] process_one_work+0x1a7/0x360
[ 906.026182] ? create_worker+0x1a0/0x1a0
[ 906.030123] worker_thread+0x30/0x390
[ 906.033802] ? create_worker+0x1a0/0x1a0
[ 906.037744] kthread+0x116/0x130
[ 906.040988] ? kthread_flush_work_fn+0x10/0x10
[ 906.045456] ret_from_fork+0x1f/0x40
Fixes: ca0f1a8055be2 ("nvmet-rdma: use new shared CQ mechanism")
Signed-off-by: Shai Malin <smalin@marvell.com>
Signed-off-by: Michal Kalderon <michal.kalderon@marvell.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
| -rw-r--r-- | drivers/nvme/target/rdma.c | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/drivers/nvme/target/rdma.c b/drivers/nvme/target/rdma.c index 6c1f3ab7649c..7d607f435e36 100644 --- a/drivers/nvme/target/rdma.c +++ b/drivers/nvme/target/rdma.c @@ -700,7 +700,7 @@ static void nvmet_rdma_send_done(struct ib_cq *cq, struct ib_wc *wc) { struct nvmet_rdma_rsp *rsp = container_of(wc->wr_cqe, struct nvmet_rdma_rsp, send_cqe); - struct nvmet_rdma_queue *queue = cq->cq_context; + struct nvmet_rdma_queue *queue = wc->qp->qp_context; nvmet_rdma_release_rsp(rsp); @@ -786,7 +786,7 @@ static void nvmet_rdma_write_data_done(struct ib_cq *cq, struct ib_wc *wc) { struct nvmet_rdma_rsp *rsp = container_of(wc->wr_cqe, struct nvmet_rdma_rsp, write_cqe); - struct nvmet_rdma_queue *queue = cq->cq_context; + struct nvmet_rdma_queue *queue = wc->qp->qp_context; struct rdma_cm_id *cm_id = rsp->queue->cm_id; u16 status; |