* [PATCH 1/1] nvme pci: fix the check of the cqe->command_id
@ 2019-01-18 5:17 Masanori Misono
2019-01-18 5:32 ` Masanori Misono
2019-01-18 15:22 ` Keith Busch
0 siblings, 2 replies; 3+ messages in thread
From: Masanori Misono @ 2019-01-18 5:17 UTC (permalink / raw)
nvme_pci_submit_async_event() sets the command_id as
NVME_AQ_BLK_MQ_DEPTH. So only call nvme_complete_async_event()
if the cqe->command_is exactly matches the value.
If the cqe->command_id is invalid due to the hardware bug,
blk_mq_tag_to_rq() returns NULL. Check that value to prevent NULL
pointer dereference. Remove the duplicate check if the command_id
is bigger than or equal to the nvmeq->q_depth.
Signed-off-by: Masanori Misono <m.misono760 at gmail.com>
---
drivers/nvme/host/pci.c | 16 ++++++++--------
1 file changed, 8 insertions(+), 8 deletions(-)
diff --git a/drivers/nvme/host/pci.c b/drivers/nvme/host/pci.c
index c33bb201b884..cc6bb7abb752 100644
--- a/drivers/nvme/host/pci.c
+++ b/drivers/nvme/host/pci.c
@@ -878,13 +878,6 @@ static inline void nvme_handle_cqe(struct nvme_queue *nvmeq, u16 idx)
volatile struct nvme_completion *cqe = &nvmeq->cqes[idx];
struct request *req;
- if (unlikely(cqe->command_id >= nvmeq->q_depth)) {
- dev_warn(nvmeq->dev->ctrl.device,
- "invalid id %d completed on queue %d\n",
- cqe->command_id, le16_to_cpu(cqe->sq_id));
- return;
- }
-
/*
* AEN requests are special as they don't time out and can
* survive any kind of queue freeze and often don't respond to
@@ -892,13 +885,20 @@ static inline void nvme_handle_cqe(struct nvme_queue *nvmeq, u16 idx)
* for them but rather special case them here.
*/
if (unlikely(nvmeq->qid == 0 &&
- cqe->command_id >= NVME_AQ_BLK_MQ_DEPTH)) {
+ cqe->command_id == NVME_AQ_BLK_MQ_DEPTH)) {
nvme_complete_async_event(&nvmeq->dev->ctrl,
cqe->status, &cqe->result);
return;
}
req = blk_mq_tag_to_rq(*nvmeq->tags, cqe->command_id);
+ if (unlikely(req == NULL)) {
+ dev_warn(nvmeq->dev->ctrl.device,
+ "invalid id %d completed on queue %d\n",
+ cqe->command_id, le16_to_cpu(cqe->sq_id));
+ return;
+ }
+
nvme_end_request(req, cqe->status, cqe->result);
}
--
2.17.1
^ permalink raw reply related [flat|nested] 3+ messages in thread
* [PATCH 1/1] nvme pci: fix the check of the cqe->command_id
2019-01-18 5:17 [PATCH 1/1] nvme pci: fix the check of the cqe->command_id Masanori Misono
@ 2019-01-18 5:32 ` Masanori Misono
2019-01-18 15:22 ` Keith Busch
1 sibling, 0 replies; 3+ messages in thread
From: Masanori Misono @ 2019-01-18 5:32 UTC (permalink / raw)
This patch is something similar to
http://lists.infradead.org/pipermail/linux-nvme/2019-January/021892.html
and tries to protect against hardware failures.
I just found this problem using my own fault injection scheme.
In fc.c and rdma.c, the return value of the blk_mq_tag_to_rq() is checked.
^ permalink raw reply [flat|nested] 3+ messages in thread
* [PATCH 1/1] nvme pci: fix the check of the cqe->command_id
2019-01-18 5:17 [PATCH 1/1] nvme pci: fix the check of the cqe->command_id Masanori Misono
2019-01-18 5:32 ` Masanori Misono
@ 2019-01-18 15:22 ` Keith Busch
1 sibling, 0 replies; 3+ messages in thread
From: Keith Busch @ 2019-01-18 15:22 UTC (permalink / raw)
On Fri, Jan 18, 2019@05:17:55AM +0000, Masanori Misono wrote:
> nvme_pci_submit_async_event() sets the command_id as
> NVME_AQ_BLK_MQ_DEPTH. So only call nvme_complete_async_event()
> if the cqe->command_is exactly matches the value.
>
> If the cqe->command_id is invalid due to the hardware bug,
> blk_mq_tag_to_rq() returns NULL. Check that value to prevent NULL
> pointer dereference. Remove the duplicate check if the command_id
> is bigger than or equal to the nvmeq->q_depth.
>
> Signed-off-by: Masanori Misono <m.misono760 at gmail.com>
This looks fine, though end result should be the same as before. We're
just trading one of the bounds checks for a NULL check, and I'm okay
with that.
Reviewed-by: Keith Busch <keith.busch at intel.com>
> drivers/nvme/host/pci.c | 16 ++++++++--------
> 1 file changed, 8 insertions(+), 8 deletions(-)
>
> diff --git a/drivers/nvme/host/pci.c b/drivers/nvme/host/pci.c
> index c33bb201b884..cc6bb7abb752 100644
> --- a/drivers/nvme/host/pci.c
> +++ b/drivers/nvme/host/pci.c
> @@ -878,13 +878,6 @@ static inline void nvme_handle_cqe(struct nvme_queue *nvmeq, u16 idx)
> volatile struct nvme_completion *cqe = &nvmeq->cqes[idx];
> struct request *req;
>
> - if (unlikely(cqe->command_id >= nvmeq->q_depth)) {
> - dev_warn(nvmeq->dev->ctrl.device,
> - "invalid id %d completed on queue %d\n",
> - cqe->command_id, le16_to_cpu(cqe->sq_id));
> - return;
> - }
> -
> /*
> * AEN requests are special as they don't time out and can
> * survive any kind of queue freeze and often don't respond to
> @@ -892,13 +885,20 @@ static inline void nvme_handle_cqe(struct nvme_queue *nvmeq, u16 idx)
> * for them but rather special case them here.
> */
> if (unlikely(nvmeq->qid == 0 &&
> - cqe->command_id >= NVME_AQ_BLK_MQ_DEPTH)) {
> + cqe->command_id == NVME_AQ_BLK_MQ_DEPTH)) {
> nvme_complete_async_event(&nvmeq->dev->ctrl,
> cqe->status, &cqe->result);
> return;
> }
>
> req = blk_mq_tag_to_rq(*nvmeq->tags, cqe->command_id);
> + if (unlikely(req == NULL)) {
> + dev_warn(nvmeq->dev->ctrl.device,
> + "invalid id %d completed on queue %d\n",
> + cqe->command_id, le16_to_cpu(cqe->sq_id));
> + return;
> + }
> +
> nvme_end_request(req, cqe->status, cqe->result);
> }
>
> --
> 2.17.1
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2019-01-18 15:22 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-01-18 5:17 [PATCH 1/1] nvme pci: fix the check of the cqe->command_id Masanori Misono
2019-01-18 5:32 ` Masanori Misono
2019-01-18 15:22 ` Keith Busch
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.