* [PATCH v3] nvme-rdma: handle nvme completion data length
@ 2020-10-25 11:51 ` zhenwei pi
0 siblings, 0 replies; 8+ messages in thread
From: zhenwei pi @ 2020-10-25 11:51 UTC (permalink / raw)
To: kbusch, axboe, hch, sagi; +Cc: linux-nvme, linux-kernel, lengchao, pizhenwei
Hit a kernel warning:
refcount_t: underflow; use-after-free.
WARNING: CPU: 0 PID: 0 at lib/refcount.c:28
RIP: 0010:refcount_warn_saturate+0xd9/0xe0
Call Trace:
<IRQ>
nvme_rdma_recv_done+0xf3/0x280 [nvme_rdma]
__ib_process_cq+0x76/0x150 [ib_core]
...
The reason is that a zero bytes message received from target, and the
host side continues to process without length checking, then the
previous CQE is processed twice.
Do sanity check on received data length, try to recovery for corrupted
CQE case.
Because zero bytes message in not defined in spec, using zero bytes
message to detect dead connections on transport layer is not
standard, currently still treat it as illegal.
Thanks to Chao Leng & Sagi for suggestions.
Signed-off-by: zhenwei pi <pizhenwei@bytedance.com>
---
drivers/nvme/host/rdma.c | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/drivers/nvme/host/rdma.c b/drivers/nvme/host/rdma.c
index aad829a2b50d..40a0a3b6476c 100644
--- a/drivers/nvme/host/rdma.c
+++ b/drivers/nvme/host/rdma.c
@@ -1768,6 +1768,14 @@ static void nvme_rdma_recv_done(struct ib_cq *cq, struct ib_wc *wc)
return;
}
+ /* sanity checking for received data length */
+ if (unlikely(wc->byte_len < len)) {
+ dev_err(queue->ctrl->ctrl.device,
+ "Unexpected nvme completion length(%d)\n", wc->byte_len);
+ nvme_rdma_error_recovery(queue->ctrl);
+ return;
+ }
+
ib_dma_sync_single_for_cpu(ibdev, qe->dma, len, DMA_FROM_DEVICE);
/*
* AEN requests are special as they don't time out and can
--
2.11.0
^ permalink raw reply related [flat|nested] 8+ messages in thread* [PATCH v3] nvme-rdma: handle nvme completion data length
@ 2020-10-25 11:51 ` zhenwei pi
0 siblings, 0 replies; 8+ messages in thread
From: zhenwei pi @ 2020-10-25 11:51 UTC (permalink / raw)
To: kbusch, axboe, hch, sagi; +Cc: pizhenwei, linux-kernel, linux-nvme, lengchao
Hit a kernel warning:
refcount_t: underflow; use-after-free.
WARNING: CPU: 0 PID: 0 at lib/refcount.c:28
RIP: 0010:refcount_warn_saturate+0xd9/0xe0
Call Trace:
<IRQ>
nvme_rdma_recv_done+0xf3/0x280 [nvme_rdma]
__ib_process_cq+0x76/0x150 [ib_core]
...
The reason is that a zero bytes message received from target, and the
host side continues to process without length checking, then the
previous CQE is processed twice.
Do sanity check on received data length, try to recovery for corrupted
CQE case.
Because zero bytes message in not defined in spec, using zero bytes
message to detect dead connections on transport layer is not
standard, currently still treat it as illegal.
Thanks to Chao Leng & Sagi for suggestions.
Signed-off-by: zhenwei pi <pizhenwei@bytedance.com>
---
drivers/nvme/host/rdma.c | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/drivers/nvme/host/rdma.c b/drivers/nvme/host/rdma.c
index aad829a2b50d..40a0a3b6476c 100644
--- a/drivers/nvme/host/rdma.c
+++ b/drivers/nvme/host/rdma.c
@@ -1768,6 +1768,14 @@ static void nvme_rdma_recv_done(struct ib_cq *cq, struct ib_wc *wc)
return;
}
+ /* sanity checking for received data length */
+ if (unlikely(wc->byte_len < len)) {
+ dev_err(queue->ctrl->ctrl.device,
+ "Unexpected nvme completion length(%d)\n", wc->byte_len);
+ nvme_rdma_error_recovery(queue->ctrl);
+ return;
+ }
+
ib_dma_sync_single_for_cpu(ibdev, qe->dma, len, DMA_FROM_DEVICE);
/*
* AEN requests are special as they don't time out and can
--
2.11.0
_______________________________________________
Linux-nvme mailing list
Linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme
^ permalink raw reply related [flat|nested] 8+ messages in thread* Re: [PATCH v3] nvme-rdma: handle nvme completion data length
2020-10-25 11:51 ` zhenwei pi
@ 2020-10-26 7:40 ` Sagi Grimberg
-1 siblings, 0 replies; 8+ messages in thread
From: Sagi Grimberg @ 2020-10-26 7:40 UTC (permalink / raw)
To: zhenwei pi, kbusch, axboe, hch; +Cc: linux-nvme, linux-kernel, lengchao
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH v3] nvme-rdma: handle nvme completion data length
2020-10-25 11:51 ` zhenwei pi
@ 2020-10-27 9:07 ` Christoph Hellwig
-1 siblings, 0 replies; 8+ messages in thread
From: Christoph Hellwig @ 2020-10-27 9:07 UTC (permalink / raw)
To: zhenwei pi; +Cc: kbusch, axboe, hch, sagi, linux-nvme, linux-kernel, lengchao
Thanks,
applied to nvme-5.10.
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH v3] nvme-rdma: handle nvme completion data length
2020-10-25 11:51 ` zhenwei pi
@ 2020-10-28 16:58 ` Max Gurtovoy
-1 siblings, 0 replies; 8+ messages in thread
From: Max Gurtovoy @ 2020-10-28 16:58 UTC (permalink / raw)
To: zhenwei pi, kbusch, axboe, hch, sagi; +Cc: linux-kernel, linux-nvme, lengchao
On 10/25/2020 1:51 PM, zhenwei pi wrote:
> Hit a kernel warning:
> refcount_t: underflow; use-after-free.
> WARNING: CPU: 0 PID: 0 at lib/refcount.c:28
>
> RIP: 0010:refcount_warn_saturate+0xd9/0xe0
> Call Trace:
> <IRQ>
> nvme_rdma_recv_done+0xf3/0x280 [nvme_rdma]
> __ib_process_cq+0x76/0x150 [ib_core]
> ...
>
> The reason is that a zero bytes message received from target, and the
> host side continues to process without length checking, then the
> previous CQE is processed twice.
>
> Do sanity check on received data length, try to recovery for corrupted
> CQE case.
>
> Because zero bytes message in not defined in spec, using zero bytes
> message to detect dead connections on transport layer is not
> standard, currently still treat it as illegal.
>
> Thanks to Chao Leng & Sagi for suggestions.
>
> Signed-off-by: zhenwei pi <pizhenwei@bytedance.com>
> ---
> drivers/nvme/host/rdma.c | 8 ++++++++
> 1 file changed, 8 insertions(+)
>
Seems strange that the targets sends zero byte packets.
Can you specify which target is this and the scenario ?
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH v3] nvme-rdma: handle nvme completion data length
@ 2020-10-28 16:58 ` Max Gurtovoy
0 siblings, 0 replies; 8+ messages in thread
From: Max Gurtovoy @ 2020-10-28 16:58 UTC (permalink / raw)
To: zhenwei pi, kbusch, axboe, hch, sagi; +Cc: linux-kernel, linux-nvme, lengchao
On 10/25/2020 1:51 PM, zhenwei pi wrote:
> Hit a kernel warning:
> refcount_t: underflow; use-after-free.
> WARNING: CPU: 0 PID: 0 at lib/refcount.c:28
>
> RIP: 0010:refcount_warn_saturate+0xd9/0xe0
> Call Trace:
> <IRQ>
> nvme_rdma_recv_done+0xf3/0x280 [nvme_rdma]
> __ib_process_cq+0x76/0x150 [ib_core]
> ...
>
> The reason is that a zero bytes message received from target, and the
> host side continues to process without length checking, then the
> previous CQE is processed twice.
>
> Do sanity check on received data length, try to recovery for corrupted
> CQE case.
>
> Because zero bytes message in not defined in spec, using zero bytes
> message to detect dead connections on transport layer is not
> standard, currently still treat it as illegal.
>
> Thanks to Chao Leng & Sagi for suggestions.
>
> Signed-off-by: zhenwei pi <pizhenwei@bytedance.com>
> ---
> drivers/nvme/host/rdma.c | 8 ++++++++
> 1 file changed, 8 insertions(+)
>
Seems strange that the targets sends zero byte packets.
Can you specify which target is this and the scenario ?
_______________________________________________
Linux-nvme mailing list
Linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme
^ permalink raw reply [flat|nested] 8+ messages in thread
end of thread, other threads:[~2020-10-28 23:21 UTC | newest]
Thread overview: 8+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-10-25 11:51 [PATCH v3] nvme-rdma: handle nvme completion data length zhenwei pi
2020-10-25 11:51 ` zhenwei pi
2020-10-26 7:40 ` Sagi Grimberg
2020-10-26 7:40 ` Sagi Grimberg
2020-10-27 9:07 ` Christoph Hellwig
2020-10-27 9:07 ` Christoph Hellwig
2020-10-28 16:58 ` Max Gurtovoy
2020-10-28 16:58 ` Max Gurtovoy
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.