[PATCH] storvsc: fix memory leak on ring buffer busy

[PATCH] storvsc: fix memory leak on ring buffer busy

[Long Li]

From: Long Li <longli@microsoft.com>

When storvsc is sending I/O to Hyper-v, it may allocate a bigger buffer
descriptor for large data payload that can't fit into a pre-allocated
buffer descriptor. This bigger buffer is freed on return path.

If I/O request to Hyper-v fails due to ring buffer busy, the storvsc allocated
buffer descriptor should also be freed.

Signed-off-by: Long Li <longli@microsoft.com>
---
 drivers/scsi/storvsc_drv.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/scsi/storvsc_drv.c b/drivers/scsi/storvsc_drv.c
index 009adb0..db52882 100644
--- a/drivers/scsi/storvsc_drv.c
+++ b/drivers/scsi/storvsc_drv.c
@@ -1657,6 +1657,8 @@ static int storvsc_queuecommand(struct Scsi_Host *host, struct scsi_cmnd *scmnd)
 	ret = storvsc_do_io(dev, cmd_request, smp_processor_id());
 
 	if (ret == -EAGAIN) {
+		if (payload_sz > sizeof(cmd_request->mpb))
+			kfree(payload);
 		/* no more space */
 		return SCSI_MLQUEUE_DEVICE_BUSY;
 	}
-- 
2.7.4

Re: [PATCH] storvsc: fix memory leak on ring buffer busy

[Martin K. Petersen]

Long,

> When storvsc is sending I/O to Hyper-v, it may allocate a bigger
> buffer descriptor for large data payload that can't fit into a
> pre-allocated buffer descriptor. This bigger buffer is freed on return
> path.
>
> If I/O request to Hyper-v fails due to ring buffer busy, the storvsc
> allocated buffer descriptor should also be freed.

Which kernel version is this patch aimed at?

-- 
Martin K. Petersen	Oracle Linux Engineering

RE: [PATCH] storvsc: fix memory leak on ring buffer busy

[Long Li]

> -----Original Message-----
> From: Martin K. Petersen [mailto:martin.petersen@oracle.com]
> Sent: Tuesday, August 29, 2017 6:31 PM
> To: Long Li <longli@microsoft.com>
> Cc: KY Srinivasan <kys@microsoft.com>; Haiyang Zhang
> <haiyangz@microsoft.com>; James E . J . Bottomley
> <JBottomley@odin.com>; devel@linuxdriverproject.org; linux-
> scsi@vger.kernel.org; linux-kernel@vger.kernel.org; Long Li
> <longli@microsoft.com>
> Subject: Re: [PATCH] storvsc: fix memory leak on ring buffer busy
> 
> 
> Long,
> 
> > When storvsc is sending I/O to Hyper-v, it may allocate a bigger
> > buffer descriptor for large data payload that can't fit into a
> > pre-allocated buffer descriptor. This bigger buffer is freed on return
> > path.
> >
> > If I/O request to Hyper-v fails due to ring buffer busy, the storvsc
> > allocated buffer descriptor should also be freed.
> 
> Which kernel version is this patch aimed at?

Martin, thanks for pointing this out. This should also go to stable trees.

Cc: stable@vger.kernel.org
> 
> --
> Martin K. Petersen	Oracle Linux Engineering

Re: [PATCH] storvsc: fix memory leak on ring buffer busy

[Martin K. Petersen]

Long,

>> Which kernel version is this patch aimed at?
>
> Martin, thanks for pointing this out. This should also go to stable
> trees.

The reason I asked is that it didn't apply to neither fixes, nor
for-next.

I applied it to 4.13/scsi-fixes by hand and added a stable tag.

-- 
Martin K. Petersen	Oracle Linux Engineering

RE: [PATCH] storvsc: fix memory leak on ring buffer busy

[Long Li]

> Long,
> 
> >> Which kernel version is this patch aimed at?
> >
> > Martin, thanks for pointing this out. This should also go to stable
> > trees.
> 
> The reason I asked is that it didn't apply to neither fixes, nor for-next.
> 
> I applied it to 4.13/scsi-fixes by hand and added a stable tag.

Thank you. I'm sorry I misunderstood your question. I just realized I was working on an experimental branch. Sorry about that.

> 
> --
> Martin K. Petersen	Oracle Linux Engineering

Re: [PATCH] storvsc: fix memory leak on ring buffer busy

[Stephen Hemminger]

On Tue, 29 Aug 2017 21:31:11 -0400
"Martin K. Petersen" <martin.petersen@oracle.com> wrote:

> Long,
> 
> > When storvsc is sending I/O to Hyper-v, it may allocate a bigger
> > buffer descriptor for large data payload that can't fit into a
> > pre-allocated buffer descriptor. This bigger buffer is freed on return
> > path.
> >
> > If I/O request to Hyper-v fails due to ring buffer busy, the storvsc
> > allocated buffer descriptor should also be freed.  
> 
> Which kernel version is this patch aimed at?
> 

Looks like this an old issue. Probably should add

Fixes: be0cf6ca301c ("scsi: storvsc: Set the tablesize based on the information given by the host")