* [PATCH] remoteproc: virtio: fix uninitialized buffer use in rproc_virtio_get()
@ 2026-01-08 19:23 Kery Qi
2026-01-12 11:02 ` Zhongqiu Han
2026-01-12 18:40 ` Mathieu Poirier
0 siblings, 2 replies; 3+ messages in thread
From: Kery Qi @ 2026-01-08 19:23 UTC (permalink / raw)
To: andersson; +Cc: linux-remoteproc, Kery Qi
rproc_virtio_get() returns early on out-of-bounds access without touching
the caller-provided buffer. Callers of virtio config accessors commonly
pass stack storage and do not get an error code back, so stale/uninit
bytes may be consumed as device configuration, leading to unpredictable
behavior and potentially leaking stack data if later exposed.
Always clear the destination buffer, reject offsets past config_len, and
clamp the read length to the available config bytes before copying.
Fixes: 92b38f851470 ("remoteproc: support virtio config space.")
Signed-off-by: Kery Qi <qikeyu2017@gmail.com>
---
drivers/remoteproc/remoteproc_virtio.c | 9 ++++++---
1 file changed, 6 insertions(+), 3 deletions(-)
diff --git a/drivers/remoteproc/remoteproc_virtio.c b/drivers/remoteproc/remoteproc_virtio.c
index c5d46a878149..8fa8c8a86b4b 100644
--- a/drivers/remoteproc/remoteproc_virtio.c
+++ b/drivers/remoteproc/remoteproc_virtio.c
@@ -298,10 +298,13 @@ static void rproc_virtio_get(struct virtio_device *vdev, unsigned int offset,
rsc = (void *)rvdev->rproc->table_ptr + rvdev->rsc_offset;
cfg = &rsc->vring[rsc->num_of_vrings];
- if (offset + len > rsc->config_len || offset + len < len) {
- dev_err(&vdev->dev, "rproc_virtio_get: access out of bounds\n");
+ memset(buf, 0, len);
+
+ if (offset > rsc->config_len)
return;
- }
+
+ if (len > rsc->config_len - offset)
+ len = rsc->config_len - offset;
memcpy(buf, cfg + offset, len);
}
--
2.34.1
^ permalink raw reply related [flat|nested] 3+ messages in thread
* Re: [PATCH] remoteproc: virtio: fix uninitialized buffer use in rproc_virtio_get()
2026-01-08 19:23 [PATCH] remoteproc: virtio: fix uninitialized buffer use in rproc_virtio_get() Kery Qi
@ 2026-01-12 11:02 ` Zhongqiu Han
2026-01-12 18:40 ` Mathieu Poirier
1 sibling, 0 replies; 3+ messages in thread
From: Zhongqiu Han @ 2026-01-12 11:02 UTC (permalink / raw)
To: Kery Qi, andersson, Mathieu Poirier; +Cc: linux-remoteproc, zhongqiu.han
On 1/9/2026 3:23 AM, Kery Qi wrote:
> rproc_virtio_get() returns early on out-of-bounds access without touching
> the caller-provided buffer. Callers of virtio config accessors commonly
> pass stack storage and do not get an error code back, so stale/uninit
> bytes may be consumed as device configuration, leading to unpredictable
> behavior and potentially leaking stack data if later exposed.
>
> Always clear the destination buffer, reject offsets past config_len, and
> clamp the read length to the available config bytes before copying.
>
> Fixes: 92b38f851470 ("remoteproc: support virtio config space.")
> Signed-off-by: Kery Qi <qikeyu2017@gmail.com>
Please use ./scripts/get_maintainer.pl to retrieve the full list of
maintainer email and cc list.
> ---
> drivers/remoteproc/remoteproc_virtio.c | 9 ++++++---
> 1 file changed, 6 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/remoteproc/remoteproc_virtio.c b/drivers/remoteproc/remoteproc_virtio.c
> index c5d46a878149..8fa8c8a86b4b 100644
> --- a/drivers/remoteproc/remoteproc_virtio.c
> +++ b/drivers/remoteproc/remoteproc_virtio.c
> @@ -298,10 +298,13 @@ static void rproc_virtio_get(struct virtio_device *vdev, unsigned int offset,
> rsc = (void *)rvdev->rproc->table_ptr + rvdev->rsc_offset;
> cfg = &rsc->vring[rsc->num_of_vrings];
Hi Kery,
You want to initialize the buffer to avoid issues caused by calling the
virtio_cread macro with an uninitialized variable buffer, which should
be fair enough,
virtio_cread(vdev, structname, member, ptr)
vdev->config->get()
OR
__virtio_cread_many()
vdev->config->get()
>
> - if (offset + len > rsc->config_len || offset + len < len) {
> - dev_err(&vdev->dev, "rproc_virtio_get: access out of bounds\n");
> + memset(buf, 0, len);
> +
The original check offset + len > rsc->config_len ensures that access to
the config space does not overflow, while offset + len < len is used to
detect unsigned integer overflow.
> + if (offset > rsc->config_len)
Any log info?
> return;
> - }
And then may I know what does this really mean? Are you suggesting a
quick fast path check or something else? Please clarify the reasoning
behind this, and if that makes sense, please split the initialization
and the boundary check into two separate patches.
> +
> + if (len > rsc->config_len - offset)
Formatting issue here.
> + len = rsc->config_len - offset;
Why is truncation allowed here? Any err log info?
>
> memcpy(buf, cfg + offset, len);
> }
--
Thx and BRs,
Zhongqiu Han
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] remoteproc: virtio: fix uninitialized buffer use in rproc_virtio_get()
2026-01-08 19:23 [PATCH] remoteproc: virtio: fix uninitialized buffer use in rproc_virtio_get() Kery Qi
2026-01-12 11:02 ` Zhongqiu Han
@ 2026-01-12 18:40 ` Mathieu Poirier
1 sibling, 0 replies; 3+ messages in thread
From: Mathieu Poirier @ 2026-01-12 18:40 UTC (permalink / raw)
To: Kery Qi; +Cc: andersson, linux-remoteproc
Good day Kery,
On Fri, Jan 09, 2026 at 03:23:36AM +0800, Kery Qi wrote:
> rproc_virtio_get() returns early on out-of-bounds access without touching
> the caller-provided buffer. Callers of virtio config accessors commonly
> pass stack storage and do not get an error code back, so stale/uninit
> bytes may be consumed as device configuration, leading to unpredictable
> behavior and potentially leaking stack data if later exposed.
>
> Always clear the destination buffer, reject offsets past config_len, and
> clamp the read length to the available config bytes before copying.
>
> Fixes: 92b38f851470 ("remoteproc: support virtio config space.")
> Signed-off-by: Kery Qi <qikeyu2017@gmail.com>
> ---
> drivers/remoteproc/remoteproc_virtio.c | 9 ++++++---
> 1 file changed, 6 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/remoteproc/remoteproc_virtio.c b/drivers/remoteproc/remoteproc_virtio.c
> index c5d46a878149..8fa8c8a86b4b 100644
> --- a/drivers/remoteproc/remoteproc_virtio.c
> +++ b/drivers/remoteproc/remoteproc_virtio.c
> @@ -298,10 +298,13 @@ static void rproc_virtio_get(struct virtio_device *vdev, unsigned int offset,
> rsc = (void *)rvdev->rproc->table_ptr + rvdev->rsc_offset;
> cfg = &rsc->vring[rsc->num_of_vrings];
>
> - if (offset + len > rsc->config_len || offset + len < len) {
> - dev_err(&vdev->dev, "rproc_virtio_get: access out of bounds\n");
> + memset(buf, 0, len);
I think this should be done in the caller, not virtio_config_ops->get(). I am
willing to reconsider if you include this change in a patchset that changes all
the other virtio_config_ops->get() function.
Thanks,
Mathieu
> +
> + if (offset > rsc->config_len)
> return;
> - }
> +
> + if (len > rsc->config_len - offset)
> + len = rsc->config_len - offset;
>
> memcpy(buf, cfg + offset, len);
> }
> --
> 2.34.1
>
>
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2026-01-12 18:40 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-01-08 19:23 [PATCH] remoteproc: virtio: fix uninitialized buffer use in rproc_virtio_get() Kery Qi
2026-01-12 11:02 ` Zhongqiu Han
2026-01-12 18:40 ` Mathieu Poirier
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox