Re: [PATCH v3 1/3] media: uvcvideo: Let uvc_parse_frame() report a skipped frame

[Natasha Klaus]

On Mon, Sep 28, 2026, Laurent Pinchart wrote:
> On what device have you seen this occurring ? How did you test this
> patch ?

On no device. Noam found the overflow by reading the code. The zero
size case came from my review of his patch, done with an LLM tool
(see my reply on 2/3). There is no user report behind either.

1/3 is the refactor Ricardo suggested on the list, so that 2/3 and
3/3 can drop one bad frame instead of the whole streaming interface.
Apart from the truncated descriptor message moving to dev_warn(), it
changes no behaviour on its own.

Testing: Noam ran the series on a UVC gadget over dummy_hcd. One
format had an over-U32_MAX frame, a zero-size frame and a valid
frame. Both bad frames were skipped and the valid one was exposed.
I built it on x86_64.

Not tested: any physical camera, the -ENODATA truncated descriptor
path, 32-bit builds.

Natasha