Re: [PATCH v2] usb: dwc3: gadget: linearize SG requests that exceed the TRB cache

[Faisal Hassan]

On 29-10-2026 03:25 pm, Greg Kroah-Hartman wrote:
> On Thu, Oct 29, 2026 at 12:40:06AM +0530, Faisal Hassan wrote:
> > Hi all,
> >
> > A gentle ping on this v2 when you get a chance.
> >
> > It addresses the sashiko review comments on v1: a linearized request is
> > now unmapped before its bounce buffer is freed, and an allocation
> > failure or a short SG list is returned from usb_ep_queue() instead of
> > leaving the request stuck on pending_list.
> >
> > Please let me know if you have any comments, or if you'd like me to
> > change anything.
>
> You seem to have missed the review comments on this patch:
> 	https://lore.kernel.org/r/sashiko-outbox-162302@xxxxxxxxxx

Hi Greg,

Sorry for the noise, and thank you for pointing me to the review. I
had read it but should have replied on the list before pinging.

Both comments look pre-existing to me:

1) The unmap/re-map leak of a request that is mapped but gets no TRB
   dates back to cdb55b39fab8 and 4a71fcb8ac5f and affects all
   requests. v2 only unmaps a linearized request before freeing its
   bounce buffer.

2) The isoc error return after the request is on pending_list comes
   from 25abad6a0584. v2 does not change it, as the new failure paths
   return before the request is queued.

I can fix 1) in a separate patch or as a first patch in a v3 series,
whichever you and Thinh prefer. For 2), please let me know if you want
it handled here.

Thanks for your patience, and sorry again for the extra mail.

Faisal