Re: [PATCH 1/2] usbcore: Add quirk for 255-bytes initial config read

[Nikhil Solanke]

On Sun, 26 Jul 2026 at 21:55, Michal Pecio <michal.pecio@xxxxxxxxx> wrote:
> > diff --git a/drivers/usb/core/config.c b/drivers/usb/core/config.c
> > index 45e20c6d76c0..442c15f92ccd 100644
> > --- a/drivers/usb/core/config.c
> > +++ b/drivers/usb/core/config.c
> > @@ -912,6 +912,17 @@ int usb_get_configuration(struct usb_device *dev)
> >       unsigned char *bigbuffer;
> >       struct usb_config_descriptor *desc;
> >       int result;
> > +     size_t usb_config_req_size;
> > +
> > +     /*
> > +      * Devices with quirky firmware will stall or reset when the initial
>
> Has anyone actually seen a device that stalls?
>


Regarding the "stall" wording, I was referring to the condition where
they fail with -EPIPE error when these devices are connected through
external usb hubs. The directly connected devices I've tested
typically fail with -EPROTO instead.  I'm happy to reword the comment
if you think that would make it clearer.


> Even if no device currently uses both quirks, I think this belongs
> after the msleep() below, to minimize interaction between the quirks.
>
> >               if (dev->quirks & USB_QUIRK_DELAY_INIT)
> >                       msleep(200);
> >


Alright I'll address this change in next revision.

For the remaining comment and log message wording, I'd prefer to keep
the current version for now unless there's a strong objection. The
patch has already gone through several review rounds and I'd like to
avoid extending the review cycle further over comment phrasing,
especially since the implementation itself now seems to be in good
shape.

Thanks,
Nikhil Solanke