On Thu, Jan 02, 2020 at 10:06:33AM -0500, Alan Stern wrote:
> On Thu, 2 Jan 2020, Takashi Iwai wrote:
> > On Thu, 02 Jan 2020 12:20:45 +0100, Johan Hovold wrote:
> > > Note that we also have this little gem in the ftdi usb-serial driver
> > > (since 2009) overriding a zero max packet size for devices with broken
> > > descriptors:
> > >
> > > 895f28badce9 ("USB: ftdi_sio: fix hi-speed device packet size calculation")
> > >
> > > Note sure how common those are but they will no longer work after the
> > > new sanity check in core. I guess we could add quirks for them (to core)
> > > in case we get any reports, but perhaps reverting the check should be
> > > considered.
> >
> > FWIW, Roger confirmed that reverting the commit d482c7bb0541 does
> > indeed fix the issue (with the latest 5.4.y kernel).
>
> All right. Suppose instead of reverting that commit, I change the code
> so that it only logs a warning when it finds an endpoint descriptor
> with maxpacket = 0 (and it skips the warning for isochronous endpoints
> in altsetting 0). At the same time, we can add a check to
> usb_submit_urb() to refuse URBs if the endpoint's maxpacket is 0.
>
> Sounds good?
Sounds good to me.
Just make sure not to add a WARN() in usb_submit_urb() so that we end up
having to add maxpacket checks to every USB driver when syzbot starts
hitting this (only driver's doing maxpacket divisions or similar should
need that).
Johan
