On Fri, Jul 06, 2018 at 01:52:57PM +0200, Landry Breuil wrote:
> On Thu, Jul 05, 2018 at 11:13:42AM +0200, Landry Breuil wrote:
> > On Thu, Jul 05, 2018 at 11:04:34AM +0200, Martin Pieuchot wrote:
> > > On 05/07/18(Thu) 09:28, Landry Breuil wrote:
> > > > On Fri, Jun 29, 2018 at 01:50:24PM +0200, Landry Breuil wrote:
> > > > > On Fri, Jun 29, 2018 at 09:05:56AM +0200, Remco wrote:
> > > > > > > > If you don't find any better solution, I'd suggest using a 
> > > > > > > > device ID
> > > > > > > > check rather than adding a quirk.  Because such quirk cannot be 
> > > > > > > > generic.
> > > > > > > > In that case you have an off-by-one, but another device might 
> > > > > > > > have a
> > > > > > > > different problem.
> > > > > > > 
> > > > > > > That was in the case we'd encounter other devices where the 
> > > > > > > descriptor
> > > > > > > is bogus and has the same issue, but i agree this is unlikely.
> > > > > > > 
> > > > > > 
> > > > > > More likely for Logitech USB webcams I'm afraid.
> > > > > > My dad's Logitech C500 showed the same issue several years back.
> > > > > > (I'll try to find some time to borrow the device and test your 
> > > > > > patch)
> > > > > > There's also a report that suggests a C200 may be affected as well.
> > > > > > (https://marc.info/?l=openbsd-misc&m=127952275021290&w=2)
> > > > > 
> > > > > Aaah, very interesting data point. Thanks for the pointer.. maybe a
> > > > > 'generic fix' would be to ignore size checks for all logitech webcams,
> > > > > or to print a warning but not error out in this case. Is there some 
> > > > > kind
> > > > > of 'usbdevs -v/lsusb -v' database somewhere ?
> > > > 
> > > > So after feedback from lots of logitech webcam users (thanks!), i got a 
> > > > list
> > > > for working vs non-working devices, and i think a new quirk would do as 
> > > > they
> > > > all expose the same issue (bLength for FEATURE_UNIT is 9 while it is 8 
> > > > on
> > > > working devices) - and they would all be 'fixed' by the attached diff.
> > > 
> > > If you know that some descriptors have an off-by-one, I'd suggest you add
> > > 1 to the corresponding buffer length instead of completely removing the
> > > check.  Removing it completely now open the door to other overflows and
> > > adding 1 auto-document the hardware bug ;)
> > 
> > Well that was the intent of the original diff. New version...
> 
> New diff adding the C250 which also exhibits the same issue. I have an
> okay from mpi@ on the previous diff, but i'd like actual reports that it
> makes the audio device actually working (mine doesnt now, i'm digging
> into it, but i'm pretty sure at some point i had it working...)
> 
> To test:
> 
> sysctl kern.audio.record=1
> aucat -f rsnd/1 -o /tmp/foo.wav (if the uaudio device attaches as audio1)
> 
> If this actually records sound (mplayer /tmp/foo.wav to check), you've
> won. if foo.wav is 68 bytes long, then recording doesnt work, interrupts
> are triggered but no transfers happen..

I should add to this testing part that you should 'revert' to usb2 for
proper uaudio(4) testing, usb3/xhci doesnt really work in this usecase
and will trigger 'error creating pipe: err=INVAL endpt=0x86' errors in
dmesg.

If you can, ensure that you use usb2 by disabling usb3 in the bios or
xhci in the kernel config.

Landry

Reply via email to