Re: usb2: scanner no longer working

On Thursday 13 November 2008, Stefan Ehmann wrote:
> On Friday 07 November 2008 19:35:17 Hans Petter Selasky wrote:
> > On Friday 07 November 2008, Stefan Ehmann wrote:
> > > On Friday 07 November 2008 18:42:49 Hans Petter Selasky wrote:
> > > > On Friday 07 November 2008, Hans Petter Selasky wrote:
> > > > > Hi,
> > > > >
> > > > > Could you dump the current config descriptor of your scanner?
> > > > >
> > > > > usbconfig -u 3 -a 2 dump_curr_config_desc
> > >
> > > # usbconfig -u 3 -a 2 dump_curr_config_desc
>
> <snip>
>
> > > > 3 and 2 are the numbers after ugen, like ugen3.2
> > >
> > > To your other mail:
> > > I'm running i386. HUB debugging didn't output any obvious errors.
> >
> > Hi,
> >
> > Try the following patch to libusb20. I suspect that it is the
> > set_configuration call that makes trouble!
> >
> > http://perforce.freebsd.org/chv.cgi?CH=152628
> >
> > Thanks for reporting.
> >
> > My private SVN repository has also been updated to include this patch if
> > you are using that.
>
> As I previously stated in my other mails. The scanner is now properly
> recognized.
>
> Now I got this problem:
> Before scanning starts, the system "locks" for several minutes, i.e.
>  - input/output to terminals is no longer working
>  - I can still switch consoles and break into ddb
>  - CTRL-t on the scanimage console is still working and shows increasing
> sys time
>
> As soon as the actual scanning process starts, everything is fine again.
> Sometimes scanning doesn't start at all (or I'm just not patient enough).
>
> To make debugging a bit more interesting: If I run scanimage with ktrace or
> enable usb debugging via sysctl, scanning starts immediately. Don't know if
> this is always the case but seemed to work a few times at least. So maybe
> it's some kind of timing problem.

Strange!

Try this before starting your scanimage:

sysctl hw.usb2.ugen.debug=15

I guess that the program is looping on a poll-syscall, because I poll on both 
stdin and the I/O descriptor. This is perhaps another bug in libusb20.

--HPS