|
From: Alan S. <st...@ro...> - 2010-06-24 16:55:29
|
On Thu, 24 Jun 2010, Ludovic Rousseau wrote: > 2010/6/24 Alan Stern <st...@ro...>: > > On Thu, 24 Jun 2010, Ludovic Rousseau wrote: > > > >> Hello, > >> > >> I try to stress my code by sending the same command on a loop. But I > >> may be hit by a kernel bug or libusb bug. > >> I am using libusb git version up-to-date and kernel 2.6.34-1-686 or > >> vmlinuz-2.6.32-5-686 (from Debian). Both kernels have the problem. > >> > >> libusb returns -99 (LIBUSB_ERROR_OTHER) and the kernel logs a line in > >> /var/log/syslog: > >> Jun 24 15:08:01 trinity kernel: [ 299.676480] usb 3-1.3: usbfs: > >> usb_submit_urb returned -121 > >> > >> I also used my previous kernel 2.6.26-2-686 from Debian stable and > >> libusb-1.0.8 and I DO NOT have the problem. So it looks a bug in the > >> recent kernels or in the way libusb is using the kernel. > > > > It is a bug in libusb. It is caused by interaction with a kernel > > feature that was added not long ago. I forget when exactly, but after > > 2.6.26. > > An already known bug. Cool. That's not what I meant. I'm not aware if this bug was known previously -- certainly I never knew about it. And if Daniel knew, he would have fixed it already. Your description made it clear what the source of the bug must be. That's all I was trying to say. Alan Stern |