[Bug 403346] New: libusb and/or kernel cause complete hang

newer
[Bug 436055] New: OpenSUSE 11. 1...

bugzilla_noreply＠novell.com

24 Jun 2008 24 Jun '08

17:04

https://bugzilla.novell.com/show_bug.cgi?id=403346 Summary: libusb and/or kernel cause complete hang Product: openSUSE 11.0 Version: Final Platform: i686 OS/Version: openSUSE 11.0 Status: NEW Severity: Major Priority: P5 - None Component: Kernel AssignedTo: bnc-team-screening@forge.provo.novell.com ReportedBy: stefan.neis@kobil.com QAContact: qa@suse.de Found By: --- We have some hardware which is mostly treated as USB-attached CDRom-drive, however for some special purposes we need to _not_ access it via the usb_storage driver, but rather via libusb. So we call usb_detach_kernel_driver_np to get access to the device. This works nicely with all SuSE-10.x versions (and also some current Ubuntu versions). However, testing it on openSuSE 11.0 results in a complete hang, the system never returns from the call to usb_detach_kernel_driver_np (or at least not within a reasonable amount of time). The system is completely unresponsive at that time, it doesn't react to keyboard or mouse input, and doesn't react to network traffic either (not even a reply to a ping). Searching for possibly related stuff I found e.g. http://www.mail-archive.com/debian-bugs-closed@lists.debian.org/msg187645.ht... and applied the patch given there to the libusb source-rpm, compiled, and installed the modified version. This somewhat improved the situation, now usb_detach_kernel_driver_np succeeds, and the software proceeds to a first call to usb_interrupt_read, which however still results in the same kind of hang. :-( ATM, my only idea is that somewhere in the USB part of the kernel, an endless loop might be hidden that's similar to the one fixed by that debian-bug mentioned above for libusb. Thanks for any ideas/suggestions on what to do here... -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

Show replies by date

bugzilla_noreply＠novell.com

29 Jun 29 Jun

11:27

New subject: [Bug 403346] libusb and/or kernel cause complete hang

https://bugzilla.novell.com/show_bug.cgi?id=403346 Cyril Hrubis changed: What |Removed |Added ---------------------------------------------------------------------------- AssignedTo|bnc-team-screening@forge.provo.novell.com |kernel-maintainers@forge.provo.novell.com -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

bugzilla_noreply＠novell.com

1 Jul 1 Jul

15:47

New subject: [Bug 403346] libusb and/or kernel cause complete hang

https://bugzilla.novell.com/show_bug.cgi?id=403346 User stefan.neis@kobil.com added comment https://bugzilla.novell.com/show_bug.cgi?id=403346#c1 --- Comment #1 from Stefan Neis 2008-07-01 09:47:16 MDT --- With some help from the USB mailing list (particularly Alan Stern), we collected a bit more information. An "strace" of our program shows this:

...

ioctl(3, USBDEVFS_SUBMITURB, 0xbf8566d4) = 0 ioctl(3, USBDEVFS_REAPURBNDELAY, 0xbf856718) = -1 EAGAIN (Resource temporarily unavailable) select(4, NULL, [3], NULL, {0, 1000}) = 0 (Timeout) write(1, "test\n", 5test) = 5 gettimeofday({1214911004, 481477}, NULL) = 0 ioctl(3, USBDEVFS_REAPURBNDELAY, 0xbf856718) = -1 EAGAIN (Resource temporarily unavailable) select(4, NULL, [3], NULL, {0, 1000}) = 0 (Timeout) write(1, "test\n", 5test) = 5 gettimeofday({1214911004, 485465}, NULL) = 0 ioctl(3, USBDEVFS_REAPURBNDELAY, 0xbf856718) = -1 EAGAIN (Resource temporarily unavailable) select(4, NULL, [3], NULL, {0, 1000}

i.e. it's hanging in a select call, which Alan commented like this:

...

It's clear now that the problem exists in the kernel, not in your program on in libusb. To make any more progress you will have to patch the kernel source. The function you want to investigate first is usbdev_poll() in drivers/usb/core/devio.c.

We didn't start that yet, however, we tried a different kernel version and found that we have the same problem with 2.6.25.9 from kernel.org, so it seems (IMHO) to be a regression introduced in 2.6.25 (since a test on an Ubuntu with 2.6.24.whatever showed no problem) and not yet fixed in the current version. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.