Bug 140453

Summary: fault in sysfs_release in kernel-xen
Product: [openSUSE] SUSE LINUX 10.0 Reporter: Charles Coffing <ccoffing>
Component: XenAssignee: Gerd Hoffmann <kraxel>
Status: VERIFIED FIXED QA Contact: Jason Douglas <jdouglas>
Severity: Normal    
Priority: P5 - None CC: jbeulich
Version: unspecifiedKeywords: Built
Target Milestone: ---   
Hardware: x86-64   
OS: SuSE Linux 10.0   
Whiteboard:
Found By: Other Services Priority:
Business Priority: Blocker: ---
Marketing QA Status: --- IT Deployment: ---
Attachments: fixing a possible reason

Description Charles Coffing 2005-12-20 23:52:26 UTC
I can cause a fault (in user context) on the (updated) kernel‑xen kernels in both 10.0 and 10.1.  I've done this same test on XenSource's 2.6.12.6 and it
doesn't fault.

1.  Create a VM.
2.  As root:  udevinfo ‑a ‑p /sys/bus/xen‑backend/drivers/vbd/vbd‑2‑769/
3.  udevinfo is terminated; look at the fault in /var/log/messages.


Dec 20 16:09:10 ccoffing3 kernel: general protection fault: 0000 [1] SMP
Dec 20 16:09:10 ccoffing3 kernel: CPU 2
Dec 20 16:09:10 ccoffing3 kernel: Modules linked in: ipt_physdev iptable_filter ip_tables af_packet bridge button battery ipv6 ac usbhid ide_cd cdrom ohci1394 ieee1394 pcnet32 mii snd_intel8x0 snd_ac97_codec snd_ac97_bus snd_pcm snd_timer snd i2c_nforce2 i2c_core ehci_hcd generic ohci_hcd usbcore forcedeth soundcore snd_page_alloc parport_pc lp parport dm_mod reiserfs fan thermal processor sata_nv libata amd74xx sd_mod scsi_mod ide_disk ide_core
Dec 20 16:09:10 ccoffing3 kernel: Pid: 9651, comm: udevinfo Not tainted 2.6.13-15.7-xen
Dec 20 16:09:10 ccoffing3 kernel: RIP: e030:[<ffffffff801cd374>] <ffffffff801cd374>{sysfs_release+84}
Dec 20 16:09:10 ccoffing3 kernel: RSP: e02b:ffff880159021ed8  EFLAGS: 00010216
Dec 20 16:09:10 ccoffing3 kernel: RAX: 36372f322f6464f6 RBX: 36372f322f646276 RCX: 0000000000000034
Dec 20 16:09:10 ccoffing3 kernel: RDX: 0000000000000000 RSI: ffffffff80235510 RDI: ffff8801ae40893c
Dec 20 16:09:10 ccoffing3 kernel: RBP: ffff8801afd5f900 R08: 0000000000001008 R09: 00002aaaaabc5c00
Dec 20 16:09:10 ccoffing3 kernel: R10: 0000000000000004 R11: ffffffffff577080 R12: ffff8801ac5f2138
Dec 20 16:09:10 ccoffing3 kernel: R13: ffff8801b03e2f00 R14: ffff8801ac5f2138 R15: ffff880152225530
Dec 20 16:09:10 ccoffing3 kernel: FS:  00002aaaaadf46d0(0000) GS:ffffffff804e2480(0000) knlGS:ffffffff804e2480
Dec 20 16:09:10 ccoffing3 kernel: CS:  e033 DS: 0000 ES: 0000
Dec 20 16:09:10 ccoffing3 kernel: Process udevinfo (pid: 9651, threadinfo ffff880159020000, task ffff8801ac5a1820)
Dec 20 16:09:10 ccoffing3 kernel: Stack: 0000000000000010 0000000000000010 ffff8801ac9f7cc0 ffffffff80188f92
Dec 20 16:09:10 ccoffing3 kernel:        ffff8801ac5f2138 ffff8801ac9f7cc0 0000000000000000 ffff8801acb6eac0
Dec 20 16:09:10 ccoffing3 kernel:        ffff8801ac9f7cc0 000000000050e880
Dec 20 16:09:10 ccoffing3 kernel: Call Trace:<ffffffff80188f92>{__fput+194} <ffffffff80185fe8>{filp_close+104}
Dec 20 16:09:10 ccoffing3 kernel:        <ffffffff80186657>{sys_close+151} <ffffffff80112c72>{system_call+134}
Dec 20 16:09:10 ccoffing3 kernel:        <ffffffff80112bec>{system_call+0}
Dec 20 16:09:10 ccoffing3 kernel:
Dec 20 16:09:10 ccoffing3 kernel: Code: ff 08 83 3b 02 75 0c 48 8b bb 90 05 00 00 e8 19 83 f6 ff 48
Dec 20 16:09:10 ccoffing3 kernel: RIP <ffffffff801cd374>{sysfs_release+84} RSP <ffff880159021ed8>
Comment 1 Greg Kroah-Hartman 2005-12-21 22:59:15 UTC
looks like a xen specific issue...
Comment 2 Gerd Hoffmann 2005-12-22 09:44:39 UTC
Jan?  IIRC you've mentioned on some list you know that one ...
Comment 3 Jan Beulich 2005-12-22 09:56:22 UTC
Created attachment 61679 [details]
fixing a possible reason

I didn't observe the bug, I just noticed two warnings during the build that pointed out a merge problem, which possibly is responsible for the observed problem. I asked Charles to test out whether the problem would be gone with the patch.
Comment 4 Charles Coffing 2005-12-22 21:19:39 UTC
Yes, that patch fixes the problem on the 10.1 (HEAD) kernel.  Check it in...

The 10.0 kernel is still broken on Xen, regardless of the patch, however...
Comment 5 Jan Beulich 2005-12-23 08:12:07 UTC
Sure, that patch shouldn't even apply on the 10.0 tree. The latter has a similar problem, but in a different place. I'll get a patch ready once I checked in the current set of changes from Charles/KY.
Comment 6 Jan Beulich 2005-12-23 10:14:23 UTC
Patches checked in to HEAD and SL100_BRANCH.
Comment 7 Lynn Bendixsen 2006-01-12 16:11:25 UTC
Marking as resolved fixed because Jan checked in the patches. 10.1 has been built with these fixes in it. (tag 10.1-2)