Bug 5210
Summary: | crash under disk stress [+4kstacks], Unable to handle kernel paging request at virtual address 08080880 @ do_page_fault | ||
---|---|---|---|
Product: | File System | Reporter: | peter gervai (grin) |
Component: | XFS | Assignee: | Diego Calleja (diegocg) |
Status: | CLOSED CODE_FIX | ||
Severity: | normal | CC: | alexn, bunk |
Priority: | P2 | ||
Hardware: | i386 | ||
OS: | Linux | ||
Kernel Version: | 2.6.13 | Subsystem: | |
Regression: | --- | Bisected commit-id: | |
Attachments: |
Serial console capture
.config Oops with requested flag |
Description
peter gervai
2005-09-09 06:58:41 UTC
Created attachment 5947 [details]
Serial console capture
Hmpf, are you using CONFIG_4KSTACKS? Could you pass over your .config please Created attachment 5974 [details]
.config
Yes, this machine happens to use 4Kstacks. Config attached.
(I'll try to crash without 4kstacks, reporting back a bit later.)
I switched off 4kstacks. I cannot freeze the machine anymore. :-/ Okay, nice hint, thank you very much; seems to be fixed (I go and switch off 4kstacks everywhere right now). I still wonder why, but that's just out of curiousity... The combination of xfs and device-mapper plus sata (I don't know how stack-hungry sata is, but it does use the scsi layer). This is unfortunate, but I'm sure there are people who are interested in this. What would be very nice is if you could turn on CONFIG_4KSTACKS and turn on "Stack utilization instrumentation" under "Kernel hacking". That should give us a trace of what makes the stack overflow (trace will be on the console). Either that or, describe your environment so that someone else can reproduce the problem. Thanks I'll try to get a stack util dump while I'm in the crashing mood. (It is fun to do on a remote server *wink* *wink* [makes operators finally work for their money].) BTW crash happened without lvm and sata, too. Maybe it's just harder to reproduce, I don't know, since I didn't try bonnie++ until I got the new sata drive, and freezes were intermittent. Created attachment 5979 [details]
Oops with requested flag
Well, I hope it's helpful, stack trace seems to shoot itself in the foot... I
can switch on any debugging flag (as long as I don't need further magic than
the serial console) if it'd help...
As far as I'm concerned, by now XFS has solved their stack issues and should be safe to use with CONFIG_4KSTACKS, so I'm closing this bug. If you can reproduce this problem with the latest stable kernel version, please reopen it. |