[Jfs-discussion] mkdir's crash kernel

[Jasmin B. <ja...@pa...>]

Hi,

I tried to benchmark some Linux filesystems but the kernel crashed when
doing lots of mkdir's on a JFS filesystem.
I did this on kernel 2.6.21-rc4-git1, so I'm not sure it's a JFS
problem. The mkdir's were done in parallel by 2 forks.

Here's the dmesg:

Bad page state in process 'mkdir'
page:ffff810001b632c0 flags:0x4000000000000818 mapping:ffff810036385ca8
mapcount:0 count:0 Trying to fix it up, but a reboot is needed
Backtrace:

Call Trace:
[<ffffffff802a7707>] bad_page+0x57/0x90
[<ffffffff8020b8a0>] free_hot_cold_page+0x60/0x130
[<ffffffff882b17ef>] :jfs:diWrite+0x50f/0x540
[<ffffffff882c162a>] :jfs:txCommit+0x1fa/0x1010
[<ffffffff882a944e>] :jfs:jfs_mkdir+0x26e/0x370
[<ffffffff8022e50f>] mntput_no_expire+0x1f/0xd0
[<ffffffff8020f1f4>] link_path_walk+0xd4/0xf0
[<ffffffff802bd0f9>] vfs_mkdir+0x79/0xd0
[<ffffffff802bd669>] sys_mkdirat+0xb9/0x130
[<ffffffff80267f8d>] error_exit+0x0/0x84
[<ffffffff80261c1e>] system_call+0x7e/0x83
...

The rest of dmesg is attached. It seems some lines at the end went
missing (it's from a logfile).

Greetings,
Jasmin Buchert