[bug report Cc'ed to linux.dev.kernel]
Our main machine, picard, crashes two times a day with Linux 2.0.0.
We get OOPses, plain hangs and kernel panics. This machine was very
stable with the 1.99.x kernels.
These were written down by hand while the machine "hung" 2 days ago.
At that time it was running Linux-2.0.1-PRE2
ll_rw_block: Trying to read nonexistant block device 00:00 (206031)
[0 00:00 2060301]
Pressing Alt-Scrollock repeatedly:
EIP 00122952: sync_buffers
EIP 00122a47: sync_buffers
EIP 00122959: sync_buffers
EIP 00122974: sync_buffers
These were logged before the machine crashed last night, while running
Linux-2.0.1-PRE4
Jun 27 01:40:16 picard kernel: VFS: Close: file count is 0
Jun 27 01:40:16 picard kernel: Unable to handle kernel paging request at virtual
address c80d0096
Jun 27 01:40:16 picard kernel: current->tss.cr3 = 01b06000, Dr3 = 01b06000
Jun 27 01:40:16 picard kernel: *pde = 00000000
Jun 27 01:40:16 picard kernel: Oops: 0000
Jun 27 01:40:16 picard kernel: CPU: 0
Jun 27 01:40:16 picard kernel: EIP: 0010:[<0012b50f>] (sys_fcntl)
Jun 27 01:40:16 picard kernel: VFS: Close: file count is 0
Jun 27 01:40:16 picard last message repeated 5 times
Jun 27 01:40:16 picard kernel: general protection: 0000
Jun 27 01:40:16 picard kernel: CPU: 0
Jun 27 01:40:16 picard kernel: EIP: 0010:[<0012c968>] (locks_remove_locks)
Jun 27 02:30:04 picard kernel: EXT2-fs error (device 08:07): ext2_find_entry: ba
d entry in directory #232574: rec_len is smaller than minimal - offset=0, inode=
2
On other machines we see even weirder behaviour ("corrupt swap space"
errors and such) - backing down to 1.99.10 for now..
Mike.
-- Miquel van | Cistron Internet Services -- Alphen aan den Rijn. Smoorenburg, | mailto:info@cistron.nl http://www.cistron.nl/ miquels@het.net | Tel: +31-172-419445 (Voice) 430979 (Fax) 442580 (Data)