2001-10-02 09:23:16

by Ryan Mack

[permalink] [raw]
Subject: [BUG] 2.4.10/0.9.10 "VFS: brelse: Trying to free free buffer"

Unfortunately I don't have an oops or a stack trace to report where this
message came from, but at least maybe this report can alert someone who
can spot the bug in the code. This occurred under light usage on an ext3
partition built on a 2 disk RAID-1 mirrored array.

The array was either two IDE disks using data-ordered mode or two SCSI
disks using data-journal mode.

Relevant files attached.

Thanks, Ryan Mack

Attachments:
dmesg (13.51 kB)
 ver_linux (723.00 B)
 Download all attachments

2001-10-02 17:02:12

by Andrew Morton

[permalink] [raw]
Subject: Re: [BUG] 2.4.10/0.9.10 "VFS: brelse: Trying to free free buffer"

Ryan Mack wrote:
>
> Unfortunately I don't have an oops or a stack trace to report where this
> message came from, but at least maybe this report can alert someone who
> can spot the bug in the code. This occurred under light usage on an ext3
> partition built on a 2 disk RAID-1 mirrored array.
>
> The array was either two IDE disks using data-ordered mode or two SCSI
> disks using data-journal mode.
>

Yes, that shouldn't be happening. If it is reproducible, could
you please add

print_buffer_trace(buf);

to the very end of fs/buffer.c:__brelse() and recompile with
CONFIG_BUFFER_TRACE=y?

Thanks.