On 30 Jan 1997 16:04:14 -0800, Steven L Baur <steve@miranova.com> said:
> ...
>> However, having said all that, we've been bitten by races in this part
>> of the code before, so if you _can_ demonstrate a real race condition
>> in the VFS itself, please do let me know...
> I don't know how I can demonstrate a problem, but something is
> definitely still amiss. I just got bitten by this bug again a few
> hours ago after eight days of using the patch in this thread and
> 2.0.28:
> Jan 30 13:04:06 altair kernel: EXT2-fs warning (device 16:02): ext2_free_inode: bit already cleared for inode 286988
> Jan 30 13:04:06 altair last message repeated 2 times
> The total lossage this time was zeroed inodes 286,985 through 286,988
> and 364,586 through 364,588. :-(
Indeed --- I doubted very much that the patch would actually fix
anything. The fact that you are seeing such regular patterns of
corruption, and that they are only appearing on one of your three
drives, really does suggest that it is a case of borderline hardware
problems rather than software. I have seen similar sporadic patterns
of bitmap corruption in the past myself, and it has always turned out
to be hardware related.
If all of your disks are IDE, then you could at least start to narrow
down the problem by moving the problem disk to the other IDE channel.
Cheers,
Stephen.