YONETANI Tomokazu qhwt+dfly at
Wed Jul 8 15:03:51 PDT 2009

On Tue, Jul 07, 2009 at 05:39:02PM -0700, Matthew Dillon wrote:
> :Running this *online* couldn't find any 'B', so I guess I have to
> :boot it with a USB memory stick and try it again?
> :
> :..
> :I'll also try memtest after that.  Thanks for the hint.
>     It's possible that HAMMER caught the problem before it wrote the
>     sector back to disk, in which case it may only have gotten corrupted
>     in-memory and the filesystem on-media might be ok.
>     You can validate the entire filesystem offline (or if it is completely
>     idle) with:
> 	hammer -f <device> show | egrep '^B'
> 	(that's carrot and B, not control-B)
>     If no lines are marked bad in the show output then the filesystem is
>     clean.  This can take a while to run since it reads every single last
>     bit of meta-data and data in the filesystem.  If any hammer cleanup
>     or reblocking op runs while the show is running the show command
>     will blow up.

Ok, this (with daily_clean_hammer_enable=NO in /etc/periodic.conf) took me
the whole night, but it gave me no B's.

>     The most likely scenario is that there's bad ram.

I haven't had time to run memtest for the obvious reason :)
I'll do it later today.

More information about the Kernel mailing list