VFS: grow_buffers:, MD, RAID-O problem.

Christian Balzer (cb@aichan.swb.de)
Thu, 12 Dec 1996 10:42:25 +0100


I'm running a MD RAID-0 stripe set for two identical 2GB disks which
hold the news spool here:
/dev/md0 raid0,8k,0,7627ab92 /dev/sdd1 /dev/sde1

Usually this works like a charm is a real speed improvement over
individual disks, including the additonal benefit of being able to
use hardlinks.
However under heavy load often (but not reproducable) the following
will happen:

Dec 11 06:41:06 aichan kernel: VFS: Wrong blocksize on device 08:61
Dec 11 06:41:06 aichan kernel: VFS: grow_buffers: size = 805307392
Dec 11 06:41:37 aichan last message repeated 2161 times
[etc, until a reboot, the device will be busy and thus unclean, however
it will not be corrupted when fsck'ed.]

I'm really out of my league trying to analyze the MD and buffer parts of the kernel for this problem, anybody out there who might be able to shed some light on this? Trying to contact the original MD author failed, BTW. What happens when kernal parts are no longer officially maintained anyway?

Thanks in advance to all who reply,


  // <CB> aka Christian Balzer, Tannenstr. 23c, D-64342 Seeheim, Germany
\X/  CB@brewhq.swb.de | Voice: +49 6257 83036, Fax/Data: +49 6257 83037
SWB  - The Software Brewery - | Team H, Germany HQ | Anime no Otaku