On Tue, 2002-05-21 at 08:45, Tang Lingbo (Allan) wrote:
> Yes. I made the CONFIG_XFS_DEBUG and CONFIG_VNODE_TRACING enable.
If you turn this on then you get a lot of verbose error reporting
from the tests which run fsstress. fsstress feeds random parameters
into system calls, including random inode numbers into bulkstat - which
results in attempts to read random inodes from places which do not
exist on the disk.
I think vnode tracing is slightly broken right now and does not free all
its memory correctly. Which leads to a slabcache which will not
free up on module unload. If you do not use the debug options both
symptoms will disappear. The XFS_DEBUG option is useful as it turns
on lots of consistancy checks, and fsstress is designed to stress
these, but the vnode tracing code will only help you if you are
doing kernel debugging.
Steve
>
> > -----Original Message-----
> > From: Stephen Lord [mailto:lord@xxxxxxx]
> > Sent: Tuesday, May 21, 2002 8:06 PM
> > To: Tang Lingbo (Allan)
> > Cc: Linux-Xfs
> > Subject: Re: xfstests/017 report cmn_err
> >
> >
> > On Tue, 2002-05-21 at 04:21, Tang Lingbo (Allan) wrote:
> > > Hi,
> > >
> > > I got the "cmn_err" while doing test with xfstests/017, it was
> > checked out
> > > yesterday from cvs.
> > >
> > > May 21 16:00:25 allan kernel: cmn_err level 1 Filesystem
> > "ide0(3,9)": xfs_diloca
> > > te: agno (2893) >= mp->m_sb.sb_agcount (8)
> >
> > Are you running with CONFIG_XFS_DEBUG enabled?
> >
> > Steve
> >
> > >
> > > The test passed without error message on console.
> > >
> > > the rpms are:
> > > acl-2.0.11
> > > attr-2.0.8
> > > xfsprogs-2.0.5
> > > xfsdump-2.0.3
> > > xfs-1.1 patch
> > >
> > > This error will lead other messages when I try to 'rmmod xfs'.
> > > Anything I missed?
> > >
> > > Allan
> >
> >
--
Steve Lord voice: +1-651-683-3511
Principal Engineer, Filesystem Software email: lord@xxxxxxx
|