ufs_hashlock mutex in lfs_vget() (lfs_vfsops.c)

For background information, see PR/33631 as well as various
posts to
the mailing lists.

I've been having problems for quite a while now with
frequent lfs
deadlocks.  As of -current of ~one day ago, I built a kernel
(i386)
and got two deadlocks while using LFS within a few hours of
each
other.  After I applied the LFS part of the patch I sent in
for this
problem about two months ago, I've yet to have another
deadlock.
*fingers crossed*

Does anyone have any comments on whether or not this patch
should be applied:

<http://bahar.aydogan.net/~blair/diffs/lfs_vfsops.c_di
ff>

Regards,

--Blair

On Mon, Jul 23, 2007 at 04:57:04AM -0400, Blair Sadewitz
wrote:

> For background information, see PR/33631 as well as
various posts to
> the mailing lists.
> 
> I've been having problems for quite a while now with
frequent lfs
> deadlocks.  As of -current of ~one day ago, I built a
kernel (i386)
> and got two deadlocks while using LFS within a few
hours of each
> other.  After I applied the LFS part of the patch I
sent in for this
> problem about two months ago, I've yet to have another
deadlock.
> *fingers crossed*
> 
> Does anyone have any comments on whether or not this
patch should be 
> applied:
> 
> <http://bahar.aydogan.net/~blair/diffs/lfs_vfsops.c_di
ff>

I checked this one in, thanks. I've been working on
abstracting the inode
hash so that it's a function of genfs. Hopefully then we can
eliminate a lot
of pasteware and fix the problem across the board. I'll send
you the diffs
when I get a chance so you can have a look.

Cheers,
Andrew