svn commit: r190655 - head/sys/kern

[Chagin Dmitry]

On Fri, Apr 03, 2009 at 12:44:32PM +0300, Kostik Belousov wrote:
> On Fri, Apr 03, 2009 at 10:05:53AM +0400, Chagin Dmitry wrote:
> > On Thu, Apr 02, 2009 at 09:16:20PM +0000, Peter Wemm wrote:
> > > Author: peter
> > > Date: Thu Apr  2 21:16:20 2009
> > > New Revision: 190655
> > > URL: http://svn.freebsd.org/changeset/base/190655
> > > 
> > > Log:
> > >   vn_vptocnp() unlocks the name cache and forgets to re-lock it before
> > >   returning in one error case, and mistakenly unlocks it for the
> > >   umount -f case.
> > > 
> > 
> > vn_vptocnp() consumers expects what in case of error returns
> > w/o CACHE_RLOCK held. so you need something like:
> Yeah, I do not quite understand this commit.
> 
> > 
> > diff --git a/sys/kern/vfs_cache.c b/sys/kern/vfs_cache.c
> > index ea19e13..b06a6c6 100644
> > --- a/sys/kern/vfs_cache.c
> > +++ b/sys/kern/vfs_cache.c
> > @@ -971,18 +971,17 @@ vn_vptocnp(struct vnode **vp, char **bp, char *buf, u_int *buflen)
> >  	vdrop(*vp);
> >  	VFS_UNLOCK_GIANT(vfslocked);
> >  	if (error) {
> > -		CACHE_RLOCK();
> >  		numfullpathfail2++;
> >  		return (error);
> >  	}
> Yes, this is reversal of the part of the r190655.
> 
> >  	*bp = buf + *buflen;
> >  	*vp = dvp;
> > -	CACHE_RLOCK();
> >  	if ((*vp)->v_iflag & VI_DOOMED) {
> >  		/* forced unmount */
> >  		vdrop(*vp);
> >  		return (ENOENT);
> >  	}
> > +	CACHE_RLOCK();
> >  	vdrop(*vp);
> >  
> >  	return (0);
> There, r190655 should be reversed too, instead of doing what you do.
> VI_DOOMED should be checked while cache is locked.
> 

whoops, it was fast hacking,
therefore make universe panicked after rebase to r190655
thnx :)

-- 
Have fun!
chd
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 196 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/svn-src-all/attachments/20090403/a436233a/attachment.pgp