Re: Problem with today's -current

[Andrew Doran <ad%netbsd.org@localhost>]

On Thu, Jan 24, 2008 at 09:09:05PM +0000, Andrew Doran wrote:

> On Thu, Jan 24, 2008 at 11:28:51AM -0800, Paul Goyette wrote:
> 
> > Built a new set of kernels right after Andy's latest commits to vfs 
> > stuff, and now I'm getting lots of the following console messages:
> > 
> > vrelel: missing VOP_CLOSE(): vnode @ 0xffff800048f80d80, flags 
> > (30<MPSAFE,LOCKSWORK>)
> >         tag VT_UFS(1), type VREG(1), usecount 1, writecount 0, holdcount 1
> >         freelisthd 0x0, mount 0xffff8000045cb000, data 0xffff800048f360f0
> >         tag VT_UFS, ino 811439, on dev 0, 0 flags 0x11, effnlink 1, nlink 1
> >         mode 0100444, owner 0, group 0, size 837 lock type vnlock: EXCL 
> >         (count 1) by pid 451.1
> > 
> > I'm also getting similar mesages for NFS files, like this:
> > 
> > vrelel: missing VOP_CLOSE(): vnode @ 0xffff8000496499d0, flags (0)
> >         tag VT_NFS(2), type VREG(1), usecount 1, writecount 0, holdcount 1
> >         freelisthd 0x0, mount 0xffff80000472f000, data 0xffff800049271348
> >         tag VT_NFS, fileid 230463 fsid 0xb01
> > 
> > 
> > Interestingly enough, this seems to be happening only on my port-amd64 
> > machine; a port-i386 configured with similar kernel options does not 
> > complain about this.
> 
> Ok, these are just spam and can be ignored. I fixed the assertion so if it
> fires again should be catching real problems. Although, a missing VOP_CLOSE
> is likely to be "mostly harmless" and doesn't mean the machine is going down
> in flames.

sys/kern/vfs_subr.c 1.321

Andrew