Re: kern/47817: procfs related panic

[David Holland <dholland-bugs%netbsd.org@localhost>]

The following reply was made to PR kern/47817; it has been noted by GNATS.

From: David Holland <dholland-bugs%netbsd.org@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: 
Subject: Re: kern/47817: procfs related panic
Date: Mon, 7 Oct 2013 03:07:09 +0000

 On Wed, May 15, 2013 at 08:10:01AM +0000, nick@zoom.localdomain wrote:
  > panic: kernel diagnostic assertion "vp->v_usecount != 0" failed: file 
"/usr/src/sys/kern/vfs_vnode.c", line 924 
 
 This panic means that a vnode with no references has been vref'd,
 which you aren't supposed to do (it requires vget instead). I think
 either this is a manifestation of one of the vnode lifecycle
 conditions that hannken@ fixed recently (the one where vget could in
 some circumstances return a blank vnode) ... or it indicates a
 vnode refcounting problem in mounts or cwd/chroot logic. The latter is
 possible of course but doesn't seem too likely.
 
 Since the call chain goes through procfs_readlink and procfs_dir, the
 vnode it's tripping on must be the target process's current dir, root
 dir, or executable image... or a mount point getcwd crosses while
 getcwd'ing one of these. It can't be a procfs vnode, as procfs refuses
 to do that. (See PR 40549.)
 
 I suspect it's the blank vnode issue and that this problem will be no
 longer reproducible -- was it ever reproducible?
 
 Another option is that it's somehow a manifestation of the same
 problem as 40549, but I don't immediately see how.
 
 -- 
 David A. Holland
 dholland%netbsd.org@localhost