Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 21 Oct 1997 01:37:21 +0000 (GMT)
From:      Terry Lambert <tlambert@primenet.com>
To:        bde@zeta.org.au (Bruce Evans)
Cc:        freebsd-current@FreeBSD.ORG, j@uriah.heep.sax.de
Subject:   Re: lockmgr panic
Message-ID:  <199710210137.SAA24623@usr08.primenet.com>
In-Reply-To: <199710160509.PAA08005@godzilla.zeta.org.au> from "Bruce Evans" at Oct 16, 97 03:09:25 pm

next in thread | previous in thread | raw e-mail | index | archive | help
> >Speaking of bandaids, would anybody mind me committing Mr. KATO's
> >fix/hack?  (Of course, unless he's going to commit it himself.)  The
> >patch seems to solve the vn(4) lockmgr problem.
> 
> I think it substitutes panics with deadlock possibilities.  Perhaps
> the deadlocks are rare.
> 
> >I haven't made a release on a 2.2.5-BETA system lately.  If the
> >problem there is the same, i would vote for moving the fix there as
> 
> I think the problem is the same but the symptoms are worse - endless
> recursion and at best a panic for a double fault.

The recursion is only one deep.  The only failure case I could sww was
if you were paging to the vn device that was locked.  That's can't happen
*AND* hit that code path, since that code path requires that a disklabel
and a filesystem be present on the vn device (and therefore not swap).

You *might* be able to trigger it by vnconfiging a vnode device on an
fs on a vnode device;  I haven't followed the code path in that case.  I
think that particular case is really unlikely.

See my analysis following the original posting of the patch; I think
the patch is good, and the problem is in the way the locking in VOP_ADVLOCK
is implemented as a call-down interface.


					Terry Lambert
					terry@lambert.org
---
Any opinions in this posting are my own and not those of my present
or previous employers.



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199710210137.SAA24623>