Date: Fri, 26 Mar 2004 14:06:04 -0500 From: John Baldwin <jhb@FreeBSD.org> To: Bruce Evans <bde@zeta.org.au> Cc: Lukas Ertl <le@FreeBSD.org> Subject: Re: panic: mutex Giant not owned Message-ID: <200403261406.04566.jhb@FreeBSD.org> In-Reply-To: <20040326114225.M39632@gamplex.bde.org> References: <20040325113619.N56656@pcle2.cc.univie.ac.at> <200403251527.27658.jhb@FreeBSD.org> <20040326114225.M39632@gamplex.bde.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thursday 25 March 2004 07:53 pm, Bruce Evans wrote:
> On Thu, 25 Mar 2004, John Baldwin wrote:
> > On Thursday 25 March 2004 05:39 am, Lukas Ertl wrote:
> > > Hi,
> > >
> > > I was dropped into DDB after a panic and tried to save a coredump with
> > > "call doadump", which resulted in a:
> > >
> > > panic: mutex Giant not owned in /usr/src/sys/cam/cam_xpt.c:3970
> > >
> > > Second call to "doadump" succeeded, but savecore(8) couldn't save the
> > > core ('bad dump header' or something like this).
> >
> > Cute. I should fix mtx_assert() to just bail and not panic if panicstr
> > is already set.
>
> Er, the panic is correct. Panic dump code must not use normal locking,
> since then it won't work if it hits an unlockable lock. panic() usually
> recursively panics earlier, near the sync() in boot(), if it hits an
> unlocakable lock. Changing mtx_assert() to not panic if panicstr is
> set would be worse for this than for the panic dump part, since sync()
> is normal file system code that depends on normal locking for correctness.
I misread it at first. mtx_assert() already doesn't recursively panic. The
current behavior does seem correct though annoying since you can't just take
an arbitrary dump it seems.
--
John Baldwin <jhb@FreeBSD.org> <>< http://www.FreeBSD.org/~jhb/
"Power Users Use the Power to Serve" = http://www.FreeBSD.org
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200403261406.04566.jhb>