Message-Id: <44EADA09.76E4.0078.0@novell.com>
Date: Tue, 22 Aug 2006 10:18:49 +0200
From: "Jan Beulich" <jbeulich@novell.com>
To: "Andrew Morton" <akpm@osdl.org>, "Andi Kleen" <ak@suse.de>
Cc: "J. Bruce Fields" <bfields@fieldses.org>, <linux-kernel@vger.kernel.org>,
       "Randy.Dunlap" <rdunlap@xenotime.net>
Subject: Re: boot failure, "DWARF2 unwinder stuck at 0xc0100199"
References: <20060820013121.GA18401@fieldses.org>
 <20060821094718.79c9a31a.rdunlap@xenotime.net>
 <20060821212043.332fdd0f.akpm@osdl.org> <200608221001.36124.ak@suse.de>
In-Reply-To: <200608221001.36124.ak@suse.de>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1126
Lines: 31

>> Has anyone even tried to reproduce Bruce's crash?
>
>I looked at it a bit, but it puzzles me. The chaining for the interrupt stacks
>on i386 -- which is what seems to be corrupted here -- shouldn't have changed at all 
>by the unwinder changes.

Not necessarily:

	if (UNW_SP(&info))
		stack = (void *)UNW_SP(&info);

is rather fragile - the minimum extra protection here should be to only use
UNW_SP() for the continuation stack pointer if it actually points into kernel
space (as is being done in one of the 2.6.19 patches), ...

>I suspect it would crash without unwinder too. Bruce, do you get the 
>same crash when you boot with "call_trace=old" ? 

... but of course I continue to agree that doing things like

	addr = *stack++;

in the legacy stack trace code cannot be good, given that this code
generally is expected to run when things are already bad in some way.

Jan
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/