linux-mips
[Top] [All Lists]

Re: crash in first printk of start_kernel

To: Ralf Baechle <ralf@linux-mips.org>
Subject: Re: crash in first printk of start_kernel
From: Attila Kinali <attila@kinali.ch>
Date: Wed, 14 Mar 2007 09:04:27 +0100
Cc: linux-mips@linux-mips.org
In-reply-to: <20070313010252.GB26119@linux-mips.org>
Organization: NERV
Original-recipient: rfc822;linux-mips@linux-mips.org
References: <20070309191354.f962e57b.attila@kinali.ch> <20070313010252.GB26119@linux-mips.org>
Sender: linux-mips-bounce@linux-mips.org
Hoi Ralf,

On Tue, 13 Mar 2007 01:02:52 +0000
Ralf Baechle <ralf@linux-mips.org> wrote:

> On Fri, Mar 09, 2007 at 07:13:54PM +0100, Attila Kinali wrote:
> > I'm using a 2.6.16.11 (an old snapshot from about last august,
> > when we started development of another board) that has slight
> > adjustments in various drivers to accomodate for our platform
> > specific stuff.
> 
> You may want to update anyway.  between the linux-2.6.16.11 tag of the
> linux-mips.org and the top of the linux-2.6.16-stable branch there are
> almost 58,000 lines of patch.  Even more if your compare the MIPS -stable
> branch to kernel.org's 2.6.16.11.  Iow a few metric buttloads.

That's already planned. But if first have to get past those
dead lines. After that i can look into makeing the whole
build system upgradeable and managable-

 
> > 0xffffffff80266134 <vsnprintf+76>:      beqz    v0,0x80266144 <vsnprintf+92>
> > 0xffffffff80266138 <vsnprintf+80>:      bltz    a0,0x802461c0 
> > <jffs2_remount_fs+144>
> 
> A branch in the delay slot of another branch is forbidden by the MIPS
> architecture.  All processors I every tried this on missbehave in very
> unobvious ways when this is attempted.
> 
> You may want to compare that against your vmlinux file.  If the vmlinux
> binary also contains this bug, try building the affected source file with
> -S to find if the bug is cause by compiler or assembler.

It turned out to be a ground bounce problem. Interestingly the
"bug" was 100% reproducable, while normale ground problems are
totaly random. Thus i thought it has to be something in the
software.

> In single stepping mode your debugger probably executes branches by
> software emulation.  Chances are the emulation does something different
> for this illegal code sequence than actual hardware.

Oh.. nice to know. Thanks.
 
Thanks a lot for your answer and help,

                        Attila Kinali
-- 
Praised are the Fountains of Shelieth, the silver harp of the waters,
But blest in my name forever this stream that stanched my thirst!
                         -- Deed of Morred

<Prev in Thread] Current Thread [Next in Thread>