Skip to contentSkip to conversation

PROBLEM: 2.4.4 oops, will not boot

DISCUSSION Gordon Sadler 1 participant 1 message May 3, 2001 1 message not in the archive

lore
Share & tools
lore
#1 lore raw
In reply to a message not in the archive
Quoted 45 lines · Andrew Morton

From: Andrew Morton <[email protected]>
X-Mailer: Mozilla 4.76 [en] (X11; U; Linux 2.4.3-ac13 i686)
To: Gordon Sadler <[email protected]>
Subject: Re: PROBLEM: 2.4.4 oops, will not boot

Quoted 27 lines · Gordon Sadler
Quoted 25 lines · Andrew Morton
Quoted 13 lines · Gordon Sadler
Quoted 11 lines · Andrew Morton
Quoted 5 lines · Gordon Sadler

Please CC on replies.
Attached is REPORTING-BUGS template from source, and a hand copied oops that I ran through ksymoops. I really hope this is resolved, anything further needed, just ask.

Unfortunately the ksymoops output doesn't show the call trace. Can you please try again? A reproducable oops is, err, rather important. The syntax to feed into ksymoops is

Quoted 7 lines
Quoted 4 lines
Call Trace: [<c0111234>] [<c0123456>] ...

Thanks.

Right, I see the problem now.. s/Calltrace/Call Trace/

Reran oops through ksymoops with change to Call Trace, output attached.

Ugh. Something seems to have corrupted your slab cache data structures, so this will be real hard to pin down.

Quoted 9 lines
Quoted 3 lines

You could try setting DEBUG to 1 in mm/slab.c, see if that catches the culprit.

As requested.. changed config in one way, removed nvidia framebuffer/replaced with vesa fb. After changeing DEBUG -> 1 in mm/slab.c it produced 3 EIPs. I rebooted to 2.2.19, copied by hand, and ran through ksymoops. The attached tar is the .config, hand copy of 3 EIPs, and the ouput from ksymoops.

Thanks for all the hard work :)

Quoted 8 lines

Something funny is happening at your end. It's dying when it's trying to return a memory mapping control structure back to the memory management pool. This code hasn't changed in a while, and if it was wrong, we'd be hearing it from lots of people.

Have you done a full rebuild? `make clean'?

This last build of 2.4.4 with mod to mm/slab.c was a fresh unpack from a src tarball from www.kernel.org. From what little I have gleaned from this list, it appears a few others are having similar problems.. specifically I hear AMD CPU + Epox 8kta3 m/b will not boot with 3dnow enabled? Some others have reported minor sucess with :
CONFIG_MK7=y
CONFIG_X86_USE_3DNOW=n

or they just compile with :
CONFIG_M686=y

I haven't seen any other ksymoops results from those individuals however, and Alan Cox made a tentative analysis, it only happens with VIA chipset -> VIA chipset bug.

I'm at a loss... I tried 2.4.4ac[23] both of them freeze during boot as well, hmm... small idea: My BIOS allows FSB of 100/133 auto or manual setting. I'm using pc100 SDRAM and the BIOS id's it as such at boot, any chance the kernel is somehow trying to set something as though it were 133 FSB?

Gordon Sadler