Immediate general protection errors on Tyan board

Immediate general protection errors on Tyan board

[Bob Richmond]

Immediately upon boot on this system, most userland programs will 
segfault, including mount. This causes the system to come up in a 
bizarre state with the root filesystem mounted read-only, and nothing 
runs without segfault. There have been numerous similar posts about this 
problem, but they also seem to point to an associated kernel message, 
"Bad page state" that I don't observe. dmesg (which runs without 
segfault) returns many similar messages to:

start_udev[576] general protection rip:2aaaaae0fc70 rsp:7fffffb23d90 error:0

This occurrs on the latest Fedora Core 4 kernel RPM 
(kernel-2.6.12-1.1447_FC4) on a Tyan S2892 motherboard with 2 Opteron 
246 processors. The BIOS is up to date with what is available on Tyan's 
website for this board.

I'm not sure if the Fedora maintainers have rolled in the latest AMD 
errata changes from 2.6.13 in the latest package, but that will be my 
next test.

Does anyone still observe this problem in their systems?

Re: Immediate general protection errors on Tyan board

[Parag Warudkar]

Bob Richmond wrote:

> Immediately upon boot on this system, most userland programs will 
> segfault, including mount. This causes the system to come up in a 
> bizarre state with the root filesystem mounted read-only, and nothing 
> runs without segfault. There have been numerous similar posts about 
> this problem, but they also seem to point to an associated kernel 
> message, "Bad page state" that I don't observe. dmesg (which runs 
> without segfault) returns many similar messages to:
>
> start_udev[576] general protection rip:2aaaaae0fc70 rsp:7fffffb23d90 
> error:0

echo 0 > /proc/sys/kernel/randomize_va_space - Seems to fix it for most 
people.

See http://bugzilla.kernel.org/show_bug.cgi?id=4851 for more details.

Parag

Re: Immediate general protection errors on Tyan board

[Sander]

Parag Warudkar wrote (ao):
> Bob Richmond wrote:
> >Immediately upon boot on this system, most userland programs will 
> >segfault, including mount. This causes the system to come up in a 
> >bizarre state with the root filesystem mounted read-only, and nothing 
> >runs without segfault. There have been numerous similar posts about 
> >this problem, but they also seem to point to an associated kernel 
> >message, "Bad page state" that I don't observe. dmesg (which runs 
> >without segfault) returns many similar messages to:
> >
> >start_udev[576] general protection rip:2aaaaae0fc70 rsp:7fffffb23d90 
> >error:0
> 
> echo 0 > /proc/sys/kernel/randomize_va_space - Seems to fix it for most 
> people.
> 
> See http://bugzilla.kernel.org/show_bug.cgi?id=4851 for more details.

I only had some programs segfault, but this got resolved by doing an
update of the OS (Debian). It also started after an update of Debian.

Both programs installed with Debian packages as self-compiled programs
crashed with the ".. general protection rip: .." and ".. segfault at
..".

This happened on a dual Opteron Tyan system running 2.6.12-rc5.

	Kind regards, Sander

-- 
Humilis IT Services and Solutions
http://www.humilis.net

Re: Immediate general protection errors on Tyan board

[Bob Richmond]

I just got it working over the weekend. The machine had an Adaptec 29320 
64-bit PCI card driving a SCSI HD. It was removed, and the drive 
replaced with a Serial ATA drive, and it came up fine.

There were probably a lot of variables that were changed by removing the 
card, but I'm wondering if anyone has experienced the same symptoms with 
any 64-bit PCI card installed on this board.

Parag Warudkar wrote:

> Bob Richmond wrote:
> 
>> Immediately upon boot on this system, most userland programs will 
>> segfault, including mount. This causes the system to come up in a 
>> bizarre state with the root filesystem mounted read-only, and nothing 
>> runs without segfault. There have been numerous similar posts about 
>> this problem, but they also seem to point to an associated kernel 
>> message, "Bad page state" that I don't observe. dmesg (which runs 
>> without segfault) returns many similar messages to:
>>
>> start_udev[576] general protection rip:2aaaaae0fc70 rsp:7fffffb23d90 
>> error:0
> 
> 
> echo 0 > /proc/sys/kernel/randomize_va_space - Seems to fix it for most 
> people.
> 
> See http://bugzilla.kernel.org/show_bug.cgi?id=4851 for more details.
> 
> Parag