Re: [PROBLEM] 2.6.21-git oops

[Srihari Vijayaraghavan <sriharivijayaraghavan@yahoo.com.au>]

--- Michal Piotrowski <michal.k.k.piotrowski@gmail.com> wrote:
> Hi Srihari,

Hello Michal,

[...]

> > It turns out the system is no go with 4 GB of RAM (4 * 1 GB modules):
> system
> > instability & SATA controller doesn't detect the drives (as I reported to
> > linux-ide:
> > http://www.mail-archive.com/linux-ide@vger.kernel.org/msg06174.html).
> >
> > With only 2 GB of RAM (with either pairs), system is very stable
> (surviving
> > many parallel kernel compiles for over 3 times longer than that of 4 GB),
> and
> > it detects & works fine with the SATA drives also.
> >
> > Any advice on getting all of 4 GB to work reliably with SATA also?

[...]

> This looks like a bad RAM problem too me. Check the RAM memory
> http://www.memtest.org/

I disagree it to be a bad RAM problem (please read on). But agree undeniably
to run the memtest a little later on (FWIW, Linux & kernel compile tests are
indeed my test case to prove the stability of the machine).

(I can assure no component is overclocked; I prefer stability over performance
with instability :-). Power & ventilation seem quite adequate, but nonetheless
I haven't eliminated yet - for stability, not for SATA detection.)

Assuming it's indeed a bad RAM (which I don't sincerely/logically believe),
may I ask then how would you explain:
1. SATA drives aren't even detected with 4 GB of RAM
2. Or the system working reliably with _both_ pairs of memory modules
individually, ie when limited to 2 GB only, when kernel compile torture test
succeeds for many hours with no problems (repeated the test case for a few
times also to prove the memory modules are themselves in very good condition),
while crashing in a few minutes reliably/repeatably under 4 GB

Logically speaking, if I were to hypothesize what the problem might be:
1. mother board is no go with 4 GB of RAM (for both system stability & SATA
drives detection) - Escalated it to Asus to confirm.
2. (my dearest) Linux kernel doesn't configure/program things in this board to
work reliably.
3. Sadly I don't know yet if there is a boot parameter for Linux or a
configuration setting in BIOS to make things work both reliably & completely.

(I vaguely remember in a thread a few weeks/months ago in LKML on AMD64/Nvidia
chipset combo, using SW IOMMU lead to improved stability on machines with
large amount of RAM?? Or was that limiting the RAM to under 4 GB that helped??
I could be completely wrong here.)

> (Removed from the list of known regressions
> http://kernelnewbies.org/known_regressions)

(Initially I believed it to be a regression wrongly, but it was not. Since
then I've eliminated a few variables in my tests. Sorry for my
misunderstanding.)

Thank you for your help so far.

Hari



Send instant messages to your online friends http://au.messenger.yahoo.com