Linux SMP kernel bug with > 512M ram

Linux SMP kernel bug with > 512M ram

[jeff]

I've been having problems with a few of our servers and I can't seem to find this 
problem mentioned anywhere else. All of the dual processor machines will not operate 
with greater than 512 megs of ram with the newer SMP kernels (2.4.7-10enterprise #1 
SMP). Two of the dual P3 1ghz machines crash after a few minutes, when the memory 
usage gets high enough, I presume. The errors they spit out vary, but its only when 
I go over 512megs of ram, and only on dual processor machines. I had a slightly 
different problem when I tried to set it up on a dual p2 266 machine, when I go over 
512 megs there, the system takes an hour to boot up, and everything crawls from 
there. I asked a friend of mine to try this newer kernel with his dual processor 
server, and he says the same thing (when I go over 512, it crashes). Has anybody had 
this problem? Is there a fix?

Regards,

Jeffrey Moss
jeff@americom.com

Re: Linux SMP kernel bug with > 512M ram

[Martin J. Bligh]

Works fine for me with up to 32Gb of RAM. Can you try a vaguely
current kernel?

Martin.

--On Friday, September 06, 2002 4:55 PM +0000 jeff@AmeriCom.com wrote:

> 
> I've been having problems with a few of our servers and I can't seem to find this 
> problem mentioned anywhere else. All of the dual processor machines will not operate 
> with greater than 512 megs of ram with the newer SMP kernels (2.4.7-10enterprise #1 
> SMP). Two of the dual P3 1ghz machines crash after a few minutes, when the memory 
> usage gets high enough, I presume. The errors they spit out vary, but its only when 
> I go over 512megs of ram, and only on dual processor machines. I had a slightly 
> different problem when I tried to set it up on a dual p2 266 machine, when I go over 
> 512 megs there, the system takes an hour to boot up, and everything crawls from 
> there. I asked a friend of mine to try this newer kernel with his dual processor 
> server, and he says the same thing (when I go over 512, it crashes). Has anybody had 
> this problem? Is there a fix?
> 
> Regards,
> 
> Jeffrey Moss
> jeff@americom.com
> 
> 
> 
> 
> 
> 
> -
> 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/
> 
> 

Re: Linux SMP kernel bug with > 512M ram

[Alan Cox]

On Fri, 2002-09-06 at 17:55, jeff@AmeriCom.com wrote:
> I've been having problems with a few of our servers and I can't seem to find this 
> problem mentioned anywhere else. All of the dual processor machines will not operate 
> with greater than 512 megs of ram with the newer SMP kernels (2.4.7-10enterprise #1 

2.4.7 is hardly "new". Red Hat has issued errata kernels going up to
2.4.9.

> SMP). Two of the dual P3 1ghz machines crash after a few minutes, when the memory 
> usage gets high enough, I presume. The errors they spit out vary, but its only when 
> I go over 512megs of ram, and only on dual processor machines. I had a slightly 

Chipset or memory hardware problems seem the most likely cause if the
errors seem random or weird

> different problem when I tried to set it up on a dual p2 266 machine, when I go over 
> 512 megs there, the system takes an hour to boot up, and everything crawls from 

Thats BIOS. Thats a well known BIOS problem where the BIOS doesnt
configure the mtrr registers properly. You can work around that one by
tweaking the settings by hand or probably by getting a newer BIOS 

Re: Linux SMP kernel bug with > 512M ram

[jeff]

I don't think its chipset or memory, the machines that crash have different brand 
motherboards with different chipsets, I ran docmem for 24 hours on each stick of ram 
and found no errors. The ram worked fine in my WindowsXP machine, and it works fine 
when I use the non-smp kernel, and/or when I take the ram down to 2 sticks (512 
meg). I'm posting here because I believe I have narrowed it down to a bug in the 
kernel.

> On Fri, 2002-09-06 at 17:55, jeff@AmeriCom.com wrote:
> > I've been having problems with a few of our servers and I can't seem to find 
this 
> > problem mentioned anywhere else. All of the dual processor machines will not 
operate 
> > with greater than 512 megs of ram with the newer SMP kernels (2.4.7-10enterprise 
#1 
> 
> 2.4.7 is hardly "new". Red Hat has issued errata kernels going up to
> 2.4.9.
> 
> > SMP). Two of the dual P3 1ghz machines crash after a few minutes, when the 
memory 
> > usage gets high enough, I presume. The errors they spit out vary, but its only 
when 
> > I go over 512megs of ram, and only on dual processor machines. I had a slightly 
> 
> Chipset or memory hardware problems seem the most likely cause if the
> errors seem random or weird
> 
> > different problem when I tried to set it up on a dual p2 266 machine, when I go 
over 
> > 512 megs there, the system takes an hour to boot up, and everything crawls from 
> 
> Thats BIOS. Thats a well known BIOS problem where the BIOS doesnt
> configure the mtrr registers properly. You can work around that one by
> tweaking the settings by hand or probably by getting a newer BIOS 
> 
> 
> 

Re: Linux SMP kernel bug with > 512M ram

[David S. Miller]

   From: jeff@AmeriCom.com
   Date: 6 Sep 2002 20:50:26 -0000

   I'm posting here because I believe I have narrowed it down to a bug in the 
   kernel.
   
And Alan also said there are errata upgrades available that
quite possibly will cure your problems.

Re: Linux SMP kernel bug with > 512M ram

[jeff]

Its a VIA motherboard, and I found the problem:
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=47160
Thanks for the help.

Regards,

Jeffrey Moss
jeff@americom.com



> Message-ID: <005a01c255ec$89be7b40$d281f6cc@WEASEL>
> From: "Steve Wolfe" <nw@codon.com>
> To: <jeff@AmeriCom.com>
> References: <fa.hqf5vev.1u6g19e@ifi.uio.no> <fa.hqeru5v.1v6u3h8@ifi.uio.no>
> Subject: Re: Linux SMP kernel bug with > 512M ram
> Date: Fri, 6 Sep 2002 15:29:29 -0600
> MIME-Version: 1.0
> 
> 

> 
> > I don't think its chipset or memory, the machines that crash have
> different brand
> > motherboards with different chipsets, I ran docmem for 24 hours on each
> stick of ram
> > and found no errors. The ram worked fine in my WindowsXP machine, and it
> works fine
> > when I use the non-smp kernel, and/or when I take the ram down to 2
> sticks (512
> > meg). I'm posting here because I believe I have narrowed it down to a
> bug in the
> > kernel.
> 
>    I've also been bit in the rear-end by memory bugs, and I can assure
> you, they can be devilish to find.  First, rather than running a memory
> test on each individual stick, you MUST run the memory test on the actual
> system, with the actual RAM.  Often times, weird, bizarre errors crop up
> in certain combinations and not in others.
> 
>   It's also a little difficult to convince people that it's a bug in the
> kernel, when there is an incredibly number of people who run SMP kernels
> with >512 MB.  I maintain a decent number of such machines, and have seen
> absolutely *zero* problems of similar natures that weren't traced back to
> defective or incompatible hardware.
> 
>    I'm curious - what kinds of motherboards are you using, and what kinds
> of RAM?  What kind of memory timings are specified in the BIOS?
> 
> steve
> 
>