a memory-related problem?

a memory-related problem?

[Ronald Bultje]

Hi,

I just added 128 MB of RAM to my machine which already had 128 MB and
which has 128 MB swap. 128 MB RAM + 128 MB swap (either the new or the
old 128 MB RAM) works, but the combination of that, 256 MB RAM + 128 MB
swap, crashes the compu during startup with either an "unresolved-
symbols in init" message (which is completely random, each boot shows
different unresolved references) or with oopses right after starting
init.

(btw is it useful to write down these oopses on paper, reboot with
only 128 MB RAM to ksymoops them?)

Any idea on what's wrong? I thought the memory could be broken but
when booting with only the new memory in the machine, it does startup
so I suppose the memory itself is okay (but I'm not a hardware expert). 
Or is this "normal behaviour" when a machine doesn't follow swap=2xRAM? 
(and then again, why did it startup with 128MB/swap?)

distro = RedHat-7.0 (with updates from redhat.com installed)
Kernel = 2.4.4
gcc = 2.96
glibc = 2.2

system = p-II 400 MHz, 128 MB swap, 440BX (abit p6b) mainboard
memory is (133 MHz) SDRAM memory (running at 100 MHz)

[please CC me, I'm currently not subscribed]

Regards,

Ronald Bultje

Re: a memory-related problem?

[Colonel]

In clouddancer.list.kernel, you wrote:
>
>Hi,
>
>I just added 128 MB of RAM to my machine which already had 128 MB and
>which has 128 MB swap. 128 MB RAM + 128 MB swap (either the new or the
>old 128 MB RAM) works, but the combination of that, 256 MB RAM + 128 MB
>swap, crashes the compu during startup with either an "unresolved-
>symbols in init" message (which is completely random, each boot shows
>different unresolved references) or with oopses right after starting
>init.

It's more likely that the two RAM sticks differ.  I had a similar
problem in a Windoze machine awhile ago.  Move one stick of RAM into
another bank, i.e. with 4 memory slots, use #1 and #3.  If you only
have 2 memory slots, return/sell what you have and buy 2 sticks at the
same time.


--My pid is Inigo Montoya.  You killed -9 my parent process.  Prepare to vi. 

Re: a memory-related problem?

[Steven Walter]

Probably what happens is that your BIOS stores some data in the top
megabyte of RAM, but doesn't set up the memory map to reflect this.
Therefore, Linux overwrites whatevers up there, causing problems.

On Sun, Jun 17, 2001 at 09:26:50PM +0200, Ronald Bultje wrote:
> P6b has three mem-slots. I would get "unresolved errors in init" if I
> had 2x64+1x128 sticks, and I would get oopses if I had 2x128M sticks. So
> there is indeed a weird difference.
> I just noticed this: if I supply "linux-2.4.4 mem=255M" instead of
> "linux-2.4.4 mem=256M" at the lilo prompt, it does work. Is this a bug
> in the code that handles options given at startup-time? (I only tried
> this for 2x128 sticks but I suppose this is the same for 2x64+1x128
> sticks - I guess I'm too lazy to try it out).
-- 
-Steven
In a time of universal deceit, telling the truth is a revolutionary act.
			-- George Orwell

Re: a memory-related problem?

[Alan Cox]

> I just noticed this: if I supply "linux-2.4.4 mem=255M" instead of
> "linux-2.4.4 mem=256M" at the lilo prompt, it does work. Is this a bug
> in the code that handles options given at startup-time? (I only tried
> this for 2x128 sticks but I suppose this is the same for 2x64+1x128
> sticks - I guess I'm too lazy to try it out).

Its common for the BIOS to reserve and use the top 1K or so

Re: a memory-related problem?

[Ronald Bultje]

On 17 Jun 2001 06:10:02 -0700, Colonel wrote:
> In clouddancer.list.kernel, you wrote:
> >
> >Hi,
> >
> >I just added 128 MB of RAM to my machine which already had 128 MB and
> >which has 128 MB swap. 128 MB RAM + 128 MB swap (either the new or the
> >old 128 MB RAM) works, but the combination of that, 256 MB RAM + 128 MB
> >swap, crashes the compu during startup with either an "unresolved-
> >symbols in init" message (which is completely random, each boot shows
> >different unresolved references) or with oopses right after starting
> >init.
> 
> It's more likely that the two RAM sticks differ.  I had a similar
> problem in a Windoze machine awhile ago.  Move one stick of RAM into
> another bank, i.e. with 4 memory slots, use #1 and #3.  If you only
> have 2 memory slots, return/sell what you have and buy 2 sticks at the
> same time.

P6b has three mem-slots. I would get "unresolved errors in init" if I
had 2x64+1x128 sticks, and I would get oopses if I had 2x128M sticks. So
there is indeed a weird difference.
I just noticed this: if I supply "linux-2.4.4 mem=255M" instead of
"linux-2.4.4 mem=256M" at the lilo prompt, it does work. Is this a bug
in the code that handles options given at startup-time? (I only tried
this for 2x128 sticks but I suppose this is the same for 2x64+1x128
sticks - I guess I'm too lazy to try it out).

Regards,

Ronald Bultje

Re: a memory-related problem?

[Andreas Bombe]

On Sun, Jun 17, 2001 at 02:37:03PM +0200, Ronald Bultje wrote:
> system = p-II 400 MHz, 128 MB swap, 440BX (abit p6b) mainboard
> memory is (133 MHz) SDRAM memory (running at 100 MHz)

The question is, does it configure your SDRAMs correctly?  I assume it's
on auto config, then the BIOS has to figure out what timings to use for
the RAMs by reading their SPD ROM.

Problem 1: you run PC133 memory in PC100 configuration.  Some PC133
DIMMs contain no configuration data for PC100, so they theoretically
couldn't legally be run as PC100.  They could be run using the PC133
config - not optimal, but should work.  The BIOS has to figure that out,
however.  This problem is more likely to occur if you're using no-name
RAMs instead of branded RAMs by a reputable vendor (you might not even
have fully valid config data for PC133 in that case).

Problem 2: your BIOS is broken and configures all DIMMs to the
configuration values of the first, causing at least one to run with the
wrong values, if they have different timing requirements.  Given your
problem (either DIMM works alone, two don't), this sounds like a more
possible candidate.

-- 
Andreas E. Bombe <andreas.bombe@munich.netsurf.de>    DSA key 0x04880A44