kernel panic with a 2nd 1GB RAM module [SOLVED]

[MasterX]

This is one thing that drives me crazy. I own a laptop with 1GB RAM ( at 533MHz), and I decided to install a second 1GB (533MHz) module. The Bios recognizes both modules and reports that I have 2048KB memory. Windows XP does the same thing. Memtest reports that my system has 2048KB memory RAM and there are free of errors.

I am running kernel 2.6.28-4, and I compiled the kernel with Highmem support. During boot, the system loads the modules and then I get an error about not being able to execute the /etc/init.d/module script. After that point is downhill and all the scripts fail to start. The good thing is that the system does not hang, and I get the command prompt. I run dmesg, and I see the same message over and over again. This message reports a kernel panic. Now, I have disable the highmem support, the system boots fine but the kernel reports only 1024KB memory RAM. I have tried booting the system with a single memory module, and it boots fine for both modules.

The memory modules two differ little bit. As I wrote above they both run at 533MHz, but they have a different number of chips on them. The old module has 16chips, and the new one has 8 chips.

I found on the Internet something about memory remapping, but the Bios does not have this option.

Besides compiling the kernel with Highmem support, what else should I do?

Thank you in advance!!

[fangorn]

Besides having half the number of chips, it is likely that the layout is different. So using the two modules together may or may not work. In your case I'd say "may not" is more likely.

Have you run a multiple hour memory torture test with both modules installed?
Did it survive multi-day runtime session in the other Operating systems?

With RAM it is a good idea to have exactly the same module for both modules. Else you have to select very carefully to get a stable system.

Memory remapping is important when having 4 GB of RAM. Because PCI bus gets mapped in the space below 4 GB, RAM between approx. 3.5 and 4 GB is not addressible otherwise. That is of no concern for machines below 3 GB.
_________________
Video Encoding scripts collection | Project page

[MasterX]

[fangorn]

I am not aware of any limit between 1 and 2 GB RAM (in kernel versions younger than five years that is). My guess was, that the BIOS does not address them correctly.

I use machines between 4 and 12 GB RAM and never had to activate something to make it work. I had problems with the memory between 3 and 4 GB from time to time, but that was because of crappy BIOSes, not because of the linux kernel.

My tip still is: get two identical modules.

I'm sorry that I could not be of more help.
_________________
Video Encoding scripts collection | Project page

[Ant P.]

Try running them at 400MHz. Sometimes the mobo itself can only handle a single stick of RAM at full speed.

[MasterX]

[Ant P.]

That's probably a bug in memtest itself. Odd that it could run overnight without errors but Linux crashes, that sounds like a kernel problem to me. Have you tried using the 64GB highmem setting instead of the 4GB one? If nothing else works, you could try one of the kernel patchsets that let you use 2GB without highmem enabled...

[eccerr0r]

Are you compiling your own kernel? Does it work with a precompiled kernel, say, the Gentoo livecd/livedvd?
Post your .config somewhere?
_________________
Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching?

[MasterX]

[energyman76b]

forget using more than 1gb on a 32bit machine?

because even if you could use the full 2gb - the slowdown makes it a questionable choice at best.
_________________
Study finds stunning lack of racial, gender, and economic diversity among middle-class white males

I identify as a dirty penismensch.

[MasterX]

[energyman76b]

yes, it is possible to use more than 2gb and yes and pcs are sold with more than 1gb.

That does not mean it is a smart choice. HP is selling pcs with 3gb - and dual channel memory controllers, it can't get more stupid like that.

If you want to access more than 1gb, you need highmem support. Highmem support means slowdown. That simple.
_________________
Study finds stunning lack of racial, gender, and economic diversity among middle-class white males

I identify as a dirty penismensch.

[eccerr0r]

[MasterX]

[MasterX]

Today, I decided to give a try to the latest vanilla source (2.6.32-rc3).

During the kernel configuration I changed several things from the old configuration. Among them, I enabled:

[MasterX]

[eccerr0r]

I cannot discern the slowdown with PAE enabled. I need to have it enabled to access all 4GB of RAM on my two 32 bit installs. (My ia64 box supports 4G directly and no need for PAE though I have a feeling it's emulated anyway for emulated x86 code.)

I've heard people see a 10% slowdown using PAE, but I don't recall what conditions these were. The worst case would be where you have to bank swap every context swap. I think it should be possible to code away from this, but if you must use a lot of memory with both processes (or go to kernel mode a lot, as in i/o, but i/o speeds tend to make PAE switch speeds seem fast), this would be the pathological case.

If PAE were as slow as ISA memory from years ago (which incurred a 70%+ penalty to access isa), it'd be noticeably slower... It's not at all that slow.
_________________
Intel Core i7 2700K/Radeon R7 250/24GB DDR3/256GB SSD
What am I supposed watching?