Too Much System Memory?

[SkipHuffman]

My NVIDIA GeForce 2 Ultra was working just fine ( so far as I could tell) until I added some system memory. I went from 256 MB to 1.25GB Now when I run starttx the screen goes black, then a small (3mmx5mm?) red box appears in the upper left corner of the screen with a smaller (2mmx2mm) green box flashing in the middle of it. and my system is hard crashed, No CTRL-C, no CTRL-ALT-F2 , no CTLR-ALT-DEL. Dead, dead, dead.

I tried adding the NVAGP option to my XF86Config-4. with a parameter of 0 or 2 the failure continues. With a parameter of 1 x will start, but kde still has a problem. The first time I tried it, the screen went grey with an X in the middle, them blue, then nothing, hang, but a CTRL-ALT-F2 let me get to a new terminal sesion so I could kill the processes. I rebooted to be sure everthing was cleared out and retried, this time my desktop came up, except for the tool bar. But only from root. From a regular user, I just get the Blue screen.

Oh, details of my system may be important :

Athlon 1.3GZ
ASUS A7V133 Motherboard
2 x512MB Micron DIMM(Added)
1x256MB Micron DIMM (original)
Elsa Gladiac Video
Linux RedHat 7.3 Latest updates as of a week ago. Kernel upgraded to 2.4-18

Any assistance would be appreciated.

Thank you,

Skip Huffman

[bwkaz]

Does Ctrl-Alt-Backspace work? That's the "kill X" key combination, and sometimes it will work when other things don't. But not always.

Anyway, that problem when you use agpgart (when NvAgp is 3 or 2) seems rampant here lately. Nobody has gotten a solution either, except one guy got it to work when he set his AGP to 1x in the BIOS (I think that was the case, anyway). You might want to try that first. You're getting the same problem at NvAgp of 0? That makes no sense, since that setting disables AGP... Oh, IIRC, they were all using RedHat kernels as well. Is there anything on RedHat's web support area about problems like this? I'm starting to wonder if they screwed up AGP with their kernel patches.

Do you feel up to installing a standard kernel, version 2.4.19? If that one works, then the problem is most likely something like that...

[SkipHuffman]

>Does Ctrl-Alt-Backspace work? That's the "kill X" key
>combination, and sometimes it will work when other things
>don't. But not always.

I did not know that key combo. I am an old hand at PCs (DOS
and OS/2, thirteen years), but new to Linux. I will try
that next time.


>Anyway, that problem when you use agpgart (when NvAgp is 3
>or 2) seems rampant here lately. Nobody has gotten a
>solution either, except one guy got it to work when he set
>his AGP to 1x in the BIOS (I think that was the case,
>anyway). You might want to try that first.

I will do that. Easy enough, but my AGP has been at 4x for
over a year without this sort of issue.

>You're getting
>the same problem at NvAgp of 0? That makes no sense, since
>that setting disables AGP...

Its possible that I tried that without doing a full reboot
between tests, but I thought that I did.

>Oh, IIRC, they were all using
>RedHat kernels as well. Is there anything on RedHat's web
>support area about problems like this? I'm starting to
>wonder if they screwed up AGP with their kernel patches.

I don't use the Red Hat kernel.

>Do you feel up to installing a standard kernel, version
>2.4.19? If that one works, then the problem is most likely
>something like that...

I use 2.4.19 I have to use a custom kernel. Red Hat, in
their infinite wisdom, chooses to compile their kernel
without Token Ring support (Yes, I am a wierdo. I have used
T/R in some form for just this side of forever, so I still
use it. Also I got the NICs and MAU for free.)

Thank you for responding.

Skip

[Klaus-P]

It seems you should check your memory options in your linux Kernel config file.
Check /usr/src/linux/.config or run 'make menuconfig' or 'make config' in that respective
directory. We had a similar sounding problem just a while ago here.
Some default kernels are set to work with 1 GB but not more unless recompiling them to work
properly with more than 1 GB.

Good luck, ... and who is really a 'super-expert' knowing all answers offhand? ;-)

[SkipHuffman]

I guess my subconcious decided that I did not have enough to do this weekend. I have just totally screwed my system. Somehow I smashed my root file system. I think that I managed to remount my root filesystem as FAT. (I have a special partiiion set aside to transfer file from Linux to OS/2 and back and I must have picked the wrong one. ) Well, now my system won't boot and I can't get the root partition to pass fsck.

CD time.

Back in a couple of hours, or maybe tomorrow.

Growl grumble.

[bwkaz]

Oh man, ouch.

Well, I do know the feeling, though. I recently, on my firewall, made a full system backup (the system was 5gigs of Linux on a 20gig drive) into another partition that wouldn't normally be mounted. But I had to create this partition first.

So I remount / read-only, after shutting down a lot of the daemons that were running, and started up GNU parted. I didn't think to set the filesystem type back to ext2 first, as I didn't remember whether or not it made a difference, and assumed it didn't.

So I resize / down by like 10 gigs, resize the extended partition, move the logical partitions down inside it, and create a new 10gig partition at the end of the drive.

parted does everything just fine, except for one minor little detail. It forgets to update the "block count" in the superblock for my root FS. So of course, after a reboot, when it tried to fsck, fsck coughed and died because "the device is smaller than the filesystem". No matter how I tried to run fsck manually (like it says to), I couldn't get either it or parted to fix the problem, so I took the numbers for the current free block count and what it should have been, and opened up hexedit on the partition. Of course, I managed to screw up the change the first time, and make the filesystem about half the size it should have been, and then fsck screwed some things up, but once I realized the problem, it was easy enough to fix. One more fsck, and (*whew*) all is back to normal. No reinstall needed, which is good, because this was LFS, and LFS takes forever to install.

So I formatted the new partition and got my full backup. Then I umounted it and tucked it away somewhere really safe. ;)

Anyway, yeah, you're not the only one that can have issues with filesystems, so don't worry too much. Sometimes fsck can fix them, and sometimes it just isn't worth the effort -- especially when you have the CDs.

Good luck!

[SkipHuffman]

At this point i am convinced that I have been hit by an old fashioned MBR virus. Both my OS/2 and my Linux side are tanked. Well 15 years of PC ownership without a loss to virii was a pretty good streak.

Time to blow away the entire system, repartition, reformat, reinstall OSes and try to recover.

Backups? Me? No, I am too smart to need backups. (ordering backup stuff today.)

Thanks for the support.

[bwkaz]

Running a good old fdisk /mbr from a Windows boot disk should be able to at least clear out the MBR. Then you can boot with the Linux boot floppy you created during the install (you did do that, right? ;) ) and re-install your Linux bootloader (either lilo or grub).

Of course, if other things have broken then that won't work, but it might be worth a shot.