nForce 4 corrupting data written to HDD

[mdmbkr]

Is this patch supposed to fix both of the issues (as krader described in post 46)? I am experiencing the version of the issue that iommu=soft does not mask.

Does anyone know if the patch has been incorporated into Dapper? I am running up-to-date Dapper on the problem system right now.

[netllama]

This issue was later determined to be kernel bug, due to a specification violation, and was fixed up stream. You should not need to use the patch attached here.

[calestyo]

I think it was included in 2.6.21, so if your kernel has this or a higher version you should be fine.

[mdmbkr]

I still see the problem when running 2.6.22 amd64 smp.

I tried to reproduce the problem using large files generated from /dev/urandom but nothing works as consistently as this particular SQL file I have. I get a different md5sum almost every time.

I've tried 2.6.22 amd64 smp with and without iommu=soft. I've also tried disabling the iommu and memhole mapping in the bios, to no effect. I've also tried using mem=1g and/or iommu=soft with 2.6.15 amd64 smp, to no effect.

I'm compiling a uni-proc amd64 kernel right now. If that doesn't work I'm thinking of removing physical memory, down to 1gb, and testing that.

[calestyo]

Originally Posted by mdmbkr

I've tried 2.6.22 amd64 smp with and without iommu=soft. I've also tried disabling the iommu and memhole mapping in the bios, to no effect. I've also tried using mem=1g and/or iommu=soft with 2.6.15 amd64 smp, to no effect.

Then you're suffering from a different problem. In each case the bug described here was solved by iommu=soft.

Chris.

[mdmbkr]

Please see post #46.

[calestyo]

Ah, I see,.. you are refering to kraders problem,.. perhaps it would be best to open a new topic for this problem, as the bug that was originaly discussed in this thread is now considered to be fixed.

[mdmbkr]

Thanks for the response. I'm going to gather a little more info about my hardware and software config first.

[mdmbkr]

Is there any way to reproduce the iommu bug without using the disk? Others have mentioned the issue might affect heavy network traffic, but would that corruption make it all the way to the user, or would intermediate drivers notice and retry?

The iommu=soft workaround fixed one of my machines, but the other uses software raid1 for its root (and only) filesystem. I have reason to suspect the ongoing issues are related to that, so I'd like to isolate the disk from the testing process to see if the system is also affected by the iommu issue. If I have to, I can install another drive in that computer.

[mdmbkr]

OK, I installed another disk and was able to isolate two separate problems, one of which is fixed by iommu=soft. I'm still seeing corruption when using the raid1 volume, even with iommu=soft.

[calestyo]

Sorry, I didn't see your messages:

Originally Posted by mdmbkr

Is there any way to reproduce the iommu bug without using the disk?

None that I'd know.

Originally Posted by mdmbkr

Others have mentioned the issue might affect heavy network traffic, but would that corruption make it all the way to the user, or would intermediate drivers notice and retry?

Well I'm definitely not an expert in device drivers and so, but I'd think that network drivers wouldn't notice such errors.

Originally Posted by mdmbkr

The iommu=soft workaround fixed one of my machines, but the other uses software raid1 for its root (and only) filesystem. I have reason to suspect the ongoing issues are related to that, so I'd like to isolate the disk from the testing process to see if the system is also affected by the iommu issue. If I have to, I can install another drive in that computer.

Well the iommu error was fixed in recent kernel versions (I think with 2.6.21).
Perhaps you suffer from the other problem reported by krader?

[tomthumb99]

Folks, I have a HP Pavilion a1610n ( ADM dual core, 4 gig RAM) with the nVidia Corporation C51 chip, running ubuntu 8.04 and 8.10 with lots of file corruption problems (think its IDE/SATA controller). I am running the iomenu=soft and meme=1g switches from grub, still getting frequent file corruption. Every third boot, I need to run fsck on linux partitions, the files being corrupted are predictable (ie files: iso-8, gnom, multiple gnome-schema files, gnome win manager corruption). I initially just had 2 SATA drives, after reading multiple ubuntu posts went ahead and installed an IDE drive as boot disk, things got better (ie: Linux started). I started to install a second SATA controller card (VIA VT6421A), I see the migrated second SATA drive, but boot-up only recognized the drives on C51 chip (1-IDE, 1-SATA). In addition, the rev on the C51 chip appears very low (IDE & SATA - rev a1; USB Controller Rev a3 )??. I am not running raid, just simple ext3 file storage and two NTFS partitions. Under one syslog booting from the 19.24 kernel, it says I have a nForece2 chip

SO, I WANTED TO SEE IF THERE WAS A FIX TO THE C51 CHIP, MAYBE A FLASH OR UPDATE FIRMWARE. This is a key before moving ahead fully with VIA card. I am running the latest HP BIOS for the box under ubuntu 8.10, it did not help.

Any suggestions?? If it involves a patch to the kernel, pls give details instructions to follow (how apply tar/gz file) Would like to have XP patch also, have dual boot OS's here.

NOTE: I think the chip issues and heavy disk load factor was pointed out in the past by an IBM person perhaps on the KML link above ( but how to fix?).

Nvidia, I also have a Geforce 7600GT on this box with dual screen, I FELL THE PAIN on both sides here. I can not get the 180 for 185 video drivers working under Ubuntu, but the 177 drivers did work fine. Why you folks and ubuntu people cannot streamline these video install is beyond me; restricted drivers, tainted kernel, Xorg.conf conflicted settings...blaahh..blaahha. (who else is buying your high end video card, besides linux uses; Mac use have them already, MS users deaf and dump...maybe Xbox folks?). Look at this Ubuntu forum posting #7446864, its 90 pages long!