Regression: system clock runs at twice the speed

[Timo Jyrinki]

Binary package hint: linux-source-2.6.20

ATI RS482 chipset, AMD64. System clock runs at twice the speed. The bug was there in Ubuntu 5.10, but was fixed in Ubuntu 6.06 (and upstream kernel 2.6.15 had some notice about it being fixed). I didn't notice the problem with 2.6.19 earlier during feisty development, but 2.6.20 shows the problem again.

The chipset is buggy, and the kernel used to detect the bugginess and select correct options automatically. Now again it doesn't work without kernel option magic, so something has again been changed in disfavor of RS480/482 chipset users.

[DaveThacker (dthacker9)]

Timo,
Thanks for the report and the diagnostic output. Could I please ask you to attach the output of dmesg and lspci as files, instead of pasting into comments. That will help move this into confirmed status. See https://wiki.ubuntu.com/DebuggingKernelProblems for the requirements.
Thanks, DaveThacker

[Timo Jyrinki (timo-jyrinki)]

Eh... looks like attachments to me :)

[Timo Jyrinki (timo-jyrinki)]

As the reply was a copy-paste reply, I'm trying to figure out what more is being asked... here's lspci -vv now. If you can be more spefic, please be. And leave the mention about "not as comments" out when copy-pasteing the answer if the bug does not have logs inserted as comments.

[Timo Jyrinki (timo-jyrinki)]

I installed 2.6.19-7 and 2.6.20-5. Confirming that is not fixed in the 2.6.20-5, but working fine in 2.6.19-7. See the attached dmesg log for the 2.6.19-7 kernel.

Note that the following is _missing_ from the 2.6.20 log attached earlier:
[ 0.000000] ATI board detected. Disabling timer routing over 8254.
as is:
[ 31.368871] ..MP-BIOS bug: 8254 timer not connected to IO-APIC

[DaveThacker (dthacker9)]

It was only half a copy/paste reply ;) Sorry, I must have missed the filenames in the attachments box. DaveThacker

[DaveThacker (dthacker9)]

I think the necessary dumps and output have been provided. Moving to confirmed. DT

[Timo Jyrinki (timo-jyrinki)]

The patch that caused this has been now reverted in Linus' tree.

[Ben Collins (ben-collins)]

Yep, saw that on the last pull. Thanks

[Ken Lewis (kenlewis)]

Ben,

Can you attach the patch to a comment here, list its git entry for your tree at kernel.org (I'm not expert with git but assume it can do that) or tell us when a new edition of the kernel will be available for download?

Thanks.
Ken.

[Timo Jyrinki (timo-jyrinki)]

The patch is here:
http://git.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=fea5f1e19611d94fbf3905875a427c4cb959cd06

[Ken Lewis (kenlewis)]

Ben Collins wrote:
> ** Changed in: linux-source-2.6.20 (Ubuntu)
> Status: Fix Committed => Fix Released
Confirmed fixed in linux-image-generic-2.6.20-6 (name may not be
formatted right).

Ken.Lewis