I'm creating a new bug as it was suggested in bug 3927. In 2.6.20, there is a regression on at least my ATI board (and thus probably others) that the clock runs at double speed. The most likely culprit is http://www.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=b026872601976f666bae77b609dc490d1834bf77 I will attach dmesg logs from 2.6.19 and 2.6.20-rc4, plus lspci outputs. Note that the following lines are missing from 2.6.20 dmesg output: [ 0.000000] ATI board detected. Disabling timer routing over 8254. [ 31.368871] ..MP-BIOS bug: 8254 timer not connected to IO-APIC
Created attachment 10028 [details] dmesg-2.6.19.log
Created attachment 10029 [details] dmesg-2.6.20.log
Created attachment 10030 [details] lspci -nvv output
Created attachment 10031 [details] lspci -vv
Created attachment 10034 [details] patch to revert the possible culprit Does applying this patch fix the issue?
Created attachment 10035 [details] dmesg-2.6.20-patched.log Confirming that using that revert patch fixes the issue and clock works normally again. ak's original git comment had "TBD Experimental. Needs a lot of testing", so if it hasn't received that testing it'd probably be good to revert this. I'm happy to test any future patches to the issue with my hardware.
The commit that broke it got reverted in Linus' tree. Thanks for your bug report!
post scriptum: the board is MSI RS482M4-ILD with the latest (and probably last) BIOS version 1.2. I'm now trying to follow linux-kernel on possible 2.6.21 patches regarding this issue.