The kernel thread kacpid crashes the system du to stack-overflow. This is caused by "recursive" event triggering when temperature treshold is reached (Affects my Acer TM660, according to upstream bug report also other Acer Latops and at least one HP laptop) Reproducible: Always Steps to Reproduce: 1. Make the CPU get hot (no cpufreq or compiling) Actual Results: System crash with trace mentionning trace_dump(), Oops #1, process kacpid. In most cases with no trace of the original cause because of corrupt stack. Eventually stacktrace with repeated sequence (including event notifier) Expected Results: System continues running Upstream has patches that solve the problem. Not loding thermal module prevents the issue as well.
Bruno, The Kernel devs have whipped up a patch for this kernel bug, but haven't merged it into the tree yet. When it does get merged, it'll make its way to the Gentoo kernel. Here are the two patches: http://bugzilla.kernel.org/attachment.cgi?id=11331&action=view http://bugzilla.kernel.org/attachment.cgi?id=11332&action=view .james
Thanks James, I'm already using them since sunday, either the first or both do fix my crashes. Matter of time when they get applied or similar patch gets applied upstream I assume. Guess they should also go into some -stable release for 2.6.21 (issue is twice in the regression list on kernelnewbies)
Looks like it's not that simple, let's wait for the real fix upstream
upstream patch is applied upstream for 2.6.24
fixed in gentoo-sources-2.6.22-r10, will leave open until 2.6.23 is fixed too
this was fixed in gentoo-sources-2.6.23-r2
