Most recent kernel where this bug did not occur:2.6.18.1 + patch http://lkml.org/lkml/2006/7/20/56 Distribution:kubuntu 6.06 LTS Hardware Environment: T60 with intel graphics. See http://www.thinkwiki.org/wiki/Hardware_Specifications BIOS at default setting - disk running in ahci mode (this is why I need a patch for 2.6.18). Software Environment: Not sure what to put here. Seems to happen both with and without X running. Problem Description: after I do a full kernel compile, my T60 stops triggering any ACPI events: tail -f /var/log/acpid does not show anything, even on Fn/F4 which is supposed to be always enabled. Restarting the acpid doesn't do anything either - ACPI starts working again, for a while, only after reboot. Works fine in 2.6.18 ( + this patch http://lkml.org/lkml/2006/7/20/56). Steps to reproduce: cd tmp git clone -s ~/scm/linux-2.6 cd linux-2.6 make oldconfig make -j 4 no new ACPI events reported after this.
Created attachment 9338 [details] Here's my .config for 2.6.19 I attached the .config for 2.6.19-rc3 with which I see the problem
Created attachment 9339 [details] Here's the config for 2.6.18.1 that works well I attached the .config for 2.6.18 (that works well) for reference
I see on my X60s too but triggered by suspend-to-ram/resume. kernel compiling did not yet show this behaviour (but I diden't really check for I usually reboot rather quickly after compiling/installing a new kernel) the problem is there since the first post 2.6.18 version that I compiled a few weeks ago. currently I run 2.6.18 with some patches, which suspends/resumes fine.
Yes, I see this too if I do not enable IB thinkpad laptop extras under PM/ACPI - in that case, kernel compile does not trigger the bug, but suspend/resume does.
> Yes, I see this too if I do not enable > IB thinkpad laptop extras under PM/ACPI - > in that case, kernel compile does not trigger the bug So the "no ACPI events after kernel build" regression happens only if ibm_acpi is loaded? > but suspend/resume does. So ACPI events used to work in 2.6.18 after suspend/resume, but do not work after suspend/resume in 2.6.19 -- and this is true not matter if ibm_acpi is loaded or not?
> > Yes, I see this too if I do not enable > > IB thinkpad laptop extras under PM/ACPI - > > in that case, kernel compile does not trigger the bug > > So the "no ACPI events after kernel build" regression > happens only if ibm_acpi is loaded? Yes, it seems so. > > but suspend/resume does. > > So ACPI events used to work in 2.6.18 after suspend/resume, > but do not work after suspend/resume in 2.6.19 -- and > this is true not matter if ibm_acpi is loaded or not? This is true if ibm_acpi is not loaded. I did not test this with ibm_acpi loaded.
Created attachment 9401 [details] patch to ibm_acpi.c There has been only 1 small patch to ibm_acpi.c between 2.6.18.1 and now. If you patch -Rp1 this commit, you'll be backing that driver down to the version that worked in 2.6.18, and the 2.6.19 run-time issue with ibm_acpi should go away -- yes? However, it isn't clear that this simple patch to clear the experimental flag could be the root of the run-time errors; and maybe we need to re-confirm the assertion that the run-time (not the suspend-time) failure... 1. does not happen in 2.6.18 with ibm_acpi loaded 2. does not happen in 2.6.19 with ibm_acpi unloaded 3. only happens in 2.6.19 with ibm_acpi loaded
OK, it seems Linus has zeroed in on the bug: http://lkml.org/lkml/2006/11/1/84 I am running f9dadfa71bc594df09044da61d1c72701121d802 and no issues yet in several days of use.