Latest working kernel version:
ACPI testing, EC fast interrupt patch on top of 2.6.27.
E.g. acpi/test tree commit v2.6.27-87-g0fbb372.
Without this fix, there's a bug where the hotkeys will stop working much sooner.
Therefore this is not a regression from 2.6.27 - instead, it extends the EeePC hotkey regression which was already present in 2.6.27.
Earliest failing kernel version: 2.6.28-rc2
(rc1 doesn't boot due to ata_piix breakage).
Distribution: Ubuntu 8.04
Hardware Environment: Asus Eee PC 701.
EC dies, hotkeys stop working, ACPI power button stops working, other strange things happen e.g. suspend/resume seems to hang, battery information probably stops being updated.
Steps to reproduce:
1. Boot on battery, no AC power. (The battery may or may not matter, but I've been playing it safe).
2. Wait for X login screen
3. Hold down "brightness up" key until full brightness is reached.
4. Hold down "brightness down" key until minimum brightness is reached.
5. Repeat steps 3&4. 25 repititions (yes, twenty five) are required to reproduce reliably.
I think I stumbled upon this using a different and more complicated sequence; this one may be tedious but it seems fairly reliable and it's nice and simple
Created attachment 18557 [details]
*** Bug 11896 has been marked as a duplicate of this bug. ***
Created attachment 18558 [details]
clean tmp before reuse
This patch is reported to fix the issue.
Thanks Alan for bisecting 3700+ patches :)
Handled-By : Alexey Starikovskiy <email@example.com>
Patch : http://bugzilla.kernel.org/attachment.cgi?id=18558&action=view
Created attachment 18602 [details]
try the patch in which the EC global lock is initialized based on the return value of _GLK object
Will you please try the attached patch and see whether the problem can be sovled by this patch?
In this patch the EC global lock is initailzed based on the return value of _GLK object.
Why? It doesn't seem to do anything different to Alex's patch.
patch in comment #3 applied to ACPI tree
There is no difference between the patch in comment #3 and my patch.
But IMO when the ACPICA interface function is evaluated, it will be better to base on the return status. If the return status is what we expected, the return value of ACPI object will be used. Otherwise it should be ignored.
just add that this patch also fixes (at least in my case/tests with a EeePC 701) the case where the hotkeys stop working after a suspend/resume cycle.
Handled-By : From: Alexey Starikovskiy <firstname.lastname@example.org>
Patch : http://marc.info/?l=linux-acpi&m=122603281422097&w=4
Not-Handled-By : From: Alexey Starikovskiy <email@example.com>
Ignore-Patch : http://bugzilla.kernel.org/attachment.cgi?id=18558&action=view
unclear what comment #11 is about, but to re-iterate comment #7,
this patch is in the acpi tree:
Author: Alexey Starikovskiy <firstname.lastname@example.org>
Date: Mon Nov 3 14:26:40 2008 -0500
ACPI EC: Fix regression due to use of uninitialized variable
shipped in 2.6.28-rc4-git3