Bug 2571

Summary:	ACPI devices lack suspend/resume support
Product:	ACPI	Reporter:	Thomas Renninger (trenn)
Component:	Power-Sleep-Wake	Assignee:	Shaohua (shaohua.li)
Status:	REJECTED INSUFFICIENT_DATA
Severity:	low	CC:	acpi-bugzilla, bunk, c-d.hailfinger.kernel.2004, pavel
Priority:	P2
Hardware:	i386
OS:	Linux
Kernel Version:	2.6.5	Subsystem:
Regression:	---	Bisected commit-id:
Attachments:	a patch test patch agains 2.6 kernel

Description Thomas Renninger 2004-04-22 13:49:39 UTC

Kernel 2.6.5 on a compaq NX5000  
  
Problem Description:  
After suspend to disk (echo 4 >/proc/acpi/sleep) the power button(FF) does not  
generate any events/interrupts any more.  
  
Steps to reproduce:  
boot: init=/bin/bash  
echo 4 >/proc/acpi/sleep  
->resume  
cat /proc/acpi/event | cat /proc/interrupts  
press the power button -> nothing

Comment 1 Thomas Renninger 2004-04-22 14:04:34 UTC

Forgot to mention that loading/unloading the button module before/after suspend 
helps -> the ad_adapter state value is updated.

Comment 2 Luming Yu 2004-04-27 01:43:49 UTC

Created attachment 2728 [details]
a patch

suspend_finish is missing from resume from Disk code path.

Comment 3 Luming Yu 2004-05-17 23:44:10 UTC

Created attachment 2892 [details]
test patch agains 2.6 kernel

Please help with testing this patch.

Comment 4 Thomas Renninger 2004-05-18 01:57:29 UTC

Sorry, this cannot be the reason.
I use swsusp (echo 4 >proc/acpi/sleep).
disk.c is based on CONFIG_PM_DISK and is not even compiled in my kernel.

Thank you very much, anyway.

> Please help with testing this patch.
Of course.

Pavel told me that swsusp doesn't really do a S4.
As far as I know _WAK function is not invoked at all after resume.
Maybe this is the problem and the button needs to get reactivated there?

However, this bug is not that sever because it's easy to workaround: un-,load
button module before/after suspend.

Comment 5 Len Brown 2004-11-03 19:06:21 UTC

same with 2.6.9?

Comment 6 Thomas Renninger 2004-11-04 01:02:18 UTC

using swusp you can now write
platform,firmware,shutdown or reboot (reboot for debugging issues)
into /sys/power/disk
1) shutdown -> no _WAK method invoked as it always has been done
2) firmware -> try to let bios do the suspend
3) platform -> _WAK method invoked

3 Fixes such little issues like wrong AC state after suspend as BIOS could cache
this value and reinitialise it in _WAK or not working buttons (the button issue
I saw on several laptops).
However, platform could crash a small amount of laptops and is therefore still
not the default.

Pavel may correct me if I am wrong in some point.

Comment 7 Carl-Daniel Hailfinger 2004-12-14 15:57:33 UTC

Seeing this with Kernel 2.6.10-rc3-bk6 on a Samsung P35

Why can't the "shutdown" method reinitialize the things by hand which the
"platform" method relies on the BIOS to reinitialize?

A simple acpi_enable_event on resume would fix the button issue for me when
using the "shutdown" method.

I'd reopen this bug if Bugzilla let me.

Comment 8 Shaohua 2004-12-14 16:58:36 UTC

acpi_enable_event looks like a workaround to me. Actually 'platform' method 
should be the default method if ACPI is robust enough.
Another solution is we provide ACPI devices suspend/resume.

Comment 9 Pavel Machek 2006-05-23 03:57:54 UTC

Thomas, does it still happen with 2.6.17-rc4? I thought we were adding some
workarounds (aha, one of them is attached), but they may be only present in
-suse or something like that.

Comment 10 Adrian Bunk 2006-07-10 13:07:05 UTC

Please reopen this bug if it's still present in kernel 2.6.17.