Bug 6857

Summary: poweroff doesn't - A7S333
Product: ACPI Reporter: Icoslau (icoslau)
Component: Power-OffAssignee: ykzhao (yakui.zhao)
Status: CLOSED CODE_FIX    
Severity: normal CC: acpi-bugzilla, bunk, dwmw2, eduardo, fdiaza, jeff, protasnb, trasgoescuro
Priority: P2    
Hardware: i386   
OS: Linux   
Kernel Version: 2.6.13 and high Subsystem:
Regression: Yes Bisected commit-id:
Attachments: Dmesg
ACPIdump
lspci
/proc/interrupts
lspci
/proc/interrupts
/var/log/boot.msg
log
dmesg output of kernel 2.6.24 that fails

Description Icoslau 2006-07-18 09:35:28 UTC
Most recent kernel where this bug did not occur:2.6.12

Distribution: SuSE Linux 10.0 and 10.1

Hardware Environment: ASUS A7S333 - Athlon XP 2400+ - 512DDR - Fireball 40Gb -
Pinnacle PCTV, Creative SB Live - Realtek 8139 - LG 52x - Samsung DVD/RW TSH552U

Software Environment: KDE from distro

Problem Description:
In the kernel 2.6.12 (Mandriva 2006) the ACPI is loaded correctly and my system
after a halt process has power off correctly, not require action on the button
of case. In the SuSE 10.0 and higher, i think the kernel 2.6.13 or higher, the
ACPI show some errors and after halt the system stay powered, only power off
with a button case. I try many different suggestions on the web, DSDT, use APM
instead of ACPI (in APM the machine power off correctly, but all times after one
hour the system hangs), kernel compiling with or not ACPI (of course the same
with APM) but the results are the same, with ACPI and kernel > 2.6.12 the
machine stay powered after halt, and with APM the machihe hangs.
Comment 1 Icoslau 2006-07-18 09:36:38 UTC
Created attachment 8576 [details]
Dmesg
Comment 2 Icoslau 2006-07-18 09:37:36 UTC
Created attachment 8577 [details]
ACPIdump
Comment 3 Icoslau 2006-07-18 09:38:09 UTC
Created attachment 8578 [details]
lspci
Comment 4 Icoslau 2006-07-18 09:38:38 UTC
Created attachment 8579 [details]
/proc/interrupts
Comment 5 Luming Yu 2006-07-19 01:55:15 UTC
Please add some printk in function acpi_enter_sleep_state to check if 
acpi_enter_sleep_state works ?
Comment 6 Icoslau 2006-07-21 16:57:47 UTC
Right, but how i can do this?
Tks.
Comment 7 Icoslau 2006-07-23 13:00:53 UTC
I think this is a real bug.

If i compiling the kernel default of distro, change my processor to Pentium Pro,
I have a Atlhon XP 2400+, the machine turns off correctly. If i compiling the
kernel, change the processor to Atlhon the problem persist. 

I think the use of processor to Pentium Pro in kernel slow my machine, but is
the only way to "power off" Asus A7S333.

Any ideia?

Tks
Comment 8 Luming Yu 2006-08-01 09:01:32 UTC
cool, you found a compiler related bug.
What gcc version you are using?
Comment 9 Len Brown 2006-08-02 19:07:02 UTC
What a surprise!
What processor was the distro (SuSE Linux 10.0 and 10.1) compiled for?

Please paste the output from gcc -v

Please boot the failing and working kernels and disassemble the
poweroff routine.  We'll be pretty lucky if there is a difference
there, but might as well start there...

# cd /boot
# gdb vmlinux-that-fails /proc/kcore
# x/15ai acpi_power_off

and do the same when running vmlinux-that-works
Comment 10 Adrian Bunk 2006-10-24 19:27:31 UTC
Please reopen this bug if:
- it is still present in kernel 2.6.18 and
- you con provide the requested information.
Comment 11 trasgo 2007-02-25 16:13:09 UTC
i have the same problem, kernel 2.6.18
Comment 12 trasgo 2007-02-25 16:25:28 UTC
Created attachment 10534 [details]
lspci
Comment 13 trasgo 2007-02-25 16:30:44 UTC
Created attachment 10535 [details]
/proc/interrupts
Comment 14 trasgo 2007-02-25 16:35:41 UTC
Created attachment 10536 [details]
/var/log/boot.msg
Comment 15 trasgo 2007-02-25 16:40:57 UTC
Created attachment 10537 [details]
log
Comment 16 trasgo 2007-02-25 16:44:46 UTC
 I use other distros like gnewsense and sabayon, they cant power off my linux 
too.
Comment 17 Zhang Rui 2007-07-11 00:46:00 UTC
Does the problem still exists in 2.6.22?
Comment 18 Fu Michael 2007-09-12 00:33:25 UTC
ping for update from bug reporter
Comment 19 Jeff Muizelaar 2007-09-19 19:25:00 UTC
It looks like I can reproduce this with 2.6.22
Comment 20 Natalie Protasevich 2007-10-22 23:30:11 UTC
Any update on this issue please. If someone can reproduce it still, can you provide requested in #9? and please mention the compiler verion too.
Thanks.
Comment 21 Fu Michael 2007-10-25 02:15:53 UTC
could you please have a test to see if the solution in bug# 6712 helps?
Comment 22 Fernando Díaz 2007-10-30 09:11:12 UTC
At work we have a PC with an A7S333 running Debian Etch, and I can confirm that the bug still happens with kernel 2.6.22 (from backports.org), at least with packages for K7 and 486. I also tried the same kernel version, custom built from the debian sources, but with ACPI_SLEEP unset (as per bug# 6712), without positive results.
I don't know if other commenters have seen this symptom, but the computer we have here does not power off when pressing briefly the power button after the "acpi_power_off called" message is displayed, responding only to pressing it for many seconds or unplugging the computer.
Comment 23 Eduardo Bacchi Kienetz 2007-12-25 18:41:22 UTC
This bug exists in kernel 2.6.23, the one I'm using. 
"acpi_power_off called" is the last message I see.
In previous kernels I always used apm, but recently it seems to have been removed, meaning I need to press the button to power the PC off.
Comment 24 Len Brown 2008-01-06 21:05:45 UTC
Jeff Muizelaar, Fernando Díaz,  Eduardo Bacchi Kienetz:

please confirm it still fails with "apm=off"

Icoslau and trasgo effectively already did by attaching dmesg:

apm: BIOS version 1.2 Flags 0x03 (Driver version 1.16ac)
apm: overridden by ACPI.

but we haven't heard from them in a long time -- so you have effectively
taken over this bug report with your failure.

If it still fails, please attach the complete dmesg -s64000
from the kernel that fails -- preferably a kernel.org kernel.
Comment 25 Fernando Díaz 2008-02-20 05:16:37 UTC
With kernel 2.6.24 (from kernel.org) the PC still doesn't turn off by itself, but now it does respond to pressing (briefly) the power button. 
Comment 26 Fernando Díaz 2008-02-20 05:25:55 UTC
Created attachment 14913 [details]
dmesg output of kernel 2.6.24 that fails
Comment 27 Eduardo Bacchi Kienetz 2008-02-20 05:39:11 UTC
I'm sorry Len, I'm unable to test since my desktop computer was left in Brazil and I will be working in Canada for at least 6 months (max a year). As soon as I go back I will test it :|
Comment 28 ykzhao 2008-04-15 02:33:50 UTC
Hi, Eduardo
    Will you please try the latest kernel and see whether the problem still exists?
    
    From the acpidump it seems that the S3 object has the same package with the S5 . IMO. It is incorrect. When the system enters the S3 state, the S3 object will be used. When the system is powered off, the S5 object will be used.

  >   Name (\S_3, Package (0x04)
   > {
        0x06, 
        0x00, 
        0x00, 
        0x00
    })
   > Name (\_S5, Package (0x04)
    {
    >    0x06, 
     >   0x00,
      >  0x00,
        0x00
    })
Comment 29 Fernando Díaz 2008-04-15 12:55:44 UTC
I have just tried the 2.6.25-rc9 kernel, and it works. The PC powers off normally.
Comment 30 ykzhao 2008-04-15 18:01:41 UTC
Hi, Fernando
    Thanks for the test. It seems that the system can work well.
    Is the bios upgraed? If the bios is upgraded, please attach the output of acpidump.
    Thanks.
Comment 31 Fernando Díaz 2008-04-16 06:41:37 UTC
No, the only modification done to the computer is the kernel change, from 2.6.24 (which didn't power off, but at least didn't block the power button as the 2.6.18) to 2.6.25-rc9. Do you need the output of acpidump anyway?
Comment 32 ykzhao 2008-04-16 22:53:33 UTC
It is unnecessary to attach the output of acpidump again.
Since the latest kernel(2.6.25-rc9) can make your laptop work well, the bug will be closed. 
Comment 33 Len Brown 2008-06-13 20:18:25 UTC
> I have just tried the 2.6.25-rc9, and it works.
> The PC powers off normally.

thanks for testing.
report closed.