Hello

With recent 2.6.25 & 2.6.26-rc1 git (around 1 week) I get occasionally
complete freeze of my T61 during suspend. (dual core, 2GB).

I'm running kernel with no_console_suspend - but all I can see is
blinking cursor on an empty screen - thus even when I run kernel with
most debug options turned on, I can't pass more details so far.  I run
suspend with with SD card in - so maybe some update in the MMC driver
might be responsible for this ?

Also - I think that option no_console_suspend doens't work correctly -
as many times with suspend I do not see any log message on my console
screen. However sometimes the log is shown.

Zdenek
--

[ message continues ]

It would be helpful if you could verify if:

(1) The problem occurs without no_console_suspend.
(2) The problem occurs without the SD card.

Thanks,
Rafael
--

[ message continues ]

Hi Rafael,

same problem here, although I was able to resume system (it's basically Intel
machine) , but it was unusable - I was able to switch between terminals and see
output from kernel. So there was:
	- Disabling irq #19;
	- some kind of lock spinning on disk:
IDE interface: Intel Corporation 82801GBM/GHM (ICH7 Family) Serial ATA Storage
Controller IDE (rev 02)
	but I can't provide more output of that lock now - no sign in logs.

I've made some successful suspend/resume all without sound card active without
problem. Those appear with sound card active, but I must take closer look - will
send info later.

-Jacek
--

[ message continues ]

Unfortunately not at my wish - usually in the least expected moment.
(Actually I've once even put notebook into the bag without checking
its sleeping :(


Will keep an eye on this - I usually keep SD card in.

Zdenek
--

[ message continues ]

Hello Rafael


The problem happens still even with -rc3. Now I've an update:

Usually I've to run the machine for couple hours to actually be able
to hit this lock.
(Usually after a day work when I want to leave)

I've also noticed that when I run the suspend after the reboot I
usually cannot see the suspend freeze - mostly because either the

I've figured out, it was caused by some weird Fedora setting, so

SD card or no_console_suspend option doesn't matter

This is what I've seen as the last thing on the screen when deadlock
appeared this time:
(no SD card inserted)

====
Freezing remaining freezable tasks ... (elapsed 0.00 seconds) done.
PM: Entering mem sleep
drm card0: class suspend
drm_sysfs_suspend
ACPI: PCI interrupt for device 0000:00:02.0 disabled
====
and this is what usually follows when the suspend works correctly
====
sd 0:0:0:0: [sda] Synchronizing SCSI cache
sd 0:0:0:0: [sda] Stopping disk
ACPI: PCI interrupt for device 0000:15:00.2 disabled
ACPI: PCI interrupt for device 0000:00:1f.1 disabled
...


lcpci:
00:02.0 VGA compatible controller: Intel Corporation Mobile
GM965/GL960 Integrated Graphics Controller (rev 0c)
15:00.2 SD Host controller: Ricoh Co Ltd R5C822 SD/SDIO/MMC/MS/MSPro
Host Adapter (rev 21)
00:1f.1 IDE interface: Intel Corporation 82801HBM/HEM (ICH8M/ICH8M-E)
IDE Controller (rev 03)

Here is also the grep from my messages log for 'hash matches device'
(gathered over some time)
(Having PM_TRACE_RTC=y  in the .config)
device:02
device:02
0000:00:1c.1:pcie03
0000:00:1c.0:pcie02
device:0a
cooling_device1
0000:15:00.1
device:07
tty47
mcelog
0000:00:02.1
device:0c
0000:0d
tty53
tty55
device:22
target0:0:0
device:03
PNP0C02:00
00:04
0000:00:1c.0:pcie00
sda
PNP0C0F:02
tty47
mcelog
0000:00:02.1
ram7
device:13
device:04
fbcon


From this list it looks pretty random :( - so I'll keep an eye if the
deadlock appears alway in the same place.
But if you have any idea what kind of debuging would help to find ...
[ message continues ]

Hm, what kind of graphics adapter is there in your box?

Rafael
--

[ message continues ]

After some more checking - I guess this information actually is not
useful at all :(.
It looks like even when the suspend is succesful - the line:
ACPI: PCI interrupt for device 0000:00:02.0 disabled
is alway the last one visible.

 I'll try to get figure out a better way how to invoke the lock.

And my graphics adapter is:
00:02.0 VGA compatible controller: Intel Corporation Mobile
GM965/GL960 Integrated Graphics Controller (rev 0c) (prog-if 00 [VGA
controller])

Zdenek
--

[ message continues ]

After some more checking - I guess this information actually is not
useful at all :(.
It looks like even when the suspend is succesful - the line:
ACPI: PCI interrupt for device 0000:00:02.0 disabled
is alway the last one visible.

 I'll try to get figure out a better way how to invoke the lock.

And my graphics adapter is:
00:02.0 VGA compatible controller: Intel Corporation Mobile
GM965/GL960 Integrated Graphics Controller (rev 0c) (prog-if 00 [VGA
controller])

Zdenek
--

[ message continues ]

If you use s2ram, please check if "echo mem > /sys/power/state" works instead.

Thanks,
Rafael
--

[ message continues ]

The problem occurs also with plain echo 'mem' - I've already checked before.
Just resently I've discovered that actually I've had some serious
problem with CONFIG_DEBUG_PAGEALLOC option. Unsure if this thing could
be related to my deadlocking problem during suspend - I'll watch, if
the freeze will be still ocurring even without this option.

I'll make probably a separate post about my debug option problem.

Zdenek
--

[ message continues ]

Hello

It looks like from the time I've disabled CONFIG_DEBUG_PAGEALLOC
option I've not seen a single freeze during the suspend - I'll
continue to watch this - but I guess it's the key part of my problem.

Zdenek
--

[ message continues ]

Try switching consoles with alt-arrows... s2ram plays with console
switches so messages may be there but on other console.
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
--

[ message continues ]

No change - logging is simply not visible anywhere - if there will be some time
I could probably bisect and find which commit randomised console logging.

Zdenek
--

[ message continues ]

Same here. Suspend to ram crashes on resume starting in 2.6.25 (didn't
test 2.6.25-rc in my production system).

Because this isn't 100% reproducible bisecting may return false
negatives.
--

[ message continues ]

Of course I've planned to bisect for the lost console logging - maybe tomorrow.

The crash in suspend is pretty random - I've never got the lock at my
will - usually it happens when I completely do not expect it ;)

Zdenek
--

[ message continues ]