Re: [patch 2.6.21-rc5-git 2/3] ACPI driver model flags and platform_enable_wake()

[TJ <linux@tjworld.net>]

On Wed, 2007-07-25 at 13:04 -0700, David Brownell wrote:

> As I recall, Mattia Dongili <malattia@linux.it> reported an error
> with a Vaio laptop; it would wake immediately after entering STR.
> It didn't reproduce for anyone else.  But that was eventually tracked
> down to a bug in iwl3945, which has since been fixed (yes?).

I'm not sure if this might inform your thinking here but today I've just
tracked down a similar issue on another Sony Vaio (PCG-SRX51) where S3
STR immediately wakes.

I enabled ACPI_FUTURE_USAGE and called acpi_get_event_status()
immediately upon resume (before the wake flag is cleared) and iterated
each wake event.

I discovered that the PM timer event is the reason the system resumes
but so far haven't discovered why. I have a tenuous theory it could be
related to the kernel timer using the ACPI PM Timer. I tried disabling
that with the boot parameter 'nopmtimer' but that didn't help.

To test this I'm currently building a kernel that disables the PM Timer
wake event during suspend and re-enables it on resume, something the
ACPI specification hints at in section 4.7.2.1.

I'm not sure if this problem occurs on a wider range of systems than the
Sony since at present, without a debug kernel with ACPI_FUTURE_USAGE and
specific additional checks, there is no way to know the reason a system
resumed.

I'm planning on submitting a patch to log the wake event since it would
greatly enhance our abilities to diagnose reasons behind these
situations if the log showed it without a custom-kernel build.

TJ.