Re: [PATCH] ACPI: don't walk tables if ACPI was disabled

[Ingo Molnar <mingo@elte.hu>]

* Vegard Nossum <vegard.nossum@gmail.com> wrote:

> Hi Ingo,
> 
> Can you see if this patch solves your problem? There might be other 
> functions that needs this guard as well, though. I wonder if maybe 
> this test should just be included at the top of every driver that uses 
> ACPI in some way. But I'm pretty sure that this lack of initialization 
> is the root of your problem in any case :-)

applied this to tip/out-of-tree for more testing, thanks Vegard.

> (By the way, I don't know why this problem popped up at this time, 
> maybe it was just bad timing/bad luck... How far back do your 
> AE_BAD_PARAMETER in the logs go?)

i have hit this warning for the first time in January 2008:

 [    0.000000] Linux version 2.6.24-rc8 (mingo@dione) (gcc version 4.2.2)
                #452 SMP Sun Jan 20 23:36:28 CET 2008

and it says:

[    0.000000] Calling initcall 0xc050758a: acpi_rtc_init+0x0/0xb8()
[    0.000000] ACPI Exception (utmutex-0263): AE_BAD_PARAMETER,
               Thread F7C22000 could not acquire Mutex [3] [20070126]
[    0.000000] initcall 0xc050758a: acpi_rtc_init+0x0/0xb8() returned 0.

the logs of my auto-tests on this box start at more than a year ago:

 Linux version 2.6.21-rc6 (mingo@dione) (gcc version 4.0.2)
               #331 SMP Fri Apr 13 10:14:12 CEST 2007

the size of the logs is 16.2 GB, covering the bootup of 58605 uniquely 
built kernels performing 67065 bootups - so it's a fairly exhaustive 
history.

that's why WARN_ON()s are so important - there's no way my automated 
tools (or even i, when taking a casual look at the logs) could have 
picked up that new ACPI Exception - if each subsystem has different 
warnings (which change frequently) then it's sheer impossible to 
automate the answer to the "does that log show any anomaly" question. 

( Even delta analysis would be of little use, due to timing related
  noise, random data variances and the impact of randconfig booting. )

The only reason i noticed it because this problem escallated into a lock 
corruption which triggered a WARN_ON().

	Ingo