[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#275572: This is ACPI related, and fixed in 2.6.9



On Thu, Nov 04, 2004 at 03:41:23PM -0400, Peter Cordes wrote:
> 
>  Version 2.6.8-4 of kernel-image-2.6.8-1-k7-smp has different symptoms:
> kacpid uses 100% CPU time when a UPS is connected to the serial port.
> On 2.4.22 and .27 kernels (and presumably all others with recent ACPI),
> keventd uses 100% CPU time.  I can run top and see kacpid/keventd jump up to
> 100% CPU or drop off when I connect/disconnect the RS-232 cable.
> 
> A tech support guy at my vendor told me ---------
>  We have seen the ACPI problem with several UPS brands/models on certain
>  kernels.  As you mentioned newer kernels do not seem to exhibit this
>  problem, and another workaround is always using acpi turned off in the BIOS. 
> -------------------------------------------------
> 
>  Disabling ACPI in the BIOS solves the problem.  So does booting with
> ACPI=off.  I don't remember if other BIOS ACPI settings (i.e. ACPI V2, or
> ACPI APIC, or inclusion of various tables) had an effect, but I don't think
> so.
> 
>  The problem doesn't happen with a 2.6.9 kernel I compiled myself from
> vanilla kernel.org source.  (This is all with i386 kernels.  64bit 2.6
> kernels will oops in kacpid or will have the keyboard freeze within a few
> minutes.  (They unfreeze when the cable is unplugged from the UPS).  I
> haven't tested a 64bit 2.6.9, but I can if you want to know whether it's
> worth pushing it into Sarge.)

ACPI seems to be a rather problematic section of code. I will take
a look but fixing 2.4.27 is likely difficult at best. The code seems
to have stabalised in 2.6.9, which is a start, but 2.6.8 is the target
for sarge.

>  This will mainly affect people who admin servers, since dual-Opteron Tyan
> boards w/ UPSes aren't typical for home users.  My point is that putting
> the acpi=off info in an install FAQ is probably good enough for people who
> might be affected by this when installing Sarge.

I agree documentation is a good idea here, certianly twidling acpi=
seems to be a common fix for various problems. And while the real
solution is to resolve the kernel problems, it is likely that acpi=
will still be a common remedy for sarge users.

Could you be a little more specific about where you think is
a good place for this documentation? Are you able to contribute
any documentation? That would be most excellent.

-- 
Horms



Reply to: