2005-03-16 21:33:39

by Andrew Morton

Subject: Re: [PATCH 1/2] Thinkpad Suspend Powersave: Fix ACPI's GFP_KERNEL allocations in contexts that can sleep

"Theodore Y. Ts'o" <[email protected]> wrote:
>
> This fixes a problem originally reported by Christian Borntraeger where
> during the wakeup from a suspend-to-ram, several "sleeping function
> called from invalid context" warning messages are issued. Unlike a
> previous patch which attempted to solve this problem, we avoid doing an
> GFP_ATOMIC kmalloc() except when explicitly necessary.
>
> Signed-off-by: "Theodore Ts'o" <[email protected]>
>
> Index: src/drivers/acpi/osl.c
> ===================================================================
> --- src.orig/drivers/acpi/osl.c 2005-03-14 09:38:15.000000000 -0500
> +++ src/drivers/acpi/osl.c 2005-03-14 09:38:18.000000000 -0500
> @@ -145,7 +145,7 @@
> void *
> acpi_os_allocate(acpi_size size)
> {
> - return kmalloc(size, GFP_KERNEL);
> + return kmalloc(size, in_atomic() ? GFP_ATOMIC : GFP_KERNEL);

We shouldn't do this. Please see
http://www.uwsg.iu.edu/hypermail/linux/kernel/0503.1/1205.html

2005-03-16 22:03:53

by Christian Borntraeger

Subject: Re: [PATCH 1/2] Thinkpad Suspend Powersave: Fix ACPI's GFP_KERNEL allocations in contexts that can sleep

Andrew Morton wrote:
> "Theodore Y. Ts'o" <[email protected]> wrote:
> > This fixes a problem originally reported by Christian Borntraeger where
> > during the wakeup from a suspend-to-ram, several "sleeping function
> > called from invalid context" warning messages are issued. Unlike a
> > previous patch which attempted to solve this problem, we avoid doing
> > an GFP_ATOMIC kmalloc() except when explicitly necessary.

Len,
you indicated that you are going to address the S3 sleeping function issue
after the development for 2.6.12 has started with an invasive but correct
fix.
(http://marc.theaimsgroup.com/?l=acpi4linux&m=110978961501752&w=2)
Can you give an overview about your progress?

cheers

Christian