On Thu, 19 Jan 2012, John Kacur wrote:

On Thu, Jan 19, 2012 at 5:38 PM, Steven Rostedt [off-list ref] wrote:

quoted

On Thu, 2012-01-19 at 17:16 +0100, John Kacur wrote:

quoted

I'm testing v3.0.14-rt32-rc1, and it greatly reduced the number of
these, but I still got one during boot.

[   11.938649] BUG: sleeping function called from invalid context at
/home/jkacur/rt.linux.git/kernel/rtmutex.c:645
[   11.938651] in_atomic(): 1, irqs_disabled(): 1, pid: 168, name: irq/9-acpi
[   11.938653] 1 lock held by irq/9-acpi/168:
[   11.938653]  #0:  (acpi_gbl_gpe_lock){......}, at:
[<ffffffff8123881d>] acpi_ev_gpe_detect+0x31/0x110

Could you test 3.2-rt. If that has the issue then we know we have more
to fix. If not, we know I missed a patch.

I tested with v3.2-rt10 and the issue is not there.

I'm surprised. I know already what the problem is and according to the
code it should be visible in 3.2 as well. I'm still twisting my head
around that acpi maze to fix that particular one.

I also examined your patches and couldn't find anything wrong with
them, and for larks tried applying
cpumask-disable-offstack-on-rt.patch from v3.2-rt10
all to no avail, the problem is still there.

What a surprise, it's completely unrelated.