Re: 2.6.14-rc3-rt2

[Steven Rostedt <rostedt@goodmis.org>]

On Wed, 5 Oct 2005, Mark Knecht wrote:

> Hi Ingo,
>    By the time I go there it was already -rt9. Unfortunately I still
> get the same message. Possibly the continued discussion is addressing
> this? I haven't tried to read it all yet.
>
> >From dmesg:
>
> PCI: Using configuration type 1
> PCI: Using MMCONFIG at e0000000
> ACPI: Subsystem revision 20050902
> BUG: swapper/1: spin-unlock irq flags assymetry?
>
> Call Trace:<ffffffff80400d19>{_spin_unlock_irqrestore+89}
> <ffffffff802694d6>{acpi_ev_create_gpe_block+642}
>        <ffffffff80263116>{acpi_os_wait_semaphore+135}
> <ffffffff80269682>{acpi_ev_gpe_initialize+109}
>        <ffffffff80266cac>{acpi_ev_initialize_events+151}
> <ffffffff802790a2>{acpi_enable_subsystem+69}
>        <ffffffff806086de>{acpi_init+86} <ffffffff80604cda>{init_bio+266}
>        <ffffffff8010b25a>{init+506} <ffffffff8010ed16>{child_rip+8}
>        <ffffffff8010b060>{init+0} <ffffffff8010ed0e>{child_rip+0}
>
> ---------------------------
> | preempt count: 00000000 ]
> | 0-level deep critical section nesting:
> ----------------------------------------
>

The acpi code is filled with macro hell (and craziness like using a void
pointer for a spinlock).  Do you have the following turned on?

CONFIG_USE_FRAME_POINTER=y
CONFIG_FRAME_POINTER=y


The back trace looks a little suspicious.  But I'm looking into the acpi
code to see if I can locate the bad flags.

-- Steve