BUG: scheduling while atomic

[Colin McCabe]

On Thu, Jul 29, 2010 at 11:32 PM, Dan Williams <dcbw at redhat.com> wrote:
> On Thu, 2010-07-29 at 17:00 +1000, Amit Hergass wrote:
>> Hi,
>> I am developing the LTIB linux kernel 2.6.34 for  Phythec board
>> phy3250.
>>
>> The libertas SPI was integrated and communicate with the external
>> world.
>
> Yeah, looks like the SPI write routines are called from main.c under
> spin_lock_irq() for data packets which is probably causing that warning.
> SDIO doesn't get it because it uses a workqueue to shovel data to the
> firmware later.  I supposed the driver shouldn't be doing this here, but
> it'll take a bit of work to make sure it doesn't.  Thoughts anyone?

Easy solution: back out change
86c34fe89e9cad9e1ba4d1a8bbf98259035f4caf, which introduced the
scheduling-while-atomic bug.

Hard solution: fix the driver core so it doesn't call hw_host_to_card
in atomic context.

There was another email thread about this last year. Unfortunately, I
don't think the issue was fixed. Also unfortunately, I no longer have
access to the hardware.

regards,
Colin McCabe