[dmaengine] [Q] jiffies value does not increase in dma_sync_wait()

[dmaengine] [Q] jiffies value does not increase in dma_sync_wait()

[javier Martin]

Hi,
I have a Visstrim_M10 board, based on i.MX27, and I'm developing a
v4l2 driver for deinterlacing video frames.

Whenever I start a new dma transfer I call the function
"dma_wait_for_async_tx()" which internally ends up calling
"dma_sync_wait()":

http://lxr.linux.no/#linux+v3.2.2/drivers/dma/dmaengine.c#L255

In this function, there is a "do while" loop, which checks for dma
completion, with a timeout. However, when the system is too loaded
this function enters this "do while" loop and never gets out of it,
blocking the system.

I've introduced a couple of printk() to check why this timeout is not
triggered and I've found that the value of jiffies does not increase
between loop iterations (i. e. it's like time didn't advance).

Does anyobody know what reasons could make jiffies not being updated?

Regards.
-- 
Javier Martin
Vista Silicon S.L.
CDTUC - FASE C - Oficina S-345
Avda de los Castros s/n
39005- Santander. Cantabria. Spain
+34 942 25 32 60
www.vista-silicon.com

[dmaengine] [Q] jiffies value does not increase in dma_sync_wait()

[Russell King - ARM Linux]

On Fri, Feb 03, 2012 at 09:37:48AM +0100, javier Martin wrote:
> I've introduced a couple of printk() to check why this timeout is not
> triggered and I've found that the value of jiffies does not increase
> between loop iterations (i. e. it's like time didn't advance).
> 
> Does anyobody know what reasons could make jiffies not being updated?

Are interrupts disabled?

[dmaengine] [Q] jiffies value does not increase in dma_sync_wait()

[javier Martin]

Hi Russell,

On 3 February 2012 10:10, Russell King - ARM Linux
<linux@arm.linux.org.uk> wrote:
> On Fri, Feb 03, 2012 at 09:37:48AM +0100, javier Martin wrote:
>> I've introduced a couple of printk() to check why this timeout is not
>> triggered and I've found that the value of jiffies does not increase
>> between loop iterations (i. e. it's like time didn't advance).
>>
>> Does anyobody know what reasons could make jiffies not being updated?
>
> Are interrupts disabled?

Apparently not but, how could I check it for sure? Is
"irqs_disabled()" suitable for that purpose?

http://lxr.linux.no/#linux+v3.2.2/include/linux/irqflags.h#L120

-- 
Javier Martin
Vista Silicon S.L.
CDTUC - FASE C - Oficina S-345
Avda de los Castros s/n
39005- Santander. Cantabria. Spain
+34 942 25 32 60
www.vista-silicon.com

[dmaengine] [Q] jiffies value does not increase in dma_sync_wait()

[Russell King - ARM Linux]

On Fri, Feb 03, 2012 at 10:22:00AM +0100, javier Martin wrote:
> Hi Russell,
> 
> On 3 February 2012 10:10, Russell King - ARM Linux
> <linux@arm.linux.org.uk> wrote:
> > On Fri, Feb 03, 2012 at 09:37:48AM +0100, javier Martin wrote:
> >> I've introduced a couple of printk() to check why this timeout is not
> >> triggered and I've found that the value of jiffies does not increase
> >> between loop iterations (i. e. it's like time didn't advance).
> >>
> >> Does anyobody know what reasons could make jiffies not being updated?
> >
> > Are interrupts disabled?
> 
> Apparently not but, how could I check it for sure? Is
> "irqs_disabled()" suitable for that purpose?

Yes.

[dmaengine] [Q] jiffies value does not increase in dma_sync_wait()

[javier Martin]

On 3 February 2012 10:24, Russell King - ARM Linux
<linux@arm.linux.org.uk> wrote:
> On Fri, Feb 03, 2012 at 10:22:00AM +0100, javier Martin wrote:
>> Hi Russell,
>>
>> On 3 February 2012 10:10, Russell King - ARM Linux
>> <linux@arm.linux.org.uk> wrote:
>> > On Fri, Feb 03, 2012 at 09:37:48AM +0100, javier Martin wrote:
>> >> I've introduced a couple of printk() to check why this timeout is not
>> >> triggered and I've found that the value of jiffies does not increase
>> >> between loop iterations (i. e. it's like time didn't advance).
>> >>
>> >> Does anyobody know what reasons could make jiffies not being updated?
>> >
>> > Are interrupts disabled?
>>
>> Apparently not but, how could I check it for sure? Is
>> "irqs_disabled()" suitable for that purpose?
>
> Yes.

"irqs_disabled()" returns 0 in every iteration of the loop. So I guess
this means IRQs are properly enabled, but jiffies keeps being fixed.


-- 
Javier Martin
Vista Silicon S.L.
CDTUC - FASE C - Oficina S-345
Avda de los Castros s/n
39005- Santander. Cantabria. Spain
+34 942 25 32 60
www.vista-silicon.com