[PATCH] ep93xx: Implement double buffering for M2M DMA channels

[H Hartley Sweeten]

On Wednesday, March 21, 2012 12:33 PM, Mika Westerberg wrote:
> On Wed, Mar 21, 2012 at 12:12:11PM -0500, H Hartley Sweeten wrote:
>> On Tuesday, March 20, 2012 1:09 AM, Rafal Prylowski wrote:
>>>
>>> Implement double buffering for M2M DMA channels.
>>>
>>> Signed-off-by: Rafal Prylowski <prylowski at metasoft.pl>
>> 
>> Hmm... This isn't working on my ep9307 board...
>> 
>> When the system boots I get stormed with messages:
>> 
>> dma dma1chan0: unknown interrupt!
>> dma dma1chan0: unknown interrupt!
>> dma dma1chan0: unknown interrupt!
>> dma dma1chan1: unknown interrupt!
>> dma dma1chan1: unknown interrupt!
>> dma dma1chan1: unknown interrupt!
>> dma dma1chan0: unknown interrupt!
>> dma dma1chan0: unknown interrupt!
>> dma dma1chan1: unknown interrupt!
>> 
>> Note, I have use_dma set for the ep93xx_spi driver and am using mmc_spi.
>> 
>> Also note, without your patch I also get a couple messages:
>> 
>> dma dma1chan1: got interrupt while active list is empty
>> 
>> But, the mmc card is working fine. I haven't tracked down why I get these
>> yet. Mika, any ideas?
>
> I've noticed the same. But they also happen without this patch applied. I've
> tracked it down to MULTI_IRQ_HANDLER support patches for ARM vic but not sure
> what might be wrong there (or is the ep93xx_dma driver behaving badly). I
> Noticed that when we get that 'got interrupt while active list is empty' we
> don't have any interrupts set in the M2M_INTERRUPT register but we still got
> an interrupt.. weird.

I think you misunderstood my comment above.

With this patch applied I get the "unknown interrupt!" storm. These messages
keep getting spewed until I power off the board.

Without this patch I get the "got interrupt while active list is empty" messages
but only occasionally. Other than the messages the mmc_spi driver seems to
be working ok with dma.

I hacked in a dump of the DMA Global Interrupt register when I get the
"got interrupt while active list is empty" messages and get this:

dma dma1chan1: got interrupt while active list is empty (00000000)

So, according to the DMAGlInt register, there are no channels with an
active interrupt.

Are we missing a write to the INTERRUPT registers somewhere to clear the
current interrupt?

Regards,
Hartley

> Below is a workaround I've been using lately but it certainly is a hack and we
> should figure out what is the root cause for this.
>
> diff --git a/arch/arm/common/vic.c b/arch/arm/common/vic.c
> index dcb004a..cb6b49a 100644
> --- a/arch/arm/common/vic.c
> +++ b/arch/arm/common/vic.c
> @@ -441,11 +441,9 @@ static int handle_one_vic(struct vic_device *vic, struct pt_regs *regs)
>  	u32 stat, irq;
>  	int handled = 0;
>  
> -	stat = readl_relaxed(vic->base + VIC_IRQ_STATUS);
> -	while (stat) {
> +	while ((stat = readl_relaxed(vic->base + VIC_IRQ_STATUS))) {
>  		irq = ffs(stat) - 1;
>  		handle_IRQ(irq_domain_to_irq(&vic->domain, irq), regs);
> -		stat &= ~(1 << irq);
>  		handled = 1;
>  	}