Robert Jarzmik <robert.jarzmik@xxxxxxx> writes:
> Guennadi Liakhovetski <g.liakhovetski@xxxxxx> writes:
>
>>> + for (i = 0; i < pcdev->channels; i++) {
>>> + pcdev->sg_tail[i] = buf->dmas[i].sg_cpu + buf->dmas[i].sglen;
>>> + pcdev->sg_tail[i]->ddadr = DDADR_STOP;
>
>> This function is now called "live" with running DMA, and you first append
>> the chain, and only then terminate it... It should be ok because it is
>> done with switched off IRQs, and DMA must be still at tail - 1 to
>> automatically continue onto the appended chain, so, you should have enough
>> time in 100% of cases, still it would look better to first terminate the
>> chain and then append it.
> Correct. I'll invert the 2 assignments.
At second thought, I think I'll change this. The first assignment doesn't append
the chain, it just moves where sg_tail points at. The "chain append" was
previously done in "pxa_dma_add_tail_buf".
So the correct thing to do is to displace the "DDADR_STOP" assignment into
"pxa_dma_add_tail_buf", to make the "STOP" on the last descriptor of the newest
tail buffer. This "STOP" should be set up _before_ the buffer is queued.
I'll amend that.
Cheers.
--
Robert
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
