Re: [PATCH] dma: shdma: transfer based runtime PM

From: Guennadi Liakhovetski
Date: Tue Aug 30 2011 - 03:12:14 EST


On Mon, 29 Aug 2011, Vinod Koul wrote:

> On Mon, 2011-08-29 at 14:21 +0200, Guennadi Liakhovetski wrote:
> > On Mon, 29 Aug 2011, Vinod Koul wrote:
> >
> > > On Fri, 2011-08-26 at 01:11 +0200, Guennadi Liakhovetski wrote:
> > > > On Thu, 25 Aug 2011, Koul, Vinod wrote:
> > > > > Wont it be easy to to do:
> > > > > - pm_runtime_get() in each submit
> > > > > - pm_runtime_put() in each callback
> > > > > Normal case above would work just fine
> > > > > - In terminate case, count the number of issued transactions, and call
> > > > > pm_runtime_put() for each canceled transaction
> > > > > (i am assuming that for each timeout error, the client will call
> > > > > terminate)
> > > >
> > > > As I said, this won't be very easy to do this in a robust way. You'd have
> > > > to scan your list of DMA blocks and see, which of them belong to one
> > > > descriptor, and once you reach the end of that descriptor, issue a put().
> > > > Perhaps, this can be done, but my choice went to the currently presented
> > > > solution.
> > > If you count the number of descriptor submitted in your submitted list
> > > and call _put for each, I see no reason why it wont be simple and better
> > > than current approach.
> >
> > Sorry, I thought, you wanted to avoid extra counting, because runtime-pm
> > counts by itself. Now you propose to count...
> see below it doesn't count but iterate thru list

Yes, sorry, iterating is a better word, but I actually meant "counting" in
a broad sense, of which "iterating" is a particular case:-)

> > > Something like:
> > > /* since callback is set for last descriptor of chain, we call runtime
> > > * put for that desc alone
> > > */
> > > list_for_each_entry_safe(desc, __desc, sh_chan->ld_queue, node) {
> > > if (desc->async_tx.callback)
> > > pm_runtime_put(device);
> >
> > Not all dma users have callbacks.
> Do you have such usage today, at least I dont :)
> Nevertheless, in tx_submit adding a simple flag in your drivers
> descriptor structure can tell you whether to call _put() or not. Agreed?

Yes, I agree, that one could make this work too. Still, I do not
understand how and why this is better to the extent, that I have to
reimplement my patch, retest and resubmit it. Maybe Dan or Paul have an
opinion on this?

Thanks
Guennadi
---
Guennadi Liakhovetski, Ph.D.
Freelance Open-Source Software Developer
http://www.open-technology.de/
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/