On Wed, Jul 16, 2014 at 06:40:01PM +0200, Sebastian Andrzej Siewior wrote:

On 07/16/2014 06:06 PM, Felipe Balbi wrote:

quoted

quoted

quoted

well, other than in probe and other functions which need to
make sure clocks are on, but it seems unnecessary to
enable/disable in every function.

What do you have in mind? Do you plan to let the uart on while
the minicom is attached but is doing nothing? In that case,
->startup() and ->shutdown() should be enough.

no the idea was to keep it on for as long as it's transferring 
characters and idle it otherwise, if that can't be done easily,
then I guess your way is the only way.

But maybe we have to add some additional logic here to keep it up for
the transfer. I've been just (maybe over)thinking: If you send 300
bytes over DMA via 300 baud it should take 10 seconds. The PM-timeout
could hit before the transfer is complete.

hmm, good point. So we _do_ need a way to get early and only put after
we know transfer has completed and I was assuming stop_tx() to be that
hint, but apparently it isn't.

Same thing with hw-flowcontrol where you could get stalled for a few
seconds.
However it doesn't seem to be a problem in current omap-serial driver.

I don't think current omap-serial driver is a great example.

-- 
balbi
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: Digital signature
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20140716/7ee27622/attachment.sig>