Messages in this thread | | | Date | Thu, 4 Dec 2008 21:03:59 +0100 (CET) | From | Guennadi Liakhovetski <> | Subject | Re: [PATCH 0/4] dmaengine redux continued |
| |
On Thu, 4 Dec 2008, Dan Williams wrote:
> On Thu, Dec 4, 2008 at 1:51 AM, Guennadi Liakhovetski > <g.liakhovetski@gmx.de> wrote: > > What about the last hunk from the patch at > > > > http://marc.info/?l=linux-arm-kernel&m=122607472721145&w=2 > > > > ? Can we add this tx_free pointer? Would you like me to submit it as a > > separate patch? I'd like to be able to reuse tx-descriptors in clients, so > > I then have to explicitly release them in the dmaengine device > > (controller) driver. > > > > I took a look but the description does not explain why the driver > needs an explicit release. In what cases is the implied release, when > the descriptor is acked && done, insufficient?
v4l. I think, the proper moment to free tx-descriptors is at VIDIOC_DQBUF time. I.e., there is an explicit ioctl from user-space: "free buffers now". That's when I want to call .tx_free().
Thanks Guennadi --- Guennadi Liakhovetski, Ph.D. Freelance Open-Source Software Developer
| |