On Thu, Nov 13, 2014 at 3:14 PM, Arnd Bergmann <arnd@xxxxxxxx> wrote:
> On Thursday 13 November 2014 15:07:06 Andrew Bresticker wrote:
>> On Thu, Nov 13, 2014 at 2:13 PM, Arnd Bergmann <arnd@xxxxxxxx> wrote:
>> > On Thursday 13 November 2014 12:58:08 Andrew Bresticker wrote:
>> >> +
>> >> +static bool mdc_filter_fn(struct dma_chan *chan, void *fn_param)
>> >> +{
>> >> + struct mdc_filter_data *data = fn_param;
>> >> + struct mdc_chan *mchan;
>> >> +
>> >> + if (chan->device->dev->driver == &mdc_dma_driver.driver) {
>> >> + mchan = to_mdc_chan(chan);
>> >> + if (!(data->mask & BIT(mchan->chan_nr)))
>> >> + return false;
>> >> + mchan->periph = data->periph;
>> >> + mchan->thread = data->thread;
>> >> + return true;
>> >> + }
>> >> + return false;
>> >> +}
>> >> +
>> >> +static struct dma_chan *mdc_of_xlate(struct of_phandle_args *dma_spec,
>> >> + struct of_dma *ofdma)
>> >> +{
>> >> + struct mdc_dma *mdma = ofdma->of_dma_data;
>> >> + struct mdc_filter_data data;
>> >> +
>> >> + if (dma_spec->args_count != 3)
>> >> + return NULL;
>> >> +
>> >> + data.periph = dma_spec->args[0];
>> >> + data.mask = dma_spec->args[1];
>> >> + data.thread = dma_spec->args[2];
>> >> +
>> >> + return dma_request_channel(mdma->dma_dev.cap_mask, mdc_filter_fn,
>> >> + &data);
>> >> +}
>> >
>> > The filter function is broken if you ever have multiple instances
>> > of the device. Better avoid calling dma_request_channel and scan
>> > the channels that the device knows about.
>>
>> It seems unlikely that there would be multiple instances of this IP in
>> a system, but it doesn't hurt to be safe. Perhaps instead of
>> iterating through the list here I could extend struct mdc_filter_data
>> to include a pointer to the struct dma_device corresponding to this
>> instance of the MDC and compare based on that.
>
> You still need to iterate through all channels of all dma engines
> in the system. Since you have the pointer to the engine here,
> just go through the channels that you already know are there.
> Unfortunately, the mask prevents you from just using
> dma_get_any_slave_channel(), that would be the easiest approach
> that we prefer in most drivers these days.
Ok, I'll do that then.
--
To unsubscribe from this list: send the line "unsubscribe dmaengine" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
