On 07-06-19, 13:38, Geert Uytterhoeven wrote:
> Currently it is not easy to find out which DMA channels are in use, and
> by which slave devices.
>
> Fix this by creating in sysfs a "slave" symlink from the DMA channel to
> the actual slave device when a channel is requested, and removing it
> again when the channel is released.
>
> For now this is limited to DT and ACPI.
>
> Signed-off-by: Geert Uytterhoeven <geert+renesas@xxxxxxxxx>
> ---
> Questions:
> 1. Do you think this is useful?
Yes for me at least :)
> 2. Should backlinks (e.g. "dma:<name>") be created from the slave
> device to the DMA channel?
> This requires storing the name in struct dma_chan, for later
> symlink removal.
that would certainly be more helpful
> 3. Should this be extended to other ways of requesting channels?
> In many cases, no device pointer is available, so a device pointer
> parameter has to be added to all DMA channel request APIs that
> don't have it yet.
I think that would need to be done.
> ---
> drivers/dma/dmaengine.c | 5 +++++
> 1 file changed, 5 insertions(+)
>
> diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c
> index 03ac4b96117cd8db..c11476f76fc96bcf 100644
> --- a/drivers/dma/dmaengine.c
> +++ b/drivers/dma/dmaengine.c
> @@ -706,6 +706,10 @@ struct dma_chan *dma_request_chan(struct device *dev, const char *name)
>
> if (chan) {
> /* Valid channel found or requester needs to be deferred */
> + if (!IS_ERR(chan) &&
> + sysfs_create_link(&chan->dev->device.kobj, &dev->kobj,
> + "slave"))
> + dev_err(dev, "Cannot create DMA slave symlink\n");
> if (!IS_ERR(chan) || PTR_ERR(chan) == -EPROBE_DEFER)
> return chan;
> }
> @@ -786,6 +790,7 @@ void dma_release_channel(struct dma_chan *chan)
> /* drop PRIVATE cap enabled by __dma_request_channel() */
> if (--chan->device->privatecnt == 0)
> dma_cap_clear(DMA_PRIVATE, chan->device->cap_mask);
> + sysfs_remove_link(&chan->dev->device.kobj, "slave");
> mutex_unlock(&dma_list_mutex);
> }
> EXPORT_SYMBOL_GPL(dma_release_channel);
> --
> 2.17.1
--
~Vinod
