Re: [PATCH v2] dmaengine: Create symlinks between DMA channels and slaves

From: Peter Ujfalusi
Date: Thu Jan 30 2020 - 05:22:10 EST


Hi Geert,

On 30/01/2020 11.51, Geert Uytterhoeven wrote:
> Hi Peter,
>
> On Thu, Jan 30, 2020 at 10:42 AM Peter Ujfalusi <peter.ujfalusi@xxxxxx> wrote:
>> On 17/01/2020 17.30, Geert Uytterhoeven wrote:
>>> Currently it is not easy to find out which DMA channels are in use, and
>>> which slave devices are using which channels.
>>>
>>> Fix this by creating two symlinks between the DMA channel and the actual
>>> slave device when a channel is requested:
>>> 1. A "slave" symlink from DMA channel to slave device,
>>> 2. A "dma:<name>" symlink slave device to DMA channel.
>>> When the channel is released, the symlinks are removed again.
>>> The latter requires keeping track of the slave device and the channel
>>> name in the dma_chan structure.
>>>
>>> Note that this is limited to channel request functions for requesting an
>>> exclusive slave channel that take a device pointer (dma_request_chan()
>>> and dma_request_slave_channel*()).
>>>
>>> Signed-off-by: Geert Uytterhoeven <geert+renesas@xxxxxxxxx>
>
>>> --- a/drivers/dma/dmaengine.c
>>> +++ b/drivers/dma/dmaengine.c
>>> @@ -60,6 +60,8 @@ static long dmaengine_ref_count;
>>>
>>> /* --- sysfs implementation --- */
>>>
>>> +#define DMA_SLAVE_NAME "slave"
>>> +
>>> /**
>>> * dev_to_dma_chan - convert a device pointer to its sysfs container object
>>> * @dev - device node
>>> @@ -730,11 +732,11 @@ struct dma_chan *dma_request_chan(struct device *dev, const char *name)
>>> if (has_acpi_companion(dev) && !chan)
>>> chan = acpi_dma_request_slave_chan_by_name(dev, name);
>>>
>>> - if (chan) {
>>> - /* Valid channel found or requester needs to be deferred */
>>> - if (!IS_ERR(chan) || PTR_ERR(chan) == -EPROBE_DEFER)
>>> - return chan;
>>> - }
>>> + if (PTR_ERR(chan) == -EPROBE_DEFER)
>>> + return chan;
>>> +
>>> + if (!IS_ERR_OR_NULL(chan))
>>> + goto found;
>>>
>>> /* Try to find the channel via the DMA filter map(s) */
>>> mutex_lock(&dma_list_mutex);
>>> @@ -754,7 +756,23 @@ struct dma_chan *dma_request_chan(struct device *dev, const char *name)
>>> }
>>> mutex_unlock(&dma_list_mutex);
>>>
>>> - return chan ? chan : ERR_PTR(-EPROBE_DEFER);
>>> + if (!IS_ERR_OR_NULL(chan))
>>> + goto found;
>>> +
>>> + return ERR_PTR(-EPROBE_DEFER);
>>> +
>>> +found:
>>> + chan->slave = dev;
>>> + chan->name = kasprintf(GFP_KERNEL, "dma:%s", name);
>>> + if (!chan->name)
>>> + return ERR_PTR(-ENOMEM);
>>
>> You will lock the channel... It is requested, but it is not released in
>> case of failure.
>
> True. Perhaps this error should just be ignored, cfr. below.
> However, if this operation fails, chances are high the system will die very soon
> anyway.

Yeah, I'll fix it up in a series I'm preparing.

>
>>> +
>>> + if (sysfs_create_link(&chan->dev->device.kobj, &dev->kobj,
>>> + DMA_SLAVE_NAME))
>>> + dev_err(dev, "Cannot create DMA %s symlink\n", DMA_SLAVE_NAME);
>>> + if (sysfs_create_link(&dev->kobj, &chan->dev->device.kobj, chan->name))
>>> + dev_err(dev, "Cannot create DMA %s symlink\n", chan->name);
>>
>> It is not a problem if these fail?
>
> IMHO, a failure to create these links is not fatal for the operation of
> the device, and thus can be ignored. Just like for debugfs.

OK, then these should not be dev_err, but dev_warn.
I'll include this is also in a fixup patch.

>
>>> + return chan;
>>> }
>>> EXPORT_SYMBOL_GPL(dma_request_chan);
>>>
>>> @@ -812,6 +830,13 @@ 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);
>>> + if (chan->slave) {
>>> + sysfs_remove_link(&chan->slave->kobj, chan->name);
>>> + kfree(chan->name);
>>> + chan->name = NULL;
>>> + chan->slave = NULL;
>>> + }
>>> + sysfs_remove_link(&chan->dev->device.kobj, DMA_SLAVE_NAME);
>>
>> If a non slave channel is released, then you remove the link you have
>> never created?
>>
>> What happens if the link creation fails and here you attempt to remove
>> the failed ones?
>
> sysfs_remove_link() should handle removing non-existent links, and just
> return -ENOENT.

True, just followed the call chain and tested as well, but the
DMA_SLAVE_NAME symlink should be also removed within the
if (chan->slave) {} block as it is never created for non slave channels.

Also including inn my fixup patch.

>
> Gr{oetje,eeting}s,
>
> Geert
>

- PÃter

Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki.
Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki