On 12/17/19 12:19 PM, Viresh Kumar wrote:
> On 17-12-19, 12:50, Peter Ujfalusi wrote:
>> dma_request_slave_channel() is a wrapper on top of dma_request_chan()
>> eating up the error code.
>>
>> By using dma_request_chan() directly the driver can support deferred
>> probing against DMA.
It doesn't seem to be the case as DMA channel is requested at the start
of the data transfer (which happens after the driver has been successfully
probed).
PS there is a typo in the patch summary (it should "pata_arasan_cf").
Best regards,
--
Bartlomiej Zolnierkiewicz
Samsung R&D Institute Poland
Samsung Electronics
>> Signed-off-by: Peter Ujfalusi <[email protected]>
>> ---
>> drivers/ata/pata_arasan_cf.c | 6 ++++--
>> 1 file changed, 4 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/ata/pata_arasan_cf.c b/drivers/ata/pata_arasan_cf.c
>> index 135173c8d138..69b555d83f68 100644
>> --- a/drivers/ata/pata_arasan_cf.c
>> +++ b/drivers/ata/pata_arasan_cf.c
>> @@ -526,9 +526,10 @@ static void data_xfer(struct work_struct *work)
>>
>> /* request dma channels */
>> /* dma_request_channel may sleep, so calling from process context */
>> - acdev->dma_chan = dma_request_slave_channel(acdev->host->dev, "data");
>> - if (!acdev->dma_chan) {
>> + acdev->dma_chan = dma_request_chan(acdev->host->dev, "data");
>> + if (IS_ERR(acdev->dma_chan)) {
>> dev_err(acdev->host->dev, "Unable to get dma_chan\n");
>> + acdev->dma_chan = NULL;
>> goto chan_request_fail;
>> }
>>
>> @@ -539,6 +540,7 @@ static void data_xfer(struct work_struct *work)
>> }
>>
>> dma_release_channel(acdev->dma_chan);
>> + acdev->dma_chan = NULL;
>>
>> /* data xferred successfully */
>> if (!ret) {
>
> Acked-by: Viresh Kumar <[email protected]>
>
On 13/01/2020 13.31, Bartlomiej Zolnierkiewicz wrote:
>
> On 12/17/19 12:19 PM, Viresh Kumar wrote:
>> On 17-12-19, 12:50, Peter Ujfalusi wrote:
>>> dma_request_slave_channel() is a wrapper on top of dma_request_chan()
>>> eating up the error code.
>>>
>>> By using dma_request_chan() directly the driver can support deferred
>>> probing against DMA.
>
> It doesn't seem to be the case as DMA channel is requested at the start
> of the data transfer (which happens after the driver has been successfully
> probed).
True, I have updated the commit message to remove the reference to
deferred probing.
If the DMA is requested upfront (at probe time, device open time?) the
driver would save quite a bit of time by not allocating and freeing the
DMA resources repeatedly for each transfer, thus most likely giving a
boost to throughput...
> PS there is a typo in the patch summary (it should "pata_arasan_cf").
Ah, fixed this as well in v2.
Thanks for catching them,
- Péter
>
> Best regards,
> --
> Bartlomiej Zolnierkiewicz
> Samsung R&D Institute Poland
> Samsung Electronics
>
>>> Signed-off-by: Peter Ujfalusi <[email protected]>
>>> ---
>>> drivers/ata/pata_arasan_cf.c | 6 ++++--
>>> 1 file changed, 4 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/drivers/ata/pata_arasan_cf.c b/drivers/ata/pata_arasan_cf.c
>>> index 135173c8d138..69b555d83f68 100644
>>> --- a/drivers/ata/pata_arasan_cf.c
>>> +++ b/drivers/ata/pata_arasan_cf.c
>>> @@ -526,9 +526,10 @@ static void data_xfer(struct work_struct *work)
>>>
>>> /* request dma channels */
>>> /* dma_request_channel may sleep, so calling from process context */
>>> - acdev->dma_chan = dma_request_slave_channel(acdev->host->dev, "data");
>>> - if (!acdev->dma_chan) {
>>> + acdev->dma_chan = dma_request_chan(acdev->host->dev, "data");
>>> + if (IS_ERR(acdev->dma_chan)) {
>>> dev_err(acdev->host->dev, "Unable to get dma_chan\n");
>>> + acdev->dma_chan = NULL;
>>> goto chan_request_fail;
>>> }
>>>
>>> @@ -539,6 +540,7 @@ static void data_xfer(struct work_struct *work)
>>> }
>>>
>>> dma_release_channel(acdev->dma_chan);
>>> + acdev->dma_chan = NULL;
>>>
>>> /* data xferred successfully */
>>> if (!ret) {
>>
>> Acked-by: Viresh Kumar <[email protected]>
>>
>
Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki.
Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki