| Message ID | 20180924130021.20530-3-peter.ujfalusi@ti.com (mailing list archive) |
|---|---|
| State | RFC |
| Headers | show |
| Series | dmaengine: ti: New DMA driver for Texas Instruments UDMA | expand |
On 24-09-18, 16:00, Peter Ujfalusi wrote: > dma_get_any_slave_channel() would skip using the filter function, which > in some cases needed to be executed before the alloc_chan_resources > callback to make sure that all parameters are provided for the slave > channel. Another request API, i though you had solved that last time around :( > > Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com> > --- > drivers/dma/dmaengine.c | 20 ++++++++++++++++++++ > include/linux/dmaengine.h | 2 ++ > 2 files changed, 22 insertions(+) > > diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c > index 27b6d7c2d8a0..b80f1bdd8813 100644 > --- a/drivers/dma/dmaengine.c > +++ b/drivers/dma/dmaengine.c > @@ -634,6 +634,26 @@ struct dma_chan *dma_get_any_slave_channel(struct dma_device *device) > } > EXPORT_SYMBOL_GPL(dma_get_any_slave_channel); > > +struct dma_chan *dmadev_get_slave_channel(struct dma_device *device, > + dma_filter_fn fn, void *fn_param) > +{ > + dma_cap_mask_t mask; > + struct dma_chan *chan; > + > + dma_cap_zero(mask); > + dma_cap_set(DMA_SLAVE, mask); > + > + /* lock against __dma_request_channel */ > + mutex_lock(&dma_list_mutex); > + > + chan = find_candidate(device, &mask, fn, fn_param); > + > + mutex_unlock(&dma_list_mutex); > + > + return IS_ERR(chan) ? NULL : chan; > +} > +EXPORT_SYMBOL_GPL(dmadev_get_slave_channel); > + > /** > * __dma_request_channel - try to allocate an exclusive channel > * @mask: capabilities that the channel must satisfy > diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h > index 16c9d021988a..32010c6ae13a 100644 > --- a/include/linux/dmaengine.h > +++ b/include/linux/dmaengine.h > @@ -1542,6 +1542,8 @@ void dma_async_device_unregister(struct dma_device *device); > void dma_run_dependencies(struct dma_async_tx_descriptor *tx); > struct dma_chan *dma_get_slave_channel(struct dma_chan *chan); > struct dma_chan *dma_get_any_slave_channel(struct dma_device *device); > +struct dma_chan *dmadev_get_slave_channel(struct dma_device *device, > + dma_filter_fn fn, void *fn_param); > #define dma_request_channel(mask, x, y) __dma_request_channel(&(mask), x, y) > #define dma_request_slave_channel_compat(mask, x, y, dev, name) \ > __dma_request_slave_channel_compat(&(mask), x, y, dev, name) > -- > Peter > > Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. > Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
On 2018-10-09 13:49, Vinod wrote: > On 24-09-18, 16:00, Peter Ujfalusi wrote: >> dma_get_any_slave_channel() would skip using the filter function, which >> in some cases needed to be executed before the alloc_chan_resources >> callback to make sure that all parameters are provided for the slave >> channel. > > Another request API, i though you had solved that last time around :( Yes, I thought so, but this time it is different. I want to drop this patch myself also, but I need to figure out a way to do what the dmadev_get_slave_channel() allows me to do. In path 4, udma_of_xlate() and the udma_dma_filter_fn() It might be possible that I look up a free udma_chan, do the setup via the filter_fn() and finally call the dma_get_slave_channel(). Might work fine, but I wanted to avoid to implement my own find free channel code when we have it already in DMAengine core. >> >> Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com> >> --- >> drivers/dma/dmaengine.c | 20 ++++++++++++++++++++ >> include/linux/dmaengine.h | 2 ++ >> 2 files changed, 22 insertions(+) >> >> diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c >> index 27b6d7c2d8a0..b80f1bdd8813 100644 >> --- a/drivers/dma/dmaengine.c >> +++ b/drivers/dma/dmaengine.c >> @@ -634,6 +634,26 @@ struct dma_chan *dma_get_any_slave_channel(struct dma_device *device) >> } >> EXPORT_SYMBOL_GPL(dma_get_any_slave_channel); >> >> +struct dma_chan *dmadev_get_slave_channel(struct dma_device *device, >> + dma_filter_fn fn, void *fn_param) >> +{ >> + dma_cap_mask_t mask; >> + struct dma_chan *chan; >> + >> + dma_cap_zero(mask); >> + dma_cap_set(DMA_SLAVE, mask); >> + >> + /* lock against __dma_request_channel */ >> + mutex_lock(&dma_list_mutex); >> + >> + chan = find_candidate(device, &mask, fn, fn_param); >> + >> + mutex_unlock(&dma_list_mutex); >> + >> + return IS_ERR(chan) ? NULL : chan; >> +} >> +EXPORT_SYMBOL_GPL(dmadev_get_slave_channel); >> + >> /** >> * __dma_request_channel - try to allocate an exclusive channel >> * @mask: capabilities that the channel must satisfy >> diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h >> index 16c9d021988a..32010c6ae13a 100644 >> --- a/include/linux/dmaengine.h >> +++ b/include/linux/dmaengine.h >> @@ -1542,6 +1542,8 @@ void dma_async_device_unregister(struct dma_device *device); >> void dma_run_dependencies(struct dma_async_tx_descriptor *tx); >> struct dma_chan *dma_get_slave_channel(struct dma_chan *chan); >> struct dma_chan *dma_get_any_slave_channel(struct dma_device *device); >> +struct dma_chan *dmadev_get_slave_channel(struct dma_device *device, >> + dma_filter_fn fn, void *fn_param); >> #define dma_request_channel(mask, x, y) __dma_request_channel(&(mask), x, y) >> #define dma_request_slave_channel_compat(mask, x, y, dev, name) \ >> __dma_request_slave_channel_compat(&(mask), x, y, dev, name) >> -- >> Peter >> >> Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. >> Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki > - Péter Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
On 2018-10-09 14:22, Peter Ujfalusi wrote: > > > On 2018-10-09 13:49, Vinod wrote: >> On 24-09-18, 16:00, Peter Ujfalusi wrote: >>> dma_get_any_slave_channel() would skip using the filter function, which >>> in some cases needed to be executed before the alloc_chan_resources >>> callback to make sure that all parameters are provided for the slave >>> channel. >> >> Another request API, i though you had solved that last time around :( > > Yes, I thought so, but this time it is different. > I want to drop this patch myself also, but I need to figure out a way to > do what the dmadev_get_slave_channel() allows me to do. > > In path 4, udma_of_xlate() and the udma_dma_filter_fn() > > It might be possible that I look up a free udma_chan, do the setup via > the filter_fn() and finally call the dma_get_slave_channel(). > Might work fine, but I wanted to avoid to implement my own find free > channel code when we have it already in DMAengine core. It is going to be racy against parallel channel requests, but >>> diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h >>> index 16c9d021988a..32010c6ae13a 100644 >>> --- a/include/linux/dmaengine.h >>> +++ b/include/linux/dmaengine.h >>> @@ -1542,6 +1542,8 @@ void dma_async_device_unregister(struct dma_device *device); >>> void dma_run_dependencies(struct dma_async_tx_descriptor *tx); >>> struct dma_chan *dma_get_slave_channel(struct dma_chan *chan); >>> struct dma_chan *dma_get_any_slave_channel(struct dma_device *device); >>> +struct dma_chan *dmadev_get_slave_channel(struct dma_device *device, >>> + dma_filter_fn fn, void *fn_param); -struct dma_chan *dma_get_any_slave_channel(struct dma_device *device); +struct dma_chan *dmadev_get_slave_channel(struct dma_device *device, dma_filter_fn fn, void *fn_param); +#define dma_get_any_slave_channel(device)\ dmadev_get_slave_channel(device, NULL, NULL) Should cut down on duplicated code at least. >>> #define dma_request_channel(mask, x, y) __dma_request_channel(&(mask), x, y) >>> #define dma_request_slave_channel_compat(mask, x, y, dev, name) \ >>> __dma_request_slave_channel_compat(&(mask), x, y, dev, name) >>> -- >>> Peter >>> >>> Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. >>> Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki >> > > - Péter > > Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. > Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki > - Péter Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c index 27b6d7c2d8a0..b80f1bdd8813 100644 --- a/drivers/dma/dmaengine.c +++ b/drivers/dma/dmaengine.c @@ -634,6 +634,26 @@ struct dma_chan *dma_get_any_slave_channel(struct dma_device *device) } EXPORT_SYMBOL_GPL(dma_get_any_slave_channel); +struct dma_chan *dmadev_get_slave_channel(struct dma_device *device, + dma_filter_fn fn, void *fn_param) +{ + dma_cap_mask_t mask; + struct dma_chan *chan; + + dma_cap_zero(mask); + dma_cap_set(DMA_SLAVE, mask); + + /* lock against __dma_request_channel */ + mutex_lock(&dma_list_mutex); + + chan = find_candidate(device, &mask, fn, fn_param); + + mutex_unlock(&dma_list_mutex); + + return IS_ERR(chan) ? NULL : chan; +} +EXPORT_SYMBOL_GPL(dmadev_get_slave_channel); + /** * __dma_request_channel - try to allocate an exclusive channel * @mask: capabilities that the channel must satisfy diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h index 16c9d021988a..32010c6ae13a 100644 --- a/include/linux/dmaengine.h +++ b/include/linux/dmaengine.h @@ -1542,6 +1542,8 @@ void dma_async_device_unregister(struct dma_device *device); void dma_run_dependencies(struct dma_async_tx_descriptor *tx); struct dma_chan *dma_get_slave_channel(struct dma_chan *chan); struct dma_chan *dma_get_any_slave_channel(struct dma_device *device); +struct dma_chan *dmadev_get_slave_channel(struct dma_device *device, + dma_filter_fn fn, void *fn_param); #define dma_request_channel(mask, x, y) __dma_request_channel(&(mask), x, y) #define dma_request_slave_channel_compat(mask, x, y, dev, name) \ __dma_request_slave_channel_compat(&(mask), x, y, dev, name)
dma_get_any_slave_channel() would skip using the filter function, which in some cases needed to be executed before the alloc_chan_resources callback to make sure that all parameters are provided for the slave channel. Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com> --- drivers/dma/dmaengine.c | 20 ++++++++++++++++++++ include/linux/dmaengine.h | 2 ++ 2 files changed, 22 insertions(+)