Messages in this thread | | | Date | Mon, 21 Jan 2013 18:14:01 +0200 | From | Mika Westerberg <> | Subject | Re: [PATCH v2 06/10] spi/pxa2xx: add support for DMA engine |
| |
On Mon, Jan 21, 2013 at 03:44:36PM +0000, Russell King - ARM Linux wrote: > On Mon, Jan 21, 2013 at 11:31:59AM +0200, Mika Westerberg wrote: > > +bool dma_is_possible(size_t len) > > +int map_dma_buffers(struct driver_data *drv_data) > > +irqreturn_t dma_transfer(struct driver_data *drv_data) > > +int dma_prepare(struct driver_data *drv_data, u32 dma_burst) > > +void dma_start(struct driver_data *drv_data) > > +int dma_setup(struct driver_data *drv_data) > > +void dma_release(struct driver_data *drv_data) > > +void dma_resume(struct driver_data *drv_data) > > +int set_dma_burst_and_threshold(struct chip_data *chip, > > + struct spi_device *spi, > > + u8 bits_per_word, u32 *burst_code, > > + u32 *threshold) > > All the above function names end up in the global namespace in the kernel > image. They're rather _too_ generic and non-specific for them to leak to > that visibility. I think this needs fixing.
Indeed, good point. I will fix this.
> Also, I'd suggest that the long term plan is for PXA to move over to the > DMA engine API, so I'd recommend that the pxa2xx SPI driver have the > DMA engine API usage built into it, and let's treat the old PXA DMA > stuff as legacy code.
In the previous version I had the old PXA DMA stuff behind #ifdef CONFIG_ARCH_PXA and otherwise it used the DMA engine stuff. Do you mean something like that (e.g keep both DMA in the single file)?
| |