Hi, Am 09.05.19 um 20:08 schrieb kernel@xxxxxxxxxxxxxxxx: >> On 09.05.2019, at 19:03, Lukas Wunner <lukas@xxxxxxxxx> wrote: >> >> Without this, the driver for the BCM2835 SPI controller uses interrupt >> mode instead of DMA mode, incurring a significant performance penalty. >> The Foundation's device tree has had these attributes for years, but for >> some reason they were never upstreamed. >> >> They were originally contributed by Noralf Trønnes and Martin Sperl: >> https://github.com/raspberrypi/linux/commit/25f3e064afc8 >> https://github.com/raspberrypi/linux/commit/e0edb52b47e6 > I believe the the original patch that I had provided has been turned down > was because: > * we do not want to “consume” dma resources automatically as long as we don't enable SPI per default, this shouldn't be a problem. I remember that the downstream MMC driver "wasted" a DMA channel at the beginning, so we were a little bit careful in the past. > * to use spi correctly you have to apply several changes to the dt already > (think GPIO mode for CS and status=enable). This is something we can't provide, since the GPIO header should be flexible as possible. But this has nothing to do with Lukas patch, because i consider the patch as SoC related and common practice. Yes, overlays are a big improvement to define a SPI bus here. Thanks Stefan > > If this mindset has changed, then I am happy. > >> Tested-by: Nicolas Saenz Julienne <nsaenzjulienne@xxxxxxx> >> Signed-off-by: Lukas Wunner <lukas@xxxxxxxxx> >> Cc: Martin Sperl <kernel@xxxxxxxxxxxxxxxx> >> Cc: Noralf Trønnes <noralf@xxxxxxxxxxx> > Reviewed-by: Martin Sperl <kernel@xxxxxxxxxxxxxxxx> > > > > > > >
