On 02/06/2017 08:38 PM, Fabio Estevam wrote: > Hi Marek, Hi, > On Mon, Feb 6, 2017 at 5:05 PM, Marek Vasut <marex@xxxxxxx> wrote: > >> Well yes, DMA is broken in 4.9 , that's why it's CC stable 4.9+ :) > > This patch is based on the fact that the UART on the imx6sx is not of > a imx6q UART type, which is not correct. > > Current imx6sx.dtsi misses to add a "fsl,imx6q-uart" compatible > string, just like all other imx6 SoC dtsi. Well given that the DMA does not work on 6SX, I guess it's not really compatible with 6Q. > When we fix the compatible uart problem on imx6sx.dtsi, then DMA will > be enabled and things will break again. > > That's why we need a better fix for this issue: > > 1) Let's see if Clemens' suggestion of using the latest SDMA firmware > can make DMA robust on your usecase. a) This will break if you don't use the latest SDMA firmware. This can be repaired by checking the SDMA firmware version and selectively disabling the DMA for broken versions. b) The DMA should not have been enabled on 4.9 , but it incorrectly is. Enabling it on 4.9 is a feature addition, which shouldn't go into stable kernel. > 2) If the system fails even with the latest SDMA, then the board dts > can disable dma by passing empty dma nodes. This might actually be the thing to do for 4.9 if you want to add the mx6q compatible to mx6sx uart . > Not all mx6 system seem to suffer from such SDMA problem. On a > imx6sx-udoo-neo we could use UART with DMA to connect to the Bluetooth > chip. Clearly some do, so it's probably also a matter of the link partner behavior. -- Best regards, Marek Vasut -- To unsubscribe from this list: send the line "unsubscribe stable" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html