The Renesas BSP has a patch to wait for the data register being empty before starting TX-DMA. This is according to the data sheet, so this small series does the same in patch 2 & 3. However, refactored, so there is less to compute and it is hopefully also easy to read. During that work, I noticed an implicit assumption about a required minimal DMA length. It is now explicit and documented :) Tested on a Renesas Salvator XS (R-Car M3N) and now regressions found when writing data with DMA. I am unsure if patches 2+3 should be squashed or not, but at least for reviewing, seperate patches is easier. Glad to hear what you think, Wolfram Wolfram Sang (3): i2c: rcar: sanity check for minimal DMA length i2c: rcar: let DMA enable routine return success status i2c: rcar: wait for data empty before starting DMA drivers/i2c/busses/i2c-rcar.c | 30 ++++++++++++++++-------------- 1 file changed, 16 insertions(+), 14 deletions(-) -- 2.11.0
