On 2024/9/9 17:49, Dmitry Baryshkov wrote: > On Mon, Sep 09, 2024 at 03:31:41PM GMT, Jinjie Ruan wrote: >> Use devm_pm_runtime_enable(), devm_request_irq() and >> devm_spi_register_controller() to simplify code. >> >> And also register a callback spi_geni_release_dma_chan() with >> devm_add_action_or_reset(), to release dma channel in both error >> and device detach path, which can make sure the release sequence is >> consistent with the original one. >> >> 1. Unregister spi controller. >> 2. Free the IRQ. >> 3. Free DMA chans >> 4. Disable runtime PM. >> >> So the remove function can also be removed. >> >> Suggested-by: Doug Anderson <dianders@xxxxxxxxxxxx> >> Signed-off-by: Jinjie Ruan <ruanjinjie@xxxxxxxxxx> >> --- >> v3: >> - Land the rest of the cleanups afterwards. >> --- >> drivers/spi/spi-geni-qcom.c | 37 +++++++++++++------------------------ >> 1 file changed, 13 insertions(+), 24 deletions(-) >> >> diff --git a/drivers/spi/spi-geni-qcom.c b/drivers/spi/spi-geni-qcom.c >> index 6f4057330444..8b0039d14605 100644 >> --- a/drivers/spi/spi-geni-qcom.c >> +++ b/drivers/spi/spi-geni-qcom.c >> @@ -632,8 +632,10 @@ static int spi_geni_grab_gpi_chan(struct spi_geni_master *mas) >> return ret; >> } >> >> -static void spi_geni_release_dma_chan(struct spi_geni_master *mas) >> +static void spi_geni_release_dma_chan(void *data) >> { >> + struct spi_geni_master *mas = data; >> + >> if (mas->rx) { >> dma_release_channel(mas->rx); >> mas->rx = NULL; >> @@ -1132,6 +1134,12 @@ static int spi_geni_probe(struct platform_device *pdev) >> if (ret) >> return ret; >> >> + ret = devm_add_action_or_reset(dev, spi_geni_release_dma_chan, spi); > > This should be mas, not spi. > > Doesn't looks like this was tested. Please correct me if I'm wrong. Yes, you are right, the data should be struct spi_geni_master, which is mas. Sorry, only compile passed. > >