On Tue, Dec 7, 2021 at 9:38 AM Baolin Wang <baolin.wang7@xxxxxxxxx> wrote:
>
> Hi Dongliang,
>
> On Mon, Dec 6, 2021 at 7:34 PM Dongliang Mu <mudongliangabcd@xxxxxxxxx> wrote:
> >
> > When pm_runtime_get_sync fails, it forgets to invoke pm_runtime_disable
> > in the label err_rpm.
> >
> > Fix this by moving pm_runtime_disable to label err_rpm.
> >
> > Fixes: 9b3b8171f7f4 ("dmaengine: sprd: Add Spreadtrum DMA driver")
> > Signed-off-by: Dongliang Mu <mudongliangabcd@xxxxxxxxx>
> > ---
>
> Thanks for your patch, but looking at the code in detail, I think we
> also should decrease the rpm counter when failing to call the
> pm_runtime_get_sync().
Thanks for your reply. There are many different pm_runtime_* API.
After I double check the usage of pm_runtime_get_sync, there are two
kinds of error handling code:
1. When pm_runtime_get_sync fails, call pm_runtime_put_sync and
pm_runtime_disable [1]
2. When pm_runtime_get_sync fails, only call pm_runtime_disable [2]
[1] https://elixir.bootlin.com/linux/latest/source/drivers/dma/ti/edma.c#L2402
[2] https://elixir.bootlin.com/linux/latest/source/drivers/dma/ti/cppi41.c#L1098
BTW, is there any standard error handling code of pm runtime API? Or
the majority wins?
>
> --- a/drivers/dma/sprd-dma.c
> +++ b/drivers/dma/sprd-dma.c
> @@ -1210,7 +1210,7 @@ static int sprd_dma_probe(struct platform_device *pdev)
> ret = dma_async_device_register(&sdev->dma_dev);
> if (ret < 0) {
> dev_err(&pdev->dev, "register dma device failed:%d\n", ret);
> - goto err_register;
> + goto err_rpm;
> }
>
> sprd_dma_info.dma_cap = sdev->dma_dev.cap_mask;
> @@ -1224,10 +1224,9 @@ static int sprd_dma_probe(struct platform_device *pdev)
>
> err_of_register:
> dma_async_device_unregister(&sdev->dma_dev);
> -err_register:
> +err_rpm:
> pm_runtime_put_noidle(&pdev->dev);
> pm_runtime_disable(&pdev->dev);
> -err_rpm:
> sprd_dma_disable(sdev);
> return ret;
> }
>
> --
> Baolin Wang
