[PATCH] dmaengine: shdma: Runtime-resume device in .shutdown()

[PATCH] dmaengine: shdma: Runtime-resume device in .shutdown()

[Geert Uytterhoeven]

During system reboot, the sh-dma-engine device may be runtime-suspended,
causing a crash:

    Unhandled fault: imprecise external abort (0x1406) at 0x0002c02c
    Internal error: : 1406 [#1] SMP ARM
    ...
    PC is at sh_dmae_ctl_stop+0x28/0x64
    LR is at sh_dmae_ctl_stop+0x24/0x64

If the sh-dma-engine is runtime-suspended, its module clock is turned
off, and its registers cannot be accessed. Runtime-resume the device in
the driver's .shutdown() callback to fix this.

Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
---
 drivers/dma/sh/shdmac.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/dma/sh/shdmac.c b/drivers/dma/sh/shdmac.c
index b65317c6ea4e722c..a13c6ba7468f12a6 100644
--- a/drivers/dma/sh/shdmac.c
+++ b/drivers/dma/sh/shdmac.c
@@ -585,7 +585,10 @@ static void sh_dmae_chan_remove(struct sh_dmae_device *shdev)
 static void sh_dmae_shutdown(struct platform_device *pdev)
 {
 	struct sh_dmae_device *shdev = platform_get_drvdata(pdev);
+
+	pm_runtime_get_sync(&pdev->dev);
 	sh_dmae_ctl_stop(shdev);
+	pm_runtime_put(&pdev->dev);
 }
 
 static int sh_dmae_runtime_suspend(struct device *dev)
-- 
1.9.1

Re: [PATCH] dmaengine: shdma: Runtime-resume device in .shutdown()

[Vinod Koul]

On Wed, Nov 19, 2014 at 04:15:46PM +0100, Geert Uytterhoeven wrote:
> During system reboot, the sh-dma-engine device may be runtime-suspended,
> causing a crash:
> 
>     Unhandled fault: imprecise external abort (0x1406) at 0x0002c02c
>     Internal error: : 1406 [#1] SMP ARM
>     ...
>     PC is at sh_dmae_ctl_stop+0x28/0x64
>     LR is at sh_dmae_ctl_stop+0x24/0x64
> 
> If the sh-dma-engine is runtime-suspended, its module clock is turned
> off, and its registers cannot be accessed. Runtime-resume the device in
> the driver's .shutdown() callback to fix this.
> 
> Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
> ---
>  drivers/dma/sh/shdmac.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/drivers/dma/sh/shdmac.c b/drivers/dma/sh/shdmac.c
> index b65317c6ea4e722c..a13c6ba7468f12a6 100644
> --- a/drivers/dma/sh/shdmac.c
> +++ b/drivers/dma/sh/shdmac.c
> @@ -585,7 +585,10 @@ static void sh_dmae_chan_remove(struct sh_dmae_device *shdev)
>  static void sh_dmae_shutdown(struct platform_device *pdev)
>  {
>  	struct sh_dmae_device *shdev = platform_get_drvdata(pdev);
> +
> +	pm_runtime_get_sync(&pdev->dev);
>  	sh_dmae_ctl_stop(shdev);
> +	pm_runtime_put(&pdev->dev);
but if you are runtime_suspended, then why should you even proceed to stop
the clock. Why not just cleanup and return when runtime suspended

-- 
~Vinod

Re: [PATCH] dmaengine: shdma: Runtime-resume device in .shutdown()

[Geert Uytterhoeven]

Hi Vinod,

On Fri, Dec 5, 2014 at 6:47 PM, Vinod Koul <vinod.koul@intel.com> wrote:
> On Wed, Nov 19, 2014 at 04:15:46PM +0100, Geert Uytterhoeven wrote:
>> During system reboot, the sh-dma-engine device may be runtime-suspended,
>> causing a crash:
>>
>>     Unhandled fault: imprecise external abort (0x1406) at 0x0002c02c
>>     Internal error: : 1406 [#1] SMP ARM
>>     ...
>>     PC is at sh_dmae_ctl_stop+0x28/0x64
>>     LR is at sh_dmae_ctl_stop+0x24/0x64
>>
>> If the sh-dma-engine is runtime-suspended, its module clock is turned
>> off, and its registers cannot be accessed. Runtime-resume the device in
>> the driver's .shutdown() callback to fix this.
>>
>> Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
>> ---
>>  drivers/dma/sh/shdmac.c | 3 +++
>>  1 file changed, 3 insertions(+)
>>
>> diff --git a/drivers/dma/sh/shdmac.c b/drivers/dma/sh/shdmac.c
>> index b65317c6ea4e722c..a13c6ba7468f12a6 100644
>> --- a/drivers/dma/sh/shdmac.c
>> +++ b/drivers/dma/sh/shdmac.c
>> @@ -585,7 +585,10 @@ static void sh_dmae_chan_remove(struct sh_dmae_device *shdev)
>>  static void sh_dmae_shutdown(struct platform_device *pdev)
>>  {
>>       struct sh_dmae_device *shdev = platform_get_drvdata(pdev);
>> +
>> +     pm_runtime_get_sync(&pdev->dev);
>>       sh_dmae_ctl_stop(shdev);
>> +     pm_runtime_put(&pdev->dev);
> but if you are runtime_suspended, then why should you even proceed to stop
> the clock. Why not just cleanup and return when runtime suspended

sh_dmae_ctl_stop() stops the DMA engine, not the clock.
Accessing the DMA engine cannot be done while the module clock is stopped.

If pm_runtime_suspended() returns true, I can indeed just return (no new
request can come in while .shutdown() is called).
However, if pm_runtime_suspended() returns false, the device may still
become runtime suspended in between the check for pm_runtime_suspended()
and accessing the DMA engine registers in sh_dmae_ctl_stop(), as runtime
suspend is an asynchronous operation, right?

So I think adding a check for pm_runtime_suspended() can only serve as
an optimization, the pm_runtime_{get_sync,put}() has to stay to prevent a
race condition.

Please correct me if I'm wrong.

Thanks!

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH] dmaengine: shdma: Runtime-resume device in .shutdown()

[Vinod Koul]

On Sun, Dec 07, 2014 at 11:35:21AM +0100, Geert Uytterhoeven wrote:
> >>  static void sh_dmae_shutdown(struct platform_device *pdev)
> >>  {
> >>       struct sh_dmae_device *shdev = platform_get_drvdata(pdev);
> >> +
> >> +     pm_runtime_get_sync(&pdev->dev);
> >>       sh_dmae_ctl_stop(shdev);
> >> +     pm_runtime_put(&pdev->dev);
> > but if you are runtime_suspended, then why should you even proceed to stop
> > the clock. Why not just cleanup and return when runtime suspended
> 
> sh_dmae_ctl_stop() stops the DMA engine, not the clock.
> Accessing the DMA engine cannot be done while the module clock is stopped.
> 
> If pm_runtime_suspended() returns true, I can indeed just return (no new
> request can come in while .shutdown() is called).
> However, if pm_runtime_suspended() returns false, the device may still
> become runtime suspended in between the check for pm_runtime_suspended()
> and accessing the DMA engine registers in sh_dmae_ctl_stop(), as runtime
> suspend is an asynchronous operation, right?
> 
> So I think adding a check for pm_runtime_suspended() can only serve as
> an optimization, the pm_runtime_{get_sync,put}() has to stay to prevent a
> race condition.
You are quite right in your observations, but this solution though correct
seems bit heavy for case like shutdown. can't we can use driver lock to
prevent race.

-- 
~Vinod