Em Sat, 24 Apr 2021 20:23:53 -0300
Ezequiel Garcia <ezequiel@xxxxxxxxxxxxx> escreveu:
> Hi Mauro,
>
> On Sat, 2021-04-24 at 08:44 +0200, Mauro Carvalho Chehab wrote:
> > Commit dd8088d5a896 ("PM: runtime: Add pm_runtime_resume_and_get to deal with usage counter")
> > added pm_runtime_resume_and_get() in order to automatically handle
> > dev->power.usage_count decrement on errors.
> >
> > Use the new API, in order to cleanup the error check logic.
> >
> > Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@xxxxxxxxxx>
> > ---
> > drivers/staging/media/hantro/hantro_drv.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/staging/media/hantro/hantro_drv.c b/drivers/staging/media/hantro/hantro_drv.c
> > index 595e82a82728..3147dcbebeb9 100644
> > --- a/drivers/staging/media/hantro/hantro_drv.c
> > +++ b/drivers/staging/media/hantro/hantro_drv.c
> > @@ -155,7 +155,7 @@ static void device_run(void *priv)
> > ret = clk_bulk_enable(ctx->dev->variant->num_clocks, ctx->dev->clocks);
> > if (ret)
> > goto err_cancel_job;
> > - ret = pm_runtime_get_sync(ctx->dev->dev);
> > + ret = pm_runtime_resume_and_get(ctx->dev->dev);
> > if (ret < 0)
> > goto err_cancel_job;
> >
>
> Seems this one needs a different fix: err_cancel_job
> will call hantro_job_finish which has a pm_runtime put.
Good point. Thanks for reviewing it!
It sounds that this is a place where the best seems
to keep using pm_runtime_get_sync(), but let's at least add a
comment explaining why it should be kept here. This should
help to avoid people to copy-and-paste the code on situations
where pm_runtime_resume_and_get() should be used instead.
See enclosed patch.
Thanks,
Mauro
[PATCH] media: hantro: document the usage of pm_runtime_get_sync()
Despite other *_get()/*_put() functions, where usage count is
incremented only if not errors, the pm_runtime_get_sync() has
a different behavior, incrementing the counter *even* on
errors.
That's an error prone behavior, as people often forget to
decrement the usage counter.
However, the hantro driver depends on this behavior, as it
will decrement the usage_count unconditionally at the m2m
job finish time, which makes sense.
So, intead of using the pm_runtime_resume_and_get() that
would decrement the counter on error, keep the current
API, but add a documentation explaining the rationale for
keep using pm_runtime_get_sync().
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@xxxxxxxxxx>
diff --git a/drivers/staging/media/hantro/hantro_drv.c b/drivers/staging/media/hantro/hantro_drv.c
index 595e82a82728..96f940c1c85c 100644
--- a/drivers/staging/media/hantro/hantro_drv.c
+++ b/drivers/staging/media/hantro/hantro_drv.c
@@ -155,6 +155,13 @@ static void device_run(void *priv)
ret = clk_bulk_enable(ctx->dev->variant->num_clocks, ctx->dev->clocks);
if (ret)
goto err_cancel_job;
+
+ /*
+ * The pm_runtime_get_sync() will increment dev->power.usage_count,
+ * even on errors. That's the expected behavior here, since the
+ * hantro_job_finish() function at the error handling code
+ * will internally call pm_runtime_put_autosuspend().
+ */
ret = pm_runtime_get_sync(ctx->dev->dev);
if (ret < 0)
goto err_cancel_job;
