Re: [PATCH v7 3/3] soc: qcom: rpmh: Invoke rpmh_flush for dirty caches

[Maulik Shah]

On 2/27/2020 4:19 AM, Stephen Boyd wrote:
> Quoting Maulik Shah (2020-02-25 21:27:13)
> > Add changes to invoke rpmh flush when the data in cache is dirty.
> >
> > This is done only if OSI is not supported in PSCI. If OSI is supported
> > rpmh_flush can get invoked when the last cpu going to power collapse
> Please write rpmh_flush() so we know it's a function and not a variable.
Done. Will update in v8.
> > deepest low power mode.
> >
> > Signed-off-by: Maulik Shah <mkshah@xxxxxxxxxxxxxx>
> > Reviewed-by: Srinivas Rao L <lsrao@xxxxxxxxxxxxxx>
> > ---
> >   drivers/soc/qcom/rpmh.c | 6 ++++++
> >   1 file changed, 6 insertions(+)
> >
> > diff --git a/drivers/soc/qcom/rpmh.c b/drivers/soc/qcom/rpmh.c
> > index 83ba4e0..839af8d 100644
> > --- a/drivers/soc/qcom/rpmh.c
> > +++ b/drivers/soc/qcom/rpmh.c
> > @@ -12,6 +12,7 @@
> >   #include <linux/module.h>
> >   #include <linux/of.h>
> >   #include <linux/platform_device.h>
> > +#include <linux/psci.h>
> >   #include <linux/slab.h>
> >   #include <linux/spinlock.h>
> >   #include <linux/types.h>
> > @@ -163,6 +164,9 @@ static struct cache_req *cache_rpm_request(struct rpmh_ctrlr *ctrlr,
> >   unlock:
> >          spin_unlock_irqrestore(&ctrlr->cache_lock, flags);
> >   
> > +       if (ctrlr->dirty && !psci_has_osi_support())
> Can we introduce a stub function for psci_has_osi_support() when
> CONFIG_ARM_PSCI_FW=n? This driver currently has:
>
>    config QCOM_RPMH
>          bool "Qualcomm RPM-Hardened (RPMH) Communication"
> 	depends on ARCH_QCOM && ARM64 || COMPILE_TEST
>
>
> which implies that this will break build testing once built on something
> that isn't arm64.
Thanks for pointing this, i think its better to remove COMPILE_TEST so 
driver only

gets build for arm64.

> > +               return rpmh_flush(ctrlr) ? ERR_PTR(-EINVAL) : req;
> > +
> >          return req;
> >   }
> >   
> > @@ -391,6 +395,8 @@ int rpmh_write_batch(const struct device *dev, enum rpmh_state state,
> >   
> >          if (state != RPMH_ACTIVE_ONLY_STATE) {
> >                  cache_batch(ctrlr, req);
> > +               if (!psci_has_osi_support())
> > +                       return rpmh_flush(ctrlr);
> While the diff is small it is also sad that we turn around after adding
> it to a list and immediately take it off the list and send it. Can't we
> do this without having to do the list add/remove dance?

No, we need to keep it in list, the target supporting OSI will get it 
off list only when last cpu

enters deepest idle mode.

> >                  return 0;
> >          }
--
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation