Re: [PATCH v3 06/10] drivers: qcom: rpmh-rsc: allow invalidation of sleep/wake TCS

[Lina Iyer]

On Thu, Mar 08 2018 at 13:40 -0700, Stephen Boyd wrote:
> Quoting Lina Iyer (2018-03-02 08:43:13)
> > diff --git a/drivers/soc/qcom/rpmh-rsc.c b/drivers/soc/qcom/rpmh-rsc.c
> > index 34e780d9670f..e9f5a1f387fd 100644
> > --- a/drivers/soc/qcom/rpmh-rsc.c
> > +++ b/drivers/soc/qcom/rpmh-rsc.c
> > @@ -170,6 +170,52 @@ static struct tcs_group *get_tcs_of_type(struct rsc_drv *drv, int type)
> >         return tcs;
> >  }
> >
> > +/**
> > + * rpmh_rsc_invalidate - Invalidate sleep and wake TCSes
> > + *
> > + * @drv: the mailbox controller
> > + */
> > +int rpmh_rsc_invalidate(struct rsc_drv *drv)
> > +{
> > +       struct tcs_group *tcs;
> > +       int m, type, ret = 0;
> > +       int inv_types[] = { WAKE_TCS, SLEEP_TCS };
> > +       unsigned long drv_flags, flags;
> > +
> > +       /* Lock the DRV and clear sleep and wake TCSes */
> > +       spin_lock_irqsave(&drv->drv_lock, drv_flags);
> > +       for (type = 0; type < ARRAY_SIZE(inv_types); type++) {
> > +               tcs = get_tcs_of_type(drv, inv_types[type]);
> > +               if (IS_ERR(tcs))
> > +                       continue;
> > +
> > +               spin_lock_irqsave(&tcs->tcs_lock, flags);
>
> Should just be a spin_lock() because we already irq saved a few lines above.
Agreed.
> > +               if (bitmap_empty(tcs->slots, MAX_TCS_SLOTS)) {
> > +                       spin_unlock_irqrestore(&tcs->tcs_lock, flags);
> > +                       continue;
> > +               }
> > +
> > +               /* Clear the enable register for each TCS of the type */
> > +               for (m = tcs->tcs_offset;
> > +                   m < tcs->tcs_offset + tcs->num_tcs; m++) {
> > +                       if (!tcs_is_free(drv, m)) {
> > +                               spin_unlock_irqrestore(&tcs->tcs_lock, flags);
> > +                               ret = -EAGAIN;
> > +                               goto drv_unlock;
> > +                       }
> > +                       write_tcs_reg_sync(drv, RSC_DRV_CMD_ENABLE, m, 0, 0);
> > +                       /* Mark the TCS slots as free */
> > +                       bitmap_zero(tcs->slots, MAX_TCS_SLOTS);
> > +               }
> > +               spin_unlock_irqrestore(&tcs->tcs_lock, flags);
>
> Maybe make another function called rpmh_tcs_invalidate(drv, enum TCS_FOO)
> and put this logic inside it? And then change from a for loop to two
> direct calls:
>
> 	lock()
> 	ret = rpmh_tcs_invalidate(drv, WAKE_TCS);
> 	if (!ret)
> 		ret = rpmh_tcs_invalidate(drv, SLEEP_TCS);
> 	unlock()
Agreed. That will be better. Will fix.

-- Lina