RE: [PATCH 1/2] scsi: ufs: core: Introduce a new clock_gating lock

[Avri Altman <Avri.Altman@wdc.com>]

> 
> On 10/27/24 1:25 AM, Avri Altman wrote:
> > Introduce a new clock gating lock to seriliaze access to the clock
>                                         ^^^^^^^^^
>                                         serialize
Done.

> 
> > diff --git a/drivers/ufs/core/ufshcd.c b/drivers/ufs/core/ufshcd.c
> > index 099373a25017..b7c7a7dd327f 100644
> > --- a/drivers/ufs/core/ufshcd.c
> > +++ b/drivers/ufs/core/ufshcd.c
> > @@ -1817,13 +1817,13 @@ static void ufshcd_ungate_work(struct
> > work_struct *work)
> >
> >       cancel_delayed_work_sync(&hba->clk_gating.gate_work);
> >
> > -     spin_lock_irqsave(hba->host->host_lock, flags);
> > +     spin_lock_irqsave(&hba->clk_gating.lock, flags);
> >       if (hba->clk_gating.state == CLKS_ON) {
> > -             spin_unlock_irqrestore(hba->host->host_lock, flags);
> > +             spin_unlock_irqrestore(&hba->clk_gating.lock, flags);
> >               return;
> >       }
> >
> > -     spin_unlock_irqrestore(hba->host->host_lock, flags);
> > +     spin_unlock_irqrestore(&hba->clk_gating.lock, flags);
> >       ufshcd_hba_vreg_set_hpm(hba);
> >       ufshcd_setup_clocks(hba, true);
> 
> This would be a great opportunity to replace the spinlock calls with
> scoped_guard(), isn't it?
Done.

> 
> > @@ -1928,7 +1928,7 @@ static void ufshcd_gate_work(struct work_struct
> *work)
> >       unsigned long flags;
> >       int ret;
> >
> > -     spin_lock_irqsave(hba->host->host_lock, flags);
> > +     spin_lock_irqsave(&hba->clk_gating.lock, flags);
> >       /*
> >        * In case you are here to cancel this work the gating state
> >        * would be marked as REQ_CLKS_ON. In this case save time by @@
> > -1946,7 +1946,7 @@ static void ufshcd_gate_work(struct work_struct
> *work)
> >       if (ufshcd_is_ufs_dev_busy(hba) || hba->ufshcd_state !=
> UFSHCD_STATE_OPERATIONAL)
> >               goto rel_lock;
> >
> > -     spin_unlock_irqrestore(hba->host->host_lock, flags);
> > +     spin_unlock_irqrestore(&hba->clk_gating.lock, flags);
> 
> Same comment here: please consider using scoped_guard().
Done.

> 
> >       /* put the link into hibern8 mode before turning off clocks */
> >       if (ufshcd_can_hibern8_during_gating(hba)) { @@ -1977,14
> > +1977,14 @@ static void ufshcd_gate_work(struct work_struct *work)
> >        * prevent from doing cancel work multiple times when there are
> >        * new requests arriving before the current cancel work is done.
> >        */
> > -     spin_lock_irqsave(hba->host->host_lock, flags);
> > +     spin_lock_irqsave(&hba->clk_gating.lock, flags);
> >       if (hba->clk_gating.state == REQ_CLKS_OFF) {
> >               hba->clk_gating.state = CLKS_OFF;
> >               trace_ufshcd_clk_gating(dev_name(hba->dev),
> >                                       hba->clk_gating.state);
> >       }
> >   rel_lock:
> > -     spin_unlock_irqrestore(hba->host->host_lock, flags);
> > +     spin_unlock_irqrestore(&hba->clk_gating.lock, flags);
> >   out:
> >       return;
> >   }
> 
> ufshcd_gate_work() can be simplified by using guard() and scoped_guard().
Done.

> 
> > @@ -2015,9 +2015,9 @@ void ufshcd_release(struct ufs_hba *hba)
> >   {
> >       unsigned long flags;
> >
> > -     spin_lock_irqsave(hba->host->host_lock, flags);
> > +     spin_lock_irqsave(&hba->clk_gating.lock, flags);
> >       __ufshcd_release(hba);
> > -     spin_unlock_irqrestore(hba->host->host_lock, flags);
> > +     spin_unlock_irqrestore(&hba->clk_gating.lock, flags);
> 
> For this function and also for later changes, please use guard().
Done.

> 
> > diff --git a/include/ufs/ufshcd.h b/include/ufs/ufshcd.h index
> > 9ea2a7411bb5..52c822fe2944 100644
> > --- a/include/ufs/ufshcd.h
> > +++ b/include/ufs/ufshcd.h
> > @@ -413,6 +413,7 @@ enum clk_gating_state {
> >    * @active_reqs: number of requests that are pending and should be
> waited for
> >    * completion before gating clocks.
> >    * @clk_gating_workq: workqueue for clock gating work.
> > + * @lock: serielize access to the clk_gating members
>               ^^^^^^^^^
>               serialize
> 
> I don't think that the added comment is correct - 'lock' is used to serialize
> access to some struct ufs_clk_gating members but not for serializing access
> to all members. Accesses to e.g. gate_work, ungate_work and
> clk_gating_workq are not serialized. Please reorder the struct ufs_clk_gating
> members as follows:
> - Members that are not serialized first.
> - Next, 'lock'.
> - Finally, the members serialized by 'lock'.
> 
> I think it is common in Linux kernel code that structure members are
> organized this way.
Done.

Thanks,
Avri

> 
> Thanks,
> 
> Bart.