linux-kernel - Re: [PATCH] clocksource/drivers/timer-mediatek: optimize systimer irq clear flow on Mediatek Socs

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <1618999672.25707.13.camel@mhfsdcap03>
Date:   Wed, 21 Apr 2021 18:07:52 +0800
From:   Fengquan Chen <fengquan.chen@...iatek.com>
To:     Daniel Lezcano <daniel.lezcano@...aro.org>,
        Thomas Gleixner <tglx@...utronix.de>,
        Matthias Brugger <matthias.bgg@...il.com>,
        <dehui.sun@...iatek.com>
CC:     <linux-kernel@...r.kernel.org>,
        <linux-arm-kernel@...ts.infradead.org>,
        <linux-mediatek@...ts.infradead.org>
Subject: Re: [PATCH] clocksource/drivers/timer-mediatek: optimize systimer
 irq clear flow on Mediatek Socs

On Mon, 2021-03-22 at 11:10 +0100, Daniel Lezcano wrote:
> On 02/03/2021 08:28, Fengquan Chen wrote:
> > 1)ensure systimer is enabled before clear and disable interrupt, which only
> > for systimer in Mediatek Socs.
> >
> > 2)clear any pending timer-irq when shutdown to keep suspend flow clean,
> > when use systimer as tick-broadcast timer
> > 
> > Change-Id: Ia3eda83324af2fdaf5cbb3569a9bf020a11f8009
> 
> Remove the above.
> 
> Add a Fixes tag.

Thanks for review, has been fixed in V4:
https://patchwork.kernel.org/project/linux-mediatek/patch/1617960162-1988-2-git-send-email-Fengquan.Chen@mediatek.com/

> 
> > Signed-off-by: Fengquan Chen <fengquan.chen@...iatek.com>
> > ---
> >  drivers/clocksource/timer-mediatek.c | 4 ++++
> >  1 file changed, 4 insertions(+)
> > 
> > diff --git a/drivers/clocksource/timer-mediatek.c b/drivers/clocksource/timer-mediatek.c
> > index 9318edc..9f1f095dc 100644
> > --- a/drivers/clocksource/timer-mediatek.c
> > +++ b/drivers/clocksource/timer-mediatek.c
> > @@ -75,6 +75,7 @@
> >  static void mtk_syst_ack_irq(struct timer_of *to)
> >  {
> >  	/* Clear and disable interrupt */
> > +	writel(SYST_CON_EN, SYST_CON_REG(to));
> 
> 	SYST_CON_EN is set below, why do you have to do it before?
> 
> Is that a hw bug ?
> 
> It is confusing what the description of the SYST_CON_EN says:
> 
> /*
>  * SYST_CON_EN: Clock enable. Shall be set to
>  *   - Start timer countdown.
>  *   - Allow timeout ticks being updated.
>  *   - Allow changing interrupt functions.
> 
> What means "interrupt functions" ?
> 
> Does writing writel(SYST_CON_EN, SYST_CON_REG(to)) before
> SYST_CON_IRQ_CLR allows to clear the interrupt flag?
> 
> Can you explain how the timer works regarding this part?
> 
> It sounds to me a bit strange.
> 
>  *
>  * SYST_CON_IRQ_EN: Set to allow interrupt.
>  *
>  * SYST_CON_IRQ_CLR: Set to clear interrupt.
>  */
> 

Thanks for review, descriptions about
SYST_CON_IRQ_CLR/SYST_CON_EN/SYST_CON_IRQ_EN have been updated in v4.

And for systimer, there's a hw limitation that before clearing pending
irqs, we must enable timer first. so we added a SYST_CON_EN write before
SYST_CON_IRQ_CLR write to ensure irq clear successfully.

Also, we cannot only write SYST_CON_IRQ_CLR without SYST_CON_EN bit,
because EN bit is also timer clock enable bit. 

this is a hardward design limitation but not a bug.

> 
> >  	writel(SYST_CON_IRQ_CLR | SYST_CON_EN, SYST_CON_REG(to));
> >  }
> >  
> > @@ -111,6 +112,9 @@ static int mtk_syst_clkevt_next_event(unsigned long ticks,
> >  
> >  static int mtk_syst_clkevt_shutdown(struct clock_event_device *clkevt)
> >  {
> > +	/* Clear any irq */
> > +	mtk_syst_ack_irq(to_timer_of(clkevt));
> > +
> >  	/* Disable timer */
> >  	writel(0, SYST_CON_REG(to_timer_of(clkevt)));
> >  
> > 
> 
>