Re: [PATCH 3/3] irqchip: renesas-rzg2l: Fix irq storm with edge trigger detection for TINT

From: Marc Zyngier <maz@kernel.org>
To: Biju Das <biju.das.jz@bp.renesas.com>
Cc: Thomas Gleixner <tglx@linutronix.de>,
	Prabhakar Mahadev Lad <prabhakar.mahadev-lad.rj@bp.renesas.com>,
	Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>,
	Geert Uytterhoeven <geert+renesas@glider.be>,
	Biju Das <biju.das.au@gmail.com>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"linux-renesas-soc@vger.kernel.org" 
	<linux-renesas-soc@vger.kernel.org>
Subject: Re: [PATCH 3/3] irqchip: renesas-rzg2l: Fix irq storm with edge trigger detection for TINT
Date: Tue, 19 Sep 2023 17:19:05 +0100	[thread overview]
Message-ID: <87cyye3zly.wl-maz@kernel.org> (raw)
In-Reply-To: <OS0PR01MB5922748F489467BE2539AA1886FAA@OS0PR01MB5922.jpnprd01.prod.outlook.com>

On Tue, 19 Sep 2023 16:24:53 +0100,
Biju Das <biju.das.jz@bp.renesas.com> wrote:
> 
> Hi Marc Zyngier,
> 
> Thanks for the feedback.
> 
> > Subject: Re: [PATCH 3/3] irqchip: renesas-rzg2l: Fix irq storm with edge
> > trigger detection for TINT
> > 
> > On Mon, 18 Sep 2023 13:24:11 +0100,
> > Biju Das <biju.das.jz@bp.renesas.com> wrote:
> > >
> > > In case of edge trigger detection, enabling the TINT source causes a
> > > phantum interrupt that leads to irq storm. So clear the phantum
> > > interrupt in rzg2l_irqc_irq_enable().
> > >
> > > This issue is observed when the irq handler disables the interrupts
> > > using
> > > disable_irq_nosync() and scheduling a work queue and in the work
> > > queue, re-enabling the interrupt with enable_irq().
> > >
> > > Fixes: 3fed09559cd8 ("irqchip: Add RZ/G2L IA55 Interrupt Controller
> > > driver")
> > > Signed-off-by: Biju Das <biju.das.jz@bp.renesas.com>
> > > Tested-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
> > > ---
> > >  drivers/irqchip/irq-renesas-rzg2l.c | 6 ++++++
> > >  1 file changed, 6 insertions(+)
> > >
> > > diff --git a/drivers/irqchip/irq-renesas-rzg2l.c
> > > b/drivers/irqchip/irq-renesas-rzg2l.c
> > > index 33a22bafedcd..78a9e90512a6 100644
> > > --- a/drivers/irqchip/irq-renesas-rzg2l.c
> > > +++ b/drivers/irqchip/irq-renesas-rzg2l.c
> > > @@ -144,6 +144,12 @@ static void rzg2l_irqc_irq_enable(struct irq_data
> > *d)
> > >  		reg = readl_relaxed(priv->base + TSSR(tssr_index));
> > >  		reg |= (TIEN | tint) << TSSEL_SHIFT(tssr_offset);
> > >  		writel_relaxed(reg, priv->base + TSSR(tssr_index));
> > > +		/*
> > > +		 * In case of edge trigger detection, enabling the TINT source
> > > +		 * cause a phantum interrupt that leads to irq storm. So clear
> > > +		 * the phantum interrupt.
> > > +		 */
> > > +		rzg2l_tint_eoi(d);
> > 
> > This looks incredibly unsafe. disable_irq()+enable_irq() with an interrupt
> > being made pending in the middle, and you've lost that interrupt.
> 
> In this driver that will never happen as it clears the TINT source
> during disable(), so there won't be any TINT source for interrupt
> detection after disable().

So you mean that you *already* lose interrupts across a disable
followed by an enable? I'm slightly puzzled...

	M.

-- 
Without deviation from the norm, progress is not possible.