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

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]

Date:   Tue, 19 Sep 2023 17:49:21 +0100
From:   Marc Zyngier <maz@...nel.org>
To:     Biju Das <biju.das.jz@...renesas.com>
Cc:     Thomas Gleixner <tglx@...utronix.de>,
        Prabhakar Mahadev Lad <prabhakar.mahadev-lad.rj@...renesas.com>,
        Claudiu Beznea <claudiu.beznea.uj@...renesas.com>,
        Geert Uytterhoeven <geert+renesas@...der.be>,
        Biju Das <biju.das.au@...il.com>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "linux-renesas-soc@...r.kernel.org" 
        <linux-renesas-soc@...r.kernel.org>
Subject: Re: [PATCH 3/3] irqchip: renesas-rzg2l: Fix irq storm with edge trigger detection for TINT

On Tue, 19 Sep 2023 17:32:05 +0100,
Biju Das <biju.das.jz@...renesas.com> wrote:

[...]

> > So you mean that you *already* lose interrupts across a disable followed by
> > an enable? I'm slightly puzzled...
> 
> There is no interrupt lost at all. 
> 
> Currently this patch addresses 2 issues.
> 
> Scenario 1: Extra interrupt when we select TINT source on enable_irq()
> 
> Getting an extra interrupt, when client drivers calls enable_irq()
> during probe()/resume(). In this case, the irq handler on the Client
> driver just clear the interrupt status bit.
>
> Issue 2: IRQ storm when we select TINT source on enable_irq()
> 
> Here as well, we are getting an extra interrupt, when client drivers
> calls enable_irq() during probe() and this Interrupts getting
> generated infinitely, when the client driver calls disable_irq() in
> irq handler and in in work queue calling enable_irq().

How do you know this is a spurious interrupt? For all you can tell,
you are just consuming an edge. I absolutely don't buy this
workaround, because you have no context that allows you to
discriminate between a real spurious interrupt and a normal interrupt
that lands while the interrupt line was masked.

> Currently we are not loosing interrupts, but we are getting additional
> Interrupt(phantom) which is causing the issue.

If you get an interrupt at probe time in the endpoint driver, that's
probably because the device is not in a quiescent state when the
interrupt is requested. And it is probably this that needs addressing.

	M.

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