Received-SPF: pass (google.com: domain of linux-kernel+bounces-41114-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.199.223 as permitted sender) client-ip=147.75.199.223;
Date: Sat, 27 Jan 2024 09:31:14 +0100
From: Nam Cao <namcao@linutronix.de>
To: Samuel Holland <samuel.holland@sifive.com>
Cc: Thomas Gleixner <tglx@linutronix.de>, Palmer Dabbelt
 <palmer@dabbelt.com>, Paul Walmsley <paul.walmsley@sifive.com>, Samuel
 Holland <samuel@sholland.org>, Marc Zyngier <maz@kernel.org>,
 linux-kernel@vger.kernel.org, linux-riscv@lists.infradead.org, Guo Ren
 <guoren@kernel.org>, stable@vger.kernel.org
Subject: Re: [PATCH] irqchip/sifive-plic: enable interrupt if needed before
 EOI
Message-ID: <20240127093114.15fdfa5c@namcao>
In-Reply-To: <071142d9-3644-4c62-a99d-40b55bf64406@sifive.com>
References: <20240126223836.202321-1-namcao@linutronix.de>
	<071142d9-3644-4c62-a99d-40b55bf64406@sifive.com>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit

On Fri, 26 Jan 2024 18:31:19 -0600 Samuel Holland <samuel.holland@sifive.com> wrote:
> On 2024-01-26 4:38 PM, Nam Cao wrote:
> > RISC-V PLIC cannot EOI disabled interrupts, as explained in the
> > description of Interrupt Completion in the PLIC spec:
> > 
> > "The PLIC signals it has completed executing an interrupt handler by
> > writing the interrupt ID it received from the claim to the claim/complete
> > register. The PLIC does not check whether the completion ID is the same
> > as the last claim ID for that target. If the completion ID does not match
> > an interrupt source that *is currently enabled* for the target, the
> > completion is silently ignored."
> > 
> > Commit 69ea463021be ("irqchip/sifive-plic: Fixup EOI failed when masked")
> > ensured that by enabling the interrupt if needed before EOI.
> > 
> > Commit a1706a1c5062 ("irqchip/sifive-plic: Separate the enable and mask
> > operations") removed the interrupt enabling code from the previous
> > commit, because it assumes that interrupt should be enabled at the point
> > of EOI. However, this is incorrect: there is a small window after a hart
> > claiming an interrupt and before irq_desc->lock getting acquired,
> > interrupt can be disabled during this window. Thus, EOI can be invoked
> > while the interrupt is disabled, effectively nullify this EOI.
> > 
> > Make sure that interrupt is really enabled before EOI.  
> 
> Could you please try the patch I previously sent for this issue[1]?

Unfortunately my system still gets frozen with the patch applied :(

I think because the patch doesn't prevent plic_irq_shutdown() from getting
called after the hart claiming the interrupt and before irq_desc is locked
up.

> It should fix the bug without complicating the IRQ hot path.

I can add an unlikely() to help that a bit, because from my experience, it
is quite rare that EOI happens with interrupt disabled.

Best regards,
Nam