Re: [PATCH 00/16] irqchip: harden initialization error paths

From: Thomas Gleixner <tglx@kernel.org>
To: Haofeng Li <lihaofeng@kylinos.cn>
Cc: linux-kernel@vger.kernel.org, Haofeng Li <13266079573@163.com>,
	Haofeng Li <lihaofeng@kylinos.cn>, Paul Walmsley <pjw@kernel.org>,
	Palmer Dabbelt <palmer@dabbelt.com>,
	Albert Ou <aou@eecs.berkeley.edu>,
	Alexandre Ghiti <alex@ghiti.fr>,
	Samuel Holland <samuel.holland@sifive.com>,
	linux-riscv@lists.infradead.org, Marc Zyngier <maz@kernel.org>,
	Radu Rendec <radu@rendec.net>
Subject: Re: [PATCH 00/16] irqchip: harden initialization error paths
Date: Thu, 20 Aug 2026 09:49:27 +0200	[thread overview]
Message-ID: <87y0e1i648.ffs@fw13> (raw)
In-Reply-To: <20260714122351.3274006-1-lihaofeng@kylinos.cn>

On Tue, Jul 14 2026 at 20:23, Haofeng Li wrote:
> An audit of irqchip initialization and allocation error paths found a
> collection of resource leaks and lifetime issues across several drivers.

A lot of these "fixes" are purely cosmetic and create a false sense of
correctness because if the initialization of the root interrupt
controller of a system fails then machine won't boot at all.

So instead of adding tons of cleanups we rather go and analyze which
controllers are actually root controllers and therefore essential for
the machine to boot. For those the only valid error handling is:

__probe(....)
{
        all setup magic
}

probe(....)
{
	if (__probe(..._))
	    panic("Failed to initialize root interrupt controller");
}

That is the proper hardening and allows to remove a boatload of
pointless cleanups.

Where cleanups actually matter are for secondary interrupt controllers
which are not essential for the machine to get up and "running",
especially those which can be built as modules.

Thanks,

        tglx

_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv