Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <YWCPyK+xotTgUMy/@arm.com>
In-Reply-To: <YWCPyK+xotTgUMy/@arm.com>
From:   Linus Torvalds <torvalds@linux-foundation.org>
Date:   Fri, 8 Oct 2021 13:25:51 -0700
Message-ID: <CAHk-=whWZ4OxfKQwKVrRc-E9=w-ygKdVFn_HcAMW-DW8SgranQ@mail.gmail.com>
Subject: Re: [GIT PULL] arm64 fixes for 5.15-rc5
To:     Catalin Marinas <catalin.marinas@arm.com>,
        Thomas Gleixner <tglx@linutronix.de>
Cc:     Will Deacon <will@kernel.org>, Marc Zyngier <maz@kernel.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        Linux ARM <linux-arm-kernel@lists.infradead.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Fri, Oct 8, 2021 at 11:37 AM Catalin Marinas <catalin.marinas@arm.com> wrote:
>
> Pingfan Liu (2):
>       kernel/irq: make irq_{enter,exit}() in handle_domain_irq() arch optional
>       arm64: entry: avoid double-accounting IRQ RCU entry

Ugh. This is *really* ugly. And it seems to be going exactly the wrong way.

I read the commit descriptions, and it still doesn't answer the
fundamental question of why arm64 needs to do the accounting in
arch-specific code, and disable the generic code.

It says

    To fix this, we must perform all the accounting from the architecture
    code. We prevent the IRQ domain code from performing any accounting by
    selecting HAVE_ARCH_IRQENTRY, and must call irq_enter_rcu() and
    irq_exit_rcu() around invoking the root IRQ handler.

but at no point does it actually explain *why* all the accounting
needs to be done by the architecture code.

Yes, yes, I read the previous paragraph. But why isn't the fix to just
stop doing the double accounting in the arm64 specific code?

Instead it doubles down on that "let's do this non-arch-specific
accounting in arch-specific code", making the common code uglier and
weaker.

I initially pulled this, and then I just unpulled in disgust.

Please explain why arm64 does this bad thing, and why the fix isn't
"fix arm64", but instead "make the generic code uglier and harder to
maintain and follow".

Really, from all the explanations those commits give, the natural
thing to do would be "just fix arm64".

So if that really isn't the answer, then the explanations are clearly lacking.

                   Linus