Received-SPF: pass (google.com: domain of linux-kernel+bounces-206447-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.199.223 as permitted sender) client-ip=147.75.199.223;
Precedence: bulk
MIME-Version: 1.0
References: <20240607074716.4068975-1-ruanjinjie@huawei.com>
In-Reply-To: <20240607074716.4068975-1-ruanjinjie@huawei.com>
From: Doug Anderson <dianders@chromium.org>
Date: Fri, 7 Jun 2024 09:02:47 -0700
Message-ID: <CAD=FV=W6Y+sUcOZ02DrsNDAkd2eOJ8VAC_ke04gOO6s5OgMvzw@mail.gmail.com>
Subject: Re: [PATCH] arm64: smp: Fix missing IPI statistics
To: Jinjie Ruan <ruanjinjie@huawei.com>
Cc: catalin.marinas@arm.com, will@kernel.org, mark.rutland@arm.com, 
	swboyd@chromium.org, sumit.garg@linaro.org, frederic@kernel.org, 
	scott@os.amperecomputing.com, misono.tomohiro@fujitsu.com, 
	linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

Hi,

On Fri, Jun 7, 2024 at 12:45=E2=80=AFAM Jinjie Ruan <ruanjinjie@huawei.com>=
 wrote:
>
> commit 83cfac95c018 ("genirq: Allow interrupts to be excluded from
> /proc/interrupts") is to avoid IPIs appear twice in /proc/interrupts.
> But the commit 331a1b3a836c ("arm64: smp: Add arch support for backtrace
> using pseudo-NMI") and commit 2f5cd0c7ffde("arm64: kgdb: Implement
> kgdb_roundup_cpus() to enable pseudo-NMI roundup") set CPU_BACKTRACE and
> KGDB_ROUNDUP IPIs "IRQ_HIDDEN" flag but not show them in
> arch_show_interrupts(), which cause the interrupt kstat_irqs accounting
> is missing in display.
>
> Fixes: 331a1b3a836c ("arm64: smp: Add arch support for backtrace using ps=
eudo-NMI")
> Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
> ---
>  arch/arm64/kernel/smp.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)

While I won't object to your patch if everyone agrees that we want it,
fully excluding "cpu backtrace" and "kgdb roundup" from
/proc/interrupts was more of a design decision than a bug. Those two
IPIs are really special cases and not something that I'd expect anyone
to care about knowing the count of. Keeping them out of
"/proc/interrupts" just avoids noise. I'd also note that I believe
arm32 makes the same design choice for "cpu backtrace".

In any case, if we truly think people want the count of these IPIs
then it feels like we should report them in arch_show_interrupts()
where we can give them a nice string.

-Doug