linux-kernel - [PATCH 0/6] genirq: Post 6.19-rc1 API cleanup

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-next>] [day] [month] [year] [list]

Message-ID: <20251210082242.360936-1-maz@kernel.org>
Date: Wed, 10 Dec 2025 08:22:36 +0000
From: Marc Zyngier <maz@...nel.org>
To: linux-kernel@...r.kernel.org,
	linux-mips@...r.kernel.org
Cc: Daniel Lezcano <daniel.lezcano@...aro.org>,
	Thomas Gleixner <tglx@...utronix.de>,
	Thomas Bogendoerfer <tsbogend@...ha.franken.de>,
	Joshua Kinard <kumba@...too.org>
Subject: [PATCH 0/6] genirq: Post 6.19-rc1 API cleanup

Fixing the issue reported at [1] pushed me to look into some of the
ugly corners of the percpu IRQ configuration, and to notice that:

- we only have 3 users left of setup_percpu_irq(), all MIPS timers

- none of the timer drivers using request_percpu_irq() are passing the
  IRQF_TIMER flag (this function doesn't take a flag argument)

- the IRQ_TIMINGS subsystem therefore cannot identify that it should
  consider these interrupts as timer interrupts, and would mistakenly
  take them into account

- but as it turns out, nobody can enable IRQ_TIMINGS the first place

- so passing flags at percpu IRQ request time is not required

The cumulative effects of the above is that we can:

- drop the IRQ_TIMINGS stuff -- it was never used, and Daniel
  indicates[2] that there is no immediate path to finishing it

- replace __request_percpu_irq() with request_percpu_irq_affinity()

- convert all three MIPS timers to request_percpu_irq()

- finally kill setup_percpu_irq() for good

Note that when it comes to MIPS, I have only tested the GIC timer
change, as I don't have any SGI HW anymore. And even that was on
QEMU. I'd appreciate someone with the necessary zoo to give it a go on
actual metal.

Patches on top of [3].

[1] https://lore.kernel.org/r/aTFozefMQRg7lYxh@aspen.lan
[2] https://lore.kernel.org/r/3f3d42d2-13e3-43ef-b839-f77424769add@linaro.org
[3] https://patch.msgid.link/20251205091814.3944205-1-maz@kernel.org

Marc Zyngier (6):
  genirq: Remove IRQ timing tracking infrastructure
  genirq: Remove __request_percpu_irq() helper
  mips: Move IP30 timer to request_percpu_irq()
  mips: Move IP27 timer to request_percpu_irq()
  clocksource: mips-gic-timer: Move GIC timer to request_percpu_irq()
  genirq: Kill setup_percpu_irq()

 arch/mips/include/asm/cevt-r4k.h     |   1 -
 arch/mips/kernel/cevt-r4k.c          |  11 -
 arch/mips/sgi-ip27/ip27-timer.c      |  10 +-
 arch/mips/sgi-ip30/ip30-timer.c      |   5 +-
 drivers/clocksource/mips-gic-timer.c |  10 +-
 include/linux/interrupt.h            |  24 +-
 include/linux/irq.h                  |   3 -
 kernel/irq/Kconfig                   |   3 -
 kernel/irq/Makefile                  |   4 -
 kernel/irq/handle.c                  |   2 -
 kernel/irq/internals.h               | 110 ---
 kernel/irq/manage.c                  |  49 +-
 kernel/irq/timings.c                 | 959 ---------------------------
 lib/Kconfig.debug                    |   8 -
 14 files changed, 16 insertions(+), 1183 deletions(-)
 delete mode 100644 kernel/irq/timings.c

-- 
2.47.3