linux-kernel - Re: [PATCH] i2c: pasemi: Add IRQ support for Apple Silicon

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>] [day] [month] [year] [list]

Message-ID: <8387D093-6220-4849-9E09-D856A4362E97@protonmail.com>
Date:   Sat, 20 Aug 2022 20:17:23 +0000
From:   Martin Povišer <povik@...tonmail.com>
To:     Arminder Singh <arminders208@...look.com>
Cc:     Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        Sven Peter <sven@...npeter.dev>,
        Alyssa Rosenzweig <alyssa@...enzweig.io>,
        Hector Martin <marcan@...can.st>,
        Michael Ellerman <mpe@...erman.id.au>,
        Benjamin Herrenschmidt <benh@...nel.crashing.org>,
        Paul Mackerras <paulus@...ba.org>,
        linux-arm-kernel@...ts.infradead.org,
        linuxppc-dev@...ts.ozlabs.org, linux-i2c@...r.kernel.org
Subject: Re: [PATCH] i2c: pasemi: Add IRQ support for Apple Silicon


> On 20. 8. 2022, at 21:45, Arminder Singh <arminders208@...look.com> wrote:
>
> This is the first time I'm interacting with the Linux mailing lists, so
> please don't eviscerate me *too much* if I get the formatting wrong.
> Of course I'm always willing to take criticism and improve my formatting
> in the future.
>
> This patch adds support for IRQs to the PASemi I2C controller driver.
> This will allow for faster performing I2C transactions on Apple Silicon
> hardware, as previously, the driver was forced to poll the SMSTA register
> for a set amount of time.
>
> With this patchset the driver on Apple silicon hardware will instead wait
> for an interrupt which will signal the completion of the I2C transaction.
> The timeout value for this completion will be the same as the current
> amount of time the I2C driver polls for.
>
> This will result in some performance improvement since the driver will be
> waiting for less time than it does right now on Apple Silicon hardware.
>
> The patch right now will only enable IRQs for Apple Silicon I2C chips,
> and only if it's able to successfully request the IRQ from the kernel.
>
> === Testing ===
>
> This patch has been tested on both the mainline Linux kernel tree and
> the Asahi branch (https://github.com/AsahiLinux/linux.git) on both an
> M1 and M2 MacBook Air, and it compiles successfully as both a module and
> built-in to the kernel itself. The patch in both trees successfully boots
> to userspace without any hitch.
>
> I do not have PASemi hardware on hand unfortunately, so I'm unable to test
> the impact of this patch on old PASemi hardware. This is also why I've
> elected to do the IRQ request and enablement on the Apple platform driver
> and not in the common file, as I'm not sure if PASemi hardware supports
> IRQs.
>
> I also fixed a quick checkpatch warning on line 303. "i ++" is now "i++".
>
> Any and all critiques of the patch would be well appreciated.
>
>
>
>
> Signed-off-by: Arminder Singh <arminders208@...look.com>

Thanks for the patch!

Tested-by: Martin Povišer <povik+lin@...ebit.org>

on Mac mini (2020) with M1, with in-kernel WIP sound drivers.

Some timing comparison on boot logs follows, three runs with
and without the patch.

With:

[    0.631034] ALSA device list:
[    0.631403]   #0: Mac mini J274 integrated audio

[    0.625559] ALSA device list:
[    0.625997]   #0: Mac mini J274 integrated audio

[    0.624561] ALSA device list:
[    0.624913]   #0: Mac mini J274 integrated audio

Without:

[    0.681599] ALSA device list:
[    0.682051]   #0: Mac mini J274 integrated audio

[    0.677538] ALSA device list:
[    0.677968]   #0: Mac mini J274 integrated audio

[    0.686037] ALSA device list:
[    0.686400]   #0: Mac mini J274 integrated audio

(After I collected these I realised I don’t know to what exactly
the timing of the print is anchored, but anyway it still suggests
there is speed-up.)

Best,
Martin