Re: [PATCH] PCI: dwc: Chain the set IRQ affinity request back to the parent

[Thomas Gleixner <tglx@linutronix.de>]

On Tue, Mar 11 2025 at 17:52, Tsai Sung-Fu wrote:

Please do not top-post and trim your replies.

> Running some basic tests with this patch (
> https://tglx.de/~tglx/patches.tar ) applied on my device, at first
> glance, the affinity feature is working.
>
> I didn't run stress test to test the stability, and the Kernel version
> we used is a bit old, so I only applied change in this 2 patches

I don't care about old kernels and what you can apply or not. Kernel
development happens against upstream and not against randomly chosen
private kernel versions.

> And adding if check on irq_chip_redirect_set_affinity() and
> irq_set_redirect_target() to avoid cpumask_first() return nr_cpu_ids

I assume you know how diff works.

> May I ask, would this patch be officially added to the 6.14 kernel ?

You may ask. But you should know the answer already, no?

The merge window for 6.14 closed on February 2nd with the release of
6.14-rc1. Anything which goes into Linus tree between rc1 and the final
release is fixes only.

This is new infrastructure, which has neither been posted nor reviewed
nor properly tested. There are also no numbers about the overhead and
no analysis whether that overhead causes regressions on existing setups.

These changes want to be:

   1) Put into a series with proper change logs

   2) Posted on the relevant mailing list

   3) Tested and proper numbers provided

So they are not even close to be ready for the 6.15 merge window, simply
because the irq tree is going to freeze at 6.14-rc7, i.e. by the end of
this week.

I'm not planning to work on them. Feel free to take the PoC patches,
polish them up and post them according to the documented process.

Thanks,

        tglx