Re: [PATCH -next v6 8/8] arm64: entry: Switch to generic IRQ entry

[Mark Rutland <mark.rutland@xxxxxxx>]

On Thu, Feb 27, 2025 at 07:08:56PM +0100, Valentin Schneider wrote:
> On 13/02/25 21:00, Jinjie Ruan wrote:
> > Currently, x86, Riscv, Loongarch use the generic entry. Convert arm64
> > to use the generic entry infrastructure from kernel/entry/*.
> > The generic entry makes maintainers' work easier and codes
> > more elegant.
> >
> > Switch arm64 to generic IRQ entry first, which removed duplicate 100+
> > LOC and make Lazy preemption on arm64 available by adding a
> > _TIF_NEED_RESCHED_LAZY bit and enabling ARCH_HAS_PREEMPT_LAZY.
> 
> Just a drive-by comment as I'm interested in lazy preemption for arm64;
> this series doesn't actually enable lazy preemption, is that for a
> follow-up with the rest of the generic entry stuff?
> 
> From a quick glance, it looks like everything is in place for enabling it.

Sorry, there's been some fractured discussion on this on the
linux-rt-users list:

  
https://lore.kernel.org/linux-rt-users/20241216190451.1c61977c@xxxxxxxxxxxxxxxxxxxx/

The TL;DR is that lazy preemption doesn't actually depend on generic
entry, and we should be able to enable it on arm64 independently of this
series. I'd posted a quick hack which Mike Galbraith cleaned up:

  
https://lore.kernel.org/linux-rt-users/a198a7dd9076f97b89d8882bb249b3bf303564ef.camel@xxxxxx/

... but that was never posted as a new thread to LAKML.

Would you be happy to take charge and take that patch, test it, and post
it here (or spin your own working version)? I was happy with the way it
looks but hadn't had the time for testing and so on.

I expect that we'll merge the generic entry code too, but having them
separate is a bit easier for bisection and backporting where people want
lazy preemption in downstream trees.

Mark.