[Intel-gfx] [patch RFC 00/15] mm/highmem: Provide a preemptible variant of kmap_atomic & friends

[Thomas Gleixner]

On Mon, Sep 21 2020 at 21:27, Thomas Gleixner wrote:
> On Mon, Sep 21 2020 at 09:24, Linus Torvalds wrote:
>> On Mon, Sep 21, 2020 at 12:39 AM Thomas Gleixner <tglx at linutronix.de> wrote:
>> Maybe we really *could* call this new kmap functionality something
>> like "kmap_percpu()" (or maybe "local" is good enough), and make it
>> act like your RT code does for spinlocks - not disable preemption, but
>> only disabling CPU migration.
>
> I"m all for it, but the scheduler people have opinions :)

I just took the latest version of migrate disable patches

  https://lore.kernel.org/r/20200921163557.234036895@infradead.org

removed the RT dependency on top of them and adopted the kmap stuff
(addressing the various comments while at it and unbreaking ARM).

I'm not going to post that until there is consensus about the general
availability of migrate disable, but for those who want to play with it
I've pushed the resulting combo out to:

   git://git.kernel.org/pub/scm/linux/kernel/git/tglx/devel.git highmem 

For testing I've tweaked a few places to use the new _local() variants
and it survived testing so far and I've verified that there is actual
preemption which means zap/restore of the thread local kmaps.

Thanks,

        tglx