Thomas Gleixner <tglx@xxxxxxxxxxxxx> writes:
> Paul!
>
> On Tue, Nov 21 2023 at 07:19, Paul E. McKenney wrote:
>> On Tue, Nov 21, 2023 at 10:00:59AM -0500, Steven Rostedt wrote:
>>> Right now, the use of cond_resched() is basically a whack-a-mole game where
>>> we need to whack all the mole loops with the cond_resched() hammer. As
>>> Thomas said, this is backwards. It makes more sense to just not preempt in
>>> areas that can cause pain (like holding a mutex or in an RCU critical
>>> section), but still have the general kernel be fully preemptable.
>>
>> Which is quite true, but that whack-a-mole game can be ended without
>> getting rid of build-time selection of the preemption model. Also,
>> that whack-a-mole game can be ended without eliminating all calls to
>> cond_resched().
>
> Which calls to cond_resched() should not be eliminated?
>
> They all suck and keeping some of them is just counterproductive as
> again people will sprinkle them all over the place for the very wrong
> reasons.
And, as Thomas alludes to here, cond_resched() is not always cost free.
Needing to call cond_resched() forces us to restructure hot paths in
ways that results in worse performance/complex code.
One example is clear_huge_page(), where removing the need to call
cond_resched() every once in a while allows the processor to optimize
differently.
*Milan* mm/clear_huge_page x86/clear_huge_page change
(GB/s) (GB/s)
pg-sz=2MB 14.55 19.29 +32.5%
pg-sz=1GB 19.34 49.60 +156.4%
(See https://lore.kernel.org/all/20230830184958.2333078-1-ankur.a.arora@xxxxxxxxxx/)
And, that's one of the simpler examples from mm. We do this kind of arbitrary
batching all over the place.
Or see the filemap_read() example that Linus gives here:
https://lore.kernel.org/lkml/CAHk-=whpYjm_AizQij6XEfTd7xvGjrVCx5gzHcHm=2Xijt+Kyg@xxxxxxxxxxxxxx/#t
Thanks
--
ankur
