Re: [PATCH 0/6] mm: Remote LRU per-cpu pagevec cache/per-cpu page list drain support

[Thomas Gleixner <tglx@linutronix.de>]

Vlastimil,

On Thu, Sep 23 2021 at 09:12, Vlastimil Babka wrote:
> On 9/23/21 00:09, Thomas Gleixner wrote:
>>   local_lock() 		-> preempt_disable()
>>   local_lock_irq()	-> local_irq_disable()
>>   ...
>
> Yes, to be clean, this would have to be a new primitive, not just an abused
> local lock. It would just look similar to the RT version (a percpu array of
> spinlocks), so for this patchset it would allow not to have two such locks
> side be side (local + spin) while only one is being used. For maximum
> flexibility the initialization would take a CONFIG_ (or something
> compile-time bool) that when false would make the !RT version an empty
> struct and "locking" would rely on preempt/irq disable (just as with !RT
> local_lock). If compile-time true it would take a static key to decide on
> boot whether the !RT version only does the preepmt/irq disable or actually
> takes the lock.
>
> But as you say below, it's too much complexity for questionable benefit.
>
> But maybe this can all be avoided anyway, as I recalled what we do for
> vmstat already (IIUC). See quiet_vmstat() - when cpu enters the nohz mode,
> it flushes per-cpu vmstat diffs and then there's no reason to further
> disturb the cpu to do that while it's on NOHZ mode. We could do the same for
> lru pagevecs and pcplists?

I'm not sure about this. I like the idea of being able to offload things
to housekeeping CPUs not only in the full isolation case.

A good example is RCU which allows to offload all RCU processing to some
other CPU(s), which is useful even w/o full isolation.

The synchronous quiescing on entering NOHZ full mode is a cute
workaround but for one it makes entering NOHZ full more expensive and it
does not necessarily provide good isolation guarantees under all
circumstances, while a full remote processing definitely does.

I think it's at least worthwhile to investigate.

Thanks,

        tglx