Re: [PATCH] mm: introduce sysctl file to flush per-cpu vmstat statistics

[Thomas Gleixner <tglx@linutronix.de>]

On Mon, Dec 07 2020 at 08:08, Christoph Lameter wrote:
> On Thu, 3 Dec 2020, Thomas Gleixner wrote:
> Again saving data may not be possible through the kernel since syscalls
> may have too much overhead and latency.
>
>>        repeat the above...
>
> There is a fundamental misunderstanding here. This is not primarily about
> compute but about I/O. In particular I/O that does not involve the kernel.
> RDMA or things like DPDK, SPDK or other low hardware level things.

The fundamental misunderstanding is that people just look at and care
about their own turf.

The above is valid in compute intensive workloads where the compute
dominates the whole scenario. Not being disturbed there is a gain
because of not losing cache locality due to interrupts, IPIs or
whatever.

>> Policy is not a binary on/off problem. It's manifold across all levels
>> of the stack and only a kernel problem when it comes down to the last
>> line of defence.
>
> This a clearly defined set of functions and I am not sure how policy fits
> into that.

It's about silencing different and largely independent parts of the OS
on a particular CPU. Just defining upfront that there is only the choice
of all or nothing _is_ policy.

There is a very wide range of use case scenarios out there and just
because the ones which you care about needs X does not mean that X is
the right thing for everybody else. You still can have X and let other
people define their own set of things they want to be protected
against.

Aside of that having it selectively is a plus for debugability, testing
etc.

All or nothing is always the worst choice if a more fine grained
approach is possible with no big effort. Granted there are things where
you only can say: all or nothing, but that's definitely not the case
here.

Thanks,

        tglx