Re: [PATCH v2 3/3] mm/slub: Use percpu partial free counter

From: Christoph Lameter
Date: Wed Mar 03 2021 - 19:01:34 EST


On Wed, 3 Mar 2021, Matthew Wilcox wrote:

> > Can this be allocated in an interrupt context?
> >
> > And I am not sure how local_t relates to that? Percpu counters can be used
> > in an interrupt context without the overhead of the address calculations
> > that are required by a local_t.
>
> As I understand the patch, this counts the number of partially free slabs.
> So if we start to free an object from a completely full slab in process
> context, as "load x, add one to x, store x" and take an interrupt
> between loading x and adding one to x, that interrupt handler might
> free a different object from another completely full slab. that would
> also load the same x, add one to it and store x, but then the process
> context would add one to the old x, overwriting the updated value from
> interrupt context.

this_cpu operations are "atomic" vs. preemption but on some platforms not
vs interrupts. That could be an issue in kmem_cache_free(). This would
need a modification to the relevant this_cpu ops so that interrupts are
disabled on those platforms.

Like this_cpu_inc_irq() or so?


> it's not the likeliest of races, and i don't know how important it is
> that these counters remain accurate. but using a local_t instead of
> a percpu long would fix the problem. i don't know why you think that
> a local_t needs "address calculations". perhaps you've misremembered
> what a local_t is?

local_t does not include the address arithmetic that the this_cpu
operation can implicitly perform on x86 f.e. with an segment register or
maybe another register on another platform thereby avoiding the need to
disable preemption or interrupts.

Therefore a manual calculation of the target address for a local_t
operation needs to be done beforehand which usually means disabling
interrupts and/or preemption for the code segment. Otherwise we may end up
on a different processor due to scheduler or other interruptions and use
the percpu counter value of a different processor which could be racy.