Re: [tip:timers/core] [posix] 1535cb8028: stress-ng.epoll.ops_per_sec 36.2% regression

[Thomas Gleixner <tglx@linutronix.de>]

On Thu, Mar 27 2025 at 10:11, Eric Dumazet wrote:
> On Thu, Mar 27, 2025 at 9:26 AM Eric Dumazet <edumazet@google.com> wrote:
>
>> We could place all these atomic fields in separate cache lines,
>> to keep read-only fields shared as much as possible.
>>
>
> Following one-liner seems good enough to separate the 4 atomics used
> to control/limit
>
> UCOUNT_RLIMIT_NPROC, UCOUNT_RLIMIT_MSGQUEUE, UCOUNT_RLIMIT_SIGPENDING,
> UCOUNT_RLIMIT_MEMLOCK,
>
>
> diff --git a/include/linux/user_namespace.h b/include/linux/user_namespace.h
> index 7183e5aca282..6cc3fbec3632 100644
> --- a/include/linux/user_namespace.h
> +++ b/include/linux/user_namespace.h
> @@ -120,7 +120,7 @@ struct ucounts {
>         kuid_t uid;
>         atomic_t count;
>         atomic_long_t ucount[UCOUNT_COUNTS];
> -       atomic_long_t rlimit[UCOUNT_RLIMIT_COUNTS];
> +       atomic_long_t ____cacheline_aligned_in_smp rlimit[UCOUNT_RLIMIT_COUNTS];
>  };

Cute. How much bloat does it cause?