On Fri, Jun 12, 2020 at 11:53 AM André Almeida via Libc-alpha
<libc-alpha@xxxxxxxxxxxxxx> wrote:
>
> Hello,
>
> This RFC is a followup to the previous discussion initiated from my last
> patch "futex: Implement mechanism to wait on any of several futexes"[1].
> As stated in the thread, the correct approach to move forward with the
> wait multiple operation would be to create a new syscall that would have
> all new cool features.
>
> The first patch adds the new interface and just translate the call for
> the old interface, without implementing new features. The goal here is
> to establish the interface and to check if everyone is happy with this
> API. The rest of patches are selftests to show the interface in action.
> I have the following questions:
>
> - Has anyone stared worked on a implementation of this interface? If
> yes, it would be nice to share the progress so we don't have duplicated
> work.
>
> - What suggestions do you have to implement this? Start from scratch or
> reuse the most code possible?
>
> - The interface seems correct and implements the requirements asked by you?
>
> - The proposed interface uses ktime_t type for absolute timeout, and I
> assumed that it should use values in a nsec resolution. If this is true,
> we have some problems with i386 ABI, please check out the
> COMPAT_32BIT_TIME implementation in patch 1 for more details. I
> haven't added a time64 implementation yet, until this is clarified.
>
> - Is expected to have a x32 ABI implementation as well? In the case of
> wait and wake, we could use the same as x86_64 ABI. However, for the
> waitv (aka wait on multiple futexes) we would need a proper x32 entry
> since we are dealing with 32bit pointers.
x32 should be able to use the same i386 compat systcall entry. Will it be
problem?
> Those are the cool new features that this syscall should address some
> day:
>
> - Operate with variable bit size futexes, not restricted to 32:
> 8, 16 and 64
>
> - Wait on multiple futexes, using the following semantics:
>
> struct futex_wait {
> void *uaddr;
> unsigned long val;
> unsigned long flags;
> };
>
> sys_futex_waitv(struct futex_wait *waiters, unsigned int nr_waiters,
> unsigned long flags, ktime_t *timo);
>
> - Have NUMA optimizations: if FUTEX_NUMA_FLAG is present, the `void *uaddr`
> argument won't be a u{8, 16, 32, 64} value anymore, but a struct
> containing a NUMA node hint:
>
> struct futex32_numa {
> u32 value __attribute__ ((aligned (8)));
> u32 hint;
> };
>
> struct futex64_numa {
> u64 value __attribute__ ((aligned (16)));
> u64 hint;
> };
>
H.J.
