* Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote: > On Wed, Oct 11, 2023 at 10:21:06AM -0600, Jens Axboe wrote: > > On 10/9/23 8:16 AM, Jens Axboe wrote: > > > On 10/9/23 8:13 AM, Peter Zijlstra wrote: > > >> On Mon, Oct 09, 2023 at 11:00:19AM +0200, Arnd Bergmann wrote: > > >>> On Mon, Oct 9, 2023, at 10:48, Peter Zijlstra wrote: > > >>>> On Mon, Oct 09, 2023 at 12:31:18PM +1100, Stephen Rothwell wrote: > > >>>>> diff --cc arch/alpha/kernel/syscalls/syscall.tbl > > >>>>> index 5d05ab716a74,b1865f9bb31e..000000000000 > > >>>>> --- a/arch/alpha/kernel/syscalls/syscall.tbl > > >>>>> +++ b/arch/alpha/kernel/syscalls/syscall.tbl > > >>>>> @@@ -492,4 -492,6 +492,7 @@@ > > >>>>> 560 common set_mempolicy_home_node sys_ni_syscall > > >>>>> 561 common cachestat sys_cachestat > > >>>>> 562 common fchmodat2 sys_fchmodat2 > > >>>>> -563 common futex_wake sys_futex_wake > > >>>>> -564 common futex_wait sys_futex_wait > > >>>>> -565 common futex_requeue sys_futex_requeue > > >>>>> +563 common map_shadow_stack sys_map_shadow_stack > > >>>>> ++564 common futex_wake sys_futex_wake > > >>>>> ++565 common futex_wait sys_futex_wait > > >>>>> ++566 common futex_requeue sys_futex_requeue > > >>>> > > >>>> So this renumbers the (futex) stuff on Alpha, does anybody care? AFAICT > > >>>> Alpha does not follow the unistd order and meh. > > >>> > > >>> Let's not make it worse for now. All the numbers since the > > >>> introduction of the time64 syscalls are offset by exactly 120 > > >>> on alpha, and I'd prefer to keep it that way for the moment. > > >>> > > >>> I still hope to eventually finish the conversion of all architectures > > >>> to a single syscall.tbl for numbers >400, and if that happens before > > >>> the end of alpha, a different ordering would just be extra pain. > > >> > > >> Fair enough; should we look at rebase those futex patches for this? (bit > > >> of a pain as that would also mean rebasing block) > > > > > > From my point of view, this isn't a huge problem if we do it now. The > > > io_uring-futex branch is a separate branch and I have nothing on top of > > > it, so I could easily just re-pull your updated branch and rebase my > > > changes on top. > > > > > >> Or do we want to keep this fixup in the merge resolution and make sure > > >> Linus is aware? > > > > > > If you're OK with it, I'd say let's rebase and save ourselves the > > > trouble at merge time. > > > > Peter, what's the verdict - do you want to rebase it, or leave it as-is? > > Ah, I looked into doing this, but tip/locking/core has since grown a > bunch of patches and has a merge commit -- I talked to Ingo yesterday > and he proposed just queueing a fix on top instead of doing a full > rebase. > > Ingo, that still your preferred solution? Yeah, that would be the best solution IMO - it's not like there's any real prospect of someone bisecting futex2 patch-enablement commits on Alpha ... and the bisection distance isn't particularly large either in any case. [ This would also document the very real historic conflict between these numbers, as it happened. ] Thanks, Ingo
