Re: [PATCH 2/2] del_timer_sync: proof of concept

* Oleg Nesterov <[EMAIL PROTECTED]> wrote: > I think that separate timer->pending field will require more changes, > because we can't read/write base+pending atomically. i think that's the killer argument in favor of the bit-trick. Being able to read/write base+pending atomically is a good excus

Re: [PATCH 2/2] del_timer_sync: proof of concept

Ingo Molnar wrote: > > * Oleg Nesterov <[EMAIL PROTECTED]> wrote: > > > New rules: > > ->_base & 1: is timer pending > > ->_base & ~1: timer's base > > how would it look like if we had a separate timer->pending field after > all? Would it be faster/cleaner? The only change visible

Re: [PATCH 2/2] del_timer_sync: proof of concept

* Oleg Nesterov <[EMAIL PROTECTED]> wrote: > New rules: > ->_base & 1: is timer pending > ->_base & ~1: timer's base how would it look like if we had a separate timer->pending field after all? Would it be faster/cleaner? (we dont need to keep them small _that_ bad - if ther