On Mon, 19 Aug 2019, Thomas Gleixner wrote: > Moving the posix cpu timers from on list to another and then expiring them > from the second list is avoiding to drop and reacquire sighand lock for > each timer expiry, but on the other hand it's more complicated code and > suboptimal for a small number of timers. > > Remove the extra list and expire them directly from the rbtree. Tests with > a large number of timers did not show a difference outside of the noise > range. > > This also allows to switch the crude heuristics of limiting the expiry of > timers to 20 for each type to a time based limitation which is way more > sensible.
This one is buggy. I know why so don't waste your time reviewing it.
