On Thu, 19 Jan 2023 17:36:22 +0100 Sebastian Andrzej Siewior > On 2023-01-19 21:59:03 [+0800], Hillf Danton wrote: > > On Thu, 19 Jan 2023 09:32:22 +0100 Sebastian Andrzej Siewior <bigeasy@xxxxxxxxxxxxx> > > > > > > As far as Mel's efforts go, I am satisfied so far. > > > > If not because you can, could you specify why 4ms fails to cure starvation? > > It does not fail to cure the starvation. I haven't tested it myself but > base on Mel's description and the patch it very much looks like it cures > the writer starvation. > > If you don't like the 4ms, it could be 1ms or 40ms - it does not really > matter. The 4ms is aligned on the generic implementation which uses the > same value. Unless there is strong evidence to use something else I > don't see the need to diverse. I am fine with either 4ms or 40ms, or a second. Given the cure, does it still work when reader bias for RT tasks is allowed? If not, why keep starving waiters after they pay the 40ms price?
