On Thu, Apr 06, 2017 at 12:47:21PM +0200, Thomas Gleixner wrote: > On Thu, 6 Apr 2017, Peter Zijlstra wrote:
> > IIRC MIPS has a case where only 1 in N cores has an FPU. And once a task > > uses FPU, it gets affined to the core that has one or something like > > that. > > > > Of course, nothing then stops someone else breaking that affinity. But I > > suspect it will simply fault on the next FPU instruction and 'reset' the > > mask or something. I've no clue and no real desire to know. > > It does nasty games with it's own storage of p->thread.user_cpus_allowed > and a fully seperate implementation of sys_sched_set|getaffinity. > > Plus a magic trap handler which forces the thread to a CPU with FPU when > the user_cpus_allowed mask intersects with the cpus_with_fpu_mask... > > Magic crap, which could all be replaced by a simple function in the > scheduler which allows to push a task to a FPU CPU and then disable > migration. If its even halfway coherent, I'd much rather let it stay where it it. I really want to limit migrate_disable() to PREEMPT_RT=y where its used to preserve spinlock semantics and not allow random other migrate_disable() usage in the kernel. Also note, that per the above, it can actually migrate to any core that has an FPU on, so its not a good match for migrate_disable() in any case.