On 2019-10-18 13:28:21 [+0200], Thomas Gleixner wrote: > The local lock is a 'sleeping' spinlock on RT (PI support) and as any other it is a "per-CPU 'sleeping' spinlock on RT". Which means that it can be acquired on multiple CPUs simultaneously (same like preempt_disable(),…). > RT substituted lock it also ensures that the task cannot be migrated when > it is held, which makes per cpu assumptions work - the kernel has lots of > them. :) … > tglx Sebastian
