On Wed, Jan 15, 2020 at 7:41 PM Sudeep Holla <sudeep.ho...@arm.com> wrote: > > On Wed, Jan 15, 2020 at 02:34:10PM +0800, Hsin-Yi Wang wrote: > > Currently system reboots uses architecture specific codes (smp_send_stop) > > to offline non reboot CPUs. Most architecture's implementation is looping > > through all non reboot online CPUs and call ipi function to each of them. > > Some > > architecture like arm64, arm, and x86... would set offline masks to cpu > > without > > really offline them. This causes some race condition and kernel warning > > comes > > out sometimes when system reboots. > > > > This patch adds a config ARCH_OFFLINE_CPUS_ON_REBOOT, which would offline > > cpus in > > migrate_to_reboot_cpu(). If non reboot cpus are all offlined here, the loop > > for > > checking online cpus would be an empty loop. If architecture don't enable > > this > > config, or some cpus somehow fails to offline, it would fallback to ipi > > function. > > > > What's the timing impact on systems with large number of CPUs(say 256 or > more) ? I remember we added some change to reduce the wait times for > offlining CPUs in system suspend path on arm64, still not negligible. >
This is not the final solution, but I would still provided some data points here: Tested on my arm64 with 4 cpu: 2 a53 and 2 a72. Offlining 3 cpu takes about 60~65 ms Offlining 2 cpu(a53+a72 or a72+a72) takes about 42~47 ms Offlining 1 cpu(a53 or a72) takes about 23~25 ms. It would take longer time for systems with large number of CPUs.
