On Tue, 25 Dec 2018, Jason Thorpe wrote:
On Dec 25, 2018, at 9:42 AM, Alexander Nasonov <alnsn%yandex.ru@localhost> wrote: File: tprof.out Number of samples: 220 percentage nsamples pid lwp cpu k address symbol ------------ -------- ------ ---- ---- - ---------------- ------ 27.727274% 61 0 0 0 1 ffffffc0000040a4 splx+0xbc 12.727272% 28 0 0 0 1 ffffffc0003c7424 mutex_enter+0x124I’m going to guess that there’s a spin mutex that’s contended.
tprof on armv7/armv8 will always looks like that, I think because the PMU interrupt handler is installed at IPL_HIGH. It would be a lot better if we used FIQ for this..