On Fri, Mar 04, 2016 at 11:11:09PM -0500, David Miller wrote: > From: Eric Dumazet <eric.duma...@gmail.com> > Date: Fri, 04 Mar 2016 16:08:30 -0800 > > > __inet_del_ifa() should probably take into account in_dev->dead (no > > promotion, no list scan...) > > Indeed, that is the real problem:
Well, tried it out. Indeed it partially released the contention but with patch applied I stuck with Samples: 20K of event 'cpu-clock', Event count (approx.): 4647374938 Overhead Shared Object Symbol 19.27% [kernel] [k] __local_bh_enable_ip 15.97% [kernel] [k] lock_acquire 15.12% [kernel] [k] fib_del_ifaddr 11.66% [kernel] [k] lock_release 7.57% [kernel] [k] lock_is_held 5.35% [kernel] [k] lock_acquired 3.26% [kernel] [k] _raw_spin_unlock_irqrestore 3.04% [kernel] [k] __local_bh_disable_ip 2.10% [kernel] [k] _raw_spin_unlock_irq 1.54% [kernel] [k] native_save_fl 1.37% [kernel] [k] ___might_sleep 0.90% [kernel] [k] do_raw_spin_trylock 0.83% [kernel] [k] nf_ct_iterate_cleanup 0.77% [kernel] [k] debug_lockdep_rcu_enabled 0.62% [kernel] [k] tick_nohz_idle_enter 0.61% [kernel] [k] _raw_spin_lock 0.58% [kernel] [k] __slab_alloc.isra.43.constprop.47 0.42% [kernel] [k] get_parent_ip 0.40% [kernel] [k] preempt_count_sub 0.36% [kernel] [k] native_save_fl 0.34% [kernel] [k] _raw_spin_unlock 0.31% [kernel] [k] do_raw_spin_unlock and until everything get cleaned up I couldn't connect to the node via ssh. I continue playing with patch maybe I find some other optimization paths. Thanks!
