ideas to find a root cause of a long rcu_wait_gp delay

ideas to find a root cause of a long rcu_wait_gp delay

[Norman]

Hello community,


  I am sorry in advance if this is not the right place to ask questions.


  I am facing a werid long delay for executing docker exec and add new named network namespace. The stack for a ip netns add command is ,


cat /proc/3843856/stack
[<0>] __wait_rcu_gp+0x134/0x160
[<0>] synchronize_sched+0x5d/0x80
[<0>] synchronize_net+0x1c/0x30
[<0>] nf_register_net_hook+0x232/0x2b0
[<0>] nf_register_net_hooks+0x41/0x80
[<0>] ebt_register_table+0x303/0x350 [ebtables]
[<0>] frame_filter_net_init+0x23/0x25 [ebtable_filter]
[<0>] ops_init+0x44/0x130
[<0>] setup_net+0xb6/0x170
[<0>] copy_net_ns+0xd1/0x210
[<0>] create_new_namespaces+0x11f/0x1d0
[<0>] unshare_nsproxy_namespaces+0x5e/0xb0
[<0>] SyS_unshare+0x213/0x3c0
[<0>] do_syscall_64+0x73/0x130
[<0>] entry_SYSCALL_64_after_hwframe+0x3d/0xa2
[<0>] 0xffffffffffffffff


Unfortuately the machine's kernel is very ancient, it is a ubuntu 20.04 system with kernel versioned 4.15.0-72-generic. I try to read references online, and I believe synchronize_sched is waiting for grace period completes for all online cpus,
but the machine does not have very high cpu usage or system load.


So I am very appreciated for ideas to help debugging or identifying the problem.


Best Regards,
Jiatong Shen