On SRX branch series, the kernel might crash after enabling the ukern commands.
If any of the ukern commands ( 'dedicated-ukern-cpu ' or ' realtime-ukern-thread ') are active in the configuration; upon reboot, login, or even committing a configuration, the kernel might crash showing the following watchdog panic messages:
'dedicated-ukern-cpu
realtime-ukern-thread
SRX-LAB (ttyu0)
login:
NMI Exception on core:0 Watchdog status, core 0: 0xfffe66ffffb FPA INT Summery: 0x0 Trapframe Register Dump: zero: 0000000000000000 at: 0000000000000000 v0: 0000000000000004 v1: ffffffffc495aaa2 Stack trace: _mtx_lock_sleep+0x158 (0x50808ce0,0x80da85e8,0xffff8010,0xffff8010) ra 0x802511b8 sz 32
For this crash, ' _mtx_lock_sleep ' was the first function in the stack trace. For other crashes it can be ' smp_rendezvous_action ', 'smp_rendezvous_action_selected' or 'smp_barrier'.
_mtx_lock_sleep
smp_rendezvous_action
'smp_rendezvous_action_selected'
'smp_barrier'.
When enabling 'set chassis dedicated-ukern-cpu' , the watchdog panic can easily run into a race condition of atomic operation than the 'set chassis realtime-ukern-thread' .
'set chassis dedicated-ukern-cpu' ,
'set chassis realtime-ukern-thread'
The panic occurs when multiple threads operate the same counter.
show chassis routing-engine' :
root@SRX-LAB> show chassis routing-engine Last reboot reason 0x1000:reboot due to panic
This issue is fixed in the following Junos versions: