Should interrupts be enabled in kernel page fault handler in netbsd?

Question

I am seeing an issue where netbsd system kernel is having watchdog timeout and the stack trace shows that kernel page fault handler got interrupted by software interrupt to process ip packets the trace looks like this. This netbsd 4.x

0xcf910db0: at m_copyback0+0x3fc

0xcf910e00: at m_makewritable+0x40

0xcf910e20: at fr_check_wrapper+0x44

0xcf910e50: at pfil_run_hooks+0xd0

0xcf910e90: at ip_input+0x3dc

0xcf910ed0: at ipintr+0x4c

0xcf910f00: at netintr+0x60

0xcf910f20: at softintr_dispatch+0x150

0xcf910f60: at ext_intr+0x2bc

0xcf910fa0: at cpu_switchto+0x768

0xf1733d90: at uvm_fault_internal+0x150

My question is, is it a good idea to have interrupts enabled inside page fault handler?

Answer 1

Looks like keeping interrupts enabled is a good way to prevent deadlock, because page fault handler can also run in user process context that can go to sleep.