Don't enable interrupts in init_secondary().

[markj]

Summary

sched_throw() expects the caller to hold a spinlock section; see the
td_md initialization in cpu_fork(). This bug can cause a deadlock when
APs are started: statclock() fires before the thread lock acquired in
sched_throw() is released, and recurses on the thread lock.

[jhb]

sys/riscv/riscv/mp_machdep.c

231

Are interrupts enabled here at all? We really depend on them not being enabled on APs until after the first context switch.

[markj]

sys/riscv/riscv/mp_machdep.c

231

Hmm, it was added here: https://svnweb.freebsd.org/changeset/base/335005

We boot in QEMU without the intr_disable(), though. I added code to read and print sstatus here, and SIE is disabled. So it seems safe to remove the intr_disable().

[markj]

• Remove unneeded intr_disable() call.