Back in early 2008, 4KiB stack support was added to reduce memory
consumption on workloads with lots of threads, using a separate IRQ
stack per CPU to alleviate stack pressure. However the SMP code added a
year and a half later didn't set up the IRQ stack when bringing up
secondary CPUs, resulting in a crash when SMP is configured with 4KiB
stacks, as soon as the first interrupt arrived on a secondary CPU.
Fix with calls to irq_ctx_exit()/irq_ctx_exit() when bringing up/down a
CPU.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: linux-metag@vger.kernel.org
{
local_irq_disable();
idle_task_exit();
+ irq_ctx_exit(smp_processor_id());
(void)cpu_report_death();
panic("No TBI found!");
per_cpu_trap_init(cpu);
+ irq_ctx_init(cpu);
preempt_disable();