x86/xen: don't let xen_pv_play_dead() return

A function called via the paravirt play_dead() hook should not return
to the caller.

xen_pv_play_dead() has a problem in this regard, as it currently will
return in case an offlined cpu is brought to life again. This can be
changed only by doing basically a longjmp() to cpu_bringup_and_idle(),
as the hypercall for bringing down the cpu will just return when the
cpu is coming up again. Just re-initializing the cpu isn't possible,
as the Xen hypervisor will deny that operation.

So introduce xen_cpu_bringup_again() resetting the stack and calling
cpu_bringup_and_idle(), which can be called after HYPERVISOR_vcpu_op()
in xen_pv_play_dead().

Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20221125063248.30256-2-jgross@suse.com
Signed-off-by: Juergen Gross <jgross@suse.com>

1 files changed, 2 insertions, 11 deletions

diff --git a/arch/x86/xen/smp_pv.c b/arch/x86/xen/smp_pv.c
index 6175f2c5c822..a97b0505a483 100644
--- a/arch/x86/xen/smp_pv.c
+++ b/arch/x86/xen/smp_pv.c
@@ -385,17 +385,8 @@ static void xen_pv_play_dead(void) /* used only with HOTPLUG_CPU */
 {
 	play_dead_common();
 	HYPERVISOR_vcpu_op(VCPUOP_down, xen_vcpu_nr(smp_processor_id()), NULL);
-	cpu_bringup();
-	/*
-	 * commit 4b0c0f294 (tick: Cleanup NOHZ per cpu data on cpu down)
-	 * clears certain data that the cpu_idle loop (which called us
-	 * and that we return from) expects. The only way to get that
-	 * data back is to call:
-	 */
-	tick_nohz_idle_enter();
-	tick_nohz_idle_stop_tick_protected();
-
-	cpuhp_online_idle(CPUHP_AP_ONLINE_IDLE);
+	xen_cpu_bringup_again((unsigned long)task_pt_regs(current));
+	BUG();
 }
 
 #else /* !CONFIG_HOTPLUG_CPU */