VIRQs come in 3 flavors, per-VPU, per-domain, and global, and the VIRQs
are tracked in per-cpu virq_to_irq arrays.

Per-domain and global VIRQs must be bound on CPU 0, and
bind_virq_to_irq() sets the per_cpu virq_to_irq at registration time
Later, the interrupt can migrate, and info->cpu is updated.  When
calling __unbind_from_irq(), the per-cpu virq_to_irq is cleared for a
different cpu.  If bind_virq_to_irq() is called again with CPU 0, the
stale irq is returned.  There won't be any irq_info for the irq, so
things break.

Make xen_rebind_evtchn_to_cpu() update the per_cpu virq_to_irq mappings
to keep them update to date with the current cpu.  This ensures the
correct virq_to_irq is cleared in __unbind_from_irq().

Fixes: e46cdb66c8fc ("xen: event channels")
Cc: stable@vger.kernel.org
Signed-off-by: Jason Andryuk <jason.andryuk@amd.com>
---
v3:
Kernel style brace placement
Delay setting old_cpu and tighten scope of variable

v2:
Different approach changing virq_to_irq
---
 drivers/xen/events/events_base.c | 13 ++++++++++++-
 1 file changed, 12 insertions(+), 1 deletion(-)

diff --git a/drivers/xen/events/events_base.c b/drivers/xen/events/events_base.c
index b060b5a95f45..9478fae014e5 100644
--- a/drivers/xen/events/events_base.c
+++ b/drivers/xen/events/events_base.c
@@ -1797,9 +1797,20 @@ static int xen_rebind_evtchn_to_cpu(struct irq_info *info, unsigned int tcpu)
 	 * virq or IPI channel, which don't actually need to be rebound. Ignore
 	 * it, but don't do the xenlinux-level rebind in that case.
 	 */
-	if (HYPERVISOR_event_channel_op(EVTCHNOP_bind_vcpu, &bind_vcpu) >= 0)
+	if (HYPERVISOR_event_channel_op(EVTCHNOP_bind_vcpu, &bind_vcpu) >= 0) {
+		int old_cpu = info->cpu;
+
 		bind_evtchn_to_cpu(info, tcpu, false);
 
+		if (info->type == IRQT_VIRQ) {
+			int virq = info->u.virq;
+			int irq = per_cpu(virq_to_irq, old_cpu)[virq];
+
+			per_cpu(virq_to_irq, old_cpu)[virq] = -1;
+			per_cpu(virq_to_irq, tcpu)[virq] = irq;
+		}
+	}
+
 	do_unmask(info, EVT_MASK_REASON_TEMPORARY);
 
 	return 0;
-- 
2.34.1

On 28.08.25 02:36, Jason Andryuk wrote:
> VIRQs come in 3 flavors, per-VPU, per-domain, and global, and the VIRQs
> are tracked in per-cpu virq_to_irq arrays.
> 
> Per-domain and global VIRQs must be bound on CPU 0, and
> bind_virq_to_irq() sets the per_cpu virq_to_irq at registration time
> Later, the interrupt can migrate, and info->cpu is updated.  When
> calling __unbind_from_irq(), the per-cpu virq_to_irq is cleared for a
> different cpu.  If bind_virq_to_irq() is called again with CPU 0, the
> stale irq is returned.  There won't be any irq_info for the irq, so
> things break.
> 
> Make xen_rebind_evtchn_to_cpu() update the per_cpu virq_to_irq mappings
> to keep them update to date with the current cpu.  This ensures the
> correct virq_to_irq is cleared in __unbind_from_irq().
> 
> Fixes: e46cdb66c8fc ("xen: event channels")
> Cc: stable@vger.kernel.org
> Signed-off-by: Jason Andryuk <jason.andryuk@amd.com>

Reviewed-by: Juergen Gross <jgross@suse.com>


Juergen