It appears that the defect outlined in 9c15eeb5362c4 ("genirq: Allow
fasteoi handler to resend interrupts on concurrent handling") also
affects some other less stellar MSI controllers, this time using
the handle_simple_irq() flow.

Teach this flow about irqd_needs_resend_when_in_progress(). Given
the invasive nature of this workaround, only this flow is updated.

Signed-off-by: Marc Zyngier <maz@kernel.org>
---
 kernel/irq/chip.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/kernel/irq/chip.c b/kernel/irq/chip.c
index 2b274007e8bab..6e789035919f7 100644
--- a/kernel/irq/chip.c
+++ b/kernel/irq/chip.c
@@ -559,7 +559,13 @@ void handle_simple_irq(struct irq_desc *desc)
 {
 	guard(raw_spinlock)(&desc->lock);
 
-	if (!irq_can_handle(desc))
+	if (!irq_can_handle_pm(desc)) {
+		if (irqd_needs_resend_when_in_progress(&desc->irq_data))
+			desc->istate |= IRQS_PENDING;
+		return;
+	}
+
+	if (!irq_can_handle_actions(desc))
 		return;
 
 	kstat_incr_irqs_this_cpu(desc);
-- 
2.39.2

On Sat, Jun 28 2025 at 18:29, Marc Zyngier wrote:

> It appears that the defect outlined in 9c15eeb5362c4 ("genirq: Allow
> fasteoi handler to resend interrupts on concurrent handling") also
> affects some other less stellar MSI controllers, this time using
> the handle_simple_irq() flow.
>
> Teach this flow about irqd_needs_resend_when_in_progress(). Given
> the invasive nature of this workaround, only this flow is updated.
>
> Signed-off-by: Marc Zyngier <maz@kernel.org>

As I have no conflicting changes pending in this area, this can go
through the PCI tree along with the rest of this lot. Therefore:

Reviewed-by: Thomas Gleixner <tglx@linutronix.de>

Thanks,

        tglx