powerpc/pseries: Pass MSI affinity to irq_create_mapping()
authorLaurent Vivier <lvivier@redhat.com>
Thu, 26 Nov 2020 08:28:52 +0000 (09:28 +0100)
committerThomas Gleixner <tglx@linutronix.de>
Mon, 30 Nov 2020 11:22:04 +0000 (12:22 +0100)
With virtio multiqueue, normally each queue IRQ is mapped to a CPU.

Commit 0d9f0a52c8b9f ("virtio_scsi: use virtio IRQ affinity") exposed
an existing shortcoming of the arch code by moving virtio_scsi to
the automatic IRQ affinity assignment.

The affinity is correctly computed in msi_desc but this is not applied
to the system IRQs.

It appears the affinity is correctly passed to rtas_setup_msi_irqs() but
lost at this point and never passed to irq_domain_alloc_descs()
(see commit 06ee6d571f0e ("genirq: Add affinity hint to irq allocation"))
because irq_create_mapping() doesn't take an affinity parameter.

Use the new irq_create_mapping_affinity() function, which allows to forward
the affinity setting from rtas_setup_msi_irqs() to irq_domain_alloc_descs().

With this change, the virtqueues are correctly dispatched between the CPUs
on pseries.

Fixes: e75eafb9b039 ("genirq/msi: Switch to new irq spreading infrastructure")
Signed-off-by: Laurent Vivier <lvivier@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Greg Kurz <groug@kaod.org>
Acked-by: Michael Ellerman <mpe@ellerman.id.au>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20201126082852.1178497-3-lvivier@redhat.com
arch/powerpc/platforms/pseries/msi.c

index 133f6ad..b3ac245 100644 (file)
@@ -458,7 +458,8 @@ again:
                        return hwirq;
                }
 
-               virq = irq_create_mapping(NULL, hwirq);
+               virq = irq_create_mapping_affinity(NULL, hwirq,
+                                                  entry->affinity);
 
                if (!virq) {
                        pr_debug("rtas_msi: Failed mapping hwirq %d\n", hwirq);