On Thu, Jun 25, 2020 at 10:42 PM Tomas Henzl <thenzl@xxxxxxxxxx> wrote:
>
> Affinity hint should be cleared before freeing irq.
Hi Tomas,
megaraid_sas driver setups IRQ affinity(non-managed) for MSI-x index
"0" to "instance->low_latency_index_start - 1".
Rest of MSI-x indices- "instance->low_latency_index_start" to
"instance->msix_vectors -1" are managed (API- pci_alloc_irq_affinity()
takes care of this)
so calling irq_set_affinity_hint() for managed IRQs does not seem
right. Please try setting affinity to NULL for only this range- "0" to
"instance->low_latency_index_start -1" and see if it solves your
problem.
Also, is it regression due to below commit:
f0b9e7bdc309 scsi: megaraid_sas: Set affinity for high IOPS reply queues
Thanks,
Sumit
>
> Signed-off-by: Tomas Henzl <thenzl@xxxxxxxxxx>
> ---
> drivers/scsi/megaraid/megaraid_sas_base.c | 5 ++++-
> 1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/scsi/megaraid/megaraid_sas_base.c b/drivers/scsi/megaraid/megaraid_sas_base.c
> index 00668335c..d5626ad8b 100644
> --- a/drivers/scsi/megaraid/megaraid_sas_base.c
> +++ b/drivers/scsi/megaraid/megaraid_sas_base.c
> @@ -5602,9 +5602,11 @@ megasas_setup_irqs_msix(struct megasas_instance *instance, u8 is_probe)
> &instance->irq_context[i])) {
> dev_err(&instance->pdev->dev,
> "Failed to register IRQ for vector %d.\n", i);
> - for (j = 0; j < i; j++)
> + for (j = 0; j < i; j++) {
> + irq_set_affinity_hint(pci_irq_vector(pdev, j), NULL);
> free_irq(pci_irq_vector(pdev, j),
> &instance->irq_context[j]);
> + }
> /* Retry irq register for IO_APIC*/
> instance->msix_vectors = 0;
> instance->msix_load_balance = false;
> @@ -5642,6 +5644,7 @@ megasas_destroy_irqs(struct megasas_instance *instance) {
>
> if (instance->msix_vectors)
> for (i = 0; i < instance->msix_vectors; i++) {
> + irq_set_affinity_hint(pci_irq_vector(instance->pdev, i), NULL);
> free_irq(pci_irq_vector(instance->pdev, i),
> &instance->irq_context[i]);
> }
> --
> 2.21.3
>
