On 11/26/20 1:19 PM, Alexander Gordeev wrote:
> The directed MSIs are delivered to CPUs whose address is
> written to the MSI message data. The current code assumes
> that a CPU logical number (as it is seen by the kernel)
> is also that CPU address.
>
> The above assumption is not correct, as the CPU address
> is rather the value returned by STAP instruction. That
> value is not necessarily matches the kernel logical CPU
> number.
I took the liberty of correcting the "is not" grammar error
above to "does not necessarily match".
>
> Fixes: e979ce7bced2 ("s390/pci: provide support for CPU directed interrupts")
> Signed-off-by: Alexander Gordeev <agordeev@xxxxxxxxxxxxx>
Still works well and checkpatches clean. I
> ---
> arch/s390/pci/pci_irq.c | 15 ++++++++++++---
> 1 file changed, 12 insertions(+), 3 deletions(-)
>
> diff --git a/arch/s390/pci/pci_irq.c b/arch/s390/pci/pci_irq.c
> index 743f257cf2cb..1309fd302f58 100644
> --- a/arch/s390/pci/pci_irq.c
> +++ b/arch/s390/pci/pci_irq.c
> @@ -103,9 +103,10 @@ static int zpci_set_irq_affinity(struct irq_data *data, const struct cpumask *de
> {
> struct msi_desc *entry = irq_get_msi_desc(data->irq);
> struct msi_msg msg = entry->msg;
> + int cpu_addr = smp_cpu_get_cpu_address(cpumask_first(dest));
>
> msg.address_lo &= 0xff0000ff;
> - msg.address_lo |= (cpumask_first(dest) << 8);
> + msg.address_lo |= (cpu_addr << 8);
> pci_write_msi_msg(data->irq, &msg);
>
> return IRQ_SET_MASK_OK;
> @@ -238,6 +239,7 @@ int arch_setup_msi_irqs(struct pci_dev *pdev, int nvec, int type)
> unsigned long bit;
> struct msi_desc *msi;
> struct msi_msg msg;
> + int cpu_addr;
> int rc, irq;
>
> zdev->aisb = -1UL;
> @@ -287,9 +289,16 @@ int arch_setup_msi_irqs(struct pci_dev *pdev, int nvec, int type)
> handle_percpu_irq);
> msg.data = hwirq - bit;
> if (irq_delivery == DIRECTED) {
> + if (msi->affinity) {
> + cpu = cpumask_first(&msi->affinity->mask);
> + cpu_addr = smp_cpu_get_cpu_address(cpu);
> + } else {
> + cpu_addr = 0;
> + }
One question I haven't really figured out from looking at the spec is
why using cpu_addr = 0; is a good fallback. Shouldn't that be smp_cpu_get_cpu_address(0) or
do we now know that the CPU addresses always start at 0?
> +
> msg.address_lo = zdev->msi_addr & 0xff0000ff;
> - msg.address_lo |= msi->affinity ?
> - (cpumask_first(&msi->affinity->mask) << 8) : 0;
> + msg.address_lo |= (cpu_addr << 8);
> +
> for_each_possible_cpu(cpu) {
> airq_iv_set_data(zpci_ibv[cpu], hwirq, irq);
> }
>
