PCI: vmd: Remove IRQ affinity so we can allocate more IRQs
authorKeith Busch <keith.busch@intel.com>
Wed, 30 Aug 2017 16:15:04 +0000 (12:15 -0400)
committerBjorn Helgaas <bhelgaas@google.com>
Wed, 30 Aug 2017 21:18:38 +0000 (16:18 -0500)
VMD hardware has to share its vectors among child devices in its PCI
domain so we should allocate as many as possible rather than just ones
that can be affinitized.

pci_alloc_irq_vectors_affinity() limits the number of affinitized IRQs to
the number of present CPUs (see irq_calc_affinity_vectors()).  But we'd
prefer to have more vectors, even if they aren't distributed across the
CPUs, so use pci_alloc_irq_vectors() instead.

Reported-by: Brad Goodman <Bradley.Goodman@dell.com>
Signed-off-by: Keith Busch <keith.busch@intel.com>
[bhelgaas: add irq_calc_affinity_vectors() reference to changelog]
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
drivers/pci/host/vmd.c

index 4fe1756af010bf32ba486bf17ee5cd6dfcca1096..509893bc3e63e910892ad0b0111c6e7a7fa65abb 100644 (file)
@@ -671,14 +671,6 @@ static int vmd_probe(struct pci_dev *dev, const struct pci_device_id *id)
        struct vmd_dev *vmd;
        int i, err;
 
-       /*
-        * The first vector is reserved for special use, so start affinity at
-        * the second vector
-        */
-       struct irq_affinity affd = {
-               .pre_vectors = 1,
-       };
-
        if (resource_size(&dev->resource[VMD_CFGBAR]) < (1 << 20))
                return -ENOMEM;
 
@@ -704,8 +696,8 @@ static int vmd_probe(struct pci_dev *dev, const struct pci_device_id *id)
        if (vmd->msix_count < 0)
                return -ENODEV;
 
-       vmd->msix_count = pci_alloc_irq_vectors_affinity(dev, 1, vmd->msix_count,
-                                       PCI_IRQ_MSIX | PCI_IRQ_AFFINITY, &affd);
+       vmd->msix_count = pci_alloc_irq_vectors(dev, 1, vmd->msix_count,
+                                       PCI_IRQ_MSIX);
        if (vmd->msix_count < 0)
                return vmd->msix_count;