PCI: hv: Specify CPU_AFFINITY_ALL for MSI affinity when >= 32 CPUs
authorK. Y. Srinivasan <kys@microsoft.com>
Fri, 24 Mar 2017 18:07:21 +0000 (11:07 -0700)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 25 May 2017 13:44:47 +0000 (15:44 +0200)
commit 433fcf6b7b31f1f233dd50aeb9d066a0f6ed4b9d upstream.

When we have 32 or more CPUs in the affinity mask, we should use a special
constant to specify that to the host. Fix this issue.

Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Long Li <longli@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/pci/host/pci-hyperv.c

index ddabf7b..dafb4cd 100644 (file)
@@ -72,6 +72,7 @@ enum {
        PCI_PROTOCOL_VERSION_CURRENT = PCI_PROTOCOL_VERSION_1_1
 };
 
+#define CPU_AFFINITY_ALL       -1ULL
 #define PCI_CONFIG_MMIO_LENGTH 0x2000
 #define CFG_PAGE_OFFSET 0x1000
 #define CFG_PAGE_SIZE (PCI_CONFIG_MMIO_LENGTH - CFG_PAGE_OFFSET)
@@ -889,9 +890,13 @@ static void hv_compose_msi_msg(struct irq_data *data, struct msi_msg *msg)
         * processors because Hyper-V only supports 64 in a guest.
         */
        affinity = irq_data_get_affinity_mask(data);
-       for_each_cpu_and(cpu, affinity, cpu_online_mask) {
-               int_pkt->int_desc.cpu_mask |=
-                       (1ULL << vmbus_cpu_number_to_vp_number(cpu));
+       if (cpumask_weight(affinity) >= 32) {
+               int_pkt->int_desc.cpu_mask = CPU_AFFINITY_ALL;
+       } else {
+               for_each_cpu_and(cpu, affinity, cpu_online_mask) {
+                       int_pkt->int_desc.cpu_mask |=
+                               (1ULL << vmbus_cpu_number_to_vp_number(cpu));
+               }
        }
 
        ret = vmbus_sendpacket(hpdev->hbus->hdev->channel, int_pkt,