spapr_pci: map the MSI window in each PHB

On sPAPR, virtio devices are connected to the PCI bus and use MSI-X.
Commit cc943c36faa192cd4b32af8fe5edb31894017d35 has modified MSI-X
so that writes are made using the bus master address space and follow
the IOMMU path.

Unfortunately, the IOMMU address space address space does not have an
MSI window: the notification is silently dropped in unassigned_mem_write
instead of reaching the guest... The most visible effect is that all
virtio devices are non-functional on sPAPR since then. :(

This patch does the following:
1) map the MSI window into the IOMMU address space for each PHB
   - since each PHB instantiates its own IOMMU address space, we
     can safely map the window at a fixed address (SPAPR_PCI_MSI_WINDOW)
   - no real need to keep the MSI window setup in a separate function,
     the spapr_pci_msi_init() code moves to spapr_phb_realize().

2) kill the global MSI window as it is not needed in the end

Signed-off-by: Greg Kurz <gkurz@linux.vnet.ibm.com>
Signed-off-by: Alexander Graf <agraf@suse.de>

1 files changed, 24 insertions, 29 deletions

diff --git a/hw/ppc/spapr_pci.c b/hw/ppc/spapr_pci.c
index e894f07f84..ad0da7fdc4 100644
--- a/hw/ppc/spapr_pci.c
+++ b/hw/ppc/spapr_pci.c
@@ -345,7 +345,7 @@ static void rtas_ibm_change_msi(PowerPCCPU *cpu, sPAPREnvironment *spapr,
     }
 
     /* Setup MSI/MSIX vectors in the device (via cfgspace or MSIX BAR) */
-    spapr_msi_setmsg(pdev, spapr->msi_win_addr, ret_intr_type == RTAS_TYPE_MSIX,
+    spapr_msi_setmsg(pdev, SPAPR_PCI_MSI_WINDOW, ret_intr_type == RTAS_TYPE_MSIX,
                      irq, req_num);
 
     /* Add MSI device to cache */
@@ -469,34 +469,6 @@ static const MemoryRegionOps spapr_msi_ops = {
     .endianness = DEVICE_LITTLE_ENDIAN
 };
 
-void spapr_pci_msi_init(sPAPREnvironment *spapr, hwaddr addr)
-{
-    uint64_t window_size = 4096;
-
-    /*
-     * As MSI/MSIX interrupts trigger by writing at MSI/MSIX vectors,
-     * we need to allocate some memory to catch those writes coming
-     * from msi_notify()/msix_notify().
-     * As MSIMessage:addr is going to be the same and MSIMessage:data
-     * is going to be a VIRQ number, 4 bytes of the MSI MR will only
-     * be used.
-     *
-     * For KVM we want to ensure that this memory is a full page so that
-     * our memory slot is of page size granularity.
-     */
-#ifdef CONFIG_KVM
-    if (kvm_enabled()) {
-        window_size = getpagesize();
-    }
-#endif
-
-    spapr->msi_win_addr = addr;
-    memory_region_init_io(&spapr->msiwindow, NULL, &spapr_msi_ops, spapr,
-                          "msi", window_size);
-    memory_region_add_subregion(get_system_memory(), spapr->msi_win_addr,
-                                &spapr->msiwindow);
-}
-
 /*
  * PHB PCI device
  */
@@ -516,6 +488,7 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
     char *namebuf;
     int i;
     PCIBus *bus;
+    uint64_t msi_window_size = 4096;
 
     if (sphb->index != -1) {
         hwaddr windows_base;
@@ -608,6 +581,28 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
     address_space_init(&sphb->iommu_as, &sphb->iommu_root,
                        sphb->dtbusname);
 
+    /*
+     * As MSI/MSIX interrupts trigger by writing at MSI/MSIX vectors,
+     * we need to allocate some memory to catch those writes coming
+     * from msi_notify()/msix_notify().
+     * As MSIMessage:addr is going to be the same and MSIMessage:data
+     * is going to be a VIRQ number, 4 bytes of the MSI MR will only
+     * be used.
+     *
+     * For KVM we want to ensure that this memory is a full page so that
+     * our memory slot is of page size granularity.
+     */
+#ifdef CONFIG_KVM
+    if (kvm_enabled()) {
+        msi_window_size = getpagesize();
+    }
+#endif
+
+    memory_region_init_io(&sphb->msiwindow, NULL, &spapr_msi_ops, spapr,
+                          "msi", msi_window_size);
+    memory_region_add_subregion(&sphb->iommu_root, SPAPR_PCI_MSI_WINDOW,
+                                &sphb->msiwindow);
+
     pci_setup_iommu(bus, spapr_pci_dma_iommu, sphb);
 
     pci_bus_set_route_irq_fn(bus, spapr_route_intx_pin_to_irq);