[Qemu-devel] [PATCH v4 06/15] spapr_pci: add PHB unrealize

Greg Kurz posted 15 patches 6 years, 8 months ago
There is a newer version of this series
[Qemu-devel] [PATCH v4 06/15] spapr_pci: add PHB unrealize
Posted by Greg Kurz 6 years, 8 months ago
To support PHB hotplug we need to clean up lingering references,
memory, child properties, etc. prior to the PHB object being
finalized. Generally this will be called as a result of calling
object_unparent() on the PHB object, which in turn would normally
be called as the result of an unplug() operation.

When the PHB is finalized, child objects will be unparented in
turn, and finalized if the PHB was the only reference holder. so
we don't bother to explicitly unparent child objects of the PHB
(spapr_iommu, spapr_drc, etc).

The formula that gives the number of DMA windows is moved to an
inline function in the hw/pci-host/spapr.h header because it
will have other users.

The unrealize function is able to cope with partially realized PHBs.
It is hence used to implement proper rollback on the realize error
path.

Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Greg Kurz <groug@kaod.org>
---
v4: - reverted to v2
v3: - don't free LSIs at unrealize
v2: - implement rollback with unrealize function
---
 hw/ppc/spapr_pci.c          |   75 +++++++++++++++++++++++++++++++++++++++++--
 include/hw/pci-host/spapr.h |    5 +++
 2 files changed, 76 insertions(+), 4 deletions(-)

diff --git a/hw/ppc/spapr_pci.c b/hw/ppc/spapr_pci.c
index d68595531d5a..e3781dd110b2 100644
--- a/hw/ppc/spapr_pci.c
+++ b/hw/ppc/spapr_pci.c
@@ -1565,6 +1565,64 @@ static void spapr_pci_unplug_request(HotplugHandler *plug_handler,
     }
 }
 
+static void spapr_phb_finalizefn(Object *obj)
+{
+    sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(obj);
+
+    g_free(sphb->dtbusname);
+    sphb->dtbusname = NULL;
+}
+
+static void spapr_phb_unrealize(DeviceState *dev, Error **errp)
+{
+    sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
+    SysBusDevice *s = SYS_BUS_DEVICE(dev);
+    PCIHostState *phb = PCI_HOST_BRIDGE(s);
+    sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(phb);
+    sPAPRTCETable *tcet;
+    int i;
+    const unsigned windows_supported = spapr_phb_windows_supported(sphb);
+
+    if (sphb->msi) {
+        g_hash_table_unref(sphb->msi);
+        sphb->msi = NULL;
+    }
+
+    /*
+     * Remove IO/MMIO subregions and aliases, rest should get cleaned
+     * via PHB's unrealize->object_finalize
+     */
+    for (i = windows_supported - 1; i >= 0; i--) {
+        tcet = spapr_tce_find_by_liobn(sphb->dma_liobn[i]);
+        if (tcet) {
+            memory_region_del_subregion(&sphb->iommu_root,
+                                        spapr_tce_get_iommu(tcet));
+        }
+    }
+
+    for (i = PCI_NUM_PINS - 1; i >= 0; i--) {
+        if (sphb->lsi_table[i].irq) {
+            spapr_irq_free(spapr, sphb->lsi_table[i].irq, 1);
+            sphb->lsi_table[i].irq = 0;
+        }
+    }
+
+    QLIST_REMOVE(sphb, list);
+
+    memory_region_del_subregion(&sphb->iommu_root, &sphb->msiwindow);
+
+    address_space_destroy(&sphb->iommu_as);
+
+    qbus_set_hotplug_handler(BUS(phb->bus), NULL, &error_abort);
+    pci_unregister_root_bus(phb->bus);
+
+    memory_region_del_subregion(get_system_memory(), &sphb->iowindow);
+    if (sphb->mem64_win_pciaddr != (hwaddr)-1) {
+        memory_region_del_subregion(get_system_memory(), &sphb->mem64window);
+    }
+    memory_region_del_subregion(get_system_memory(), &sphb->mem32window);
+}
+
 static void spapr_phb_realize(DeviceState *dev, Error **errp)
 {
     /* We don't use SPAPR_MACHINE() in order to exit gracefully if the user
@@ -1582,8 +1640,7 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
     PCIBus *bus;
     uint64_t msi_window_size = 4096;
     sPAPRTCETable *tcet;
-    const unsigned windows_supported =
-        sphb->ddw_enabled ? SPAPR_PCI_DMA_MAX_WINDOWS : 1;
+    const unsigned windows_supported = spapr_phb_windows_supported(sphb);
 
     if (!spapr) {
         error_setg(errp, TYPE_SPAPR_PCI_HOST_BRIDGE " needs a pseries machine");
@@ -1740,6 +1797,10 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
             if (local_err) {
                 error_propagate_prepend(errp, local_err,
                                         "can't allocate LSIs: ");
+                /*
+                 * Older machines will never support PHB hotplug, ie, this is an
+                 * init only path and QEMU will terminate. No need to rollback.
+                 */
                 return;
             }
 
@@ -1749,7 +1810,7 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
         spapr_irq_claim(spapr, irq, &local_err);
         if (local_err) {
             error_propagate_prepend(errp, local_err, "can't allocate LSIs: ");
-            return;
+            goto unrealize;
         }
 
         sphb->lsi_table[i].irq = irq;
@@ -1769,13 +1830,17 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
         if (!tcet) {
             error_setg(errp, "Creating window#%d failed for %s",
                        i, sphb->dtbusname);
-            return;
+            goto unrealize;
         }
         memory_region_add_subregion(&sphb->iommu_root, 0,
                                     spapr_tce_get_iommu(tcet));
     }
 
     sphb->msi = g_hash_table_new_full(g_int_hash, g_int_equal, g_free, g_free);
+    return;
+
+unrealize:
+    spapr_phb_unrealize(dev, NULL);
 }
 
 static int spapr_phb_children_reset(Object *child, void *opaque)
@@ -1974,6 +2039,7 @@ static void spapr_phb_class_init(ObjectClass *klass, void *data)
 
     hc->root_bus_path = spapr_phb_root_bus_path;
     dc->realize = spapr_phb_realize;
+    dc->unrealize = spapr_phb_unrealize;
     dc->props = spapr_phb_properties;
     dc->reset = spapr_phb_reset;
     dc->vmsd = &vmstate_spapr_pci;
@@ -1989,6 +2055,7 @@ static const TypeInfo spapr_phb_info = {
     .name          = TYPE_SPAPR_PCI_HOST_BRIDGE,
     .parent        = TYPE_PCI_HOST_BRIDGE,
     .instance_size = sizeof(sPAPRPHBState),
+    .instance_finalize = spapr_phb_finalizefn,
     .class_init    = spapr_phb_class_init,
     .interfaces    = (InterfaceInfo[]) {
         { TYPE_HOTPLUG_HANDLER },
diff --git a/include/hw/pci-host/spapr.h b/include/hw/pci-host/spapr.h
index 51d81c4b7ce8..7cfce54a9449 100644
--- a/include/hw/pci-host/spapr.h
+++ b/include/hw/pci-host/spapr.h
@@ -163,4 +163,9 @@ static inline void spapr_phb_vfio_reset(DeviceState *qdev)
 
 void spapr_phb_dma_reset(sPAPRPHBState *sphb);
 
+static inline unsigned spapr_phb_windows_supported(sPAPRPHBState *sphb)
+{
+    return sphb->ddw_enabled ? SPAPR_PCI_DMA_MAX_WINDOWS : 1;
+}
+
 #endif /* PCI_HOST_SPAPR_H */


Re: [Qemu-devel] [PATCH v4 06/15] spapr_pci: add PHB unrealize
Posted by David Gibson 6 years, 8 months ago
On Tue, Feb 12, 2019 at 07:24:33PM +0100, Greg Kurz wrote:
> To support PHB hotplug we need to clean up lingering references,
> memory, child properties, etc. prior to the PHB object being
> finalized. Generally this will be called as a result of calling
> object_unparent() on the PHB object, which in turn would normally
> be called as the result of an unplug() operation.
> 
> When the PHB is finalized, child objects will be unparented in
> turn, and finalized if the PHB was the only reference holder. so
> we don't bother to explicitly unparent child objects of the PHB
> (spapr_iommu, spapr_drc, etc).
> 
> The formula that gives the number of DMA windows is moved to an
> inline function in the hw/pci-host/spapr.h header because it
> will have other users.
> 
> The unrealize function is able to cope with partially realized PHBs.
> It is hence used to implement proper rollback on the realize error
> path.
> 
> Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
> Signed-off-by: Greg Kurz <groug@kaod.org>

Reviewed-by: David Gibson <david@gibson.dropbear.id.au>

> ---
> v4: - reverted to v2
> v3: - don't free LSIs at unrealize
> v2: - implement rollback with unrealize function
> ---
>  hw/ppc/spapr_pci.c          |   75 +++++++++++++++++++++++++++++++++++++++++--
>  include/hw/pci-host/spapr.h |    5 +++
>  2 files changed, 76 insertions(+), 4 deletions(-)
> 
> diff --git a/hw/ppc/spapr_pci.c b/hw/ppc/spapr_pci.c
> index d68595531d5a..e3781dd110b2 100644
> --- a/hw/ppc/spapr_pci.c
> +++ b/hw/ppc/spapr_pci.c
> @@ -1565,6 +1565,64 @@ static void spapr_pci_unplug_request(HotplugHandler *plug_handler,
>      }
>  }
>  
> +static void spapr_phb_finalizefn(Object *obj)
> +{
> +    sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(obj);
> +
> +    g_free(sphb->dtbusname);
> +    sphb->dtbusname = NULL;
> +}
> +
> +static void spapr_phb_unrealize(DeviceState *dev, Error **errp)
> +{
> +    sPAPRMachineState *spapr = SPAPR_MACHINE(qdev_get_machine());
> +    SysBusDevice *s = SYS_BUS_DEVICE(dev);
> +    PCIHostState *phb = PCI_HOST_BRIDGE(s);
> +    sPAPRPHBState *sphb = SPAPR_PCI_HOST_BRIDGE(phb);
> +    sPAPRTCETable *tcet;
> +    int i;
> +    const unsigned windows_supported = spapr_phb_windows_supported(sphb);
> +
> +    if (sphb->msi) {
> +        g_hash_table_unref(sphb->msi);
> +        sphb->msi = NULL;
> +    }
> +
> +    /*
> +     * Remove IO/MMIO subregions and aliases, rest should get cleaned
> +     * via PHB's unrealize->object_finalize
> +     */
> +    for (i = windows_supported - 1; i >= 0; i--) {
> +        tcet = spapr_tce_find_by_liobn(sphb->dma_liobn[i]);
> +        if (tcet) {
> +            memory_region_del_subregion(&sphb->iommu_root,
> +                                        spapr_tce_get_iommu(tcet));
> +        }
> +    }
> +
> +    for (i = PCI_NUM_PINS - 1; i >= 0; i--) {
> +        if (sphb->lsi_table[i].irq) {
> +            spapr_irq_free(spapr, sphb->lsi_table[i].irq, 1);
> +            sphb->lsi_table[i].irq = 0;
> +        }
> +    }
> +
> +    QLIST_REMOVE(sphb, list);
> +
> +    memory_region_del_subregion(&sphb->iommu_root, &sphb->msiwindow);
> +
> +    address_space_destroy(&sphb->iommu_as);
> +
> +    qbus_set_hotplug_handler(BUS(phb->bus), NULL, &error_abort);
> +    pci_unregister_root_bus(phb->bus);
> +
> +    memory_region_del_subregion(get_system_memory(), &sphb->iowindow);
> +    if (sphb->mem64_win_pciaddr != (hwaddr)-1) {
> +        memory_region_del_subregion(get_system_memory(), &sphb->mem64window);
> +    }
> +    memory_region_del_subregion(get_system_memory(), &sphb->mem32window);
> +}
> +
>  static void spapr_phb_realize(DeviceState *dev, Error **errp)
>  {
>      /* We don't use SPAPR_MACHINE() in order to exit gracefully if the user
> @@ -1582,8 +1640,7 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
>      PCIBus *bus;
>      uint64_t msi_window_size = 4096;
>      sPAPRTCETable *tcet;
> -    const unsigned windows_supported =
> -        sphb->ddw_enabled ? SPAPR_PCI_DMA_MAX_WINDOWS : 1;
> +    const unsigned windows_supported = spapr_phb_windows_supported(sphb);
>  
>      if (!spapr) {
>          error_setg(errp, TYPE_SPAPR_PCI_HOST_BRIDGE " needs a pseries machine");
> @@ -1740,6 +1797,10 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
>              if (local_err) {
>                  error_propagate_prepend(errp, local_err,
>                                          "can't allocate LSIs: ");
> +                /*
> +                 * Older machines will never support PHB hotplug, ie, this is an
> +                 * init only path and QEMU will terminate. No need to rollback.
> +                 */
>                  return;
>              }
>  
> @@ -1749,7 +1810,7 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
>          spapr_irq_claim(spapr, irq, &local_err);
>          if (local_err) {
>              error_propagate_prepend(errp, local_err, "can't allocate LSIs: ");
> -            return;
> +            goto unrealize;
>          }
>  
>          sphb->lsi_table[i].irq = irq;
> @@ -1769,13 +1830,17 @@ static void spapr_phb_realize(DeviceState *dev, Error **errp)
>          if (!tcet) {
>              error_setg(errp, "Creating window#%d failed for %s",
>                         i, sphb->dtbusname);
> -            return;
> +            goto unrealize;
>          }
>          memory_region_add_subregion(&sphb->iommu_root, 0,
>                                      spapr_tce_get_iommu(tcet));
>      }
>  
>      sphb->msi = g_hash_table_new_full(g_int_hash, g_int_equal, g_free, g_free);
> +    return;
> +
> +unrealize:
> +    spapr_phb_unrealize(dev, NULL);
>  }
>  
>  static int spapr_phb_children_reset(Object *child, void *opaque)
> @@ -1974,6 +2039,7 @@ static void spapr_phb_class_init(ObjectClass *klass, void *data)
>  
>      hc->root_bus_path = spapr_phb_root_bus_path;
>      dc->realize = spapr_phb_realize;
> +    dc->unrealize = spapr_phb_unrealize;
>      dc->props = spapr_phb_properties;
>      dc->reset = spapr_phb_reset;
>      dc->vmsd = &vmstate_spapr_pci;
> @@ -1989,6 +2055,7 @@ static const TypeInfo spapr_phb_info = {
>      .name          = TYPE_SPAPR_PCI_HOST_BRIDGE,
>      .parent        = TYPE_PCI_HOST_BRIDGE,
>      .instance_size = sizeof(sPAPRPHBState),
> +    .instance_finalize = spapr_phb_finalizefn,
>      .class_init    = spapr_phb_class_init,
>      .interfaces    = (InterfaceInfo[]) {
>          { TYPE_HOTPLUG_HANDLER },
> diff --git a/include/hw/pci-host/spapr.h b/include/hw/pci-host/spapr.h
> index 51d81c4b7ce8..7cfce54a9449 100644
> --- a/include/hw/pci-host/spapr.h
> +++ b/include/hw/pci-host/spapr.h
> @@ -163,4 +163,9 @@ static inline void spapr_phb_vfio_reset(DeviceState *qdev)
>  
>  void spapr_phb_dma_reset(sPAPRPHBState *sphb);
>  
> +static inline unsigned spapr_phb_windows_supported(sPAPRPHBState *sphb)
> +{
> +    return sphb->ddw_enabled ? SPAPR_PCI_DMA_MAX_WINDOWS : 1;
> +}
> +
>  #endif /* PCI_HOST_SPAPR_H */
> 

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson