Skip to contentSkip to conversation

pnv-phb related cleanups

PATCH In review Daniel Henrique Barboza 4 participants 34 messages Jun 13, 2022 – Jun 17, 2022

lore
Share & tools
lore
#1 lore raw

PATCH 1/11 ppc/pnv: move root port attach to pnv_phb4_realize()

In reply to Daniel Henrique Barboza

Creating a root port is something related to the PHB, not the PEC. It also makes the logic more in line with what pnv-phb3 does.

---
 hw/pci-host/pnv_phb4.c     | 4 ++++
 hw/pci-host/pnv_phb4_pec.c | 3 ---
 2 files changed, 4 insertions(+), 3 deletions(-)
hw/pci-host/pnv_phb4.c +4 -0
index 6594016121..23ad8de7ee 100644
@@ -1547,6 +1547,7 @@ static void pnv_phb4_instance_init(Object *obj)
 static void pnv_phb4_realize(DeviceState *dev, Error **errp)
 {
     PnvPHB4 *phb = PNV_PHB4(dev);
+    PnvPhb4PecClass *pecc = PNV_PHB4_PEC_GET_CLASS(phb->pec);
     PCIHostState *pci = PCI_HOST_BRIDGE(dev);
     XiveSource *xsrc = &phb->xsrc;
     int nr_irqs;
@@ -1583,6 +1584,9 @@ static void pnv_phb4_realize(DeviceState *dev, Error **errp)
     pci_setup_iommu(pci->bus, pnv_phb4_dma_iommu, phb);
     pci->bus->flags |= PCI_BUS_EXTENDED_CONFIG_SPACE;
 
+    /* Add a single Root port if running with defaults */
+    pnv_phb_attach_root_port(pci, pecc->rp_model);
+
     /* Setup XIVE Source */
     if (phb->big_phb) {
         nr_irqs = PNV_PHB4_MAX_INTs;
hw/pci-host/pnv_phb4_pec.c +0 -3
index 8b7e823fa5..c9aaf1c28e 100644
@@ -130,9 +130,6 @@ static void pnv_pec_default_phb_realize(PnvPhb4PecState *pec,
     if (!sysbus_realize(SYS_BUS_DEVICE(phb), errp)) {
         return;
     }
-
-    /* Add a single Root port if running with defaults */
-    pnv_phb_attach_root_port(PCI_HOST_BRIDGE(phb), pecc->rp_model);
 }
 
 static void pnv_pec_realize(DeviceState *dev, Error **errp)
Signature · 1 line
2.36.1
#9 lore raw

PATCH 3/11 ppc/pnv: use dev->parent_bus->parent to get the PHB

In reply to Daniel Henrique Barboza

It is not advisable to execute an object_dynamic_cast() to poke into bus->qbus.parent and follow it up with a C cast into the PnvPHB type we think we got.

A better way is to access the PnvPHB object via a QOM macro accessing the existing parent links of the DeviceState. For a given pnv-phb3/4-root-port 'dev', dev->parent_bus will give us the PHB bus, and dev->parent_bus->parent is the PHB. Use the adequate QOM macro to assert the type, and keep the NULL check in case we didn't get the object we were expecting.

---
 hw/pci-host/pnv_phb3.c | 10 +++++++---
 hw/pci-host/pnv_phb4.c | 10 +++++++---
 2 files changed, 14 insertions(+), 6 deletions(-)
hw/pci-host/pnv_phb3.c +7 -3
index 4ba660f8b9..7901d8172c 100644
@@ -1139,12 +1139,16 @@ static void pnv_phb3_root_port_realize(DeviceState *dev, Error **errp)
 {
     PCIERootPortClass *rpc = PCIE_ROOT_PORT_GET_CLASS(dev);
     PCIDevice *pci = PCI_DEVICE(dev);
-    PCIBus *bus = pci_get_bus(pci);
     PnvPHB3 *phb = NULL;
     Error *local_err = NULL;
 
-    phb = (PnvPHB3 *) object_dynamic_cast(OBJECT(bus->qbus.parent),
-                                          TYPE_PNV_PHB3);
+    /*
+     * dev->parent_bus gives access to the pnv-phb-root bus.
+     * The PnvPHB3 is the owner (parent) of the bus.
+     */
+    if (dev && dev->parent_bus) {
+        phb = PNV_PHB3(dev->parent_bus->parent);
+    }
 
     if (!phb) {
         error_setg(errp,
hw/pci-host/pnv_phb4.c +7 -3
index ffd9d8a947..bae9398d86 100644
@@ -1782,12 +1782,16 @@ static void pnv_phb4_root_port_realize(DeviceState *dev, Error **errp)
 {
     PCIERootPortClass *rpc = PCIE_ROOT_PORT_GET_CLASS(dev);
     PCIDevice *pci = PCI_DEVICE(dev);
-    PCIBus *bus = pci_get_bus(pci);
     PnvPHB4 *phb = NULL;
     Error *local_err = NULL;
 
-    phb = (PnvPHB4 *) object_dynamic_cast(OBJECT(bus->qbus.parent),
-                                          TYPE_PNV_PHB4);
+    /*
+     * dev->parent_bus gives access to the pnv-phb-root bus.
+     * The PnvPHB4 is the owner (parent) of the bus.
+     */
+    if (dev && dev->parent_bus) {
+        phb = PNV_PHB4(dev->parent_bus->parent);
+    }
 
     if (!phb) {
         error_setg(errp, "%s must be connected to pnv-phb4 buses", dev->id);
Signature · 1 line
2.36.1
#5 lore raw

PATCH 2/11 ppc/pnv: attach phb3/phb4 root ports in QOM tree

In reply to Daniel Henrique Barboza

At this moment we leave the pnv-phb3(4)-root-port unattached in QOM:

/unattached (container)
(...)

    /device[2] (pnv-phb3-root-port)
      /bus master container[0] (memory-region)
      /bus master[0] (memory-region)
      /pci_bridge_io[0] (memory-region)
      /pci_bridge_io[1] (memory-region)
      /pci_bridge_mem[0] (memory-region)
      /pci_bridge_pci[0] (memory-region)
      /pci_bridge_pref_mem[0] (memory-region)
      /pci_bridge_vga_io_hi[0] (memory-region)
      /pci_bridge_vga_io_lo[0] (memory-region)
      /pci_bridge_vga_mem[0] (memory-region)
      /pcie.0 (PCIE)

Let's make changes in pnv_phb_attach_root_port() to attach the created root ports to its corresponding PHB.

This is the result afterwards:

    /pnv-phb3[0] (pnv-phb3)
      /lsi (ics)
      /msi (phb3-msi)
      /msi32[0] (memory-region)
      /msi64[0] (memory-region)
      /pbcq (pnv-pbcq)
    (...)
      /phb3_iommu[0] (pnv-phb3-iommu-memory-region)
      /pnv-phb3-root.0 (pnv-phb3-root)
        /pnv-phb3-root-port[0] (pnv-phb3-root-port)
          /bus master container[0] (memory-region)
          /bus master[0] (memory-region)
          /pci_bridge_io[0] (memory-region)
          /pci_bridge_io[1] (memory-region)
          /pci_bridge_mem[0] (memory-region)
          /pci_bridge_pci[0] (memory-region)
          /pci_bridge_pref_mem[0] (memory-region)
          /pci_bridge_vga_io_hi[0] (memory-region)
          /pci_bridge_vga_io_lo[0] (memory-region)
          /pci_bridge_vga_mem[0] (memory-region)
          /pcie.0 (PCIE)
---
 hw/pci-host/pnv_phb3.c | 2 +-
 hw/pci-host/pnv_phb4.c | 2 +-
 hw/ppc/pnv.c           | 7 ++++++-
 include/hw/ppc/pnv.h   | 2 +-
 4 files changed, 9 insertions(+), 4 deletions(-)
hw/pci-host/pnv_phb3.c +1 -1
index 26ac9b7123..4ba660f8b9 100644
@@ -1052,7 +1052,7 @@ static void pnv_phb3_realize(DeviceState *dev, Error **errp)
 
     pci_setup_iommu(pci->bus, pnv_phb3_dma_iommu, phb);
 
-    pnv_phb_attach_root_port(PCI_HOST_BRIDGE(phb), TYPE_PNV_PHB3_ROOT_PORT);
+    pnv_phb_attach_root_port(pci, TYPE_PNV_PHB3_ROOT_PORT, phb->phb_id);
 }
 
 void pnv_phb3_update_regions(PnvPHB3 *phb)
hw/pci-host/pnv_phb4.c +1 -1
index 23ad8de7ee..ffd9d8a947 100644
@@ -1585,7 +1585,7 @@ static void pnv_phb4_realize(DeviceState *dev, Error **errp)
     pci->bus->flags |= PCI_BUS_EXTENDED_CONFIG_SPACE;
 
     /* Add a single Root port if running with defaults */
-    pnv_phb_attach_root_port(pci, pecc->rp_model);
+    pnv_phb_attach_root_port(pci, pecc->rp_model, phb->phb_id);
 
     /* Setup XIVE Source */
     if (phb->big_phb) {
hw/ppc/pnv.c +6 -1
index 7c08a78d6c..40e0cbd84d 100644
@@ -1190,9 +1190,14 @@ static void pnv_chip_icp_realize(Pnv8Chip *chip8, Error **errp)
 }
 
 /* Attach a root port device */
-void pnv_phb_attach_root_port(PCIHostState *pci, const char *name)
+void pnv_phb_attach_root_port(PCIHostState *pci, const char *name, int index)
 {
     PCIDevice *root = pci_new(PCI_DEVFN(0, 0), name);
+    g_autofree char *default_id = g_strdup_printf("%s[%d]", name, index);
+    const char *dev_id = DEVICE(root)->id;
+
+    object_property_add_child(OBJECT(pci->bus), dev_id ? dev_id : default_id,
+                              OBJECT(root));
 
     pci_realize_and_unref(root, pci->bus, &error_fatal);
 }
include/hw/ppc/pnv.h +1 -1
index 86cb7d7f97..033890a23f 100644
@@ -189,7 +189,7 @@ DECLARE_INSTANCE_CHECKER(PnvChip, PNV_CHIP_POWER10,
                          TYPE_PNV_CHIP_POWER10)
 
 PowerPCCPU *pnv_chip_find_cpu(PnvChip *chip, uint32_t pir);
-void pnv_phb_attach_root_port(PCIHostState *pci, const char *name);
+void pnv_phb_attach_root_port(PCIHostState *pci, const char *name, int index);
 
 #define TYPE_PNV_MACHINE       MACHINE_TYPE_NAME("powernv")
 typedef struct PnvMachineClass PnvMachineClass;
Signature · 1 line
2.36.1
#0 lore raw

cover pnv-phb related cleanups

Hi,

I decided to make some cleanups/fixes after the feedback received in the v2 of the pnv-phb proxy series [1]. A lot of the reviews and comment were related to the current state of the code instead of what was being done there.

With this series we want to:

- provide a better base for proxy pnv-phb series, fixing and changing code that is already in place;
- make a common logic that applies to both default and user created devices. There's no reason to make distinctions between them aside from what we will already have to deal with.

After these cleanups we'll be able to simplify the work required in [1].

[1] https://lists.gnu.org/archive/html/qemu-devel/2022-05/msg06254.html

Daniel Henrique Barboza (11):
ppc/pnv: move root port attach to pnv_phb4_realize() ppc/pnv: attach phb3/phb4 root ports in QOM tree ppc/pnv: use dev->parent_bus->parent to get the PHB ppc/pnv: use dev instead of pci->qdev in root_port_realize() ppc/pnv: make pnv_ics_get() use the chip8->phbs[] array ppc/pnv: make pnv_ics_resend() use chip8->phbs[] ppc/pnv: make pnv_chip_power8_pic_print_info() use chip8->phbs[] ppc/pnv: turn chip8->phbs[] into a PnvPHB3* array ppc/pnv: add PHB object/bus parenting helpers ppc/pnv: move PHB3 initialization to realize time ppc/pnv: move PHB4 parent fixup to phb4_realize()

 hw/pci-host/pnv_phb3.c     |  30 +++++++--
 hw/pci-host/pnv_phb4.c     |  31 +++++++--
 hw/pci-host/pnv_phb4_pec.c |   4 --
 hw/ppc/pnv.c               | 127 +++++++++++++++++++++----------------
 include/hw/ppc/pnv.h       |  10 ++-
 5 files changed, 128 insertions(+), 74 deletions(-)
Signature · 1 line
2.36.1
#19 lore raw

PATCH 6/11 ppc/pnv: make pnv_ics_resend() use chip8->phbs[]

pnv_ics_resend() is scrolling through all the child objects of the chip to search for the PHBs. It's faster and simpler to just use the phbs[] array.

pnv_ics_resend_child() was folded into pnv_ics_resend() since it's too simple to justify its own function.

---
 hw/ppc/pnv.c | 22 +++++++---------------
 1 file changed, 7 insertions(+), 15 deletions(-)
hw/ppc/pnv.c +7 -15
index 05a8d5034f..d70deffa1d 100644
@@ -1993,28 +1993,20 @@ PnvChip *pnv_get_chip(PnvMachineState *pnv, uint32_t chip_id)
     return NULL;
 }
 
-static int pnv_ics_resend_child(Object *child, void *opaque)
-{
-    PnvPHB3 *phb3 = (PnvPHB3 *) object_dynamic_cast(child, TYPE_PNV_PHB3);
-
-    if (phb3) {
-        ics_resend(&phb3->lsis);
-        ics_resend(ICS(&phb3->msis));
-    }
-    return 0;
-}
-
 static void pnv_ics_resend(XICSFabric *xi)
 {
     PnvMachineState *pnv = PNV_MACHINE(xi);
-    int i;
+    int i, j;
 
     for (i = 0; i < pnv->num_chips; i++) {
-        PnvChip *chip = pnv->chips[i];
         Pnv8Chip *chip8 = PNV8_CHIP(pnv->chips[i]);
 
-        ics_resend(&chip8->psi.ics);
-        object_child_foreach(OBJECT(chip), pnv_ics_resend_child, NULL);
+        for (j = 0; j < chip8->num_phbs; j++) {
+            PnvPHB3 *phb3 = &chip8->phbs[j];
+
+            ics_resend(&phb3->lsis);
+            ics_resend(ICS(&phb3->msis));
+        }
     }
 }
 
Signature · 1 line
2.36.1
#25 lore raw

PATCH 8/11 ppc/pnv: turn chip8->phbs[] into a PnvPHB3* array

In reply to Daniel Henrique Barboza

When enabling user created PHBs (a change reverted by commit 9c10d86fee) we were handling PHBs created by default versus by the user in different manners. The only difference between these PHBs is that one will have a valid phb3->chip that is assigned during pnv_chip_power8_realize(), while the user created needs to search which chip it belongs to.

Aside from that there shouldn't be any difference. Making the default PHBs behave in line with the user created ones will make it easier to re-introduce them later on. It will also make the code easier to follow since we are dealing with them in equal manner.

The first step is to turn chip8->phbs[] into a PnvPHB3 pointer array. This will allow us to assign user created PHBs into it later on. The way we initilize the default case is now more in line with that would happen with the user created case: the object is created, parented by the chip because pnv_xscom_dt() relies on it, and then assigned to the array.

---
 hw/ppc/pnv.c         | 19 ++++++++++++++-----
 include/hw/ppc/pnv.h |  6 +++++-
 2 files changed, 19 insertions(+), 6 deletions(-)
hw/ppc/pnv.c +14 -5
index 5e3323e950..6ce9e94e05 100644
@@ -660,7 +660,7 @@ static void pnv_chip_power8_pic_print_info(PnvChip *chip, Monitor *mon)
     ics_pic_print_info(&chip8->psi.ics, mon);
 
     for (i = 0; i < chip8->num_phbs; i++) {
-        PnvPHB3 *phb3 = &chip8->phbs[i];
+        PnvPHB3 *phb3 = chip8->phbs[i];
 
         pnv_phb3_msi_pic_print_info(&phb3->msis, mon);
         ics_pic_print_info(&phb3->lsis, mon);
@@ -1149,7 +1149,16 @@ static void pnv_chip_power8_instance_init(Object *obj)
     chip8->num_phbs = pcc->num_phbs;
 
     for (i = 0; i < chip8->num_phbs; i++) {
-        object_initialize_child(obj, "phb[*]", &chip8->phbs[i], TYPE_PNV_PHB3);
+        PnvPHB3 *phb3 = PNV_PHB3(object_new(TYPE_PNV_PHB3));
+
+        /*
+         * We need the chip to parent the PHB to allow the DT
+         * to build correctly (via pnv_xscom_dt()).
+         *
+         * TODO: the PHB should be parented by a PEC device.
+         */
+        object_property_add_child(obj, "phb[*]", OBJECT(phb3));
+        chip8->phbs[i] = phb3;
     }
 
 }
@@ -1278,7 +1287,7 @@ static void pnv_chip_power8_realize(DeviceState *dev, Error **errp)
 
     /* PHB3 controllers */
     for (i = 0; i < chip8->num_phbs; i++) {
-        PnvPHB3 *phb = &chip8->phbs[i];
+        PnvPHB3 *phb = chip8->phbs[i];
 
         object_property_set_int(OBJECT(phb), "index", i, &error_fatal);
         object_property_set_int(OBJECT(phb), "chip-id", chip->chip_id,
@@ -1963,7 +1972,7 @@ static ICSState *pnv_ics_get(XICSFabric *xi, int irq)
         }
 
         for (j = 0; j < chip8->num_phbs; j++) {
-            pnv_ics_get_phb_ics(&chip8->phbs[j], &args);
+            pnv_ics_get_phb_ics(chip8->phbs[j], &args);
 
             if (args.ics) {
                 return args.ics;
@@ -1996,7 +2005,7 @@ static void pnv_ics_resend(XICSFabric *xi)
         Pnv8Chip *chip8 = PNV8_CHIP(pnv->chips[i]);
 
         for (j = 0; j < chip8->num_phbs; j++) {
-            PnvPHB3 *phb3 = &chip8->phbs[j];
+            PnvPHB3 *phb3 = chip8->phbs[j];
 
             ics_resend(&phb3->lsis);
             ics_resend(ICS(&phb3->msis));
include/hw/ppc/pnv.h +5 -1
index 033890a23f..11f1089289 100644
@@ -80,7 +80,11 @@ struct Pnv8Chip {
     PnvHomer     homer;
 
 #define PNV8_CHIP_PHB3_MAX 4
-    PnvPHB3      phbs[PNV8_CHIP_PHB3_MAX];
+    /*
+     * The array is used to allow quick access to the phbs by
+     * pnv_ics_get_child() and pnv_ics_resend_child().
+     */
+    PnvPHB3      *phbs[PNV8_CHIP_PHB3_MAX];
     uint32_t     num_phbs;
 
     XICSFabric    *xics;
Signature · 1 line
2.36.1
#16 lore raw

PATCH 5/11 ppc/pnv: make pnv_ics_get() use the chip8->phbs[] array

In reply to Daniel Henrique Barboza

The function is working today by getting all the child objects of the chip, interacting with each of them to check whether the child is a PHB, and then doing what needs to be done.

We have all the chip PHBs in the phbs[] array so interacting with all child objects is unneeded.

---
 hw/ppc/pnv.c | 30 ++++++++++++++----------------
 1 file changed, 14 insertions(+), 16 deletions(-)
hw/ppc/pnv.c +14 -16
index 40e0cbd84d..05a8d5034f 100644
@@ -1944,41 +1944,39 @@ typedef struct ForeachPhb3Args {
     ICSState *ics;
 } ForeachPhb3Args;
 
-static int pnv_ics_get_child(Object *child, void *opaque)
+static void pnv_ics_get_phb_ics(PnvPHB3 *phb3, ForeachPhb3Args *args)
 {
-    ForeachPhb3Args *args = opaque;
-    PnvPHB3 *phb3 = (PnvPHB3 *) object_dynamic_cast(child, TYPE_PNV_PHB3);
+    if (ics_valid_irq(&phb3->lsis, args->irq)) {
+        args->ics = &phb3->lsis;
+    }
 
-    if (phb3) {
-        if (ics_valid_irq(&phb3->lsis, args->irq)) {
-            args->ics = &phb3->lsis;
-        }
-        if (ics_valid_irq(ICS(&phb3->msis), args->irq)) {
-            args->ics = ICS(&phb3->msis);
-        }
+    if (ics_valid_irq(ICS(&phb3->msis), args->irq)) {
+        args->ics = ICS(&phb3->msis);
     }
-    return args->ics ? 1 : 0;
 }
 
 static ICSState *pnv_ics_get(XICSFabric *xi, int irq)
 {
     PnvMachineState *pnv = PNV_MACHINE(xi);
     ForeachPhb3Args args = { irq, NULL };
-    int i;
+    int i, j;
 
     for (i = 0; i < pnv->num_chips; i++) {
-        PnvChip *chip = pnv->chips[i];
         Pnv8Chip *chip8 = PNV8_CHIP(pnv->chips[i]);
 
         if (ics_valid_irq(&chip8->psi.ics, irq)) {
             return &chip8->psi.ics;
         }
 
-        object_child_foreach(OBJECT(chip), pnv_ics_get_child, &args);
-        if (args.ics) {
-            return args.ics;
+        for (j = 0; j < chip8->num_phbs; j++) {
+            pnv_ics_get_phb_ics(&chip8->phbs[j], &args);
+
+            if (args.ics) {
+                return args.ics;
+            }
         }
     }
+
     return NULL;
 }
 
Signature · 1 line
2.36.1
#14 lore raw

PATCH 4/11 ppc/pnv: use dev instead of pci->qdev in root_port_realize()

In reply to Daniel Henrique Barboza

We already have access to the 'dev' object.

---
 hw/pci-host/pnv_phb3.c | 4 ++--
 hw/pci-host/pnv_phb4.c | 5 ++---
 2 files changed, 4 insertions(+), 5 deletions(-)
hw/pci-host/pnv_phb3.c +2 -2
index 7901d8172c..bda23fd20b 100644
@@ -1157,8 +1157,8 @@ static void pnv_phb3_root_port_realize(DeviceState *dev, Error **errp)
     }
 
     /* Set unique chassis/slot values for the root port */
-    qdev_prop_set_uint8(&pci->qdev, "chassis", phb->chip_id);
-    qdev_prop_set_uint16(&pci->qdev, "slot", phb->phb_id);
+    qdev_prop_set_uint8(dev, "chassis", phb->chip_id);
+    qdev_prop_set_uint16(dev, "slot", phb->phb_id);
 
     rpc->parent_realize(dev, &local_err);
     if (local_err) {
hw/pci-host/pnv_phb4.c +2 -3
index bae9398d86..bfec8b9f6d 100644
@@ -1781,7 +1781,6 @@ static void pnv_phb4_root_port_reset(DeviceState *dev)
 static void pnv_phb4_root_port_realize(DeviceState *dev, Error **errp)
 {
     PCIERootPortClass *rpc = PCIE_ROOT_PORT_GET_CLASS(dev);
-    PCIDevice *pci = PCI_DEVICE(dev);
     PnvPHB4 *phb = NULL;
     Error *local_err = NULL;
 
@@ -1799,8 +1798,8 @@ static void pnv_phb4_root_port_realize(DeviceState *dev, Error **errp)
     }
 
     /* Set unique chassis/slot values for the root port */
-    qdev_prop_set_uint8(&pci->qdev, "chassis", phb->chip_id);
-    qdev_prop_set_uint16(&pci->qdev, "slot", phb->phb_id);
+    qdev_prop_set_uint8(dev, "chassis", phb->chip_id);
+    qdev_prop_set_uint16(dev, "slot", phb->phb_id);
 
     rpc->parent_realize(dev, &local_err);
     if (local_err) {
Signature · 1 line
2.36.1
#30 lore raw

PATCH 9/11 ppc/pnv: add PHB object/bus parenting helpers

In reply to Daniel Henrique Barboza

These helpers are inspired by the changes that were reverted in commit 9c10d86fee "ppc/pnv: Remove user-created PHB{3,4,5} devices". We'll use them to handle the default case we already support.

---
 hw/ppc/pnv.c         | 31 +++++++++++++++++++++++++++++++
 include/hw/ppc/pnv.h |  2 ++
 2 files changed, 33 insertions(+)
hw/ppc/pnv.c +31 -0
index 6ce9e94e05..d77c90d64a 100644
@@ -1983,6 +1983,37 @@ static ICSState *pnv_ics_get(XICSFabric *xi, int irq)
     return NULL;
 }
 
+
+/*
+ * Set the QOM parent of an object child. If the device state
+ * associated with the child has an id, use it as QOM id. Otherwise
+ * use object_typename[index] as QOM id.
+ */
+void pnv_parent_qom_fixup(Object *parent, Object *child, int index)
+{
+    g_autofree char *default_id =
+        g_strdup_printf("%s[%d]", object_get_typename(child), index);
+    const char *dev_id = DEVICE(child)->id;
+
+    if (child->parent == parent) {
+        return;
+    }
+
+    object_ref(child);
+    object_unparent(child);
+    object_property_add_child(parent, dev_id ? dev_id : default_id, child);
+    object_unref(child);
+}
+
+void pnv_parent_bus_fixup(DeviceState *parent, DeviceState *phb_dev)
+{
+    BusState *s = qdev_get_parent_bus(parent);
+
+    if (!qdev_set_parent_bus(phb_dev, s, &error_fatal)) {
+        return;
+    }
+}
+
 PnvChip *pnv_get_chip(PnvMachineState *pnv, uint32_t chip_id)
 {
     int i;
include/hw/ppc/pnv.h +2 -0
index 11f1089289..c091f23039 100644
@@ -194,6 +194,8 @@ DECLARE_INSTANCE_CHECKER(PnvChip, PNV_CHIP_POWER10,
 
 PowerPCCPU *pnv_chip_find_cpu(PnvChip *chip, uint32_t pir);
 void pnv_phb_attach_root_port(PCIHostState *pci, const char *name, int index);
+void pnv_parent_qom_fixup(Object *parent, Object *child, int index);
+void pnv_parent_bus_fixup(DeviceState *parent, DeviceState *child);
 
 #define TYPE_PNV_MACHINE       MACHINE_TYPE_NAME("powernv")
 typedef struct PnvMachineClass PnvMachineClass;
Signature · 1 line
2.36.1
#23 lore raw

PATCH 7/11 ppc/pnv: make pnv_chip_power8_pic_print_info() use chip8->phbs[]

In reply to Daniel Henrique Barboza

It's inneficient to scroll all child objects when we have all PHBs available in chip8->phbs[].

pnv_chip_power8_pic_print_info_child() ended up folded into pic_print_info() for simplicity.

---
 hw/ppc/pnv.c | 22 ++++++++--------------
 1 file changed, 8 insertions(+), 14 deletions(-)
hw/ppc/pnv.c +8 -14
index d70deffa1d..5e3323e950 100644
@@ -652,25 +652,19 @@ static ISABus *pnv_isa_create(PnvChip *chip, Error **errp)
     return PNV_CHIP_GET_CLASS(chip)->isa_create(chip, errp);
 }
 
-static int pnv_chip_power8_pic_print_info_child(Object *child, void *opaque)
-{
-    Monitor *mon = opaque;
-    PnvPHB3 *phb3 = (PnvPHB3 *) object_dynamic_cast(child, TYPE_PNV_PHB3);
-
-    if (phb3) {
-        pnv_phb3_msi_pic_print_info(&phb3->msis, mon);
-        ics_pic_print_info(&phb3->lsis, mon);
-    }
-    return 0;
-}
-
 static void pnv_chip_power8_pic_print_info(PnvChip *chip, Monitor *mon)
 {
     Pnv8Chip *chip8 = PNV8_CHIP(chip);
+    int i;
 
     ics_pic_print_info(&chip8->psi.ics, mon);
-    object_child_foreach(OBJECT(chip),
-                         pnv_chip_power8_pic_print_info_child, mon);
+
+    for (i = 0; i < chip8->num_phbs; i++) {
+        PnvPHB3 *phb3 = &chip8->phbs[i];
+
+        pnv_phb3_msi_pic_print_info(&phb3->msis, mon);
+        ics_pic_print_info(&phb3->lsis, mon);
+    }
 }
 
 static int pnv_chip_power9_pic_print_info_child(Object *child, void *opaque)
Signature · 1 line
2.36.1
#31 lore raw

PATCH 10/11 ppc/pnv: move PHB3 initialization to realize time

In reply to Daniel Henrique Barboza

There's nothing special that is being done in pnv_chip_power8_instance_init() that can't be done during pnv_chip_power8_realize(). Move the PHB creating and phbs[] assignment to power8_realize().

We also need to assign a proper phb->chip parent and bus. This is done by the PHB itself, in pnv_phb3_realize(), in a similar fashion that user created PHB3s are going to do.

After all this we're left with logic that, aside from phb chip assignment that are still being done in power8_realize(), behaves the same for default and user created PHB3s.

---
 hw/pci-host/pnv_phb3.c | 14 ++++++++++++++
 hw/ppc/pnv.c           | 24 +++++-------------------
 2 files changed, 19 insertions(+), 19 deletions(-)
hw/pci-host/pnv_phb3.c +14 -0
index bda23fd20b..c1c73fb88d 100644
@@ -998,6 +998,20 @@ static void pnv_phb3_realize(DeviceState *dev, Error **errp)
         return;
     }
 
+    /*
+     * We need the chip to parent the PHB to allow the DT
+     * to build correctly (via pnv_xscom_dt()).
+     *
+     * TODO: the PHB should be parented by a PHB3 PEC device.
+     */
+    pnv_parent_qom_fixup(OBJECT(phb->chip), OBJECT(phb), phb->phb_id);
+
+    /*
+     * pnv-phb3 buses are child of the main-system-bus, same as
+     * the chip.
+     */
+    pnv_parent_bus_fixup(DEVICE(phb->chip), dev);
+
     /* LSI sources */
     object_property_set_link(OBJECT(&phb->lsis), "xics", OBJECT(pnv),
                              &error_abort);
hw/ppc/pnv.c +5 -19
index d77c90d64a..e4080a98e1 100644
@@ -1130,8 +1130,6 @@ static void pnv_chip_power10_intc_print_info(PnvChip *chip, PowerPCCPU *cpu,
 static void pnv_chip_power8_instance_init(Object *obj)
 {
     Pnv8Chip *chip8 = PNV8_CHIP(obj);
-    PnvChipClass *pcc = PNV_CHIP_GET_CLASS(obj);
-    int i;
 
     object_property_add_link(obj, "xics", TYPE_XICS_FABRIC,
                              (Object **)&chip8->xics,
@@ -1145,22 +1143,6 @@ static void pnv_chip_power8_instance_init(Object *obj)
     object_initialize_child(obj, "occ", &chip8->occ, TYPE_PNV8_OCC);
 
     object_initialize_child(obj, "homer", &chip8->homer, TYPE_PNV8_HOMER);
-
-    chip8->num_phbs = pcc->num_phbs;
-
-    for (i = 0; i < chip8->num_phbs; i++) {
-        PnvPHB3 *phb3 = PNV_PHB3(object_new(TYPE_PNV_PHB3));
-
-        /*
-         * We need the chip to parent the PHB to allow the DT
-         * to build correctly (via pnv_xscom_dt()).
-         *
-         * TODO: the PHB should be parented by a PEC device.
-         */
-        object_property_add_child(obj, "phb[*]", OBJECT(phb3));
-        chip8->phbs[i] = phb3;
-    }
-
 }
 
 static void pnv_chip_icp_realize(Pnv8Chip *chip8, Error **errp)
@@ -1286,8 +1268,12 @@ static void pnv_chip_power8_realize(DeviceState *dev, Error **errp)
                                 &chip8->homer.regs);
 
     /* PHB3 controllers */
+    chip8->num_phbs = pcc->num_phbs;
+
     for (i = 0; i < chip8->num_phbs; i++) {
-        PnvPHB3 *phb = chip8->phbs[i];
+        PnvPHB3 *phb = PNV_PHB3(object_new(TYPE_PNV_PHB3));
+
+        chip8->phbs[i] = phb;
 
         object_property_set_int(OBJECT(phb), "index", i, &error_fatal);
         object_property_set_int(OBJECT(phb), "chip-id", chip->chip_id,
Signature · 1 line
2.36.1
#33 lore raw

PATCH 11/11 ppc/pnv: move PHB4 parent fixup to phb4_realize()

In reply to Daniel Henrique Barboza

After the code cleanups done in the past, default PHB4 code is already fairly close to user created PHB4s. What we need to do to make it equal is move the QOM and bus parenting changes from pnv_pec_default_phb_realize() to pnv_phb4_realize().

Using the same logic for both cases (aside from PEC assigning) will make our lives easier when re-enabling user created PHB4s.

---
 hw/pci-host/pnv_phb4.c     | 12 ++++++++++++
 hw/pci-host/pnv_phb4_pec.c |  1 -
 2 files changed, 12 insertions(+), 1 deletion(-)
hw/pci-host/pnv_phb4.c +12 -0
index bfec8b9f6d..fd6fac21ac 100644
@@ -1553,6 +1553,18 @@ static void pnv_phb4_realize(DeviceState *dev, Error **errp)
     int nr_irqs;
     char name[32];
 
+    /*
+     * We need the PEC to parent the PHB to allow the DT
+     * to build correctly (via pnv_xscom_dt()).
+     */
+    pnv_parent_qom_fixup(OBJECT(phb->pec), OBJECT(phb), phb->phb_id);
+
+    /*
+     * pnv-phb4 buses are child of the main-system-bus, same as
+     * the chip.
+     */
+    pnv_parent_bus_fixup(DEVICE(phb->pec->chip), dev);
+
     /* Set the "big_phb" flag */
     phb->big_phb = phb->phb_id == 0 || phb->phb_id == 3;
 
hw/pci-host/pnv_phb4_pec.c +0 -1
index c9aaf1c28e..eb47b50737 100644
@@ -119,7 +119,6 @@ static void pnv_pec_default_phb_realize(PnvPhb4PecState *pec,
     PnvPHB4 *phb = PNV_PHB4(qdev_new(pecc->phb_type));
     int phb_id = pnv_phb4_pec_get_phb_id(pec, stack_no);
 
-    object_property_add_child(OBJECT(pec), "phb[*]", OBJECT(phb));
     object_property_set_link(OBJECT(phb), "pec", OBJECT(pec),
                              &error_abort);
     object_property_set_int(OBJECT(phb), "chip-id", pec->chip_id,
Signature · 1 line
2.36.1
#17 lore raw

Re: [PATCH 05/11] ppc/pnv: make pnv_ics_get() use the chip8->phbs[] array

In reply to Daniel Henrique Barboza
Quoted 31 lines · Daniel Henrique Barboza

The function is working today by getting all the child objects of the chip, interacting with each of them to check whether the child is a PHB, and then doing what needs to be done.

We have all the chip PHBs in the phbs[] array so interacting with all child objects is unneeded.

---
  hw/ppc/pnv.c | 30 ++++++++++++++----------------
  1 file changed, 14 insertions(+), 16 deletions(-)
hw/ppc/pnv.c +4 -5
index 40e0cbd84d..05a8d5034f 100644
@@ -1944,41 +1944,39 @@ typedef struct ForeachPhb3Args {
      ICSState *ics;
  } ForeachPhb3Args;
 
-static int pnv_ics_get_child(Object *child, void *opaque)
+static void pnv_ics_get_phb_ics(PnvPHB3 *phb3, ForeachPhb3Args *args)
  {
-    ForeachPhb3Args *args = opaque;
-    PnvPHB3 *phb3 = (PnvPHB3 *) object_dynamic_cast(child, TYPE_PNV_PHB3);
+    if (ics_valid_irq(&phb3->lsis, args->irq)) {
+        args->ics = &phb3->lsis;
+    }
 
-    if (phb3) {
-        if (ics_valid_irq(&phb3->lsis, args->irq)) {

hw/ppc/pnv.c typedef struct ForeachPhb3Args { · line 1956

Quoted 10 lines
hw/ppc/pnv.c +2 -6
-            args->ics = &phb3->lsis;
-        }
-        if (ics_valid_irq(ICS(&phb3->msis), args->irq)) {
-            args->ics = ICS(&phb3->msis);
-        }
+    if (ics_valid_irq(ICS(&phb3->msis), args->irq)) {
+        args->ics = ICS(&phb3->msis);
      }
-    return args->ics ? 1 : 0;
  }

It seems that we could gain in readability by dropping the ForeachPhb3Args structure completely. The 'irq' member can just be an input argument to the function instead of the full structure. The 'ics' member is no longer needed, it can be the returned value of the function (instead of void)

Fred

Quoted 31 lines
hw/ppc/pnv.c +8 -5
  static ICSState *pnv_ics_get(XICSFabric *xi, int irq)
  {
      PnvMachineState *pnv = PNV_MACHINE(xi);
      ForeachPhb3Args args = { irq, NULL };
-    int i;
+    int i, j;
 
      for (i = 0; i < pnv->num_chips; i++) {
-        PnvChip *chip = pnv->chips[i];
          Pnv8Chip *chip8 = PNV8_CHIP(pnv->chips[i]);
 
          if (ics_valid_irq(&chip8->psi.ics, irq)) {
              return &chip8->psi.ics;
          }
 
-        object_child_foreach(OBJECT(chip), pnv_ics_get_child, &args);
-        if (args.ics) {
-            return args.ics;
+        for (j = 0; j < chip8->num_phbs; j++) {
+            pnv_ics_get_phb_ics(&chip8->phbs[j], &args);
+
+            if (args.ics) {
+                return args.ics;
+            }
          }
      }
+
      return NULL;
  }
 
Reviewed-by on v1 1/11
#2 lore raw

Re: [PATCH 01/11] ppc/pnv: move root port attach to pnv_phb4_realize()

In reply to Daniel Henrique Barboza
Quoted 5 lines · Daniel Henrique Barboza

Creating a root port is something related to the PHB, not the PEC. It also makes the logic more in line with what pnv-phb3 does.

---

LGTM,

Fred

Quoted 40 lines
  hw/pci-host/pnv_phb4.c     | 4 ++++
  hw/pci-host/pnv_phb4_pec.c | 3 ---
  2 files changed, 4 insertions(+), 3 deletions(-)
hw/pci-host/pnv_phb4.c +4 -0
index 6594016121..23ad8de7ee 100644
@@ -1547,6 +1547,7 @@ static void pnv_phb4_instance_init(Object *obj)
  static void pnv_phb4_realize(DeviceState *dev, Error **errp)
  {
      PnvPHB4 *phb = PNV_PHB4(dev);
+    PnvPhb4PecClass *pecc = PNV_PHB4_PEC_GET_CLASS(phb->pec);
      PCIHostState *pci = PCI_HOST_BRIDGE(dev);
      XiveSource *xsrc = &phb->xsrc;
      int nr_irqs;
@@ -1583,6 +1584,9 @@ static void pnv_phb4_realize(DeviceState *dev, Error **errp)
      pci_setup_iommu(pci->bus, pnv_phb4_dma_iommu, phb);
      pci->bus->flags |= PCI_BUS_EXTENDED_CONFIG_SPACE;
 
+    /* Add a single Root port if running with defaults */
+    pnv_phb_attach_root_port(pci, pecc->rp_model);
+
      /* Setup XIVE Source */
      if (phb->big_phb) {
          nr_irqs = PNV_PHB4_MAX_INTs;
hw/pci-host/pnv_phb4_pec.c +0 -3
index 8b7e823fa5..c9aaf1c28e 100644
@@ -130,9 +130,6 @@ static void pnv_pec_default_phb_realize(PnvPhb4PecState *pec,
      if (!sysbus_realize(SYS_BUS_DEVICE(phb), errp)) {
          return;
      }
-
-    /* Add a single Root port if running with defaults */
-    pnv_phb_attach_root_port(PCI_HOST_BRIDGE(phb), pecc->rp_model);
  }
 
  static void pnv_pec_realize(DeviceState *dev, Error **errp)
#10 lore raw

Re: [PATCH 03/11] ppc/pnv: use dev->parent_bus->parent to get the PHB

In reply to Daniel Henrique Barboza
Quoted 50 lines · Daniel Henrique Barboza

It is not advisable to execute an object_dynamic_cast() to poke into bus->qbus.parent and follow it up with a C cast into the PnvPHB type we think we got.

A better way is to access the PnvPHB object via a QOM macro accessing the existing parent links of the DeviceState. For a given pnv-phb3/4-root-port 'dev', dev->parent_bus will give us the PHB bus, and dev->parent_bus->parent is the PHB. Use the adequate QOM macro to assert the type, and keep the NULL check in case we didn't get the object we were expecting.

---
  hw/pci-host/pnv_phb3.c | 10 +++++++---
  hw/pci-host/pnv_phb4.c | 10 +++++++---
  2 files changed, 14 insertions(+), 6 deletions(-)
hw/pci-host/pnv_phb3.c +7 -3
index 4ba660f8b9..7901d8172c 100644
@@ -1139,12 +1139,16 @@ static void pnv_phb3_root_port_realize(DeviceState *dev, Error **errp)
  {
      PCIERootPortClass *rpc = PCIE_ROOT_PORT_GET_CLASS(dev);
      PCIDevice *pci = PCI_DEVICE(dev);
-    PCIBus *bus = pci_get_bus(pci);
      PnvPHB3 *phb = NULL;
      Error *local_err = NULL;
 
-    phb = (PnvPHB3 *) object_dynamic_cast(OBJECT(bus->qbus.parent),
-                                          TYPE_PNV_PHB3);
+    /*
+     * dev->parent_bus gives access to the pnv-phb-root bus.
+     * The PnvPHB3 is the owner (parent) of the bus.
+     */
+    if (dev && dev->parent_bus) {
+        phb = PNV_PHB3(dev->parent_bus->parent);
+    }
 
      if (!phb) {
          error_setg(errp,
hw/pci-host/pnv_phb4.c +0 -1
index ffd9d8a947..bae9398d86 100644
@@ -1782,12 +1782,16 @@ static void pnv_phb4_root_port_realize(DeviceState *dev, Error **errp)
  {
      PCIERootPortClass *rpc = PCIE_ROOT_PORT_GET_CLASS(dev);
      PCIDevice *pci = PCI_DEVICE(dev);
-    PCIBus *bus = pci_get_bus(pci);

hw/pci-host/pnv_phb4.c static void pnv_phb4_root_port_realize(DeviceState *dev, Error **errp) · line 1792

Quoted 10 lines
hw/pci-host/pnv_phb4.c +5 -2
      PnvPHB4 *phb = NULL;
      Error *local_err = NULL;
 
-    phb = (PnvPHB4 *) object_dynamic_cast(OBJECT(bus->qbus.parent),
-                                          TYPE_PNV_PHB4);
+    /*
+     * dev->parent_bus gives access to the pnv-phb-root bus.
+     * The PnvPHB4 is the owner (parent) of the bus.
+     */
+    if (dev && dev->parent_bus) {

Does it make sense to test 'dev' first when it's the device being realized?

Fred

Quoted 5 lines
hw/pci-host/pnv_phb4.c +2 -0
+        phb = PNV_PHB4(dev->parent_bus->parent);
+    }
 
      if (!phb) {
          error_setg(errp, "%s must be connected to pnv-phb4 buses", dev->id);
#20 lore raw

Re: [PATCH 06/11] ppc/pnv: make pnv_ics_resend() use chip8->phbs[]

In reply to Daniel Henrique Barboza
Quoted 32 lines · Daniel Henrique Barboza

pnv_ics_resend() is scrolling through all the child objects of the chip to search for the PHBs. It's faster and simpler to just use the phbs[] array.

pnv_ics_resend_child() was folded into pnv_ics_resend() since it's too simple to justify its own function.

---
  hw/ppc/pnv.c | 22 +++++++---------------
  1 file changed, 7 insertions(+), 15 deletions(-)
hw/ppc/pnv.c +0 -11
index 05a8d5034f..d70deffa1d 100644
@@ -1993,28 +1993,20 @@ PnvChip *pnv_get_chip(PnvMachineState *pnv, uint32_t chip_id)
      return NULL;
  }
 
-static int pnv_ics_resend_child(Object *child, void *opaque)
-{
-    PnvPHB3 *phb3 = (PnvPHB3 *) object_dynamic_cast(child, TYPE_PNV_PHB3);
-
-    if (phb3) {
-        ics_resend(&phb3->lsis);
-        ics_resend(ICS(&phb3->msis));
-    }
-    return 0;
-}
-
  static void pnv_ics_resend(XICSFabric *xi)

hw/ppc/pnv.c PnvChip *pnv_get_chip(PnvMachineState *pnv, uint32_t chip_id) · line 2016

Quoted 10 lines
hw/ppc/pnv.c +1 -3
  {
      PnvMachineState *pnv = PNV_MACHINE(xi);
-    int i;
+    int i, j;
 
      for (i = 0; i < pnv->num_chips; i++) {
-        PnvChip *chip = pnv->chips[i];
          Pnv8Chip *chip8 = PNV8_CHIP(pnv->chips[i]);
 
-        ics_resend(&chip8->psi.ics);

That line shouldn't be dropped, right?

Fred

Quoted 10 lines
hw/ppc/pnv.c +6 -1
-        object_child_foreach(OBJECT(chip), pnv_ics_resend_child, NULL);
+        for (j = 0; j < chip8->num_phbs; j++) {
+            PnvPHB3 *phb3 = &chip8->phbs[j];
+
+            ics_resend(&phb3->lsis);
+            ics_resend(ICS(&phb3->msis));
+        }
      }
  }
 
Reviewed-by on v1 2/11
#6 lore raw

Re: [PATCH 02/11] ppc/pnv: attach phb3/phb4 root ports in QOM tree

In reply to Daniel Henrique Barboza
Quoted 44 lines · Daniel Henrique Barboza

At this moment we leave the pnv-phb3(4)-root-port unattached in QOM:

/unattached (container)
(...)

     /device[2] (pnv-phb3-root-port)
       /bus master container[0] (memory-region)
       /bus master[0] (memory-region)
       /pci_bridge_io[0] (memory-region)
       /pci_bridge_io[1] (memory-region)
       /pci_bridge_mem[0] (memory-region)
       /pci_bridge_pci[0] (memory-region)
       /pci_bridge_pref_mem[0] (memory-region)
       /pci_bridge_vga_io_hi[0] (memory-region)
       /pci_bridge_vga_io_lo[0] (memory-region)
       /pci_bridge_vga_mem[0] (memory-region)
       /pcie.0 (PCIE)

Let's make changes in pnv_phb_attach_root_port() to attach the created root ports to its corresponding PHB.

This is the result afterwards:

     /pnv-phb3[0] (pnv-phb3)
       /lsi (ics)
       /msi (phb3-msi)
       /msi32[0] (memory-region)
       /msi64[0] (memory-region)
       /pbcq (pnv-pbcq)
     (...)
       /phb3_iommu[0] (pnv-phb3-iommu-memory-region)
       /pnv-phb3-root.0 (pnv-phb3-root)
         /pnv-phb3-root-port[0] (pnv-phb3-root-port)
           /bus master container[0] (memory-region)
           /bus master[0] (memory-region)
           /pci_bridge_io[0] (memory-region)
           /pci_bridge_io[1] (memory-region)
           /pci_bridge_mem[0] (memory-region)
           /pci_bridge_pci[0] (memory-region)
           /pci_bridge_pref_mem[0] (memory-region)
           /pci_bridge_vga_io_hi[0] (memory-region)
           /pci_bridge_vga_io_lo[0] (memory-region)
           /pci_bridge_vga_mem[0] (memory-region)
           /pcie.0 (PCIE)
Quoted 2 lines

---

I've always wondered if there was a good reason to have them detached. Glad to see there was none :-)

Fred

Quoted 65 lines
  hw/pci-host/pnv_phb3.c | 2 +-
  hw/pci-host/pnv_phb4.c | 2 +-
  hw/ppc/pnv.c           | 7 ++++++-
  include/hw/ppc/pnv.h   | 2 +-
  4 files changed, 9 insertions(+), 4 deletions(-)
hw/pci-host/pnv_phb3.c +1 -1
index 26ac9b7123..4ba660f8b9 100644
@@ -1052,7 +1052,7 @@ static void pnv_phb3_realize(DeviceState *dev, Error **errp)
 
      pci_setup_iommu(pci->bus, pnv_phb3_dma_iommu, phb);
 
-    pnv_phb_attach_root_port(PCI_HOST_BRIDGE(phb), TYPE_PNV_PHB3_ROOT_PORT);
+    pnv_phb_attach_root_port(pci, TYPE_PNV_PHB3_ROOT_PORT, phb->phb_id);
  }
 
  void pnv_phb3_update_regions(PnvPHB3 *phb)
hw/pci-host/pnv_phb4.c +1 -1
index 23ad8de7ee..ffd9d8a947 100644
@@ -1585,7 +1585,7 @@ static void pnv_phb4_realize(DeviceState *dev, Error **errp)
      pci->bus->flags |= PCI_BUS_EXTENDED_CONFIG_SPACE;
 
      /* Add a single Root port if running with defaults */
-    pnv_phb_attach_root_port(pci, pecc->rp_model);
+    pnv_phb_attach_root_port(pci, pecc->rp_model, phb->phb_id);
 
      /* Setup XIVE Source */
      if (phb->big_phb) {
hw/ppc/pnv.c +6 -1
index 7c08a78d6c..40e0cbd84d 100644
@@ -1190,9 +1190,14 @@ static void pnv_chip_icp_realize(Pnv8Chip *chip8, Error **errp)
  }
 
  /* Attach a root port device */
-void pnv_phb_attach_root_port(PCIHostState *pci, const char *name)
+void pnv_phb_attach_root_port(PCIHostState *pci, const char *name, int index)
  {
      PCIDevice *root = pci_new(PCI_DEVFN(0, 0), name);
+    g_autofree char *default_id = g_strdup_printf("%s[%d]", name, index);
+    const char *dev_id = DEVICE(root)->id;
+
+    object_property_add_child(OBJECT(pci->bus), dev_id ? dev_id : default_id,
+                              OBJECT(root));
 
      pci_realize_and_unref(root, pci->bus, &error_fatal);
  }
include/hw/ppc/pnv.h +1 -1
index 86cb7d7f97..033890a23f 100644
@@ -189,7 +189,7 @@ DECLARE_INSTANCE_CHECKER(PnvChip, PNV_CHIP_POWER10,
                           TYPE_PNV_CHIP_POWER10)
 
  PowerPCCPU *pnv_chip_find_cpu(PnvChip *chip, uint32_t pir);
-void pnv_phb_attach_root_port(PCIHostState *pci, const char *name);
+void pnv_phb_attach_root_port(PCIHostState *pci, const char *name, int index);
 
  #define TYPE_PNV_MACHINE       MACHINE_TYPE_NAME("powernv")
  typedef struct PnvMachineClass PnvMachineClass;
Reviewed-by on v1 7/11
#24 lore raw

Re: [PATCH 07/11] ppc/pnv: make pnv_chip_power8_pic_print_info() use chip8->phbs[]

In reply to Daniel Henrique Barboza
Quoted 8 lines · Daniel Henrique Barboza

It's inneficient to scroll all child objects when we have all PHBs available in chip8->phbs[].

pnv_chip_power8_pic_print_info_child() ended up folded into pic_print_info() for simplicity.

---

Fred

Quoted 41 lines
  hw/ppc/pnv.c | 22 ++++++++--------------
  1 file changed, 8 insertions(+), 14 deletions(-)
hw/ppc/pnv.c +8 -14
index d70deffa1d..5e3323e950 100644
@@ -652,25 +652,19 @@ static ISABus *pnv_isa_create(PnvChip *chip, Error **errp)
      return PNV_CHIP_GET_CLASS(chip)->isa_create(chip, errp);
  }
 
-static int pnv_chip_power8_pic_print_info_child(Object *child, void *opaque)
-{
-    Monitor *mon = opaque;
-    PnvPHB3 *phb3 = (PnvPHB3 *) object_dynamic_cast(child, TYPE_PNV_PHB3);
-
-    if (phb3) {
-        pnv_phb3_msi_pic_print_info(&phb3->msis, mon);
-        ics_pic_print_info(&phb3->lsis, mon);
-    }
-    return 0;
-}
-
  static void pnv_chip_power8_pic_print_info(PnvChip *chip, Monitor *mon)
  {
      Pnv8Chip *chip8 = PNV8_CHIP(chip);
+    int i;
 
      ics_pic_print_info(&chip8->psi.ics, mon);
-    object_child_foreach(OBJECT(chip),
-                         pnv_chip_power8_pic_print_info_child, mon);
+
+    for (i = 0; i < chip8->num_phbs; i++) {
+        PnvPHB3 *phb3 = &chip8->phbs[i];
+
+        pnv_phb3_msi_pic_print_info(&phb3->msis, mon);
+        ics_pic_print_info(&phb3->lsis, mon);
+    }
  }
 
  static int pnv_chip_power9_pic_print_info_child(Object *child, void *opaque)
Reviewed-by on v1 4/11
#15 lore raw

Re: [PATCH 04/11] ppc/pnv: use dev instead of pci->qdev in root_port_realize()

In reply to Daniel Henrique Barboza
Quoted 4 lines · Daniel Henrique Barboza

We already have access to the 'dev' object.

---

Fred

Quoted 42 lines
  hw/pci-host/pnv_phb3.c | 4 ++--
  hw/pci-host/pnv_phb4.c | 5 ++---
  2 files changed, 4 insertions(+), 5 deletions(-)
hw/pci-host/pnv_phb3.c +2 -2
index 7901d8172c..bda23fd20b 100644
@@ -1157,8 +1157,8 @@ static void pnv_phb3_root_port_realize(DeviceState *dev, Error **errp)
      }
 
      /* Set unique chassis/slot values for the root port */
-    qdev_prop_set_uint8(&pci->qdev, "chassis", phb->chip_id);
-    qdev_prop_set_uint16(&pci->qdev, "slot", phb->phb_id);
+    qdev_prop_set_uint8(dev, "chassis", phb->chip_id);
+    qdev_prop_set_uint16(dev, "slot", phb->phb_id);
 
      rpc->parent_realize(dev, &local_err);
      if (local_err) {
hw/pci-host/pnv_phb4.c +2 -3
index bae9398d86..bfec8b9f6d 100644
@@ -1781,7 +1781,6 @@ static void pnv_phb4_root_port_reset(DeviceState *dev)
  static void pnv_phb4_root_port_realize(DeviceState *dev, Error **errp)
  {
      PCIERootPortClass *rpc = PCIE_ROOT_PORT_GET_CLASS(dev);
-    PCIDevice *pci = PCI_DEVICE(dev);
      PnvPHB4 *phb = NULL;
      Error *local_err = NULL;
 
@@ -1799,8 +1798,8 @@ static void pnv_phb4_root_port_realize(DeviceState *dev, Error **errp)
      }
 
      /* Set unique chassis/slot values for the root port */
-    qdev_prop_set_uint8(&pci->qdev, "chassis", phb->chip_id);
-    qdev_prop_set_uint16(&pci->qdev, "slot", phb->phb_id);
+    qdev_prop_set_uint8(dev, "chassis", phb->chip_id);
+    qdev_prop_set_uint16(dev, "slot", phb->phb_id);
 
      rpc->parent_realize(dev, &local_err);
      if (local_err) {
#18 lore raw

Re: [PATCH 05/11] ppc/pnv: make pnv_ics_get() use the chip8->phbs[] array

In reply to Daniel Henrique Barboza
Quoted 52 lines · Daniel Henrique Barboza

The function is working today by getting all the child objects of the chip, interacting with each of them to check whether the child is a PHB, and then doing what needs to be done.

We have all the chip PHBs in the phbs[] array so interacting with all child objects is unneeded.

---
  hw/ppc/pnv.c | 30 ++++++++++++++----------------
  1 file changed, 14 insertions(+), 16 deletions(-)
hw/ppc/pnv.c +7 -13
index 40e0cbd84d..05a8d5034f 100644
@@ -1944,41 +1944,39 @@ typedef struct ForeachPhb3Args {
      ICSState *ics;
  } ForeachPhb3Args;
 
-static int pnv_ics_get_child(Object *child, void *opaque)
+static void pnv_ics_get_phb_ics(PnvPHB3 *phb3, ForeachPhb3Args *args)
  {
-    ForeachPhb3Args *args = opaque;
-    PnvPHB3 *phb3 = (PnvPHB3 *) object_dynamic_cast(child, TYPE_PNV_PHB3);
+    if (ics_valid_irq(&phb3->lsis, args->irq)) {
+        args->ics = &phb3->lsis;
+    }
 
-    if (phb3) {
-        if (ics_valid_irq(&phb3->lsis, args->irq)) {
-            args->ics = &phb3->lsis;
-        }
-        if (ics_valid_irq(ICS(&phb3->msis), args->irq)) {
-            args->ics = ICS(&phb3->msis);
-        }
+    if (ics_valid_irq(ICS(&phb3->msis), args->irq)) {
+        args->ics = ICS(&phb3->msis);
      }
-    return args->ics ? 1 : 0;
  }
 
  static ICSState *pnv_ics_get(XICSFabric *xi, int irq)
  {
      PnvMachineState *pnv = PNV_MACHINE(xi);
      ForeachPhb3Args args = { irq, NULL };
-    int i;
+    int i, j;
 
      for (i = 0; i < pnv->num_chips; i++) {
-        PnvChip *chip = pnv->chips[i];
          Pnv8Chip *chip8 = PNV8_CHIP(pnv->chips[i]);

hw/ppc/pnv.c typedef struct ForeachPhb3Args { · line 1972

Quoted 10 lines
hw/ppc/pnv.c +2 -3
 
          if (ics_valid_irq(&chip8->psi.ics, irq)) {
              return &chip8->psi.ics;
          }
 
-        object_child_foreach(OBJECT(chip), pnv_ics_get_child, &args);
-        if (args.ics) {
-            return args.ics;
+        for (j = 0; j < chip8->num_phbs; j++) {
+            pnv_ics_get_phb_ics(&chip8->phbs[j], &args);

If we don't need this function elsewhere, why keep it ?

Thanks,

C.

Quoted 10 lines
hw/ppc/pnv.c +5 -0
+
+            if (args.ics) {
+                return args.ics;
+            }
          }
      }
+
      return NULL;
  }
 
#7 lore raw

Re: [PATCH 02/11] ppc/pnv: attach phb3/phb4 root ports in QOM tree

In reply to Frederic Barrat
Quoted 47 lines · Frederic Barrat
Quoted 44 lines · Daniel Henrique Barboza

At this moment we leave the pnv-phb3(4)-root-port unattached in QOM:

   /unattached (container)
(...)
     /device[2] (pnv-phb3-root-port)
       /bus master container[0] (memory-region)
       /bus master[0] (memory-region)
       /pci_bridge_io[0] (memory-region)
       /pci_bridge_io[1] (memory-region)
       /pci_bridge_mem[0] (memory-region)
       /pci_bridge_pci[0] (memory-region)
       /pci_bridge_pref_mem[0] (memory-region)
       /pci_bridge_vga_io_hi[0] (memory-region)
       /pci_bridge_vga_io_lo[0] (memory-region)
       /pci_bridge_vga_mem[0] (memory-region)
       /pcie.0 (PCIE)

Let's make changes in pnv_phb_attach_root_port() to attach the created root ports to its corresponding PHB.

This is the result afterwards:

     /pnv-phb3[0] (pnv-phb3)
       /lsi (ics)
       /msi (phb3-msi)
       /msi32[0] (memory-region)
       /msi64[0] (memory-region)
       /pbcq (pnv-pbcq)
     (...)
       /phb3_iommu[0] (pnv-phb3-iommu-memory-region) /pnv-phb3-root.0 (pnv-phb3-root) /pnv-phb3-root-port[0] (pnv-phb3-root-port) /bus master container[0] (memory-region) /bus master[0] (memory-region) /pci_bridge_io[0] (memory-region) /pci_bridge_io[1] (memory-region) /pci_bridge_mem[0] (memory-region) /pci_bridge_pci[0] (memory-region) /pci_bridge_pref_mem[0] (memory-region) /pci_bridge_vga_io_hi[0] (memory-region) /pci_bridge_vga_io_lo[0] (memory-region) /pci_bridge_vga_mem[0] (memory-region) /pcie.0 (PCIE)

Quoted 5 lines
Quoted 2 lines

---

I've always wondered if there was a good reason to have them detached. Glad to see there was none :-)

Wasn't it for libvirt integration ?

C.

Quoted 62 lines

  Fred

Quoted 56 lines

  hw/pci-host/pnv_phb3.c | 2 +-
  hw/pci-host/pnv_phb4.c | 2 +-
  hw/ppc/pnv.c           | 7 ++++++-
  include/hw/ppc/pnv.h   | 2 +-

  4 files changed, 9 insertions(+), 4 deletions(-)
hw/pci-host/pnv_phb3.c +1 -1
index 26ac9b7123..4ba660f8b9 100644
@@ -1052,7 +1052,7 @@ static void pnv_phb3_realize(DeviceState *dev, Error **errp)
       pci_setup_iommu(pci->bus, pnv_phb3_dma_iommu, phb);
-    pnv_phb_attach_root_port(PCI_HOST_BRIDGE(phb), TYPE_PNV_PHB3_ROOT_PORT);
+    pnv_phb_attach_root_port(pci, TYPE_PNV_PHB3_ROOT_PORT, phb->phb_id);
   }
   void pnv_phb3_update_regions(PnvPHB3 *phb)
hw/pci-host/pnv_phb4.c +1 -1
index 23ad8de7ee..ffd9d8a947 100644
@@ -1585,7 +1585,7 @@ static void pnv_phb4_realize(DeviceState *dev, Error **errp)
       pci->bus->flags |= PCI_BUS_EXTENDED_CONFIG_SPACE;
       /* Add a single Root port if running with defaults */
-    pnv_phb_attach_root_port(pci, pecc->rp_model);
+    pnv_phb_attach_root_port(pci, pecc->rp_model, phb->phb_id);
       /* Setup XIVE Source */
       if (phb->big_phb) {
hw/ppc/pnv.c +6 -1
index 7c08a78d6c..40e0cbd84d 100644
@@ -1190,9 +1190,14 @@ static void pnv_chip_icp_realize(Pnv8Chip *chip8, Error **errp)
   }
   /* Attach a root port device */
-void pnv_phb_attach_root_port(PCIHostState *pci, const char *name)
+void pnv_phb_attach_root_port(PCIHostState *pci, const char *name, int index)
   {
       PCIDevice *root = pci_new(PCI_DEVFN(0, 0), name);
+    g_autofree char *default_id = g_strdup_printf("%s[%d]", name, index);
+    const char *dev_id = DEVICE(root)->id;
+
+    object_property_add_child(OBJECT(pci->bus), dev_id ? dev_id : default_id,
+                              OBJECT(root));
       pci_realize_and_unref(root, pci->bus, &error_fatal);
   }
include/hw/ppc/pnv.h +1 -1
index 86cb7d7f97..033890a23f 100644
@@ -189,7 +189,7 @@ DECLARE_INSTANCE_CHECKER(PnvChip, PNV_CHIP_POWER10,
                            TYPE_PNV_CHIP_POWER10)
   PowerPCCPU *pnv_chip_find_cpu(PnvChip *chip, uint32_t pir);
-void pnv_phb_attach_root_port(PCIHostState *pci, const char *name);
+void pnv_phb_attach_root_port(PCIHostState *pci, const char *name, int index);
   #define TYPE_PNV_MACHINE       MACHINE_TYPE_NAME("powernv")
   typedef struct PnvMachineClass PnvMachineClass;
Reviewed-by on v1 8/11
#26 lore raw

Re: [PATCH 08/11] ppc/pnv: turn chip8->phbs[] into a PnvPHB3* array

In reply to Daniel Henrique Barboza
Quoted 11 lines · Daniel Henrique Barboza

When enabling user created PHBs (a change reverted by commit 9c10d86fee) we were handling PHBs created by default versus by the user in different manners. The only difference between these PHBs is that one will have a valid phb3->chip that is assigned during pnv_chip_power8_realize(), while the user created needs to search which chip it belongs to.

Aside from that there shouldn't be any difference. Making the default PHBs behave in line with the user created ones will make it easier to re-introduce them later on. It will also make the code easier to follow since we are dealing with them in equal manner.

Quoted 8 lines

The first step is to turn chip8->phbs[] into a PnvPHB3 pointer array. This will allow us to assign user created PHBs into it later on. The way we initilize the default case is now more in line with that would happen with the user created case: the object is created, parented by the chip because pnv_xscom_dt() relies on it, and then assigned to the array.

---

This patch is more prep work for the user-created device instead of general cleanup like the previous ones, but I don't see anything wrong with it. So:

Fred

Quoted 79 lines
  hw/ppc/pnv.c         | 19 ++++++++++++++-----
  include/hw/ppc/pnv.h |  6 +++++-
  2 files changed, 19 insertions(+), 6 deletions(-)
hw/ppc/pnv.c +14 -5
index 5e3323e950..6ce9e94e05 100644
@@ -660,7 +660,7 @@ static void pnv_chip_power8_pic_print_info(PnvChip *chip, Monitor *mon)
      ics_pic_print_info(&chip8->psi.ics, mon);
 
      for (i = 0; i < chip8->num_phbs; i++) {
-        PnvPHB3 *phb3 = &chip8->phbs[i];
+        PnvPHB3 *phb3 = chip8->phbs[i];
 
          pnv_phb3_msi_pic_print_info(&phb3->msis, mon);
          ics_pic_print_info(&phb3->lsis, mon);
@@ -1149,7 +1149,16 @@ static void pnv_chip_power8_instance_init(Object *obj)
      chip8->num_phbs = pcc->num_phbs;
 
      for (i = 0; i < chip8->num_phbs; i++) {
-        object_initialize_child(obj, "phb[*]", &chip8->phbs[i], TYPE_PNV_PHB3);
+        PnvPHB3 *phb3 = PNV_PHB3(object_new(TYPE_PNV_PHB3));
+
+        /*
+         * We need the chip to parent the PHB to allow the DT
+         * to build correctly (via pnv_xscom_dt()).
+         *
+         * TODO: the PHB should be parented by a PEC device.
+         */
+        object_property_add_child(obj, "phb[*]", OBJECT(phb3));
+        chip8->phbs[i] = phb3;
      }
 
  }
@@ -1278,7 +1287,7 @@ static void pnv_chip_power8_realize(DeviceState *dev, Error **errp)
 
      /* PHB3 controllers */
      for (i = 0; i < chip8->num_phbs; i++) {
-        PnvPHB3 *phb = &chip8->phbs[i];
+        PnvPHB3 *phb = chip8->phbs[i];
 
          object_property_set_int(OBJECT(phb), "index", i, &error_fatal);
          object_property_set_int(OBJECT(phb), "chip-id", chip->chip_id,
@@ -1963,7 +1972,7 @@ static ICSState *pnv_ics_get(XICSFabric *xi, int irq)
          }
 
          for (j = 0; j < chip8->num_phbs; j++) {
-            pnv_ics_get_phb_ics(&chip8->phbs[j], &args);
+            pnv_ics_get_phb_ics(chip8->phbs[j], &args);
 
              if (args.ics) {
                  return args.ics;
@@ -1996,7 +2005,7 @@ static void pnv_ics_resend(XICSFabric *xi)
          Pnv8Chip *chip8 = PNV8_CHIP(pnv->chips[i]);
 
          for (j = 0; j < chip8->num_phbs; j++) {
-            PnvPHB3 *phb3 = &chip8->phbs[j];
+            PnvPHB3 *phb3 = chip8->phbs[j];
 
              ics_resend(&phb3->lsis);
              ics_resend(ICS(&phb3->msis));
include/hw/ppc/pnv.h +5 -1
index 033890a23f..11f1089289 100644
@@ -80,7 +80,11 @@ struct Pnv8Chip {
      PnvHomer     homer;
 
  #define PNV8_CHIP_PHB3_MAX 4
-    PnvPHB3      phbs[PNV8_CHIP_PHB3_MAX];
+    /*
+     * The array is used to allow quick access to the phbs by
+     * pnv_ics_get_child() and pnv_ics_resend_child().
+     */
+    PnvPHB3      *phbs[PNV8_CHIP_PHB3_MAX];
      uint32_t     num_phbs;
 
      XICSFabric    *xics;
#21 lore raw

Re: [PATCH 06/11] ppc/pnv: make pnv_ics_resend() use chip8->phbs[]

In reply to Frederic Barrat
Quoted 37 lines · Frederic Barrat
Quoted 34 lines · Daniel Henrique Barboza

pnv_ics_resend() is scrolling through all the child objects of the chip to search for the PHBs. It's faster and simpler to just use the phbs[] array.

pnv_ics_resend_child() was folded into pnv_ics_resend() since it's too simple to justify its own function.

---
  hw/ppc/pnv.c | 22 +++++++---------------

  1 file changed, 7 insertions(+), 15 deletions(-)
hw/ppc/pnv.c +0 -12
index 05a8d5034f..d70deffa1d 100644
@@ -1993,28 +1993,20 @@ PnvChip *pnv_get_chip(PnvMachineState *pnv, uint32_t chip_id)
       return NULL;
   }
-static int pnv_ics_resend_child(Object *child, void *opaque)
-{
-    PnvPHB3 *phb3 = (PnvPHB3 *) object_dynamic_cast(child, TYPE_PNV_PHB3);
-
-    if (phb3) {
-        ics_resend(&phb3->lsis);
-        ics_resend(ICS(&phb3->msis));
-    }
-    return 0;
-}
-
   static void pnv_ics_resend(XICSFabric *xi)
   {
       PnvMachineState *pnv = PNV_MACHINE(xi);
-    int i;
Quoted 8 lines

hw/ppc/pnv.c PnvChip *pnv_get_chip(PnvMachineState *pnv, uint32_t chip_id) · line 2016

Quoted 5 lines
hw/ppc/pnv.c +1 -2
+    int i, j;
       for (i = 0; i < pnv->num_chips; i++) {
-        PnvChip *chip = pnv->chips[i];
           Pnv8Chip *chip8 = PNV8_CHIP(pnv->chips[i]);
-        ics_resend(&chip8->psi.ics);

That line shouldn't be dropped, right?

yes. that's a typo, which should break the console.

C.

Quoted 13 lines

  Fred

Quoted 9 lines
hw/ppc/pnv.c +6 -1
-        object_child_foreach(OBJECT(chip), pnv_ics_resend_child, NULL);
+        for (j = 0; j < chip8->num_phbs; j++) {
+            PnvPHB3 *phb3 = &chip8->phbs[j];
+
+            ics_resend(&phb3->lsis);
+            ics_resend(ICS(&phb3->msis));
+        }
       }
   }
#32 lore raw

Re: [PATCH 10/11] ppc/pnv: move PHB3 initialization to realize time

In reply to Daniel Henrique Barboza
Quoted 25 lines · Daniel Henrique Barboza

There's nothing special that is being done in pnv_chip_power8_instance_init() that can't be done during pnv_chip_power8_realize(). Move the PHB creating and phbs[] assignment to power8_realize().

We also need to assign a proper phb->chip parent and bus. This is done by the PHB itself, in pnv_phb3_realize(), in a similar fashion that user created PHB3s are going to do.

After all this we're left with logic that, aside from phb chip assignment that are still being done in power8_realize(), behaves the same for default and user created PHB3s.

---
  hw/pci-host/pnv_phb3.c | 14 ++++++++++++++
  hw/ppc/pnv.c           | 24 +++++-------------------
  2 files changed, 19 insertions(+), 19 deletions(-)
hw/pci-host/pnv_phb3.c +0 -0
index bda23fd20b..c1c73fb88d 100644
@@ -998,6 +998,20 @@ static void pnv_phb3_realize(DeviceState *dev, Error **errp)
          return;

hw/pci-host/pnv_phb3.c static void pnv_phb3_realize(DeviceState *dev, Error **errp) · line 1008

Quoted 10 lines
hw/pci-host/pnv_phb3.c +8 -0
      }
 
+    /*
+     * We need the chip to parent the PHB to allow the DT
+     * to build correctly (via pnv_xscom_dt()).
+     *
+     * TODO: the PHB should be parented by a PHB3 PEC device.
+     */
+    pnv_parent_qom_fixup(OBJECT(phb->chip), OBJECT(phb), phb->phb_id);
+

Wouldn't we get the same result in a cleaner way by adding the phb as a child of the chip in pnv_chip_power8_realize() ? Right next to when the PnvPHB3 object pointer is added to the chip8->phbs array

Fred

Quoted 59 lines
hw/pci-host/pnv_phb3.c +6 -0
+    /*
+     * pnv-phb3 buses are child of the main-system-bus, same as
+     * the chip.
+     */
+    pnv_parent_bus_fixup(DEVICE(phb->chip), dev);
+
      /* LSI sources */
      object_property_set_link(OBJECT(&phb->lsis), "xics", OBJECT(pnv),
                               &error_abort);
hw/ppc/pnv.c +5 -19
index d77c90d64a..e4080a98e1 100644
@@ -1130,8 +1130,6 @@ static void pnv_chip_power10_intc_print_info(PnvChip *chip, PowerPCCPU *cpu,
  static void pnv_chip_power8_instance_init(Object *obj)
  {
      Pnv8Chip *chip8 = PNV8_CHIP(obj);
-    PnvChipClass *pcc = PNV_CHIP_GET_CLASS(obj);
-    int i;
 
      object_property_add_link(obj, "xics", TYPE_XICS_FABRIC,
                               (Object **)&chip8->xics,
@@ -1145,22 +1143,6 @@ static void pnv_chip_power8_instance_init(Object *obj)
      object_initialize_child(obj, "occ", &chip8->occ, TYPE_PNV8_OCC);
 
      object_initialize_child(obj, "homer", &chip8->homer, TYPE_PNV8_HOMER);
-
-    chip8->num_phbs = pcc->num_phbs;
-
-    for (i = 0; i < chip8->num_phbs; i++) {
-        PnvPHB3 *phb3 = PNV_PHB3(object_new(TYPE_PNV_PHB3));
-
-        /*
-         * We need the chip to parent the PHB to allow the DT
-         * to build correctly (via pnv_xscom_dt()).
-         *
-         * TODO: the PHB should be parented by a PEC device.
-         */
-        object_property_add_child(obj, "phb[*]", OBJECT(phb3));
-        chip8->phbs[i] = phb3;
-    }
-
  }
 
  static void pnv_chip_icp_realize(Pnv8Chip *chip8, Error **errp)
@@ -1286,8 +1268,12 @@ static void pnv_chip_power8_realize(DeviceState *dev, Error **errp)
                                  &chip8->homer.regs);
 
      /* PHB3 controllers */
+    chip8->num_phbs = pcc->num_phbs;
+
      for (i = 0; i < chip8->num_phbs; i++) {
-        PnvPHB3 *phb = chip8->phbs[i];
+        PnvPHB3 *phb = PNV_PHB3(object_new(TYPE_PNV_PHB3));
+
+        chip8->phbs[i] = phb;
 
          object_property_set_int(OBJECT(phb), "index", i, &error_fatal);
          object_property_set_int(OBJECT(phb), "chip-id", chip->chip_id,
Reviewed-by on v1 1/11
#3 lore raw

Re: [PATCH 01/11] ppc/pnv: move root port attach to pnv_phb4_realize()

In reply to Daniel Henrique Barboza
Quoted 4 lines · Daniel Henrique Barboza

Creating a root port is something related to the PHB, not the PEC. It also makes the logic more in line with what pnv-phb3 does.

So the root port is back where it was.

Could we avoid the pci_new() and use object_initialize_child() instead ?

Thanks,

C.

Quoted 41 lines
---
  hw/pci-host/pnv_phb4.c     | 4 ++++
  hw/pci-host/pnv_phb4_pec.c | 3 ---
  2 files changed, 4 insertions(+), 3 deletions(-)
hw/pci-host/pnv_phb4.c +4 -0
index 6594016121..23ad8de7ee 100644
@@ -1547,6 +1547,7 @@ static void pnv_phb4_instance_init(Object *obj)
  static void pnv_phb4_realize(DeviceState *dev, Error **errp)
  {
      PnvPHB4 *phb = PNV_PHB4(dev);
+    PnvPhb4PecClass *pecc = PNV_PHB4_PEC_GET_CLASS(phb->pec);
      PCIHostState *pci = PCI_HOST_BRIDGE(dev);
      XiveSource *xsrc = &phb->xsrc;
      int nr_irqs;
@@ -1583,6 +1584,9 @@ static void pnv_phb4_realize(DeviceState *dev, Error **errp)
      pci_setup_iommu(pci->bus, pnv_phb4_dma_iommu, phb);
      pci->bus->flags |= PCI_BUS_EXTENDED_CONFIG_SPACE;
 
+    /* Add a single Root port if running with defaults */
+    pnv_phb_attach_root_port(pci, pecc->rp_model);
+
      /* Setup XIVE Source */
      if (phb->big_phb) {
          nr_irqs = PNV_PHB4_MAX_INTs;
hw/pci-host/pnv_phb4_pec.c +0 -3
index 8b7e823fa5..c9aaf1c28e 100644
@@ -130,9 +130,6 @@ static void pnv_pec_default_phb_realize(PnvPhb4PecState *pec,
      if (!sysbus_realize(SYS_BUS_DEVICE(phb), errp)) {
          return;
      }
-
-    /* Add a single Root port if running with defaults */
-    pnv_phb_attach_root_port(PCI_HOST_BRIDGE(phb), pecc->rp_model);
  }
 
  static void pnv_pec_realize(DeviceState *dev, Error **errp)
Reviewed-by on v1 2/11
#8 lore raw

Re: [PATCH 02/11] ppc/pnv: attach phb3/phb4 root ports in QOM tree

In reply to Daniel Henrique Barboza
Quoted 44 lines · Daniel Henrique Barboza

At this moment we leave the pnv-phb3(4)-root-port unattached in QOM:

/unattached (container)
(...)

     /device[2] (pnv-phb3-root-port)
       /bus master container[0] (memory-region)
       /bus master[0] (memory-region)
       /pci_bridge_io[0] (memory-region)
       /pci_bridge_io[1] (memory-region)
       /pci_bridge_mem[0] (memory-region)
       /pci_bridge_pci[0] (memory-region)
       /pci_bridge_pref_mem[0] (memory-region)
       /pci_bridge_vga_io_hi[0] (memory-region)
       /pci_bridge_vga_io_lo[0] (memory-region)
       /pci_bridge_vga_mem[0] (memory-region)
       /pcie.0 (PCIE)

Let's make changes in pnv_phb_attach_root_port() to attach the created root ports to its corresponding PHB.

This is the result afterwards:

     /pnv-phb3[0] (pnv-phb3)
       /lsi (ics)
       /msi (phb3-msi)
       /msi32[0] (memory-region)
       /msi64[0] (memory-region)
       /pbcq (pnv-pbcq)
     (...)
       /phb3_iommu[0] (pnv-phb3-iommu-memory-region)
       /pnv-phb3-root.0 (pnv-phb3-root)
         /pnv-phb3-root-port[0] (pnv-phb3-root-port)
           /bus master container[0] (memory-region)
           /bus master[0] (memory-region)
           /pci_bridge_io[0] (memory-region)
           /pci_bridge_io[1] (memory-region)
           /pci_bridge_mem[0] (memory-region)
           /pci_bridge_pci[0] (memory-region)
           /pci_bridge_pref_mem[0] (memory-region)
           /pci_bridge_vga_io_hi[0] (memory-region)
           /pci_bridge_vga_io_lo[0] (memory-region)
           /pci_bridge_vga_mem[0] (memory-region)
           /pcie.0 (PCIE)
Quoted 1 line

Thanks,

C.

Quoted 66 lines
---
  hw/pci-host/pnv_phb3.c | 2 +-
  hw/pci-host/pnv_phb4.c | 2 +-
  hw/ppc/pnv.c           | 7 ++++++-
  include/hw/ppc/pnv.h   | 2 +-
  4 files changed, 9 insertions(+), 4 deletions(-)
hw/pci-host/pnv_phb3.c +1 -1
index 26ac9b7123..4ba660f8b9 100644
@@ -1052,7 +1052,7 @@ static void pnv_phb3_realize(DeviceState *dev, Error **errp)
 
      pci_setup_iommu(pci->bus, pnv_phb3_dma_iommu, phb);
 
-    pnv_phb_attach_root_port(PCI_HOST_BRIDGE(phb), TYPE_PNV_PHB3_ROOT_PORT);
+    pnv_phb_attach_root_port(pci, TYPE_PNV_PHB3_ROOT_PORT, phb->phb_id);
  }
 
  void pnv_phb3_update_regions(PnvPHB3 *phb)
hw/pci-host/pnv_phb4.c +1 -1
index 23ad8de7ee..ffd9d8a947 100644
@@ -1585,7 +1585,7 @@ static void pnv_phb4_realize(DeviceState *dev, Error **errp)
      pci->bus->flags |= PCI_BUS_EXTENDED_CONFIG_SPACE;
 
      /* Add a single Root port if running with defaults */
-    pnv_phb_attach_root_port(pci, pecc->rp_model);
+    pnv_phb_attach_root_port(pci, pecc->rp_model, phb->phb_id);
 
      /* Setup XIVE Source */
      if (phb->big_phb) {
hw/ppc/pnv.c +6 -1
index 7c08a78d6c..40e0cbd84d 100644
@@ -1190,9 +1190,14 @@ static void pnv_chip_icp_realize(Pnv8Chip *chip8, Error **errp)
  }
 
  /* Attach a root port device */
-void pnv_phb_attach_root_port(PCIHostState *pci, const char *name)
+void pnv_phb_attach_root_port(PCIHostState *pci, const char *name, int index)
  {
      PCIDevice *root = pci_new(PCI_DEVFN(0, 0), name);
+    g_autofree char *default_id = g_strdup_printf("%s[%d]", name, index);
+    const char *dev_id = DEVICE(root)->id;
+
+    object_property_add_child(OBJECT(pci->bus), dev_id ? dev_id : default_id,
+                              OBJECT(root));
 
      pci_realize_and_unref(root, pci->bus, &error_fatal);
  }
include/hw/ppc/pnv.h +1 -1
index 86cb7d7f97..033890a23f 100644
@@ -189,7 +189,7 @@ DECLARE_INSTANCE_CHECKER(PnvChip, PNV_CHIP_POWER10,
                           TYPE_PNV_CHIP_POWER10)
 
  PowerPCCPU *pnv_chip_find_cpu(PnvChip *chip, uint32_t pir);
-void pnv_phb_attach_root_port(PCIHostState *pci, const char *name);
+void pnv_phb_attach_root_port(PCIHostState *pci, const char *name, int index);
 
  #define TYPE_PNV_MACHINE       MACHINE_TYPE_NAME("powernv")
  typedef struct PnvMachineClass PnvMachineClass;
#12 lore raw

Re: [PATCH 03/11] ppc/pnv: use dev->parent_bus->parent to get the PHB

In reply to Daniel Henrique Barboza
Quoted 29 lines · Daniel Henrique Barboza

It is not advisable to execute an object_dynamic_cast() to poke into bus->qbus.parent and follow it up with a C cast into the PnvPHB type we think we got.

A better way is to access the PnvPHB object via a QOM macro accessing the existing parent links of the DeviceState. For a given pnv-phb3/4-root-port 'dev', dev->parent_bus will give us the PHB bus, and dev->parent_bus->parent is the PHB. Use the adequate QOM macro to assert the type, and keep the NULL check in case we didn't get the object we were expecting.

---
  hw/pci-host/pnv_phb3.c | 10 +++++++---
  hw/pci-host/pnv_phb4.c | 10 +++++++---
  2 files changed, 14 insertions(+), 6 deletions(-)
hw/pci-host/pnv_phb3.c +0 -1
index 4ba660f8b9..7901d8172c 100644
@@ -1139,12 +1139,16 @@ static void pnv_phb3_root_port_realize(DeviceState *dev, Error **errp)
  {
      PCIERootPortClass *rpc = PCIE_ROOT_PORT_GET_CLASS(dev);
      PCIDevice *pci = PCI_DEVICE(dev);
-    PCIBus *bus = pci_get_bus(pci);
      PnvPHB3 *phb = NULL;
      Error *local_err = NULL;
 

hw/pci-host/pnv_phb3.c static void pnv_phb3_root_port_realize(DeviceState *dev, Error **errp) · line 1152

Quoted 10 lines
hw/pci-host/pnv_phb3.c +7 -2
-    phb = (PnvPHB3 *) object_dynamic_cast(OBJECT(bus->qbus.parent),
-                                          TYPE_PNV_PHB3);
+    /*
+     * dev->parent_bus gives access to the pnv-phb-root bus.
+     * The PnvPHB3 is the owner (parent) of the bus.
+     */
+    if (dev && dev->parent_bus) {
+        phb = PNV_PHB3(dev->parent_bus->parent);
+    }
 

Couldn't we simply use :

      phb = PNV_PHB3(bus);

?

Thanks,

C.

Quoted 26 lines
hw/pci-host/pnv_phb3.c +0 -0
      if (!phb) {
          error_setg(errp,
hw/pci-host/pnv_phb4.c +7 -3
index ffd9d8a947..bae9398d86 100644
@@ -1782,12 +1782,16 @@ static void pnv_phb4_root_port_realize(DeviceState *dev, Error **errp)
  {
      PCIERootPortClass *rpc = PCIE_ROOT_PORT_GET_CLASS(dev);
      PCIDevice *pci = PCI_DEVICE(dev);
-    PCIBus *bus = pci_get_bus(pci);
      PnvPHB4 *phb = NULL;
      Error *local_err = NULL;
 
-    phb = (PnvPHB4 *) object_dynamic_cast(OBJECT(bus->qbus.parent),
-                                          TYPE_PNV_PHB4);
+    /*
+     * dev->parent_bus gives access to the pnv-phb-root bus.
+     * The PnvPHB4 is the owner (parent) of the bus.
+     */
+    if (dev && dev->parent_bus) {
+        phb = PNV_PHB4(dev->parent_bus->parent);
+    }
 
      if (!phb) {
          error_setg(errp, "%s must be connected to pnv-phb4 buses", dev->id);
#4 lore raw

Re: [PATCH 01/11] ppc/pnv: move root port attach to pnv_phb4_realize()

In reply to Cédric Le Goater
Quoted 11 lines · Cédric Le Goater
Quoted 4 lines · Daniel Henrique Barboza

Creating a root port is something related to the PHB, not the PEC. It also makes the logic more in line with what pnv-phb3 does.

So the root port is back where it was.

Could we avoid the pci_new() and use object_initialize_child() instead ?

We could but then we would need to deal with yet another difference with default versus user created devices, given that for user devices we can't initialize_child(). And since we're also unifying the root ports later on I'd rather wait to see how it turns out when everything is finished.

Tanks,

Daniel

Quoted 46 lines

Thanks,

C.

Quoted 39 lines

---
  hw/pci-host/pnv_phb4.c     | 4 ++++
  hw/pci-host/pnv_phb4_pec.c | 3 ---

  2 files changed, 4 insertions(+), 3 deletions(-)
hw/pci-host/pnv_phb4.c +4 -0
index 6594016121..23ad8de7ee 100644
@@ -1547,6 +1547,7 @@ static void pnv_phb4_instance_init(Object *obj)
   static void pnv_phb4_realize(DeviceState *dev, Error **errp)
   {
       PnvPHB4 *phb = PNV_PHB4(dev);
+    PnvPhb4PecClass *pecc = PNV_PHB4_PEC_GET_CLASS(phb->pec);
       PCIHostState *pci = PCI_HOST_BRIDGE(dev);
       XiveSource *xsrc = &phb->xsrc;
       int nr_irqs;
@@ -1583,6 +1584,9 @@ static void pnv_phb4_realize(DeviceState *dev, Error **errp)
       pci_setup_iommu(pci->bus, pnv_phb4_dma_iommu, phb);
       pci->bus->flags |= PCI_BUS_EXTENDED_CONFIG_SPACE;
+    /* Add a single Root port if running with defaults */
+    pnv_phb_attach_root_port(pci, pecc->rp_model);
+
       /* Setup XIVE Source */
       if (phb->big_phb) {
           nr_irqs = PNV_PHB4_MAX_INTs;
hw/pci-host/pnv_phb4_pec.c +0 -3
index 8b7e823fa5..c9aaf1c28e 100644
@@ -130,9 +130,6 @@ static void pnv_pec_default_phb_realize(PnvPhb4PecState *pec,
       if (!sysbus_realize(SYS_BUS_DEVICE(phb), errp)) {
           return;
       }
-
-    /* Add a single Root port if running with defaults */
-    pnv_phb_attach_root_port(PCI_HOST_BRIDGE(phb), pecc->rp_model);
   }
   static void pnv_pec_realize(DeviceState *dev, Error **errp)
#22 lore raw

Re: [PATCH 06/11] ppc/pnv: make pnv_ics_resend() use chip8->phbs[]

In reply to Frederic Barrat
Quoted 37 lines · Frederic Barrat
Quoted 34 lines · Daniel Henrique Barboza

pnv_ics_resend() is scrolling through all the child objects of the chip to search for the PHBs. It's faster and simpler to just use the phbs[] array.

pnv_ics_resend_child() was folded into pnv_ics_resend() since it's too simple to justify its own function.

---
  hw/ppc/pnv.c | 22 +++++++---------------

  1 file changed, 7 insertions(+), 15 deletions(-)
hw/ppc/pnv.c +0 -12
index 05a8d5034f..d70deffa1d 100644
@@ -1993,28 +1993,20 @@ PnvChip *pnv_get_chip(PnvMachineState *pnv, uint32_t chip_id)
       return NULL;
   }
-static int pnv_ics_resend_child(Object *child, void *opaque)
-{
-    PnvPHB3 *phb3 = (PnvPHB3 *) object_dynamic_cast(child, TYPE_PNV_PHB3);
-
-    if (phb3) {
-        ics_resend(&phb3->lsis);
-        ics_resend(ICS(&phb3->msis));
-    }
-    return 0;
-}
-
   static void pnv_ics_resend(XICSFabric *xi)
   {
       PnvMachineState *pnv = PNV_MACHINE(xi);
-    int i;
Quoted 8 lines

hw/ppc/pnv.c PnvChip *pnv_get_chip(PnvMachineState *pnv, uint32_t chip_id) · line 2016

Quoted 5 lines
hw/ppc/pnv.c +1 -2
+    int i, j;
       for (i = 0; i < pnv->num_chips; i++) {
-        PnvChip *chip = pnv->chips[i];
           Pnv8Chip *chip8 = PNV8_CHIP(pnv->chips[i]);
-        ics_resend(&chip8->psi.ics);

That line shouldn't be dropped, right?

oooffff. It shouldn't. I'll fix it in the v2.

It didn't break anything I could see though. OS boots with network with ping ....

Daniel

Quoted 14 lines

  Fred

Quoted 9 lines
hw/ppc/pnv.c +6 -1
-        object_child_foreach(OBJECT(chip), pnv_ics_resend_child, NULL);
+        for (j = 0; j < chip8->num_phbs; j++) {
+            PnvPHB3 *phb3 = &chip8->phbs[j];
+
+            ics_resend(&phb3->lsis);
+            ics_resend(ICS(&phb3->msis));
+        }
       }
   }
#27 lore raw

Re: [PATCH 08/11] ppc/pnv: turn chip8->phbs[] into a PnvPHB3* array

In reply to Frederic Barrat
Quoted 20 lines · Frederic Barrat
Quoted 17 lines · Daniel Henrique Barboza

When enabling user created PHBs (a change reverted by commit 9c10d86fee) we were handling PHBs created by default versus by the user in different manners. The only difference between these PHBs is that one will have a valid phb3->chip that is assigned during pnv_chip_power8_realize(), while the user created needs to search which chip it belongs to.

Aside from that there shouldn't be any difference. Making the default PHBs behave in line with the user created ones will make it easier to re-introduce them later on. It will also make the code easier to follow since we are dealing with them in equal manner.

The first step is to turn chip8->phbs[] into a PnvPHB3 pointer array. This will allow us to assign user created PHBs into it later on. The way we initilize the default case is now more in line with that would happen with the user created case: the object is created, parented by the chip because pnv_xscom_dt() relies on it, and then assigned to the array.

Quoted 7 lines
Quoted 2 lines

---

This patch is more prep work for the user-created device instead of general cleanup like the previous ones, but I don't see anything wrong with it. So:

I've been thinking about it and I guess I could do better with this and the proxy pnv-phb series that is already in v2. What I'm thinking is:

- crop patches 8-11 from this series. Patches 1-7 would be the prep cleanup series;

- split the pnv-phb series in two:

- first series will just introduce the pnv-phb devices and consolidate the
root ports. We're going to deal just with default devices. No consideration about future user-created devices will be made;

- a second series would then re-introduce user creatable phbs and root ports.
Patches 8-11 of this series would be handled in this second patch set since it's closely related to user devices.

Does that sound fair?

Thanks,

Daniel

Quoted 72 lines

  Fred

Quoted 67 lines

  hw/ppc/pnv.c         | 19 ++++++++++++++-----
  include/hw/ppc/pnv.h |  6 +++++-

  2 files changed, 19 insertions(+), 6 deletions(-)
hw/ppc/pnv.c +14 -5
index 5e3323e950..6ce9e94e05 100644
@@ -660,7 +660,7 @@ static void pnv_chip_power8_pic_print_info(PnvChip *chip, Monitor *mon)
       ics_pic_print_info(&chip8->psi.ics, mon);
       for (i = 0; i < chip8->num_phbs; i++) {
-        PnvPHB3 *phb3 = &chip8->phbs[i];
+        PnvPHB3 *phb3 = chip8->phbs[i];
           pnv_phb3_msi_pic_print_info(&phb3->msis, mon);
           ics_pic_print_info(&phb3->lsis, mon);
@@ -1149,7 +1149,16 @@ static void pnv_chip_power8_instance_init(Object *obj)
       chip8->num_phbs = pcc->num_phbs;
       for (i = 0; i < chip8->num_phbs; i++) {
-        object_initialize_child(obj, "phb[*]", &chip8->phbs[i], TYPE_PNV_PHB3);
+        PnvPHB3 *phb3 = PNV_PHB3(object_new(TYPE_PNV_PHB3));
+
+        /*
+         * We need the chip to parent the PHB to allow the DT
+         * to build correctly (via pnv_xscom_dt()).
+         *
+         * TODO: the PHB should be parented by a PEC device.
+         */
+        object_property_add_child(obj, "phb[*]", OBJECT(phb3));
+        chip8->phbs[i] = phb3;
       }
   }
@@ -1278,7 +1287,7 @@ static void pnv_chip_power8_realize(DeviceState *dev, Error **errp)
       /* PHB3 controllers */
       for (i = 0; i < chip8->num_phbs; i++) {
-        PnvPHB3 *phb = &chip8->phbs[i];
+        PnvPHB3 *phb = chip8->phbs[i];
           object_property_set_int(OBJECT(phb), "index", i, &error_fatal);
           object_property_set_int(OBJECT(phb), "chip-id", chip->chip_id,
@@ -1963,7 +1972,7 @@ static ICSState *pnv_ics_get(XICSFabric *xi, int irq)
           }
           for (j = 0; j < chip8->num_phbs; j++) {
-            pnv_ics_get_phb_ics(&chip8->phbs[j], &args);
+            pnv_ics_get_phb_ics(chip8->phbs[j], &args);
               if (args.ics) {
                   return args.ics;
@@ -1996,7 +2005,7 @@ static void pnv_ics_resend(XICSFabric *xi)
           Pnv8Chip *chip8 = PNV8_CHIP(pnv->chips[i]);
           for (j = 0; j < chip8->num_phbs; j++) {
-            PnvPHB3 *phb3 = &chip8->phbs[j];
+            PnvPHB3 *phb3 = chip8->phbs[j];
               ics_resend(&phb3->lsis);
               ics_resend(ICS(&phb3->msis));
include/hw/ppc/pnv.h +5 -1
index 033890a23f..11f1089289 100644
@@ -80,7 +80,11 @@ struct Pnv8Chip {
       PnvHomer     homer;
   #define PNV8_CHIP_PHB3_MAX 4
-    PnvPHB3      phbs[PNV8_CHIP_PHB3_MAX];
+    /*
+     * The array is used to allow quick access to the phbs by
+     * pnv_ics_get_child() and pnv_ics_resend_child().
+     */
+    PnvPHB3      *phbs[PNV8_CHIP_PHB3_MAX];
       uint32_t     num_phbs;
       XICSFabric    *xics;
#28 lore raw

Re: [PATCH 08/11] ppc/pnv: turn chip8->phbs[] into a PnvPHB3* array

Quoted 14 lines · Daniel Henrique Barboza

I've been thinking about it and I guess I could do better with this and the proxy pnv-phb series that is already in v2. What I'm thinking is:

- crop patches 8-11 from this series. Patches 1-7 would be the prep cleanup series;

- split the pnv-phb series in two:

  - first series will just introduce the pnv-phb devices and
consolidate the
root ports. We're going to deal just with default devices. No consideration about future user-created devices will be made;

Quoted 7 lines

  - a second series would then re-introduce user creatable phbs and
root ports.
Patches 8-11 of this series would be handled in this second patch set since it's closely related to user devices.

Does that sound fair?

Sounds good to me. That should keep series smaller and easier to review and merge.

Fred

#29 lore raw

Re: [PATCH 08/11] ppc/pnv: turn chip8->phbs[] into a PnvPHB3* array

In reply to Daniel Henrique Barboza
Quoted 36 lines · Daniel Henrique Barboza
Quoted 20 lines · Frederic Barrat
Quoted 17 lines · Daniel Henrique Barboza

When enabling user created PHBs (a change reverted by commit 9c10d86fee) we were handling PHBs created by default versus by the user in different manners. The only difference between these PHBs is that one will have a valid phb3->chip that is assigned during pnv_chip_power8_realize(), while the user created needs to search which chip it belongs to.

Aside from that there shouldn't be any difference. Making the default PHBs behave in line with the user created ones will make it easier to re-introduce them later on. It will also make the code easier to follow since we are dealing with them in equal manner.

The first step is to turn chip8->phbs[] into a PnvPHB3 pointer array. This will allow us to assign user created PHBs into it later on. The way we initilize the default case is now more in line with that would happen with the user created case: the object is created, parented by the chip because pnv_xscom_dt() relies on it, and then assigned to the array.

Quoted 7 lines
Quoted 2 lines

---

This patch is more prep work for the user-created device instead of general cleanup like the previous ones, but I don't see anything wrong with it. So:

I've been thinking about it and I guess I could do better with this and the proxy pnv-phb series that is already in v2. What I'm thinking is:

Quoted 8 lines

- crop patches 8-11 from this series. Patches 1-7 would be the prep cleanup series;

- split the pnv-phb series in two:

  - first series will just introduce the pnv-phb devices and consolidate the
root ports. We're going to deal just with default devices. No consideration about future user-created devices will be made;

Yes. From what I have read, this looks very feasible with a v2.

Thanks,

C.

Quoted 89 lines

  - a second series would then re-introduce user creatable phbs and root ports.
Patches 8-11 of this series would be handled in this second patch set since it's closely related to user devices.

Does that sound fair?

Thanks,

Daniel

Quoted 72 lines

   Fred

Quoted 67 lines

  hw/ppc/pnv.c         | 19 ++++++++++++++-----
  include/hw/ppc/pnv.h |  6 +++++-

  2 files changed, 19 insertions(+), 6 deletions(-)
hw/ppc/pnv.c +14 -5
index 5e3323e950..6ce9e94e05 100644
@@ -660,7 +660,7 @@ static void pnv_chip_power8_pic_print_info(PnvChip *chip, Monitor *mon)
       ics_pic_print_info(&chip8->psi.ics, mon);
       for (i = 0; i < chip8->num_phbs; i++) {
-        PnvPHB3 *phb3 = &chip8->phbs[i];
+        PnvPHB3 *phb3 = chip8->phbs[i];
           pnv_phb3_msi_pic_print_info(&phb3->msis, mon);
           ics_pic_print_info(&phb3->lsis, mon);
@@ -1149,7 +1149,16 @@ static void pnv_chip_power8_instance_init(Object *obj)
       chip8->num_phbs = pcc->num_phbs;
       for (i = 0; i < chip8->num_phbs; i++) {
-        object_initialize_child(obj, "phb[*]", &chip8->phbs[i], TYPE_PNV_PHB3);
+        PnvPHB3 *phb3 = PNV_PHB3(object_new(TYPE_PNV_PHB3));
+
+        /*
+         * We need the chip to parent the PHB to allow the DT
+         * to build correctly (via pnv_xscom_dt()).
+         *
+         * TODO: the PHB should be parented by a PEC device.
+         */
+        object_property_add_child(obj, "phb[*]", OBJECT(phb3));
+        chip8->phbs[i] = phb3;
       }
   }
@@ -1278,7 +1287,7 @@ static void pnv_chip_power8_realize(DeviceState *dev, Error **errp)
       /* PHB3 controllers */
       for (i = 0; i < chip8->num_phbs; i++) {
-        PnvPHB3 *phb = &chip8->phbs[i];
+        PnvPHB3 *phb = chip8->phbs[i];
           object_property_set_int(OBJECT(phb), "index", i, &error_fatal);
           object_property_set_int(OBJECT(phb), "chip-id", chip->chip_id,
@@ -1963,7 +1972,7 @@ static ICSState *pnv_ics_get(XICSFabric *xi, int irq)
           }
           for (j = 0; j < chip8->num_phbs; j++) {
-            pnv_ics_get_phb_ics(&chip8->phbs[j], &args);
+            pnv_ics_get_phb_ics(chip8->phbs[j], &args);
               if (args.ics) {
                   return args.ics;
@@ -1996,7 +2005,7 @@ static void pnv_ics_resend(XICSFabric *xi)
           Pnv8Chip *chip8 = PNV8_CHIP(pnv->chips[i]);
           for (j = 0; j < chip8->num_phbs; j++) {
-            PnvPHB3 *phb3 = &chip8->phbs[j];
+            PnvPHB3 *phb3 = chip8->phbs[j];
               ics_resend(&phb3->lsis);
               ics_resend(ICS(&phb3->msis));
include/hw/ppc/pnv.h +5 -1
index 033890a23f..11f1089289 100644
@@ -80,7 +80,11 @@ struct Pnv8Chip {
       PnvHomer     homer;
   #define PNV8_CHIP_PHB3_MAX 4
-    PnvPHB3      phbs[PNV8_CHIP_PHB3_MAX];
+    /*
+     * The array is used to allow quick access to the phbs by
+     * pnv_ics_get_child() and pnv_ics_resend_child().
+     */
+    PnvPHB3      *phbs[PNV8_CHIP_PHB3_MAX];
       uint32_t     num_phbs;
       XICSFabric    *xics;
#11 lore raw

Re: [PATCH 03/11] ppc/pnv: use dev->parent_bus->parent to get the PHB

In reply to Frederic Barrat
Quoted 55 lines · Frederic Barrat
Quoted 52 lines · Daniel Henrique Barboza

It is not advisable to execute an object_dynamic_cast() to poke into bus->qbus.parent and follow it up with a C cast into the PnvPHB type we think we got.

A better way is to access the PnvPHB object via a QOM macro accessing the existing parent links of the DeviceState. For a given pnv-phb3/4-root-port 'dev', dev->parent_bus will give us the PHB bus, and dev->parent_bus->parent is the PHB. Use the adequate QOM macro to assert the type, and keep the NULL check in case we didn't get the object we were expecting.

---
  hw/pci-host/pnv_phb3.c | 10 +++++++---
  hw/pci-host/pnv_phb4.c | 10 +++++++---

  2 files changed, 14 insertions(+), 6 deletions(-)
hw/pci-host/pnv_phb3.c +7 -3
index 4ba660f8b9..7901d8172c 100644
@@ -1139,12 +1139,16 @@ static void pnv_phb3_root_port_realize(DeviceState *dev, Error **errp)
   {
       PCIERootPortClass *rpc = PCIE_ROOT_PORT_GET_CLASS(dev);
       PCIDevice *pci = PCI_DEVICE(dev);
-    PCIBus *bus = pci_get_bus(pci);
       PnvPHB3 *phb = NULL;
       Error *local_err = NULL;
-    phb = (PnvPHB3 *) object_dynamic_cast(OBJECT(bus->qbus.parent),
-                                          TYPE_PNV_PHB3);
+    /*
+     * dev->parent_bus gives access to the pnv-phb-root bus.
+     * The PnvPHB3 is the owner (parent) of the bus.
+     */
+    if (dev && dev->parent_bus) {
+        phb = PNV_PHB3(dev->parent_bus->parent);
+    }
       if (!phb) {
           error_setg(errp,
hw/pci-host/pnv_phb4.c +0 -3
index ffd9d8a947..bae9398d86 100644
@@ -1782,12 +1782,16 @@ static void pnv_phb4_root_port_realize(DeviceState *dev, Error **errp)
   {
       PCIERootPortClass *rpc = PCIE_ROOT_PORT_GET_CLASS(dev);
       PCIDevice *pci = PCI_DEVICE(dev);
-    PCIBus *bus = pci_get_bus(pci);
       PnvPHB4 *phb = NULL;
       Error *local_err = NULL;
-    phb = (PnvPHB4 *) object_dynamic_cast(OBJECT(bus->qbus.parent),
-                                          TYPE_PNV_PHB4);
Quoted 8 lines

hw/pci-host/pnv_phb4.c static void pnv_phb4_root_port_realize(DeviceState *dev, Error **errp) · line 1792

Quoted 5 lines
hw/pci-host/pnv_phb4.c +5 -0
+    /*
+     * dev->parent_bus gives access to the pnv-phb-root bus.
+     * The PnvPHB4 is the owner (parent) of the bus.
+     */
+    if (dev && dev->parent_bus) {

Does it make sense to test 'dev' first when it's the device being realized?

Hmmm not really. I got overzealous here it seems.

I'll keep just the check for dev->parent in v2.

Thanks,

Daniel

Quoted 11 lines

  Fred

Quoted 4 lines
hw/pci-host/pnv_phb4.c +2 -0
+        phb = PNV_PHB4(dev->parent_bus->parent);
+    }
       if (!phb) {
           error_setg(errp, "%s must be connected to pnv-phb4 buses", dev->id);
#13 lore raw

Re: [PATCH 03/11] ppc/pnv: use dev->parent_bus->parent to get the PHB

In reply to Cédric Le Goater
Quoted 42 lines · Cédric Le Goater
Quoted 28 lines · Daniel Henrique Barboza

It is not advisable to execute an object_dynamic_cast() to poke into bus->qbus.parent and follow it up with a C cast into the PnvPHB type we think we got.

A better way is to access the PnvPHB object via a QOM macro accessing the existing parent links of the DeviceState. For a given pnv-phb3/4-root-port 'dev', dev->parent_bus will give us the PHB bus, and dev->parent_bus->parent is the PHB. Use the adequate QOM macro to assert the type, and keep the NULL check in case we didn't get the object we were expecting.

---
  hw/pci-host/pnv_phb3.c | 10 +++++++---
  hw/pci-host/pnv_phb4.c | 10 +++++++---

  2 files changed, 14 insertions(+), 6 deletions(-)
hw/pci-host/pnv_phb3.c +0 -1
index 4ba660f8b9..7901d8172c 100644
@@ -1139,12 +1139,16 @@ static void pnv_phb3_root_port_realize(DeviceState *dev, Error **errp)
   {
       PCIERootPortClass *rpc = PCIE_ROOT_PORT_GET_CLASS(dev);
       PCIDevice *pci = PCI_DEVICE(dev);
-    PCIBus *bus = pci_get_bus(pci);
       PnvPHB3 *phb = NULL;
       Error *local_err = NULL;

hw/pci-host/pnv_phb3.c static void pnv_phb3_root_port_realize(DeviceState *dev, Error **errp) · line 1151

Quoted 10 lines
hw/pci-host/pnv_phb3.c +7 -2
-    phb = (PnvPHB3 *) object_dynamic_cast(OBJECT(bus->qbus.parent),
-                                          TYPE_PNV_PHB3);
+    /*
+     * dev->parent_bus gives access to the pnv-phb-root bus.
+     * The PnvPHB3 is the owner (parent) of the bus.
+     */
+    if (dev && dev->parent_bus) {
+        phb = PNV_PHB3(dev->parent_bus->parent);
+    }
 

Couldn't we simply use :

Quoted 3 lines

      phb = PNV_PHB3(bus);

?

No. This will give us a reference to a pnv-phb3-root object.

Getting a reference to the PHB by using bus->parent happens in other parts of code, such as:

hw/pci-host/gpex-acpi.c: crs = build_crs(PCI_HOST_BRIDGE(BUS(bus)->parent), &crs_range_set,
hw/pci-bridge/pci_expander_bridge.c: main_host = PCI_HOST_BRIDGE(pxb_dev_base->parent_bus->parent);

So I believe we're not out of line here.

Thanks,

Daniel

Quoted 31 lines

Thanks,

C.

Quoted 24 lines
hw/pci-host/pnv_phb3.c +0 -0
       if (!phb) {
           error_setg(errp,
hw/pci-host/pnv_phb4.c +7 -3
index ffd9d8a947..bae9398d86 100644
@@ -1782,12 +1782,16 @@ static void pnv_phb4_root_port_realize(DeviceState *dev, Error **errp)
   {
       PCIERootPortClass *rpc = PCIE_ROOT_PORT_GET_CLASS(dev);
       PCIDevice *pci = PCI_DEVICE(dev);
-    PCIBus *bus = pci_get_bus(pci);
       PnvPHB4 *phb = NULL;
       Error *local_err = NULL;
-    phb = (PnvPHB4 *) object_dynamic_cast(OBJECT(bus->qbus.parent),
-                                          TYPE_PNV_PHB4);
+    /*
+     * dev->parent_bus gives access to the pnv-phb-root bus.
+     * The PnvPHB4 is the owner (parent) of the bus.
+     */
+    if (dev && dev->parent_bus) {
+        phb = PNV_PHB4(dev->parent_bus->parent);
+    }
       if (!phb) {
           error_setg(errp, "%s must be connected to pnv-phb4 buses", dev->id);