[PATCH v3 00/15] s390x/pci: Implement migration for emulated devices

[PATCH v3 00/15] s390x/pci: Implement migration for emulated devices

[Konstantin Shkolnyy]

Up until now QEMU marked any zPCI device as "unmigratable."

This patch series adds support for migrating emulated devices, which are
simpler to migrate. It leaves VFIO devices still unmigratable.

To enable migration, the device state needs to be saved/restored to/from
the migration stream.

There are 2 kinds of emulated devices - those that use the zPCI IOMMU page
table emulation in QEMU (e.g., Intel IGB NIC), and those that don't (virtio).
This is important to note for testing purposes.

This change was tested on IGB, virtio-net and virtio-blk devices by running
I/O on them while performing "virsh managedsave, virsh start" and also live migration to another host and back.

Changes in v3:
 - patch 1: provided proper error handling.
 - patch 14: Fixed zpci_table corruption during migration that occurred if
 devices were hot-unplugged and plugged before migration.
 - patch 14: Provided UID-based migration stream IDs for zpci devices to avoid
 unstable auto-generated IDs that caused migration problems if devices
 were hot-unplugged and plugged before migration.

Konstantin Shkolnyy (14):
  s390x/pci: Create function to contain translation status check
  s390x/pci: Move iommu_mr from S390PCIIOMMU to S390PCIBusDevice
  s390x/pci: Move dm_mr from S390PCIIOMMU to S390PCIBusDevice
  s390x/pci: Move iotlb from S390PCIIOMMU to S390PCIBusDevice
  s390x/pci: Remove a ptr to S390PCIBusDevice from S390PCIIOMMU
  s390x/pci: Move/rename enabled from S390PCIIOMMU to S390PCIBusDevice
  s390x/pci: Move dma_limit from S390PCIIOMMU to S390PCIBusDevice
  s390x/pci: Move g_iota from S390PCIIOMMU to S390PCIBusDevice
  s390x/pci: Move pba from S390PCIIOMMU to S390PCIBusDevice
  s390x/pci: Move pal from S390PCIIOMMU to S390PCIBusDevice
  s390x/pci: Move max_dma_limit from S390PCIIOMMU to S390PCIBusDevice
  s390x/pci: Add a comment explaining S390PCIIOMMU purpose
  s390x/pci: Implement migration for emulated devices
  s390x/pci: Create function to contain fmb_timer start

Matthew Rosato (1):
  s390x/pci: implement IOMMU replay

 hw/s390x/s390-pci-bus.c          | 338 ++++++++++++++++++++++++++-----
 hw/s390x/s390-pci-inst.c         | 116 ++++++-----
 hw/s390x/s390-pci-vfio.c         |   4 +-
 hw/s390x/s390-virtio-ccw.c       |   5 +
 include/hw/s390x/s390-pci-bus.h  |  39 ++--
 include/hw/s390x/s390-pci-inst.h |   3 +-
 6 files changed, 380 insertions(+), 125 deletions(-)

-- 
2.34.1

[PATCH v3 01/15] s390x/pci: implement IOMMU replay

[Konstantin Shkolnyy]

From: Matthew Rosato <mjrosato@linux.ibm.com>

There are a few scenarios where IOMMU replay can potentially be needed
for zPCI device, namely VFIO device reset scenarios where the guest
continues running and expects the contents of its IOMMU to be replayed
upon IOAT re-registration and migration scenarios where the destination
must reconstruct the IOMMU on the destination.

zPCI migration is not supported yet, but the IOMMU replay function is
implemented so that it can be called both from IOMMUMemoryRegionClass
now and migration post_load later.

Signed-off-by: Matthew Rosato <mjrosato@linux.ibm.com>
Signed-off-by: Konstantin Shkolnyy <kshk@linux.ibm.com>
---
 hw/s390x/s390-pci-bus.c          | 62 ++++++++++++++++++++++++++++----
 hw/s390x/s390-pci-inst.c         |  4 +--
 include/hw/s390x/s390-pci-inst.h |  1 +
 3 files changed, 59 insertions(+), 8 deletions(-)

diff --git a/hw/s390x/s390-pci-bus.c b/hw/s390x/s390-pci-bus.c
index 4de7b587e8..a104e550b1 100644
--- a/hw/s390x/s390-pci-bus.c
+++ b/hw/s390x/s390-pci-bus.c
@@ -592,14 +592,64 @@ err:
     return ret;
 }
 
-static void s390_pci_iommu_replay(IOMMUMemoryRegion *iommu,
+static void s390_pci_ioat_replay(S390PCIIOMMU *iommu)
+{
+    S390IOTLBEntry entry;
+    uint16_t error = 0;
+    uint32_t dma_avail;
+    hwaddr curr, end;
+
+    curr = iommu->pba;
+    end = iommu->pal;
+
+    if (iommu->dm_mr) {
+        /* If direct mapping is used, there are no guest tables to replay */
+        return;
+    }
+
+    if (iommu->dma_limit) {
+        dma_avail = iommu->dma_limit->avail;
+    } else {
+        dma_avail = 1;
+    }
+
+    while (curr < end) {
+        error = s390_guest_io_table_walk(iommu->g_iota, curr, &entry);
+        if (error) {
+            pbdev->state = ZPCI_FS_ERROR;
+            s390_pci_generate_error_event(error, pbdev->fh, pbdev->fid, curr,
+                                          0);
+            error_report("Failure to walk table during iommu remap");
+            return;
+        }
+
+        if (entry.perm != IOMMU_NONE) {
+            if (dma_avail > 0) {
+                dma_avail = s390_pci_update_iotlb(iommu, &entry);
+            } else {
+                /*
+                 * There is no reliable method to request the guest to release
+                 * mappings other than in response to a RPCIT instruction;
+                 * generate a permanent error condition and require the device
+                 * to be completely re-initialized from the guest side.
+                 */
+                pbdev->state = ZPCI_FS_ERROR;
+                s390_pci_generate_error_event(ERR_EVENT_PERMERR, pbdev->fh,
+                                              pbdev->fid, 0, 0);
+                error_report("DMA mappings exhausted: iommu remap failed");
+                return;
+            }
+        }
+        curr += entry.len;
+    }
+}
+
+static void s390_pci_iommu_replay(IOMMUMemoryRegion *mr,
                                   IOMMUNotifier *notifier)
 {
-    /* It's impossible to plug a pci device on s390x that already has iommu
-     * mappings which need to be replayed, that is due to the "one iommu per
-     * zpci device" construct. But when we support migration of vfio-pci
-     * devices in future, we need to revisit this.
-     */
+    S390PCIIOMMU *iommu = container_of(mr, S390PCIIOMMU, iommu_mr);
+
+    s390_pci_ioat_replay(iommu);
 }
 
 static S390PCIIOMMU *s390_pci_get_iommu(S390pciState *s, PCIBus *bus,
diff --git a/hw/s390x/s390-pci-inst.c b/hw/s390x/s390-pci-inst.c
index 10066ca618..1834596076 100644
--- a/hw/s390x/s390-pci-inst.c
+++ b/hw/s390x/s390-pci-inst.c
@@ -613,8 +613,8 @@ int pcistg_service_call(S390CPU *cpu, uint8_t r1, uint8_t r2, uintptr_t ra)
     return 0;
 }
 
-static uint32_t s390_pci_update_iotlb(S390PCIIOMMU *iommu,
-                                      S390IOTLBEntry *entry)
+uint32_t s390_pci_update_iotlb(S390PCIIOMMU *iommu,
+                               S390IOTLBEntry *entry)
 {
     S390IOTLBEntry *cache = g_hash_table_lookup(iommu->iotlb, &entry->iova);
     IOMMUTLBEvent event = {
diff --git a/include/hw/s390x/s390-pci-inst.h b/include/hw/s390x/s390-pci-inst.h
index 5cb8da540b..c782990e3b 100644
--- a/include/hw/s390x/s390-pci-inst.h
+++ b/include/hw/s390x/s390-pci-inst.h
@@ -111,6 +111,7 @@ int mpcifc_service_call(S390CPU *cpu, uint8_t r1, uint64_t fiba, uint8_t ar,
 int stpcifc_service_call(S390CPU *cpu, uint8_t r1, uint64_t fiba, uint8_t ar,
                          uintptr_t ra);
 void fmb_timer_free(S390PCIBusDevice *pbdev);
+uint32_t s390_pci_update_iotlb(S390PCIIOMMU *iommu, S390IOTLBEntry *entry);
 
 #define ZPCI_IO_BAR_MIN 0
 #define ZPCI_IO_BAR_MAX 5
-- 
2.34.1

[PATCH v3 02/15] s390x/pci: Create function to contain translation status check

[Konstantin Shkolnyy]

Make it more clear what the bit means, and the new function will be called
from yet another place in the future.

Reviewed-by: Christian Borntraeger <borntraeger@linux.ibm.com>
Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Signed-off-by: Konstantin Shkolnyy <kshk@linux.ibm.com>
---
 hw/s390x/s390-pci-inst.c        | 7 ++++++-
 include/hw/s390x/s390-pci-bus.h | 1 +
 2 files changed, 7 insertions(+), 1 deletion(-)

diff --git a/hw/s390x/s390-pci-inst.c b/hw/s390x/s390-pci-inst.c
index 1834596076..7c784b31f3 100644
--- a/hw/s390x/s390-pci-inst.c
+++ b/hw/s390x/s390-pci-inst.c
@@ -993,6 +993,11 @@ int pci_dereg_irqs(S390PCIBusDevice *pbdev)
     return 0;
 }
 
+bool s390_pci_is_translation_enabled(uint64_t g_iota)
+{
+    return ((g_iota >> 11) & 0x1) != 0; /* "T" bit */
+}
+
 static int reg_ioat(CPUS390XState *env, S390PCIBusDevice *pbdev, ZpciFib fib,
                     uintptr_t ra)
 {
@@ -1001,7 +1006,7 @@ static int reg_ioat(CPUS390XState *env, S390PCIBusDevice *pbdev, ZpciFib fib,
     uint64_t pal = ldq_be_p(&fib.pal);
     uint64_t g_iota = ldq_be_p(&fib.iota);
     uint8_t dt = (g_iota >> 2) & 0x7;
-    uint8_t t = (g_iota >> 11) & 0x1;
+    bool t = s390_pci_is_translation_enabled(g_iota);
 
     pba &= ~0xfff;
     pal |= 0xfff;
diff --git a/include/hw/s390x/s390-pci-bus.h b/include/hw/s390x/s390-pci-bus.h
index 9228523ce8..eb15cb8b2d 100644
--- a/include/hw/s390x/s390-pci-bus.h
+++ b/include/hw/s390x/s390-pci-bus.h
@@ -390,6 +390,7 @@ int pci_chsc_sei_nt2_get_event(void *res);
 int pci_chsc_sei_nt2_have_event(void);
 void s390_pci_sclp_configure(SCCB *sccb);
 void s390_pci_sclp_deconfigure(SCCB *sccb);
+bool s390_pci_is_translation_enabled(uint64_t g_iota);
 void s390_pci_iommu_enable(S390PCIIOMMU *iommu);
 void s390_pci_iommu_direct_map_enable(S390PCIIOMMU *iommu);
 void s390_pci_iommu_disable(S390PCIIOMMU *iommu);
-- 
2.34.1

[PATCH v3 03/15] s390x/pci: Move iommu_mr from S390PCIIOMMU to S390PCIBusDevice

[Konstantin Shkolnyy]

This field is only used when S390PCIBusDevice exists, so it can be moved
there to simplify S390PCIIOMMU which purpose is just to store the "root"
AddressSpace.

Signed-off-by: Konstantin Shkolnyy <kshk@linux.ibm.com>
---
 hw/s390x/s390-pci-bus.c          | 32 ++++++++++++++++++--------------
 hw/s390x/s390-pci-inst.c         | 28 +++++++++++++++-------------
 include/hw/s390x/s390-pci-bus.h  |  6 +++---
 include/hw/s390x/s390-pci-inst.h |  4 ++--
 4 files changed, 38 insertions(+), 32 deletions(-)

diff --git a/hw/s390x/s390-pci-bus.c b/hw/s390x/s390-pci-bus.c
index a104e550b1..98f626b57f 100644
--- a/hw/s390x/s390-pci-bus.c
+++ b/hw/s390x/s390-pci-bus.c
@@ -206,7 +206,7 @@ void s390_pci_sclp_deconfigure(SCCB *sccb)
             pci_dereg_irqs(pbdev);
         }
         if (pbdev->iommu->enabled) {
-            pci_dereg_ioat(pbdev->iommu);
+            pci_dereg_ioat(pbdev);
         }
         pbdev->state = ZPCI_FS_STANDBY;
         rc = SCLP_RC_NORMAL_COMPLETION;
@@ -538,7 +538,8 @@ uint16_t s390_guest_io_table_walk(uint64_t g_iota, hwaddr addr,
 static IOMMUTLBEntry s390_translate_iommu(IOMMUMemoryRegion *mr, hwaddr addr,
                                           IOMMUAccessFlags flag, int iommu_idx)
 {
-    S390PCIIOMMU *iommu = container_of(mr, S390PCIIOMMU, iommu_mr);
+    S390PCIBusDevice *pbdev = container_of(mr, S390PCIBusDevice, iommu_mr);
+    S390PCIIOMMU *iommu = pbdev->iommu;
     S390IOTLBEntry *entry;
     uint64_t iova = addr & TARGET_PAGE_MASK;
     uint16_t error = 0;
@@ -592,12 +593,13 @@ err:
     return ret;
 }
 
-static void s390_pci_ioat_replay(S390PCIIOMMU *iommu)
+static void s390_pci_ioat_replay(S390PCIBusDevice *pbdev)
 {
     S390IOTLBEntry entry;
     uint16_t error = 0;
     uint32_t dma_avail;
     hwaddr curr, end;
+    S390PCIIOMMU *iommu = pbdev->iommu;
 
     curr = iommu->pba;
     end = iommu->pal;
@@ -625,7 +627,7 @@ static void s390_pci_ioat_replay(S390PCIIOMMU *iommu)
 
         if (entry.perm != IOMMU_NONE) {
             if (dma_avail > 0) {
-                dma_avail = s390_pci_update_iotlb(iommu, &entry);
+                dma_avail = s390_pci_update_iotlb(pbdev, &entry);
             } else {
                 /*
                  * There is no reliable method to request the guest to release
@@ -647,9 +649,9 @@ static void s390_pci_ioat_replay(S390PCIIOMMU *iommu)
 static void s390_pci_iommu_replay(IOMMUMemoryRegion *mr,
                                   IOMMUNotifier *notifier)
 {
-    S390PCIIOMMU *iommu = container_of(mr, S390PCIIOMMU, iommu_mr);
+    S390PCIBusDevice *pbdev = container_of(mr, S390PCIBusDevice, iommu_mr);
 
-    s390_pci_ioat_replay(iommu);
+    s390_pci_ioat_replay(pbdev);
 }
 
 static S390PCIIOMMU *s390_pci_get_iommu(S390pciState *s, PCIBus *bus,
@@ -771,19 +773,20 @@ static const MemoryRegionOps s390_msi_ctrl_ops = {
     .endianness = DEVICE_LITTLE_ENDIAN,
 };
 
-void s390_pci_iommu_enable(S390PCIIOMMU *iommu)
+void s390_pci_iommu_enable(S390PCIBusDevice *pbdev)
 {
+    S390PCIIOMMU *iommu = pbdev->iommu;
     /*
      * The iommu region is initialized against a 0-mapped address space,
      * so the smallest IOMMU region we can define runs from 0 to the end
      * of the PCI address space.
      */
     char *name = g_strdup_printf("iommu-s390-%04x", iommu->pbdev->uid);
-    memory_region_init_iommu(&iommu->iommu_mr, sizeof(iommu->iommu_mr),
+    memory_region_init_iommu(&pbdev->iommu_mr, sizeof(pbdev->iommu_mr),
                              TYPE_S390_IOMMU_MEMORY_REGION, OBJECT(&iommu->mr),
                              name, iommu->pal + 1);
     iommu->enabled = true;
-    memory_region_add_subregion(&iommu->mr, 0, MEMORY_REGION(&iommu->iommu_mr));
+    memory_region_add_subregion(&iommu->mr, 0, MEMORY_REGION(&pbdev->iommu_mr));
     g_free(name);
 }
 
@@ -809,8 +812,9 @@ void s390_pci_iommu_direct_map_enable(S390PCIIOMMU *iommu)
                                 iommu->dm_mr);
 }
 
-void s390_pci_iommu_disable(S390PCIIOMMU *iommu)
+void s390_pci_iommu_disable(S390PCIBusDevice *pbdev)
 {
+    S390PCIIOMMU *iommu = pbdev->iommu;
     iommu->enabled = false;
     g_hash_table_remove_all(iommu->iotlb);
     if (iommu->dm_mr) {
@@ -820,8 +824,8 @@ void s390_pci_iommu_disable(S390PCIIOMMU *iommu)
         iommu->dm_mr = NULL;
     } else {
         memory_region_del_subregion(&iommu->mr,
-                                    MEMORY_REGION(&iommu->iommu_mr));
-        object_unparent(OBJECT(&iommu->iommu_mr));
+                                    MEMORY_REGION(&pbdev->iommu_mr));
+        object_unparent(OBJECT(&pbdev->iommu_mr));
     }
 }
 
@@ -1420,7 +1424,7 @@ static void s390_pcihost_reset(DeviceState *dev)
                 pci_dereg_irqs(pbdev);
             }
             if (pbdev->iommu->enabled) {
-                pci_dereg_ioat(pbdev->iommu);
+                pci_dereg_ioat(pbdev);
             }
             pbdev->state = ZPCI_FS_STANDBY;
             s390_pci_perform_unplug(pbdev);
@@ -1561,7 +1565,7 @@ static void s390_pci_device_reset(DeviceState *dev)
         pci_dereg_irqs(pbdev);
     }
     if (pbdev->iommu->enabled) {
-        pci_dereg_ioat(pbdev->iommu);
+        pci_dereg_ioat(pbdev);
     }
 
     fmb_timer_free(pbdev);
diff --git a/hw/s390x/s390-pci-inst.c b/hw/s390x/s390-pci-inst.c
index 7c784b31f3..ed60e6100e 100644
--- a/hw/s390x/s390-pci-inst.c
+++ b/hw/s390x/s390-pci-inst.c
@@ -613,9 +613,10 @@ int pcistg_service_call(S390CPU *cpu, uint8_t r1, uint8_t r2, uintptr_t ra)
     return 0;
 }
 
-uint32_t s390_pci_update_iotlb(S390PCIIOMMU *iommu,
+uint32_t s390_pci_update_iotlb(S390PCIBusDevice *pbdev,
                                S390IOTLBEntry *entry)
 {
+    S390PCIIOMMU *iommu = pbdev->iommu;
     S390IOTLBEntry *cache = g_hash_table_lookup(iommu->iotlb, &entry->iova);
     IOMMUTLBEvent event = {
         .type = entry->perm ? IOMMU_NOTIFIER_MAP : IOMMU_NOTIFIER_UNMAP,
@@ -645,7 +646,7 @@ uint32_t s390_pci_update_iotlb(S390PCIIOMMU *iommu,
 
             event.type = IOMMU_NOTIFIER_UNMAP;
             event.entry.perm = IOMMU_NONE;
-            memory_region_notify_iommu(&iommu->iommu_mr, 0, event);
+            memory_region_notify_iommu(&pbdev->iommu_mr, 0, event);
             event.type = IOMMU_NOTIFIER_MAP;
             event.entry.perm = entry->perm;
         }
@@ -663,13 +664,13 @@ uint32_t s390_pci_update_iotlb(S390PCIIOMMU *iommu,
      * All associated iotlb entries have already been cleared, trigger the
      * unmaps.
      */
-    memory_region_notify_iommu(&iommu->iommu_mr, 0, event);
+    memory_region_notify_iommu(&pbdev->iommu_mr, 0, event);
 
 out:
     return iommu->dma_limit ? iommu->dma_limit->avail : 1;
 }
 
-static void s390_pci_batch_unmap(S390PCIIOMMU *iommu, uint64_t iova,
+static void s390_pci_batch_unmap(S390PCIBusDevice *pbdev, uint64_t iova,
                                  uint64_t len)
 {
     uint64_t remain = len, start = iova, end = start + len - 1, mask, size;
@@ -687,7 +688,7 @@ static void s390_pci_batch_unmap(S390PCIIOMMU *iommu, uint64_t iova,
         size = mask + 1;
         event.entry.iova = start;
         event.entry.addr_mask = mask;
-        memory_region_notify_iommu(&iommu->iommu_mr, 0, event);
+        memory_region_notify_iommu(&pbdev->iommu_mr, 0, event);
         start += size;
         remain -= size;
     }
@@ -778,14 +779,14 @@ int rpcit_service_call(S390CPU *cpu, uint8_t r1, uint8_t r2, uintptr_t ra)
             coalesce += entry.len;
         } else if (coalesce > 0) {
             /* Unleash the coalesced unmap before processing a new map */
-            s390_pci_batch_unmap(iommu, iova, coalesce);
+            s390_pci_batch_unmap(pbdev, iova, coalesce);
             coalesce = 0;
         }
 
         start += entry.len;
         while (entry.iova < start && entry.iova < end) {
             if (dma_avail > 0 || entry.perm == IOMMU_NONE) {
-                dma_avail = s390_pci_update_iotlb(iommu, &entry);
+                dma_avail = s390_pci_update_iotlb(pbdev, &entry);
                 entry.iova += TARGET_PAGE_SIZE;
                 entry.translated_addr += TARGET_PAGE_SIZE;
             } else {
@@ -801,7 +802,7 @@ int rpcit_service_call(S390CPU *cpu, uint8_t r1, uint8_t r2, uintptr_t ra)
     }
     if (coalesce) {
             /* Unleash the coalesced unmap before finishing rpcit */
-            s390_pci_batch_unmap(iommu, iova, coalesce);
+            s390_pci_batch_unmap(pbdev, iova, coalesce);
             coalesce = 0;
     }
     if (again && dma_avail > 0)
@@ -1031,7 +1032,7 @@ static int reg_ioat(CPUS390XState *env, S390PCIBusDevice *pbdev, ZpciFib fib,
     iommu->g_iota = g_iota;
 
     if (t) {
-        s390_pci_iommu_enable(iommu);
+        s390_pci_iommu_enable(pbdev);
     } else {
         s390_pci_iommu_direct_map_enable(iommu);
     }
@@ -1039,9 +1040,10 @@ static int reg_ioat(CPUS390XState *env, S390PCIBusDevice *pbdev, ZpciFib fib,
     return 0;
 }
 
-void pci_dereg_ioat(S390PCIIOMMU *iommu)
+void pci_dereg_ioat(S390PCIBusDevice *pbdev)
 {
-    s390_pci_iommu_disable(iommu);
+    S390PCIIOMMU *iommu = pbdev->iommu;
+    s390_pci_iommu_disable(pbdev);
     iommu->pba = 0;
     iommu->pal = 0;
     iommu->g_iota = 0;
@@ -1265,7 +1267,7 @@ int mpcifc_service_call(S390CPU *cpu, uint8_t r1, uint64_t fiba, uint8_t ar,
             cc = ZPCI_PCI_LS_ERR;
             s390_set_status_code(env, r1, ZPCI_MOD_ST_SEQUENCE);
         } else {
-            pci_dereg_ioat(pbdev->iommu);
+            pci_dereg_ioat(pbdev);
         }
         break;
     case ZPCI_MOD_FC_REREG_IOAT:
@@ -1276,7 +1278,7 @@ int mpcifc_service_call(S390CPU *cpu, uint8_t r1, uint64_t fiba, uint8_t ar,
             cc = ZPCI_PCI_LS_ERR;
             s390_set_status_code(env, r1, ZPCI_MOD_ST_SEQUENCE);
         } else {
-            pci_dereg_ioat(pbdev->iommu);
+            pci_dereg_ioat(pbdev);
             if (reg_ioat(env, pbdev, fib, ra)) {
                 cc = ZPCI_PCI_LS_ERR;
                 s390_set_status_code(env, r1, ZPCI_MOD_ST_INSUF_RES);
diff --git a/include/hw/s390x/s390-pci-bus.h b/include/hw/s390x/s390-pci-bus.h
index eb15cb8b2d..a71f562dfc 100644
--- a/include/hw/s390x/s390-pci-bus.h
+++ b/include/hw/s390x/s390-pci-bus.h
@@ -276,7 +276,6 @@ struct S390PCIIOMMU {
     S390PCIBusDevice *pbdev;
     AddressSpace as;
     MemoryRegion mr;
-    IOMMUMemoryRegion iommu_mr;
     MemoryRegion *dm_mr;
     bool enabled;
     uint64_t g_iota;
@@ -354,6 +353,7 @@ struct S390PCIBusDevice {
     S390MsixInfo msix;
     AdapterRoutes routes;
     S390PCIIOMMU *iommu;
+    IOMMUMemoryRegion iommu_mr;
     MemoryRegion msix_notify_mr;
     IndAddr *summary_ind;
     IndAddr *indicator;
@@ -391,9 +391,9 @@ int pci_chsc_sei_nt2_have_event(void);
 void s390_pci_sclp_configure(SCCB *sccb);
 void s390_pci_sclp_deconfigure(SCCB *sccb);
 bool s390_pci_is_translation_enabled(uint64_t g_iota);
-void s390_pci_iommu_enable(S390PCIIOMMU *iommu);
+void s390_pci_iommu_enable(S390PCIBusDevice *pbdev);
 void s390_pci_iommu_direct_map_enable(S390PCIIOMMU *iommu);
-void s390_pci_iommu_disable(S390PCIIOMMU *iommu);
+void s390_pci_iommu_disable(S390PCIBusDevice *pbdev);
 void s390_pci_generate_error_event(uint16_t pec, uint32_t fh, uint32_t fid,
                                    uint64_t faddr, uint32_t e);
 uint16_t s390_guest_io_table_walk(uint64_t g_iota, hwaddr addr,
diff --git a/include/hw/s390x/s390-pci-inst.h b/include/hw/s390x/s390-pci-inst.h
index c782990e3b..38268c256e 100644
--- a/include/hw/s390x/s390-pci-inst.h
+++ b/include/hw/s390x/s390-pci-inst.h
@@ -99,7 +99,7 @@ typedef struct ZpciFib {
 } QEMU_PACKED ZpciFib;
 
 int pci_dereg_irqs(S390PCIBusDevice *pbdev);
-void pci_dereg_ioat(S390PCIIOMMU *iommu);
+void pci_dereg_ioat(S390PCIBusDevice *pbdev);
 int clp_service_call(S390CPU *cpu, uint8_t r2, uintptr_t ra);
 int pcilg_service_call(S390CPU *cpu, uint8_t r1, uint8_t r2, uintptr_t ra);
 int pcistg_service_call(S390CPU *cpu, uint8_t r1, uint8_t r2, uintptr_t ra);
@@ -111,7 +111,7 @@ int mpcifc_service_call(S390CPU *cpu, uint8_t r1, uint64_t fiba, uint8_t ar,
 int stpcifc_service_call(S390CPU *cpu, uint8_t r1, uint64_t fiba, uint8_t ar,
                          uintptr_t ra);
 void fmb_timer_free(S390PCIBusDevice *pbdev);
-uint32_t s390_pci_update_iotlb(S390PCIIOMMU *iommu, S390IOTLBEntry *entry);
+uint32_t s390_pci_update_iotlb(S390PCIBusDevice *pbdev, S390IOTLBEntry *entry);
 
 #define ZPCI_IO_BAR_MIN 0
 #define ZPCI_IO_BAR_MAX 5
-- 
2.34.1

[PATCH v3 04/15] s390x/pci: Move dm_mr from S390PCIIOMMU to S390PCIBusDevice

[Konstantin Shkolnyy]

This field is only used when S390PCIBusDevice exists, so it can be moved
there to simplify S390PCIIOMMU which purpose is just to store the "root"
AddressSpace.

Signed-off-by: Konstantin Shkolnyy <kshk@linux.ibm.com>
---
 hw/s390x/s390-pci-bus.c         | 21 +++++++++++----------
 hw/s390x/s390-pci-inst.c        |  2 +-
 include/hw/s390x/s390-pci-bus.h |  4 ++--
 3 files changed, 14 insertions(+), 13 deletions(-)

diff --git a/hw/s390x/s390-pci-bus.c b/hw/s390x/s390-pci-bus.c
index 98f626b57f..f09d85da8f 100644
--- a/hw/s390x/s390-pci-bus.c
+++ b/hw/s390x/s390-pci-bus.c
@@ -604,7 +604,7 @@ static void s390_pci_ioat_replay(S390PCIBusDevice *pbdev)
     curr = iommu->pba;
     end = iommu->pal;
 
-    if (iommu->dm_mr) {
+    if (pbdev->dm_mr) {
         /* If direct mapping is used, there are no guest tables to replay */
         return;
     }
@@ -790,8 +790,9 @@ void s390_pci_iommu_enable(S390PCIBusDevice *pbdev)
     g_free(name);
 }
 
-void s390_pci_iommu_direct_map_enable(S390PCIIOMMU *iommu)
+void s390_pci_iommu_direct_map_enable(S390PCIBusDevice *pbdev)
 {
+    S390PCIIOMMU *iommu = pbdev->iommu;
     MachineState *ms = MACHINE(qdev_get_machine());
     S390CcwMachineState *s390ms = S390_CCW_MACHINE(ms);
 
@@ -803,13 +804,13 @@ void s390_pci_iommu_direct_map_enable(S390PCIIOMMU *iommu)
     g_autofree char *name = g_strdup_printf("iommu-dm-s390-%04x",
                                             iommu->pbdev->uid);
 
-    iommu->dm_mr = g_malloc0(sizeof(*iommu->dm_mr));
-    memory_region_init_alias(iommu->dm_mr, OBJECT(&iommu->mr), name,
+    pbdev->dm_mr = g_malloc0(sizeof(*pbdev->dm_mr));
+    memory_region_init_alias(pbdev->dm_mr, OBJECT(&iommu->mr), name,
                              get_system_memory(), 0,
                              s390_get_memory_limit(s390ms));
     iommu->enabled = true;
     memory_region_add_subregion(&iommu->mr, iommu->pbdev->zpci_fn.sdma,
-                                iommu->dm_mr);
+                                pbdev->dm_mr);
 }
 
 void s390_pci_iommu_disable(S390PCIBusDevice *pbdev)
@@ -817,11 +818,11 @@ void s390_pci_iommu_disable(S390PCIBusDevice *pbdev)
     S390PCIIOMMU *iommu = pbdev->iommu;
     iommu->enabled = false;
     g_hash_table_remove_all(iommu->iotlb);
-    if (iommu->dm_mr) {
-        memory_region_del_subregion(&iommu->mr, iommu->dm_mr);
-        object_unparent(OBJECT(iommu->dm_mr));
-        g_free(iommu->dm_mr);
-        iommu->dm_mr = NULL;
+    if (pbdev->dm_mr) {
+        memory_region_del_subregion(&iommu->mr, pbdev->dm_mr);
+        object_unparent(OBJECT(pbdev->dm_mr));
+        g_free(pbdev->dm_mr);
+        pbdev->dm_mr = NULL;
     } else {
         memory_region_del_subregion(&iommu->mr,
                                     MEMORY_REGION(&pbdev->iommu_mr));
diff --git a/hw/s390x/s390-pci-inst.c b/hw/s390x/s390-pci-inst.c
index ed60e6100e..834a0aec95 100644
--- a/hw/s390x/s390-pci-inst.c
+++ b/hw/s390x/s390-pci-inst.c
@@ -1034,7 +1034,7 @@ static int reg_ioat(CPUS390XState *env, S390PCIBusDevice *pbdev, ZpciFib fib,
     if (t) {
         s390_pci_iommu_enable(pbdev);
     } else {
-        s390_pci_iommu_direct_map_enable(iommu);
+        s390_pci_iommu_direct_map_enable(pbdev);
     }
 
     return 0;
diff --git a/include/hw/s390x/s390-pci-bus.h b/include/hw/s390x/s390-pci-bus.h
index a71f562dfc..f310672584 100644
--- a/include/hw/s390x/s390-pci-bus.h
+++ b/include/hw/s390x/s390-pci-bus.h
@@ -276,7 +276,6 @@ struct S390PCIIOMMU {
     S390PCIBusDevice *pbdev;
     AddressSpace as;
     MemoryRegion mr;
-    MemoryRegion *dm_mr;
     bool enabled;
     uint64_t g_iota;
     uint64_t pba;
@@ -354,6 +353,7 @@ struct S390PCIBusDevice {
     AdapterRoutes routes;
     S390PCIIOMMU *iommu;
     IOMMUMemoryRegion iommu_mr;
+    MemoryRegion *dm_mr;
     MemoryRegion msix_notify_mr;
     IndAddr *summary_ind;
     IndAddr *indicator;
@@ -392,7 +392,7 @@ void s390_pci_sclp_configure(SCCB *sccb);
 void s390_pci_sclp_deconfigure(SCCB *sccb);
 bool s390_pci_is_translation_enabled(uint64_t g_iota);
 void s390_pci_iommu_enable(S390PCIBusDevice *pbdev);
-void s390_pci_iommu_direct_map_enable(S390PCIIOMMU *iommu);
+void s390_pci_iommu_direct_map_enable(S390PCIBusDevice *pbdev);
 void s390_pci_iommu_disable(S390PCIBusDevice *pbdev);
 void s390_pci_generate_error_event(uint16_t pec, uint32_t fh, uint32_t fid,
                                    uint64_t faddr, uint32_t e);
-- 
2.34.1

[PATCH v3 05/15] s390x/pci: Move iotlb from S390PCIIOMMU to S390PCIBusDevice

[Konstantin Shkolnyy]

This field is only used when S390PCIBusDevice exists, so it can be moved
there to simplify S390PCIIOMMU which purpose is just to store the "root"
AddressSpace.

Signed-off-by: Konstantin Shkolnyy <kshk@linux.ibm.com>
---
 hw/s390x/s390-pci-bus.c         | 10 +++++-----
 hw/s390x/s390-pci-inst.c        |  6 +++---
 include/hw/s390x/s390-pci-bus.h |  2 +-
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/hw/s390x/s390-pci-bus.c b/hw/s390x/s390-pci-bus.c
index f09d85da8f..34dcc7c3bd 100644
--- a/hw/s390x/s390-pci-bus.c
+++ b/hw/s390x/s390-pci-bus.c
@@ -569,7 +569,7 @@ static IOMMUTLBEntry s390_translate_iommu(IOMMUMemoryRegion *mr, hwaddr addr,
         goto err;
     }
 
-    entry = g_hash_table_lookup(iommu->iotlb, &iova);
+    entry = g_hash_table_lookup(pbdev->iotlb, &iova);
     if (entry) {
         ret.iova = entry->iova;
         ret.translated_addr = entry->translated_addr;
@@ -681,8 +681,6 @@ static S390PCIIOMMU *s390_pci_get_iommu(S390pciState *s, PCIBus *bus,
                                         PCI_FUNC(devfn));
         memory_region_init(&iommu->mr, OBJECT(iommu), mr_name, UINT64_MAX);
         address_space_init(&iommu->as, &iommu->mr, as_name);
-        iommu->iotlb = g_hash_table_new_full(g_int64_hash, g_int64_equal,
-                                             NULL, g_free);
         table->iommu[PCI_SLOT(devfn)] = iommu;
 
         g_free(mr_name);
@@ -817,7 +815,7 @@ void s390_pci_iommu_disable(S390PCIBusDevice *pbdev)
 {
     S390PCIIOMMU *iommu = pbdev->iommu;
     iommu->enabled = false;
-    g_hash_table_remove_all(iommu->iotlb);
+    g_hash_table_remove_all(pbdev->iotlb);
     if (pbdev->dm_mr) {
         memory_region_del_subregion(&iommu->mr, pbdev->dm_mr);
         object_unparent(OBJECT(pbdev->dm_mr));
@@ -841,7 +839,6 @@ static void s390_pci_iommu_free(S390pciState *s, PCIBus *bus, int32_t devfn)
     }
 
     table->iommu[PCI_SLOT(devfn)] = NULL;
-    g_hash_table_destroy(iommu->iotlb);
     /*
      * An attached PCI device may have memory listeners, eg. VFIO PCI.
      * The associated subregion will already have been unmapped in
@@ -1263,6 +1260,8 @@ static void s390_pcihost_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
         /* the allocated idx is actually getting used */
         s->next_idx = (pbdev->idx + 1) & FH_MASK_INDEX;
         pbdev->fh = pbdev->idx;
+        pbdev->iotlb = g_hash_table_new_full(g_int64_hash, g_int64_equal,
+                                             NULL, g_free);
         QTAILQ_INSERT_TAIL(&s->zpci_devs, pbdev, link);
         g_hash_table_insert(s->zpci_table, &pbdev->idx, pbdev);
     } else {
@@ -1305,6 +1304,7 @@ static void s390_pcihost_unplug(HotplugHandler *hotplug_dev, DeviceState *dev,
         if (pbdev->iommu && pbdev->iommu->dma_limit) {
             s390_pci_end_dma_count(s, pbdev->iommu->dma_limit);
         }
+        g_hash_table_destroy(pbdev->iotlb);
         qdev_unrealize(dev);
     }
 }
diff --git a/hw/s390x/s390-pci-inst.c b/hw/s390x/s390-pci-inst.c
index 834a0aec95..8349b44dd4 100644
--- a/hw/s390x/s390-pci-inst.c
+++ b/hw/s390x/s390-pci-inst.c
@@ -617,7 +617,7 @@ uint32_t s390_pci_update_iotlb(S390PCIBusDevice *pbdev,
                                S390IOTLBEntry *entry)
 {
     S390PCIIOMMU *iommu = pbdev->iommu;
-    S390IOTLBEntry *cache = g_hash_table_lookup(iommu->iotlb, &entry->iova);
+    S390IOTLBEntry *cache = g_hash_table_lookup(pbdev->iotlb, &entry->iova);
     IOMMUTLBEvent event = {
         .type = entry->perm ? IOMMU_NOTIFIER_MAP : IOMMU_NOTIFIER_UNMAP,
         .entry = {
@@ -633,7 +633,7 @@ uint32_t s390_pci_update_iotlb(S390PCIBusDevice *pbdev,
         if (!cache) {
             goto out;
         }
-        g_hash_table_remove(iommu->iotlb, &entry->iova);
+        g_hash_table_remove(pbdev->iotlb, &entry->iova);
         inc_dma_avail(iommu);
         /* Don't notify the iommu yet, maybe we can bundle contiguous unmaps */
         goto out;
@@ -656,7 +656,7 @@ uint32_t s390_pci_update_iotlb(S390PCIBusDevice *pbdev,
         cache->translated_addr = entry->translated_addr;
         cache->len = TARGET_PAGE_SIZE;
         cache->perm = entry->perm;
-        g_hash_table_replace(iommu->iotlb, &cache->iova, cache);
+        g_hash_table_replace(pbdev->iotlb, &cache->iova, cache);
         dec_dma_avail(iommu);
     }
 
diff --git a/include/hw/s390x/s390-pci-bus.h b/include/hw/s390x/s390-pci-bus.h
index f310672584..d9bafd6859 100644
--- a/include/hw/s390x/s390-pci-bus.h
+++ b/include/hw/s390x/s390-pci-bus.h
@@ -281,7 +281,6 @@ struct S390PCIIOMMU {
     uint64_t pba;
     uint64_t pal;
     uint64_t max_dma_limit;
-    GHashTable *iotlb;
     S390PCIDMACount *dma_limit;
 };
 
@@ -354,6 +353,7 @@ struct S390PCIBusDevice {
     S390PCIIOMMU *iommu;
     IOMMUMemoryRegion iommu_mr;
     MemoryRegion *dm_mr;
+    GHashTable *iotlb;
     MemoryRegion msix_notify_mr;
     IndAddr *summary_ind;
     IndAddr *indicator;
-- 
2.34.1

[PATCH v3 06/15] s390x/pci: Remove a ptr to S390PCIBusDevice from S390PCIIOMMU

[Konstantin Shkolnyy]

This pointer is no longer used, after fields were moved from S390PCIIOMMU
to S390PCIBusDevice.

Signed-off-by: Konstantin Shkolnyy <kshk@linux.ibm.com>
---
 hw/s390x/s390-pci-bus.c         | 15 +++++++--------
 include/hw/s390x/s390-pci-bus.h |  1 -
 2 files changed, 7 insertions(+), 9 deletions(-)

diff --git a/hw/s390x/s390-pci-bus.c b/hw/s390x/s390-pci-bus.c
index 34dcc7c3bd..21455d9233 100644
--- a/hw/s390x/s390-pci-bus.c
+++ b/hw/s390x/s390-pci-bus.c
@@ -551,7 +551,7 @@ static IOMMUTLBEntry s390_translate_iommu(IOMMUMemoryRegion *mr, hwaddr addr,
         .perm = IOMMU_NONE,
     };
 
-    switch (iommu->pbdev->state) {
+    switch (pbdev->state) {
     case ZPCI_FS_ENABLED:
     case ZPCI_FS_BLOCKED:
         if (!iommu->enabled) {
@@ -586,9 +586,9 @@ static IOMMUTLBEntry s390_translate_iommu(IOMMUMemoryRegion *mr, hwaddr addr,
     }
 err:
     if (error) {
-        iommu->pbdev->state = ZPCI_FS_ERROR;
-        s390_pci_generate_error_event(error, iommu->pbdev->fh,
-                                      iommu->pbdev->fid, addr, 0);
+        pbdev->state = ZPCI_FS_ERROR;
+        s390_pci_generate_error_event(error, pbdev->fh,
+                                      pbdev->fid, addr, 0);
     }
     return ret;
 }
@@ -779,7 +779,7 @@ void s390_pci_iommu_enable(S390PCIBusDevice *pbdev)
      * so the smallest IOMMU region we can define runs from 0 to the end
      * of the PCI address space.
      */
-    char *name = g_strdup_printf("iommu-s390-%04x", iommu->pbdev->uid);
+    char *name = g_strdup_printf("iommu-s390-%04x", pbdev->uid);
     memory_region_init_iommu(&pbdev->iommu_mr, sizeof(pbdev->iommu_mr),
                              TYPE_S390_IOMMU_MEMORY_REGION, OBJECT(&iommu->mr),
                              name, iommu->pal + 1);
@@ -800,14 +800,14 @@ void s390_pci_iommu_direct_map_enable(S390PCIBusDevice *pbdev)
      * IOVA X + SDMA.  VFIO will handle pinning via its memory listener.
      */
     g_autofree char *name = g_strdup_printf("iommu-dm-s390-%04x",
-                                            iommu->pbdev->uid);
+                                            pbdev->uid);
 
     pbdev->dm_mr = g_malloc0(sizeof(*pbdev->dm_mr));
     memory_region_init_alias(pbdev->dm_mr, OBJECT(&iommu->mr), name,
                              get_system_memory(), 0,
                              s390_get_memory_limit(s390ms));
     iommu->enabled = true;
-    memory_region_add_subregion(&iommu->mr, iommu->pbdev->zpci_fn.sdma,
+    memory_region_add_subregion(&iommu->mr, pbdev->zpci_fn.sdma,
                                 pbdev->dm_mr);
 }
 
@@ -1199,7 +1199,6 @@ static void s390_pcihost_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
 
         pbdev->pdev = pdev;
         pbdev->iommu = s390_pci_get_iommu(s, pci_get_bus(pdev), pdev->devfn);
-        pbdev->iommu->pbdev = pbdev;
         pbdev->state = ZPCI_FS_DISABLED;
         set_pbdev_info(pbdev);
 
diff --git a/include/hw/s390x/s390-pci-bus.h b/include/hw/s390x/s390-pci-bus.h
index d9bafd6859..fcaaf19512 100644
--- a/include/hw/s390x/s390-pci-bus.h
+++ b/include/hw/s390x/s390-pci-bus.h
@@ -273,7 +273,6 @@ typedef struct S390PCIDMACount {
 
 struct S390PCIIOMMU {
     Object parent_obj;
-    S390PCIBusDevice *pbdev;
     AddressSpace as;
     MemoryRegion mr;
     bool enabled;
-- 
2.34.1

[PATCH v3 07/15] s390x/pci: Move/rename enabled from S390PCIIOMMU to S390PCIBusDevice

[Konstantin Shkolnyy]

This field is only used when S390PCIBusDevice exists, so it can be moved
there to simplify S390PCIIOMMU which purpose is just to store the "root"
AddressSpace.

This also allows to save/restore this field during migration.

Signed-off-by: Konstantin Shkolnyy <kshk@linux.ibm.com>
---
 hw/s390x/s390-pci-bus.c         | 14 +++++++-------
 hw/s390x/s390-pci-inst.c        |  8 ++++----
 include/hw/s390x/s390-pci-bus.h |  2 +-
 3 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/hw/s390x/s390-pci-bus.c b/hw/s390x/s390-pci-bus.c
index 21455d9233..ccbccc8972 100644
--- a/hw/s390x/s390-pci-bus.c
+++ b/hw/s390x/s390-pci-bus.c
@@ -205,7 +205,7 @@ void s390_pci_sclp_deconfigure(SCCB *sccb)
         } else if (pbdev->summary_ind) {
             pci_dereg_irqs(pbdev);
         }
-        if (pbdev->iommu->enabled) {
+        if (pbdev->iommu_enabled) {
             pci_dereg_ioat(pbdev);
         }
         pbdev->state = ZPCI_FS_STANDBY;
@@ -554,7 +554,7 @@ static IOMMUTLBEntry s390_translate_iommu(IOMMUMemoryRegion *mr, hwaddr addr,
     switch (pbdev->state) {
     case ZPCI_FS_ENABLED:
     case ZPCI_FS_BLOCKED:
-        if (!iommu->enabled) {
+        if (!pbdev->iommu_enabled) {
             return ret;
         }
         break;
@@ -783,7 +783,7 @@ void s390_pci_iommu_enable(S390PCIBusDevice *pbdev)
     memory_region_init_iommu(&pbdev->iommu_mr, sizeof(pbdev->iommu_mr),
                              TYPE_S390_IOMMU_MEMORY_REGION, OBJECT(&iommu->mr),
                              name, iommu->pal + 1);
-    iommu->enabled = true;
+    pbdev->iommu_enabled = true;
     memory_region_add_subregion(&iommu->mr, 0, MEMORY_REGION(&pbdev->iommu_mr));
     g_free(name);
 }
@@ -806,7 +806,7 @@ void s390_pci_iommu_direct_map_enable(S390PCIBusDevice *pbdev)
     memory_region_init_alias(pbdev->dm_mr, OBJECT(&iommu->mr), name,
                              get_system_memory(), 0,
                              s390_get_memory_limit(s390ms));
-    iommu->enabled = true;
+    pbdev->iommu_enabled = true;
     memory_region_add_subregion(&iommu->mr, pbdev->zpci_fn.sdma,
                                 pbdev->dm_mr);
 }
@@ -814,7 +814,7 @@ void s390_pci_iommu_direct_map_enable(S390PCIBusDevice *pbdev)
 void s390_pci_iommu_disable(S390PCIBusDevice *pbdev)
 {
     S390PCIIOMMU *iommu = pbdev->iommu;
-    iommu->enabled = false;
+    pbdev->iommu_enabled = false;
     g_hash_table_remove_all(pbdev->iotlb);
     if (pbdev->dm_mr) {
         memory_region_del_subregion(&iommu->mr, pbdev->dm_mr);
@@ -1423,7 +1423,7 @@ static void s390_pcihost_reset(DeviceState *dev)
             } else if (pbdev->summary_ind) {
                 pci_dereg_irqs(pbdev);
             }
-            if (pbdev->iommu->enabled) {
+            if (pbdev->iommu_enabled) {
                 pci_dereg_ioat(pbdev);
             }
             pbdev->state = ZPCI_FS_STANDBY;
@@ -1564,7 +1564,7 @@ static void s390_pci_device_reset(DeviceState *dev)
     } else if (pbdev->summary_ind) {
         pci_dereg_irqs(pbdev);
     }
-    if (pbdev->iommu->enabled) {
+    if (pbdev->iommu_enabled) {
         pci_dereg_ioat(pbdev);
     }
 
diff --git a/hw/s390x/s390-pci-inst.c b/hw/s390x/s390-pci-inst.c
index 8349b44dd4..4458eae681 100644
--- a/hw/s390x/s390-pci-inst.c
+++ b/hw/s390x/s390-pci-inst.c
@@ -1251,7 +1251,7 @@ int mpcifc_service_call(S390CPU *cpu, uint8_t r1, uint64_t fiba, uint8_t ar,
         if (dmaas != 0) {
             cc = ZPCI_PCI_LS_ERR;
             s390_set_status_code(env, r1, ZPCI_MOD_ST_DMAAS_INVAL);
-        } else if (pbdev->iommu->enabled) {
+        } else if (pbdev->iommu_enabled) {
             cc = ZPCI_PCI_LS_ERR;
             s390_set_status_code(env, r1, ZPCI_MOD_ST_SEQUENCE);
         } else if (reg_ioat(env, pbdev, fib, ra)) {
@@ -1263,7 +1263,7 @@ int mpcifc_service_call(S390CPU *cpu, uint8_t r1, uint64_t fiba, uint8_t ar,
         if (dmaas != 0) {
             cc = ZPCI_PCI_LS_ERR;
             s390_set_status_code(env, r1, ZPCI_MOD_ST_DMAAS_INVAL);
-        } else if (!pbdev->iommu->enabled) {
+        } else if (!pbdev->iommu_enabled) {
             cc = ZPCI_PCI_LS_ERR;
             s390_set_status_code(env, r1, ZPCI_MOD_ST_SEQUENCE);
         } else {
@@ -1274,7 +1274,7 @@ int mpcifc_service_call(S390CPU *cpu, uint8_t r1, uint64_t fiba, uint8_t ar,
         if (dmaas != 0) {
             cc = ZPCI_PCI_LS_ERR;
             s390_set_status_code(env, r1, ZPCI_MOD_ST_DMAAS_INVAL);
-        } else if (!pbdev->iommu->enabled) {
+        } else if (!pbdev->iommu_enabled) {
             cc = ZPCI_PCI_LS_ERR;
             s390_set_status_code(env, r1, ZPCI_MOD_ST_SEQUENCE);
         } else {
@@ -1404,7 +1404,7 @@ int stpcifc_service_call(S390CPU *cpu, uint8_t r1, uint64_t fiba, uint8_t ar,
         /* fallthrough */
     case ZPCI_FS_ENABLED:
         fib.fc |= 0x80;
-        if (pbdev->iommu->enabled) {
+        if (pbdev->iommu_enabled) {
             fib.fc |= 0x10;
         }
         if (!(fh & FH_MASK_ENABLE)) {
diff --git a/include/hw/s390x/s390-pci-bus.h b/include/hw/s390x/s390-pci-bus.h
index fcaaf19512..24f6e60786 100644
--- a/include/hw/s390x/s390-pci-bus.h
+++ b/include/hw/s390x/s390-pci-bus.h
@@ -275,7 +275,6 @@ struct S390PCIIOMMU {
     Object parent_obj;
     AddressSpace as;
     MemoryRegion mr;
-    bool enabled;
     uint64_t g_iota;
     uint64_t pba;
     uint64_t pal;
@@ -350,6 +349,7 @@ struct S390PCIBusDevice {
     S390MsixInfo msix;
     AdapterRoutes routes;
     S390PCIIOMMU *iommu;
+    bool iommu_enabled;
     IOMMUMemoryRegion iommu_mr;
     MemoryRegion *dm_mr;
     GHashTable *iotlb;
-- 
2.34.1

[PATCH v3 08/15] s390x/pci: Move dma_limit from S390PCIIOMMU to S390PCIBusDevice

[Konstantin Shkolnyy]

This field is only used when S390PCIBusDevice exists, so it can be moved
there to simplify S390PCIIOMMU which purpose is just to store the "root"
AddressSpace.

Signed-off-by: Konstantin Shkolnyy <kshk@linux.ibm.com>
---
 hw/s390x/s390-pci-bus.c         | 10 +++++-----
 hw/s390x/s390-pci-inst.c        | 23 +++++++++++------------
 include/hw/s390x/s390-pci-bus.h |  2 +-
 3 files changed, 17 insertions(+), 18 deletions(-)

diff --git a/hw/s390x/s390-pci-bus.c b/hw/s390x/s390-pci-bus.c
index ccbccc8972..f70aaf4425 100644
--- a/hw/s390x/s390-pci-bus.c
+++ b/hw/s390x/s390-pci-bus.c
@@ -609,8 +609,8 @@ static void s390_pci_ioat_replay(S390PCIBusDevice *pbdev)
         return;
     }
 
-    if (iommu->dma_limit) {
-        dma_avail = iommu->dma_limit->avail;
+    if (pbdev->dma_limit) {
+        dma_avail = pbdev->dma_limit->avail;
     } else {
         dma_avail = 1;
     }
@@ -1222,7 +1222,7 @@ static void s390_pcihost_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
                     pbdev->forwarding_assist = false;
                 }
             }
-            pbdev->iommu->dma_limit = s390_pci_start_dma_count(s, pbdev);
+            pbdev->dma_limit = s390_pci_start_dma_count(s, pbdev);
             /* Fill in CLP information passed via the vfio region */
             s390_pci_get_clp_info(pbdev);
             if (!pbdev->interp) {
@@ -1300,8 +1300,8 @@ static void s390_pcihost_unplug(HotplugHandler *hotplug_dev, DeviceState *dev,
         pbdev->fid = 0;
         QTAILQ_REMOVE(&s->zpci_devs, pbdev, link);
         g_hash_table_remove(s->zpci_table, &pbdev->idx);
-        if (pbdev->iommu && pbdev->iommu->dma_limit) {
-            s390_pci_end_dma_count(s, pbdev->iommu->dma_limit);
+        if (pbdev->dma_limit) {
+            s390_pci_end_dma_count(s, pbdev->dma_limit);
         }
         g_hash_table_destroy(pbdev->iotlb);
         qdev_unrealize(dev);
diff --git a/hw/s390x/s390-pci-inst.c b/hw/s390x/s390-pci-inst.c
index 4458eae681..ac4e7f412d 100644
--- a/hw/s390x/s390-pci-inst.c
+++ b/hw/s390x/s390-pci-inst.c
@@ -28,17 +28,17 @@
 
 #include "trace.h"
 
-static inline void inc_dma_avail(S390PCIIOMMU *iommu)
+static inline void inc_dma_avail(S390PCIBusDevice *pbdev)
 {
-    if (iommu->dma_limit) {
-        iommu->dma_limit->avail++;
+    if (pbdev->dma_limit) {
+        pbdev->dma_limit->avail++;
     }
 }
 
-static inline void dec_dma_avail(S390PCIIOMMU *iommu)
+static inline void dec_dma_avail(S390PCIBusDevice *pbdev)
 {
-    if (iommu->dma_limit) {
-        iommu->dma_limit->avail--;
+    if (pbdev->dma_limit) {
+        pbdev->dma_limit->avail--;
     }
 }
 
@@ -616,7 +616,6 @@ int pcistg_service_call(S390CPU *cpu, uint8_t r1, uint8_t r2, uintptr_t ra)
 uint32_t s390_pci_update_iotlb(S390PCIBusDevice *pbdev,
                                S390IOTLBEntry *entry)
 {
-    S390PCIIOMMU *iommu = pbdev->iommu;
     S390IOTLBEntry *cache = g_hash_table_lookup(pbdev->iotlb, &entry->iova);
     IOMMUTLBEvent event = {
         .type = entry->perm ? IOMMU_NOTIFIER_MAP : IOMMU_NOTIFIER_UNMAP,
@@ -634,7 +633,7 @@ uint32_t s390_pci_update_iotlb(S390PCIBusDevice *pbdev,
             goto out;
         }
         g_hash_table_remove(pbdev->iotlb, &entry->iova);
-        inc_dma_avail(iommu);
+        inc_dma_avail(pbdev);
         /* Don't notify the iommu yet, maybe we can bundle contiguous unmaps */
         goto out;
     } else {
@@ -657,7 +656,7 @@ uint32_t s390_pci_update_iotlb(S390PCIBusDevice *pbdev,
         cache->len = TARGET_PAGE_SIZE;
         cache->perm = entry->perm;
         g_hash_table_replace(pbdev->iotlb, &cache->iova, cache);
-        dec_dma_avail(iommu);
+        dec_dma_avail(pbdev);
     }
 
     /*
@@ -667,7 +666,7 @@ uint32_t s390_pci_update_iotlb(S390PCIBusDevice *pbdev,
     memory_region_notify_iommu(&pbdev->iommu_mr, 0, event);
 
 out:
-    return iommu->dma_limit ? iommu->dma_limit->avail : 1;
+    return pbdev->dma_limit ? pbdev->dma_limit->avail : 1;
 }
 
 static void s390_pci_batch_unmap(S390PCIBusDevice *pbdev, uint64_t iova,
@@ -744,8 +743,8 @@ int rpcit_service_call(S390CPU *cpu, uint8_t r1, uint8_t r2, uintptr_t ra)
     }
 
     iommu = pbdev->iommu;
-    if (iommu->dma_limit) {
-        dma_avail = iommu->dma_limit->avail;
+    if (pbdev->dma_limit) {
+        dma_avail = pbdev->dma_limit->avail;
     } else {
         dma_avail = 1;
     }
diff --git a/include/hw/s390x/s390-pci-bus.h b/include/hw/s390x/s390-pci-bus.h
index 24f6e60786..126a50c98b 100644
--- a/include/hw/s390x/s390-pci-bus.h
+++ b/include/hw/s390x/s390-pci-bus.h
@@ -279,7 +279,6 @@ struct S390PCIIOMMU {
     uint64_t pba;
     uint64_t pal;
     uint64_t max_dma_limit;
-    S390PCIDMACount *dma_limit;
 };
 
 typedef struct S390PCIIOMMUTable {
@@ -353,6 +352,7 @@ struct S390PCIBusDevice {
     IOMMUMemoryRegion iommu_mr;
     MemoryRegion *dm_mr;
     GHashTable *iotlb;
+    S390PCIDMACount *dma_limit;
     MemoryRegion msix_notify_mr;
     IndAddr *summary_ind;
     IndAddr *indicator;
-- 
2.34.1

[PATCH v3 09/15] s390x/pci: Move g_iota from S390PCIIOMMU to S390PCIBusDevice

[Konstantin Shkolnyy]

This field is only used when S390PCIBusDevice exists, so it can be moved
there to simplify S390PCIIOMMU which purpose is just to store the "root"
AddressSpace.

This also allows to save/restore this field during migration.

Signed-off-by: Konstantin Shkolnyy <kshk@linux.ibm.com>
---
 hw/s390x/s390-pci-bus.c         |  2 +-
 hw/s390x/s390-pci-inst.c        | 10 +++++-----
 include/hw/s390x/s390-pci-bus.h |  2 +-
 3 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/hw/s390x/s390-pci-bus.c b/hw/s390x/s390-pci-bus.c
index f70aaf4425..3acd15ee0a 100644
--- a/hw/s390x/s390-pci-bus.c
+++ b/hw/s390x/s390-pci-bus.c
@@ -616,7 +616,7 @@ static void s390_pci_ioat_replay(S390PCIBusDevice *pbdev)
     }
 
     while (curr < end) {
-        error = s390_guest_io_table_walk(iommu->g_iota, curr, &entry);
+        error = s390_guest_io_table_walk(pbdev->g_iota, curr, &entry);
         if (error) {
             pbdev->state = ZPCI_FS_ERROR;
             s390_pci_generate_error_event(error, pbdev->fh, pbdev->fid, curr,
diff --git a/hw/s390x/s390-pci-inst.c b/hw/s390x/s390-pci-inst.c
index ac4e7f412d..f4ec299614 100644
--- a/hw/s390x/s390-pci-inst.c
+++ b/hw/s390x/s390-pci-inst.c
@@ -748,7 +748,7 @@ int rpcit_service_call(S390CPU *cpu, uint8_t r1, uint8_t r2, uintptr_t ra)
     } else {
         dma_avail = 1;
     }
-    if (!iommu->g_iota) {
+    if (!pbdev->g_iota) {
         error = ERR_EVENT_INVALAS;
         goto err;
     }
@@ -762,7 +762,7 @@ int rpcit_service_call(S390CPU *cpu, uint8_t r1, uint8_t r2, uintptr_t ra)
     start = sstart;
     again = false;
     while (start < end) {
-        error = s390_guest_io_table_walk(iommu->g_iota, start, &entry);
+        error = s390_guest_io_table_walk(pbdev->g_iota, start, &entry);
         if (error) {
             break;
         }
@@ -1028,7 +1028,7 @@ static int reg_ioat(CPUS390XState *env, S390PCIBusDevice *pbdev, ZpciFib fib,
 
     iommu->pba = pba;
     iommu->pal = pal;
-    iommu->g_iota = g_iota;
+    pbdev->g_iota = g_iota;
 
     if (t) {
         s390_pci_iommu_enable(pbdev);
@@ -1045,7 +1045,7 @@ void pci_dereg_ioat(S390PCIBusDevice *pbdev)
     s390_pci_iommu_disable(pbdev);
     iommu->pba = 0;
     iommu->pal = 0;
-    iommu->g_iota = 0;
+    pbdev->g_iota = 0;
 }
 
 void fmb_timer_free(S390PCIBusDevice *pbdev)
@@ -1418,7 +1418,7 @@ int stpcifc_service_call(S390CPU *cpu, uint8_t r1, uint64_t fiba, uint8_t ar,
 
     stq_be_p(&fib.pba, pbdev->iommu->pba);
     stq_be_p(&fib.pal, pbdev->iommu->pal);
-    stq_be_p(&fib.iota, pbdev->iommu->g_iota);
+    stq_be_p(&fib.iota, pbdev->g_iota);
     stq_be_p(&fib.aibv, pbdev->routes.adapter.ind_addr);
     stq_be_p(&fib.aisb, pbdev->routes.adapter.summary_addr);
     stq_be_p(&fib.fmb_addr, pbdev->fmb_addr);
diff --git a/include/hw/s390x/s390-pci-bus.h b/include/hw/s390x/s390-pci-bus.h
index 126a50c98b..a7d1f1ade2 100644
--- a/include/hw/s390x/s390-pci-bus.h
+++ b/include/hw/s390x/s390-pci-bus.h
@@ -275,7 +275,6 @@ struct S390PCIIOMMU {
     Object parent_obj;
     AddressSpace as;
     MemoryRegion mr;
-    uint64_t g_iota;
     uint64_t pba;
     uint64_t pal;
     uint64_t max_dma_limit;
@@ -352,6 +351,7 @@ struct S390PCIBusDevice {
     IOMMUMemoryRegion iommu_mr;
     MemoryRegion *dm_mr;
     GHashTable *iotlb;
+    uint64_t g_iota;
     S390PCIDMACount *dma_limit;
     MemoryRegion msix_notify_mr;
     IndAddr *summary_ind;
-- 
2.34.1

[PATCH v3 10/15] s390x/pci: Move pba from S390PCIIOMMU to S390PCIBusDevice

[Konstantin Shkolnyy]

This field is only used when S390PCIBusDevice exists, so it can be moved
there to simplify S390PCIIOMMU which purpose is just to store the "root"
AddressSpace.

This also allows to save/restore this field during migration.

Signed-off-by: Konstantin Shkolnyy <kshk@linux.ibm.com>
---
 hw/s390x/s390-pci-bus.c         | 4 ++--
 hw/s390x/s390-pci-inst.c        | 8 ++++----
 include/hw/s390x/s390-pci-bus.h | 2 +-
 3 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/hw/s390x/s390-pci-bus.c b/hw/s390x/s390-pci-bus.c
index 3acd15ee0a..b7d350606c 100644
--- a/hw/s390x/s390-pci-bus.c
+++ b/hw/s390x/s390-pci-bus.c
@@ -564,7 +564,7 @@ static IOMMUTLBEntry s390_translate_iommu(IOMMUMemoryRegion *mr, hwaddr addr,
 
     trace_s390_pci_iommu_xlate(addr);
 
-    if (addr < iommu->pba || addr > iommu->pal) {
+    if (addr < pbdev->pba || addr > iommu->pal) {
         error = ERR_EVENT_OORANGE;
         goto err;
     }
@@ -601,7 +601,7 @@ static void s390_pci_ioat_replay(S390PCIBusDevice *pbdev)
     hwaddr curr, end;
     S390PCIIOMMU *iommu = pbdev->iommu;
 
-    curr = iommu->pba;
+    curr = pbdev->pba;
     end = iommu->pal;
 
     if (pbdev->dm_mr) {
diff --git a/hw/s390x/s390-pci-inst.c b/hw/s390x/s390-pci-inst.c
index f4ec299614..7680d87080 100644
--- a/hw/s390x/s390-pci-inst.c
+++ b/hw/s390x/s390-pci-inst.c
@@ -753,7 +753,7 @@ int rpcit_service_call(S390CPU *cpu, uint8_t r1, uint8_t r2, uintptr_t ra)
         goto err;
     }
 
-    if (end < iommu->pba || start > iommu->pal) {
+    if (end < pbdev->pba || start > iommu->pal) {
         error = ERR_EVENT_OORANGE;
         goto err;
     }
@@ -1026,7 +1026,7 @@ static int reg_ioat(CPUS390XState *env, S390PCIBusDevice *pbdev, ZpciFib fib,
         return -EINVAL;
     }
 
-    iommu->pba = pba;
+    pbdev->pba = pba;
     iommu->pal = pal;
     pbdev->g_iota = g_iota;
 
@@ -1043,7 +1043,7 @@ void pci_dereg_ioat(S390PCIBusDevice *pbdev)
 {
     S390PCIIOMMU *iommu = pbdev->iommu;
     s390_pci_iommu_disable(pbdev);
-    iommu->pba = 0;
+    pbdev->pba = 0;
     iommu->pal = 0;
     pbdev->g_iota = 0;
 }
@@ -1416,7 +1416,7 @@ int stpcifc_service_call(S390CPU *cpu, uint8_t r1, uint64_t fiba, uint8_t ar,
         return 0;
     }
 
-    stq_be_p(&fib.pba, pbdev->iommu->pba);
+    stq_be_p(&fib.pba, pbdev->pba);
     stq_be_p(&fib.pal, pbdev->iommu->pal);
     stq_be_p(&fib.iota, pbdev->g_iota);
     stq_be_p(&fib.aibv, pbdev->routes.adapter.ind_addr);
diff --git a/include/hw/s390x/s390-pci-bus.h b/include/hw/s390x/s390-pci-bus.h
index a7d1f1ade2..42f39a5cf0 100644
--- a/include/hw/s390x/s390-pci-bus.h
+++ b/include/hw/s390x/s390-pci-bus.h
@@ -275,7 +275,6 @@ struct S390PCIIOMMU {
     Object parent_obj;
     AddressSpace as;
     MemoryRegion mr;
-    uint64_t pba;
     uint64_t pal;
     uint64_t max_dma_limit;
 };
@@ -352,6 +351,7 @@ struct S390PCIBusDevice {
     MemoryRegion *dm_mr;
     GHashTable *iotlb;
     uint64_t g_iota;
+    uint64_t pba;
     S390PCIDMACount *dma_limit;
     MemoryRegion msix_notify_mr;
     IndAddr *summary_ind;
-- 
2.34.1

[PATCH v3 11/15] s390x/pci: Move pal from S390PCIIOMMU to S390PCIBusDevice

[Konstantin Shkolnyy]

This field is only used when S390PCIBusDevice exists, so it can be moved
there to simplify S390PCIIOMMU which purpose is just to store the "root"
AddressSpace.

This also allows to save/restore this field during migration.

Signed-off-by: Konstantin Shkolnyy <kshk@linux.ibm.com>
---
 hw/s390x/s390-pci-bus.c         |  8 +++-----
 hw/s390x/s390-pci-inst.c        | 12 ++++--------
 include/hw/s390x/s390-pci-bus.h |  2 +-
 3 files changed, 8 insertions(+), 14 deletions(-)

diff --git a/hw/s390x/s390-pci-bus.c b/hw/s390x/s390-pci-bus.c
index b7d350606c..8505abcc90 100644
--- a/hw/s390x/s390-pci-bus.c
+++ b/hw/s390x/s390-pci-bus.c
@@ -539,7 +539,6 @@ static IOMMUTLBEntry s390_translate_iommu(IOMMUMemoryRegion *mr, hwaddr addr,
                                           IOMMUAccessFlags flag, int iommu_idx)
 {
     S390PCIBusDevice *pbdev = container_of(mr, S390PCIBusDevice, iommu_mr);
-    S390PCIIOMMU *iommu = pbdev->iommu;
     S390IOTLBEntry *entry;
     uint64_t iova = addr & TARGET_PAGE_MASK;
     uint16_t error = 0;
@@ -564,7 +563,7 @@ static IOMMUTLBEntry s390_translate_iommu(IOMMUMemoryRegion *mr, hwaddr addr,
 
     trace_s390_pci_iommu_xlate(addr);
 
-    if (addr < pbdev->pba || addr > iommu->pal) {
+    if (addr < pbdev->pba || addr > pbdev->pal) {
         error = ERR_EVENT_OORANGE;
         goto err;
     }
@@ -599,10 +598,9 @@ static void s390_pci_ioat_replay(S390PCIBusDevice *pbdev)
     uint16_t error = 0;
     uint32_t dma_avail;
     hwaddr curr, end;
-    S390PCIIOMMU *iommu = pbdev->iommu;
 
     curr = pbdev->pba;
-    end = iommu->pal;
+    end = pbdev->pal;
 
     if (pbdev->dm_mr) {
         /* If direct mapping is used, there are no guest tables to replay */
@@ -782,7 +780,7 @@ void s390_pci_iommu_enable(S390PCIBusDevice *pbdev)
     char *name = g_strdup_printf("iommu-s390-%04x", pbdev->uid);
     memory_region_init_iommu(&pbdev->iommu_mr, sizeof(pbdev->iommu_mr),
                              TYPE_S390_IOMMU_MEMORY_REGION, OBJECT(&iommu->mr),
-                             name, iommu->pal + 1);
+                             name, pbdev->pal + 1);
     pbdev->iommu_enabled = true;
     memory_region_add_subregion(&iommu->mr, 0, MEMORY_REGION(&pbdev->iommu_mr));
     g_free(name);
diff --git a/hw/s390x/s390-pci-inst.c b/hw/s390x/s390-pci-inst.c
index 7680d87080..191f549ba7 100644
--- a/hw/s390x/s390-pci-inst.c
+++ b/hw/s390x/s390-pci-inst.c
@@ -700,7 +700,6 @@ int rpcit_service_call(S390CPU *cpu, uint8_t r1, uint8_t r2, uintptr_t ra)
     uint32_t fh;
     uint16_t error = 0;
     S390PCIBusDevice *pbdev;
-    S390PCIIOMMU *iommu;
     S390IOTLBEntry entry;
     hwaddr start, end, sstart;
     uint32_t dma_avail;
@@ -742,7 +741,6 @@ int rpcit_service_call(S390CPU *cpu, uint8_t r1, uint8_t r2, uintptr_t ra)
         break;
     }
 
-    iommu = pbdev->iommu;
     if (pbdev->dma_limit) {
         dma_avail = pbdev->dma_limit->avail;
     } else {
@@ -753,7 +751,7 @@ int rpcit_service_call(S390CPU *cpu, uint8_t r1, uint8_t r2, uintptr_t ra)
         goto err;
     }
 
-    if (end < pbdev->pba || start > iommu->pal) {
+    if (end < pbdev->pba || start > pbdev->pal) {
         error = ERR_EVENT_OORANGE;
         goto err;
     }
@@ -1001,7 +999,6 @@ bool s390_pci_is_translation_enabled(uint64_t g_iota)
 static int reg_ioat(CPUS390XState *env, S390PCIBusDevice *pbdev, ZpciFib fib,
                     uintptr_t ra)
 {
-    S390PCIIOMMU *iommu = pbdev->iommu;
     uint64_t pba = ldq_be_p(&fib.pba);
     uint64_t pal = ldq_be_p(&fib.pal);
     uint64_t g_iota = ldq_be_p(&fib.iota);
@@ -1027,7 +1024,7 @@ static int reg_ioat(CPUS390XState *env, S390PCIBusDevice *pbdev, ZpciFib fib,
     }
 
     pbdev->pba = pba;
-    iommu->pal = pal;
+    pbdev->pal = pal;
     pbdev->g_iota = g_iota;
 
     if (t) {
@@ -1041,10 +1038,9 @@ static int reg_ioat(CPUS390XState *env, S390PCIBusDevice *pbdev, ZpciFib fib,
 
 void pci_dereg_ioat(S390PCIBusDevice *pbdev)
 {
-    S390PCIIOMMU *iommu = pbdev->iommu;
     s390_pci_iommu_disable(pbdev);
     pbdev->pba = 0;
-    iommu->pal = 0;
+    pbdev->pal = 0;
     pbdev->g_iota = 0;
 }
 
@@ -1417,7 +1413,7 @@ int stpcifc_service_call(S390CPU *cpu, uint8_t r1, uint64_t fiba, uint8_t ar,
     }
 
     stq_be_p(&fib.pba, pbdev->pba);
-    stq_be_p(&fib.pal, pbdev->iommu->pal);
+    stq_be_p(&fib.pal, pbdev->pal);
     stq_be_p(&fib.iota, pbdev->g_iota);
     stq_be_p(&fib.aibv, pbdev->routes.adapter.ind_addr);
     stq_be_p(&fib.aisb, pbdev->routes.adapter.summary_addr);
diff --git a/include/hw/s390x/s390-pci-bus.h b/include/hw/s390x/s390-pci-bus.h
index 42f39a5cf0..1f5d8d4bd4 100644
--- a/include/hw/s390x/s390-pci-bus.h
+++ b/include/hw/s390x/s390-pci-bus.h
@@ -275,7 +275,6 @@ struct S390PCIIOMMU {
     Object parent_obj;
     AddressSpace as;
     MemoryRegion mr;
-    uint64_t pal;
     uint64_t max_dma_limit;
 };
 
@@ -352,6 +351,7 @@ struct S390PCIBusDevice {
     GHashTable *iotlb;
     uint64_t g_iota;
     uint64_t pba;
+    uint64_t pal;
     S390PCIDMACount *dma_limit;
     MemoryRegion msix_notify_mr;
     IndAddr *summary_ind;
-- 
2.34.1

[PATCH v3 12/15] s390x/pci: Move max_dma_limit from S390PCIIOMMU to S390PCIBusDevice

[Konstantin Shkolnyy]

This field is only used when S390PCIBusDevice exists, so it can be moved
there to simplify S390PCIIOMMU which purpose is just to store the "root"
AddressSpace.

This also allows to save/restore this field during migration.

Signed-off-by: Konstantin Shkolnyy <kshk@linux.ibm.com>
---
 hw/s390x/s390-pci-vfio.c        | 4 ++--
 include/hw/s390x/s390-pci-bus.h | 2 +-
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/hw/s390x/s390-pci-vfio.c b/hw/s390x/s390-pci-vfio.c
index 7c754b656d..093743f2f0 100644
--- a/hw/s390x/s390-pci-vfio.c
+++ b/hw/s390x/s390-pci-vfio.c
@@ -90,7 +90,7 @@ S390PCIDMACount *s390_pci_start_dma_count(S390pciState *s,
     cnt->users = 1;
     cnt->avail = avail;
     QTAILQ_INSERT_TAIL(&s->zpci_dma_limit, cnt, link);
-    pbdev->iommu->max_dma_limit = avail;
+    pbdev->max_dma_limit = avail;
     return cnt;
 }
 
@@ -151,7 +151,7 @@ static void s390_pci_read_base(S390PCIBusDevice *pbdev,
      * to request that the guest free DMA mappings as necessary.
      */
     if (!pbdev->rtr_avail) {
-        vfio_size = pbdev->iommu->max_dma_limit << qemu_target_page_bits();
+        vfio_size = pbdev->max_dma_limit << qemu_target_page_bits();
         if (vfio_size > 0 && vfio_size < cap->end_dma - cap->start_dma + 1) {
             pbdev->zpci_fn.edma = cap->start_dma + vfio_size - 1;
         }
diff --git a/include/hw/s390x/s390-pci-bus.h b/include/hw/s390x/s390-pci-bus.h
index 1f5d8d4bd4..e3cbee2695 100644
--- a/include/hw/s390x/s390-pci-bus.h
+++ b/include/hw/s390x/s390-pci-bus.h
@@ -275,7 +275,6 @@ struct S390PCIIOMMU {
     Object parent_obj;
     AddressSpace as;
     MemoryRegion mr;
-    uint64_t max_dma_limit;
 };
 
 typedef struct S390PCIIOMMUTable {
@@ -352,6 +351,7 @@ struct S390PCIBusDevice {
     uint64_t g_iota;
     uint64_t pba;
     uint64_t pal;
+    uint64_t max_dma_limit;
     S390PCIDMACount *dma_limit;
     MemoryRegion msix_notify_mr;
     IndAddr *summary_ind;
-- 
2.34.1

[PATCH v3 13/15] s390x/pci: Add a comment explaining S390PCIIOMMU purpose

[Konstantin Shkolnyy]

Signed-off-by: Konstantin Shkolnyy <kshk@linux.ibm.com>
---
 include/hw/s390x/s390-pci-bus.h | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/include/hw/s390x/s390-pci-bus.h b/include/hw/s390x/s390-pci-bus.h
index e3cbee2695..2fab28e6e0 100644
--- a/include/hw/s390x/s390-pci-bus.h
+++ b/include/hw/s390x/s390-pci-bus.h
@@ -271,6 +271,14 @@ typedef struct S390PCIDMACount {
     QTAILQ_ENTRY(S390PCIDMACount) link;
 } S390PCIDMACount;
 
+/*
+ * This structure holds the PCI device AddressSpace that QEMU needs to link
+ * into its internal structures before the zPCI and PCI devices are fully
+ * initialized. It's a QEMU requirement to provide this "root" AddressSpace
+ * early. The AddressSpace is only actually used for I/O while the PCI
+ * device is plugged in and configured by the guest, at which time it gets
+ * additional memory subregions from zPCI device, that can do real work.
+ */
 struct S390PCIIOMMU {
     Object parent_obj;
     AddressSpace as;
-- 
2.34.1

[PATCH v3 14/15] s390x/pci: Implement migration for emulated devices

[Konstantin Shkolnyy]

Implement zPCI device state migration, consequently enabling migration
of VMs that have emulated PCI devices, whether virtio or not.
Migration is allowed for devices whose function handle has the
FH_SHM_EMUL bit set. For these devices QEMU will save and restore the
state of its zPCI emulator.

This will enable emulated PCI migration starting with s390-ccw-virtio-11.1.

Passthrough devices will continue to block migration.

Signed-off-by: Konstantin Shkolnyy <kshk@linux.ibm.com>
---
 hw/s390x/s390-pci-bus.c         | 187 +++++++++++++++++++++++++++++++-
 hw/s390x/s390-pci-inst.c        |   2 +-
 hw/s390x/s390-virtio-ccw.c      |   5 +
 include/hw/s390x/s390-pci-bus.h |   4 +
 4 files changed, 192 insertions(+), 6 deletions(-)

diff --git a/hw/s390x/s390-pci-bus.c b/hw/s390x/s390-pci-bus.c
index 8505abcc90..b05c4b43d1 100644
--- a/hw/s390x/s390-pci-bus.c
+++ b/hw/s390x/s390-pci-bus.c
@@ -26,6 +26,7 @@
 #include "hw/pci/pci_bridge.h"
 #include "hw/pci/msi.h"
 #include "exec/cpu-common.h"
+#include "migration/blocker.h"
 #include "qemu/error-report.h"
 #include "qemu/module.h"
 #include "system/reset.h"
@@ -33,6 +34,11 @@
 
 #include "trace.h"
 
+static const Property phb_props[] = {
+    DEFINE_PROP_BOOL("x-zpci-emul-dev-migr-enabled", S390pciState,
+                     emul_dev_migr_enabled, true),
+};
+
 S390pciState *s390_get_phb(void)
 {
     static S390pciState *phb;
@@ -916,6 +922,23 @@ static void set_pbdev_info(S390PCIBusDevice *pbdev)
     pbdev->pci_group = s390_group_find(ZPCI_DEFAULT_FN_GRP);
 }
 
+static int s390_set_emul_dev_migration_blocker(S390pciState *s, Error **errp)
+{
+    if (s->emul_dev_migr_enabled) {
+        return 0;
+    }
+    error_setg(&s->emul_dev_migr_blocker,
+               "Migration disabled for emulated zPCI devices on this machine type");
+    return migrate_add_blocker(&s->emul_dev_migr_blocker, errp);
+}
+
+static void s390_clear_emul_dev_migration_blocker(S390pciState *s)
+{
+    if (s->emul_dev_migr_blocker) {
+        migrate_del_blocker(&s->emul_dev_migr_blocker);
+    }
+}
+
 static void s390_pcihost_realize(DeviceState *dev, Error **errp)
 {
     PCIBus *b;
@@ -951,6 +974,7 @@ static void s390_pcihost_realize(DeviceState *dev, Error **errp)
     css_register_io_adapters(CSS_IO_ADAPTER_PCI, true, false,
                              S390_ADAPTER_SUPPRESSIBLE, errp);
     s390_pcihost_kvm_realize();
+    s390_set_emul_dev_migration_blocker(s, errp);
 }
 
 static void s390_pcihost_unrealize(DeviceState *dev)
@@ -958,6 +982,8 @@ static void s390_pcihost_unrealize(DeviceState *dev)
     S390PCIGroup *group;
     S390pciState *s = S390_PCI_HOST_BRIDGE(dev);
 
+    s390_clear_emul_dev_migration_blocker(s);
+
     while (!QTAILQ_EMPTY(&s->zpci_groups)) {
         group = QTAILQ_FIRST(&s->zpci_groups);
         QTAILQ_REMOVE(&s->zpci_groups, group, link);
@@ -1128,6 +1154,27 @@ static int s390_pci_interp_plug(S390pciState *s, S390PCIBusDevice *pbdev)
     return 0;
 }
 
+static int s390_set_passthrough_migration_blocker(S390PCIBusDevice *pbdev,
+                                            Error **errp)
+{
+    pbdev->passthrough_migr_blocker = NULL;
+
+    if (pbdev->fh & FH_SHM_EMUL) {
+        return 0;
+    }
+    error_setg(&pbdev->passthrough_migr_blocker,
+               "Migration blocked by passthrough zPCI device "
+               "fh 0x%x uid %d fid %d", pbdev->fh, pbdev->uid, pbdev->fid);
+    return migrate_add_blocker(&pbdev->passthrough_migr_blocker, errp);
+}
+
+static void s390_clear_passthrough_migration_blocker(S390PCIBusDevice *pbdev)
+{
+    if (pbdev->passthrough_migr_blocker) {
+        migrate_del_blocker(&pbdev->passthrough_migr_blocker);
+    }
+}
+
 static void s390_pcihost_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
                               Error **errp)
 {
@@ -1247,6 +1294,11 @@ static void s390_pcihost_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
             return;
         }
 
+        if (s390_set_passthrough_migration_blocker(pbdev, errp)) {
+            s390_pci_msix_free(pbdev);
+            return;
+        }
+
         if (dev->hotplugged) {
             s390_pci_generate_plug_event(HP_EVENT_TO_CONFIGURED ,
                                          pbdev->fh, pbdev->fid);
@@ -1283,6 +1335,8 @@ static void s390_pcihost_unplug(HotplugHandler *hotplug_dev, DeviceState *dev,
             return;
         }
 
+        s390_clear_passthrough_migration_blocker(pbdev);
+
         s390_pci_generate_plug_event(HP_EVENT_STANDBY_TO_RESERVED,
                                      pbdev->fh, pbdev->fid);
         bus = pci_get_bus(pci_dev);
@@ -1450,6 +1504,7 @@ static void s390_pcihost_class_init(ObjectClass *klass, const void *data)
     hc->unplug_request = s390_pcihost_unplug_request;
     hc->unplug = s390_pcihost_unplug;
     msi_nonbroken = true;
+    device_class_set_props(dc, phb_props);
 }
 
 static const TypeInfo s390_pcihost_info = {
@@ -1463,10 +1518,33 @@ static const TypeInfo s390_pcihost_info = {
     }
 };
 
+/* Return a unique bus "path" for zpci device */
+static char *s390_pci_bus_get_dev_path(DeviceState *dev)
+{
+    S390PCIBusDevice *pbdev = S390_PCI_DEVICE(dev);
+    return g_strdup_printf("uid-%04x", pbdev->uid);
+}
+
+static void s390_pcibus_class_init(ObjectClass *oc, const void *data)
+{
+    BusClass *bc = BUS_CLASS(oc);
+    bc->get_dev_path = s390_pci_bus_get_dev_path;
+}
+
 static const TypeInfo s390_pcibus_info = {
     .name = TYPE_S390_PCI_BUS,
     .parent = TYPE_BUS,
     .instance_size = sizeof(S390PCIBus),
+    /*
+     * Implement get_dev_path() to provide each zpci device with a unique
+     * stable UID-based bus "path". The "path" is used as part of idstr in the
+     ^ migration stream, making idstr unique and instance_id always 0.
+     * For migration to succeed, (idstr+instance_id) must match those generated
+     * during QEMU start. Without unique idstr, QEMU will generate variable
+     * instance_id to distinquish devices, and that instance_id can change
+     * if a device is unplugged and plugged back, preventing migration.
+     */
+    .class_init = s390_pcibus_class_init,
 };
 
 static uint16_t s390_pci_generate_uid(S390pciState *s)
@@ -1612,13 +1690,112 @@ static const Property s390_pci_device_properties[] = {
                      true),
 };
 
-static const VMStateDescription s390_pci_device_vmstate = {
-    .name = TYPE_S390_PCI_DEVICE,
+static int s390_pci_device_pre_load(void *opaque)
+{
+    S390PCIBusDevice *pbdev = S390_PCI_DEVICE(opaque);
+    S390PCIBusDevice *found_pbdev;
+
+    /*
+     * Make sure pbdev is removed from the table before state load. The change
+     * of pbdev->idx means it needs to be moved to a different position anyway,
+     * and is illegal while in the table. But be careful to not remove
+     * instead another pbdev whose state might have been loaded earlier and
+     * that has then replaced our pbdev. (post_load() will put our pbdev back.)
+     */
+    found_pbdev = g_hash_table_lookup(s390_get_phb()->zpci_table, &pbdev->idx);
+    assert(found_pbdev);
+    if (found_pbdev == pbdev) {
+        g_hash_table_remove(s390_get_phb()->zpci_table, &pbdev->idx);
+    }
+
+    return 0;
+}
+
+static int s390_pci_device_post_load(void *opaque, int version_id)
+{
+    S390PCIBusDevice *pbdev = S390_PCI_DEVICE(opaque);
+
     /*
-     * TODO: add state handling here, so migration works at least with
-     * emulated pci devices on s390x
+     * Now that pbdev->idx has been loaded, use it to place pbdev back into
+     * the table. This may replace a different not-yet-state-loaded pbdev,
+     * but pre_load() handles this case.
      */
-    .unmigratable = 1,
+    g_hash_table_replace(s390_get_phb()->zpci_table, &pbdev->idx, pbdev);
+
+    /*
+     * Regenerate IOMMU state, including IOTLB contents and QEMU memory regions.
+     */
+    if (pbdev->iommu_enabled) {
+        assert(pbdev->iommu);
+        if (s390_pci_is_translation_enabled(pbdev->g_iota)) {
+            s390_pci_iommu_enable(pbdev);
+            s390_pci_ioat_replay(pbdev);
+        } else {
+            s390_pci_iommu_direct_map_enable(pbdev);
+        }
+    }
+
+    /*
+     * Guest sets fmb_addr by mpcifc.ZPCI_MOD_FC_SET_MEASURE instruction,
+     * whose handler consequently starts fmb_timer. We may need to restart it.
+     */
+    if (pbdev->fmb_addr) {
+        assert(!pbdev->fmb_timer);
+        assert(pbdev->pci_group);
+        pbdev->fmb_timer = timer_new_ms(QEMU_CLOCK_VIRTUAL,
+                                        fmb_update, pbdev);
+        timer_mod(pbdev->fmb_timer,
+                  qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) +
+                                    pbdev->pci_group->zpci_group.mui);
+    }
+    return 0;
+}
+
+static const VMStateDescription s390_pci_device_vmstate = {
+    .name = TYPE_S390_PCI_DEVICE,
+    .version_id = 1,
+    .minimum_version_id = 1,
+    .pre_load = s390_pci_device_pre_load,
+    .post_load = s390_pci_device_post_load,
+    .fields = (const VMStateField[]) {
+        VMSTATE_UINT32(state, S390PCIBusDevice),
+        VMSTATE_UINT16(uid, S390PCIBusDevice),
+        VMSTATE_UINT32(idx, S390PCIBusDevice),
+        VMSTATE_UINT32(fh, S390PCIBusDevice),
+        VMSTATE_UINT32(fid, S390PCIBusDevice),
+        VMSTATE_BOOL(fid_defined, S390PCIBusDevice),
+        VMSTATE_UINT64(fmb_addr, S390PCIBusDevice),
+        VMSTATE_UINT32(fmb.format, S390PCIBusDevice),
+        VMSTATE_UINT32(fmb.sample, S390PCIBusDevice),
+        VMSTATE_UINT64(fmb.last_update, S390PCIBusDevice),
+        VMSTATE_UINT64_ARRAY(fmb.counter, S390PCIBusDevice,
+                ARRAY_SIZE(((S390PCIBusDevice *)0)->fmb.counter)),
+        VMSTATE_UINT64(fmb.fmt0.dma_rbytes, S390PCIBusDevice),
+        VMSTATE_UINT64(fmb.fmt0.dma_wbytes, S390PCIBusDevice),
+        VMSTATE_UINT8(isc, S390PCIBusDevice),
+        VMSTATE_UINT16(noi, S390PCIBusDevice),
+        VMSTATE_UINT8(sum, S390PCIBusDevice),
+        VMSTATE_UINT8(pft, S390PCIBusDevice),
+        VMSTATE_UINT64(routes.adapter.ind_addr, S390PCIBusDevice),
+        VMSTATE_UINT64(routes.adapter.summary_addr, S390PCIBusDevice),
+        VMSTATE_UINT64(routes.adapter.ind_offset, S390PCIBusDevice),
+        VMSTATE_UINT32(routes.adapter.summary_offset, S390PCIBusDevice),
+        VMSTATE_UINT32(routes.adapter.adapter_id, S390PCIBusDevice),
+        VMSTATE_BOOL(iommu_enabled, S390PCIBusDevice),
+        VMSTATE_UINT64(g_iota, S390PCIBusDevice),
+        VMSTATE_UINT64(pba, S390PCIBusDevice),
+        VMSTATE_UINT64(pal, S390PCIBusDevice),
+        VMSTATE_UINT64(max_dma_limit, S390PCIBusDevice),
+        VMSTATE_PTR_TO_IND_ADDR(summary_ind, S390PCIBusDevice),
+        VMSTATE_PTR_TO_IND_ADDR(indicator, S390PCIBusDevice),
+        VMSTATE_BOOL(pci_unplug_request_processed, S390PCIBusDevice),
+        VMSTATE_BOOL(unplug_requested, S390PCIBusDevice),
+        VMSTATE_BOOL(interp, S390PCIBusDevice),
+        VMSTATE_BOOL(forwarding_assist, S390PCIBusDevice),
+        VMSTATE_BOOL(aif, S390PCIBusDevice),
+        VMSTATE_BOOL(rtr_avail, S390PCIBusDevice),
+        VMSTATE_END_OF_LIST()
+    }
 };
 
 static void s390_pci_device_class_init(ObjectClass *klass, const void *data)
diff --git a/hw/s390x/s390-pci-inst.c b/hw/s390x/s390-pci-inst.c
index 191f549ba7..a9c0a4effb 100644
--- a/hw/s390x/s390-pci-inst.c
+++ b/hw/s390x/s390-pci-inst.c
@@ -1094,7 +1094,7 @@ static int fmb_do_update(S390PCIBusDevice *pbdev, int offset, uint64_t val,
     return ret;
 }
 
-static void fmb_update(void *opaque)
+void fmb_update(void *opaque)
 {
     S390PCIBusDevice *pbdev = opaque;
     int64_t t = qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL);
diff --git a/hw/s390x/s390-virtio-ccw.c b/hw/s390x/s390-virtio-ccw.c
index cc768daeb0..5900d919d2 100644
--- a/hw/s390x/s390-virtio-ccw.c
+++ b/hw/s390x/s390-virtio-ccw.c
@@ -923,8 +923,13 @@ static void ccw_machine_11_0_instance_options(MachineState *machine)
 
 static void ccw_machine_11_0_class_options(MachineClass *mc)
 {
+    static GlobalProperty compat[] = {
+        { TYPE_S390_PCI_HOST_BRIDGE, "x-zpci-emul-dev-migr-enabled", "off" },
+    };
+
     ccw_machine_11_1_class_options(mc);
     compat_props_add(mc->compat_props, hw_compat_11_0, hw_compat_11_0_len);
+    compat_props_add(mc->compat_props, compat, G_N_ELEMENTS(compat));
 }
 DEFINE_CCW_MACHINE(11, 0);
 
diff --git a/include/hw/s390x/s390-pci-bus.h b/include/hw/s390x/s390-pci-bus.h
index 2fab28e6e0..3840f6a48f 100644
--- a/include/hw/s390x/s390-pci-bus.h
+++ b/include/hw/s390x/s390-pci-bus.h
@@ -337,6 +337,7 @@ struct S390PCIBusDevice {
     uint16_t uid;
     uint32_t idx;
     uint32_t fh;
+    Error *passthrough_migr_blocker;
     uint32_t fid;
     bool fid_defined;
     uint64_t fmb_addr;
@@ -390,6 +391,8 @@ struct S390pciState {
     QTAILQ_HEAD(, S390PCIDMACount) zpci_dma_limit;
     QTAILQ_HEAD(, S390PCIGroup) zpci_groups;
     uint8_t next_sim_grp;
+    bool emul_dev_migr_enabled;
+    Error *emul_dev_migr_blocker;
 };
 
 S390pciState *s390_get_phb(void);
@@ -415,5 +418,6 @@ S390PCIBusDevice *s390_pci_find_dev_by_pci(S390pciState *s,
 S390PCIBusDevice *s390_pci_find_next_avail_dev(S390pciState *s,
                                                S390PCIBusDevice *pbdev);
 void s390_pci_ism_reset(void);
+void fmb_update(void *opaque);
 
 #endif
-- 
2.34.1

[PATCH v3 15/15] s390x/pci: Create function to contain fmb_timer start

[Konstantin Shkolnyy]

fmb_timer is now started in 3 different places. The new function will encapsulate
that to make sure mui is added in all cases.

Signed-off-by: Konstantin Shkolnyy <kshk@linux.ibm.com>
---
 hw/s390x/s390-pci-bus.c         |  5 ++---
 hw/s390x/s390-pci-inst.c        | 14 ++++++++++----
 include/hw/s390x/s390-pci-bus.h |  1 +
 3 files changed, 13 insertions(+), 7 deletions(-)

diff --git a/hw/s390x/s390-pci-bus.c b/hw/s390x/s390-pci-bus.c
index b05c4b43d1..49efed2724 100644
--- a/hw/s390x/s390-pci-bus.c
+++ b/hw/s390x/s390-pci-bus.c
@@ -1744,9 +1744,8 @@ static int s390_pci_device_post_load(void *opaque, int version_id)
         assert(pbdev->pci_group);
         pbdev->fmb_timer = timer_new_ms(QEMU_CLOCK_VIRTUAL,
                                         fmb_update, pbdev);
-        timer_mod(pbdev->fmb_timer,
-                  qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) +
-                                    pbdev->pci_group->zpci_group.mui);
+        s390_pci_schedule_fmb_timer(pbdev,
+                                    qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL));
     }
     return 0;
 }
diff --git a/hw/s390x/s390-pci-inst.c b/hw/s390x/s390-pci-inst.c
index a9c0a4effb..62ecd50dcb 100644
--- a/hw/s390x/s390-pci-inst.c
+++ b/hw/s390x/s390-pci-inst.c
@@ -1094,9 +1094,16 @@ static int fmb_do_update(S390PCIBusDevice *pbdev, int offset, uint64_t val,
     return ret;
 }
 
+void s390_pci_schedule_fmb_timer(S390PCIBusDevice *pbdev, uint64_t start)
+{
+    timer_mod(pbdev->fmb_timer, start + pbdev->pci_group->zpci_group.mui);
+}
+
 void fmb_update(void *opaque)
 {
     S390PCIBusDevice *pbdev = opaque;
+
+    /* Must be read before updating U bit */
     int64_t t = qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL);
     int i;
 
@@ -1133,7 +1140,7 @@ void fmb_update(void *opaque)
                       sizeof(pbdev->fmb.last_update))) {
         return;
     }
-    timer_mod(pbdev->fmb_timer, t + pbdev->pci_group->zpci_group.mui);
+    s390_pci_schedule_fmb_timer(pbdev, t);
 }
 
 static int mpcifc_reg_int_interp(S390PCIBusDevice *pbdev, ZpciFib *fib)
@@ -1326,9 +1333,8 @@ int mpcifc_service_call(S390CPU *cpu, uint8_t r1, uint64_t fiba, uint8_t ar,
             timer_del(pbdev->fmb_timer);
         }
         pbdev->fmb_addr = fmb_addr;
-        timer_mod(pbdev->fmb_timer,
-                  qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) +
-                                    pbdev->pci_group->zpci_group.mui);
+        s390_pci_schedule_fmb_timer(pbdev,
+                                    qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL));
         break;
     }
     default:
diff --git a/include/hw/s390x/s390-pci-bus.h b/include/hw/s390x/s390-pci-bus.h
index 3840f6a48f..0385c1aba8 100644
--- a/include/hw/s390x/s390-pci-bus.h
+++ b/include/hw/s390x/s390-pci-bus.h
@@ -419,5 +419,6 @@ S390PCIBusDevice *s390_pci_find_next_avail_dev(S390pciState *s,
                                                S390PCIBusDevice *pbdev);
 void s390_pci_ism_reset(void);
 void fmb_update(void *opaque);
+void s390_pci_schedule_fmb_timer(S390PCIBusDevice *pbdev, uint64_t start);
 
 #endif
-- 
2.34.1

Re: [PATCH v3 01/15] s390x/pci: implement IOMMU replay

[Farhan Ali]

On 6/4/2026 7:17 PM, Konstantin Shkolnyy wrote:
> From: Matthew Rosato <mjrosato@linux.ibm.com>
>
> There are a few scenarios where IOMMU replay can potentially be needed
> for zPCI device, namely VFIO device reset scenarios where the guest
> continues running and expects the contents of its IOMMU to be replayed
> upon IOAT re-registration and migration scenarios where the destination
> must reconstruct the IOMMU on the destination.
>
> zPCI migration is not supported yet, but the IOMMU replay function is
> implemented so that it can be called both from IOMMUMemoryRegionClass
> now and migration post_load later.
>
> Signed-off-by: Matthew Rosato <mjrosato@linux.ibm.com>
> Signed-off-by: Konstantin Shkolnyy <kshk@linux.ibm.com>
> ---
>   hw/s390x/s390-pci-bus.c          | 62 ++++++++++++++++++++++++++++----
>   hw/s390x/s390-pci-inst.c         |  4 +--
>   include/hw/s390x/s390-pci-inst.h |  1 +
>   3 files changed, 59 insertions(+), 8 deletions(-)
>
> diff --git a/hw/s390x/s390-pci-bus.c b/hw/s390x/s390-pci-bus.c
> index 4de7b587e8..a104e550b1 100644
> --- a/hw/s390x/s390-pci-bus.c
> +++ b/hw/s390x/s390-pci-bus.c
> @@ -592,14 +592,64 @@ err:
>       return ret;
>   }
>   
> -static void s390_pci_iommu_replay(IOMMUMemoryRegion *iommu,
> +static void s390_pci_ioat_replay(S390PCIIOMMU *iommu)
> +{
> +    S390IOTLBEntry entry;
> +    uint16_t error = 0;
> +    uint32_t dma_avail;
> +    hwaddr curr, end;
> +
> +    curr = iommu->pba;
> +    end = iommu->pal;
> +
> +    if (iommu->dm_mr) {
> +        /* If direct mapping is used, there are no guest tables to replay */
> +        return;
> +    }

I am curious, how would migration work if direct mapping is used? How is 
the IOMMU state replicated on the target machine?


> +
> +    if (iommu->dma_limit) {
> +        dma_avail = iommu->dma_limit->avail;
> +    } else {
> +        dma_avail = 1;
> +    }
> +
> +    while (curr < end) {
> +        error = s390_guest_io_table_walk(iommu->g_iota, curr, &entry);
> +        if (error) {
> +            pbdev->state = ZPCI_FS_ERROR;
> +            s390_pci_generate_error_event(error, pbdev->fh, pbdev->fid, curr,
> +                                          0);
> +            error_report("Failure to walk table during iommu remap");
> +            return;
> +        }
> +
> +        if (entry.perm != IOMMU_NONE) {
> +            if (dma_avail > 0) {
> +                dma_avail = s390_pci_update_iotlb(iommu, &entry);
> +            } else {
> +                /*
> +                 * There is no reliable method to request the guest to release
> +                 * mappings other than in response to a RPCIT instruction;
> +                 * generate a permanent error condition and require the device
> +                 * to be completely re-initialized from the guest side.
> +                 */
> +                pbdev->state = ZPCI_FS_ERROR;
> +                s390_pci_generate_error_event(ERR_EVENT_PERMERR, pbdev->fh,
> +                                              pbdev->fid, 0, 0);
> +                error_report("DMA mappings exhausted: iommu remap failed");
> +                return;
> +            }
> +        }
> +        curr += entry.len;
> +    }
> +}
> +
> +static void s390_pci_iommu_replay(IOMMUMemoryRegion *mr,
>                                     IOMMUNotifier *notifier)
>   {
> -    /* It's impossible to plug a pci device on s390x that already has iommu
> -     * mappings which need to be replayed, that is due to the "one iommu per
> -     * zpci device" construct. But when we support migration of vfio-pci
> -     * devices in future, we need to revisit this.
> -     */
> +    S390PCIIOMMU *iommu = container_of(mr, S390PCIIOMMU, iommu_mr);
> +
> +    s390_pci_ioat_replay(iommu);
>   }
>   
>   static S390PCIIOMMU *s390_pci_get_iommu(S390pciState *s, PCIBus *bus,
> diff --git a/hw/s390x/s390-pci-inst.c b/hw/s390x/s390-pci-inst.c
> index 10066ca618..1834596076 100644
> --- a/hw/s390x/s390-pci-inst.c
> +++ b/hw/s390x/s390-pci-inst.c
> @@ -613,8 +613,8 @@ int pcistg_service_call(S390CPU *cpu, uint8_t r1, uint8_t r2, uintptr_t ra)
>       return 0;
>   }
>   
> -static uint32_t s390_pci_update_iotlb(S390PCIIOMMU *iommu,
> -                                      S390IOTLBEntry *entry)
> +uint32_t s390_pci_update_iotlb(S390PCIIOMMU *iommu,
> +                               S390IOTLBEntry *entry)
>   {
>       S390IOTLBEntry *cache = g_hash_table_lookup(iommu->iotlb, &entry->iova);
>       IOMMUTLBEvent event = {
> diff --git a/include/hw/s390x/s390-pci-inst.h b/include/hw/s390x/s390-pci-inst.h
> index 5cb8da540b..c782990e3b 100644
> --- a/include/hw/s390x/s390-pci-inst.h
> +++ b/include/hw/s390x/s390-pci-inst.h
> @@ -111,6 +111,7 @@ int mpcifc_service_call(S390CPU *cpu, uint8_t r1, uint64_t fiba, uint8_t ar,
>   int stpcifc_service_call(S390CPU *cpu, uint8_t r1, uint64_t fiba, uint8_t ar,
>                            uintptr_t ra);
>   void fmb_timer_free(S390PCIBusDevice *pbdev);
> +uint32_t s390_pci_update_iotlb(S390PCIIOMMU *iommu, S390IOTLBEntry *entry);
>   
>   #define ZPCI_IO_BAR_MIN 0
>   #define ZPCI_IO_BAR_MAX 5

Re: [PATCH v3 01/15] s390x/pci: implement IOMMU replay

[Matthew Rosato]

On 6/9/26 6:07 PM, Farhan Ali wrote:
> 
> On 6/4/2026 7:17 PM, Konstantin Shkolnyy wrote:
>> From: Matthew Rosato <mjrosato@linux.ibm.com>
>>
>> There are a few scenarios where IOMMU replay can potentially be needed
>> for zPCI device, namely VFIO device reset scenarios where the guest
>> continues running and expects the contents of its IOMMU to be replayed
>> upon IOAT re-registration and migration scenarios where the destination
>> must reconstruct the IOMMU on the destination.
>>
>> zPCI migration is not supported yet, but the IOMMU replay function is
>> implemented so that it can be called both from IOMMUMemoryRegionClass
>> now and migration post_load later.
>>
>> Signed-off-by: Matthew Rosato <mjrosato@linux.ibm.com>
>> Signed-off-by: Konstantin Shkolnyy <kshk@linux.ibm.com>
>> ---
>>   hw/s390x/s390-pci-bus.c          | 62 ++++++++++++++++++++++++++++----
>>   hw/s390x/s390-pci-inst.c         |  4 +--
>>   include/hw/s390x/s390-pci-inst.h |  1 +
>>   3 files changed, 59 insertions(+), 8 deletions(-)
>>
>> diff --git a/hw/s390x/s390-pci-bus.c b/hw/s390x/s390-pci-bus.c
>> index 4de7b587e8..a104e550b1 100644
>> --- a/hw/s390x/s390-pci-bus.c
>> +++ b/hw/s390x/s390-pci-bus.c
>> @@ -592,14 +592,64 @@ err:
>>       return ret;
>>   }
>>   -static void s390_pci_iommu_replay(IOMMUMemoryRegion *iommu,
>> +static void s390_pci_ioat_replay(S390PCIIOMMU *iommu)
>> +{
>> +    S390IOTLBEntry entry;
>> +    uint16_t error = 0;
>> +    uint32_t dma_avail;
>> +    hwaddr curr, end;
>> +
>> +    curr = iommu->pba;
>> +    end = iommu->pal;
>> +
>> +    if (iommu->dm_mr) {
>> +        /* If direct mapping is used, there are no guest tables to replay */
>> +        return;
>> +    }
> 
> I am curious, how would migration work if direct mapping is used? How is the IOMMU state replicated on the target machine?

Hm, good question.

For this series, it's not really an issue; emulated devices are always
disallowed from direct mapping and only emulated devices are being
enabled for migration here (passthrough devices are still fenced) --
so you'll never take this path during a migration with this series.

When I wrote this patch I was testing the VFIO reset case where we will
get a new MPCFIC from the guest to drive reg_ioat().  But you're right,
that won't happen with a migration; rather migration code will need
specific handling to re-trigger the mapping done in reg_ioat() for 
direct-mapped devices in order to fill the host IOMMU without a guest
IOMMU to replay.

Actually, I believe Konstantin is doing the necessary work already in 
patch 14 via a call to s390_pci_iommu_direct_map_enable() during
s390_pci_device_post_load().

Thanks,
Matt

Re: [PATCH v3 15/15] s390x/pci: Create function to contain fmb_timer start

[Matthew Rosato]

On 6/4/26 10:17 PM, Konstantin Shkolnyy wrote:
> fmb_timer is now started in 3 different places. The new function will encapsulate
> that to make sure mui is added in all cases.
> 

Nit: move 'encapsulate' to the next line to keep git log output in the
80 char range.

> Signed-off-by: Konstantin Shkolnyy <kshk@linux.ibm.com>

Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>

Re: [PATCH v3 13/15] s390x/pci: Add a comment explaining S390PCIIOMMU purpose

[Matthew Rosato]

On 6/4/26 10:17 PM, Konstantin Shkolnyy wrote:
> Signed-off-by: Konstantin Shkolnyy <kshk@linux.ibm.com>

Commit message needs something here -- You don't need to include the
explanation of the structure (that's in the commit itself) but maybe
describe the context (e.g. you're adding the comment because you just
moved most of the contents of the structure out in support of migration
and are now describing the remaining purpose)

> ---
>  include/hw/s390x/s390-pci-bus.h | 8 ++++++++
>  1 file changed, 8 insertions(+)
> 
> diff --git a/include/hw/s390x/s390-pci-bus.h b/include/hw/s390x/s390-pci-bus.h
> index e3cbee2695..2fab28e6e0 100644
> --- a/include/hw/s390x/s390-pci-bus.h
> +++ b/include/hw/s390x/s390-pci-bus.h
> @@ -271,6 +271,14 @@ typedef struct S390PCIDMACount {
>      QTAILQ_ENTRY(S390PCIDMACount) link;
>  } S390PCIDMACount;
>  
> +/*
> + * This structure holds the PCI device AddressSpace that QEMU needs to link
> + * into its internal structures before the zPCI and PCI devices are fully
> + * initialized. It's a QEMU requirement to provide this "root" AddressSpace
> + * early. The AddressSpace is only actually used for I/O while the PCI
> + * device is plugged in and configured by the guest, at which time it gets
> + * additional memory subregions from zPCI device, that can do real work.
> + */

To be more clear, it's not that it is required 'early' so much as that
it is required to be there at the time the PCI device is plugged.

The lifecycle has always been a bit odd here; it is an attribute of the
zPCI device, but is required by the PCI device in order to function and
the necessary linkage between the 2 (PCI->zPCI) is established via the
iommu_table in the S390pciState.  Would be good to work some of that
into this explanation.

>  struct S390PCIIOMMU {
>      Object parent_obj;
>      AddressSpace as;

Re: [PATCH v3 03/15] s390x/pci: Move iommu_mr from S390PCIIOMMU to S390PCIBusDevice

[Farhan Ali]

<..snip..>

On 6/4/2026 7:17 PM, Konstantin Shkolnyy wrote:
> diff --git a/include/hw/s390x/s390-pci-bus.h b/include/hw/s390x/s390-pci-bus.h
> index eb15cb8b2d..a71f562dfc 100644
> --- a/include/hw/s390x/s390-pci-bus.h
> +++ b/include/hw/s390x/s390-pci-bus.h
> @@ -276,7 +276,6 @@ struct S390PCIIOMMU {
>       S390PCIBusDevice *pbdev;
>       AddressSpace as;
>       MemoryRegion mr;
> -    IOMMUMemoryRegion iommu_mr;
>       MemoryRegion *dm_mr;
>       bool enabled;
>       uint64_t g_iota;
> @@ -354,6 +353,7 @@ struct S390PCIBusDevice {
>       S390MsixInfo msix;
>       AdapterRoutes routes;
>       S390PCIIOMMU *iommu;
> +    IOMMUMemoryRegion iommu_mr;
>       MemoryRegion msix_notify_mr;
>       IndAddr *summary_ind;
>       IndAddr *indicator;

Maybe I am missing something, but why do we need to remove iommu_mr from 
S390PCIIOMMU? S390PCIBusDevice already has a reference to the 
S390PCIIOMMU in its definition. So is there something about iommu_mr 
defined in S390PCIIOMMU makes it difficult to migrate? Its hard to 
understand from the commit message.

Thanks

Farhan

Re: [PATCH v3 13/15] s390x/pci: Add a comment explaining S390PCIIOMMU purpose

[Farhan Ali]

On 6/19/2026 10:39 AM, Matthew Rosato wrote:
> On 6/4/26 10:17 PM, Konstantin Shkolnyy wrote:
>> Signed-off-by: Konstantin Shkolnyy <kshk@linux.ibm.com>
> Commit message needs something here -- You don't need to include the
> explanation of the structure (that's in the commit itself) but maybe
> describe the context (e.g. you're adding the comment because you just
> moved most of the contents of the structure out in support of migration
> and are now describing the remaining purpose)
>
>> ---
>>   include/hw/s390x/s390-pci-bus.h | 8 ++++++++
>>   1 file changed, 8 insertions(+)
>>
>> diff --git a/include/hw/s390x/s390-pci-bus.h b/include/hw/s390x/s390-pci-bus.h
>> index e3cbee2695..2fab28e6e0 100644
>> --- a/include/hw/s390x/s390-pci-bus.h
>> +++ b/include/hw/s390x/s390-pci-bus.h
>> @@ -271,6 +271,14 @@ typedef struct S390PCIDMACount {
>>       QTAILQ_ENTRY(S390PCIDMACount) link;
>>   } S390PCIDMACount;
>>   
>> +/*
>> + * This structure holds the PCI device AddressSpace that QEMU needs to link
>> + * into its internal structures before the zPCI and PCI devices are fully
>> + * initialized. It's a QEMU requirement to provide this "root" AddressSpace
>> + * early. The AddressSpace is only actually used for I/O while the PCI
>> + * device is plugged in and configured by the guest, at which time it gets
>> + * additional memory subregions from zPCI device, that can do real work.
>> + */
> To be more clear, it's not that it is required 'early' so much as that
> it is required to be there at the time the PCI device is plugged.
>
> The lifecycle has always been a bit odd here; it is an attribute of the
> zPCI device, but is required by the PCI device in order to function and
> the necessary linkage between the 2 (PCI->zPCI) is established via the
> iommu_table in the S390pciState.  Would be good to work some of that
> into this explanation.

+1 I think we moving few fields from the struct, so having an 
explanation why do we need the remaining fields would be helpful. Also 
can this struct be extended in the future without any issues and 
breaking? If not then maybe adding something about would that would also 
be helpful.

Thanks

Farhan


>>   struct S390PCIIOMMU {
>>       Object parent_obj;
>>       AddressSpace as;