s390x/pci: Honor DMA limits set by vfio

When an s390 guest is using lazy unmapping, it can result in a very
large number of oustanding DMA requests, far beyond the default
limit configured for vfio.  Let's track DMA usage similar to vfio
in the host, and trigger the guest to flush their DMA mappings
before vfio runs out.

Signed-off-by: Matthew Rosato <mjrosato@linux.ibm.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
[aw: non-Linux build fixes]
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>

3 files changed, 92 insertions, 11 deletions

diff --git a/hw/s390x/s390-pci-bus.c b/hw/s390x/s390-pci-bus.c
index a929340688..218717397a 100644
--- a/hw/s390x/s390-pci-bus.c
+++ b/hw/s390x/s390-pci-bus.c
@@ -17,6 +17,7 @@
 #include "cpu.h"
 #include "hw/s390x/s390-pci-bus.h"
 #include "hw/s390x/s390-pci-inst.h"
+#include "hw/s390x/s390-pci-vfio.h"
 #include "hw/pci/pci_bus.h"
 #include "hw/qdev-properties.h"
 #include "hw/pci/pci_bridge.h"
@@ -764,6 +765,7 @@ static void s390_pcihost_realize(DeviceState *dev, Error **errp)
     s->bus_no = 0;
     QTAILQ_INIT(&s->pending_sei);
     QTAILQ_INIT(&s->zpci_devs);
+    QTAILQ_INIT(&s->zpci_dma_limit);
 
     css_register_io_adapters(CSS_IO_ADAPTER_PCI, true, false,
                              S390_ADAPTER_SUPPRESSIBLE, errp);
@@ -941,17 +943,18 @@ 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;
+
         if (object_dynamic_cast(OBJECT(dev), "vfio-pci")) {
             pbdev->fh |= FH_SHM_VFIO;
+            pbdev->iommu->dma_limit = s390_pci_start_dma_count(s, pbdev);
         } else {
             pbdev->fh |= FH_SHM_EMUL;
         }
 
-        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;
-
         if (s390_pci_msix_init(pbdev)) {
             error_setg(errp, "MSI-X support is mandatory "
                        "in the S390 architecture");
@@ -1004,6 +1007,9 @@ 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->dma_limit) {
+            s390_pci_end_dma_count(s, pbdev->iommu->dma_limit);
+        }
         qdev_unrealize(dev);
     }
 }
diff --git a/hw/s390x/s390-pci-inst.c b/hw/s390x/s390-pci-inst.c
index 639b13c8d6..4eadd9e794 100644
--- a/hw/s390x/s390-pci-inst.c
+++ b/hw/s390x/s390-pci-inst.c
@@ -32,6 +32,20 @@
         }                                                          \
     } while (0)
 
+static inline void inc_dma_avail(S390PCIIOMMU *iommu)
+{
+    if (iommu->dma_limit) {
+        iommu->dma_limit->avail++;
+    }
+}
+
+static inline void dec_dma_avail(S390PCIIOMMU *iommu)
+{
+    if (iommu->dma_limit) {
+        iommu->dma_limit->avail--;
+    }
+}
+
 static void s390_set_status_code(CPUS390XState *env,
                                  uint8_t r, uint64_t status_code)
 {
@@ -572,7 +586,8 @@ int pcistg_service_call(S390CPU *cpu, uint8_t r1, uint8_t r2, uintptr_t ra)
     return 0;
 }
 
-static void s390_pci_update_iotlb(S390PCIIOMMU *iommu, S390IOTLBEntry *entry)
+static uint32_t s390_pci_update_iotlb(S390PCIIOMMU *iommu,
+                                      S390IOTLBEntry *entry)
 {
     S390IOTLBEntry *cache = g_hash_table_lookup(iommu->iotlb, &entry->iova);
     IOMMUTLBEntry notify = {
@@ -585,14 +600,15 @@ static void s390_pci_update_iotlb(S390PCIIOMMU *iommu, S390IOTLBEntry *entry)
 
     if (entry->perm == IOMMU_NONE) {
         if (!cache) {
-            return;
+            goto out;
         }
         g_hash_table_remove(iommu->iotlb, &entry->iova);
+        inc_dma_avail(iommu);
     } else {
         if (cache) {
             if (cache->perm == entry->perm &&
                 cache->translated_addr == entry->translated_addr) {
-                return;
+                goto out;
             }
 
             notify.perm = IOMMU_NONE;
@@ -606,9 +622,13 @@ static void s390_pci_update_iotlb(S390PCIIOMMU *iommu, S390IOTLBEntry *entry)
         cache->len = PAGE_SIZE;
         cache->perm = entry->perm;
         g_hash_table_replace(iommu->iotlb, &cache->iova, cache);
+        dec_dma_avail(iommu);
     }
 
     memory_region_notify_iommu(&iommu->iommu_mr, 0, notify);
+
+out:
+    return iommu->dma_limit ? iommu->dma_limit->avail : 1;
 }
 
 int rpcit_service_call(S390CPU *cpu, uint8_t r1, uint8_t r2, uintptr_t ra)
@@ -620,6 +640,7 @@ int rpcit_service_call(S390CPU *cpu, uint8_t r1, uint8_t r2, uintptr_t ra)
     S390PCIIOMMU *iommu;
     S390IOTLBEntry entry;
     hwaddr start, end;
+    uint32_t dma_avail;
 
     if (env->psw.mask & PSW_MASK_PSTATE) {
         s390_program_interrupt(env, PGM_PRIVILEGED, ra);
@@ -658,6 +679,11 @@ 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;
+    } else {
+        dma_avail = 1;
+    }
     if (!iommu->g_iota) {
         error = ERR_EVENT_INVALAS;
         goto err;
@@ -675,8 +701,9 @@ int rpcit_service_call(S390CPU *cpu, uint8_t r1, uint8_t r2, uintptr_t ra)
         }
 
         start += entry.len;
-        while (entry.iova < start && entry.iova < end) {
-            s390_pci_update_iotlb(iommu, &entry);
+        while (entry.iova < start && entry.iova < end &&
+               (dma_avail > 0 || entry.perm == IOMMU_NONE)) {
+            dma_avail = s390_pci_update_iotlb(iommu, &entry);
             entry.iova += PAGE_SIZE;
             entry.translated_addr += PAGE_SIZE;
         }
@@ -689,7 +716,13 @@ err:
         s390_pci_generate_error_event(error, pbdev->fh, pbdev->fid, start, 0);
     } else {
         pbdev->fmb.counter[ZPCI_FMB_CNT_RPCIT]++;
-        setcc(cpu, ZPCI_PCI_LS_OK);
+        if (dma_avail > 0) {
+            setcc(cpu, ZPCI_PCI_LS_OK);
+        } else {
+            /* vfio DMA mappings are exhausted, trigger a RPCIT */
+            setcc(cpu, ZPCI_PCI_LS_ERR);
+            s390_set_status_code(env, r1, ZPCI_RPCIT_ST_INSUFF_RES);
+        }
     }
     return 0;
 }
diff --git a/hw/s390x/s390-pci-vfio.c b/hw/s390x/s390-pci-vfio.c
index cb3f4d98ad..0621fa386c 100644
--- a/hw/s390x/s390-pci-vfio.c
+++ b/hw/s390x/s390-pci-vfio.c
@@ -12,7 +12,9 @@
 #include <sys/ioctl.h>
 
 #include "qemu/osdep.h"
+#include "hw/s390x/s390-pci-bus.h"
 #include "hw/s390x/s390-pci-vfio.h"
+#include "hw/vfio/pci.h"
 #include "hw/vfio/vfio-common.h"
 
 /*
@@ -52,3 +54,43 @@ retry:
     return vfio_get_info_dma_avail(info, avail);
 }
 
+S390PCIDMACount *s390_pci_start_dma_count(S390pciState *s,
+                                          S390PCIBusDevice *pbdev)
+{
+    S390PCIDMACount *cnt;
+    uint32_t avail;
+    VFIOPCIDevice *vpdev = container_of(pbdev->pdev, VFIOPCIDevice, pdev);
+    int id;
+
+    assert(vpdev);
+
+    id = vpdev->vbasedev.group->container->fd;
+
+    if (!s390_pci_update_dma_avail(id, &avail)) {
+        return NULL;
+    }
+
+    QTAILQ_FOREACH(cnt, &s->zpci_dma_limit, link) {
+        if (cnt->id  == id) {
+            cnt->users++;
+            return cnt;
+        }
+    }
+
+    cnt = g_new0(S390PCIDMACount, 1);
+    cnt->id = id;
+    cnt->users = 1;
+    cnt->avail = avail;
+    QTAILQ_INSERT_TAIL(&s->zpci_dma_limit, cnt, link);
+    return cnt;
+}
+
+void s390_pci_end_dma_count(S390pciState *s, S390PCIDMACount *cnt)
+{
+    assert(cnt);
+
+    cnt->users--;
+    if (cnt->users == 0) {
+        QTAILQ_REMOVE(&s->zpci_dma_limit, cnt, link);
+    }
+}