qemu: Build HMAT command line

Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1786303

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>

src/conf/numa_conf.c

@@ -1904,6 +1904,13 @@ virDomainNumaGetNodeInitiator(const virDomainNuma *numa,
     if (!numa || node >= numa->nmem_nodes)
         return -1;
 
+    /* A NUMA node which has at least one vCPU is initiator to itself by
+     * definition. */
+    if (numa->mem_nodes[node].cpumask)
+        return node;
+
+    /* For the rest, "NUMA node that has best performance (the lowest
+     * latency or largest bandwidth) to this NUMA node." */
     for (i = 0; i < numa->ninterconnects; i++) {
         const virDomainNumaInterconnect *l = &numa->interconnects[i];
 

src/qemu/qemu_command.c

@@ -6933,6 +6933,9 @@ qemuBuildMachineCommandLine(virCommandPtr cmd,
             virBufferAsprintf(&buf, ",pflash1=%s", priv->pflash1->nodeformat);
     }
 
+    if (virDomainNumaHasHMAT(def->numa))
+        virBufferAddLit(&buf, ",hmat=on");
+
     virCommandAddArgBuffer(cmd, &buf);
 
     return 0;
@@ -7115,6 +7118,134 @@ qemuBuildIOThreadCommandLine(virCommandPtr cmd,
 }
 
 
+static int
+qemuBuilNumaCellCache(virCommandPtr cmd,
+                      const virDomainDef *def,
+                      size_t cell)
+{
+    size_t ncaches = virDomainNumaGetNodeCacheCount(def->numa, cell);
+    size_t i;
+
+    if (ncaches == 0)
+        return 0;
+
+    for (i = 0; i < ncaches; i++) {
+        g_auto(virBuffer) buf = VIR_BUFFER_INITIALIZER;
+        unsigned int level;
+        unsigned int size;
+        unsigned int line;
+        virDomainCacheAssociativity associativity;
+        virDomainCachePolicy policy;
+
+        if (virDomainNumaGetNodeCache(def->numa, cell, i,
+                                      &level, &size, &line,
+                                      &associativity, &policy) < 0) {
+            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
+                           _("Unable to format NUMA node cache"));
+            return -1;
+        }
+
+        virBufferAsprintf(&buf,
+                          "hmat-cache,node-id=%zu,size=%uK,level=%u",
+                          cell, size, level);
+
+        switch (associativity) {
+        case VIR_DOMAIN_CACHE_ASSOCIATIVITY_NONE:
+            virBufferAddLit(&buf, ",associativity=none");
+            break;
+        case VIR_DOMAIN_CACHE_ASSOCIATIVITY_DIRECT:
+            virBufferAddLit(&buf, ",associativity=direct");
+            break;
+        case VIR_DOMAIN_CACHE_ASSOCIATIVITY_FULL:
+            virBufferAddLit(&buf, ",associativity=complex");
+            break;
+        case VIR_DOMAIN_CACHE_ASSOCIATIVITY_LAST:
+            break;
+        }
+
+        switch (policy) {
+        case VIR_DOMAIN_CACHE_POLICY_NONE:
+            virBufferAddLit(&buf, ",policy=none");
+            break;
+        case VIR_DOMAIN_CACHE_POLICY_WRITEBACK:
+            virBufferAddLit(&buf, ",policy=write-back");
+            break;
+        case VIR_DOMAIN_CACHE_POLICY_WRITETHROUGH:
+            virBufferAddLit(&buf, ",policy=write-through");
+            break;
+        case VIR_DOMAIN_CACHE_POLICY_LAST:
+            break;
+        }
+
+        if (line > 0)
+            virBufferAsprintf(&buf, ",line=%u", line);
+
+        virCommandAddArg(cmd, "-numa");
+        virCommandAddArgBuffer(cmd, &buf);
+    }
+
+    return 0;
+}
+
+
+VIR_ENUM_DECL(qemuDomainMemoryHierarchy);
+VIR_ENUM_IMPL(qemuDomainMemoryHierarchy,
+              4, /* Maximum level of cache */
+              "memory", /* Special case, whole memory not specific cache */
+              "first-level",
+              "second-level",
+              "third-level");
+
+static int
+qemuBuildNumaHMATCommandLine(virCommandPtr cmd,
+                             const virDomainDef *def)
+{
+    size_t nlatencies;
+    size_t i;
+
+    if (!def->numa)
+        return 0;
+
+    nlatencies = virDomainNumaGetInterconnectsCount(def->numa);
+    for (i = 0; i < nlatencies; i++) {
+        g_auto(virBuffer) buf = VIR_BUFFER_INITIALIZER;
+        virDomainNumaInterconnectType type;
+        unsigned int initiator;
+        unsigned int target;
+        unsigned int cache;
+        virDomainMemoryLatency accessType;
+        unsigned long value;
+        const char *hierarchyStr;
+        const char *accessStr;
+
+        if (virDomainNumaGetInterconnect(def->numa, i,
+                                         &type, &initiator, &target,
+                                         &cache, &accessType, &value) < 0)
+            return -1;
+
+        hierarchyStr = qemuDomainMemoryHierarchyTypeToString(cache);
+        accessStr = virDomainMemoryLatencyTypeToString(accessType);
+        virBufferAsprintf(&buf,
+                          "hmat-lb,initiator=%u,target=%u,hierarchy=%s,data-type=%s-",
+                          initiator, target, hierarchyStr, accessStr);
+
+        switch (type) {
+        case VIR_DOMAIN_NUMA_INTERCONNECT_TYPE_LATENCY:
+            virBufferAsprintf(&buf, "latency,latency=%lu", value);
+            break;
+        case VIR_DOMAIN_NUMA_INTERCONNECT_TYPE_BANDWIDTH:
+            virBufferAsprintf(&buf, "bandwidth,bandwidth=%luK", value);
+            break;
+        }
+
+        virCommandAddArg(cmd, "-numa");
+        virCommandAddArgBuffer(cmd, &buf);
+    }
+
+    return 0;
+}
+
+
 static int
 qemuBuildNumaCommandLine(virQEMUDriverConfigPtr cfg,
                          virDomainDefPtr def,
@@ -7127,9 +7258,11 @@ qemuBuildNumaCommandLine(virQEMUDriverConfigPtr cfg,
     char *next = NULL;
     virBufferPtr nodeBackends = NULL;
     bool needBackend = false;
+    bool hmat = false;
     int rc;
     int ret = -1;
     size_t ncells = virDomainNumaGetNodeCount(def->numa);
+    ssize_t masterInitiator = -1;
 
     if (!virDomainNumatuneNodesetIsAvailable(def->numa, priv->autoNodeset))
         goto cleanup;
@@ -7139,6 +7272,11 @@ qemuBuildNumaCommandLine(virQEMUDriverConfigPtr cfg,
                                                def->os.machine))
         needBackend = true;
 
+    if (virDomainNumaHasHMAT(def->numa)) {
+        needBackend = true;
+        hmat = true;
+    }
+
     if (VIR_ALLOC_N(nodeBackends, ncells) < 0)
         goto cleanup;
 
@@ -7167,8 +7305,22 @@ qemuBuildNumaCommandLine(virQEMUDriverConfigPtr cfg,
         qemuBuildMemPathStr(def, cmd, priv) < 0)
         goto cleanup;
 
+    for (i = 0; i < ncells; i++) {
+        if (virDomainNumaGetNodeCpumask(def->numa, i)) {
+            masterInitiator = i;
+            break;
+        }
+    }
+
+    if (masterInitiator) {
+        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
+                       _("At least one NUMA node has to have CPUs"));
+        goto cleanup;
+    }
+
     for (i = 0; i < ncells; i++) {
         virBitmapPtr cpumask = virDomainNumaGetNodeCpumask(def->numa, i);
+        ssize_t initiator = virDomainNumaGetNodeInitiator(def->numa, i);
 
         if (needBackend) {
             virCommandAddArg(cmd, "-object");
@@ -7193,6 +7345,13 @@ qemuBuildNumaCommandLine(virQEMUDriverConfigPtr cfg,
             }
         }
 
+        if (hmat) {
+            if (initiator < 0)
+                initiator = masterInitiator;
+
+            virBufferAsprintf(&buf, ",initiator=%zd", initiator);
+        }
+
         if (needBackend)
             virBufferAsprintf(&buf, ",memdev=ram-node%zu", i);
         else
@@ -7218,6 +7377,18 @@ qemuBuildNumaCommandLine(virQEMUDriverConfigPtr cfg,
         }
     }
 
+    if (hmat) {
+        if (qemuBuildNumaHMATCommandLine(cmd, def) < 0)
+            goto cleanup;
+
+        /* This can't be moved into any of the loops above,
+         * because hmat-cache can be specified only after hmat-lb. */
+        for (i = 0; i < ncells; i++) {
+            if (qemuBuilNumaCellCache(cmd, def, i) < 0)
+                goto cleanup;
+        }
+    }
+
     ret = 0;
 
  cleanup:

src/qemu/qemu_validate.c

@@ -888,6 +888,13 @@ qemuValidateDomainDef(const virDomainDef *def,
         }
     }
 
+    if (virDomainNumaHasHMAT(def->numa) &&
+        !virQEMUCapsGet(qemuCaps, QEMU_CAPS_NUMA_HMAT)) {
+        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
+                       _("HMAT is not supported with this QEMU"));
+        return -1;
+    }
+
     if (def->genidRequested &&
         !virQEMUCapsGet(qemuCaps, QEMU_CAPS_DEVICE_VMGENID)) {
         virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",

tests/qemuxml2argvdata/numatune-hmat.x86_64-latest.args

@@ -0,0 +1,53 @@
+LC_ALL=C \
+PATH=/bin \
+HOME=/tmp/lib/domain--1-QEMUGuest \
+USER=test \
+LOGNAME=test \
+XDG_DATA_HOME=/tmp/lib/domain--1-QEMUGuest/.local/share \
+XDG_CACHE_HOME=/tmp/lib/domain--1-QEMUGuest/.cache \
+XDG_CONFIG_HOME=/tmp/lib/domain--1-QEMUGuest/.config \
+QEMU_AUDIO_DRV=none \
+/usr/bin/qemu-system-x86_64 \
+-name guest=QEMUGuest,debug-threads=on \
+-S \
+-object secret,id=masterKey0,format=raw,\
+file=/tmp/lib/domain--1-QEMUGuest/master-key.aes \
+-machine pc,accel=tcg,usb=off,dump-guest-core=off,hmat=on \
+-cpu qemu64 \
+-m 12288 \
+-overcommit mem-lock=off \
+-smp 12,sockets=12,cores=1,threads=1 \
+-object memory-backend-ram,id=ram-node0,size=2147483648 \
+-numa node,nodeid=0,cpus=0-3,initiator=0,memdev=ram-node0 \
+-object memory-backend-ram,id=ram-node1,size=2147483648 \
+-numa node,nodeid=1,cpus=4-7,initiator=1,memdev=ram-node1 \
+-object memory-backend-ram,id=ram-node2,size=2147483648 \
+-numa node,nodeid=2,cpus=8-11,initiator=2,memdev=ram-node2 \
+-object memory-backend-ram,id=ram-node3,size=2147483648 \
+-numa node,nodeid=3,initiator=0,memdev=ram-node3 \
+-object memory-backend-ram,id=ram-node4,size=2147483648 \
+-numa node,nodeid=4,initiator=0,memdev=ram-node4 \
+-object memory-backend-ram,id=ram-node5,size=2147483648 \
+-numa node,nodeid=5,initiator=0,memdev=ram-node5 \
+-numa hmat-lb,initiator=0,target=0,hierarchy=memory,data-type=access-latency,\
+latency=5 \
+-numa hmat-lb,initiator=0,target=0,hierarchy=first-level,\
+data-type=access-latency,latency=10 \
+-numa hmat-lb,initiator=0,target=0,hierarchy=memory,data-type=access-bandwidth,\
+bandwidth=204800K \
+-numa hmat-cache,node-id=0,size=10K,level=1,associativity=direct,\
+policy=write-back,line=8 \
+-uuid c7a5fdb2-cdaf-9455-926a-d65c16db1809 \
+-display none \
+-no-user-config \
+-nodefaults \
+-chardev socket,id=charmonitor,fd=1729,server,nowait \
+-mon chardev=charmonitor,id=monitor,mode=control \
+-rtc base=utc \
+-no-shutdown \
+-boot strict=on \
+-device piix3-usb-uhci,id=usb,bus=pci.0,addr=0x1.0x2 \
+-device virtio-balloon-pci,id=balloon0,bus=pci.0,addr=0x3 \
+-sandbox on,obsolete=deny,elevateprivileges=deny,spawn=deny,\
+resourcecontrol=deny \
+-msg timestamp=on

tests/qemuxml2argvtest.c

@@ -1946,6 +1946,7 @@ mymain(void)
 
     DO_TEST("numatune-distances", QEMU_CAPS_NUMA, QEMU_CAPS_NUMA_DIST);
     DO_TEST("numatune-no-vcpu", NONE);
+    DO_TEST_CAPS_LATEST("numatune-hmat");
 
     DO_TEST("numatune-auto-nodeset-invalid", NONE);
     DO_TEST("numatune-auto-prefer", QEMU_CAPS_OBJECT_MEMORY_RAM,

tests/qemuxml2xmltest.c

@@ -1127,7 +1127,7 @@ mymain(void)
     DO_TEST("numatune-memnode-no-memory", QEMU_CAPS_OBJECT_MEMORY_FILE);
     DO_TEST("numatune-distances", QEMU_CAPS_NUMA, QEMU_CAPS_NUMA_DIST);
     DO_TEST("numatune-no-vcpu", QEMU_CAPS_NUMA);
-    DO_TEST("numatune-hmat", NONE);
+    DO_TEST("numatune-hmat", QEMU_CAPS_NUMA_HMAT);
 
     DO_TEST("bios-nvram", NONE);
     DO_TEST("bios-nvram-os-interleave", NONE);