[libvirt] [PATCH v1 6/7] qemu: Implement ./hugepages/page/[@size, @unit, @nodeset]

Michal Privoznik mprivozn at redhat.com
Thu Jul 17 16:12:47 UTC 2014


Signed-off-by: Michal Privoznik <mprivozn at redhat.com>
---
 src/qemu/qemu_capabilities.c                       |  2 +
 src/qemu/qemu_capabilities.h                       |  1 +
 src/qemu/qemu_command.c                            | 91 +++++++++++++++++++---
 .../qemuxml2argv-hugepages-pages.args              | 16 ++++
 tests/qemuxml2argvdata/qemuxml2argv-hugepages.args |  2 +-
 tests/qemuxml2argvtest.c                           | 10 ++-
 6 files changed, 109 insertions(+), 13 deletions(-)
 create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-hugepages-pages.args

diff --git a/src/qemu/qemu_capabilities.c b/src/qemu/qemu_capabilities.c
index 07306e5..f69c4d0 100644
--- a/src/qemu/qemu_capabilities.c
+++ b/src/qemu/qemu_capabilities.c
@@ -263,6 +263,7 @@ VIR_ENUM_IMPL(virQEMUCaps, QEMU_CAPS_LAST,
 
               "memory-backend-ram", /* 170 */
               "numa",
+              "memory-backend-file",
     );
 
 
@@ -1481,6 +1482,7 @@ struct virQEMUCapsStringFlags virQEMUCapsObjectTypes[] = {
     { "pvpanic", QEMU_CAPS_DEVICE_PANIC },
     { "usb-kbd", QEMU_CAPS_DEVICE_USB_KBD },
     { "memory-backend-ram", QEMU_CAPS_OBJECT_MEMORY_RAM },
+    { "memory-backend-file", QEMU_CAPS_OBJECT_MEMORY_FILE },
 };
 
 static struct virQEMUCapsStringFlags virQEMUCapsObjectPropsVirtioBlk[] = {
diff --git a/src/qemu/qemu_capabilities.h b/src/qemu/qemu_capabilities.h
index 4332633..e80a377 100644
--- a/src/qemu/qemu_capabilities.h
+++ b/src/qemu/qemu_capabilities.h
@@ -211,6 +211,7 @@ typedef enum {
     QEMU_CAPS_CHANGE_BACKING_FILE = 169, /* change name of backing file in metadata */
     QEMU_CAPS_OBJECT_MEMORY_RAM  = 170, /* -object memory-backend-ram */
     QEMU_CAPS_NUMA               = 171, /* newer -numa handling with disjoint cpu ranges */
+    QEMU_CAPS_OBJECT_MEMORY_FILE = 172, /* -object memory-backend-file */
 
     QEMU_CAPS_LAST,                   /* this must always be the last item */
 } virQEMUCapsFlags;
diff --git a/src/qemu/qemu_command.c b/src/qemu/qemu_command.c
index 0b8cef5..cb35727 100644
--- a/src/qemu/qemu_command.c
+++ b/src/qemu/qemu_command.c
@@ -6381,24 +6381,36 @@ qemuBuildSmpArgStr(const virDomainDef *def,
 }
 
 static int
-qemuBuildNumaArgStr(const virDomainDef *def,
+qemuBuildNumaArgStr(virQEMUDriverConfigPtr cfg,
+                    const virDomainDef *def,
                     virCommandPtr cmd,
                     virQEMUCapsPtr qemuCaps)
 {
-    size_t i;
+    size_t i, j;
     virBuffer buf = VIR_BUFFER_INITIALIZER;
+    virDomainHugePagePtr master_hugepage = NULL;
     char *cpumask = NULL, *tmpmask = NULL, *next = NULL;
     char *nodemask = NULL;
+    char *mem_path = NULL;
     int ret = -1;
 
     if (virDomainNumatuneHasPerNodeBinding(def->numatune) &&
-        !virQEMUCapsGet(qemuCaps, QEMU_CAPS_OBJECT_MEMORY_RAM)) {
+        !(virQEMUCapsGet(qemuCaps, QEMU_CAPS_OBJECT_MEMORY_RAM) ||
+          virQEMUCapsGet(qemuCaps, QEMU_CAPS_OBJECT_MEMORY_FILE))) {
         virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                        _("Per-node memory binding is not supported "
                          "with this QEMU"));
         goto cleanup;
     }
 
+    if (def->mem.nhugepages && def->mem.hugepages[0].size &&
+        !virQEMUCapsGet(qemuCaps, QEMU_CAPS_OBJECT_MEMORY_FILE)) {
+        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
+                       _("huge pages pre NUMA node are not "
+                         "supported with this QEMU"));
+        goto cleanup;
+    }
+
     for (i = 0; i < def->cpu->ncells; i++) {
         int cellmem = VIR_DIV_UP(def->cpu->cells[i].mem, 1024);
         def->cpu->cells[i].mem = cellmem * 1024;
@@ -6417,15 +6429,74 @@ qemuBuildNumaArgStr(const virDomainDef *def,
             goto cleanup;
         }
 
-        if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_OBJECT_MEMORY_RAM)) {
+        if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_OBJECT_MEMORY_RAM) ||
+            virQEMUCapsGet(qemuCaps, QEMU_CAPS_OBJECT_MEMORY_FILE)) {
             virDomainNumatuneMemMode mode;
+            virDomainHugePagePtr hugepage = NULL;
             const char *policy = NULL;
 
             mode = virDomainNumatuneGetMode(def->numatune, i);
             policy = qemuNumaPolicyTypeToString(mode);
 
-            virBufferAsprintf(&buf, "memory-backend-ram,size=%dM,id=ram-node%zu",
-                              cellmem, i);
+            /* Find the huge page size we want to use */
+            for (j = 0; j < def->mem.nhugepages; j++) {
+                bool thisHugepage = false;
+
+                hugepage = &def->mem.hugepages[j];
+
+                if (!hugepage->nodemask) {
+                    master_hugepage = hugepage;
+                    continue;
+                }
+
+                if (virBitmapGetBit(hugepage->nodemask, i, &thisHugepage) < 0) {
+                    /* Ignore this error. It's not an error after all. Well,
+                     * the nodemask for this <page/> can contain lower NUMA
+                     * nodes than we are querying in here. */
+                    continue;
+                }
+
+                if (thisHugepage) {
+                    /* Hooray, we've found the page size */
+                    break;
+                }
+            }
+
+            if (j == def->mem.nhugepages) {
+                /* We have not found specific huge page to be used with this
+                 * NUMA node. Use the generic setting then (<page/> without any
+                 * @nodemask) if possible. */
+                hugepage = master_hugepage;
+            }
+
+            if (hugepage) {
+                /* Now lets see, if the huge page we want to use is even mounted
+                 * and ready to use */
+
+                for (j = 0; j < cfg->nhugetlbfs; j++) {
+                    if (cfg->hugetlbfs[j].size == hugepage->size)
+                        break;
+                }
+
+                if (j == cfg->nhugetlbfs) {
+                    virReportError(VIR_ERR_INTERNAL_ERROR,
+                                   _("Unable to find any usable hugetlbfs mount for %llu KiB"),
+                                   hugepage->size);
+                    goto cleanup;
+                }
+
+                VIR_FREE(mem_path);
+                if (!(mem_path = qemuGetHugepagePath(&cfg->hugetlbfs[j])))
+                    goto cleanup;
+
+                virBufferAsprintf(&buf,
+                                  "memory-backend-file,prealloc=yes,mem-path=%s",
+                                  mem_path);
+            } else {
+                virBufferAddLit(&buf, "memory-backend-ram");
+            }
+
+            virBufferAsprintf(&buf, ",size=%dM,id=ram-node%zu", cellmem, i);
 
             if (virDomainNumatuneMaybeFormatNodeset(def->numatune, NULL,
                                                     &nodemask, i) < 0)
@@ -6464,7 +6535,8 @@ qemuBuildNumaArgStr(const virDomainDef *def,
             virBufferAdd(&buf, tmpmask, -1);
         }
 
-        if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_OBJECT_MEMORY_RAM)) {
+        if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_OBJECT_MEMORY_RAM) ||
+            virQEMUCapsGet(qemuCaps, QEMU_CAPS_OBJECT_MEMORY_FILE)) {
             virBufferAsprintf(&buf, ",memdev=ram-node%zu", i);
         } else {
             virBufferAsprintf(&buf, ",mem=%d", cellmem);
@@ -6477,6 +6549,7 @@ qemuBuildNumaArgStr(const virDomainDef *def,
  cleanup:
     VIR_FREE(cpumask);
     VIR_FREE(nodemask);
+    VIR_FREE(mem_path);
     virBufferFreeAndReset(&buf);
     return ret;
 }
@@ -7332,7 +7405,7 @@ qemuBuildCommandLine(virConnectPtr conn,
     virCommandAddArg(cmd, "-m");
     def->mem.max_balloon = VIR_DIV_UP(def->mem.max_balloon, 1024) * 1024;
     virCommandAddArgFormat(cmd, "%llu", def->mem.max_balloon / 1024);
-    if (def->mem.nhugepages) {
+    if (def->mem.nhugepages && !def->mem.hugepages[0].size) {
         char *mem_path;
 
         if (!cfg->nhugetlbfs) {
@@ -7376,7 +7449,7 @@ qemuBuildCommandLine(virConnectPtr conn,
     VIR_FREE(smp);
 
     if (def->cpu && def->cpu->ncells)
-        if (qemuBuildNumaArgStr(def, cmd, qemuCaps) < 0)
+        if (qemuBuildNumaArgStr(cfg, def, cmd, qemuCaps) < 0)
             goto error;
 
     if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_UUID))
diff --git a/tests/qemuxml2argvdata/qemuxml2argv-hugepages-pages.args b/tests/qemuxml2argvdata/qemuxml2argv-hugepages-pages.args
new file mode 100644
index 0000000..042683a
--- /dev/null
+++ b/tests/qemuxml2argvdata/qemuxml2argv-hugepages-pages.args
@@ -0,0 +1,16 @@
+LC_ALL=C PATH=/bin HOME=/home/test USER=test LOGNAME=test QEMU_AUDIO_DRV=none \
+/usr/bin/qemu -S -M pc -m 4096 -smp 4 \
+-object memory-backend-file,prealloc=yes,mem-path=/dev/hugepages1G/libvirt/qemu,\
+size=1024M,id=ram-node0,host-nodes=0-3,policy=bind \
+-numa node,nodeid=0,cpus=0,memdev=ram-node0 \
+-object memory-backend-file,prealloc=yes,mem-path=/dev/hugepages2M/libvirt/qemu,\
+size=1024M,id=ram-node1,host-nodes=0-3,policy=bind \
+-numa node,nodeid=1,cpus=1,memdev=ram-node1 \
+-object memory-backend-file,prealloc=yes,mem-path=/dev/hugepages1G/libvirt/qemu,\
+size=1024M,id=ram-node2,host-nodes=0-3,policy=bind \
+-numa node,nodeid=2,cpus=2,memdev=ram-node2 \
+-object memory-backend-file,prealloc=yes,mem-path=/dev/hugepages1G/libvirt/qemu,\
+size=1024M,id=ram-node3,host-nodes=3,policy=bind \
+-numa node,nodeid=3,cpus=3,memdev=ram-node3 \
+-nographic -monitor unix:/tmp/test-monitor,server,nowait -no-acpi -boot c -usb \
+-hda /dev/HostVG/QEMUGuest1 -net none -serial none -parallel none
diff --git a/tests/qemuxml2argvdata/qemuxml2argv-hugepages.args b/tests/qemuxml2argvdata/qemuxml2argv-hugepages.args
index d42d9fc..51c5d62 100644
--- a/tests/qemuxml2argvdata/qemuxml2argv-hugepages.args
+++ b/tests/qemuxml2argvdata/qemuxml2argv-hugepages.args
@@ -1,5 +1,5 @@
 LC_ALL=C PATH=/bin HOME=/home/test USER=test LOGNAME=test QEMU_AUDIO_DRV=none \
 /usr/bin/qemu -S -M \
-pc -m 214 -mem-prealloc -mem-path /dev/hugepages/libvirt/qemu -smp 1 \
+pc -m 214 -mem-prealloc -mem-path /dev/hugepages2M/libvirt/qemu -smp 1 \
 -nographic -monitor unix:/tmp/test-monitor,server,nowait -no-acpi -boot c -usb -hda \
 /dev/HostVG/QEMUGuest1 -net none -serial none -parallel none
diff --git a/tests/qemuxml2argvtest.c b/tests/qemuxml2argvtest.c
index 1a5a4b0..63c9c4b 100644
--- a/tests/qemuxml2argvtest.c
+++ b/tests/qemuxml2argvtest.c
@@ -525,13 +525,15 @@ mymain(void)
     if (VIR_STRDUP_QUIET(driver.config->stateDir, "/nowhere") < 0)
         return EXIT_FAILURE;
     VIR_FREE(driver.config->hugetlbfs);
-    if (VIR_ALLOC_N(driver.config->hugetlbfs, 1) < 0)
+    if (VIR_ALLOC_N(driver.config->hugetlbfs, 2) < 0)
         return EXIT_FAILURE;
-    driver.config->nhugetlbfs = 1;
-    if (VIR_STRDUP(driver.config->hugetlbfs[0].mnt_dir, "/dev/hugepages") < 0)
+    driver.config->nhugetlbfs = 2;
+    if (VIR_STRDUP(driver.config->hugetlbfs[0].mnt_dir, "/dev/hugepages2M") < 0 ||
+        VIR_STRDUP(driver.config->hugetlbfs[1].mnt_dir, "/dev/hugepages1G") < 0)
         return EXIT_FAILURE;
     driver.config->hugetlbfs[0].size = 2048;
     driver.config->hugetlbfs[0].deflt = true;
+    driver.config->hugetlbfs[1].size = 1048576;
     driver.config->spiceTLS = 1;
     if (VIR_STRDUP_QUIET(driver.config->spicePassword, "123456") < 0)
         return EXIT_FAILURE;
@@ -665,6 +667,8 @@ mymain(void)
     DO_TEST("hyperv-off", NONE);
 
     DO_TEST("hugepages", QEMU_CAPS_MEM_PATH);
+    DO_TEST("hugepages-pages", QEMU_CAPS_MEM_PATH, QEMU_CAPS_OBJECT_MEMORY_RAM,
+            QEMU_CAPS_OBJECT_MEMORY_FILE);
     DO_TEST("nosharepages", QEMU_CAPS_MACHINE_OPT, QEMU_CAPS_MEM_MERGE);
     DO_TEST("disk-cdrom", NONE);
     DO_TEST("disk-cdrom-network-http", QEMU_CAPS_KVM, QEMU_CAPS_DEVICE,
-- 
1.8.5.5




More information about the libvir-list mailing list