Support configuration of huge pages in guests

Add option to domain XML for

     <memoryBacking>
        <hugepages/>
     </memoryBacking>

* configure.in: Add check for mntent.h
* qemud/libvirtd_qemu.aug, qemud/test_libvirtd_qemu.aug, src/qemu.conf
  Add 'hugetlbfs_mount' config parameter
* src/qemu_conf.c, src/qemu_conf.h: Check for -mem-path flag in QEMU,
  and pass it when hugepages are requested.
  Load hugetlbfs_mount config parameter, search for mount if not given.
* src/qemu_driver.c: Free hugetlbfs_mount/path parameter in driver shutdown.
  Create directory for QEMU hugepage usage, chowning if required.
* docs/formatdomain.html.in: Document memoryBacking/hugepages elements
* docs/schemas/domain.rng: Add memoryBacking/hugepages elements to schema
* src/util.c, src/util.h, src/libvirt_private.syms: Add virFileFindMountPoint
  helper API
* tests/qemuhelptest.c: Add -mem-path constants
* tests/qemuxml2argvtest.c, tests/qemuxml2xmltest.c: Add tests for hugepage
  handling
* tests/qemuxml2argvdata/qemuxml2argv-hugepages.xml,
  tests/qemuxml2argvdata/qemuxml2argv-hugepages.args: Data files for
  hugepage tests

configure.in

@@ -83,7 +83,7 @@ dnl Availability of various not common threadsafe functions
 AC_CHECK_FUNCS([strerror_r strtok_r getmntent_r getgrnam_r getpwuid_r])
 
 dnl Availability of various common headers (non-fatal if missing).
-AC_CHECK_HEADERS([pwd.h paths.h regex.h sys/syslimits.h sys/utsname.h sys/wait.h winsock2.h sched.h termios.h sys/poll.h syslog.h])
+AC_CHECK_HEADERS([pwd.h paths.h regex.h sys/syslimits.h sys/utsname.h sys/wait.h winsock2.h sched.h termios.h sys/poll.h syslog.h mntent.h])
 
 dnl Where are the XDR functions?
 dnl If portablexdr is installed, prefer that.

docs/formatdomain.html

@@ -338,13 +338,19 @@
         ...
 	<memory>524288</memory>
 	<currentMemory>524288</currentMemory>
+	<memoryBacking>
+	  <hugepages/>
+        </memoryBacking>
 	<vcpu>1</vcpu>
 	...</pre>
         <dl><dt><code>memory</code></dt><dd>The maximum allocation of memory for the guest at boot time.
 	The units for this value are kilobytes (i.e. blocks of 1024 bytes)</dd><dt><code>currentMemory</code></dt><dd>The actual allocation of memory for the guest. This value
 	be less than the maximum allocation, to allow for ballooning
 	up the guests memory on the fly. If this is omitted, it defaults
-	to the same value as the <code>memory<code> element</code></code></dd><dt><code>vcpu</code></dt><dd>The content of this element defines the number of virtual
+	to the same value as the <code>memory<code> element</code></code></dd><dt><code>memoryBacking</code></dt><dd>The optional <code>memoryBacking</code> element, may have an
+	<code>hugepages</code> element set within it. This tells the
+	hypervisor that the guest should have its memory allocated using
+	hugepages instead of the normal native page size.</dd><dt><code>vcpu</code></dt><dd>The content of this element defines the number of virtual
 	CPUs allocated for the guest OS.</dd></dl>
         <h3>
           <a name="elementsLifecycle" id="elementsLifecycle">Lifecycle control</a>

docs/formatdomain.html.in

@@ -176,6 +176,9 @@
         ...
 	<memory>524288</memory>
 	<currentMemory>524288</currentMemory>
+	<memoryBacking>
+	  <hugepages/>
+        </memoryBacking>
 	<vcpu>1</vcpu>
 	...</pre>
 
@@ -188,6 +191,11 @@
 	be less than the maximum allocation, to allow for ballooning
 	up the guests memory on the fly. If this is omitted, it defaults
 	to the same value as the <code>memory<code> element</dd>
+      <dt><code>memoryBacking</code></dt>
+      <dd>The optional <code>memoryBacking</code> element, may have an
+	<code>hugepages</code> element set within it. This tells the
+	hypervisor that the guest should have its memory allocated using
+	hugepages instead of the normal native page size.</dd>
       <dt><code>vcpu</code></dt>
       <dd>The content of this element defines the number of virtual
 	CPUs allocated for the guest OS.</dd>

docs/schemas/domain.rng

@@ -245,6 +245,15 @@
           <ref name="memoryKB"/>
         </element>
       </optional>
+      <optional>
+	<element name="memoryBacking">
+	  <optional>
+	    <element name="hugepages">
+	      <empty/>
+	    </element>
+	  </optional>
+	</element>
+      </optional>
       <optional>
         <element name="vcpu">
           <optional>

qemud/libvirtd_qemu.aug

@@ -35,6 +35,7 @@ module Libvirtd_qemu =
                  | str_array_entry "cgroup_controllers"
                  | str_array_entry "cgroup_device_acl"
                  | str_entry "save_image_format"
+                 | str_entry "hugetlbfs_mount"
 
    (* Each enty in the config is one of the following three ... *)
    let entry = vnc_entry

qemud/test_libvirtd_qemu.aug

@@ -91,6 +91,8 @@ cgroup_controllers = [ \"cpu\", \"devices\" ]
 cgroup_device_acl = [ \"/dev/null\", \"/dev/full\", \"/dev/zero\" ]
 
 save_image_format = \"gzip\"
+
+hugetlbfs_mount = \"/dev/hugepages\"
 "
 
    test Libvirtd_qemu.lns get conf =
@@ -192,3 +194,5 @@ save_image_format = \"gzip\"
 }
 { "#empty" }
 { "save_image_format" = "gzip" }
+{ "#empty" }
+{ "hugetlbfs_mount" = "/dev/hugepages" }
\ No newline at end of file

src/domain_conf.c

@@ -2537,6 +2537,10 @@ static virDomainDefPtr virDomainDefParseXML(virConnectPtr conn,
     if (virXPathULong(conn, "string(./currentMemory[1])", ctxt, &def->memory) < 0)
         def->memory = def->maxmem;
 
+    node = virXPathNode(conn, "./memoryBacking/hugepages", ctxt);
+    if (node)
+        def->hugepage_backed = 1;
+
     if (virXPathULong(conn, "string(./vcpu[1])", ctxt, &def->vcpus) < 0)
         def->vcpus = 1;
 
@@ -4161,7 +4165,11 @@ char *virDomainDefFormat(virConnectPtr conn,
     virBufferVSprintf(&buf, "  <memory>%lu</memory>\n", def->maxmem);
     virBufferVSprintf(&buf, "  <currentMemory>%lu</currentMemory>\n",
                       def->memory);
-
+    if (def->hugepage_backed) {
+        virBufferAddLit(&buf, "  <memoryBacking>\n");
+        virBufferAddLit(&buf, "    <hugepages/>\n");
+        virBufferAddLit(&buf, "  </memoryBacking>\n");
+    }
     for (n = 0 ; n < def->cpumasklen ; n++)
         if (def->cpumask[n] != 1)
             allones = 0;

src/domain_conf.h

@@ -535,6 +535,7 @@ struct _virDomainDef {
 
     unsigned long memory;
     unsigned long maxmem;
+    unsigned char hugepage_backed;
     unsigned long vcpus;
     int cpumasklen;
     char *cpumask;

src/libvirt_private.syms

@@ -408,6 +408,7 @@ virGetUserDirectory;
 virGetUserName;
 virGetUserID;
 virGetGroupID;
+virFileFindMountPoint;
 
 
 # uuid.h

src/qemu.conf

@@ -139,3 +139,15 @@
 # the process of saving a domain in order to save disk space.
 #
 # save_image_format = "raw"
+
+# If provided by the host and a hugetlbfs mount point is configured,
+# a guest may request huge page backing.  When this mount point is
+# unspecified here, determination of a host mount point in /proc/mounts
+# will be attempted.  Specifying an explicit mount overrides detection
+# of the same in /proc/mounts.  Setting the mount point to "" will
+# disable guest hugepage backing.
+#
+# NB, within this mount point, guests will create memory backing files
+# in a location of  $MOUNTPOINT/libvirt/qemu
+
+# hugetlbfs_mount = "/dev/hugepages"

src/qemu_conf.c

@@ -35,6 +35,7 @@
 #include <sys/wait.h>
 #include <arpa/inet.h>
 #include <sys/utsname.h>
+#include <mntent.h>
 
 #include "c-ctype.h"
 #include "virterror_internal.h"
@@ -87,6 +88,7 @@ VIR_ENUM_IMPL(qemuVideo, VIR_DOMAIN_VIDEO_TYPE_LAST,
               NULL, /* no arg needed for xen */
               NULL /* don't support vbox */);
 
+#define PROC_MOUNT_BUF_LEN 255
 
 int qemudLoadDriverConfig(struct qemud_driver *driver,
                           const char *filename) {
@@ -106,6 +108,21 @@ int qemudLoadDriverConfig(struct qemud_driver *driver,
         return -1;
     }
 
+#ifdef HAVE_MNTENT_H
+    /* For privileged driver, try and find hugepage mount automatically.
+     * Non-privileged driver requires admin to create a dir for the
+     * user, chown it, and then let user configure it manually */
+    if (driver->privileged &&
+        !(driver->hugetlbfs_mount = virFileFindMountPoint("hugetlbfs"))) {
+        if (errno != ENOENT) {
+            virReportSystemError(NULL, errno, "%s",
+                                 _("unable to find hugetlbfs mountpoint"));
+            return -1;
+        }
+    }
+#endif
+
+
     /* Just check the file is readable before opening it, otherwise
      * libvirt emits an error.
      */
@@ -290,6 +307,17 @@ int qemudLoadDriverConfig(struct qemud_driver *driver,
         }
     }
 
+     p = virConfGetValue (conf, "hugetlbfs_mount");
+     CHECK_TYPE ("hugetlbfs_mount", VIR_CONF_STRING);
+     if (p && p->str) {
+         VIR_FREE(driver->hugetlbfs_mount);
+         if (!(driver->hugetlbfs_mount = strdup(p->str))) {
+             virReportOOMError(NULL);
+             virConfFree(conf);
+             return -1;
+         }
+     }
+
     virConfFree (conf);
     return 0;
 }
@@ -784,6 +812,8 @@ static unsigned int qemudComputeCmdFlags(const char *help,
         flags |= QEMUD_CMD_FLAG_DRIVE_BOOT;
     if (strstr(help, "-pcidevice"))
         flags |= QEMUD_CMD_FLAG_PCIDEVICE;
+    if (strstr(help, "-mem-path"))
+        flags |= QEMUD_CMD_FLAG_MEM_PATH;
 
     if (version >= 9000)
         flags |= QEMUD_CMD_FLAG_VNC_COLON;
@@ -1583,6 +1613,26 @@ int qemudBuildCommandLine(virConnectPtr conn,
         ADD_ARG_LIT("-no-kvm");
     ADD_ARG_LIT("-m");
     ADD_ARG_LIT(memory);
+    if (def->hugepage_backed) {
+        if (!driver->hugetlbfs_mount) {
+            qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
+                             "%s", _("hugetlbfs filesystem is not mounted"));
+            goto error;
+        }
+        if (!driver->hugepage_path) {
+            qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
+                             "%s", _("hugepages are disabled by administrator config"));
+            goto error;
+        }
+        if (!(qemuCmdFlags & QEMUD_CMD_FLAG_MEM_PATH)) {
+            qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
+                             _("hugepage backing not supported by '%s'"),
+                             def->emulator);
+            goto error;
+        }
+        ADD_ARG_LIT("-mem-path");
+        ADD_ARG_LIT(driver->hugepage_path);
+    }
     ADD_ARG_LIT("-smp");
     ADD_ARG_LIT(vcpus);
 

src/qemu_conf.h

@@ -67,6 +67,7 @@ enum qemud_cmd_flags {
     QEMUD_CMD_FLAG_HOST_NET_ADD = QEMUD_CMD_FLAG_0_10, /* host_net_add monitor command */
 
     QEMUD_CMD_FLAG_PCIDEVICE     = (1 << 17), /* PCI device assignment only supported by qemu-kvm */
+    QEMUD_CMD_FLAG_MEM_PATH      = (1 << 18), /* mmap'ped guest backing supported */
 };
 
 /* Main driver state */
@@ -99,6 +100,8 @@ struct qemud_driver {
     char *vncListen;
     char *vncPassword;
     char *vncSASLdir;
+    char *hugetlbfs_mount;
+    char *hugepage_path;
 
     virCapsPtr caps;
 

src/qemu_driver.c

@@ -555,6 +555,37 @@ qemudStartup(int privileged) {
         goto error;
     }
 
+    /* If hugetlbfs is present, then we need to create a sub-directory within
+     * it, since we can't assume the root mount point has permissions that
+     * will let our spawned QEMU instances use it.
+     *
+     * NB the check for '/', since user may config "" to disable hugepages
+     * even when mounted
+     */
+    if (qemu_driver->hugetlbfs_mount &&
+        qemu_driver->hugetlbfs_mount[0] == '/') {
+        char *mempath = NULL;
+        if (virAsprintf(&mempath, "%s/libvirt/qemu", qemu_driver->hugetlbfs_mount) < 0)
+            goto out_of_memory;
+
+        if ((rc = virFileMakePath(mempath)) != 0) {
+            virReportSystemError(NULL, rc,
+                                 _("unable to create hugepage path %s"), mempath);
+            VIR_FREE(mempath);
+            goto error;
+        }
+        if (qemu_driver->privileged &&
+            chown(mempath, qemu_driver->user, qemu_driver->group) < 0) {
+            virReportSystemError(NULL, errno,
+                                 _("unable to set ownership on %s to %d:%d"),
+                                 mempath, qemu_driver->user, qemu_driver->group);
+            VIR_FREE(mempath);
+            goto error;
+        }
+
+        qemu_driver->hugepage_path = mempath;
+    }
+
     /* Get all the running persistent or transient configs first */
     if (virDomainLoadAllConfigs(NULL,
                                 qemu_driver->caps,
@@ -686,6 +717,8 @@ qemudShutdown(void) {
     VIR_FREE(qemu_driver->vncPassword);
     VIR_FREE(qemu_driver->vncSASLdir);
     VIR_FREE(qemu_driver->saveImageFormat);
+    VIR_FREE(qemu_driver->hugetlbfs_mount);
+    VIR_FREE(qemu_driver->hugepage_path);
 
     /* Free domain callback list */
     virDomainEventCallbackListFree(qemu_driver->domainEventCallbacks);

src/util.c

@@ -60,7 +60,9 @@
 #if HAVE_CAPNG
 #include <cap-ng.h>
 #endif
-
+#ifdef HAVE_MNTENT_H
+#include <mntent.h>
+#endif
 
 #include "virterror_internal.h"
 #include "logging.h"
@@ -1983,3 +1985,37 @@ int virGetGroupID(virConnectPtr conn,
     return 0;
 }
 #endif
+
+
+#ifdef HAVE_MNTENT_H
+/* search /proc/mounts for mount point of *type; return pointer to
+ * malloc'ed string of the path if found, otherwise return NULL
+ * with errno set to an appropriate value.
+ */
+char *virFileFindMountPoint(const char *type)
+{
+    FILE *f;
+    struct mntent mb;
+    char mntbuf[1024];
+    char *ret = NULL;
+
+    f = setmntent("/proc/mounts", "r");
+    if (!f)
+        return NULL;
+
+    while (getmntent_r(f, &mb, mntbuf, sizeof(mntbuf))) {
+        if (STREQ(mb.mnt_type, type)) {
+            ret = strdup(mb.mnt_dir);
+            goto cleanup;
+        }
+    }
+
+    if (!ret)
+        errno = ENOENT;
+
+cleanup:
+    endmntent(f);
+
+    return ret;
+}
+#endif

src/util.h

@@ -233,4 +233,8 @@ int virGetGroupID(virConnectPtr conn,
 int virRandomInitialize(unsigned int seed);
 int virRandom(int max);
 
+#ifdef HAVE_MNTENT_H
+char *virFileFindMountPoint(const char *type);
+#endif
+
 #endif /* __VIR_UTIL_H__ */

tests/qemuhelptest.c

@@ -105,7 +105,8 @@ mymain(int argc, char **argv)
             QEMUD_CMD_FLAG_VNET_HDR |
             QEMUD_CMD_FLAG_MIGRATE_KVM_STDIO |
             QEMUD_CMD_FLAG_KVM |
-            QEMUD_CMD_FLAG_DRIVE_FORMAT,
+            QEMUD_CMD_FLAG_DRIVE_FORMAT |
+            QEMUD_CMD_FLAG_MEM_PATH,
             9001,  1, 74);
     DO_TEST("qemu-0.10.5",
             QEMUD_CMD_FLAG_KQEMU |
@@ -136,7 +137,8 @@ mymain(int argc, char **argv)
             QEMUD_CMD_FLAG_DRIVE_FORMAT |
             QEMUD_CMD_FLAG_VGA |
             QEMUD_CMD_FLAG_0_10 |
-            QEMUD_CMD_FLAG_PCIDEVICE,
+            QEMUD_CMD_FLAG_PCIDEVICE |
+            QEMUD_CMD_FLAG_MEM_PATH,
             10005, 1,  0);
     DO_TEST("kvm-86",
             QEMUD_CMD_FLAG_VNC_COLON |

tests/qemuxml2argvdata/qemuxml2argv-hugepages.args

+LC_ALL=C PATH=/bin HOME=/home/test USER=test LOGNAME=test /usr/bin/qemu -S -M pc -m 214 -mem-path /dev/hugepages/libvirt/qemu -smp 1 -nographic -monitor unix:/tmp/test-monitor,server,nowait -no-acpi -boot c -hda /dev/HostVG/QEMUGuest1 -net none -serial none -parallel none -usb

tests/qemuxml2argvdata/qemuxml2argv-hugepages.xml

+<domain type='qemu'>
+  <name>QEMUGuest1</name>
+  <uuid>c7a5fdbd-edaf-9455-926a-d65c16db1809</uuid>
+  <memory>219200</memory>
+  <currentMemory>219200</currentMemory>
+  <memoryBacking>
+    <hugepages/>
+  </memoryBacking>
+  <vcpu>1</vcpu>
+  <os>
+    <type arch='i686' machine='pc'>hvm</type>
+    <boot dev='hd'/>
+  </os>
+  <clock offset='utc'/>
+  <on_poweroff>destroy</on_poweroff>
+  <on_reboot>restart</on_reboot>
+  <on_crash>destroy</on_crash>
+  <devices>
+    <emulator>/usr/bin/qemu</emulator>
+    <disk type='block' device='disk'>
+      <source dev='/dev/HostVG/QEMUGuest1'/>
+      <target dev='hda' bus='ide'/>
+    </disk>
+  </devices>
+</domain>

tests/qemuxml2argvtest.c

@@ -160,7 +160,11 @@ mymain(int argc, char **argv)
 
     if ((driver.caps = testQemuCapsInit()) == NULL)
         return EXIT_FAILURE;
-    if((driver.stateDir = strdup("/nowhere")) == NULL)
+    if ((driver.stateDir = strdup("/nowhere")) == NULL)
+        return EXIT_FAILURE;
+    if ((driver.hugetlbfs_mount = strdup("/dev/hugepages")) == NULL)
+        return EXIT_FAILURE;
+    if ((driver.hugepage_path = strdup("/dev/hugepages/libvirt/qemu")) == NULL)
         return EXIT_FAILURE;
 
 #define DO_TEST_FULL(name, extraFlags, migrateFrom)                     \
@@ -189,6 +193,7 @@ mymain(int argc, char **argv)
     DO_TEST("bootloader", 0);
     DO_TEST("clock-utc", 0);
     DO_TEST("clock-localtime", 0);
+    DO_TEST("hugepages", QEMUD_CMD_FLAG_MEM_PATH);
     DO_TEST("disk-cdrom", 0);
     DO_TEST("disk-cdrom-empty", QEMUD_CMD_FLAG_DRIVE);
     DO_TEST("disk-floppy", 0);

tests/qemuxml2xmltest.c

@@ -92,6 +92,7 @@ mymain(int argc, char **argv)
     DO_TEST("bootloader");
     DO_TEST("clock-utc");
     DO_TEST("clock-localtime");
+    DO_TEST("hugepages");
     DO_TEST("disk-cdrom");
     DO_TEST("disk-floppy");
     DO_TEST("disk-many");