Introduce support for QEMU's new mapped-ram stream format [1].
mapped-ram can be enabled by setting the 'save_image_format'
setting in qemu.conf to 'sparse'.
To use mapped-ram with QEMU:
- The 'mapped-ram' migration capability must be set to true
- The 'multifd' migration capability must be set to true and
the 'multifd-channels' migration parameter must set to 1
- QEMU must be provided an fdset containing the migration fd
- The 'migrate' qmp command is invoked with a URI referencing the
fdset and an offset where to start reading or writing the data
stream, e.g.
{"execute":"migrate",
"arguments":{"detach":true,"resume":false,
"uri":"file:/dev/fdset/0,offset=0x11921"}}
The mapped-ram stream, in conjunction with direct IO and multifd
support provided by subsequent patches, can significantly improve
the time required to save VM memory state. The following tables
compare mapped-ram with the existing, sequential save stream. In
all cases, the save and restore operations are to/from a block
device comprised of two NVMe disks in RAID0 configuration with
xfs (~8600MiB/s). The values in the 'save time' and 'restore time'
columns were scraped from the 'real' time reported by time(1). The
'Size' and 'Blocks' columns were provided by the corresponding
outputs of stat(1).
VM: 32G RAM, 1 vcpu, idle (shortly after boot)
| save | restore |
| time | time | Size | Blocks
-----------------------+---------+---------+--------------+--------
legacy | 6.193s | 4.399s | 985744812 | 1925288
-----------------------+---------+---------+--------------+--------
mapped-ram | 5.109s | 1.176s | 34368554354 | 1774472
-----------------------+---------+---------+--------------+--------
legacy + direct IO | 5.725s | 4.512s | 985765251 | 1925328
-----------------------+---------+---------+--------------+--------
mapped-ram + direct IO | 4.627s | 1.490s | 34368554354 | 1774304
-----------------------+---------+---------+--------------+--------
mapped-ram + direct IO | | | |
+ multifd-channels=8 | 4.421s | 0.845s | 34368554318 | 1774312
-------------------------------------------------------------------
VM: 32G RAM, 30G dirty, 1 vcpu in tight loop dirtying memory
| save | restore |
| time | time | Size | Blocks
-----------------------+---------+---------+--------------+---------
legacy | 25.800s | 14.332s | 33154309983 | 64754512
-----------------------+---------+---------+--------------+---------
mapped-ram | 18.742s | 15.027s | 34368559228 | 64617160
-----------------------+---------+---------+--------------+---------
legacy + direct IO | 13.115s | 18.050s | 33154310496 | 64754520
-----------------------+---------+---------+--------------+---------
mapped-ram + direct IO | 13.623s | 15.959s | 34368557392 | 64662040
-----------------------+-------- +---------+--------------+---------
mapped-ram + direct IO | | | |
+ multifd-channels=8 | 6.994s | 6.470s | 34368554980 | 64665776
--------------------------------------------------------------------
As can be seen from the tables, one caveat of mapped-ram is the logical
file size of a saved image is basically equivalent to the VM memory size.
Note however that mapped-ram typically uses fewer blocks on disk, hence
the name 'sparse' for 'save_image_format'.
Also note the mapped-ram stream is incompatible with the existing stream
format, hence mapped-ram cannot be used to restore an image saved with
the existing format and vice versa.
[1] https://gitlab.com/qemu-project/qemu/-/blob/master/docs/devel/migration/mapped-ram.rst?ref_type=heads
Signed-off-by: Jim Fehlig <jfehlig@xxxxxxxx>
---
src/qemu/qemu_driver.c | 23 +++++-
src/qemu/qemu_migration.c | 149 ++++++++++++++++++++++++++------------
src/qemu/qemu_migration.h | 4 +-
src/qemu/qemu_monitor.c | 34 +++++++++
src/qemu/qemu_monitor.h | 4 +
src/qemu/qemu_saveimage.c | 31 +++++---
src/qemu/qemu_saveimage.h | 1 +
src/qemu/qemu_snapshot.c | 7 +-
8 files changed, 191 insertions(+), 62 deletions(-)
diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c
index 53bdae402a..2b2dda9d26 100644
--- a/src/qemu/qemu_driver.c
+++ b/src/qemu/qemu_driver.c
@@ -2599,6 +2599,7 @@ qemuDomainSaveInternal(virQEMUDriver *driver,
qemuDomainObjPrivate *priv = vm->privateData;
virQEMUSaveData *data = NULL;
g_autoptr(qemuDomainSaveCookie) cookie = NULL;
+ g_autoptr(qemuMigrationParams) saveParams = NULL;
if (virDomainObjBeginAsyncJob(vm, VIR_ASYNC_JOB_SAVE,
VIR_DOMAIN_JOB_OPERATION_SAVE, flags) < 0)
@@ -2607,6 +2608,14 @@ qemuDomainSaveInternal(virQEMUDriver *driver,
if (!qemuMigrationSrcIsAllowed(vm, false, VIR_ASYNC_JOB_SAVE, 0))
goto endjob;
+ if (format == QEMU_SAVE_FORMAT_SPARSE &&
+ !qemuMigrationCapsGet(vm, QEMU_MIGRATION_CAP_MAPPED_RAM)) {
+ virReportError(VIR_ERR_OPERATION_UNSUPPORTED,
+ _("save image format %1$s is not supported by this QEMU binary"),
+ qemuSaveFormatTypeToString(format));
+ goto endjob;
+ }
+
if (!virDomainObjIsActive(vm)) {
virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("guest unexpectedly quit"));
@@ -2670,8 +2679,11 @@ qemuDomainSaveInternal(virQEMUDriver *driver,
goto endjob;
xml = NULL;
+ if (!(saveParams = qemuMigrationParamsForSave(format == QEMU_SAVE_FORMAT_SPARSE)))
+ goto endjob;
+
ret = qemuSaveImageCreate(driver, vm, path, data, compressor,
- flags, VIR_ASYNC_JOB_SAVE);
+ saveParams, flags, VIR_ASYNC_JOB_SAVE);
if (ret < 0)
goto endjob;
@@ -3103,6 +3115,8 @@ doCoreDump(virQEMUDriver *driver,
memory_dump_format) < 0)
goto cleanup;
} else {
+ g_autoptr(qemuMigrationParams) dump_params = NULL;
+
if (dumpformat != VIR_DOMAIN_CORE_DUMP_FORMAT_RAW) {
virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
_("kdump-compressed format is only supported with memory-only dump"));
@@ -3112,8 +3126,11 @@ doCoreDump(virQEMUDriver *driver,
if (!qemuMigrationSrcIsAllowed(vm, false, VIR_ASYNC_JOB_DUMP, 0))
goto cleanup;
- if (qemuMigrationSrcToFile(driver, vm, fd, compressor,
- VIR_ASYNC_JOB_DUMP) < 0)
+ if (!(dump_params = qemuMigrationParamsNew()))
+ goto cleanup;
+
+ if (qemuMigrationSrcToFile(driver, vm, &fd, compressor,
+ dump_params, dump_flags, VIR_ASYNC_JOB_DUMP) < 0)
goto cleanup;
}
diff --git a/src/qemu/qemu_migration.c b/src/qemu/qemu_migration.c
index 8753fb5358..f692986056 100644
--- a/src/qemu/qemu_migration.c
+++ b/src/qemu/qemu_migration.c
@@ -7118,46 +7118,17 @@ qemuMigrationProcessUnattended(virQEMUDriver *driver,
}
-/* Helper function called while vm is active. */
-int
-qemuMigrationSrcToFile(virQEMUDriver *driver, virDomainObj *vm,
- int fd,
- virCommand *compressor,
- virDomainAsyncJob asyncJob)
+static int
+qemuMigrationSrcToLegacyFile(virQEMUDriver *driver,
+ virDomainObj *vm,
+ int fd,
+ virCommand *compressor,
+ virDomainAsyncJob asyncJob)
{
qemuDomainObjPrivate *priv = vm->privateData;
- int rc;
int ret = -1;
int pipeFD[2] = { -1, -1 };
- unsigned long saveMigBandwidth = priv->migMaxBandwidth;
char *errbuf = NULL;
- virErrorPtr orig_err = NULL;
- g_autoptr(qemuMigrationParams) migParams = NULL;
-
- if (qemuMigrationSetDBusVMState(driver, vm) < 0)
- return -1;
-
- /* Increase migration bandwidth to unlimited since target is a file.
- * Failure to change migration speed is not fatal. */
- if (!(migParams = qemuMigrationParamsForSave(false)))
- return -1;
-
- if (qemuMigrationParamsSetULL(migParams,
- QEMU_MIGRATION_PARAM_MAX_BANDWIDTH,
- QEMU_DOMAIN_MIG_BANDWIDTH_MAX * 1024 * 1024) < 0)
- return -1;
-
- if (qemuMigrationParamsApply(vm, asyncJob, migParams, 0) < 0)
- return -1;
-
- priv->migMaxBandwidth = QEMU_DOMAIN_MIG_BANDWIDTH_MAX;
-
- if (!virDomainObjIsActive(vm)) {
- virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
- _("guest unexpectedly quit"));
- /* nothing to tear down */
- return -1;
- }
if (compressor && virPipe(pipeFD) < 0)
return -1;
@@ -7174,7 +7145,7 @@ qemuMigrationSrcToFile(virQEMUDriver *driver, virDomainObj *vm,
goto cleanup;
if (!compressor) {
- rc = qemuMonitorMigrateToFd(priv->mon, 0, fd);
+ ret = qemuMonitorMigrateToFd(priv->mon, 0, fd);
} else {
virCommandSetInputFD(compressor, pipeFD[0]);
virCommandSetOutputFD(compressor, &fd);
@@ -7190,12 +7161,98 @@ qemuMigrationSrcToFile(virQEMUDriver *driver, virDomainObj *vm,
qemuDomainObjExitMonitor(vm);
goto cleanup;
}
- rc = qemuMonitorMigrateToFd(priv->mon, 0, pipeFD[1]);
+ ret = qemuMonitorMigrateToFd(priv->mon, 0, pipeFD[1]);
if (VIR_CLOSE(pipeFD[0]) < 0 ||
VIR_CLOSE(pipeFD[1]) < 0)
VIR_WARN("failed to close intermediate pipe");
}
qemuDomainObjExitMonitor(vm);
+
+ cleanup:
+ VIR_FORCE_CLOSE(pipeFD[0]);
+ VIR_FORCE_CLOSE(pipeFD[1]);
+
+ if (errbuf) {
+ VIR_DEBUG("Compression binary stderr: %s", NULLSTR(errbuf));
+ VIR_FREE(errbuf);
+ }
+
+ return ret;
+}
+
+
+static int
+qemuMigrationSrcToSparseFile(virQEMUDriver *driver,
+ virDomainObj *vm,
+ int *fd,
+ unsigned int flags,
+ virDomainAsyncJob asyncJob)
+{
+ int ret;
+
+ /* mapped-ram does not support directIO */
+ if ((flags & VIR_DOMAIN_SAVE_BYPASS_CACHE)) {
+ virReportError(VIR_ERR_OPERATION_FAILED, "%s",
+ _("bypass cache unsupported by this system"));
+ return -1;
+ }
+
+ if (qemuSecuritySetImageFDLabel(driver->securityManager, vm->def, *fd) < 0)
+ return -1;
+
+ if (qemuDomainObjEnterMonitorAsync(vm, asyncJob) < 0)
+ return -1;
+
+ ret = qemuMonitorMigrateToFdSet(vm, 0, fd);
+ qemuDomainObjExitMonitor(vm);
+ return ret;
+}
+
+
+/* Helper function called while vm is active. */
+int
+qemuMigrationSrcToFile(virQEMUDriver *driver, virDomainObj *vm,
+ int *fd,
+ virCommand *compressor,
+ qemuMigrationParams *migParams,
+ unsigned int flags,
+ virDomainAsyncJob asyncJob)
+{
+ qemuDomainObjPrivate *priv = vm->privateData;
+ int rc;
+ int ret = -1;
+ unsigned long saveMigBandwidth = priv->migMaxBandwidth;
+ virErrorPtr orig_err = NULL;
+
+ if (qemuMigrationSetDBusVMState(driver, vm) < 0)
+ return -1;
+
+ /* Increase migration bandwidth to unlimited since target is a file.
+ * Failure to change migration speed is not fatal. */
+ if (migParams &&
+ qemuMigrationParamsSetULL(migParams,
+ QEMU_MIGRATION_PARAM_MAX_BANDWIDTH,
+ QEMU_DOMAIN_MIG_BANDWIDTH_MAX * 1024 * 1024) < 0)
+ return -1;
+
+ if (qemuMigrationParamsApply(vm, asyncJob, migParams, 0) < 0)
+ return -1;
+
+ priv->migMaxBandwidth = QEMU_DOMAIN_MIG_BANDWIDTH_MAX;
+
+ if (!virDomainObjIsActive(vm)) {
+ virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
+ _("guest unexpectedly quit"));
+ /* nothing to tear down */
+ return -1;
+ }
+
+ if (migParams &&
+ qemuMigrationParamsCapEnabled(migParams, QEMU_MIGRATION_CAP_MAPPED_RAM))
+ rc = qemuMigrationSrcToSparseFile(driver, vm, fd, flags, asyncJob);
+ else
+ rc = qemuMigrationSrcToLegacyFile(driver, vm, *fd, compressor, asyncJob);
+
if (rc < 0)
goto cleanup;
@@ -7221,8 +7278,17 @@ qemuMigrationSrcToFile(virQEMUDriver *driver, virDomainObj *vm,
if (ret < 0 && !orig_err)
virErrorPreserveLast(&orig_err);
- /* Restore max migration bandwidth */
+ /* Remove fdset passed to qemu and restore max migration bandwidth */
if (qemuDomainObjIsActive(vm)) {
+ if (qemuDomainObjEnterMonitorAsync(vm, asyncJob) == 0) {
+ qemuFDPass *fdPass =
+ qemuFDPassNewFromMonitor("libvirt-outgoing-migrate", priv->mon);
+
+ if (fdPass)
+ qemuFDPassTransferMonitorRollback(fdPass, priv->mon);
+ qemuDomainObjExitMonitor(vm);
+ }
+
if (qemuMigrationParamsSetULL(migParams,
QEMU_MIGRATION_PARAM_MAX_BANDWIDTH,
saveMigBandwidth * 1024 * 1024) == 0)
@@ -7231,13 +7297,6 @@ qemuMigrationSrcToFile(virQEMUDriver *driver, virDomainObj *vm,
priv->migMaxBandwidth = saveMigBandwidth;
}
- VIR_FORCE_CLOSE(pipeFD[0]);
- VIR_FORCE_CLOSE(pipeFD[1]);
- if (errbuf) {
- VIR_DEBUG("Compression binary stderr: %s", NULLSTR(errbuf));
- VIR_FREE(errbuf);
- }
-
virErrorRestore(&orig_err);
return ret;
diff --git a/src/qemu/qemu_migration.h b/src/qemu/qemu_migration.h
index efe1b9e88a..9fa007b949 100644
--- a/src/qemu/qemu_migration.h
+++ b/src/qemu/qemu_migration.h
@@ -238,8 +238,10 @@ qemuMigrationSrcIsAllowed(virDomainObj *vm,
int
qemuMigrationSrcToFile(virQEMUDriver *driver,
virDomainObj *vm,
- int fd,
+ int *fd,
virCommand *compressor,
+ qemuMigrationParams *migParams,
+ unsigned int flags,
virDomainAsyncJob asyncJob)
ATTRIBUTE_NONNULL(1) ATTRIBUTE_NONNULL(2) G_GNUC_WARN_UNUSED_RESULT;
diff --git a/src/qemu/qemu_monitor.c b/src/qemu/qemu_monitor.c
index c069d17265..8d244c6f4a 100644
--- a/src/qemu/qemu_monitor.c
+++ b/src/qemu/qemu_monitor.c
@@ -2232,6 +2232,40 @@ qemuMonitorMigrateToFd(qemuMonitor *mon,
}
+int
+qemuMonitorMigrateToFdSet(virDomainObj *vm,
+ unsigned int flags,
+ int *fd)
+{
+ qemuDomainObjPrivate *priv = vm->privateData;
+ qemuMonitor *mon = priv->mon;
+ off_t offset;
+ g_autoptr(qemuFDPass) fdPassMigrate = NULL;
+ g_autofree char *uri = NULL;
+ int ret;
+
+ VIR_DEBUG("fd=%d flags=0x%x", *fd, flags);
+
+ QEMU_CHECK_MONITOR(mon);
+
+ if ((offset = lseek(*fd, 0, SEEK_CUR)) == -1) {
+ virReportSystemError(errno,
+ "%s", _("failed to seek on file descriptor"));
+ return -1;
+ }
+
+ fdPassMigrate = qemuFDPassNew("libvirt-outgoing-migrate", priv);
+ qemuFDPassAddFD(fdPassMigrate, fd, "-fd");
+ qemuFDPassTransferMonitor(fdPassMigrate, mon);
+
+ uri = g_strdup_printf("file:%s,offset=%#lx",
+ qemuFDPassGetPath(fdPassMigrate), offset);
+ ret = qemuMonitorJSONMigrate(mon, flags, uri);
+
+ return ret;
+}
+
+
int
qemuMonitorMigrateToHost(qemuMonitor *mon,
unsigned int flags,
diff --git a/src/qemu/qemu_monitor.h b/src/qemu/qemu_monitor.h
index f7b9263b64..4022bc3c7b 100644
--- a/src/qemu/qemu_monitor.h
+++ b/src/qemu/qemu_monitor.h
@@ -859,6 +859,10 @@ int qemuMonitorMigrateToFd(qemuMonitor *mon,
unsigned int flags,
int fd);
+int qemuMonitorMigrateToFdSet(virDomainObj *vm,
+ unsigned int flags,
+ int *fd);
+
int qemuMonitorMigrateToHost(qemuMonitor *mon,
unsigned int flags,
const char *protocol,
diff --git a/src/qemu/qemu_saveimage.c b/src/qemu/qemu_saveimage.c
index e3f6a0ad0f..cad982f450 100644
--- a/src/qemu/qemu_saveimage.c
+++ b/src/qemu/qemu_saveimage.c
@@ -430,6 +430,7 @@ qemuSaveImageCreateFd(virQEMUDriver *driver,
virDomainObj *vm,
const char *path,
virFileWrapperFd *wrapperFd,
+ bool sparse,
bool *needUnlink,
unsigned int flags)
{
@@ -439,7 +440,7 @@ qemuSaveImageCreateFd(virQEMUDriver *driver,
int directFlag = 0;
unsigned int wrapperFlags = VIR_FILE_WRAPPER_NON_BLOCKING;
- if (flags & VIR_DOMAIN_SAVE_BYPASS_CACHE) {
+ if (!sparse && flags & VIR_DOMAIN_SAVE_BYPASS_CACHE) {
wrapperFlags |= VIR_FILE_WRAPPER_BYPASS_CACHE;
directFlag = virFileDirectFdFlag();
if (directFlag < 0) {
@@ -459,7 +460,7 @@ qemuSaveImageCreateFd(virQEMUDriver *driver,
if (qemuSecuritySetImageFDLabel(driver->securityManager, vm->def, fd) < 0)
return -1;
- if (!(wrapperFd = virFileWrapperFdNew(&fd, path, wrapperFlags)))
+ if (!sparse && !(wrapperFd = virFileWrapperFdNew(&fd, path, wrapperFlags)))
return -1;
ret = fd;
@@ -478,6 +479,7 @@ qemuSaveImageCreate(virQEMUDriver *driver,
const char *path,
virQEMUSaveData *data,
virCommand *compressor,
+ qemuMigrationParams *saveParams,
unsigned int flags,
virDomainAsyncJob asyncJob)
{
@@ -486,9 +488,10 @@ qemuSaveImageCreate(virQEMUDriver *driver,
int ret = -1;
int fd = -1;
virFileWrapperFd *wrapperFd = NULL;
+ bool sparse = data->header.format == QEMU_SAVE_FORMAT_SPARSE;
/* Obtain the file handle. */
- fd = qemuSaveImageCreateFd(driver, vm, path, wrapperFd, &needUnlink, flags);
+ fd = qemuSaveImageCreateFd(driver, vm, path, wrapperFd, sparse, &needUnlink, flags);
if (fd < 0)
goto cleanup;
@@ -497,7 +500,7 @@ qemuSaveImageCreate(virQEMUDriver *driver,
goto cleanup;
/* Perform the migration */
- if (qemuMigrationSrcToFile(driver, vm, fd, compressor, asyncJob) < 0)
+ if (qemuMigrationSrcToFile(driver, vm, &fd, compressor, saveParams, flags, asyncJob) < 0)
goto cleanup;
/* Touch up file header to mark image complete. */
@@ -505,14 +508,18 @@ qemuSaveImageCreate(virQEMUDriver *driver,
/* Reopen the file to touch up the header, since we aren't set
* up to seek backwards on wrapperFd. The reopened fd will
* trigger a single page of file system cache pollution, but
- * that's acceptable. */
- if (VIR_CLOSE(fd) < 0) {
- virReportSystemError(errno, _("unable to close %1$s"), path);
- goto cleanup;
- }
+ * that's acceptable.
+ * If using mapped-ram, the fd was passed to qemu, so no need
+ * to close it. */
+ if (!sparse) {
+ if (VIR_CLOSE(fd) < 0) {
+ virReportSystemError(errno, _("unable to close %1$s"), path);
+ goto cleanup;
+ }
- if (qemuDomainFileWrapperFDClose(vm, wrapperFd) < 0)
- goto cleanup;
+ if (qemuDomainFileWrapperFDClose(vm, wrapperFd) < 0)
+ goto cleanup;
+ }
if ((fd = qemuDomainOpenFile(cfg, vm->def, path, O_WRONLY, NULL)) < 0 ||
virQEMUSaveDataFinish(data, &fd, path) < 0)
@@ -552,7 +559,7 @@ qemuSaveImageGetCompressionProgram(int format,
*compressor = NULL;
- if (format == QEMU_SAVE_FORMAT_RAW)
+ if (format == QEMU_SAVE_FORMAT_RAW || format == QEMU_SAVE_FORMAT_SPARSE)
return 0;
if (!(prog = virFindFileInPath(imageFormat))) {
diff --git a/src/qemu/qemu_saveimage.h b/src/qemu/qemu_saveimage.h
index 0411f3140a..495d773ea5 100644
--- a/src/qemu/qemu_saveimage.h
+++ b/src/qemu/qemu_saveimage.h
@@ -136,6 +136,7 @@ qemuSaveImageCreate(virQEMUDriver *driver,
const char *path,
virQEMUSaveData *data,
virCommand *compressor,
+ qemuMigrationParams *saveParams,
unsigned int flags,
virDomainAsyncJob asyncJob);
diff --git a/src/qemu/qemu_snapshot.c b/src/qemu/qemu_snapshot.c
index 891e67e98b..fa814e051c 100644
--- a/src/qemu/qemu_snapshot.c
+++ b/src/qemu/qemu_snapshot.c
@@ -1661,6 +1661,8 @@ qemuSnapshotCreateActiveExternal(virQEMUDriver *driver,
/* do the memory snapshot if necessary */
if (memory) {
+ g_autoptr(qemuMigrationParams) snap_params = NULL;
+
/* check if migration is possible */
if (!qemuMigrationSrcIsAllowed(vm, false, VIR_ASYNC_JOB_SNAPSHOT, 0))
goto cleanup;
@@ -1691,8 +1693,11 @@ qemuSnapshotCreateActiveExternal(virQEMUDriver *driver,
memory_existing = virFileExists(snapdef->memorysnapshotfile);
+ if (!(snap_params = qemuMigrationParamsNew()))
+ goto cleanup;
+
if ((ret = qemuSaveImageCreate(driver, vm, snapdef->memorysnapshotfile,
- data, compressor, 0,
+ data, compressor, snap_params, 0,
VIR_ASYNC_JOB_SNAPSHOT)) < 0)
goto cleanup;
--
2.43.0
