- Devel - libvirt List Archives

[libvirt] [PATCH] virsh: fix memory leak in cmdVolPath code
by ajia＠redhat.com 28 Jul '11

28 Jul '11

* tools/virsh.c: avoid memory leak in cmdVolPath. * how to reproduce? % dd if=/dev/zero of=/var/lib/libvirt/images/foo.img count=1 bs=10M % virsh pool-refresh default % valgrind -v --leak-check=full virsh vol-path --vol /var/lib/libvirt/images/foo.img * actual results: Detected in valgrind run: ==16436== 32 bytes in 1 blocks are definitely lost in loss record 7 of 22 ==16436== at 0x4A05FDE: malloc (vg_replace_malloc.c:236) ==16436== by 0x386A314B3D: xdr_string (in /lib64/libc-2.12.so) ==16436== by 0x3DF8CD770D: xdr_remote_nonnull_string (remote_protocol.c:30) ==16436== by 0x3DF8CD7EC8: xdr_remote_storage_vol_get_path_ret (remote_protocol.c:2952) ==16436== by 0x3DF8CDF161: virNetMessageDecodePayload (virnetmessage.c:286) ==16436== by 0x3DF8CDE9E5: virNetClientProgramCall (virnetclientprogram.c:318) ==16436== by 0x3DF8CC28A2: call (remote_driver.c:3929) ==16436== by 0x3DF8CC8412: remoteStorageVolGetPath (remote_client_bodies.h:5219) ==16436== by 0x3DF8C9BF14: virStorageVolGetPath (libvirt.c:11389) ==16436== by 0x418ECA: cmdVolPath (virsh.c:8754) ==16436== by 0x410CC2: vshCommandRun (virsh.c:12758) ==16436== by 0x41F286: main (virsh.c:14110) ==16436== ==16436== LEAK SUMMARY: ==16436== definitely lost: 32 bytes in 1 blocks Signed-off-by: Alex Jia <ajia(a)redhat.com> --- tools/virsh.c | 9 ++++++++- 1 files changed, 8 insertions(+), 1 deletions(-) diff --git a/tools/virsh.c b/tools/virsh.c index 841df61..d194a8b 100644 --- a/tools/virsh.c +++ b/tools/virsh.c @@ -9326,6 +9326,7 @@ cmdVolPath(vshControl *ctl, const vshCmd *cmd) { virStorageVolPtr vol; const char *name = NULL; + char * StorageVolPath; if (!vshConnectionUsability(ctl, ctl->conn)) return false; @@ -9334,7 +9335,13 @@ cmdVolPath(vshControl *ctl, const vshCmd *cmd) return false; } - vshPrint(ctl, "%s\n", virStorageVolGetPath(vol)); + if ((StorageVolPath = virStorageVolGetPath(vol)) == NULL) { + virStorageVolFree(vol); + return false; + } + + vshPrint(ctl, "%s\n", StorageVolPath); + VIR_FREE(StorageVolPath); virStorageVolFree(vol); return true; } -- 1.7.1

3 3

[libvirt] Memory leak of the remoteDomainSet* functions
by Osier Yang 28 Jul '11

28 Jul '11

Hello, there Per bug https://bugzilla.redhat.com/show_bug.cgi?id=725322, should we change the remoteDomainSet* functions into "skipgen", and fix the leaks like below? (NB, new VIR_FREE statements) static int remoteDomainSetBlkioParameters(virDomainPtr dom, virTypedParameterPtr params, int nparams, unsigned int flags) { int rv = -1; int i; struct private_data *priv = dom->conn->privateData; remote_domain_set_blkio_parameters_args args; remote_domain_set_blkio_parameters_args args; remoteDriverLock(priv); make_nonnull_domain(&args.dom, dom); args.flags = flags; if (remoteSerializeTypedParameters(params, nparams, &args.params.params_val, &args.params.params_len) < 0) { for (i = 0; i < nparams; i++) { VIR_FREE(args.params.params_val[i].field); } VIR_FREE(args.params.params_val); xdr_free((xdrproc_t)xdr_remote_domain_set_blkio_parameters_args, (char *)&args); goto done; } if (call(dom->conn, priv, 0, REMOTE_PROC_DOMAIN_SET_BLKIO_PARAMETERS, (xdrproc_t)xdr_remote_domain_set_blkio_parameters_args, (char *)&args, (xdrproc_t)xdr_void, (char *)NULL) == -1) { goto done; } rv = 0; done: if (args.params.params_val) { for (i = 0; i < nparams; i++) { VIR_FREE(args.params.params_val[i].field); } VIR_FREE(args.params.params_val); } remoteDriverUnlock(priv); return rv; } Regards, Osier

2 3

[libvirt] [PATCH] qemu: Remove bogus codes in function getCompressionType
by Osier Yang 28 Jul '11

28 Jul '11

The error will never be reported, remove the codes, and also improve the docs in qemu.conf to tell user the truth. --- src/qemu/qemu.conf | 4 ++++ src/qemu/qemu_driver.c | 16 +++++----------- 2 files changed, 9 insertions(+), 11 deletions(-) diff --git a/src/qemu/qemu.conf b/src/qemu/qemu.conf index 145062c..75d945b 100644 --- a/src/qemu/qemu.conf +++ b/src/qemu/qemu.conf @@ -199,6 +199,10 @@ # save_image_format is used when you use 'virsh save' at scheduled saving. # dump_image_format is used when you use 'virsh dump' at emergency crashdump. # +# If the specified format is not valid (the valid formats are "raw", "lzop", +# "gzip", "bzip2", and "xz"), or the compress program is not available, "raw" +# will be used as the format silently without error or warning. +# # save_image_format = "raw" # dump_image_format = "raw" diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c index b928004..48fc46a 100644 --- a/src/qemu/qemu_driver.c +++ b/src/qemu/qemu_driver.c @@ -2712,19 +2712,13 @@ getCompressionType(struct qemud_driver *driver) */ if (driver->dumpImageFormat) { compress = qemudSaveCompressionTypeFromString(driver->dumpImageFormat); - if (compress < 0) { - qemuReportError(VIR_ERR_OPERATION_FAILED, "%s", - _("Invalid dump image format specified in " - "configuration file, using raw")); + /* Use "raw" as the format if the specified format is not valid, + * or the compress program is not available, + */ + if (compress < 0) return QEMUD_SAVE_FORMAT_RAW; - } - if (!qemudCompressProgramAvailable(compress)) { - qemuReportError(VIR_ERR_OPERATION_FAILED, - "%s", _("Compression program for dump image format " - "in configuration file isn't available, " - "using raw")); + if (!qemudCompressProgramAvailable(compress)) return QEMUD_SAVE_FORMAT_RAW; - } } return compress; } -- 1.7.6

2 4

[libvirt] [PATCH] qemu: improve thread documentation
by Eric Blake 28 Jul '11

28 Jul '11

* src/qemu/THREADS.txt: Fix problems with typos, grammar, and outdated examples. --- Doc only, so pushing under the trivial rule. src/qemu/THREADS.txt | 30 ++++++++++++++++-------------- 1 files changed, 16 insertions(+), 14 deletions(-) diff --git a/src/qemu/THREADS.txt b/src/qemu/THREADS.txt index 3a27a85..e73076c 100644 --- a/src/qemu/THREADS.txt +++ b/src/qemu/THREADS.txt @@ -6,7 +6,7 @@ the QEMU driver. The criteria for this model are: - Objects must never be exclusively locked for any prolonged time - Code which sleeps must be able to time out after suitable period - - Must be safe against dispatch asynchronous events from monitor + - Must be safe against dispatch of asynchronous events from monitor Basic locking primitives @@ -19,7 +19,7 @@ There are a number of locks on various objects This is the top level lock on the entire driver. Every API call in the QEMU driver is blocked while this is held, though some internal callbacks may still run asynchronously. This lock must never be held - for anything which sleeps/waits (ie monitor commands) + for anything which sleeps/waits (i.e. monitor commands) When obtaining the driver lock, under *NO* circumstances must any lock be held on a virDomainObjPtr. This *WILL* result in @@ -44,7 +44,7 @@ There are a number of locks on various objects to have the driver locked when re-acquiring the dropped locked, since the reference count prevents it being freed by another thread. - This lock must not be held for anything which sleeps/waits (ie monitor + This lock must not be held for anything which sleeps/waits (i.e. monitor commands). @@ -80,7 +80,7 @@ There are a number of locks on various objects whenever it hits a piece of code which may sleep/wait, and re-acquire it after the sleep/wait. Whenever an asynchronous job wants to talk to the monitor, it needs to acquire nested job (a - special kind of normla job) to obtain exclusive access to the + special kind of normal job) to obtain exclusive access to the monitor. Since the virDomainObjPtr lock was dropped while waiting for the @@ -139,7 +139,7 @@ To acquire the normal job condition - Increments ref count on virDomainObjPtr - Waits until the job is compatible with current async job or no async job is running - - Waits job.cond condition 'job.active != 0' using virDomainObjPtr + - Waits for job.cond condition 'job.active != 0' using virDomainObjPtr mutex - Rechecks if the job is still compatible and repeats waiting if it isn't @@ -150,7 +150,7 @@ To acquire the normal job condition - Unlocks driver - Waits until the job is compatible with current async job or no async job is running - - Waits job.cond condition 'job.active != 0' using virDomainObjPtr + - Waits for job.cond condition 'job.active != 0' using virDomainObjPtr mutex - Rechecks if the job is still compatible and repeats waiting if it isn't @@ -160,7 +160,7 @@ To acquire the normal job condition - Locks virDomainObjPtr NB: this variant is required in order to comply with lock ordering - rules for virDomainObjPtr vs driver + rules for virDomainObjPtr vs. driver qemuDomainObjEndJob() @@ -175,7 +175,7 @@ To acquire the asynchronous job condition qemuDomainObjBeginAsyncJob() (if driver is unlocked) - Increments ref count on virDomainObjPtr - Waits until no async job is running - - Waits job.cond condition 'job.active != 0' using virDomainObjPtr + - Waits for job.cond condition 'job.active != 0' using virDomainObjPtr mutex - Rechecks if any async job was started while waiting on job.cond and repeats waiting in that case @@ -185,7 +185,7 @@ To acquire the asynchronous job condition - Increments ref count on virDomainObjPtr - Unlocks driver - Waits until no async job is running - - Waits job.cond condition 'job.active != 0' using virDomainObjPtr + - Waits for job.cond condition 'job.active != 0' using virDomainObjPtr mutex - Rechecks if any async job was started while waiting on job.cond and repeats waiting in that case @@ -271,7 +271,7 @@ Design patterns - * Accessing something directly todo with a virDomainObjPtr + * Accessing something directly to do with a virDomainObjPtr virDomainObjPtr obj; @@ -285,7 +285,7 @@ Design patterns - * Accessing something directly todo with a virDomainObjPtr and driver + * Accessing something directly to do with a virDomainObjPtr and driver virDomainObjPtr obj; @@ -299,11 +299,11 @@ Design patterns - * Updating something directly todo with a virDomainObjPtr + * Updating something directly to do with a virDomainObjPtr virDomainObjPtr obj; - qemuDriverLockRO(driver); + qemuDriverLock(driver); obj = virDomainFindByUUID(driver->domains, dom->uuid); qemuDriverUnlock(driver); @@ -324,7 +324,7 @@ Design patterns virDomainObjPtr obj; qemuDomainObjPrivatePtr priv; - qemuDriverLockRO(driver); + qemuDriverLock(driver); obj = virDomainFindByUUID(driver->domains, dom->uuid); qemuDriverUnlock(driver); @@ -333,6 +333,7 @@ Design patterns ...do prep work... if (virDomainObjIsActive(vm)) { + /* using ignore_value is safe since vm is active */ ignore_value(qemuDomainObjEnterMonitor(obj)); qemuMonitorXXXX(priv->mon); qemuDomainObjExitMonitor(obj); @@ -360,6 +361,7 @@ Design patterns ...do prep work... if (virDomainObjIsActive(vm)) { + /* using ignore_value is safe since vm is active */ ignore_value(qemuDomainObjEnterMonitorWithDriver(driver, obj)); qemuMonitorXXXX(priv->mon); qemuDomainObjExitMonitorWithDriver(driver, obj); -- 1.7.4.4

1 0

[libvirt] [PATCH] virsh: expose change-protection during migration
by Eric Blake 27 Jul '11

27 Jul '11

* tools/virsh.c (doMigrate): Add --change-protection flag. * tools/virsh.pod (migrate): Document it. --- As promised here: https://www.redhat.com/archives/libvir-list/2011-July/msg01919.html tools/virsh.c | 4 ++++ tools/virsh.pod | 9 +++++++-- 2 files changed, 11 insertions(+), 2 deletions(-) diff --git a/tools/virsh.c b/tools/virsh.c index 113124f..34036f9 100644 --- a/tools/virsh.c +++ b/tools/virsh.c @@ -4830,6 +4830,8 @@ static const vshCmdOptDef opts_migrate[] = { {"suspend", VSH_OT_BOOL, 0, N_("do not restart the domain on the destination host")}, {"copy-storage-all", VSH_OT_BOOL, 0, N_("migration with non-shared storage with full disk copy")}, {"copy-storage-inc", VSH_OT_BOOL, 0, N_("migration with non-shared storage with incremental copy (same base image shared between source and destination)")}, + {"change-protection", VSH_OT_BOOL, 0, + N_("prevent any configuration changes to source until migration ends)")}, {"verbose", VSH_OT_BOOL, 0, N_("display the progress of migration")}, {"domain", VSH_OT_DATA, VSH_OFLAG_REQ, N_("domain name, id or uuid")}, {"desturi", VSH_OT_DATA, VSH_OFLAG_REQ, N_("connection URI of the destination host as seen from the client(normal migration) or source(p2p migration)")}, @@ -4906,6 +4908,8 @@ doMigrate (void *opaque) if (vshCommandOptBool (cmd, "copy-storage-inc")) flags |= VIR_MIGRATE_NON_SHARED_INC; + if (vshCommandOptBool (cmd, "change-protection")) + flags |= VIR_MIGRATE_CHANGE_PROTECTION; if (xmlfile && virFileReadAll(xmlfile, 8192, &xml) < 0) diff --git a/tools/virsh.pod b/tools/virsh.pod index e9aaa80..6c3c960 100644 --- a/tools/virsh.pod +++ b/tools/virsh.pod @@ -564,8 +564,9 @@ type attribute for the <domain> element of XML. =item B<migrate> [I<--live>] [I<--direct>] [I<--p2p> [I<--tunnelled>]] [I<--persistent>] [I<--undefinesource>] [I<--suspend>] [I<--copy-storage-all>] -[I<--copy-storage-inc>] [I<--verbose>] I<domain-id> I<desturi> [I<migrateuri>] -[I<dname>] [I<--timeout> B<seconds>] [I<--xml> B<file>] +[I<--copy-storage-inc>] [I<--change-protection>] [I<--verbose>] +I<domain-id> I<desturi> [I<migrateuri>] [I<dname>] +[I<--timeout> B<seconds>] [I<--xml> B<file>] Migrate domain to another host. Add I<--live> for live migration; I<--p2p> for peer-2-peer migration; I<--direct> for direct migration; or I<--tunnelled> @@ -575,6 +576,10 @@ and I<--suspend> leaves the domain paused on the destination host. I<--copy-storage-all> indicates migration with non-shared storage with full disk copy, I<--copy-storage-inc> indicates migration with non-shared storage with incremental copy (same base image shared between source and destination). +I<--change-protection> enforces that no incompatible configuration changes +will be made to the domain while the migration is underway; this flag is +implicitly enabled when supported by the hypervisor, but can be used to +reject the migration if the hypervisor lacks change protection support. I<--verbose> displays the progress of migration. The I<desturi> is the connection URI of the destination host, and -- 1.7.4.4

2 2

Re: [libvirt] [virt-tools-list] virt-manager/libvirt backwards compatibility problem?
by Cole Robinson 27 Jul '11

27 Jul '11

On 07/27/2011 12:28 PM, Whit Blauvelt wrote: > Looks like my real problem may be not incorporating a Debian/Ubuntu patch > before building 0.9.x, since netcat differs: > > https://bugs.launchpad.net/ubuntu/+source/libvirt/+bug/517478 > Hmm, what was the error message you were getting? We've made efforts at the libvirt level to have useful error reporting here, so that the cause of failure is more obvious. Also, I patched virt-manager a while ago to use a shell script to handle both RH and debian 'nc' versions. Script is here: http://git.fedorahosted.org/git?p=virt-manager.git;a=blob;f=src/virtManager… Yeah, it's complete and utter crack, but this has been a constant frustration for users, so maybe worth revisiting at the libvirt level (and hey, launching nc over ssh is kinda crack to being with :) ). Even if we solved this in a clean way by providing our own /usr/libexec/libvirt_nc binary, we'd still probably have to play shell script games over SSH to handle connecting to older libvirt (or what if libvirt_nc was installed somewhere other than libexec?) - Cole

3 2

[libvirt] [PATCH] rpc: Fix memory leak in remoteDomainSet*Parameters functions
by Matthias Bolte 27 Jul '11

27 Jul '11

Add a new helper remoteFreeTypedParameters and teach the generator to add it to the cleanup section. https://bugzilla.redhat.com/show_bug.cgi?id=725322 --- src/remote/remote_driver.c | 22 +++++++++++++++++----- src/rpc/gendispatch.pl | 5 +++++ 2 files changed, 22 insertions(+), 5 deletions(-) diff --git a/src/remote/remote_driver.c b/src/remote/remote_driver.c index 0652e0d..e5bfa4b 100644 --- a/src/remote/remote_driver.c +++ b/src/remote/remote_driver.c @@ -1208,6 +1208,22 @@ done: return rv; } +/* Helper to free typed parameters. */ +static void +remoteFreeTypedParameters(remote_typed_param *args_params_val, + u_int args_params_len) +{ + u_int i; + + if (args_params_val == NULL) + return; + + for (i = 0; i < args_params_len; i++) + VIR_FREE(args_params_val[i].field); + + VIR_FREE(args_params_val); +} + /* Helper to serialize typed parameters. */ static int remoteSerializeTypedParameters(virTypedParameterPtr params, @@ -1264,11 +1280,7 @@ remoteSerializeTypedParameters(virTypedParameterPtr params, rv = 0; cleanup: - if (val) { - for (i = 0; i < nparams; i++) - VIR_FREE(val[i].field); - VIR_FREE(val); - } + remoteFreeTypedParameters(val, nparams); return rv; } diff --git a/src/rpc/gendispatch.pl b/src/rpc/gendispatch.pl index ceeb1f5..0d344e8 100755 --- a/src/rpc/gendispatch.pl +++ b/src/rpc/gendispatch.pl @@ -979,6 +979,7 @@ elsif ($opt_k) { my @args_check_list = (); my @setters_list = (); my @setters_list2 = (); + my @free_list = (); my $priv_src = "conn"; my $priv_name = "privateData"; my $call_args = "&args"; @@ -1105,6 +1106,7 @@ elsif ($opt_k) { " xdr_free((xdrproc_t)xdr_$call->{args}, (char *)&args);\n" . " goto done;\n" . " }"); + push(@free_list, " remoteFreeTypedParameters(args.params.params_val, args.params.params_len);\n"); } elsif ($args_member =~ m/^((?:unsigned )?int) (\S+);\s*\/\*\s*call-by-reference\s*\*\//) { my $type_name = "$1 *"; my $arg_name = $2; @@ -1500,6 +1502,9 @@ elsif ($opt_k) { print "\n"; print "done:\n"; + + print join("\n", @free_list); + print " remoteDriverUnlock(priv);\n"; print " return rv;\n"; print "}\n"; -- 1.7.4.1

2 2

[libvirt] Environmental Variables inside a container
by Devendra K. Modium 27 Jul '11

27 Jul '11

Hi All I am trying to pass an environmental flag to the LXC container when it is started.I found libvirt_lxc is already doing it in lxc_container.c as shown below static virCommandPtr lxcContainerBuildInitCmd(virDomainDefPtr vmDef) { char uuidstr[VIR_UUID_STRING_BUFLEN]; virCommandPtr cmd; virUUIDFormat(vmDef->uuid, uuidstr); cmd = virCommandNew(vmDef->os.init); virCommandAddEnvString(cmd, "PATH=/bin:/sbin"); virCommandAddEnvString(cmd, "TERM=linux-deva"); virCommandAddEnvPair(cmd, "LIBVIRT_LXC_UUID", uuidstr); virCommandAddEnvPair(cmd, "LIBVIRT_LXC_NAME", vmDef->name); return cmd; } But only TERM flag is getting visible inside the container not remaining. Does it depends on how /sbin/init is receiving them? Is /sbin/init only recognizing TERM flag. If so what is intention of sending remaining flags in current libvirt. Please let me know if there is any elegant way to send environmental variable to container. Thanks for your time. Regards Devendra

2 2

[libvirt] Add support for image write mode in virtualbox driver
by Boris FELD 27 Jul '11

27 Jul '11

Hi everyone, We currently using libvirt for PYTI GSOC Project (it works very well!) and we encounter a problem. We use VM with two disk images, one which must to be snapshoted/rollbacked and the other one which must be not. We use VirtualBox as hypervisor and to do so, VirtualBox define "Special image write modes" (http://www.virtualbox.org/manual/ch05.html#hdimagewrites) What we need is to set the type of one image to "writethrough". This information seems defined in VM config file. For instance: <HardDisk uuid="{402aef07-4df8-4518-831c-eeb9ae2a32d2}" location="/home/boris/VirtualBox VMs/PYTI.vdi" format="VDI" type="Normal"/> <HardDisk uuid="{63acf6d9-b7db-4d8d-8389-1276c3d500bb}" location="/home/boris/VirtualBox VMs/adhock.vdi" format="VDI" type="Writethrough"/> I found two pages mentioning the type of medium, this one: http://www.virtualbox.org/sdkref/_virtual_box_8idl.html#b4abb9d1b9e44b997d1… and this one: http://www.virtualbox.org/sdkref/interface_i_medium_attachment.html. Hope it will help. Cheers, FELD Boris

1 0

[libvirt] [PATCH 00/10] Rollback migration when libvirtd restarts
by Jiri Denemark 27 Jul '11

27 Jul '11

This is the rest of the original 19 patch series updated with some bugfixes and rebased on current master, which is also available at https://gitorious.org/~jirka/libvirt/jirka-staging/commits/migration-recove… I didn't manage to run this through the libvirt-tck migration test but I did some additional testing with the following combinations of libvirt (current contains this series): source destination client ------------------------------------------- current current current current current 0.9.2-1.el6 current 0.9.2-2.fc15 current 0.9.2-2.fc15 current 0.9.2-1.el6 In all combinations I did - normal migration from src to dst - normal migration back - p2p migration from src to dst - p2p migration back To test failure recovery, I aborted the migration or libvirtd deamons at various places. All this was done with a single running domain without restarting it. Jiri Denemark (10): qemu: Implement migration job phases qemu: Migration job on destination daemon qemu: Migration job on source daemon qemu: Recover from interrupted migration qemu: Remove special case for virDomainGetBlockInfo qemu: Remove special case for virDomainBlockStats qemu: Remove special case for virDomainMigrateSetMaxSpeed qemu: Remove special case for virDomainMigrateSetMaxDowntime qemu: Remove special case for virDomainSuspend qemu: Remove special case for virDomainAbortJob include/libvirt/libvirt.h.in | 3 + src/libvirt.c | 27 ++- src/libvirt_internal.h | 6 + src/qemu/MIGRATION.txt | 55 ++++ src/qemu/qemu_domain.c | 18 +- src/qemu/qemu_domain.h | 31 +-- src/qemu/qemu_driver.c | 285 ++++++++++++--------- src/qemu/qemu_migration.c | 574 ++++++++++++++++++++++++------------------ src/qemu/qemu_migration.h | 40 +++- src/qemu/qemu_process.c | 121 +++++++++- 10 files changed, 750 insertions(+), 410 deletions(-) create mode 100644 src/qemu/MIGRATION.txt -- 1.7.6

3 24