September 2016 - Devel - libvirt List Archives

[libvirt] [PATCH] Make --postcopy flag mandatory with --postcopy-after-precopy
by Kothapally Madhu Pavan 02 Sep '16

02 Sep '16

--postcopy-after-precopy is just an aditional flag for postcopy migration. Signed-off-by: Kothapally Madhu Pavan <kmp(a)linux.vnet.ibm.com> --- tools/virsh-domain.c | 6 +++--- tools/virsh.pod | 4 ++-- 2 files changed, 5 insertions(+), 5 deletions(-) diff --git a/tools/virsh-domain.c b/tools/virsh-domain.c index c9ad0d7..a614512 100644 --- a/tools/virsh-domain.c +++ b/tools/virsh-domain.c @@ -10318,10 +10318,10 @@ cmdMigrate(vshControl *ctl, const vshCmd *cmd) } if (vshCommandOptBool(cmd, "postcopy-after-precopy")) { - if (!live_flag) { + if (!vshCommandOptBool(cmd, "postcopy")) { virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s", - _("post-copy migration is not supported with " - "non-live or paused migration")); + _("--postcopy-after-precopy can only be used with " + "--postcopy")); goto cleanup; } iterEvent = virConnectDomainEventRegisterAny( diff --git a/tools/virsh.pod b/tools/virsh.pod index 1e36ee1..3670ace 100644 --- a/tools/virsh.pod +++ b/tools/virsh.pod @@ -1617,8 +1617,8 @@ migration. I<--postcopy> enables post-copy logic in migration, but does not actually start post-copy, i.e., migration is started in pre-copy mode. Once migration is running, the user may switch to post-copy using the B<migrate-postcopy> command sent from another virsh instance or use -I<--postcopy-after-precopy> to let libvirt automatically switch to -post-copy after the first pass of pre-copy is finished. +I<--postcopy-after-precopy> along with I<--postcopy> to let libvirt automatically +switch to post-copy after the first pass of pre-copy is finished. I<--auto-converge> forces convergence during live migration. The initial guest CPU throttling rate can be set with I<auto-converge-initial>. If the

2 2

[libvirt] Entering freeze for libvirt-2.2.0
by Daniel Veillard 02 Sep '16

02 Sep '16

Nearly that the time scheduled, I just tagged 2.2.0 release candidate 1 in git and masigned talbal and rpms at the usual place: ftp://libvirt.org/libvirt/ Seems to be working fine with my limited testing, it's also the first time I see *everything* green in ci.centos.org/view/libvirt-project/ ! :-) Nonetheless this should get some more serious testing, for example on other architectures, and assuming everything goes fine, rc2 will go out on Tuesday and 2.2.0 final should be available on Sep 1st, so thanks in advance for your testing ! Daniel -- Daniel Veillard | Open Source and Standards, Red Hat veillard(a)redhat.com | libxml Gnome XML XSLT toolkit http://xmlsoft.org/ http://veillard.com/ | virtualization library http://libvirt.org/

3 3

[libvirt] [libvirt-php][PATCH] m4: Check for php modules more wisely
by Michal Privoznik 02 Sep '16

02 Sep '16

I've got two version of PHP installed on my system, however one of them has imagick the other one doesn't. During configure I've noticed that wrong assumption has been made. Configure script wrongly assumed the plugin missing. This is because for detecting php plugins we use plain 'php -m | grep $module'. On my system, php is a symlink and it defaults to the version without the plugin. However, I am explicitly supplying 'php-config' from the version that has the plugin installed: ./configure --with-php-config=/usr/lib64/php5.6/bin/php-config The trick is to query provided php-config for path to php binary. Signed-off-by: Michal Privoznik <mprivozn(a)redhat.com> --- Pushed under php-auto-push rule. m4/virt-php-extension.m4 | 13 ++++++++++++- 1 file changed, 12 insertions(+), 1 deletion(-) diff --git a/m4/virt-php-extension.m4 b/m4/virt-php-extension.m4 index 9040eb1..5aa3cb7 100644 --- a/m4/virt-php-extension.m4 +++ b/m4/virt-php-extension.m4 @@ -25,7 +25,18 @@ dnl AC_DEFUN([LIBVIRT_CHECK_PHP_EXTENSION],[ AC_MSG_CHECKING([for php module $1]) - module="$(php -m | grep $1)" + phpbinary="$($PHPCONFIG --php-binary)" + if test "x$phpbinary" = "x"; then + phpbinary="$($PHPCONFIG --prefix)/bin/php" + fi + + if test ! -x "$phpbinary"; then + AC_MSG_ERROR([php binary not found]) + fi + + AC_SUBST([phpbinary]) + + module="$($phpbinary -m | grep $1)" if test "x$module" = "x"; then AC_MSG_ERROR([php module $1 not found]) -- 2.8.4

2 2

[libvirt] [libvirt-glib 1/2] config: Fix gvir_config_xml_node_to_string() leak
by Christophe Fergeau 02 Sep '16

02 Sep '16

If xmlNodeDump() fails, we would be leaking the xmlBuffer we created. This commit ensures we don't return early before this buffer is freed. --- libvirt-gconfig/libvirt-gconfig-helpers.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/libvirt-gconfig/libvirt-gconfig-helpers.c b/libvirt-gconfig/libvirt-gconfig-helpers.c index 0314a72..e8f9664 100644 --- a/libvirt-gconfig/libvirt-gconfig-helpers.c +++ b/libvirt-gconfig/libvirt-gconfig-helpers.c @@ -293,7 +293,7 @@ gvir_config_xml_node_to_string(xmlNodePtr node) xmlbuf = xmlBufferCreate(); if (xmlNodeDump(xmlbuf, node->doc, node, 0, 1) < 0) - return NULL; + xml = NULL; else xml = g_strndup((gchar *)xmlBufferContent(xmlbuf), xmlBufferLength(xmlbuf)); -- 2.7.4

1 1

[libvirt] [PATCH] qemu: fix state change lock held by remoteDispatchDomainBlockJobAbort forever
by Xiubo Li 02 Sep '16

02 Sep '16

When doing the snapshot using the script below: ========================================= !#/bin/bash virsh blockjob $instance_name vdX --abort virsh undefine $instance_name qemu-img create -f qcow2 -o backing_file=$diskfile,size=$size $path/$uuid.dlta virsh blockcopy --domain $instance_name vdX $path/$uuid.dlta --wait --reuse-external --verbose virsh blockjob $instance_name vdX --abort qemu-img convert -f qcow2 -O qcow2 $path/$uuid.dlta $path/$uuid.qcow2 echo "start uploading at $(date)" ...... ========================================= Sometimes you can encounter the following warning and error logs: +++++++++++++++++++++++++++++++++++++++++ 2016-08-26 07:51:05.685+0000: 8916: warning : qemuDomainObjBeginJobInternal:1571 : Cannot start job (query, none) for domain instance-00001b3a; current job is (modify, none) owned by (8914 remoteDispatchDomainBlockJobAbort, 0 <null>) for (30s, 0s) 2016-08-26 07:51:05.685+0000: 8916: error : qemuDomainObjBeginJobInternal:1583 : Timed out during operation: cannot acquire state change lock (held by remoteDispatchDomainBlockJobAbort) ----------------------------------------- Mostly it will be okay later. But sometimes the state change lock maybe hold by remoteDispatchDomainBlockJobAbort forever, then the instance couldn't be connected through the VNC or the network(ping,ssh...), expect after rebooting the instance. >From the code, we find that after sending the --abort command, there will be two steps by condition waiting the two reply signals: +++++++++++++++++++++++++++++++++++++++++ The first condition wait is: In qemuMonitorBlockJobCancel() --> qemuMonitoJSONBlockJobCancel() --> qemuMonitorJSONCommand() --> qemuMonitorJSONCommandWithFd() --> qemuMonitorSend() --> virCondWait(&mon->notify, &mon->parent.lock). The second condition wait is: In virDomainCondWait(vm) --> virCondWait(&vm->cond, &vm->parent.lock). ----------------------------------------- The two corresponding replies are: +++++++++++++++++++++++++++++++++++++++++ The "return" reply is: QEMU_MONITOR_RECV_REPLY: mon=0x7ff3fc001020 reply={"return": [], "id": "libvirt-338"} With the condition wakeup signal: virCondBroadcast(&mon->notify) The "event" reply is: QEMU_MONITOR_RECV_EVENT: mon=0x7fe08401a3f0 event={"timestamp": {"seconds": 1472524518, "microseconds": 360135}, "event": "BLOCK_JOB_CANCELLED", "data": {"device": "drive-virtio-disk0", "len": 42949672960, "offset": 10485760, "speed": 0, "type": "mirror"}} With the condition wakeup signal: virDomainObjBroadcast(vm) --> virCondBroadcast(&vm->cond) ----------------------------------------- But sometimes the qemu daemon will reply like: +++++++++++++++++++++++++++++++++++++++++ The "event" reply is: QEMU_MONITOR_RECV_EVENT: mon=0x7fe08401a3f0 event={"timestamp": {"seconds": 1472524518, "microseconds": 360135}, "event": "BLOCK_JOB_CANCELLED", "data": {"device": "drive-virtio-disk0", "len": 42949672960, "offset": 10485760, "speed": 0, "type": "mirror"}} With the condition wakeup signal: virDomainObjBroadcast(vm) --> virCondBroadcast(&vm->cond) The "return" reply is: QEMU_MONITOR_RECV_REPLY: mon=0x7ff3fc001020 reply={"return": [], "id": "libvirt-338"} With the condition wakeup signal: virCondBroadcast(&mon->notify) ----------------------------------------- In this case, when in the first condition wait, the received "event" reply & signal will be lost and still waiting for the "return" reply & signal. So the when in the second condition wait step, it will wait forever by holding the lock. This patch fix this bug. Signed-off-by: Xiubo Li <lixiubo(a)cmss.chinamobile.com> Signed-off-by: Zhuoyu Zhang <zhangzhuoyu(a)cmss.chinamobile.com> Signed-off-by: Wei Tang <tangwei(a)cmss.chinamobile.com> Signed-off-by: Yaowei Bai <baiyaowei(a)cmss.chinamobile.com> Signed-off-by: Qide Chen <chenqide(a)cmss.chinamobile.com> --- src/qemu/qemu_domain.h | 2 ++ src/qemu/qemu_driver.c | 8 +++++++- 2 files changed, 9 insertions(+), 1 deletion(-) diff --git a/src/qemu/qemu_domain.h b/src/qemu/qemu_domain.h index c49f31c..2624c6d 100644 --- a/src/qemu/qemu_domain.h +++ b/src/qemu/qemu_domain.h @@ -61,6 +61,8 @@ (JOB_MASK(QEMU_JOB_DESTROY) | \ JOB_MASK(QEMU_JOB_ASYNC)) +# define BLOCK_JOB_ABORT_TIMEOUT 5000 + /* Only 1 job is allowed at any time * A job includes *all* monitor commands, even those just querying * information, not merely actions */ diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c index f9a3b15..6ebaf6b 100644 --- a/src/qemu/qemu_driver.c +++ b/src/qemu/qemu_driver.c @@ -15929,6 +15929,7 @@ qemuDomainBlockJobAbort(virDomainPtr dom, bool pivot = !!(flags & VIR_DOMAIN_BLOCK_JOB_ABORT_PIVOT); bool async = !!(flags & VIR_DOMAIN_BLOCK_JOB_ABORT_ASYNC); virDomainObjPtr vm; + unsigned long long now; int ret = -1; virCheckFlags(VIR_DOMAIN_BLOCK_JOB_ABORT_ASYNC | @@ -16018,7 +16019,12 @@ qemuDomainBlockJobAbort(virDomainPtr dom, qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk); qemuBlockJobUpdate(driver, vm, disk); while (diskPriv->blockjob) { - if (virDomainObjWait(vm) < 0) { + if (virTimeMillisNow(&now) < 0) { + ret = -1; + goto endjob; + } + + if (virDomainObjWaitUntil(vm, now + BLOCK_JOB_ABORT_TIMEOUT) < 0) { ret = -1; goto endjob; } -- 1.8.3.1

2 2

[libvirt] [PATCH v5 0/8] perf: add more perf events support
by John Ferlan 02 Sep '16

02 Sep '16

v4: http://www.redhat.com/archives/libvir-list/2016-July/msg00607.html Reworked/reworded the series slightly. The end result is mostly the same as the original, but with the suggested tweaks. John Ferlan (3): virsh: Add a forward reference to perf command from domstats --perf virsh: Rework the perf event names into a table. util: Move virPerfNew and virPerfFree Qiaowei Ren (5): perf: rename qemuDomainGetStatsPerfRdt() perf: Remove the switch from qemuDomainGetStatsPerf util: Add some comment details for virPerfEventType perf: Adjust the perf initialization perf: add more perf events support docs/formatdomain.html.in | 24 +++ docs/schemas/domaincommon.rng | 4 + include/libvirt/libvirt-domain.h | 39 +++++ src/libvirt-domain.c | 9 ++ src/qemu/qemu_driver.c | 23 ++- src/util/virperf.c | 230 +++++++++++++++++----------- src/util/virperf.h | 14 +- tests/genericxml2xmlindata/generic-perf.xml | 4 + tools/virsh.pod | 40 +++-- 9 files changed, 275 insertions(+), 112 deletions(-) -- 2.7.4

2 11

[libvirt] DUID-LLT (DHCPv6 with time-based UUID generation) - how does it work in libvirt?
by Yaniv Kaul 02 Sep '16

02 Sep '16

Reading the examples, I'm not sure how to get DUID-LLT working (which is unfortunate, as it seems to be the default of dhclient). I got DUID-LL working, but DUID-LLT has a timestamp - which is generated, AFAIK, by the client - how can the XML contain the correct DUID-LLT entry? The examples in the docs: <host id="0:1:0:1:18:aa:62:fe:0:16:3e:44:55:66" ip="2001:db8:ca2:2:3::2" /> Somehow assumes the time is 18:aa - no idea how it 'guessed' this is what the client would use. DUID-LL was easy and worked, as it just a prefix + the MAC address. TIA, Y.

1 0

[libvirt] [PATCH v2] virsh: use virConnectGetDomainCapabilities with maxvcpus
by Shivaprasad G Bhat 01 Sep '16

01 Sep '16

virsh maxvcpus --type kvm output is useless on PPC. Also, in commit e6806d79 we documented not rely on virConnectGetMaxVcpus output. Fix the maxvcpus to use virConnectGetDomainCapabilities now to make it useful. The call is made to use the default emulator binary and to check for the host machine and arch which is what the command intends to show anyway. Signed-off-by: Shivaprasad G Bhat <sbhat(a)linux.vnet.ibm.com> --- tools/virsh-host.c | 26 ++++++++++++++++++++++++-- 1 file changed, 24 insertions(+), 2 deletions(-) diff --git a/tools/virsh-host.c b/tools/virsh-host.c index 57f0c0e..dd6ff4e 100644 --- a/tools/virsh-host.c +++ b/tools/virsh-host.c @@ -606,15 +606,37 @@ static bool cmdMaxvcpus(vshControl *ctl, const vshCmd *cmd) { const char *type = NULL; - int vcpus; + int vcpus = -1; + char *caps = NULL; + const unsigned int flags = 0; /* No flags so far */ + xmlDocPtr xml = NULL; + xmlXPathContextPtr ctxt = NULL; virshControlPtr priv = ctl->privData; if (vshCommandOptStringReq(ctl, cmd, "type", &type) < 0) return false; + caps = virConnectGetDomainCapabilities(priv->conn, NULL, NULL, NULL, type, flags); + if (!caps) + goto fallback; + + xml = virXMLParseStringCtxt(caps, _("(domainCapabilities)"), &ctxt); + if (!xml) { + VIR_FREE(caps); + goto fallback; + } + + virXPathInt("string(./vcpu[1]/@max)", ctxt, &vcpus); + + xmlXPathFreeContext(ctxt); + xmlFreeDoc(xml); + if (vcpus > 0) + goto exit; + + fallback: if ((vcpus = virConnectGetMaxVcpus(priv->conn, type)) < 0) return false; - + exit: vshPrint(ctl, "%d\n", vcpus); return true;

2 1

[libvirt] [PATCH v2 0/3] Introduce support for rx_queue_size
by Michal Privoznik 01 Sep '16

01 Sep '16

v2 of: https://www.redhat.com/archives/libvir-list/2016-August/msg00960.html diff to v1: - rename attribute to rx_queue_size - added more documentation and test cases - Other small nits John found Michal Privoznik (3): conf: Add support for virtio-net.rx_queue_size qemu_capabilities: Introduce virtio-net-*.rx_queue_size qemu: Implement virtio-net rx_queue_size docs/formatdomain.html.in | 16 ++++++++- docs/schemas/domaincommon.rng | 5 +++ src/conf/domain_conf.c | 16 +++++++++ src/conf/domain_conf.h | 1 + src/qemu/qemu_capabilities.c | 3 ++ src/qemu/qemu_capabilities.h | 3 ++ src/qemu/qemu_command.c | 8 +++++ src/qemu/qemu_domain.c | 7 ++++ ...ml2argv-net-virtio-rxqueuesize-invalid-size.xml | 29 +++++++++++++++ .../qemuxml2argv-net-virtio-rxqueuesize.args | 25 +++++++++++++ .../qemuxml2argv-net-virtio-rxqueuesize.xml | 29 +++++++++++++++ tests/qemuxml2argvtest.c | 3 ++ .../qemuxml2xmlout-net-virtio-rxqueuesize.xml | 41 ++++++++++++++++++++++ tests/qemuxml2xmltest.c | 1 + 14 files changed, 186 insertions(+), 1 deletion(-) create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-net-virtio-rxqueuesize-invalid-size.xml create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-net-virtio-rxqueuesize.args create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-net-virtio-rxqueuesize.xml create mode 100644 tests/qemuxml2xmloutdata/qemuxml2xmlout-net-virtio-rxqueuesize.xml -- 2.8.4

2 5

[libvirt] [PATCH v2] libxl: add memory attach support
by Bob Liu 01 Sep '16

01 Sep '16

Support for VIR_DOMAIN_DEVICE_MEMORY on domainAttachDeviceFlags API in libxl driver, using libxl_set_memory_target in xen libxl. With "virsh attach-device" command we can then hotplug memory to a domain: <memory model='dimm'> <target> <size unit='MiB'>128</size> <node>0</node> </target> </memory> $ virsh attach-device domain_name this.xml --live Signed-off-by: Bob Liu <bob.liu(a)oracle.com> --- v2: * Unlock virDomainObj while attaching. * Fix memory leak of mem. --- src/libxl/libxl_domain.c | 1 + src/libxl/libxl_driver.c | 35 +++++++++++++++++++++++++++++++++++ 2 files changed, 36 insertions(+) diff --git a/src/libxl/libxl_domain.c b/src/libxl/libxl_domain.c index f529a2e..3924ba0 100644 --- a/src/libxl/libxl_domain.c +++ b/src/libxl/libxl_domain.c @@ -414,6 +414,7 @@ virDomainDefParserConfig libxlDomainDefParserConfig = { .macPrefix = { 0x00, 0x16, 0x3e }, .devicesPostParseCallback = libxlDomainDeviceDefPostParse, .domainPostParseCallback = libxlDomainDefPostParse, + .features = VIR_DOMAIN_DEF_FEATURE_MEMORY_HOTPLUG }; diff --git a/src/libxl/libxl_driver.c b/src/libxl/libxl_driver.c index a34eb02..b23c1d4 100644 --- a/src/libxl/libxl_driver.c +++ b/src/libxl/libxl_driver.c @@ -3085,6 +3085,34 @@ libxlDomainAttachHostUSBDevice(libxlDriverPrivatePtr driver, #endif static int +libxlDomainAttachMemory(libxlDriverPrivatePtr driver, + virDomainObjPtr vm, + virDomainMemoryDefPtr mem) +{ + int res = -1; + libxlDriverConfigPtr cfg = libxlDriverConfigGet(driver); + + if (mem->targetNode != 0) { + virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s", + _("Non-zero target node is not supported.")); + goto out; + } + + /* Unlock virDomainObj while attaching memory */ + virObjectUnlock(vm); + res = libxl_set_memory_target(cfg->ctx, vm->def->id, mem->size, 1, 1); + virObjectLock(vm); + if (res < 0) + virReportError(VIR_ERR_INTERNAL_ERROR, + _("Failed to attach %lluKB memory for domain %d"), + mem->size, vm->def->id); + + out: + virDomainMemoryDefFree(mem); + return res; +} + +static int libxlDomainAttachHostDevice(libxlDriverPrivatePtr driver, virDomainObjPtr vm, virDomainHostdevDefPtr hostdev) @@ -3284,6 +3312,13 @@ libxlDomainAttachDeviceLive(libxlDriverPrivatePtr driver, dev->data.hostdev = NULL; break; + case VIR_DOMAIN_DEVICE_MEMORY: + /* Note that libxlDomainAttachMemory always consumes + * dev->data.memory. */ + ret = libxlDomainAttachMemory(driver, vm, dev->data.memory); + dev->data.memory = NULL; + break; + default: virReportError(VIR_ERR_CONFIG_UNSUPPORTED, _("device type '%s' cannot be attached"), -- 2.6.5

2 1