September 2015 - Devel - libvirt List Archives

[libvirt] [PATCHv3 0/2] Added waiting for DAD to finish for bridge address.
by Maxim Perevedentsev 19 Oct '15

19 Oct '15

This is a fix for commit db488c79173b240459c7754f38c3c6af9b432970 dnsmasq main process which is relied on when waiting for DAD to complete exits without actually waiting for DAD. This is dnsmasq daemon's task. It seems to be a race that DAD finished before dnsmasq main process exited. The above commit needs the execution to block until DAD finishes for bridge IPv6 address because then it closes dummy tap device. Thus we need to ensure this ourselves. So we periodically poll the kernel using netlink and check whether there are any IPv6 addresses assigned to bridge which have 'tentative' state. After DAD is finished, execution continues. I guess that is what dnsmasq was assumed to do. We use netlink to dump information about existing IPv6 addresses. Netlink's response is a multi-part message. Unfortunately, the current implementation of virNetlink treats such messages as faulty and throws an error. So the patch 2/2 adds multi-part nelink response support. Update v2: fixed syntax. Update v3: moved to virnetdev. Resend: These patches were ignored and buried long ago :-( Maxim Perevedentsev (2): network: added waiting for DAD to finish for bridge address. netlink: add support for multi-part netlink messages. src/libvirt_private.syms | 1 + src/network/bridge_driver.c | 35 +++++++++- src/util/virnetdev.c | 160 ++++++++++++++++++++++++++++++++++++++++++++ src/util/virnetdev.h | 2 + src/util/virnetlink.c | 4 +- 5 files changed, 200 insertions(+), 2 deletions(-) -- 1.8.3.1

3 10

[libvirt] [PATCH] docs: event impl. registration before hypervisor connection
by Dominik Perpeet 16 Oct '15

16 Oct '15

Event implementations need to be registered before a connection to the Hypervisor is opened, otherwise event handling can be impaired (e.g. delayed messages). This fact is referenced in an e-mail [1], but should also be noted in the documentation of the registration functions. [1] https://www.redhat.com/archives/libvirt-users/2014-April/msg00011.html --- src/util/virevent.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/src/util/virevent.c b/src/util/virevent.c index 54b6396..e0fd35e 100644 --- a/src/util/virevent.c +++ b/src/util/virevent.c @@ -204,6 +204,9 @@ virEventRemoveTimeout(int timer) * to integrate with the libglib2 event loop, or libevent * or the QT event loop. * + * For proper event handling, it is important that the event implementation + * is registered before a connection to the Hypervisor is opened. + * * Use of the virEventAddHandle() and similar APIs require that the * corresponding handler is registered. Use of the * virConnectDomainEventRegisterAny() and similar APIs requires that @@ -247,6 +250,9 @@ void virEventRegisterImpl(virEventAddHandleFunc addHandle, * not have a need to integrate with an external event * loop impl. * + * For proper event handling, it is important that the event implementation + * is registered before a connection to the Hypervisor is opened. + * * Once registered, the application has to invoke virEventRunDefaultImpl() in * a loop to process events. Failure to do so may result in connections being * closed unexpectedly as a result of keepalive timeout. The default -- 1.8.3.1

3 2

[libvirt] [PATCH v5 0/5] vz: add migration support
by Nikolay Shirokovskiy 15 Oct '15

15 Oct '15

NOTE that minimal command to migrate vz domain is like next: virsh -c vz:///system migrate 200 vz+ssh://shiny0/system --p2p --live --compressed --persistent --undefinesource Difference from v4: 1. move preparation of the migration uri from src to dst as dst intended to do it. 2. change hypervisor migration scheme from 'tcp' to 'vzmigr' 3. declare this scheme in host caps src/vz/vz_driver.c | 356 ++++++++++++++++++++++++++++++++++++++++++++++++++++ src/vz/vz_sdk.c | 86 +++++++++++-- src/vz/vz_sdk.h | 6 + src/vz/vz_utils.h | 4 +- 4 files changed, 437 insertions(+), 15 deletions(-)

4 12

[libvirt] crash in virDomainNumaGetMemorySize
by Olaf Hering 15 Oct '15

15 Oct '15

With current master (56945e1), while toying around with a WS2008R2 Hyper-V host: root@probook:~ # gdb --quiet -ex 'r -c hv dumpxml all-sles12-dev' -ex bt -ex detach -ex quit virsh Reading symbols from virsh...Reading symbols from /usr/lib/debug/usr/bin/virsh.debug...done. done. Starting program: /usr/bin/virsh -c hv dumpxml all-sles12-dev [Thread debugging using libthread_db enabled] Using host libthread_db library "/lib64/libthread_db.so.1". Enter Administrator's password for optiplex.fritz.box: [New Thread 0x7fffee601700 (LWP 26921)] Program received signal SIGSEGV, Segmentation fault. virDomainNumaGetMemorySize (numa=0x0) at conf/numa_conf.c:971 971 for (i = 0; i < numa->nmem_nodes; i++) #0 virDomainNumaGetMemorySize (numa=0x0) at conf/numa_conf.c:971 #1 0x00007ffff793bce0 in virDomainDefGetMemoryActual (def=def@entry=0x55555582ead0) at conf/domain_conf.c:7857 #2 0x00007ffff7948978 in virDomainDefFormatInternal (def=0x55555582ead0, flags=0, buf=buf@entry=0x7fffffffdde0) at conf/domain_conf.c:21677 #3 0x00007ffff794b8bc in virDomainDefFormat (def=<optimized out>, flags=<optimized out>) at conf/domain_conf.c:22507 #4 0x00007ffff7a94c33 in hypervDomainGetXMLDesc (domain=<optimized out>, flags=0) at hyperv/hyperv_driver.c:882 #5 0x00007ffff79b5351 in virDomainGetXMLDesc (domain=domain@entry=0x5555558237a0, flags=0) at libvirt-domain.c:2591 #6 0x000055555558b978 in cmdDumpXML (ctl=0x7fffffffe1c0, cmd=<optimized out>) at virsh-domain.c:9634 #7 0x000055555557d04f in vshCommandRun (ctl=0x7fffffffe1c0, cmd=0x555555822580) at vsh.c:1212 #8 0x0000555555579a4d in main (argc=5, argv=0x7fffffffe3c8) at virsh.c:921 Detaching from program: /usr/bin/virsh, process 26917 root@probook:~ # rpm -qf `which virsh` libvirt-client-20150917T085913.56945e1-3.xen_unstable.1.x86_64 Olaf

2 3

[libvirt] cpu affinity, isolcpus and cgroups
by Henning Schild 14 Oct '15

14 Oct '15

Hi, i am currently looking into realtime VMs using libvirt. My first starting point was reserving a couple of cores using isolcpus and later tuning the affinity to place my vcpus on the reserved pcpus. My first observation was that libvirt ignores isolcpus. Affinity masks of new qemus will default to all cpus and will not be inherited from libvirtd. A comment in the code suggests that this is done on purpose. After that i changed the code to use only the available cpus by default. But taskset was still showing all 'f's on my qemus. Then i traced my change down to sched_setaffinity assuming that some other mechanism might have reverted my hack, but it is still in place. Libvirt is setting up cgroups and now my suspicion is that cgroups and taskset might not work well together. > /sys/fs/cgroup/cpu/machine.slice/machine-qemu\x2dvm1.scope/vcpu0# > cpuacct.usage_percpu > 247340587 50851635 89631114 23383025 412639264 1241965 55442753 19923 > 14093629 15863859 27403280 1292195745 82031088 53690508 135826421 > 124915000 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Looks like the last 16 cores are not used. But if i use taskset to ask for the affinity mask i get all 32 cpus. > taskset -p `cat tasks` > pid 12905's current affinity mask: ffffffff I know that is not strictly libvirt but also a kernel question, still you guys are probably able to point me to what i am missing here. > Linux 3.18.11+ #4 SMP PREEMPT RT regards, Henning

2 4

[libvirt] [PATCH RFC 0/7] libxl: domain statistics support
by Joao Martins 14 Oct '15

14 Oct '15

Hey Jim, This series bring support for various statistics about domains regarding CPU, Memory, Network Interfaces and BlockStats. Not all of the statistics are implemented: qdisk support is missing in this series and some of the memory statistics aren't available. With this series we further implement 7 more functions of libvirt APIs. It is organized as follows: * Patch 1, 2: implements cpu/memory statistics. * Patch 3, 4: implements (netback) network statistics and VBD block statistics. QDisk will follow up in a separate series regarding QEMU monitor integration. * Patch 5: implement fetching all domain statistics * Patch 6, 7: implements Job information. Overall it looks big but 70% of the patch is due to #4 and #5 but doesn't add necessarily more complexity to the driver I believe. Patch #6 and #7 are of special importance because GetJobInfo and GetJobStats are now used in Openstack Kilo to monitor live-migration progress. This two patches together with an earlier series [0] I sent before let us sucessfully live-migrate with Openstack Kilo. Further with this series we get to support nova diagnostics. Tested this series on 4.4.3 and 4.5 setups plus Openstack Kilo. Any comments or suggestions are welcome, Thanks! Joao Joao Martins (7): libxl: implement virDomainGetCPUStats libxl: implement virDomainMemorystats libxl: implement virDomainInterfaceStats libxl: implement virDomainBlockStats libxl: implement virConnectGetAllDomainStats libxl: implement virDomainGetJobInfo libxl: implement virDomainGetJobStats configure.ac | 2 +- src/libxl/libxl_domain.c | 29 ++ src/libxl/libxl_domain.h | 6 + src/libxl/libxl_driver.c | 1010 ++++++++++++++++++++++++++++++++++++++++++++++ 4 files changed, 1046 insertions(+), 1 deletion(-) -- 2.1.4

3 27

[libvirt] [RFC PATCH 0/8] Make loading domains with invalid XML possible
by Martin Kletzander 13 Oct '15

13 Oct '15

We always had to deal with new parsing errors in a weird way. All of them needed to go into functions starting the domains. That messes up the code, it's confusing to newcomers and so on. I once had an idea that we can handle this stuff. We know what failed, we have the XML that failed parsing and if the problem is not in the domain name nor UUID, we can create a domain object out of that as well. This way we are able to do something we weren't with this series applied. Example follows. Assume "dummy" is a domain with invalid XML (I just modified the file). Now, just for the purpose of this silly example, let's say we allowed domains with archtecture x86_*, but now we've realized that x86_64 is the only one we want to allow, but there already is a domain defined that has <type arch='x86_256' .../>. With the current master, the domain would be lost, so we would need to modify the funstion starting the domain (e.g. qemuProcessStart for qemu domain). However, with this series, it behaves like this: # virsh list --all Id Name State ---------------------------------------------------- - dummy shut off # virsh domstate --reason dummy shut off (invalid XML) # virsh start dummy error: Failed to start domain dummy error: XML error: domain 'dummy' was not loaded due to an XML error (unsupported configuration: Unknown architecture x86_256), please redefine it # VISUAL='sed -i s/x86_256/x86_64/' virsh edit dummy Domain dummy XML configuration edited. # virsh domstate --reason dummy shut off (unknown) # virsh start dummy Domain dummy started This is a logical next step for us to clean and separate parsing and starting, getting rid of some old code without sacrifying compatibility and maybe generating parser code in the future. Why is this an RFC? - I can certainly easily imagine some people who might not like this - It doesn't work for all drivers (yet) - It does handle status XMLs, but only slightly. All the handling is done in a commit message that says something along the lines of "beware, we should still be careful, but not as much as before". - The error type used for the message is one that is already used for something else and we might want a new type of error for exactly this one error message, although it might be enough for some to have the possibility of checking this by querying the domain state and checking the reason. - Just so you know I came up with something new for which there's no BZ (or at least yet) =) Martin Kletzander (8): conf, virsh: Add new domain shutoff reason qemu: Few whitespace cleanups conf: Extract name-parsing into its own function conf: Extract UUID parsing into its own function conf: Optionally keep domains with invalid XML qemu: Don't lookup invalid domains unless specified otherwise qemu: Prepare basic APIs to handle invalid defs qemu: Load domains with invalid XML on start include/libvirt/libvirt-domain.h | 2 + src/bhyve/bhyve_driver.c | 2 + src/conf/domain_conf.c | 180 +++++++++++++++++++++++++++++++++------ src/conf/domain_conf.h | 5 ++ src/libxl/libxl_driver.c | 3 + src/lxc/lxc_driver.c | 3 + src/qemu/qemu_driver.c | 64 +++++++++++--- src/uml/uml_driver.c | 2 + tools/virsh-domain-monitor.c | 3 +- 9 files changed, 223 insertions(+), 41 deletions(-) -- 2.5.3

4 24

[libvirt] [PATCH 0/2] improve the error in 2 function
by Luyao Huang 12 Oct '15

12 Oct '15

*** BLURB HERE *** Luyao Huang (2): util: split the virNumaGetHugePageInfoPath into separate function util: Produce friendlier error message to user src/util/virnuma.c | 54 ++++++++++++++++++++++++++++++------------------------ 1 file changed, 30 insertions(+), 24 deletions(-) -- 1.8.3.1

3 10

[libvirt] [libvirt-test-API] [PATCH] fix some typo
by Shanzhi Yu 09 Oct '15

09 Oct '15

Signed-off-by: Shanzhi Yu <shyu(a)redhat.com> --- repos/domain/fsinfo.py | 2 +- repos/domain/install_linux_check.py | 2 +- repos/domain/install_linux_net.py | 2 +- repos/interface/define.py | 2 +- repos/interface/undefine.py | 4 ++-- 5 files changed, 6 insertions(+), 6 deletions(-) diff --git a/repos/domain/fsinfo.py b/repos/domain/fsinfo.py index e6d1bf0..d4614f5 100644 --- a/repos/domain/fsinfo.py +++ b/repos/domain/fsinfo.py @@ -81,7 +81,7 @@ def fsinfo(params): return 1 fsinfo = vm.fsInfo() - logger.info("get geust filesystem information") + logger.info("get guest filesystem information") mac = get_guest_mac(vm) if not mac: diff --git a/repos/domain/install_linux_check.py b/repos/domain/install_linux_check.py index ab1e7db..bb3cc2a 100644 --- a/repos/domain/install_linux_check.py +++ b/repos/domain/install_linux_check.py @@ -99,7 +99,7 @@ def install_linux_check(params): Test_Result = 1 return Test_Result - # Check whether vcpu equals the value set in geust config xml + # Check whether vcpu equals the value set in guest config xml logger.info("check point3: check cpu number in guest equals to \ the value set in domain config xml") vcpunum_expect = int(utils.get_num_vcpus(domain_name)) diff --git a/repos/domain/install_linux_net.py b/repos/domain/install_linux_net.py index 4367d97..7173294 100644 --- a/repos/domain/install_linux_net.py +++ b/repos/domain/install_linux_net.py @@ -239,7 +239,7 @@ def install_linux_net(params): logger.info("booting guest vm off harddisk failed") return 1 else: - logger.info("geust is booting up") + logger.info("guest is booting up") else: interval = 0 while(interval < 3600): diff --git a/repos/interface/define.py b/repos/interface/define.py index 18ba71b..e0dbcf4 100644 --- a/repos/interface/define.py +++ b/repos/interface/define.py @@ -42,7 +42,7 @@ def define(params): try: conn.interfaceDefineXML(xmlstr, 0) if check_define_interface(ifacename): - logger.info("define a interface form xml is successful") + logger.info("define a interface from xml is successful") else: logger.error("fail to check define interface") return 1 diff --git a/repos/interface/undefine.py b/repos/interface/undefine.py index f9b2d74..8b250de 100644 --- a/repos/interface/undefine.py +++ b/repos/interface/undefine.py @@ -39,14 +39,14 @@ def undefine(params): try: ifaceobj.undefine() if check_undefine_interface(ifacename): - logger.info("undefine a interface form xml is successful") + logger.info("undefine a interface is successful") else: logger.error("fail to check undefine interface") return 1 except libvirtError, e: logger.error("API error message: %s, error code is %s" \ % (e.message, e.get_error_code())) - logger.error("fail to undefine a interface from xml") + logger.error("fail to undefine a interface") return 1 return 0 -- 1.8.3.1

2 1

[libvirt] [libvirt-test-API] [PATCH] fix two error in script pin_iothread.py and connection_getAllDomainStats.py
by Shanzhi Yu 09 Oct '15

09 Oct '15

Signed-off-by: Shanzhi Yu <shyu(a)redhat.com> --- repos/domain/pin_iothread.py | 2 +- repos/virconn/connection_getAllDomainStats.py | 4 ++-- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/repos/domain/pin_iothread.py b/repos/domain/pin_iothread.py index 28ee2c4..d2d812f 100644 --- a/repos/domain/pin_iothread.py +++ b/repos/domain/pin_iothread.py @@ -116,7 +116,7 @@ def pin_iothread(params): if not find_iothreadid_fromxml(vm, 1, 1): logger.info("add iothread %d to running guest" % 1) - vm.addIOThread(i, libvirt.VIR_DOMAIN_AFFECT_LIVE) + vm.addIOThread(1, libvirt.VIR_DOMAIN_AFFECT_LIVE) vm.pinIOThread(1, tu_cpu, libvirt.VIR_DOMAIN_AFFECT_LIVE) cpuset = find_iothreadpin_fromxml(vm, 1, 1) diff --git a/repos/virconn/connection_getAllDomainStats.py b/repos/virconn/connection_getAllDomainStats.py index d95004f..3362444 100644 --- a/repos/virconn/connection_getAllDomainStats.py +++ b/repos/virconn/connection_getAllDomainStats.py @@ -164,7 +164,7 @@ def check_each_vcpu(logger,dom_name,dom_active,dom_eles): if not vcpu_cur: for i in range(0,vcpu_max): vcpu_pre = "vcpu."+ str(i) + "." - logger.debug("Checking %sstate: %d" \ + logger.debug("Checking %sstate: %s" \ %(vcpu_pre, dom_eles.get(vcpu_pre + "state"))) if not compare_value(logger,vcpu_stat, \ dom_eles.get(vcpu_pre + "state")): @@ -172,7 +172,7 @@ def check_each_vcpu(logger,dom_name,dom_active,dom_eles): elif int(vcpu_cur.nodeValue) <= vcpu_max: for i in range(0,int(vcpu_cur.nodeValue)): vcpu_pre = "vcpu."+ str(i) + "." - logger.debug("Checking %sstate: %d" \ + logger.debug("Checking %sstate: %s" \ %(vcpu_pre, dom_eles.get(vcpu_pre + "state"))) if not compare_value(logger,vcpu_stat, \ dom_eles.get(vcpu_pre + "state")): -- 1.8.3.1

2 1