- Devel - libvirt List Archives

[libvirt] [PATCH] qemu: match controller index for LIVE+CONFIG when doing hotplug
by Laine Stump 27 Jun '16

27 Jun '16

An attempt to attach a new scsi controller with both --live and --config but without specifying an index, e.g.: <controller model="virtio-scsi" type="scsi" /> led to this error: internal error: Cannot parse controller index -1 This was because unspecified indexes are auto-assigned during virDomainDefPostParse(), which doesn't happen for hotplugged devices until after the device has been added to the domainDef, but qemuDomainAttachFlags() makes a copy of the device (for feeding to qemuDomainAttachDeviceLive() *before* it's added to the config, and the copying function actually formats the device object and then re-parses it into a new object. Since qemuDomainAttachDeviceConfig() consumes the device object pointer (i.e. sets it to NULL in the original virDomainDeviceDef) we can't just wait to make the copy. Instead, we need to make a *shallow* copy of the virDomainDeviceDef prior to qemuDomainAttachDeviceConfig(), then make a deep copy of the shallow copy. This resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1344899 --- src/qemu/qemu_driver.c | 31 ++++++++++++++++++++----------- 1 file changed, 20 insertions(+), 11 deletions(-) diff --git a/src/qemu/qemu_driver.c b/src/qemu/qemu_driver.c index fa05046..f3e17e2 100644 --- a/src/qemu/qemu_driver.c +++ b/src/qemu/qemu_driver.c @@ -8199,6 +8199,7 @@ qemuDomainAttachDeviceFlags(virDomainPtr dom, virDomainObjPtr vm = NULL; virDomainDefPtr vmdef = NULL; virDomainDeviceDefPtr dev = NULL, dev_copy = NULL; + virDomainDeviceDef dev_shallow; int ret = -1; unsigned int parse_flags = VIR_DOMAIN_DEF_PARSE_INACTIVE | VIR_DOMAIN_DEF_PARSE_ABI_UPDATE; @@ -8237,22 +8238,19 @@ qemuDomainAttachDeviceFlags(virDomainPtr dom, if (dev == NULL) goto endjob; - if (flags & VIR_DOMAIN_AFFECT_CONFIG && - flags & VIR_DOMAIN_AFFECT_LIVE) { - /* If we are affecting both CONFIG and LIVE - * create a deep copy of device as adding - * to CONFIG takes one instance. - */ - dev_copy = virDomainDeviceDefCopy(dev, vm->def, caps, driver->xmlopt); - if (!dev_copy) - goto endjob; - } - if (priv->qemuCaps) qemuCaps = virObjectRef(priv->qemuCaps); else if (!(qemuCaps = virQEMUCapsCacheLookup(driver->qemuCapsCache, vm->def->emulator))) goto endjob; + /* Save away the pointer to the device object before it is + * potentially swallowed up by qemuDomainAttachDeviceConfig(). + * This will allow us to make a copy of the device after any + * modifications made by virDomainDefPostParse() (which is called + * after the new device is added to the config + */ + dev_shallow = *dev; + if (flags & VIR_DOMAIN_AFFECT_CONFIG) { /* Make a copy for updated domain. */ vmdef = virDomainObjCopyPersistentDef(vm, caps, driver->xmlopt); @@ -8270,6 +8268,17 @@ qemuDomainAttachDeviceFlags(virDomainPtr dom, } if (flags & VIR_DOMAIN_AFFECT_LIVE) { + if (flags & VIR_DOMAIN_AFFECT_CONFIG) { + /* If we are affecting both CONFIG and LIVE create a deep + * copy of device from the saved shallow copy, as adding + * it to CONFIG has already consumed the original. + */ + dev_copy = virDomainDeviceDefCopy(&dev_shallow, vm->def, + caps, driver->xmlopt); + if (!dev_copy) + goto endjob; + } + if (virDomainDefCompatibleDevice(vm->def, dev_copy, VIR_DOMAIN_DEVICE_ACTION_ATTACH) < 0) goto endjob; -- 2.5.5

5 6

[libvirt] [PATCH] libvirtd.conf: Fix invalid default of max_anonymous_clients
by Michal Privoznik 27 Jun '16

27 Jun '16

https://bugzilla.redhat.com/show_bug.cgi?id=1343442 When a client connects, it is placed into a queue. As soon as it authenticate, it is taken out of that queue and placed into a different one. Now, we have a setting in the daemon config file that allows users to control the length of the queue of yet not authenticated clients. By default, it has a value 20 but in the description to the config knob we clam it's zero. Signed-off-by: Michal Privoznik <mprivozn(a)redhat.com> --- daemon/libvirtd.conf | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/daemon/libvirtd.conf b/daemon/libvirtd.conf index 3b957e5..8466616 100644 --- a/daemon/libvirtd.conf +++ b/daemon/libvirtd.conf @@ -283,8 +283,8 @@ #max_queued_clients = 1000 # The maximum length of queue of accepted but not yet -# authenticated clients. The default value is zero, meaning -# the feature is disabled. +# authenticated clients. The default value is 20. Set this to +# zero to turn this feature off. #max_anonymous_clients = 20 # The minimum limit sets the number of workers to start up -- 2.8.4

2 1

[libvirt] how to make syntax-check on tar-tall released-version
by Zhangbo (Oscar) 27 Jun '16

27 Jun '16

Hi all: I downloaded libvirt v1.3.4 from http://libvirt.org/sources, and try to make syntax-check inside the code tree. But I encounted with many problems, first, I need to have file .gitignore and directory .git, and then have gnulib, many problems still comes out after that. I still have not fixed the problems until now. So what's the right way to make syntax-check on released-version libvirt project? Thanks in advance. Oscar.

3 4

[libvirt] [PATCH 0/2] Make a generic version of virDomainDefMetadataSanitize()
by Laine Stump 27 Jun '16

27 Jun '16

Peter pointed out that Brandon's patch to add metadata to the network XML should "sanitize" it (to remove duplicate namespaces) in the same manner as we do (thanks to Peter!) for the domain metadata. In order to make that simpler, I made a patch that turns virDomainDefMetadataSanitize() into the generic function virXMLNodeSanitizeNamespaces(), then (in a separate patch) changed the domain parser to use that new generic function. I will also add a call to that function into Brandon's patch before pushing it. Laine Stump (2): util: new function virXMLNodeSanitizeNamespaces() qemu: use virXMLNodeSanitizeNamespaces() src/conf/domain_conf.c | 56 ++---------------------------------------------- src/libvirt_private.syms | 1 + src/util/virxml.c | 52 ++++++++++++++++++++++++++++++++++++++++++++ src/util/virxml.h | 2 ++ 4 files changed, 57 insertions(+), 54 deletions(-) -- 2.5.5

2 4

[libvirt] [PATCH 00/28] IP peer address and host-side address config support
by Laine Stump 26 Jun '16

26 Jun '16

This (admittedly very long, but most of the patches are either very small or just code movement/renaming, really!) series started out to be just a replacement for a single patch posted by Vasiliy Tolstov back in April (originally pushed upstream as commit 690969af, which was reverted in commit 1d14b13f when I noticed an inconsistency in behavior between lxc and qemu). But each step forward seemed to require (or strongly suggest) a step sideways to reorganize existing code, or eliminate annoying problems I found, and I ended up with 28 patches. The patches that actually add the new functionality are 24-28 (and they are quite small), all the rest just setting up for that. In particular, the three largest patches (03, 13, 14) are purely moving code and renaming identifiers. Patches 07, 08, 12, and 18 fix minor bugs I found during all the changes. (12 is a bit strange, as it removes an undocumented behavior from the openvz driver that has been there since 2010). The one possible controversy in this series (other than the modification to openvz configuration in patch 12) is my placement of the config info for the host-side network device. I could have either put the IP info for the host-side interface as a subelement of <target> where the host-side interface is named, or I could follow the convention of other device types and make it a subelement of the <source> element of interface (since there is where you're supposed to describe how a device connects to the host side resources). <target> element is used to give the name of the tap interface itself. I went back and forth about this, and in the end decided that since the IP address is configuring how the device connects to the outside world (just as with the "network" and "bridge" attributes of <source> for those other network types), that it makes most sense to put it under <source> Please don't be put off by the large number of patches - most of them are fairly straightforward and should be quick reviews. I promise! :-) Laine Stump (27): util: move virNetDevLinkDump to virnetlink.c util: move virInterface(State|Link)/virNetDevFeature from conf to util global: consistently use IP rather than Ip in identifiers conf: clean up virDomainNetIPParseXML() tests: mock virNetDevSetIPAddress util: allow calling virSocketAddrGetIPPrefix with NULL netmask or address lxc: eliminate extraneous free of netDef->ifname_guest lxc: use correct prefix when setting veth IP address conf: new function virDomainNetDefClear qemu: don't set/clear NetDef IP addresses in qemuConnectDomainXMLToNative() qemu: eliminate memory leaks when converting NetDefs to type='ethernet' conf/openvz: eliminate incorrect/undocumented use of <source dev='blah'/> util: new files virnetdevip.[ch] for IP-related netdev functions util: move IP route & address object-related functions to virnetdevip.c conf: single object containing list of IP addresses, list of routes conf: use virNetDevIPInfo in virDomainHostdevCaps conf: use virNetDevIPInfo for guest-side <interface> config qemu: forbid setting guest-side IP address/route info of <interface> conf: clean up after adding calls to virNetDevIPInfo helpers lxc: move debug/error log when adding IP addresses to virNetDevIPAddrAdd util: new function virNetDevIPInfoAddToDev util: provide default destination IP in alternate virNetDevIPRouteAdd() util: make messages consistent in both versions of virNetDevIPRouteAdd() conf: support host-side IP/route information in <interface> util: support setting peer for virNetDevIPInfo addresses lxc: support setting host-side IP addresses/routes qemu: support setting host-side IP addresses/routes Vasiliy Tolstov (1): conf: allow setting peer address in <ip> element of <interface> docs/formatdomain.html.in | 60 +- docs/schemas/domaincommon.rng | 82 +- po/POTFILES.in | 1 + src/Makefile.am | 1 + src/conf/device_conf.c | 31 +- src/conf/device_conf.h | 44 +- src/conf/domain_conf.c | 399 +++++---- src/conf/domain_conf.h | 28 +- src/conf/interface_conf.c | 38 +- src/conf/interface_conf.h | 10 +- src/conf/network_conf.c | 96 +- src/conf/network_conf.h | 24 +- src/conf/networkcommon_conf.c | 107 +-- src/conf/networkcommon_conf.h | 55 +- src/conf/node_device_conf.h | 2 +- src/esx/esx_driver.c | 44 +- src/esx/esx_interface_driver.c | 4 +- src/esx/esx_vi.c | 4 +- src/esx/esx_vi.h | 2 +- src/libvirt_private.syms | 50 +- src/libxl/libxl_conf.c | 10 +- src/libxl/libxl_domain.c | 2 +- src/lxc/lxc_container.c | 92 +- src/lxc/lxc_controller.c | 2 +- src/lxc/lxc_native.c | 42 +- src/lxc/lxc_process.c | 8 + src/network/bridge_driver.c | 103 +-- src/network/bridge_driver_linux.c | 76 +- src/nwfilter/nwfilter_ebiptables_driver.c | 24 +- src/openvz/openvz_conf.c | 2 +- src/openvz/openvz_driver.c | 15 +- src/qemu/qemu_domain.c | 34 + src/qemu/qemu_driver.c | 76 +- src/qemu/qemu_hotplug.c | 6 +- src/qemu/qemu_interface.c | 6 +- src/uml/uml_conf.c | 4 +- src/util/virnetdev.c | 876 +------------------ src/util/virnetdev.h | 46 +- src/util/virnetdevip.c | 971 +++++++++++++++++++++ src/util/virnetdevip.h | 93 ++ src/util/virnetdevvportprofile.c | 4 +- src/util/virnetlink.c | 135 +++ src/util/virnetlink.h | 5 + src/util/virsocketaddr.c | 12 +- src/util/virsocketaddr.h | 3 +- src/vbox/vbox_common.c | 6 +- src/vbox/vbox_network.c | 8 +- src/vz/vz_sdk.c | 2 +- src/xenconfig/xen_common.c | 14 +- src/xenconfig/xen_sxpr.c | 16 +- tests/lxcxml2xmldata/lxc-ethernet-hostip.xml | 44 + tests/lxcxml2xmltest.c | 1 + .../qemuxml2argv-net-eth-hostip.args | 23 + .../qemuxml2argv-net-eth-hostip.xml | 39 + tests/qemuxml2argvmock.c | 10 +- tests/qemuxml2argvtest.c | 1 + .../qemuxml2xmlout-net-eth-hostip.xml | 44 + tests/qemuxml2xmltest.c | 1 + tests/virnetdevtest.c | 14 +- tests/xml2sexprdata/xml2sexpr-net-routed.xml | 1 - 60 files changed, 2193 insertions(+), 1760 deletions(-) create mode 100644 src/util/virnetdevip.c create mode 100644 src/util/virnetdevip.h create mode 100644 tests/lxcxml2xmldata/lxc-ethernet-hostip.xml create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-net-eth-hostip.args create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-net-eth-hostip.xml create mode 100644 tests/qemuxml2xmloutdata/qemuxml2xmlout-net-eth-hostip.xml -- 2.5.5

2 72

[libvirt] [PATCH 0/3] Enable admin API functionality and tar distribution
by Erik Skultety 26 Jun '16

26 Jun '16

Finally enable admin API in our code and start distributing it both via a tarball and as part of an rpm package. I also added some usage examples, however the html documentation is missing, because of our current xsl stylesheet, more specifically because of how xsltproc works. It takes a stylesheet and an xml file which it processes and produces index.html for each xml file on its input. This way, we either get TOC for libvirt APIs or just the admin. The same story goes for libvirt-qemu and libvirt-lxc which we neither generate, nor distribute. I talked to Jirka and we thought about merging the xmls somehow to a temporary file and then generate the index.html but that requires some more digging, especially into xslt capabilities. So I haven't figured this one out yet, but once I do, I'll enable it for all of our *.xml API 'recipes'. Erik Skultety (3): admin: enable both admin API functionality and tarball distribution examples: admin: Add some examples for the new admin APIs spec: distribute admin API within libvirt-client package .gitignore | 6 ++ Makefile.am | 2 +- cfg.mk | 4 +- daemon/libvirtd.c | 2 +- docs/Makefile.am | 5 +- examples/Makefile.am | 14 +++- examples/admin/client_close.c | 57 +++++++++++++ examples/admin/client_info.c | 165 +++++++++++++++++++++++++++++++++++++ examples/admin/client_limits.c | 63 ++++++++++++++ examples/admin/list_clients.c | 111 +++++++++++++++++++++++++ examples/admin/list_servers.c | 42 ++++++++++ examples/admin/threadpool_params.c | 65 +++++++++++++++ include/libvirt/Makefile.am | 6 +- libvirt.spec.in | 14 ++-- src/libvirt_admin_public.syms | 2 +- 15 files changed, 538 insertions(+), 20 deletions(-) create mode 100644 examples/admin/client_close.c create mode 100644 examples/admin/client_info.c create mode 100644 examples/admin/client_limits.c create mode 100644 examples/admin/list_clients.c create mode 100644 examples/admin/list_servers.c create mode 100644 examples/admin/threadpool_params.c -- 2.5.5

2 5

[libvirt] [PATCH 00/18] vz: support disk/net/graphics devices updates
by Nikolay Shirokovskiy 25 Jun '16

25 Jun '16

It makes sense to support other devices too as update can be on persistent config only which is always possible, but as AFAIK the main purpose of this API is to change live config the set of devices is just set of devices supported by qemu for live updates. Nikolay Shirokovskiy (18): vz: remove disk cache mode hunk vz: make prlsdkGetDisk more generic vz: reuse edit config frame in for attach/detach functions vz: add device updates vz: leverage disks parameters check on disks updates too vz: move disks checks to device post parse vz: fix memory leaks in attach/detach functions vz: give nice report if network device not found vz: dump ip addresses to domain xml vz: dump route info in domain xml vz: fix updating to no gateways vz: fix minor type safey issues with net union usage vz: trustGuestRxFilters fixes vz: move getting container video devices out from vnc code vz: support attach/detach/update/ of graphics device vz: remove exlicitly setting zeros in dumping graphics vz: support vnc password vz: always pass graphics address to sdk src/conf/networkcommon_conf.c | 2 +- src/conf/networkcommon_conf.h | 2 +- src/libvirt_private.syms | 1 + src/vz/vz_driver.c | 115 +++--- src/vz/vz_sdk.c | 897 +++++++++++++++++++++++------------------- src/vz/vz_sdk.h | 10 +- src/vz/vz_utils.c | 204 ++++++++-- src/vz/vz_utils.h | 7 +- 8 files changed, 740 insertions(+), 498 deletions(-) -- 1.8.3.1

2 36

[libvirt] [Bug] make check fails when LIBVIRT_DEBUG is defined
by Tomasz Flendrich 25 Jun '16

25 Jun '16

Hello everybody, I am resending this email as I suppose that the previous one wasn’t accepted due to a big attachment inside. If I create a double post, please excuse me, I didn’t mean to. I set up a clean libvirt repo with: git clone ./autogen.sh make and then executed tests with various values of LIBVIRT_DEBUG using the following command: LIBVIRT_DEBUG=0…5 VIR_TEST_DEBUG=1 make check With LIBVIRT_DEBUG=0 (or =5 or probably any invalid value), the logs say: "Ignoring invalid log level setting” but still, 10 tests fail due to various reasons. With LIBVIRT_DEBUG=1, there is a slight difference. The number of tests that fail is also 10, but here, commandtest fails instead of eventtest. With LIBVIRT_DEBUG=2, only 9 tests fail, because commandtest passes. With LIBVIRT_DEBUG=3, only 1 test fails: qemuargv2xmltest. The reason is the same as in the previous tests. With LIBVIRT_DEBUG=4, it’s just qemuargv2xmltest that is failing, but for a different reason than before: "qemuParseCommandLineString should have logged a warning”. Links to tests/test-syslog.log, indexed by value of LIBVIRT_DEBUG: 0: http://pastebin.com/bZrDV1K2 <http://pastebin.com/bZrDV1K2> 1: https://gist.github.com/Darge/d180ebdd497573ed633625cf136de73c <https://gist.github.com/Darge/d180ebdd497573ed633625cf136de73c> 2: http://pastebin.com/bhsq1MtC <http://pastebin.com/bhsq1MtC> 3: http://pastebin.com/BSRtR9Rq <http://pastebin.com/BSRtR9Rq> 4: http://pastebin.com/FQn79her <http://pastebin.com/FQn79her> Perhaps it’s okay that these tests fail, but still there's no harm in letting people know. Thank you rbogorodskiy for guidance on the mailing list :) Have a nice day, Tomasz

1 0

[libvirt] [PATCH 0/3] Drop some useless CPU code
by Jiri Denemark 25 Jun '16

25 Jun '16

Jiri Denemark (3): cpu: Consolidate ARM drivers cpu: Drop generic driver cpu: Drop NR_DRIVERS macro po/POTFILES.in | 1 - src/Makefile.am | 2 - src/cpu/cpu.c | 16 ++-- src/cpu/cpu_aarch64.c | 133 ----------------------------- src/cpu/cpu_aarch64.h | 31 ------- src/cpu/cpu_arm.c | 37 +++++---- src/cpu/cpu_generic.c | 226 -------------------------------------------------- src/cpu/cpu_generic.h | 32 ------- 8 files changed, 26 insertions(+), 452 deletions(-) delete mode 100644 src/cpu/cpu_aarch64.c delete mode 100644 src/cpu/cpu_aarch64.h delete mode 100644 src/cpu/cpu_generic.c delete mode 100644 src/cpu/cpu_generic.h -- 2.9.0

3 9

[libvirt] [PATCH v2 1/1] cpu_map.xml: add cmt/mbm feature to x86
by Qiaowei Ren 25 Jun '16

25 Jun '16

Some Intel processor families (e.g. the Intel Xeon processor E5 v3 family) introduced some PQos (Platform Qos) features, including CMT (Cache Monitoring technology) and MBM (Memory Bandwidth Monitoring), to monitor or control shared resource. This patch add them into x86 part of cpu_map.xml to be used for applications based on libvirt to get cpu capabilities. For example, Nova in OpenStack schedules guests based on the CPU features that the host has. Signed-off-by: Qiaowei Ren <qiaowei.ren(a)intel.com> --- src/cpu/cpu_map.xml | 11 +++++++++++ 1 file changed, 11 insertions(+) diff --git a/src/cpu/cpu_map.xml b/src/cpu/cpu_map.xml index bf2dfc6..6da8321 100644 --- a/src/cpu/cpu_map.xml +++ b/src/cpu/cpu_map.xml @@ -224,6 +224,9 @@ <feature name='rtm'> <cpuid eax_in='0x07' ebx='0x00000800'/> </feature> + <feature name='cmt'> + <cpuid eax_in='0x07' ebx='0x00001000'/> + </feature> <feature name='mpx'> <cpuid eax_in='0x07' ebx='0x00004000'/> </feature> @@ -266,6 +269,14 @@ <cpuid eax_in='0x0d' ecx_in='0x01' eax='0x00000008'/> </feature> +  + <feature name='mbm_total'>  + <cpuid eax_in='0x0f' ecx_in='0x01' edx='0x00000002'/> + </feature> + <feature name='mbm_local'>  + <cpuid eax_in='0x0f' ecx_in='0x01' edx='0x00000004'/> + </feature> +  <feature name='syscall'>  <cpuid eax_in='0x80000001' edx='0x00000800'/> -- 1.9.1

2 1