January 2021 - Devel - libvirt List Archives

[libvirt PATCH 0/2] Propose a simple mechanism on how to disable pipeline jobs
by Erik Skultety 11 Jan '21

11 Jan '21

>From time to time it happens that some of the distros for which we run our pipelines break (or the images we pull break). Because we don't have dedicated maintainers for the jobs/runners, we can only employ the best effort approach wrt to making the pipeline green again when we're positive the problem doesn't lie in libvirt. For jobs that are broken for quite some time we should opt to disable them unconditionally unless a volunteer dedicates the time to either fix it or the matter gets fixed on its own in time (e.g. updated container images). GitLab only provides the "only/except" keywords to conditionally run or skip jobs. However, the problem with this is that since the job is not scheduled, = it won't appear in the pipeline output, so it's by no means clear to an outsider looking at the pipeline that something "is missing" in the UI. There have been a few open feature requests to incorporate exit status reporting for jobs and their interpretation in the pipeline output [1], but the community has been waiting for this for over 4 years already. As for patch 2, I spent some time this morning to figure out why the FreeBSD jobs were failing, not knowing the qemu-utils package is not outdated and surpassed by the qemu package in port. As a result I intended to disable the FreeBSD jobs until the maintainers flip the qemu-utils package from BROKEN to available again. Now that Dan solved the problem in libvirt-ci, I only left it there for trivial demonstration purposes. [1] https://gitlab.com/gitlab-org/gitlab-foss/-/issues/25738 Erik Skultety (2): gitlab-ci.yaml: Introduce a variable to disable long broken jobs DO NOT MERGE: Showcase a disabled job .gitlab-ci.yml | 14 ++++++++++++++ 1 file changed, 14 insertions(+) --=20 2.29.2

3 8

[PATCH 0/2] qemu: migration: Don't use stale block job state when migrating disks
by Peter Krempa 11 Jan '21

11 Jan '21

Peter Krempa (2): qemuMigrationSrcNBDStorageCopyReady: Use ready-state of mirror from qemuBlockJobData qemuBlockJobEventProcess: Always clear 'mirrorState' when a job finishes src/qemu/qemu_blockjob.c | 4 ++++ src/qemu/qemu_migration.c | 6 +++--- 2 files changed, 7 insertions(+), 3 deletions(-) -- 2.29.2

2 3

[PATCH] docs: Mention KubeVirt in apps using Libvirt
by Michal Privoznik 11 Jan '21

11 Jan '21

KubeVirt uses Libvirt to manage virtual machines run in containers. Mention this in our apps page. Signed-off-by: Michal Privoznik <mprivozn(a)redhat.com> --- docs/apps.html.in | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/docs/apps.html.in b/docs/apps.html.in index 99687c758d..bac1e36365 100644 --- a/docs/apps.html.in +++ b/docs/apps.html.in @@ -253,6 +253,13 @@ life-cycle, monitoring and so on. </dd> + <dt><a href="https://kubevirt.io/">KubeVirt</a></dt> + <dd> + KubeVirt is a virtual machine management add-on for Kubernetes. The aim + is to provide a common ground for virtualization solutions on top of + Kubernetes. + </dd> + <dt><a href="https://github.com/gustavfranssonnyvell/cherrypop">Cherrypop</a></dt> <dd> A cloud software with no masters or central points. Nodes -- 2.26.2

2 1

[PATCH] openvswitch: Check if OVS_VSCTL exists when getting interface name
by Michal Privoznik 11 Jan '21

11 Jan '21

So far we assumed that any vhostuser interface is plugged into an OVS bridge and thus 'ovs-vsctl' exists. But this is not always true. In testing scenarios it is possible to create a vhostuser interface with this tool dpdk-testpmd (part of dpdk RPM) which creates/connects to UNIX socket needed for vhostuser. Of course, since there is no OVS then there is no interface name in which case virNetDevOpenvswitchGetVhostuserIfname() should return 0. The rest of APIs that assume OVS are not 'fixed' because we still want them to fail (e.g. getting statistics, plugging interface into an OVS bridge, unplugging it from an OVS bridge, ...). The only API that is fixed is virNetDevOpenvswitchGetVhostuserIfname() because it is called explicitly when starting a guest (and callers are okay if no name was found). The other way to fix this bug seems to be to simply require 'ovs-vsctl' on spec file level, but that is too heavy gun given that vhostuser is used by a small set of our users (assumption made on requirements for vhostuser). Also, this way would drag in yet another dependency for all users (even those who want minimal libvirt). Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1913156 Signed-off-by: Michal Privoznik <mprivozn(a)redhat.com> --- src/util/virnetdevopenvswitch.c | 13 ++++++++++++- 1 file changed, 12 insertions(+), 1 deletion(-) diff --git a/src/util/virnetdevopenvswitch.c b/src/util/virnetdevopenvswitch.c index 896c1bd917..f9b3369b2a 100644 --- a/src/util/virnetdevopenvswitch.c +++ b/src/util/virnetdevopenvswitch.c @@ -29,6 +29,7 @@ #include "virstring.h" #include "virlog.h" #include "virjson.h" +#include "virfile.h" #define VIR_FROM_THIS VIR_FROM_NONE @@ -529,9 +530,19 @@ virNetDevOpenvswitchGetVhostuserIfname(const char *path, bool server, char **ifname) { - g_autoptr(virCommand) cmd = virNetDevOpenvswitchCreateCmd(); + g_autoptr(virCommand) cmd = NULL; + g_autofree char *absoluteOvsVsctlPath = NULL; int status; + if (!(absoluteOvsVsctlPath = virFindFileInPath(OVS_VSCTL))) { + /* If there is no 'ovs-vsctl' then the interface is + * probably not an OpenVSwitch interface and the @path to + * socket was created by some DPDK testing script (e.g. + * dpdk-testpmd). */ + return 0; + } + + cmd = virNetDevOpenvswitchCreateCmd(); if (server) { virCommandAddArgList(cmd, "--no-headings", "--columns=name", "find", -- 2.26.2

2 1

[PATCH 00/11] hyperv: implement defining/undefining domains and
by Matt Coleman 11 Jan '21

11 Jan '21

This series starts by fixing a some dangerous behavior due to ambiguous VM names, since Hyper-V allows multiple VMs to be defined with the same name. That meant that `virsh dumpxml` could return XML for the wrong domain. Additionally, when I implemented `undefine`, it would just select one of the domains with the given name. The majority of the series implements defining and undefining domains, as well as attaching storage devices to domains. Networking functionality will be in the next patch series. Here's a GitLab MR, if you'd prefer to review the changes there: https://gitlab.com/iammattcoleman/libvirt/-/merge_requests/14 Matt Coleman (11): hyperv: ambiguous VM names will throw an error hyperv: implement domainUndefine and domainUndefineFlags hyperv: implement domainCreateXML and domainDefineXML hyperv: add hypervMsvmVSMSAddResourceSettings hyperv: create SCSI controllers when defining domains hyperv: attach virtual disks when defining domains hyperv: attach physical disks when defining domains hyperv: attach virtual optical disks when defining domains hyperv: attach floppy disks when defining domains hyperv: implement domainAttachDevice and domainAttachDeviceFlags news: defining/undefining domains & device attachment for Hyper-V NEWS.rst | 7 +- include/libvirt/virterror.h | 1 + src/hyperv/hyperv_driver.c | 881 ++++++++++++++++++++++++++++++++ src/hyperv/hyperv_wmi.c | 52 ++ src/hyperv/hyperv_wmi.h | 5 + src/hyperv/hyperv_wmi_classes.h | 1 + src/util/virerror.c | 6 +- 7 files changed, 949 insertions(+), 4 deletions(-) -- 2.27.0

2 23

[PATCH] NEWS: Mention stuff I've dealt with for 7.0
by Peter Krempa 11 Jan '21

11 Jan '21

Signed-off-by: Peter Krempa <pkrempa(a)redhat.com> --- NEWS.rst | 37 +++++++++++++++++++++++++++++++++++++ 1 file changed, 37 insertions(+) diff --git a/NEWS.rst b/NEWS.rst index da88b19d0a..cb111ccb0f 100644 --- a/NEWS.rst +++ b/NEWS.rst @@ -11,6 +11,15 @@ For a more fine-grained view, use the `git log`_. v7.0.0 (unreleased) =================== +* **Project governance** + + * Formal handover of release tarball signing + + Starting from libvirt-6.6 the release tarballs are signed by Jiří Denemark. + Releases starting with 7.0 contain a note from the previous maintainer + Daniel Veillard officaly handing over the signing of packages so that the + transition can be verified. + * **New features** * nodedev: Add node device driver support for AP devices @@ -19,10 +28,38 @@ v7.0.0 (unreleased) queues and AP matrix devices (which are capable of MDEV) of a KVM host system in libvirt node device driver with correct object relationships. + * qemu: Allow control of ``qcow2`` metadata cache + + In specific usecases such as when massive storage images are used it's + possible to achieve better performance with increasing the metadata cache + size. The new knob allows advanced users setting the size according to + qemu's documentation to suit their image. + * **Improvements** + * qemu: Discourage users from polling ``virDomainGetBlockJobInfo`` for block + job completion + + Document that waiting for events is a more robust solution. + + * secret: Relax XML schema for the ``usage`` name of a ``secret`` + + Various bits of documentation of how to use libvirt with RBD volumes used + an usage name which would not pass the XML validation. Relax the requirement + to make such XMLs valid. + * **Bug fixes** + * qemu: Fix logic bug in inactive snapshot deletion + + This release fixes a bug introduced in ``libvirt-6.9`` where libvirt's + snapshot metadata would not be deleted on successful snapshot deletion. + + * qemu: Fix VM's with ``<iotune>`` on an empty cdrom + + Specifying ``<iotune>`` for an empty cdrom would prevent the VM from + starting as qemu doesn't accept the tuning for an empty drive. We now + postpone the setting of the parameters until a new media is inserted. v6.10.0 (2020-12-01) ==================== -- 2.29.2

2 1

[libvirt PATCH 0/1] Adjust the SEV limitations kbase article section
by Erik Skultety 11 Jan '21

11 Jan '21

virtio-blk has been available for use with boot disks for some time now, so update the text. Erik Skultety (1): docs: kbase: sev: Adjust the claims that virtio-blk doesn't work docs/kbase/launch_security_sev.rst | 19 +++++++++---------- 1 file changed, 9 insertions(+), 10 deletions(-) -- 2.29.2

2 2

[PATCH 0/2] downloads: Establish handover of signing releases to Jirka
by Peter Krempa 11 Jan '21

11 Jan '21

Peter Krempa (2): docs: downloads: Move 'signatures' section to the end of the document docs: downloads: Establish handover of package signing docs/downloads.html.in | 64 +++++++++++++++++++++++++++++++++--------- 1 file changed, 50 insertions(+), 14 deletions(-) -- 2.29.2

2 4

[PATCH v2] rpm: adjust xenlight requirements
by Olaf Hering 11 Jan '21

11 Jan '21

According to meson.build, and the actual code, Xen 4.6+ is required. Drop the version, they is mentioned in meson.build. Signed-off-by: Olaf Hering <olaf(a)aepfle.de> --- libvirt.spec.in | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/libvirt.spec.in b/libvirt.spec.in index 0a8b0ebad4..470631f7bd 100644 --- a/libvirt.spec.in +++ b/libvirt.spec.in @@ -126,7 +126,7 @@ %endif # RHEL doesn't ship OpenVZ, VBox, PowerHypervisor, -# VMware, libxenlight (Xen 4.1 and newer), +# VMware, libxenlight # or HyperV. %if 0%{?rhel} %define with_openvz 0 @@ -272,7 +272,7 @@ BuildRequires: perl BuildRequires: python3 BuildRequires: systemd-units %if %{with_libxl} -BuildRequires: xen-devel +BuildRequires: pkgconfig(xenlight) %endif BuildRequires: glib2-devel >= 2.48 BuildRequires: libxml2-devel

3 3

[PATCH RFC v2 2/2] qemu: suppor dirty ring feature
by huangy81＠chinatelecom.cn 10 Jan '21

10 Jan '21

From: Hyman <huangy81(a)chinatelecom.cn> QEMU introduced a dirty ring feature, this patch add a new KVM feature 'dirty-ring' to set this feature for kvm guests. To enable the feature, libvirt add "-accel dirty-gfn-count=xxx" to QEMU command line, the following XML needs to be added to the guest's domain description: <features> <kvm> <dirty-ring state='on' size='xxx'> </kvm> </features> Signed-off-by: Hyman <huangy81(a)chinatelecom.cn> --- docs/formatdomain.rst | 16 +++++++++------- docs/schemas/domaincommon.rng | 10 ++++++++++ src/conf/domain_conf.c | 34 ++++++++++++++++++++++++++++++++++ src/conf/domain_conf.h | 4 ++++ src/qemu/qemu_command.c | 25 +++++++++++++++++++++++++ 5 files changed, 82 insertions(+), 7 deletions(-) diff --git a/docs/formatdomain.rst b/docs/formatdomain.rst index 6100b88..16338be 100644 --- a/docs/formatdomain.rst +++ b/docs/formatdomain.rst @@ -1767,6 +1767,7 @@ Hypervisors may allow certain CPU / machine features to be toggled on/off. <hidden state='on'/> <hint-dedicated state='on'/> <poll-control='on'/> + <dirty-ring state='on' size='4096'/> </kvm> <xen> <e820_host state='on'/> @@ -1849,13 +1850,14 @@ are: ``kvm`` Various features to change the behavior of the KVM hypervisor. - ============== ============================================================================ ======= ============================ - Feature Description Value Since - ============== ============================================================================ ======= ============================ - hidden Hide the KVM hypervisor from standard MSR based discovery on, off :since:`1.2.8 (QEMU 2.1.0)` - hint-dedicated Allows a guest to enable optimizations when running on dedicated vCPUs on, off :since:`5.7.0 (QEMU 2.12.0)` - poll-control Decrease IO completion latency by introducing a grace period of busy waiting on, off :since:`6.10.0 (QEMU 4.2)` - ============== ============================================================================ ======= ============================ + ============== ============================================================================ ================================== ============================ + Feature Description Value Since + ============== ============================================================================ ================================== ============================ + hidden Hide the KVM hypervisor from standard MSR based discovery on, off :since:`1.2.8 (QEMU 2.1.0)` + hint-dedicated Allows a guest to enable optimizations when running on dedicated vCPUs on, off :since:`5.7.0 (QEMU 2.12.0)` + poll-control Decrease IO completion latency by introducing a grace period of busy waiting on, off :since:`6.10.0 (QEMU 4.2)` + dirty-ring Enable dirty ring feature on, off; size - must be power of 2 :since:`6.10.1 (QEMU 5.2.1)` + ============== ============================================================================ ================================== ============================ ``xen`` Various features to change the behavior of the Xen hypervisor. diff --git a/docs/schemas/domaincommon.rng b/docs/schemas/domaincommon.rng index 7dc419b..5af4bbe 100644 --- a/docs/schemas/domaincommon.rng +++ b/docs/schemas/domaincommon.rng @@ -6569,6 +6569,16 @@ <ref name="featurestate"/> </element> </optional> + <optional> + <element name="dirty-ring"> + <ref name="featurestate"/> + <optional> + <attribute name="size"> + <data type="unsignedInt"/> + </attribute> + </optional> + </element> + </optional> </interleave> </element> </define> diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c index 349fc28..e617b95 100644 --- a/src/conf/domain_conf.c +++ b/src/conf/domain_conf.c @@ -186,6 +186,7 @@ VIR_ENUM_IMPL(virDomainKVM, "hidden", "hint-dedicated", "poll-control", + "dirty-ring", ); VIR_ENUM_IMPL(virDomainXen, @@ -18379,6 +18380,7 @@ virDomainFeaturesDefParse(virDomainDefPtr def, if (def->features[VIR_DOMAIN_FEATURE_KVM] == VIR_TRISTATE_SWITCH_ON) { int feature; int value; + node = ctxt->node; if ((n = virXPathNodeSet("./features/kvm/*", ctxt, &nodes)) < 0) return -1; @@ -18395,6 +18397,7 @@ virDomainFeaturesDefParse(virDomainDefPtr def, case VIR_DOMAIN_KVM_HIDDEN: case VIR_DOMAIN_KVM_DEDICATED: case VIR_DOMAIN_KVM_POLLCONTROL: + case VIR_DOMAIN_KVM_DIRTY_RING: if (!(tmp = virXMLPropString(nodes[i], "state"))) { virReportError(VIR_ERR_XML_ERROR, _("missing 'state' attribute for " @@ -18413,6 +18416,26 @@ virDomainFeaturesDefParse(virDomainDefPtr def, VIR_FREE(tmp); def->kvm_features[feature] = value; + + /* only for dirty ring case */ + if (((virDomainKVM) feature) == VIR_DOMAIN_KVM_DIRTY_RING && + value == VIR_TRISTATE_SWITCH_ON) { + ctxt->node = nodes[i]; + if (virXPathUInt("string(./@size)", ctxt, + &def->dirty_gfn_count) < 0) { + virReportError(VIR_ERR_XML_ERROR, "%s", + _("invalid number of dirty GFNs")); + return -1; + } + + if ((def->dirty_gfn_count & (def->dirty_gfn_count - 1)) || + def->dirty_gfn_count > 65536) { + virReportError(VIR_ERR_XML_ERROR, "%s", + _("max number of dirty GFNs is 65536 " + "and must be power of 2")); + return -1; + } + } break; /* coverity[dead_error_begin] */ @@ -18421,6 +18444,7 @@ virDomainFeaturesDefParse(virDomainDefPtr def, } } VIR_FREE(nodes); + ctxt->node = node; } if (def->features[VIR_DOMAIN_FEATURE_XEN] == VIR_TRISTATE_SWITCH_ON) { @@ -22521,6 +22545,7 @@ virDomainDefFeaturesCheckABIStability(virDomainDefPtr src, case VIR_DOMAIN_KVM_HIDDEN: case VIR_DOMAIN_KVM_DEDICATED: case VIR_DOMAIN_KVM_POLLCONTROL: + case VIR_DOMAIN_KVM_DIRTY_RING: if (src->kvm_features[i] != dst->kvm_features[i]) { virReportError(VIR_ERR_CONFIG_UNSUPPORTED, _("State of KVM feature '%s' differs: " @@ -28271,6 +28296,15 @@ virDomainDefFormatFeatures(virBufferPtr buf, def->kvm_features[j])); break; + case VIR_DOMAIN_KVM_DIRTY_RING: + if (def->kvm_features[j]) + virBufferAsprintf(&childBuf, "<%s state='%s' size='%d'/>\n", + virDomainKVMTypeToString(j), + virTristateSwitchTypeToString( + def->kvm_features[j]), + def->dirty_gfn_count); + break; + /* coverity[dead_error_begin] */ case VIR_DOMAIN_KVM_LAST: break; diff --git a/src/conf/domain_conf.h b/src/conf/domain_conf.h index ec43bbe..120d490 100644 --- a/src/conf/domain_conf.h +++ b/src/conf/domain_conf.h @@ -1917,6 +1917,7 @@ typedef enum { VIR_DOMAIN_KVM_HIDDEN = 0, VIR_DOMAIN_KVM_DEDICATED, VIR_DOMAIN_KVM_POLLCONTROL, + VIR_DOMAIN_KVM_DIRTY_RING, VIR_DOMAIN_KVM_LAST } virDomainKVM; @@ -2728,6 +2729,9 @@ struct _virDomainDef { callbacks failed for a non-critical reason (was not able to fill in some data) and thus should be re-run before starting */ + + /* Number of dirty GFNs per ring */ + unsigned int dirty_gfn_count; }; diff --git a/src/qemu/qemu_command.c b/src/qemu/qemu_command.c index 9a64473..17487f3 100644 --- a/src/qemu/qemu_command.c +++ b/src/qemu/qemu_command.c @@ -6525,6 +6525,9 @@ qemuBuildCpuCommandLine(virCommandPtr cmd, virBufferAddLit(&buf, ",kvm-poll-control=on"); break; + case VIR_DOMAIN_KVM_DIRTY_RING: + break; + /* coverity[dead_error_begin] */ case VIR_DOMAIN_KVM_LAST: break; @@ -6911,6 +6914,26 @@ qemuBuildMachineCommandLine(virCommandPtr cmd, return 0; } + +static void +qemuBuildAccelCommandLineTcgOptions(void) +{ + /* TODO: build command line tcg accelerator + * property like tb-size */ +} + + +static void +qemuBuildAccelCommandLineKvmOptions(virBuffer *buf, + const virDomainDef *def) +{ + if (def->features[VIR_DOMAIN_FEATURE_KVM] == VIR_TRISTATE_SWITCH_ON && + def->kvm_features[VIR_DOMAIN_KVM_DIRTY_RING] == VIR_TRISTATE_SWITCH_ON) { + virBufferAsprintf(buf, ",dirty-gfn-count=%d", def->dirty_gfn_count); + } +} + + static int qemuBuildAccelCommandLine(virCommandPtr cmd, const virDomainDef *def) @@ -6923,10 +6946,12 @@ qemuBuildAccelCommandLine(virCommandPtr cmd, switch ((virDomainVirtType)def->virtType) { case VIR_DOMAIN_VIRT_QEMU: virBufferAddLit(&buf, "tcg"); + qemuBuildAccelCommandLineTcgOptions(); break; case VIR_DOMAIN_VIRT_KVM: virBufferAddLit(&buf, "kvm"); + qemuBuildAccelCommandLineKvmOptions(&buf, def); break; case VIR_DOMAIN_VIRT_KQEMU: -- 1.8.3.1

1 0