September 2018 - Devel - libvirt List Archives

[libvirt] [PATCH 0/2] build: fix CI builds
by Jim Fehlig 20 Sep '18

20 Sep '18

Fix CI builds broken by commit 5bdcef13. While doing so I noticed more unused variables, which are removed by patch1. Jim Fehlig (2): build: remove unused variables from virt-driver-libxl.m4 libxl: fallback to lib probe if pkgconfig file not found m4/virt-driver-libxl.m4 | 23 ++++++++++++++++++++--- 1 file changed, 20 insertions(+), 3 deletions(-) -- 2.18.0

4 8

[libvirt] what determines the type of virtual NIC in libvirt (veth or tap)?
by Frank Wang 20 Sep '18

20 Sep '18

Hi All, I'd like to know what determines the type of virtual NIC in libvirt ? There was a problem I encountered, the vm can't get the mirror traffic while the virtual port in openvswitch is veth, If the virtual port is tun, the everything is fine. [root@allinone-247 ~]# ethtool -i vn62f717800 driver: veth version: 1.0 firmware-version: expansion-rom-version: bus-info: supports-statistics: yes supports-test: no supports-eeprom-access: no supports-register-dump: no supports-priv-flags: no I uninstall the veth kernel module, then restart vm, then the vm can receive the mirror traffic [root@allinone-247 ~]# ethtool -i vn62f717800 driver: tun version: 1.6 firmware-version: expansion-rom-version: bus-info: tap supports-statistics: no supports-test: no supports-eeprom-access: no supports-register-dump: no supports-priv-flags: no so, I'm wondering if there are configuration options which can be configured to use which type of virtual NIC(tap or veth). Any comments would be appreciate.

2 1

[libvirt] [PATCH] virhostdev: Fix PCI devices are still attatched to stub driver bug
by Wu Zongyong 20 Sep '18

20 Sep '18

Currently, PCI devices will not be rebound to host drivers but attached to the stub driver when: 1) use libvirt to start a virtual machine with PCI devices assigned, then stop libvirtd process and shutdown the virtual machine. Finally, PCI devices are still bound to the stub driver instead of host drivers after libvirt start again. 2) use libvirt to shutdown a virtual machine wtih PCI devices assigned, then stop libvirtd process before libvirt try to rebind PCI devices to host drivers. Finally, PCI devices are still bound to the stub driver after libvirt start again. Notice that the comment on the top of virPCIDeviceDetach as follows: activeDevs should be a list of all PCI devices currently in use by a domain.inactiveDevs is a list of all PCI devices that libvirt has detached from the host driver + attached to the stub driver, but hasn't yet assigned to a domain. It's not reasonable that libvirt filter out devices that are either not active or not used by the current domain and driver. For devices belong to domains that has been shutdown before libvirt start, we should put them into inactiveDevs if then meet the condition that PCI devices have been detached from the host driver + attached to the stub driver. Moreover, we set orignal states of PCI devices when build PCI devices list in virHostdevGetPCIHostDeviceList if states has been set in struct virDomainHostdevDefPtr. Signed-off-by: Wu Zongyong <cordius.wu(a)huawei.com> --- src/util/virhostdev.c | 20 ++++++++++++++++++-- 1 file changed, 18 insertions(+), 2 deletions(-) diff --git a/src/util/virhostdev.c b/src/util/virhostdev.c index ca79c37..ecf95e3 100644 --- a/src/util/virhostdev.c +++ b/src/util/virhostdev.c @@ -235,6 +235,7 @@ virHostdevGetPCIHostDeviceList(virDomainHostdevDefPtr *hostdevs, int nhostdevs) for (i = 0; i < nhostdevs; i++) { virDomainHostdevDefPtr hostdev = hostdevs[i]; virDomainHostdevSubsysPCIPtr pcisrc = &hostdev->source.subsys.u.pci; + virDomainHostdevOrigStatesPtr origstates = &hostdev->origstates; virPCIDevicePtr pci; if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS) @@ -262,6 +263,10 @@ virHostdevGetPCIHostDeviceList(virDomainHostdevDefPtr *hostdevs, int nhostdevs) virPCIDeviceSetStubDriver(pci, VIR_PCI_STUB_DRIVER_XEN); else virPCIDeviceSetStubDriver(pci, VIR_PCI_STUB_DRIVER_KVM); + + virPCIDeviceSetUnbindFromStub(pci, origstates->states.pci.unbind_from_stub); + virPCIDeviceSetRemoveSlot(pci, origstates->states.pci.remove_slot); + virPCIDeviceSetReprobe(pci, origstates->states.pci.reprobe); } return pcidevs; @@ -1008,8 +1013,19 @@ virHostdevReAttachPCIDevices(virHostdevManagerPtr mgr, continue; } } else { - virPCIDeviceListDel(pcidevs, pci); - continue; + int stub = virPCIDeviceGetStubDriver(pci); + if (stub > VIR_PCI_STUB_DRIVER_NONE && + stub < VIR_PCI_STUB_DRIVER_LAST) { + /* The device is bound to a known stub driver: add a copy + * to the inactive list */ + VIR_DEBUG("Adding PCI device %s to inactive list", + virPCIDeviceGetName(pci)); + if (virPCIDeviceListAddCopy(mgr->inactivePCIHostdevs, pci) < 0) { + VIR_ERROR(_("Failed to add PCI device %s to the inactive list"), + virGetLastErrorMessage()); + virResetLastError(); + } + } } i++; -- 1.9.1

3 3

[libvirt] [jenkins-ci PATCH] Forcefully remove build/ directory for autotools builds
by Andrea Bolognani 20 Sep '18

20 Sep '18

Normally this shouldn't be needed, because 'git clean -xdf' is executed after updating the git repository and before starting the build; however, some RPM builds (notably libvirt's) internally use git to apply patches, and if one of those fails it will leave a git repository inside of build/ behind, which 'git clean' dutifully refuses to remove and whose presence will in turn cause 'mkdir build', and thus the entire build, to fail. Solve the issue by forcefully removing build/ ourselves before starting an autotools build. Signed-off-by: Andrea Bolognani <abologna(a)redhat.com> --- guests/playbooks/build/jobs/autotools-build-job.yml | 1 + jobs/autotools.yaml | 1 + 2 files changed, 2 insertions(+) diff --git a/guests/playbooks/build/jobs/autotools-build-job.yml b/guests/playbooks/build/jobs/autotools-build-job.yml index bf7a616..01b6ff5 100644 --- a/guests/playbooks/build/jobs/autotools-build-job.yml +++ b/guests/playbooks/build/jobs/autotools-build-job.yml @@ -6,6 +6,7 @@ {{ global_env }} {{ local_env }} + rm -rf build mkdir build cd build ../autogen.sh --prefix=$VIRT_PREFIX {{ autogen_args }} diff --git a/jobs/autotools.yaml b/jobs/autotools.yaml index 8956855..9349b8c 100644 --- a/jobs/autotools.yaml +++ b/jobs/autotools.yaml @@ -41,6 +41,7 @@ - shell: | {global_env} {local_env} + rm -rf build mkdir build cd build ../autogen.sh --prefix=$VIRT_PREFIX {autogen_args} -- 2.17.1

2 1

[libvirt] [PATCH] Remove ignore_value or void from unlink calls
by John Ferlan 20 Sep '18

20 Sep '18

There seems to be no need to add the ignore_value wrapper or caste with (void) to the unlink() calls, so let's just remove them. I assume at one point in time Coverity complained. So, let's just be consistent - those that care to check the return status can and those that don't can just have the naked unlink. Signed-off-by: John Ferlan <jferlan(a)redhat.com> --- As a result of Michal's review comment on my recent storage driver error path series... src/conf/virsecretobj.c | 2 +- src/nwfilter/nwfilter_dhcpsnoop.c | 2 +- src/qemu/qemu_capabilities.c | 2 +- src/storage/storage_driver.c | 4 ++-- src/util/virfilecache.c | 2 +- src/util/virnetdev.c | 2 +- 6 files changed, 7 insertions(+), 7 deletions(-) diff --git a/src/conf/virsecretobj.c b/src/conf/virsecretobj.c index 2ff2998c15..78911c0908 100644 --- a/src/conf/virsecretobj.c +++ b/src/conf/virsecretobj.c @@ -666,7 +666,7 @@ virSecretObjDeleteData(virSecretObjPtr obj) { /* The configFile will already be removed, so secret won't be * loaded again if this fails */ - (void)unlink(obj->base64File); + unlink(obj->base64File); } diff --git a/src/nwfilter/nwfilter_dhcpsnoop.c b/src/nwfilter/nwfilter_dhcpsnoop.c index 2330ba0479..6d114557c7 100644 --- a/src/nwfilter/nwfilter_dhcpsnoop.c +++ b/src/nwfilter/nwfilter_dhcpsnoop.c @@ -1923,7 +1923,7 @@ virNWFilterSnoopLeaseFileRefresh(void) if (rename(TMPLEASEFILE, LEASEFILE) < 0) { virReportSystemError(errno, _("rename(\"%s\", \"%s\")"), TMPLEASEFILE, LEASEFILE); - ignore_value(unlink(TMPLEASEFILE)); + unlink(TMPLEASEFILE); } virAtomicIntSet(&virNWFilterSnoopState.wLeases, 0); diff --git a/src/qemu/qemu_capabilities.c b/src/qemu/qemu_capabilities.c index 04c2adcfb5..394202942f 100644 --- a/src/qemu/qemu_capabilities.c +++ b/src/qemu/qemu_capabilities.c @@ -4304,7 +4304,7 @@ virQEMUCapsInitQMPCommandAbort(virQEMUCapsInitQMPCommandPtr cmd) cmd->cmd = NULL; if (cmd->monpath) - ignore_value(unlink(cmd->monpath)); + unlink(cmd->monpath); virDomainObjEndAPI(&cmd->vm); diff --git a/src/storage/storage_driver.c b/src/storage/storage_driver.c index bf43d77c8b..7c22c43584 100644 --- a/src/storage/storage_driver.c +++ b/src/storage/storage_driver.c @@ -84,7 +84,7 @@ storagePoolRefreshFailCleanup(virStorageBackendPtr backend, virErrorPtr orig_err = virSaveLastError(); if (stateFile) - ignore_value(unlink(stateFile)); + unlink(stateFile); if (backend->stopPool) backend->stopPool(obj); if (orig_err) { @@ -142,7 +142,7 @@ storagePoolUpdateStateCallback(virStoragePoolObjPtr obj, virReportError(VIR_ERR_INTERNAL_ERROR, _("Failed to initialize storage pool '%s': %s"), def->name, virGetLastErrorMessage()); - ignore_value(unlink(stateFile)); + unlink(stateFile); active = false; } diff --git a/src/util/virfilecache.c b/src/util/virfilecache.c index 2927c68358..15c0d99fd9 100644 --- a/src/util/virfilecache.c +++ b/src/util/virfilecache.c @@ -161,7 +161,7 @@ virFileCacheLoad(virFileCachePtr cache, if (!cache->handlers.isValid(loadData, cache->priv)) { VIR_DEBUG("Outdated cached capabilities '%s' for '%s'", file, name); - ignore_value(unlink(file)); + unlink(file); ret = 0; goto cleanup; } diff --git a/src/util/virnetdev.c b/src/util/virnetdev.c index 39c91313de..e94960c9da 100644 --- a/src/util/virnetdev.c +++ b/src/util/virnetdev.c @@ -2161,7 +2161,7 @@ virNetDevReadNetConfig(const char *linkdev, int vf, } /* we won't need the file again */ - ignore_value(unlink(filePath)); + unlink(filePath); ret = 0; cleanup: -- 2.17.1

2 1

[libvirt] [PATCH 0/5] Cleanup some storage driver paths
by John Ferlan 20 Sep '18

20 Sep '18

See the patches for details - everything leads to the last one. John Ferlan (5): storage: Clean up stateFile if refreshPool fails storage: Clean up storagePoolUpdateStateCallback processing storage: Create error label path for storagePoolCreateXML storage: Introduce storagePoolRefreshFailCleanup storage: Save error during refresh failure processing src/storage/storage_driver.c | 73 ++++++++++++++++++++---------------- 1 file changed, 40 insertions(+), 33 deletions(-) -- 2.17.1

2 9

[libvirt] [libvirt PATCH v6 00/15] Finish the conversion to virConfGetValue* functions libvirt
by Fabiano Fidêncio 20 Sep '18

20 Sep '18

This patchset finishes the conversion to virConfGetValue* functions, started by Daniel Berrange a few months ago. Please, mind that although we could make virConfGetValue* functions more generic in order to support numbers and booleans as strings, that doesn't seem the safest path to take. The side-effect of this is that we will have to live with some specific code doing that as part of vmx and xen_common. Once this patchset gets merged, https://wiki.libvirt.org/page/BiteSizedTasks#Finish_conversion_to_virConfGe… can be removed. - Changes since v1: All the "values" from virConfGetValueString() are freed - Changes since v2: All comments from Ján Tomko have been addressed; A few leaks were (possibly) found and they're addressed in the last patch of the series; - Changes since v3: All comments from Ján Tomko have been addressed in order to lower the non-whitespace changes in the first patch; - Changes since v4: All comments from Ján Tomko have been addressed (hopefully, they actually were this time :-)); - Changes since v5: All comments from John Ferlan have been addressed, including splitting the patches per "function" touched. Some new patches were also needed due to a behaviour change introduced by the "xen_common: Change xenParsePCIList to using virConfGetValueStringList" patch. Fabiano Fidêncio (15): xen_common: Change xenConfigCopyStringInternal to using virConfGetValueString xen_common: Change xenConfigGetUUID to using virConfGetValueString xen_common: Change xenConfigGetString to using virConfGetValueString xen_xl: Adapt xenParseCmdline due to changes in xenConfigGetString xen_xl: Adapt xenParseXLOS due to changes in xenConfigGetString xen_xl: adapt xenParseXLCPUID due to changes in xenConfigGetString xen_common: Adapt xenParseEventsActions due to changes in xenConfigGetString xen_common: Adapt xenParseCPUFeatures due to changes in xenConfigGetString xen_common: Adapt xenParseEmulatedDevices due to changes in xenConfigGetString xen_common: Adapt xenParseGeneralMeta due to changes in xenConfigGetString xen_xm: Adapt xenParseXMOS due to changes in xenConfigGetString xen_xm: Adapt xenParseXMInputDevs due to changes in xenConfigGetString xen_common: Change xenParsePCIList to using virConfGetValueStringList xen_common: Change xenParseVfbs to using virConfGetValueStringList xen_common: Change xenParseCharDev to using virConfGetValueStringList src/xenconfig/xen_common.c | 223 +++++++++++++++++++------------------ src/xenconfig/xen_xl.c | 71 +++++++----- src/xenconfig/xen_xm.c | 48 ++++---- 3 files changed, 186 insertions(+), 156 deletions(-) -- 2.17.1

2 22

[libvirt] [PATCH 0/2] Tiny system header cleanup
by Erik Skultety 20 Sep '18

20 Sep '18

2 6

[libvirt] [PATCH v3] qemu: Ignore nwfilter binding instantiation issues during reconnect
by John Ferlan 20 Sep '18

20 Sep '18

https://bugzilla.redhat.com/show_bug.cgi?id=1607202 It's essentially stated in the nwfilterBindingDelete that we will allow the admin to shoot themselves in the foot by deleting the nwfilter binding which then allows them to undefine the nwfilter that is in use for the running guest... However, by allowing this we cause a problem for libvirtd restart reconnect processing which would then try to recreate the missing binding attempting to use the deleted filter resulting in an error and thus shutting the guest down. So rather than keep adding virDomainConfNWFilterInstantiate flags to "ignore" specific error conditions, modify the logic to ignore, but VIR_WARN errors other than ignoreExists. This will at least allow the guest to not shutdown for only nwfilter binding errors that we can now perhaps recover from since we have the binding create/delete capability. Signed-off-by: John Ferlan <jferlan(a)redhat.com> --- v2: https://www.redhat.com/archives/libvir-list/2018-August/msg01567.html Differences to v2. Leave the ignoreExists bool, but just allow and VIR_WARN other errors from virDomainConfNWFilterInstantiate. Continue processing all filters from error point too. src/qemu/qemu_process.c | 24 ++++++++++++++++-------- 1 file changed, 16 insertions(+), 8 deletions(-) diff --git a/src/qemu/qemu_process.c b/src/qemu/qemu_process.c index ab749389ee..61a277f468 100644 --- a/src/qemu/qemu_process.c +++ b/src/qemu/qemu_process.c @@ -3160,20 +3160,29 @@ qemuProcessNotifyNets(virDomainDefPtr def) } } -static int -qemuProcessFiltersInstantiate(virDomainDefPtr def, bool ignoreExists) +/* Attempt to instantiate the filters. Ignore failures because it's + * possible that someone deleted a filter binding and the associated + * filter while the guest was running and we don't want that action + * to cause failure to keep the guest running during the reconnection + * processing. Nor do we necessarily want other failures to do the + * same. We'll just log the error conditions other than of course + * ignoreExists possibility (e.g. the true flag) */ +static void +qemuProcessFiltersInstantiate(virDomainDefPtr def) { size_t i; for (i = 0; i < def->nnets; i++) { virDomainNetDefPtr net = def->nets[i]; if ((net->filter) && (net->ifname)) { - if (virDomainConfNWFilterInstantiate(def->name, def->uuid, net, ignoreExists) < 0) - return 1; + if (virDomainConfNWFilterInstantiate(def->name, def->uuid, net, + true) < 0) { + VIR_WARN("filter '%s' instantiation for '%s' failed '%s'", + net->filter, net->ifname, virGetLastErrorMessage()); + virResetLastError(); + } } } - - return 0; } static int @@ -7892,8 +7901,7 @@ qemuProcessReconnect(void *opaque) qemuProcessNotifyNets(obj->def); - if (qemuProcessFiltersInstantiate(obj->def, true)) - goto error; + qemuProcessFiltersInstantiate(obj->def); if (qemuProcessRefreshDisks(driver, obj, QEMU_ASYNC_JOB_NONE) < 0) goto error; -- 2.17.1

2 4

[libvirt] [PATCH V3 0/2] libxl: drop support for Xen 4.4 and 4.5
by Jim Fehlig 19 Sep '18

19 Sep '18

V2: https://www.redhat.com/archives/libvir-list/2018-September/msg00383.html New in V3: - Change definition of LIBXL_API_VERSION to 0x040500, the API version set when Xen 4.6 was released - Drop the 'true' parameter to LIBVIRT_CHECK_PKG macro so configure fails if libxl is requested but the devel packages are not installed - Remove a now unused variable and a redundant 'if test with_libxl' check - Moved the news file addition to 'Removed features' section Jim Fehlig (2): libxl: drop support for Xen < 4.6 news: Announce dropping support for Xen 4.4 and 4.5 docs/drvxen.html.in | 2 +- docs/news.xml | 11 +++++++++++ m4/virt-driver-libxl.m4 | 27 ++------------------------- src/libxl/libxl_domain.c | 2 +- src/libxl/libxl_driver.c | 2 +- 5 files changed, 16 insertions(+), 28 deletions(-) -- 2.18.0

3 10