- Devel - libvirt List Archives

[libvirt] [PATCH v5 0/2] Avoid slash characteres for 'domrename' command.
by Julio Faracco 03 May '18

03 May '18

The commits enhace the 'domrename' command of QEMU and Test drivers to avoid slashes chars inside the new domain name. Julio Faracco (2): qemu: avoid slash characters to the new domain name. test: avoid slash characters to the new domain name. src/qemu/qemu_driver.c | 6 ++++++ src/test/test_driver.c | 6 ++++++ 2 files changed, 12 insertions(+) -- 2.17.0

2 6

[libvirt] [PATCH 0/6] Complete cleanup of domain object usage
by John Ferlan 03 May '18

03 May '18

This should be the "last time" (famous last words) needing to alter this processing. I hope to get more than one pair of eyes looking at the series. I've looked at it long enough to feel comfortable with the changes, but I also know I probably have looked at it so long that I could have missed something! The first 3 patches set up the necessary infrastructure. Patches 4 & 5 are just something I found in libxl while doing this - they could be extracted and considered separately if need be; however, since I was thinking about @vm referencing - it just seemed to fit. Since I altered @vm it only seemed right to do the same for @conn. While it could be said that neither condition could happen - I figured better safe than sorry and I think it follows how this type of logic has been handled elsewhere in qemu. Patch 6 is where all the magic happens. I tried a few different ways to rework and split the Add/Remove processing, but either way I was left with rather ugly patches that required touching the same places, so I capitulated and combined them since there really is a delicate balance between those paths that have added a virObjectRef after the virDomainObjListAdd and those that just use virObjectUnlock on the returned ListAdd objects. Converting the latter to use Ref caused more "busy work" and equally large patches. The busy work was mosting around the processing required during ListAdd if something happened which resulted in ListRemove needing to be called. The "general change" is that instead of having ListAdd return an object with 2 refs and 1 lock it will now return an object with 3 refs and 1 lock. With that returned object, some drivers would perform an ObjectRef while others did not. For those that did not, they would use ObjectUnlock when done with the object so that when leaving whatever Create routine there was there would be at least 2 Refs on the object. The change here is to have them all be consistent. What follows is a "general description" of what was done for each. For bhyve, openvz, test, uml, and vmware (e.g. domain drivers that do not use virObjectRef on the returned virDomainObjListAdd object): Create/Add processing: -> Use EndAPI instead of ObjectUnlock -> Remove setting vm = NULL if the ListRemove was called to allow EndAPI processing to perform last Unref and Unlock NB: ListRemove would remove 2 refs and unlock, so it "worked", but would "consume" the returned object Undefine/Destroy processing: -> Alter ListRemove and ObjectLock to just be ListRemove allowing the EndAPI processing from a FindBy* to perform last Unref and Unlock For libxl, lxc, qemu, and vz (e.g. domain drivers that use virObjectRef on the returned virDomainObjListAdd object): Create/Add processing: -> Remove the ObjectRef -> Remove the ObjectLock when needing to call ListRemove to allow EndAPI processing to perform the last Unref and Unlock NB: ListRemove no longer Unlock's the returned object Undefine/Destroy processing: -> Alter ListRemove and ObjectLock to just be ListRemove allowing the EndAPI processing from a FindBy* to perform last Unref and Unlock NB: For the AutoDestroy type logic, no change is required since in the long run the object was left with the correct number of refs after create processing ran. The issue is more the direct mgmt of object during Add/Remove rather than mgmt of object once defined. John Ferlan (6): conf: Split FindBy{UUID|Name} into locked helpers conf: Use virDomainObjListFindBy*Locked for virDomainObjListAdd conf: Move and use virDomainObjListRemoveLocked libxl: Add refcnt for args->vm during migration libxl: Add refcnt for args->conn during migration conf: Clean up object referencing for Add and Remove src/bhyve/bhyve_driver.c | 21 ++---- src/conf/virdomainobjlist.c | 165 ++++++++++++++++++++++++++++---------------- src/libxl/libxl_driver.c | 64 ++++------------- src/libxl/libxl_migration.c | 41 ++++------- src/lxc/lxc_driver.c | 21 ++---- src/lxc/lxc_process.c | 17 +++-- src/openvz/openvz_conf.c | 2 +- src/openvz/openvz_driver.c | 20 ++---- src/qemu/qemu_domain.c | 17 +---- src/qemu/qemu_driver.c | 6 +- src/qemu/qemu_migration.c | 3 +- src/test/test_driver.c | 56 +++++---------- src/uml/uml_driver.c | 37 +++------- src/vmware/vmware_conf.c | 3 +- src/vmware/vmware_driver.c | 17 ++--- src/vz/vz_driver.c | 1 - src/vz/vz_sdk.c | 3 - 17 files changed, 192 insertions(+), 302 deletions(-) -- 2.13.6

3 20

[libvirt] [PATCH 0/3] Ditch qemu-img help scraping
by Ján Tomko 03 May '18

03 May '18

Ján Tomko (3): tests: delete most qemu-img test cases assuming FMT_OPTIONS tests: assume FMT_COMPAT for qemu-img tests storage: remove qemu-img help scraping src/storage/storage_util.c | 73 ++------------------- src/storage/storage_util.h | 1 - .../qcow2-convert-nobacking.argv | 2 - .../storagevolxml2argvdata/qcow2-from-logical.argv | 2 - .../qcow2-nobacking-convert-prealloc.argv | 2 - .../qcow2-nobacking-prealloc.argv | 2 - .../qcow2-nocapacity-convert-prealloc.argv | 2 +- tests/storagevolxml2argvdata/qcow2-nocapacity.argv | 2 +- tests/storagevolxml2argvdata/qcow2.argv | 2 - tests/storagevolxml2argvtest.c | 74 ++++++---------------- 10 files changed, 24 insertions(+), 138 deletions(-) delete mode 100644 tests/storagevolxml2argvdata/qcow2-convert-nobacking.argv delete mode 100644 tests/storagevolxml2argvdata/qcow2-from-logical.argv delete mode 100644 tests/storagevolxml2argvdata/qcow2-nobacking-convert-prealloc.argv delete mode 100644 tests/storagevolxml2argvdata/qcow2-nobacking-prealloc.argv delete mode 100644 tests/storagevolxml2argvdata/qcow2.argv -- 2.13.6

2 9

[libvirt] [PATCH] configure: remove bogus LIBVIRT_GETTEXT macro reference
by Daniel P. Berrangé 03 May '18

03 May '18

The LIBVIRT_GETTEXT macro was an artifact of patch development and was later renamed to LIBVIRT_CHECK_NLS. This cruft causes configure to print out ./configure: line 75084: LIBVIRT_GETTEXT: command not found but fortunately this is non-fatal Reported-by: Martin Kletzander <mkletzan(a)redhat.com> Signed-off-by: Daniel P. Berrangé <berrange(a)redhat.com> --- Pushed as trivial fix configure.ac | 2 -- 1 file changed, 2 deletions(-) diff --git a/configure.ac b/configure.ac index 419a1aa2f5..cc005ea9e8 100644 --- a/configure.ac +++ b/configure.ac @@ -734,8 +734,6 @@ if test "$enable_test_locking" = "yes"; then fi AM_CONDITIONAL([WITH_CIL],[test "$enable_test_locking" = "yes"]) -LIBVIRT_GETTEXT - dnl Cygwin, MinGW and MSVC checks LIBVIRT_WIN_CHECK_COMMON LIBVIRT_WIN_CHECK_CYGWIN -- 2.14.3

1 0

[libvirt] [PATCH 00/14] nwfilter: refactor the driver to make it independent of virt drivers
by Daniel P. Berrangé 03 May '18

03 May '18

Today the nwfilter driver is entangled with the virt drivers in both directions. At various times when rebuilding filters nwfilter will call out to the virt driver to iterate over running guest's NICs. This has caused very complicated lock ordering rules to be required. If we are to split the virt drivers out into separate daemons we need to get rid of this coupling since we don't want the separate daemons calling each other, as that risks deadlock if all of the RPC workers are busy. The obvious way to solve this is to have the nwfilter driver remember all the filters it has active, avoiding the need to iterate over running guests. NB, these patches are all ready for review, but the last patch really should not be merged at this time. I need to do more work to be able to serialize the filter state to disk, so the nwfilter driver can keep track of it across daemon restarts. All except the last patch should be ok to merge though. Daniel P. Berrangé (14): nwfilter: remove pointless virNWFilterHashTable struct nwfilter: remove methods that are trivial wrappers for virHash APIs nwfilter: remove virNWFilterHashTable typedefs entirely nwfilter: make virNWFilterIPAddrLearnReq type private nwfilter: remove obsolete code related to firewalld nwfilter: fix leaking of filter parameters upon error nwfilter: introduce virNWFilterBinding to decouple from virDomainNet nwfilter: pass vm name in when instantiating filters nwfilter: convert the gentech driver code to use virNWFilterBinding nwfilter: convert IP address learning code to virNWFilterBinding nwfilter: convert DHCP address snooping code to virNWFilterBinding nwfilter: report an error if nic needs filtering by no driver is present nwfilter: keep track of active filter bindings nwfilter: remove virt driver callback layer for rebuilding filters src/conf/domain_conf.c | 8 +- src/conf/domain_conf.h | 2 +- src/conf/domain_nwfilter.c | 14 +- src/conf/domain_nwfilter.h | 6 +- src/conf/nwfilter_conf.c | 224 ++++++++--------- src/conf/nwfilter_conf.h | 67 ++--- src/conf/nwfilter_ipaddrmap.c | 15 +- src/conf/nwfilter_params.c | 127 +++------- src/conf/nwfilter_params.h | 33 +-- src/conf/virnwfilterobj.c | 4 +- src/conf/virnwfilterobj.h | 4 +- src/libvirt_private.syms | 8 +- src/lxc/lxc_driver.c | 28 --- src/lxc/lxc_process.c | 2 +- src/nwfilter/nwfilter_dhcpsnoop.c | 153 +++++------- src/nwfilter/nwfilter_dhcpsnoop.h | 7 +- src/nwfilter/nwfilter_driver.c | 97 +++++--- src/nwfilter/nwfilter_driver.h | 2 - src/nwfilter/nwfilter_gentech_driver.c | 432 ++++++++++++++++++--------------- src/nwfilter/nwfilter_gentech_driver.h | 28 +-- src/nwfilter/nwfilter_learnipaddr.c | 113 ++++----- src/nwfilter/nwfilter_learnipaddr.h | 25 +- src/nwfilter/nwfilter_tech_driver.h | 2 +- src/qemu/qemu_driver.c | 25 -- src/qemu/qemu_hotplug.c | 6 +- src/qemu/qemu_interface.c | 4 +- src/qemu/qemu_process.c | 2 +- src/uml/uml_conf.c | 2 +- src/uml/uml_driver.c | 29 --- tests/nwfilterxml2firewalltest.c | 36 +-- 30 files changed, 642 insertions(+), 863 deletions(-) -- 2.14.3

4 30

[libvirt] [PATCH v3 0/3] Support tpm-crb TPM interface for QEMU
by Stefan Berger 03 May '18

03 May '18

This patch series adds support for the recently added tpm-crb TPM interface for QEMU (2.12). Stefan Stefan Berger (3): tpm: Enable TPM CRB interface in the domain XML and test it qemu: Extend the capabilities with tpm-crb device tests: add test case for tpm-crb QEMU device command line docs/formatdomain.html.in | 6 +++- docs/schemas/domaincommon.rng | 5 +++- src/conf/domain_conf.c | 5 ++-- src/conf/domain_conf.h | 1 + src/qemu/qemu_capabilities.c | 5 ++++ src/qemu/qemu_capabilities.h | 1 + src/qemu/qemu_command.c | 16 +++++++++- tests/qemucapabilitiesdata/caps_2.12.0.x86_64.xml | 1 + tests/qemuxml2argvdata/tpm-passthrough-crb.args | 26 ++++++++++++++++ tests/qemuxml2argvdata/tpm-passthrough-crb.xml | 32 ++++++++++++++++++++ tests/qemuxml2argvtest.c | 2 ++ tests/qemuxml2xmloutdata/tpm-passthrough-crb.xml | 36 +++++++++++++++++++++++ tests/qemuxml2xmltest.c | 1 + 13 files changed, 131 insertions(+), 6 deletions(-) create mode 100644 tests/qemuxml2argvdata/tpm-passthrough-crb.args create mode 100644 tests/qemuxml2argvdata/tpm-passthrough-crb.xml create mode 100644 tests/qemuxml2xmloutdata/tpm-passthrough-crb.xml -- 2.5.5

2 10

[libvirt] [perl PATCH] maint: Drop autobuild.sh
by Andrea Bolognani 03 May '18

03 May '18

The Test-AutoBuild project, that this script is supposed to be used with, hasn't seen any activity in ~7 years; these days, libvirt-perl CI builds are happening on the Jenkins-based CentOS CI environment under the libvirt umbrella[1], and in that context the script is not used at all. [1] https://ci.centos.org/view/libvirt/ Signed-off-by: Andrea Bolognani <abologna(a)redhat.com> --- autobuild.sh | 74 ------------------------------------------------------------ 1 file changed, 74 deletions(-) delete mode 100755 autobuild.sh diff --git a/autobuild.sh b/autobuild.sh deleted file mode 100755 index 8a6d102..0000000 --- a/autobuild.sh +++ /dev/null @@ -1,74 +0,0 @@ -#!/bin/sh -# -# This script is used to Test::AutoBuild (http://www.autobuild.org) -# to perform automated builds of the Sys-Virt module - -NAME=Sys-Virt - -set -e - -test -n "$1" && RESULTS=$1 || RESULTS=results.log -: ${AUTOBUILD_INSTALL_ROOT=$HOME/builder} - -make -k realclean ||: -rm -rf MANIFEST blib pm_to_blib - -export TEST_MAINTAINER=1 - -perl Makefile.PL PREFIX=$AUTOBUILD_INSTALL_ROOT - -rm -f MANIFEST - -# Build the RPM. -make -make manifest - -if [ -z "$USE_COVER" ]; then - perl -MDevel::Cover -e '' 1>/dev/null 2>&1 && USE_COVER=1 || USE_COVER=0 -fi - -if [ -z "$SKIP_TESTS" -o "$SKIP_TESTS" = "0" ]; then - if [ "$USE_COVER" = "1" ]; then - cover -delete - export HARNESS_PERL_SWITCHES=-MDevel::Cover - fi - - # set -o pipefail is a bashism; this use of exec is the POSIX alternative - exec 3>&1 - st=$( - exec 4>&1 >&3 - { make test 2>&1 3>&- 4>&-; echo $? >&4; } | tee "$RESULTS" - ) - exec 3>&- - test "$st" = 0 - - if [ "$USE_COVER" = "1" ]; then - cover - mkdir blib/coverage - cp -a cover_db/*.html cover_db/*.css blib/coverage - mv blib/coverage/coverage.html blib/coverage/index.html - fi -fi - -make install - -rm -f $NAME-*.tar.gz -make dist - -if [ -f /usr/bin/rpmbuild ]; then - if [ -n "$AUTOBUILD_COUNTER" ]; then - EXTRA_RELEASE=".auto$AUTOBUILD_COUNTER" - else - NOW=`date +"%s"` - EXTRA_RELEASE=".$USER$NOW" - fi - rpmbuild --nodeps -ta --define "extra_release $EXTRA_RELEASE" --clean $NAME-*.tar.gz -fi - -# Skip debian pkg for now -exit 0 - -if [ -f /usr/bin/fakeroot ]; then - fakeroot debian/rules clean - fakeroot debian/rules DESTDIR=$HOME/packages/debian binary -fi -- 2.14.3

2 1

[libvirt] [python PATCH] maint: Drop autobuild.sh
by Andrea Bolognani 03 May '18

03 May '18

The Test-AutoBuild project, that this script is supposed to be used with, hasn't seen any activity in ~7 years; these days, libvirt-python CI builds are happening on the Jenkins-based CentOS CI environment under the libvirt umbrella[1], and in that context the script is not used at all. [1] https://ci.centos.org/view/libvirt/ Signed-off-by: Andrea Bolognani <abologna(a)redhat.com> --- autobuild.sh | 30 ------------------------------ 1 file changed, 30 deletions(-) delete mode 100755 autobuild.sh diff --git a/autobuild.sh b/autobuild.sh deleted file mode 100755 index 5dffe96..0000000 --- a/autobuild.sh +++ /dev/null @@ -1,30 +0,0 @@ -#!/bin/sh - -set -ve - -: ${AUTOBUILD_INSTALL_ROOT="$HOME/builder"} - -rm -rf MANIFEST dist build - - -python2 setup.py sdist - -python2 setup.py build -python2 setup.py test -python2 setup.py install --root="$AUTOBUILD_INSTALL_ROOT" - -if test -f /usr/bin/python3 ; then - python3 setup.py build - python3 setup.py test - python3 setup.py install --root="$AUTOBUILD_INSTALL_ROOT" -fi - -type -p /usr/bin/rpmbuild > /dev/null 2>&1 || exit 0 - -if [ -n "$AUTOBUILD_COUNTER" ]; then - EXTRA_RELEASE=".auto$AUTOBUILD_COUNTER" -else - NOW=`date +"%s"` - EXTRA_RELEASE=".$USER$NOW" -fi -rpmbuild --nodeps --define "extra_release $EXTRA_RELEASE" -ta --clean dist/*.tar.gz -- 2.14.3

2 1

[libvirt] [jenkins-ci PATCH 0/4] Introduce Fedora 28; drop Fedora 26
by Andrea Bolognani 03 May '18

03 May '18

It's that time of the year again :) Andrea Bolognani (4): guests: Add libvirt-fedora-28 projects: Add libvirt-fedora-28 projects: Remove libvirt-fedora-26 guests: Remove libvirt-fedora-26 .../host_vars/{libvirt-fedora-26 => libvirt-fedora-28}/install.yml | 2 +- guests/host_vars/{libvirt-fedora-26 => libvirt-fedora-28}/main.yml | 0 guests/inventory | 2 +- jobs/defaults.yaml | 4 ++-- projects/libvirt-dbus.yaml | 4 ++-- projects/libvirt-sandbox.yaml | 4 ++-- projects/libvirt-tck.yaml | 4 ++-- projects/libvirt.yaml | 4 ++-- projects/virt-manager.yaml | 6 +++--- 9 files changed, 15 insertions(+), 15 deletions(-) rename guests/host_vars/{libvirt-fedora-26 => libvirt-fedora-28}/install.yml (74%) rename guests/host_vars/{libvirt-fedora-26 => libvirt-fedora-28}/main.yml (100%) -- 2.14.3

2 6

[libvirt] [PATCH v3 0/9] tests: Fix 'make check' on FreeBSD
by Andrea Bolognani 03 May '18

03 May '18

Changes from [v2]: * introduce a wrapper around canonicalize_file_name() instead of moving to realpath(); * fix more usages of mode_t with va_arg(). Changes from [v1]: * drop the virnetsockettest changes, since the relationship between the errors seen on FreeBSD and IPv4 clients connecting to IPv6 sockets has turned out to be just a red herring; * fix the mocking instead of dropping test cases that don't work without it. [v2] https://www.redhat.com/archives/libvir-list/2018-April/msg02685.html [v1] https://www.redhat.com/archives/libvir-list/2018-April/msg02611.html Andrea Bolognani (9): util: Introduce virFileCanonicalizePath() tests: Mock virFileCanonicalizePath() all: Use virFileCanonicalizePath() instead of canonicalize_file_name() tests: Stop mocking canonicalize_file_name() syntax-check: Prohibit canonicalize_file_name() tests: Fix mode_t usage with va_arg() tests: Build virpcimock on non-Linux tests: Extend NUMA mocking tests: Drop DO_TEST_LINUX() cfg.mk | 10 ++++ src/libvirt_private.syms | 1 + src/storage/storage_backend_fs.c | 2 +- src/util/virfile.c | 15 +++++- src/util/virfile.h | 1 + src/util/virpci.c | 2 +- tests/nssmock.c | 2 +- tests/qemuxml2argvmock.c | 38 ++++++++++---- tests/qemuxml2argvtest.c | 46 ++++++----------- tests/vircgroupmock.c | 2 +- tests/virpcimock.c | 104 ++++++++++++++++++++++----------------- tests/virstoragetest.c | 10 ++-- tests/virtestmock.c | 8 +-- 13 files changed, 141 insertions(+), 100 deletions(-) -- 2.14.3

2 20