March 2021 - Devel - libvirt List Archives

[libvirt PATCH 0/2] glib: Avoid yet another crash due to race in event loop code
by Martin Kletzander 04 Mar '21

04 Mar '21

Similarly to commit 0db4743645b7 we need to fix the other event loop as well. To be completely honest I do not completely understand all the details in the original workaround, so I just hope this is the best way to fix the issue. Hopefully-resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1931331 Martin Kletzander (2): util: Move glib event loop workaround to glibcompat glib: Use safe glib event workaround in other event loops src/qemu/qemu_agent.c | 2 +- src/qemu/qemu_monitor.c | 2 +- src/util/glibcompat.c | 35 +++++++++++++++++++++++++++++++++++ src/util/glibcompat.h | 11 +++++++++++ src/util/vireventglib.c | 30 ------------------------------ src/util/vireventthread.c | 2 +- 6 files changed, 49 insertions(+), 33 deletions(-) -- 2.30.1

2 5

[PATCH 0/4] Refactoring node_device_conf
by Kristina Hanicova 04 Mar '21

04 Mar '21

This refactors file node_device_conf in order to remove VIR_FREE (if possible) and removes redundant variables and labels. Kristina Hanicova (4): node_device_conf: Use g_autofree where possible node_device_conf: Refactoring functions in order to remove VIR_FREE node_device_conf: use g_autoptr in virNodeDevCapPCIDevIommuGroupParseXML() node_device_conf: Remove redundant variables/labels src/conf/node_device_conf.c | 324 +++++++++++++----------------------- 1 file changed, 118 insertions(+), 206 deletions(-) -- 2.29.2

2 6

[libvirt PATCH 0/5] Some fixes in domain_conf.c: virDomainFeaturesDefParse
by Tim Wiederhake 04 Mar '21

04 Mar '21

Patches 1-4 are minor cleanups, patch 5 is the actual one that I want out of the way before continuing down my rabbit hole of the week. Tim Wiederhake (5): domain_conf: Reduce scope of tmp in virDomainFeaturesDefParse domain_conf: Reduce scope of gic_version in virDomainFeaturesDefParse domain_conf: Reduce scope of node in virDomainFeaturesDefParse domain_conf: Add missing break in switch domain_conf: Fix check for hyperv stimer src/conf/domain_conf.c | 36 ++++++++++++++---------------------- 1 file changed, 14 insertions(+), 22 deletions(-) -- 2.26.2

2 7

[libvirt PATCH] cpu_map: Fix spelling of svme-addr-chk feature
by Jiri Denemark 04 Mar '21

04 Mar '21

Commit a208176ca1d9eedf8aa6bf12fde6a7a9579ab549 introduced this feature with an incorrect "svme-addr-check" spelling. Signed-off-by: Jiri Denemark <jdenemar(a)redhat.com> --- src/cpu_map/sync_qemu_i386.py | 2 +- src/cpu_map/x86_EPYC-Milan.xml | 2 +- src/cpu_map/x86_features.xml | 2 +- 3 files changed, 3 insertions(+), 3 deletions(-) diff --git a/src/cpu_map/sync_qemu_i386.py b/src/cpu_map/sync_qemu_i386.py index 3febecdfd1..1a98fa70d7 100755 --- a/src/cpu_map/sync_qemu_i386.py +++ b/src/cpu_map/sync_qemu_i386.py @@ -142,7 +142,7 @@ def translate_feature(name): "CPUID_SS": "ss", "CPUID_SVM_NPT": "npt", "CPUID_SVM_NRIPSAVE": "nrip-save", - "CPUID_SVM_SVME_ADDR_CHK": "svme-addr-check", + "CPUID_SVM_SVME_ADDR_CHK": "svme-addr-chk", "CPUID_TSC": "tsc", "CPUID_VME": "vme", "CPUID_XSAVE_XGETBV1": "xgetbv1", diff --git a/src/cpu_map/x86_EPYC-Milan.xml b/src/cpu_map/x86_EPYC-Milan.xml index 53f0cd6aac..3055e175fa 100644 --- a/src/cpu_map/x86_EPYC-Milan.xml +++ b/src/cpu_map/x86_EPYC-Milan.xml @@ -76,7 +76,7 @@ <feature name='sse4a'/> <feature name='ssse3'/> <feature name='svm'/> - <feature name='svme-addr-check'/> + <feature name='svme-addr-chk'/> <feature name='syscall'/> <feature name='tsc'/> <feature name='umip'/> diff --git a/src/cpu_map/x86_features.xml b/src/cpu_map/x86_features.xml index c43520d08f..e98b84f3ef 100644 --- a/src/cpu_map/x86_features.xml +++ b/src/cpu_map/x86_features.xml @@ -554,7 +554,7 @@ <feature name='pfthreshold'> <cpuid eax_in='0x8000000a' edx='0x00001000'/> </feature> - <feature name='svme-addr-check'> + <feature name='svme-addr-chk'> <cpuid eax_in='0x8000000a' edx='0x10000000'/> </feature> -- 2.30.0

2 1

[PATCH 0/4] use g_autofree and g_autoptr
by Kristina Hanicova 03 Mar '21

03 Mar '21

This uses g_autofree (if possible) in files: vsh-table, netdev_bandwidth_conf and g_autoptr instead of virNetDevBandwidthFree where possible. Kristina Hanicova (4): vsh-table: Use g_autofree where possible netdev_bandwidth_conf: Use g_autofree where possible Use g_autoptr instead of virNetDevBandwidthFree where possible netdev_bandwidth_conf: Remove redundant variables/labels src/conf/netdev_bandwidth_conf.c | 67 ++++++++++++-------------------- src/network/bridge_driver.c | 6 +-- src/qemu/qemu_driver.c | 11 ++---- src/test/test_driver.c | 3 +- src/util/virnetdevbandwidth.h | 2 + tests/virnetdevbandwidthtest.c | 3 +- tools/vsh-table.c | 16 +++----- 7 files changed, 40 insertions(+), 68 deletions(-) -- 2.29.2

4 11

[PATCH 0/5] Clean up various places using strcpy()
by Peter Krempa 03 Mar '21

03 Mar '21

This series + https://listman.redhat.com/archives/libvir-list/2021-March/msg00081.html removes use of strcpy(). Peter Krempa (5): virIndexToDiskName: Make 'idx' unsigned and remove check virIndexToDiskName: Use g_string_prepend(_c) to improve readability remote_daemon_dispatch: Replace g_new + strcpy with g_strdup virHostCPUGetStatsLinux: Avoid 'strcpy' commandhelper: printCwd: Print result directly instead of copying it src/remote/remote_daemon_dispatch.c | 9 +++------ src/util/virhostcpu.c | 6 +++--- src/util/virutil.c | 29 +++++++---------------------- src/util/virutil.h | 2 +- src/vmx/vmx.c | 12 ------------ src/vz/vz_sdk.c | 3 --- tests/commandhelper.c | 7 ++++--- 7 files changed, 18 insertions(+), 50 deletions(-) -- 2.29.2

3 7

[libvirt PATCH 0/3] cgroup cpu.shares cleanup
by Pavel Hrdina 03 Mar '21

03 Mar '21

After the conversion to using systemd DBus APIs the behavior of changed for values outside of valid range. This series fixes the issue by enforcing the range on libvirt side instead of the previous magic conversion done by kernel. Pavel Hrdina (3): vircgroup: enforce range limit for cpu.shares cgroup: use virCgroupSetCpuShares instead of virCgroupSetupCpuShares vircgroup: drop unused function virCgroupSetupCpuShares docs/formatdomain.rst | 3 ++- docs/manpages/virsh.rst | 5 +---- src/conf/domain_validate.c | 10 ++++++++++ src/libvirt_private.syms | 1 - src/lxc/lxc_cgroup.c | 4 +--- src/lxc/lxc_driver.c | 6 ++---- src/qemu/qemu_cgroup.c | 21 +-------------------- src/qemu/qemu_driver.c | 7 +++---- src/util/vircgroup.c | 20 -------------------- src/util/vircgroup.h | 4 ++-- src/util/vircgroupv1.c | 10 ++++++++++ src/util/vircgroupv2.c | 10 ++++++++++ 12 files changed, 42 insertions(+), 59 deletions(-) -- 2.29.2

2 4

[PATCH] gitpublish: Drop custom prefix
by Andrea Bolognani 03 Mar '21

03 Mar '21

Now that all projects except the main library are using GitLab merge requests, all the patches that end up on libvir-list are by definition intended for libvirt itself rather than one of the subprojects, and so it's no longer necessary to include additional information in the mail subject. Signed-off-by: Andrea Bolognani <abologna(a)redhat.com> --- .gitpublish | 1 - 1 file changed, 1 deletion(-) diff --git a/.gitpublish b/.gitpublish index d37bf4174a..857f0d552c 100644 --- a/.gitpublish +++ b/.gitpublish @@ -1,4 +1,3 @@ [gitpublishprofile "default"] base = master to = libvir-list(a)redhat.com -prefix = libvirt PATCH -- 2.26.2

2 3

[PATCH] build: Require glib >= 2.56
by Jim Fehlig 03 Mar '21

03 Mar '21

There are a few uses of g_autoslist in the qemu driver and likely more will come throughout the codebase in the future. g_autoslist first appeared in glib 2.56, so bump the minimum version. https://developer.gnome.org/glib/stable/glib-Miscellaneous-Macros.html#g-au… Signed-off-by: Jim Fehlig <jfehlig(a)suse.com> --- libvirt.spec.in | 2 +- meson.build | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/libvirt.spec.in b/libvirt.spec.in index 29b476184d..8d8b900fbb 100644 --- a/libvirt.spec.in +++ b/libvirt.spec.in @@ -278,7 +278,7 @@ BuildRequires: systemd-units %if %{with_libxl} BuildRequires: xen-devel %endif -BuildRequires: glib2-devel >= 2.48 +BuildRequires: glib2-devel >= 2.56 BuildRequires: libxml2-devel BuildRequires: libxslt BuildRequires: readline-devel diff --git a/meson.build b/meson.build index daa1475c99..c225df42d4 100644 --- a/meson.build +++ b/meson.build @@ -1038,7 +1038,7 @@ if fuse_dep.found() conf.set('WITH_FUSE', 1) endif -glib_version = '2.48.0' +glib_version = '2.56.0' glib_dep = dependency('glib-2.0', version: '>=' + glib_version) gobject_dep = dependency('gobject-2.0', version: '>=' + glib_version) if host_machine.system() == 'windows' -- 2.29.2

5 7

Some confusion about lsilogic controller
by xingchaochao 03 Mar '21

03 Mar '21

Hello, I have been confused by such a phenomenon recently. Libvirt is the master branch , and the VM is centos8.2(kernel is 4.18.0-193.el8.aarch64). When I hot-plug the scsi disk for a virtual machine without a virtio-scsi controller, libvirt will automatically generate an lsilogic controller for the scsi disk. <disk type='file' device='disk'> <driver name='qemu' type='qcow2' cache='none' io='native'/> <source file='/Images/xcc/tmp.img'/> <backingStore/> <target dev='sdt' bus='scsi'/> <address type='drive' controller='0' bus='0' target='0' unit='0'/> </disk> linux-upcHIq:/Images/xcc # virsh list Id Name State ---------------------- 12 g1 running linux-upcHIq:/Images/xcc # virsh attach-device g1 disk.xml Device attached successfully linux-upcHIq:/Images/xcc # virsh dumpxml g1 | grep scsi <target dev='sdt' bus='scsi'/> <alias name='scsi0-0-0'/> <controller type='scsi' index='0' model='lsilogic'> <alias name='scsi0'/> But this scsi disk cannot be found through the lsblk command inside the virtual machine. [root@localhost ~]# lsblk NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT vda 252:0 0 20G 0 disk ├─vda1 252:1 0 600M 0 part /boot/efi ├─vda2 252:2 0 1G 0 part /boot └─vda3 252:3 0 18.4G 0 part ├─cl-root 253:0 0 16.4G 0 lvm / └─cl-swap 253:1 0 2G 0 lvm [SWAP] After hot unplugging the scsi disk, I performed the hot unplug operation of the lsilogic controller. libvirt shows "Device detached successfully", but in fact, the lsilogic controller is not removed from the live XML and persistent XML. Through "virsh dumpxml vmname" and "virsh edit vmname", I can see <controller type='scsi' index='0' model='lsilogic'> is always there. linux-upcHIq:/Images/xcc # virsh detach-device g1 disk.xml Device detached successfully linux-upcHIq:/Images/xcc # virsh dumpxml g1 | grep scsi <controller type='scsi' index='0' model='lsilogic'> <alias name='scsi0'/> linux-upcHIq:/Images/xcc # linux-upcHIq:/Images/xcc # cat lsi.xml <controller type='scsi' index='0' model='lsilogic'> <alias name='scsi0'/> <address type='pci' domain='0x0000' bus='0x03' slot='0x05' function='0x0'/> </controller> linux-upcHIq:/Images/xcc # virsh detach-device g1 lsi.xml Device detached successfully linux-upcHIq:/Images/xcc # virsh dumpxml g1 | grep scsi <controller type='scsi' index='0' model='lsilogic'> <alias name='scsi0'/> I am confused, why libvirt chooses to generate an lsilogic controller for the scsi disk when there is no scsi controller, instead of directly reporting an error and exiting the hot plug operation. After all, the scsi disk based on the lsilogic controller is not perceived inside the virtual machine, and lsilogic will remain in the XML file of the virtual machine.

2 2