libvirt releases
================
# v5.0.0 (2019-01-15)
* New features
- Xen: Add support for openvswitch
The libxl driver now supports virtual interfaces that connect to an
openvswitch bridge, including interfaces with VLAN tagging and trunking
configuration.
- qemu: Report whether KVM nesting is available
Running nested KVM guests requires specific configuration steps to be
performed on the host; libvirt will now report in the host capabilities
whether KVM nesting support is available.
* Removed features
- Drop UML driver
The UML driver was unmaintained and not tested for quite some time now.
Worse, there is a bug that causes it to deadlock on some very basic
operations (e.g. dumping domain XML). These facts make us believe no
one uses it.
* Improvements
- qemu: Add support for ARMv6l guests
- Support more NVDIMM configuration options
Introduce more configuration options. For the source element, add the
'alignsize' and 'pmem' subelements. For the target element, add the
'readonly' subelement.
- cpu: Add support for "stibp" x86_64 feature
Add cpu flag stibp (Single Thread Indirect Branch Predictors) to
prevent indirect branch predictions from being controlled by the
sibling Hyperthread.
- libxl: Handle external domain destroy
Historically, if a domain was destroyed using xl rather than through
libvirt APIs, libvirt would not be aware of the fact and keep
considering it as running. This is no longer the case.
- Start selecting the first available DRI device for OpenGL operations
If OpenGL support is needed (either with SPICE gl enabled or with
egl-headless), libvirt is now able to pick the first available DRI
device for the job. At the same time, this improvement is also a bugfix
as it prevents permission-related issues with regards to our mount
namespaces and the default DRI render node's permissions which would
normally prevent QEMU from accessing such a device.
- qemu: Add support for postcopy-requests migration statistics
The virDomainJobInfo can get number page requests received from the
destination host during post-copy migration.
* Bug fixes
- lxc: Don't forbid interfaces with type=direct
Such interfaces are supported by lxc and should be allowed.
- qemu: Fully clean up RNG devices on detach
Some RNG device types, such as those using EGD, might need extra clean
up on the host in addition to removing the guest-side device.
# v4.10.0 (2018-12-03)
* New features
- qemu: Add Hyper-V PV IPI and Enlightened VMCS support
The QEMU driver now has support for Hyper-V PV IPI and Enlightened VMCS
for Windows and Hyper-V guests.
- qemu: Added support for PCI devices on S390
PCI addresses can now include the new zpci element which contains uid
(user-defined identifier) and fid (PCI function identifier) attributes
and makes the corresponding devices usable by S390 guests.
- Support changing IOThread polling parameters for a live guest
Introduced virDomainSetIOThreadParams which allows dynamically setting
the IOThread polling parameters used by QEMU to manage the thread
polling interval and the algorithm for growth or shrink of the polling
time. The values only affect a running guest with IOThreads. The
guest's IOThread polling values can be viewed via the domain
statistics.
- Xen: Add support for PVH
The libxl driver now supports Xen's PVH virtual machine type. PVH
machines are enabled with the new "xenpvh" OS type, e.g.
<os><type>xenpvh</type></os>
- qemu: Added support for CMT (Cache Monitoring Technology)
Introduced cache monitoring using the monitor element in cachetune for
vCPU threads. Added interfaces to get and display the cache utilization
statistics through the command 'virsh domstats' via the
virConnectGetAllDomainStats API.
- qemu: Add support for nested HV for pSeries guests
Nested HV support makes it possible to run nested (L2) guests with
minimal performance penalty when compared to regular (L1) guests on
ppc64 hardware.
* Improvements
* Bug fixes
- Xen: Handle soft reset shutdown event
The pvops Linux kernel uses soft reset to handle the crash machine
operation. The libxl driver now supports the soft reset shutdown event,
allowing proper crash handling of pvops-based HVM domains.
# v4.9.0 (2018-11-04)
* New features
- util: Add cgroup v2 support
cgroup v2 support has been implemented in libvirt, with both "unified"
(v2 only) and "hybrid" (v2 + v1) setups being usable; existing "legacy"
(v1 only) setups will keep working.
- qemu: Add vfio AP support
The QEMU driver now has support to passthrough adjunct processors into
QEMU guests on S390.
* Improvements
- rpc: Make 'genprotocol' output reproducible
This is another step towards making libvirt builds fully reproducible.
* Bug fixes
- security: Fix permissions for UNIX sockets
Since 4.5.0, libvirt is using FD passing to hand sockets over to QEMU,
which in theory removes the need for them to be accessible by the user
under which the QEMU process is running; however, other processes such
as vdsm need to access the sockets as well, which means adjusting
permissions is still necessary.
- cpu_map: Add Icelake model definitions
These CPU models will be available in the upcoming 3.1.0 QEMU release.
- util: Properly parse URIs with missing trailing slash
Some storage URIs were not parsed correctly, in which case libvirt
ended up emitting XML that it would then refuse to parse back.
# v4.8.0 (2018-10-01)
* New features
- Xen: Support PM Suspend and Wakeup
The libxl driver now supports the virDomainPMSuspendForDuration and
virDomainPMWakeup APIs.
* Removed features
- Xen: Drop support for Xen 4.4 and 4.5
Xen 4.4 and 4.5 are no longer supported by the Xen community. Drop
support for these older versions and require Xen >= 4.6.
- nwfilter: Disallow binding creation in session mode
Ensure that a filter binding creation is not attempted in session mode
and generates a proper error message.
* Improvements
- qemu: Retrieve guest hostname through QEMU Guest Agent command
QEMU is now able to retrieve the guest hostname using a new QEMU-GA
command called 'guest-get-host-name'. Virsh users can execute
'domhostname' for QEMU driver for domains configured to use the Guest
Agent.
- virsh: Implement vsh-table in virsh and virsh-admin
The new API fixes problems with table-alignment, making the tables more
readable and deals with unicode.
* Bug fixes
- storage: Allow inputvol to be encrypted
When creating a storage volume based on another volume, the base input
volume is allowed to be encrypted.
- virsh: Require explicit --domain for domxml-to-native
The --domain option for domxml-to-native virsh command has always been
documented as required, but commit v4.3.0-127-gd86531daf2 accidentally
made it optional.
- lxc_monitor: Avoid AB / BA lock race
A deadlock situation could occur when autostarting a LXC domain 'guest'
due to two threads attempting to take opposing locks while holding
opposing locks (AB BA problem).
# v4.7.0 (2018-09-03)
* New features
- storage: add storage pool iscsi-direct
Introduce a new storage pool backend that uses libiscsi instead of
iscsiadm. It support basic pool operations: checkPool and refreshPool.
- Add support for MBA (Memory Bandwidth Allocation technology)
Domain vCPU threads can now have allocated some parts of host memory
bandwidth by using the memorytune element in cputune.
- qemu: Add support for RISC-V guests
riscv32 and riscv64 guest architectures are now supported.
* Improvements
- qemu: Add ccw support for vhost-vsock
Support the vhost-vsock-ccw device on S390.
- qemu: Make default machine type independent of QEMU
We can't control whether or not QEMU will change its default machine
type in the future, or whether downstream distributions will decide to
compile out some machine types, so our only option to provide a
predictable behavior is taking care of the default ourselves;
management applications and users are encouraged to explicitly pick a
machine type when creating new guests.
- apparmor: Various improvements
Rules have been added to deal with a number of scenarios that didn't
work correctly.
* Bug fixes
- esx: Truncate CPU model name
Some CPU model names are too long to be stored into the corresponding
property, and should be explicitly truncated to avoid unexpected
behavior in users of the virNodeGetInfo() API such as virsh nodeinfo.
- utils: Remove arbitrary limit on socket_id/core_id
Both values were assumed to be smaller than 4096, but in fact they are
entirely hardware-dependent and there have been reports of machines
presenting much bigger values, preventing libvirt from working
correctly; all such limits have now been removed.
# v4.6.0 (2018-08-06)
* New features
- qemu: Implement the HTM pSeries feature
Users can now decide whether HTM (Hardware Transactional Memory)
support should be available to the guest.
- qemu: Enable VNC console for mediated devices
Host devices now support a new atribute 'display' which can be used to
turn on frame buffer rendering on a vgpu mediated device instead of on
an emulated GPU, like QXL.
* Improvements
- qemu: Introduce a new video model of type 'none'
Introduce a new video model type that disables the automatic addition
of a video device to domains with 'graphics' specified in their XML.
This can be useful with GPU mediated devices which can serve as the
only rendering devices within the guest.
- virsh: Add --alias to attach-disk and attach-interface commands
Add option --alias to set customized device alias name when using
attach-disk or attach-interface commands.
- virsh: Support usb and sata address to attach-disk
Usb or sata address could be used when attach-disk with --address. For
example, use usb address as usb:<bus>.<port>, use sata address as
<controller>.<bus>.<unit>.
* Bug fixes
# v4.5.0 (2018-07-02)
* New features
- qemu: Provide TPM emulator support
Support QEMU's TPM emulator based on swtpm. Each QEMU guest gets its
own virtual TPM.
- bhyve: Support specifying guest CPU topology
Bhyve's guest CPU topology could be specified using the <cpu><topology
../></cpu> element.
- qemu: Add support for extended TSEG size
Support specifying extended TSEG size for SMM in QEMU.
- qemu: Add support for SEV guests
SEV (Secure Encrypted Virtualization) is a feature available on AMD
CPUs that encrypts the guest memory and makes it inaccessible even to
the host OS.
* Removed features
- Remove support for qcow/default encrypted volumes
Disallow using a qcow encrypted volume for the guest and disallow
creation of the qcow or default encrypted volume from the storage
driver. Support for qcow encrypted volumes has been phasing out since
QEMU 2.3 and by QEMU 2.9 creation of a qcow encrypted volume via
qemu-img required usage of secret objects, but that support was never
added to libvirt.
- Make GnuTLS mandatory
Building without GnuTLS is no longer possible.
- qemu: Remove allow_disk_format_probing configuration option
The option represented a security risk when used with malicious disk
images, so users were recommended against enabling it; with this
release, it's been removed altogether.
* Improvements
- capabilities: Provide info about host IOMMU support
Capabilities XML now provide information about host IOMMU support.
- virsh: Add --all to domblkinfo command
Alter the domblkinfo command to add the option --all in order to
display the size details of each domain block device from one command
in a output table.
- qemu: Allow concurrent access to monitor and guest agent
Historically libvirt prevented concurrent accesses to the qemu monitor
and the guest agent. Therefore two independent calls (one querying the
monitor and the other querying guest agent) would serialize which hurts
performance. The code was reworked to allow two independent calls run
at the same time.
- qemu: Allow configuring the page size for HPT pSeries guests
For HPT pSeries guests, the size of the host pages used to back guest
memory and the usable guest page sizes are connected; the new setting
can be used to request that a certain page size is available in the
guest.
- Add support to use an raw input volume for encryption
It is now possible to provide a raw input volume as input for to
generate a luks encrypted volume via either virsh vol-create-from or
virStorageVolCreateXMLFrom.
- qemu: Add support for vsock hot (un)plug and cold (un)plug
- qemu: Add support for NBD over TLS
NBD volumes can now be accessed securely.
- qemu: Implement FD passing for Unix sockets
Instead of having QEMU open the socket and then connecting to it, which
is inherently racy, starting with QEMU 2.12 we can open the socket
ourselves and pass it to QEMU, avoiding race conditions.
- virsh: Introduce --nowait option for domstat command
When this option is specified, virsh will try to fetch the guest stats
but abort instead of stalling if they can't be retrieved right away.
* Bug fixes
- qemu: Fix a potential libvirtd crash on VM reconnect
Initialization of the driver worker pool needs to come before libvirtd
trying to reconnect to all machines, since one of the QEMU processes
migh have already emitted events which need to be handled prior to us
getting to the worker pool initialization.
- qemu: Fix domain resume after failed migration
Recent versions of QEMU activate block devices before the guest CPU has
been started, which makes it impossible to roll back a failed
migration. Use the late-block-activate migration capability if
supported to avoid the issue.
- vmx: Permit guests to have an odd number of vCPUs
An odd number of vCPUs greater than 1 was forbidden in the past, but
current versions of ESXi have lifted that restriction.
# v4.4.0 (2018-06-04)
* New features
- bhyve: Support locking guest memory
Bhyve's guest memory may be wired using the
<memoryBacking><locked/></memoryBacking> element.
- qemu: Provide VFIO channel I/O passthrough support
Support passthrough devices that use channel I/O based mechanism in a
QEMU virtual machine.
- qemu: Add support for migration of VMs with non-shared storage over TLS
It's now possible to use the VIR_MIGRATE_TLS flag together with
VIR_MIGRATE_NON_SHARED_DISK. The connection is then secured using the
TLS environment which is setup for the migration connection.
- Add support for VM Generation ID
The VM Generatation ID exposes a 128-bit, cryptographically random,
integer value identifier, referred to as a Globally Unique Identifier
(GUID) to the guest in order to notify the guest operating system when
the virtual machine is executed with a different configuration. Add a
new domain XML processing and a domain capabilities feature.
- Introduce virDomainDetachDeviceAlias
This new API enables users to detach device using only its alias.
- Introduce new virConnectCompareHypervisorCPU and
virConnectBaselineHypervisorCPU APIs
Unlike the old virConnectCompareCPU and virConnectBaselineCPU APIs,
both new APIs consider capabilities of a specific hypervisor.
- Introduce SCSI persistent reservations support
The QEMU driver gained support for qemu-pr-helper which enables guests
to issue SCSI commands for persistent reservation.
- qemu: Implement multiple screen support for virDomainScreenshot
While the virDomainScreenshot API supported multihead video cards, the
implementation was missing. But now that QEMU implemented it libvirt
has done as well.
- qemu: add support for vhost-vsock-device
A new vsock device was introduced, allowing communication between the
guest and the host via the AF_VSOCK family.
* Improvements
- qemu: Add suport for OpenGL rendering with SDL
Domains using SDL as a graphics backend will now be able to use OpenGL
accelerated rendering.
- qemu: Add support for 'output' audio codec
Support QEMU's 'hda-output' codec advertising only a line-out for ich6
and ich9 sound devices.
- virsh: Enhance event name completion
Implement event name completion for some commands (e.g. event,
secret-event, pool-event and nodedev-event)
* Bug fixes
# v4.3.0 (2018-05-02)
* New features
- qemu: Add support for the pcie-to-pci-bridge controller
Pure PCIe guests such as x86_64/q35 and aarch64/virt will now add this
controller when traditional PCI devices are in use.
- Xen: Support setting CPU features for host-passthrough model
The CPU model presented to Xen HVM domains is equivalent to libvirt's
host-passthrough model, although individual features can be enabled and
disabled via the cpuid setting. The libvirt libxl driver now supports
enabling and disabling individual features of the host-passthrough CPU
model.
* Removed features
- Xen: Drop the legacy xend-based driver
The xm/xend toolstack was deprecated in Xen 4.2 and removed from the
Xen sources in the 4.5 development cycle. The libvirt driver based on
xend is now removed from the libvirt sources.
* Improvements
- qemu: Support hot plug and hot unplug of mediated devices
Libvirt now allows mediated devices to be hot plugged and hot unplugged
from a guest rather than reporting an error that this isn't supported.
In fact, kernel has been supporting this since 4.10.
* Bug fixes
- Improve handling of device mapper targets
When starting a domain with a disk backed by a device mapper volume
libvirt also needs to allow the storage backing the device mapper in
CGroups. In the past kernel did not care, but starting from 4.16
CGroups are consulted on each access to the device mapper target.
# v4.2.0 (2018-04-01)
* New features
- Support building with Python 3
Python is required to build libvirt, and up until now only Python 2
could be used as an interpreter. All scripts used during build have now
been made compatible with Python 3, which means both major releases of
the language are fully supported.
- qemu: Provide ccw address support for graphics and input devices
Support the virtio-gpu-ccw device as a video device and
virtio-{keyboard, mouse, tablet}-ccw devices as input devices on S390.
* Improvements
- qemu: Add logging of guest crash information on S390
On S390, when the guest crashes and QEMU exposes the guest crash
information, log the relevant data to the domain log file.
- qemu: use arp table of host to get the IP address of guests
Find IP address of a VM by arp table on hosts. If someone customizing
IP address inside VM, it will be helpful.
- Xen: Remove hard-coded scheduler weight
The libxl driver was accidentally hard-coding the per-domain scheduler
weight to 1000, silently ignoring any user-provided <shares> in
<cputune>. The driver now honors <shares>, and defers setting a default
value to Xen. Note that the Xen default is 256, so any domains started
after this improvement will have one fourth the shares of previously
started domains. If all domains must have equal CPU shares,
administrators must manually set the weight of previously started
domains to 256, or restart them.
* Bug fixes
- qemu: TLS migration now enforces use of TLS for the NBD connection
When the VIR_MIGRATE_TLS flag was used with the migration API libvirt
did not ensure that the NBD connection was using TLS as well. The code
now rejects such migration as the TLS transport for NBD is not ready
yet, but prevents a false sense of security that TLS would be used. The
support TLS for NBD will be added soon.
# v4.1.0 (2018-03-05)
* New features
- Added support for CAT (Cache allocation Technology)
Domain vCPU threads can now have allocated some parts of host cache
using the cachetune element in cputune.
- Allow opening secondary drivers
Up until now it was possible to connect to only hypervisor drivers
(e.g. qemu:///system, lxc:///, vbox:///system, and so on). The internal
drivers (like network driver, node device driver, etc.) were hidden
from users and users could use them only indirectly. Starting with this
release new connection URIs are accepted. For instance
network:///system, storage:///system and so on.
- virtlogd, virtlockd: Add support for admin protocol
These two daemons now support admin protocol through which some admin
info can be gathered or some configuration tweaked on the fly.
* Improvements
- virsh: Enhance bash completion
Implement more bash completions so that basic libvirt objects can be
auto-completed (e.g. networks, interfaces, NWFilters, and so on).
- qemu: Use VIR_ERR_DEVICE_MISSING for various hotplug/detach messages
- qemu: Allow showing the dump progress for memory only dump
Alter the QEMU dump-guest-memory command processing to check for and
allow asynchronous completion which then allows for the virsh dump
--memory-only --verbose command to display percent completion data.
- conf: add support for setting Chassis SMBIOS data fields
- libxl: add support for setting clock offset and adjustment
- Make port allocator global
Up until now each driver had their own port allocator module. This
meant that info on port usage was not shared. Starting with this
release, the port allocator module is made global and therefore drivers
allocate ports from global pool.
- Fixed some compiler warnings that appear with GCC 8
* Bug fixes
- qemu: Check for unsafe migration more thoroughly
If a domain disk is stored on local filesystem (e.g. ext4) but is not
being migrated it is very likely that domain is not able to run on
destination. Regardless of share/cache mode.
- qemu: Fix updating device with boot order
Starting with 3.7.0 release updating any device with boot order would
fail with 'boot order X is already used by another device' while in
fact it was the very same device.
- virlog: determine the hostname on startup CVE-2018-6764
At later point it might not be possible or even safe to use
getaddrinfo(). It can in turn result in a load of NSS module which can
even be loaded from unsage guest filesystem and thus escape the
confinment of its container.
- qemu: Rework vCPU statistics fetching
Fetching vCPU statistics was very expensive because it lead to waking
up vCPU threads in QEMU and thus it degraded performance. The code was
reworked so that fetching statistics does not wake up halted vCPUs.
- qemu: unlink memory backing file on domain shutdown
Depending on the filesystem where domain memory is stored, some files
might have been left behind. This is not a problem on hugetlbfs, but it
is a problem on regular filesystems like ext4.
- qemu: Fix shutting down domains in parallel
If multiple domains were being shut down in parallel, libvirtd might
have deadlocked.
- nodedev: Update PCI mdev capabilities dynamically
PCI devices may have other nested capabilities, like SRIOV and mdev
which depend on the device being plugged into the native vendor driver.
However, in case such a device is directly assigned to a guest using
VFIO driver, the device will naturally lose these capabilities and
libvirt needs to reflect that.
# v4.0.0 (2018-01-19)
* New features
- tools: Provide bash completion support
Both virsh and virt-admin now implement basic bash completion support.
- qemu: Refresh capabilities on host microcode update
A microcode update can cause the CPUID bits to change; therefore, the
capabilities cache should be rebuilt when such an update is detected on
the host.
- lxc: Set hostname based on container name
* Improvements
- CPU frequency reporting improvements
The CPU frequency will now be reported by virsh nodeinfo and other
tools for s390 hosts; at the same time; CPU frequency has been disabled
on aarch64 hosts because there's no way to detect it reliably.
- libxl: Mark domain0 as persistent
- Xen: Add support for multiple IP addresses on interface devices
- qemu: Add support for hot unplugging redirdev device
* Bug fixes
- qemu: Enforce vCPU hotplug granularity constraints
QEMU 2.7 and newer don't allow guests to start unless the initial vCPUs
count is a multiple of the vCPU hotplug granularity, so validate it and
report an error if needed.
# v3.10.0 (2017-12-04)
* New features
- conf: Support defining distances between virtual NUMA cells
A NUMA hardware architecture supports the notion of distances between
NUMA cells. This can now be specified using the <distances> element
within the NUMA cell configuration. Drivers which support this include
Xen and QEMU.
- Xen: Support defining vNUMA topology
Xen now supports defining a virtual NUMA topology for VMs, including
specifying distances between NUMA cells.
- qemu: Add the ability to configure HPT resizing for pSeries guests
The user can now decide whether HPT (Hash Page Table) resizing should
be enabled, disabled or required instead of leaving it up to hypervisor
defaults and negotiation between the guest and the host.
- qemu: Add vmcoreinfo feature
Starting with QEMU 2.11, the guest can save kernel debug details when
this feature is enabled and the kernel supports it. It is useful to
process kernel dump with KASLR enabled, and also provides various
kernel details to crash tools.
- conf: Move the auth and encryption definitions to disk source
Allow parsing and formatting of the auth and encryption sub-elements to
be a child of the source element. This will allow adding an auth
sub-element to a backingStore or mirror elements as a means to track
specific authentication and/or encryption needs.
* Improvements
- vbox: Add VirtualBox 5.2 support
- vbox: Add support for configuring storage controllers
The VirtualBox driver now supports the <controller> element in the
domain XML for configuring storage controllers in VBOX VMs.
Additionally, libvirt's domain XML schema was updated to allow optional
model attribute for <controller type='ide'> which is used by the VBOX
driver to set the IDE controller model to be one of 'piix4', 'piix4'
(default), or 'ich6'. Finally, with this change dumpxml generates
<controller> elements that correspond to current VBOX VM storage
controller configuration.
- vbox: Add support for attaching empty removable disks
The VirutalBox driver now supports adding CD-ROM and floppy disk
devices that do not have the disk source specified. Previously such
devices were silently ignored.
- vbox: Add support for attaching SAS storage controllers
In VirtualBox, SCSI and SAS are distinct controller types whereas
libvirt does not make such distinction. Therefore, the VBOX driver was
updated to allow attaching SAS controllers via <controller type='scsi'
model='lsisas1068'> element. If there are both SCSI and SAS controllers
present in the VBOX VM, the domain XML can associate the disk device
using the <address> element with the controller attribute, and
optionally, set the port via unit attribute.
- qemu: Generate predictable paths for qemu memory backends
In some cases management applications need to know paths passed to
memory-backend-file objects upfront. Libvirt now generates predictable
paths so applications can prepare the files if they need to do so.
- Shareable disks work properly with recent qemu
Recent qemu versions added image locking to avoid potential corruption
of disk images. This broke shareable disks with libvirt since the
feature was turned on by default in qemu. Libvirt now enables sharing
of those disks in qemu so that the image locking is not applied in that
case. Additionally libvirt now checks that shareable disks have
supported format (raw) to avoid metadata corruption.
- Improve serial console behavior on non-x86 architectures
ppc64, aarch64 and s390x guests were treating the <serial> and
<console> elements differently from x86, in some cases presenting
misleading information to the user. The behavior is now consistent
across all architectures and the information reported is always
accurate.
* Bug fixes
- vbox: Do not ignore failures to attach disk devices when defining
The define now fails and reports an error if any of the controller or
disk devices specified in the domain XML fail to attach to the
VirtualBox VM.
- vbox: Fix dumpxml to always output disk devices
The VirtualBox driver was ignoring any disk devices in dumpxml output
if there was a SAS storage controller attached to the VM.
- vbox: Fix dumpxml to always generate valid domain XML
When a VirtualBox VM has multiple disks attached, each to a different
storage controller that uses 'sd' prefix for block device names e.g.
one disk attached to SATA and one to SCSI controller, it no longer
generates XML where both would have 'sda' device name assigned. Instead
it properly assigns 'sda' and 'sdb' to those disks in the order of
appearance.
- Securely pass iSCSI authentication data
Rather than supplying the authentication data as part of the iSCSI URL
for a disk or host device, utilize the encrypted secret object to
securely pass the authentication data.
# v3.9.0 (2017-11-02)
* New features
- Add capability to allow hot (un)plug of a domain watchdog device
- Allow users to set device aliases
Users can set aliases to domain devices and thus identify them easily.
- qemu: Support multiqueue for virtio-blk
Multiqueue support for virtio-blk has been available in QEMU ever since
2.7.0, and now libvirt guests can enable it.
- Add virDomainSetLifecycleAction API
Provided a new API to allow dynamic guest lifecycle control for guest
reactions to poweroff, restart, or crash type events related to the
domain XML on_poweroff, on_reboot, and on_crash elements. The virsh
set-lifecycle-action command was created to control the actions.
- qemu: Allow cold(un)plugging and hot(un)plugging input devices
- net: Implement QoS for vhostuser
* Improvements
- Allow a logical volume to be create using LUKS
A logical volume may be created using an encryption element using
"luks" format. This does require a previously created secret to store
the passphrase used to encrypt the volume Adding the volume to a domain
can then either provide the secret or allow the consumer in the guest
to provide the passphrase in order to decrypt the volume.
- net: Ignore auto-generated MAC address when detaching an interface
If the MAC address has not been specified by the user, libvirt will try
and fill in the gaps by generating one; however, for some error paths
that led to some confusing error messages, so when an auto-generated
MAC address is specified the error message will not include the
auto-generated MAC.
- net: Enable MAC address lookup for virDomainInterfaceStats
- apparmor: Several improvements
Changes include permitting access to data about USB devices and dnsmasq
instances, allowing spaces in guest names and many more.
- cpu: Use CPU information obtained from QEMU when possible
Recent QEMU versions can expose information about which CPU models are
available and usable on the host; libvirt will now make use of such
information whenever possible.
- hyperv: Various improvements
The error reported when clients can't connect to Hyper-V has been made
more descriptive, and memory limits for guests are now mapped to more
appropriate libvirt equivalents.
- qemu: Report QEMU error on failed migration
Instead of reporting a generic error, ask QEMU for a more detailed and
thus hopefully more helpful one.
- vbox: Implement autoport for RDP
libvirt will now obtain the (dynamically allocated) RDP port number
from VirtualBox itself, avoiding conflicts between multiple guests
wanting to use RDP at the same time.
- qemu: Allow rotation of small logs
On a host where numerous unique instances are executed per day, it's
quite possible that, even though each of the single log files are
fairly small, collectively the quantity and volume may add tens of
thousands of log files to the /var/log/libvirt/qemu/ directory.
Removing the constraints that log have to be bigger than 100 KiB before
they can be rotated solves the issue.
* Bug fixes
- Fix swapped interface statistics and QoS
Due to internal implementation, reported statistics for some types of
interfaces were swapped (RX appeared in TX and vice versa). Similarly,
QoS was set in reversed way.
- Properly resize local LUKS encrypted volume
Resizing of a local LUKS encrypted volume will now use qemu-img to
resize the volume. This will require configuring a secret for the LUKS
encrypted volume.
- qemu: Reserve PCI addresses for implicit i440fx devices
Failing to do so causes the addresses to be considered usable by
libvirt, which means they could be assigned to more than one device
resulting in the guest failing to start.
- spec: Restart libvirtd only at the end of the upgrade process
Use %posttrans to make sure libvirtd is not restarted before all other
components, such as the library itself and storage / hypervisor
drivers, have already been upgraded.
* Security
- qemu: Ensure TLS clients always verify the server certificate
While it's reasonable to turn off client certificate validation, as
setting it up can be non-trivial, clients should always verify the
server certificate to avoid MITM attacks. However, libvirt was using
the same knob to control both checks, leading to CVE-2017-1000256 /
LSN-2017-0002.
# v3.8.0 (2017-10-04)
* New features
- qemu: Added support for cold-(un)plug of watchdog devices
- qemu: Added support for setting IP address os usernet interfaces
- qemu: Added support for Veritas Hyperscale (VxHS) block devices
- storage: Added new events for pool-build and pool-delete
* Improvements
- qemu: Set DAC permissions properly for spice rendernode
When a rendernode path is set for SPICE GL on qemu:///system, we now
correctly set DAC permissions on the device at VM startup. This is the
last remaining hurdle to let SPICE GL work for qemu:///system without
any external host changes.
- nodedev: Add switchdev offload query to NIC capabilities
Allow querying the NIC interface capabilities for the availability of
switchdev offloading (also known as kernel-forward-plane-offload).
- New CPU models for AMD and Intel
AMD EPYC and Intel Skylake-Server CPU models were added together with
their features
- Improve long waiting when saving a domain
While waiting for a write to disk to be finished, e.g. during save,
even simple operations like virsh list would be blocking due to domain
lock. This is now resolved by unlocking the domain in places where it
is not needed.
* Bug fixes
- Proper units are now used in virsh manpage for dom(mem)stats
Previously the documentation used multiples of 1000, but now it is
fixed to use multiples of 1024.
- qemu: Fix error reporting when disk attachment fails
There was a possibility for the actual error to be overridden or
cleared during the rollback.
- qemu: Fix assignment of graphics ports after daemon restart
This could be seen with newer kernels that have bug regarding
SO_REUSEADDR. After libvirtd was restarted it could assign already used
address to new guests which would make them fail to start. This is
fixed by marking used ports unavailable when reconnecting to running
QEMU domains.
- Fix message decoding which was causing a very strange bug
When parsing an RPC message with file descriptors was interrupted and
had to restart, the offset of the payload was calculated badly causing
strange issues like not being able to find a domain that was not
requested.
# v3.7.0 (2017-09-04)
* New features
- qemu: Add managedsave-edit commands
Using managedsave-dumpxml, managedsave-define and managedsave-edit
commands, now we can dump and edit the XML configuration of domain
which has managedsave image.
- qemu: Add migrate-getmaxdowntime command
Currently, the maximum tolerable downtime for a domain being migrated
is write-only from libvirt, via migrate-setmaxdowntime. This implements
a complementary migrate-getmaxdowntime command
- bhyve: Support autoport for VNC ports
It's no longer necessary to explicitly specify VNC port for the bhyve
guests. With the autoport feature it will be allocated automatically.
Please refer to the bhyve driver documentation for examples.
- qemu: Added support for setting heads of virtio GPU
- qemu: Added support to configure reconnect timeout for chardev devices
When you have a TCP or UNIX chardev device and it's connected somewhere
you can configure reconnect timeout if the connection is closed.
* Improvements
- qemu: Report a clear error when dropping a VM during startup
"Failed to load config for domain 'DOMNAME'" is now reported if a VM
config can't be parsed for some reason, and thus provides a clear
indication for users (and devs).
- apparmor: Update for QEMU 2.10 compatibility
Starting with QEMU 2.10, disk images and NVRAM files get automatically
locked to prevent them from being corrupted; however, file locking
needs to be explicitly allowed through virt-aa-helper or AppArmor will
reject the requests and the guest will not be able to run.
- virsh: List Unix sockets in 'domdisplay' output
VNC and SPICE graphics can use Unix sockets instead of TCP/IP sockets
as connection endpoints, but such a configuration was not handled
correctly by virsh domdisplay, causing the respective endpoints to be
missing from the output.
- qemu: Don't check whether offline migration is safe
Since offline migration only copies the guest definition to the
destination host, data corruption is not a concern and the operation
can always be performed safely.
- virt-host-validate: Fix IOMMU detection on ppc64
* Bug fixes
- qemu: Better support for international domain names (with wide
characters)
There were some issues with multi-byte domains getting lost on daemon
restart due to truncation, so the code now handles multi-byte names a
bit better.
- qemu: Support long domain names with namespaces
Domains with extremely long names would fail to start due to temporary
namespace paths being created with the whole name. The path is now
generated with shortened name instead.
- qemu: Tolerate missing emulator binary during libvirtd restart
For some time libvirt required qemu capabilities being present when
parsing VM configs during startup. As a side effect VM configs would
fail to parse and thus vanish, if the emulator binary would be
uninstalled or broken. Libvirt now tolerates when capabilities are
missing during startup.
- qemu: Prevent pSeries guests from disappearing in some situations
pSeries guest would disappear if any of the host devices they were
configured to use was not available during libvirtd startup, which
could easily happen for SR-IOV Virtual Functions. This scenario is now
handled correctly.
- qemu: Honor <on_reboot/> setting
The setting was accepted by the parser, but not actually implemented.
- Fix --verbose option for all daemons
Since v3.0.0, the option had been ignored by all libvirt daemons
(libvirtd, virtlogd and virtlockd); it's now working as intended once
again.
# v3.6.0 (2017-08-02)
* New features
- hyperv: Implement virDomainSetMemory and virDomainSendKey APIs
- qemu: Support multiple PHBs for pSeries guests
pSeries guests can now have multiple PHBs (PCI Host Bridges), which
show up as separate PCI domains in the guest. To create additional
PHBs, simply add PCI controllers with model pci-root to the guest
configuration.
- qemu: Isolate hostdevs on pSeries guests
To enable better error reporting and recovery, unrelated hostdevs will
now be automatically isolated on pSeries guests by placing them on
separate PHBs (PCI Host Bridges).
* Improvements
- qemu: platform serial devices can now use chardev features
QEMU VMs that depend on platform serial devices can now use QEMU's
-chardev option, which enables access to advanced features like log
file configuration. This applies to the default serial devices for arm,
aarch64, and some ppc configurations.
- Require use of GCC 4.4 or Clang compilers
We only ever test libvirt with GCC or Clang (which provides a GCC
compatible compilation environment). Between them, these compilers
cover every supported operating system platform, including Windows.
- qemu: shared disks with directsync cache should be safe for migration
At present shared disks can be migrated with either readonly or
cache=none. But cache=directsync should be safe for migration, because
both cache=directsync and cache=none don't use the host page cache, and
cache=direct write through qemu block layer cache.
- Handle hotplug change on VLAN configuration using OVS
Libvirt was taught to handle VLAN change for running OVS interface.
* Bug fixes
- qemu: Use vCPU 'node-id' property and pass it back to qemu
vCPU properties gathered from query-hotpluggable-cpus need to be passed
back to QEMU. As QEMU did not use the node-id property until now and
libvirt forgot to pass it back properly (it was parsed but not passed
around) we did not honor this.
- Miscellaneous stream fixes
After introducing sparse stream features there were still some known
bugs left. Those are fixed in this release.
- qemu: Miscellaneous domain NS fixes
Libvirt starts qemu domains in separate Linux namespaces for a while
now. However, there were still some bugs lingering. For instance
libvirt did not know how to handle file based bind mounts.
- Various CPU driver improvements
There were some minor bugs when using 'host-model' CPU.
# v3.5.0 (2017-07-04)
* New features
- qemu: Add support for loadparm for a boot device
Add an optional boot parameter 'loadparm' for a boot device. Loadparm
is an 8 byte parameter that, when present, is queried by S390 guests
via sclp or diag 308. Linux guests on S390 use it to select a boot
entry.
- Support controlling how video devices are exposed to the bhyve guests
The vgaconf attribute was added to video's driver element. Possible
values are: on, off, and io. It controls the way how bhyve exposes
video devices to its guests; various guest OSes might require different
settings to boot properly.
- qemu: Add support for live updates of coalesce settings
Users can now use virsh update-device to change the coalesce settings
of an interfaces while the domain is running.
- qemu: Allow VirtIO devices to use vIOMMU
It is now possible to turn on IOTBL for the vIOMMU and have VirtIO
devices use it, provided they have been configured appropriately.
* Improvements
- qemu: block copy job can be used with persistent domains
Until now it was not possible to use block copy with persistent VMs. In
use cases where it's not required to recover the job after VM shutdown,
it's possible to specify VIR_DOMAIN_BLOCK_COPY_TRANSIENT_JOB flag to
start the copy job.
- JSON pseudo-protocol backing store parser supports new format of qemu 2.9
QEMU 2.9 modified a few structures corresponding to the JSON format of
specifying a backing store for a disk image. Libvirt now implements the
new format.
- Capabilities now include info about host's CAT settings
Various information about resource control from the host is gathered
and presented in capabilities if available.
- apparmor: Several improvements
Allow access to Ceph config, EFI firmware on both x86_64 and aarch64,
device tree on ppc64 and more.
- qemu: Support host-model on POWER9 machines
* Bug fixes
- qemu: snapshot: retrieve image metadata from user provided files
Disk images of an external snapshot created with
VIR_DOMAIN_SNAPSHOT_CREATE_REUSE_EXT flag specified would not be
scanned for metadata after recent changes. The metadata is necessary to
allow keeping relative paths between images when doing a block-commit.
- Parse decimal numbers in a locale-independent way
Some locales, such as de_DE and pt_BR, use comma rather than dot to
separate the integer part from the fractional part of a decimal number;
however, several data sources such as the kernel use a
locale-independent representation and need to be treated accordingly.
- Support compilation with newer compiler and libc versions
Several fixes have been included to make compilation with Clang 4.0.0,
GCC 7.1 and glibc >= 2.25.90 possible.
- qemu: Query name for vhost-user interfaces at runtime
This makes it possible to use virsh subcommands such as domiflist and
domifstat on vhost-user interfaces.
- qemu: Set MTU for hotplugged interfaces correctly
When hotplugging a network interface, the MTU was only set on the guest
side. Set it on the host side as well.
- qemu: Forbid updating MTU for interfaces of running guests
The MTU setting can't be modified while the guest is running, so any
attempt to alter it at runtime will now result in an error rather than
being silently ignored.
- qemu: Fix specifying QXL heads with older QEMU releases
Specifying the number of QXL heads was not working correctly for QEMU
releases older than 1.6.
- qemu: Fix migration to older libvirt/QEMU versions
When the guest is started, libvirt updates the CPU definition to
reflect the actual CPU features to enforce ABI. We need to send
original and updated CPU definition in order to support migration to
older libvirt/QEMU versions. Only the updated CPU definition was sent
to destination.
# v3.4.0 (2017-06-02)
* New features
- Improved streams to efficiently transfer sparseness
New extension to virStream was implemented so that
virStorageVolDownload and virStorageVolUpload can preserve file
sparseness.
- I/O APIC type can be specified for QEMU/KVM domains
The ioapic tag was added to domain features, so the type of the I/O
APIC can now be specified (e.g. putting it in userspace for KVM
domains).
- The reason for VM shutdown is reported, if known
QEMU 2.10 will be able to report the reason for shutting down (whether
that was caused by the guest or not), and libvirt is prepared for that
and reports that information in its shutdown event as well, if it is
known.
* Improvements
- Repository now has new README.md file
The new file uses markdown syntax, so it looks better on github and
possibly other web pages, but it has also more useful information. The
old README is now symlink to the new file.
- qemu: Use GICv2 by default for aarch64/virt TCG guests
The emulated GICv3 has some limitations that make it unusable as a
default; use GICv2 until they're sorted out. This change makes it once
again possible to run aarch64/virt guests on a x86_64 host without
having to tweak their configuration.
- Additional capabilities for the node_device module
Introduce two new capabilities to the node_device module. The first is
for CCW devices, most common on the S390 architecture. The second is
for fibre channel-backed SCSI devices and exposes the fc_remote_port
sub-capability to SCSI target devices.
- Node devices now report Mediated device capabilities
Endpoint devices support new mdev capability and their parents now
report the supported types in new mdev_types capability.
- Capabilities now report information about host caches
If supported in the kernel, host capabilities will now list L3 caches.
The code for other levels was added as well, but only L3 caches are
reported currently.
- POWER9 CPU model was added
It is now properly reported in host capabilities.
- libxl: NUMA sibling distances are now reported in host capabilities
- VMDK version 3 files are now properly detected
- Interrupt remapping and Extended interrupt mode for IOMMU devices
These two new features can now be controlled with new <driver
intremap='on/off' eim='on/off'/> tag for iommu devices.
- Graphics in libxl domains now have default addresses
Even though there were default addresses before this change, they were
not saved in the XML. It is now possible to see and control the listen
addresses properly.
- Default USB controllers are now added for devices in libxl domains
Even though they were added automatically when USB device was attached,
they could've been missing in some other cases. The logic is now fixed
so there are always USB controllers, even if there was none of them in
the specified XML.
- Limits for RPC messages were increased
Hitting the RPC limits we have is easier every day, so they were
increased once again and some guessing logic was improved as well. It
is now possible to get more stats than ever using the
virConnectGetAllDomainStats() call and push through even bigger
requests and replies for all APIs.
* Bug fixes
- qemu: Create memory_backing_dir on startup
Libvirt's policy is that directories are created on startup if they
don't exist. We've missed this one.
- PCIe 4.0 cards now report proper link speeds
It could happen that the link speed for PCIe devices was not properly
reported or the nodedev-dumpxml just failed. That was due to mistake in
the field width, but should now work properly.
- qemu: Do not report errors on shutdown
For some users, in some rare cases, it could happen that there was an
error message "internal error: End of file from qemu monitor" in the
logs even though no problem happened. The detection of these false
positives was improved and such errors should not show any more.
- User-specified UNIX socket paths for virtio channels should not be reset
It could happen, in some cases, that libvirt would mistake a
user-specified path for its own generated one and thus remove it from
the XML. The detection of such addresses was improved now.
- Fix address reservation during RNG hot-plug
When error occurred in a specific point in time during the hot-plug of
an RNG device, it could happen that an address was released even though
another device was already using it, making it possible to hot-plug
another device with that address, effectively having duplicated
addresses in the XML.
# v3.3.0 (2017-05-05)
* New features
- net: Add support for coalesce settings
Enabling data batching through these settings can improve network
performance for guests.
- qemu: Add support for guest CPU cache specification
This features allows fine-grained control of the cache behavior of the
guest CPU.
- qemu: Add support for the qemu-xhci USB controller
* Improvements
- hyperv: Support Hyper-V 2012 and newer
Starting with Hyper-V 2012 the API has changed causing the existing
driver to be unable to send and process requests properly. This has
been resolved by adding abstractions to handle the differences and ease
handling such breaks if they happen in the future.
- libxl: Add support for nested HVM domains
Xen has supported nested HVM domains since version 4.4. The libvirt
libxl driver now supports nested HVM domains by specifying the
host-passthrough CPU mode when defining a domain.
- qemu: Implement ACPI support for aarch64 guests
Up until this point, ACPI support was only advertised for x86_64 guests
and disabling it for aarch64 guests was not possible at all.
- vz: Add support for changing the number of vCPUs
- qemu: Automatically choose the best USB controller for guests
The recently introduced qemu-xhci USB controller is the best choice for
both ppc64 and aarch64 guests, so use it by default on those
architectures if available.
- daemon: Increase default task limit for libvirtd
The default number of tasks for the pids cgroup controller is 512,
which libvirtd can quickly bump into when starting lots of guests.
Raise the limit to a more reasonable 32768.
- docs: Include man pages describing key code names and values
- virsh: Report initialization errors
Sometimes virsh might be unable to start: when that happens, report
useful diagnostics instead of failing silently.
* Bug fixes
- nss: Don't require a network restart for libvirt_guest
Previously, the libvirt_guest NSS module would only work properly after
the corresponding network had been restarted; now newly started guests
will be reported correctly right away.
- storage: Remove unavailable transient pools after restart
Solve an issue where transient storage pools would be stuck in an
unmanageable state if the source disappeared and libvirtd was
subsequently restarted.
- storage: Fix capacity value for LUKS encrypted volumes
The 'capacity' value (e.g. guest logical size) for a LUKS volume is
smaller than the 'physical' value of the file in the file system, so we
need to account for that.
- qemu: Fix regression when hyperv/vendor_id feature is used
Guests using the feature would not be started at all; it is now
possible to start them as expected.
- qemu: Do not crash on USB address with no port and invalid bus
- crypto: Always pad data before encrypting it
If this step is not performed, when the data length matches the chunk
size the decryption routines will misinterpret the last byte of data as
the padding length and fail to decode it correctly.
# v3.2.0 (2017-04-02)
* New features
- The virt-host-validate tool now supports bhyve hypervisor
- Introduce NVDIMM memory model
NVDIMM is new type of memory introduced into QEMU 2.6. The idea is that
we have a non-volatile memory module that keeps the data persistent
across domain reboots and offers much faster data accesses. However,
due to a bug in QEMU, this feature is not enabled for QEMUs older than
2.9.0.
- qemu: Introduce support for generic PCIe Root Ports
For new controllers, a generic device (pcie-root-port) will be used by
default instead of the Intel-specific device (ioh3420), provided the
QEMU binary supports it.
- qemu: Add support for checking guest CPU ABI compatibility
When migrating a domain to a different host, restoring a domain from a
file or reverting a snapshot libvirt will make sure the guest CPU QEMU
presents to the guest OS exactly matches the one provided on the source
host (or before the domain's state was saved). This enhanced check may
also be requested when starting a new domain to ensure the virtual CPU
exactly matches the one specified in the XML.
- qemu: Add support to migrate using TLS
Add the ability to migrate QEMU guests using TLS via a new flag
VIR_MIGRATE_TLS or virsh migrate '--tls' option. Requires using at
least QEMU 2.9.0 in order to work properly.
- qemu: add mediated devices framework support
Recent kernel version introduced new mediated device framework, so
provide an initial support of this framework for libvirt, mainly by
introducing a new host device type in the XML.
- qemu: Add support for setting TSC frequency
Setting TSC frequency is required to enable migration for domains with
'invtsc' CPU feature turned on.
- Add support for block device threshold event
When using thin provisioning, management tools need to resize the disk
in certain cases. To avoid having them to poll disk usage this version
introduces an event which will be fired when a given offset of the
storage is written by the hypervisor. Together with the API it allows
registering thresholds for given storage backing volumes and this event
will then notify management if the threshold is exceeded. Currently
only the qemu driver supports this.
- bhyve: Add support for UEFI boot ROM, VNC, and USB tablet
The bhyve driver now supports booting using the UEFI boot ROM, so
non-FreeBSD guests that support UEFI could be booted without using an
external boot loader like grub-bhyve. Video is also supported now,
allowing to connect to guests via VNC and use an USB tablet as an input
device. Please refer to the driver page for domain XML examples.
* Improvements
- qemu: Detect host CPU model by asking QEMU on x86_64
Previously, libvirt detected the host CPU model using CPUID
instruction, which caused libvirt to detect a lot of CPU features that
are not supported by QEMU/KVM. Asking QEMU makes sure we don't start it
with unsupported features.
- perf: Add more perf statistics
Add support to get the count of cpu clock time, task clock time, page
faults, context switches, cpu migrations, minor page faults, major page
faults, alignment faults, emulation faults by applications running on
the platform.
- Write hyperv crash information into vm log
qemu's implementation of the hyperv panic notifier now reports
information about the crash from the guest os. Starting with this
version, libvirt logs the information to the vm log file for possible
debugging.
* Bug fixes
- QEMU: Use adaptive timeout for connecting to monitor
When starting qemu, libvirt waits for qemu to create the monitor socket
which libvirt connects to. Historically, there was sharp 30 second
timeout after which the qemu process was killed. This approach is
suboptimal as in some scenarios with huge amounts of guest RAM it can
take a minute or more for kernel to allocate and zero out pages for
qemu. The timeout is now flexible and computed by libvirt at domain
startup.
- Overwrite (clear) 2 KB instead of just 512 bytes when initializing
logical device
- Describe the logical backend requirements better for pool-create-as
# v3.1.0 (2017-03-03)
* New features
- storage: Add Virtuozzo storage backend storage pool
Add new storage backend to support pool and volume management within
the Virtuozzo Storage environment. Virtuozzo Storage is a highly
available distributed software defined storage with built-in
replication and disaster recovery.
- qemu: Add support for memory backing with file source
Add support in numa topology for file source inside memory backing
(hugepages are not needed) Three new elements <source/>,<access/> and
<allocation/> were added to <memoryBacking/> element. Also new
configuration parameter memory_backing_dir was added to qemu.conf.
- network: make openvswitch call timeout configurable
Adding the ability to specify the timeout value in seconds for
openvswitch calls in the libvirtd configuration file.
- bhyve: add e1000 NIC support
Add support for e1000 NIC. Previously, the only available option was
virtio-net.
- libxl: add tunneled migration support
Add tunneled migration to libxl driver, which is always capable of
strong encryption and doesn't require any extra network connection
other than what's required for remote access of libvirtd.
- qemu: add rendernode argument
Add a new attribute 'rendernode' to <gl> spice element.
- nodedev: add drm capability
Add a new 'drm' capability for Direct Rendering Manager (DRM) devices,
providing device type information.
- Add API for individual/specific vCPU hotplug
The new API allows selecting specific vCPUs to be added/removed from
the VM. The existing APIs allowed only adding/removing from the end
which did not play well with NUMA.
* Improvements
- virsh: pool-list: allow both --uuid and --name in one command
Adjust the virsh-pool command to support the --uuid and/or --name
options in order to print just the --name and/or --uuid of pools.
- Introduce MTU to domain <interface/> and <network>
Allow setting MTU size for some types of domain interface and network.
- libxl: improve support for <timer> configurations
Add support for multiple timers. Extend the tsc timer to support the
emulate mode. Improve conversion of timer XML to/from xl.cfg.
- storage: modularize the storage driver
Split up the storage driver backends into loadable modules so that
binary distributions don't have to compromise on shipping the storage
driver with all backends which may pull in too many dependencies.
* Bug fixes
- nodedev: Fabric name must not be required for fc_host capability
fabric_name is one of many fc_host attributes in Linux that is optional
and left to the low-level driver to decide if it is implemented. For
example the zfcp device driver does not provide a fabric name for an
fcp host. The requirement for the existence of a fabric name has been
removed by making it optional.
- bhyve: change address allocation schema for SATA disks
Previously, the bhyve driver assigned PCI addresses to SATA disks
directly rather than assigning that to a controller and using SATA
addresses for disks. It was implemented this way because bhyve has no
notion of an explicit SATA controller. However, as this doesn't match
libvirt's understanding of disk addresses, the bhyve driver was changed
to follow the common schema and have PCI addresses for SATA controllers
and SATA addresses for disks. If you're having issues because of this,
it's recommended to edit the domain's XML and remove <address
type='pci'> from the <disk> elements with <target bus='sata'/> and let
libvirt regenerate it properly.
- libxl: maximum memory fixes
Fix reporting of domain maximum memory. Fix setting dom0 maximum
memory.
- libxl: fix disk detach when <driver> not specified
- libxl: fix dom0 autoballooning with Xen 4.8
- qemu: Allow empty script path to <interface/>
Historically, this was always allowed. Unfortunately, due to some
rework done for 1.3.2 release a bug was dragged in which suddenly stop
allowing domain with such configuration to start.
# v3.0.0 (2017-01-17)
* New features
- Domain events for metadata content changes
The domain events framework has a new event ID that can be used to get
notifications when domain metadata content changes.
- Event notifications for the secret object
The secret object now supports event notifications, covering lifcycle
changes and secret value changes.
- New localPtr attribute for "ip" element in network XML
- qemu: Support QEMU group I/O throttling
Add the capability to allow group I/O throttling via a new domain
<disk> <iotune> subelement "group_name" to allow sharing I/O throttling
quota between multiple drives.
- nss: Introduce libvirt_guest
New libvirt_guest nss module that translates libvirt guest names into
IP addresses.
- daemon: Add support for runtime logging settings adjustment
Logging-related settings like log outputs and filters can now be
adjusted during runtime using the admin interface without the necessity
of the daemon's restart.
- storage: Add virStorageVolInfoFlags API
Add the API to support using the VIR_STORAGE_VOL_GET_PHYSICAL flag in
order to return the host physical size in bytes of the image container
in the allocation field of the _virStorageVolInfo structure. The
--physical flag has been added to the virsh vol-info command to access
the data.
- libxl: Implement virDomainGetMaxVcpus API
- storage: Add overwrite flag checking for logical pool
Add support for the OVERWRITE flags for the logical storage backend
including checking for existing data on the target volumes when
building a new logical pool on target volume(s).
- qemu: Add support for guest CPU configuration on s390(x)
* Improvements
- perf: Add more perf statistics
Add support to get the count of branch instructions executed, branch
misses, bus cycles, stalled frontend cpu cycles, stalled backend cpu
cycles, and ref cpu cycles by applications running on the platform.
- conf: Display <physical> for volume xml
Add a display of the <physical> size of a disk volume in the output of
the volume XML.
- qemu: Use virtio-pci by default for aarch64 mach-virt guests
virtio-pci provides several advantages over virtio-mmio, such as the
ability to hotplug devices and improved performance. While opting in to
virtio-pci has been possible for a while, newly-defined guests will now
use it automatically.
- vbox: remove support for VirtualBox 3.x and older
Those old VirtualBox versions have been unsupported by upstream for a
long time and the API of 4.0 and newer has diverged enough to require
code abstractions to handle differences. Removing support for those old
versions drops lots of code from the driver and simplifies the logic to
ease implementation of new features going forward.
- virsh: pool-info: introduce option --bytes
Add option --bytes to virsh pool-info in order ti allow display of
units in bytes rather than default of human readable output.
- scsi: Add parent wwnn/wwpn or fabric capability for createVport
Improve the algorithm searching for the parent scsi_host device for
vHBA/NPIV scsi_host creation. Rather than supplying the "parent" by
name, it's now possible to define the parent by it's wwnn/wwpn or
fabric_wwn in the node device create XML or the storage pool XML.
- qemu: aggregate pcie-root-ports onto multiple functions of a slot
When pcie-root-ports are added to pcie-root in order to provide a place
to connect PCI Express endpoint devices, libvirt now aggregates
multiple root ports together onto the same slot (up to 8 per slot) in
order to conserve slots. Using this method, it's possible to connect
more than 200 endpoint devices to a guest that uses PCIe without
requiring setup of any PCIe switches.
* Bug fixes
- lxc: fix accidental killing of containers during libvirtd restart
The libvirt_lxc process was previously not moved into the container
scope. As a result, if systemd reloads its config after a container is
started, when libvirtd is later restarted it will accidentally kill the
containers.
- qemu: Correct GetBlockInfo values
For an active domain, correct the physical value provided for a raw
sparse file backed storage and the allocation value provided for a
qcow2 file backed storage that hasn't yet been opened on the domain.
- qemu: Make virtio console usable on ppc64 guests
The chardev detection code has been improved and can now handle this
configuration properly.
- qemu: Enable mount namespace
To avoid funny races with udev relabelling devices under our hands and
to enhance security, libvirt now spawns each qemu process with its own
/dev.
- storage: Fix implementation of no-overwrite for file system backend
Fix file system storage backend implementation of the OVERWRITE flags
to be consistent between code and documentation. Add checks to ensure
that when building a new file system on a target volume that there is
not something already on the disk in a format that libvirt can
recognize.
- qemu: Create hugepage path on per domain basis
Historically, all hugepage enabled domains shared the same path under
hugetlbfs. This left libvirt unable to correctly set security labels on
it. With this release, however, each domain is put into a separate path
which is also correctly labeled.
- conf: Reject domains with duplicate drive addresses
Reject duplicate drive addresses for disks and hostdevs at domain
definition.
- libxl: reverse defaults on HVM net device attach
Fixes network interface attach for HVM domains when no model is
specified. Emulated hotplug isn't yet supported and hence we should
default to the general working scenario.
- libxl: always enable pae for x86_64 HVM
By default pae is disabled in libxl. Without an explicit <pae/> setting
in the domain <features> configuration, an x86_64 HVM domain would be
get an i686 environment. pae should always be enabled for x86_64 HVM
domains.
- qemu: Fix XML dump of autogenerated websocket
As a result autogenerated websocket port is regenerated on domain
restore, migration and such as it should be.
# v2.5.0 (2016-12-04)
* New features
- shmem: Add support for additional models
The shmem device can now utilize QEMU's ivshmem-plain and
ivshmem-doorbell, more modern versions of ivshmem.
- vbox: Add VirtualBox 5.1 support
- libssh: New transport
The new libssh transport allows one to connect to a running libvirtd
via SSH, using the libssh library; for example:
qemu+libssh://server/system.
- vhost-scsi: Add support scsi_host hostdev passthrough
Add the capability to pass through a scsi_host HBA and the associated
LUNs to the guest.
- Allow debugging of gluster volumes in qemu
Users can now enable debug logging for native gluster volumes in qemu
using the "gluster_debug_level" option in qemu.conf
- Pre-allocate memory slots for memory hotplug
Slot numbers for memory devices are now automatically allocated and
thus persistent. In addition slot numbers can be specified without
providing a base address, which simplifies user configuration
- qemu: Express devices will be placed on PCIe bus by default
For machine types that use a PCI Express root bus (e.g. x86_64/Q35 and
aarch64/virt), any unaddressed PCI device that is an Express device
(all virtio-1.0 devices, e1000e, nec-xhci, vfio assigned devices) will
be placed on an Express controller (i.e. a pcie-root-port) instead of a
legacy PCI controller (i.e. pci-bridge) with the root ports added as
needed.
* Improvements
- docs: Better documentation for migration APIs and flags
- vbox: Address thread safety issues
- virsh: Add support for passing an alternative persistent XML to migrate
command
- vhostuser: Allow hotplug of multiqueue devices
- NEWS: Switch to an improved format
List user-visible changes instead of single commits for a better
high-level overview of differences between libvirt releases.
- website: Modernize layout and branding
The libvirt website looked very cluttered and outdated; it has now been
completely overhauled, resulting in a design that's better organized
and more pleasant to look at.
* Bug fixes
- vz: Fix migration in P2P mode
- Forbid newline character in names of some libvirt objects
- Fix compilation on macOS
==============================================================================
Older libvirt releases didn't have proper release notes: if you are interested
in changes between them, you should check out ChangeLog* and docs/news-*.html.