Init a shared log instance in virtinst/logger.py, and use that
throughout the code base, so we aren't calling directly into
'logging'. This helps protect our logging output from being
cluttered with other library output, as happens with some
'requests' usage
qemu nowadays should reject the common cases here with its default
image locking. Our impl wasn't very good anyways so let's just leave
it up to the lower layers
Libvirt nowadays will throw an explicit warning about this
exact situation. It doesn't guide people to virt-manager UI, but
then again it should be obscure enough that it doesn't warrant this
special corner case code to inform the user
Having this at set_uefi time complicates the domain XML building
machinery, where we don't want things to have osinfo access.
Rearrange it so that editing cases call this explicitly, and
the XML builder just deals with it at the set_defaults time
It's hard to validate whether something like --extra-args or
--initrd-inject is supported based on the command line arguments. It's
easier to let the installer.py figure it out because it's the
authoritative source
This adds a field in gsettings and preferences UI to enable
XML editing. It's off by default. The XML tab page is still visible,
but the textview is not editable, and there's a warning at the top:
XML editing is disabled in 'Preferences'.
Only enable it if you know what you are doing.
When the setting is enabled, the warning goes away and the textview
is editable. This puts a roadblock up for people that don't know what
they are doing, but still advertises the feature and keeps the
UI surface difference fairly small between on/off states.
Handling this is a bit different from other bits, because:
1) the <device> editing paradigm is unique. We need to replace the
device in line in the XML which is a new operation
2) the New VM customize pattern is tricky and needs lots of
special handling
The XML editor ui is a two tabbed notebook, one 'Details' tab
and one 'XML' tab. The latter has a gtksourceview and allows editing
the raw libvirt XML for whatever the selected object is.
API users will programmatically insert the xmleditor notebook into
their UI, with the existing UI under the details tab.
SupportCache.check_support(SUPPORT_FOOBAR, args) becomes
SupportCache.foobar(args)
And SupportCache absorbs the caching infrastructure from
VirtinstConnection.
For now we add some hackery to hide the API change from callers, but
this will be undone in the next patch
Will take a string of comma separated options which we can use to
tweak app behavior, for testing. Convert the existing --test-* options
to use the same abstraction internally. This will make it easier to
add new test options in the future
Re-use CharSource, just like libvirt does internally. Adjust all
callers to match. Rename type -> backend_model while we are here,
because type is ambiguous
Move all ./source handling into CharSource, which will be reused by
other device classes as well. This requires us to add ../ handling
into our xmlapi xpath engine
It's not a pattern I think is worth extending in the future, and
make internal refactorings more difficult. Drop it, and drop it
from tpm and char devices since it is now unused
Whether we show something in the UI should be more dependent on
whether it's actually set in the XML, not some internal hardcoded
list which can go out of date.
Right now we overload the DEVICE column for non-device pages
like overview or memory. Add a separate column for KEY, and have
DEVICE==None for non-device pages
vmmVMWindow handles all the menuing, and coordinating between the
console, snapshots, and details panel. Simplifies the details
code a bit which will help when we add xmlediting
It's a mix of stuff libvirt already catches, UI sanity checking
like if user specified an empty field, and prompting for really
obscure corner cases. This stuff isn't important enough to carry
around all this code IMO
Name and forward mode config are always visible. ipv4, ipv6, and
domain name are under their own expanders which are collapsed by
default.
This will fit better with the XML editor pattern and reduce the
urge to squeeze more UI elements into the now smaller wizard
Rather than a mix of radio buttons and other combo boxes.
This follows the pattern we more commonly use in other UI, and
makes it easier to hide UI elements that aren't relevant for
specific choices, like the possibly large SR-IOV selector
This wizard is sufficiently obscure that I don't think it's
really valuable to try to explain networking concepts with
UI labels. If users don't know what they are trying to create
by using this wizard, there's no way we are going to adequately
explain to them what they are looking at. The example values
should be self explanatory enough anyways
This only applies for inter VM traffic when ipv6 networking is
disabled, which IMO is pretty obscure. If users want ipv6
connectivity, just enabling ipv6 will handle it appropriately
Right now we have:
* memory -> ./currentMemory
* maxmem -> ./memory
* hotplugmaxmem -> ./maxMemory
Which is just a mess to know what we might be really setting behind
the scenes. Rename the properties to match the XML element name, and
adjust all users to the new names. cli options aren't changed though
The current code already prefers the "display name" over the "name" of
each application; hence, use "name" only if the "display name" is not
available.
Reviewed-by: Cole Robinson <crobinso@redhat.com>
Use the non-deprecated g.inspect_list_applications2() call to list the
installed applications. It is available since libguestfs >= 1.20, which
is lower than the current requirement (i.e. 1.22).
Reviewed-by: Cole Robinson <crobinso@redhat.com>
Instead of passing around the raw results of
g.inspect_list_applications(), create an helper vmmInspectionApplication
object with the data of an inspected application that we use. This is
done for different reasons:
- when using the data, it is easier to use member variables instead of
looking up values in a dictionary
- we keep only the data needed, slightly lowering the memory/objects
used for the inspected applications
- it will be easier to switch from g.inspect_list_applications() to
g.inspect_list_applications2() without changing code outside the
inspection code
Reviewed-by: Cole Robinson <crobinso@redhat.com>
Save the package format of a guest as part of its inspection data, so
later on it can be used to tweak other results (like the version of
installed packages).
Reviewed-by: Cole Robinson <crobinso@redhat.com>
The old code basically always set filesystems_mounted to True, even when
no filesystems were mounted successfully, unless
g.inspect_get_mountpoints() failed (very unlikely).
Instead, set it when at least one filesystem is mounted; considering
that the first filesystem to be mounted is usually /, then failing to
mount it will usually prevent the mounting of the others. In any case,
we can try to extract data even when only / is mounted, which can work
depending on the mount points of the guest.
Reviewed-by: Cole Robinson <crobinso@redhat.com>
Rely on the Python 3 sorting facilities to sort the mount points using
a key based on the length of the mount point, doing the same effect as
the old compare function.
As side change required by this, enable python_return_dict on the
GuestFS handle, so we get proper hashes instead of lists. This requires
libguestfs 1.22, which is 6 years old by now (and other virt-manager
requires are way more recent than that).
Reviewed-by: Cole Robinson <crobinso@redhat.com>
This block only catches failures in g.inspect_get_mountpoints(), as the
g.mount_ro() calls are already within own try/catch blocks. Considering
that:
- g.inspect_get_mountpoints() is a simple API to query one of the
results of the inspection, it is very unlikely that it fails
- the whole _inspect_vm function (that contains the inspection code) is
already run within an own try/catch block, so even a failure in
g.inspect_get_mountpoints() will not crash virt-manager
then just remove this extra try/catch block.
Reviewed-by: Cole Robinson <crobinso@redhat.com>
Return the generated virtinst device up through the call chain.
Makes the flow a lot more sensible, and will be needed for separating
device building from extra UI validation/prompting
By default we copy CPU security features to the guest if specific CPU
model is selected. However, this may break migration and will affect
performance of the guest. This adds an option to disable this default
behavior.
The checkbox is clickable only on x86 and only on host where we can
detect any CPU security features, otherwise a tooltip is set to notify
users that there is nothing to copy.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
If user selects specific CPU model to be configured for guest we will
automatically add CPU security features to make sure that the guest is
not vulnerable to these CPU HW bugs.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1582667
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Cole Robinson <crobinso@redhat.com>
We will need to pass another variable into the setter so we cannot use
the property setter.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Cole Robinson <crobinso@redhat.com>
Currently we just call libvirt API which will return all CPU models for
specific architecture known to libvirt and we offer all of them to users
in GUI. Let's switch to domain capabilities where we have more details
about these CPUs such as whether each model is usable with current QEMU
binary. If libvirt can detect the usability we will offer only CPU
models that QEMU can actually run.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Cole Robinson <crobinso@redhat.com>
This doesn't seem to do anything, as we end up with with a
virtio-scsi controller anyway both when using virt-install and
when using the GUI, and it's not correct anyway because there's
nothing preventing ppc64/pseries guests from using virtio-scsi.
Signed-off-by: Andrea Bolognani <abologna@redhat.com>
Function do_we_default returns only if we want to default to clone disk
even if we know how to create default clone path. Only in case that the
storage pool is TYPE_DISK we don't know how to create default path and
we cannot default to clone that disk. In all other cases as ReadOnly
disk or Shareable and so on we can prepare the default path for user if
they decide to clone it.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1565106
Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
If user renames guest in VM details page before installation starts
there is no need to rename NVRAM file as well.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1667308
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
When cloning a guest in virt-manager the GUI shows a list of disks and
select default cloning policy for every disk. For storage pools where
we know that cloning is not possible we should not select that option
as default one.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1463066
Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Instead of duplicating the code into CLI and GUI move it into graphics
device file which is used from both places. This also fixes a bug in
virt-xml where changing listen to address was not working.
This also changes behavior to always configure one listen type when
using CLI listen option or GUI. If user wants to modify only specific
listen type they can use listens[] options from CLI.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1565968
Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Move set boot order responsibility to guest XML builder.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Boris Fiuczynski <fiuczy@linux.ibm.com>
Add 'get_boot_order' method. This method has an optional parameter for
getting the legacy boot order defined in the OS node of the domain
definition.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Boris Fiuczynski <fiuczy@linux.ibm.com>
Move get boot order methods to guest XML builder. For this we must
replace self.get_bootable_devices() with
self.get_bootable_devices(exclude_redirdev=True) in the
_get_device_boot_order and set_device_boot_order to have no
functionality change.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Boris Fiuczynski <fiuczy@linux.ibm.com>
Move use_device_boot_order to guest XML builder and rename it into
'set_device_boot_order'.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Boris Fiuczynski <fiuczy@linux.ibm.com>
Operate only on @boot_order and @guest and not on @self. This allows
us to refactor this functionality in the next patch. Additionally,
simplify the used algorithm without functionality change.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Boris Fiuczynski <fiuczy@linux.ibm.com>
Try to avoid self access since this function operates on the guest
definition only. This is needed for later refactoring as it will
reduce side-effects.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Boris Fiuczynski <fiuczy@linux.ibm.com>
Add a method to the Guest XMLBuilder for getting all bootable devices
of the guest definition with the option to exclude the redirdevs in
the returned list.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Boris Fiuczynski <fiuczy@linux.ibm.com>
Move (find_compare)_device functions to guest.py as it will be used
there in an upcoming patch.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Boris Fiuczynski <fiuczy@linux.ibm.com>
The Python binding of libguestfs errors out in case any of the details
(e.g. summary, description) of an application are not proper UTF-8.
This seems a rare situation, which nevertheless causes the failure of
the inspection of a guest, making the rest of the metadata not used.
While these encoding issues are tackled, and fixed in python-libguestfs
(or libguestfs itself), gracefully handle the failure of
inspect_list_applications(), so at least the rest of the details of a
guest are used/shown.
When a VM is using interface type=network with a forward mode=bridge
network, on qemu domain startup the runtime XML will be changed to
type=bridge and both source/@bridge and source/@network will be filled
in, which breaks some assumptions virt-manager had internally.
Unfortunately this isn't easy to test with the unittest driver either.
https://bugzilla.redhat.com/show_bug.cgi?id=1669265
Whether we check for a vol-format value shouldn't be dependent
on whether the UI is on screen (is_visible), but whether the UI
is requested to be shown (get_visible)
https://bugzilla.redhat.com/show_bug.cgi?id=1670266
Recently the Python binding of libguestfs was adapted to properly use
bytes in APIs that return data, instead of (ab)using strings [1].
This change was done only when built for Python 3, which has this
distinct bytes and strings.
Because of that, now the 'icon == ""' (empty string) checks fail, using
whatever inspect_get_icon() returns, including empty arrays of bytes.
Hence, change the checks to use the length of the data as condition, as
also the libguestfs Python API documentation says. Leave also the
checks for None, in the remote case the API will return None in the
future for no data.
[1] 0ee02e0117
This avoids the access of non-local variables of the closure.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Boris Fiuczynski <fiuczy@linux.ibm.com>
This has the advantage that we can access the attributes by name and
it's immutable by design.
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com>
Reviewed-by: Boris Fiuczynski <fiuczy@linux.ibm.com>
Add support for vsock devices to Hardware Details UI so that vsock devices can
be configured or removed.
Signed-off-by: Slavomir Kaslev <kaslevs@vmware.com>
100 is a bit too short for managing/creating vms over the corporate VPN. About half the time it times out and I have to exit the virtmgr app and restart it to see the new vm.
Now we have only one Installer class, and callers don't need to
worry about choosing a particular class type depending on their
needs, just pass cdrom vs. location to the installer init and
we figure out everything behind the scenes.
Besides simplifying the callers this makes the control flow a
lot easier to follow whether looking at InstallerTreeMedia or
Installer classes
It's a common requirement for VMs to send SCSI PR commands in VM cluster
environment. This patch adds the managed mode support of scsi persistent
reservation in details page.
Signed-off-by: Lin Ma <lma@suse.com>
Otherwise if we hit an error in connect.tick, connection can be
automatically closed in the UI, plus other parts of the code
were implicitly depending on this ordering, like VM rename support
Class will work above connection. Right know, it is possible to
provide stats for domains only, but could be expanded.
To retrieve stats, you can use get_vm_stats(vm).
This class uses new virConnectGetAllDomainStats call, which reduces
number of calls needed to poll stats.
Stats are refreshed with every connection tick.
Signed-off-by: Simon Kobyda <skobyda@redhat.com>
(crobinso: squash commits, move statsmanager to its own file,
fix pylint)
For one, we don't offer to install it anymore because we dropped
packagekit stuff. But even having a single package here isn't
really the whole story because different desktops have their
own askpass providers, like Fedora has openssh-askpass for GNOME,
but ksshaskpass for KDE
So drop it all, make the error more generic, and point people
to setting up SSH keys too
In the future, the only option for systray icons is going to
be statusnotifier/libappindicator. statusicon doesn't work on
wayland and gtk4 is dropping the APIs. KDE however is keeping
statusnotifier's around, and ubuntu is maintaining a gnome-shell
plugin to make them work
Revive support, but add a dbus check to make it smarter about
only using libappindicator when there's a listener present.
Better architect the code to separate the differences between
statusicon and libappindicator. Clean up the menu building
On first run of the app we will check to see if libvirt and qemu
are installed, and if not, offer to install them. In theory anyways.
In practice this stuff breaks repeatedly and is a pain to test because
every desktop has their own API provider with subtly different behavior.
My last round of testing about 12 months ago: apper on KDE was completely
busted and apparently unmaintained (although that may have changed lately),
gnome-software is the latest packagekit provider on gnome and completely
changes the semantics of the API compared to old style gnome-packagekit
that break a lot of virt-manager assumptions.
So I'm tired of it and want it all gone. Still use systemd to try and
check if libvirtd is running, and provide error messages at startup
to guide people.
Libvirtd can advertise itself over avahi. The feature is disabled by
default though and in practice I hear of no one actually using it
and frankly I don't think it's all that useful
The 'Open Connection' wizard has a disproportionate amount of code
devoted to this feature, but I don't think it's useful or worth
maintaining, so let's drop it
This switch says: if we detect a rhel host, use special version
checks that match rhel backports. This pattern sucks. The way
forward is to have libvirt advertise the bits that are supported,
through domcapabilities. Then virt-manager/virtinst can react
as appropriate.
The new UI is handled in mediacombo. It's a combobox+entry. The
combobox is prepopulated with host cdrom/floppy devices, and
previously used media paths from gsettings
The new VM wizard no longer has separate UI for cdrom device vs
ISO media. The choosecd dialog is gone all together, and media
is changed with the 'apply' button like all other details changes
There is a potential race of a backend to be opened and being true on
self._backend.is_open but self._hostinfo not being set yet.
Avoid spurial bugs due to that by also checking against None before
accessing subelements ot self._hostinfo
Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
This is just a big nasty commit.
Turn the OS inspection page into an always available page that
shows the libosinfo name from the domain metadata XML. Use oslist.py
and have it absorb more of the common behavior needed by create.py
and details.py. Add UI tests for it all
Rather than forcing API users to go through the capabilities APIs.
This lets us simplify things in virt-install quite a bit, and is
needed for smarter machine type defaults
I know openstack uses tcp consoles but for end users I've never
really heard about it. RHEL compiles out udp as well. I'm fine telling
users to go to the cli and use virt-xml for this use case.
Use this opportunity to drop a lot of code that only simplified the
case when there are tons of char options we need to consider
It simply makes vmdk disk images not selectable. I don't even know
if that's relevant for RHEL7+ anymore, but if it is, maintaining
code for disabling this in the UI doesn't make the situation any
better IMO
* Tweak the UI
* Add accelerator for the refresh button
* Make the IP labels selectable
* Drop the IP prefix from the UI, it's not the important bit
* Call DHCPLeases on the network instead to support this for more
drivers, like LXC
* Cache the IP results in the domain/network object wrappers
* Catch and log errors
* Poll for IP address when first visiting the interface page
If the interface's type is 'network', then code uses lease as one of
parameters of interface_addresses to get ipv4 and ipv6 address.
If the result is negative, then uses agent to try again, If the result
is still negative, uses arp for final try.
Signed-off-by: Lin Ma <lma@suse.com>
This changes all the callers to invoke start_install directly on the
Installer object. We still stash the installer instance inside the
guest object in create.py, just for simplicity
Cole Robinson