This adds:
VirtCLIArgument: a single foo=bar mapping
VirtOptionString: A collection of VirtCLIArguments, that parses the whole thing
VirtCLIParser: Represents a single cli option like --disk, --network, etc.
Centralizing this infrastructure opens up a lot of doors for future
improvements, like cli option introspection.
There were multiple problems with the setting and usage of
hide_unsupported_rhel_options. Due to the fact that the option has
several diferent namings throughout the code, this patch is renaming
it to stable_defaults, which basically says what the option does and
makes it possible to use it without need for more than one negation
(where the old code had up to 4 of them in some places), which also
helps understanding it.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
On first run, the remote URL install handling creates a storage pool
for /var/lib/libvirt/boot on the remote host. After this, it clears
the VirtualConnection's object cache, so the next time all pools are
fetched, it returns an accurate list.
However that clear_cache call wasn't propagated up to virt-manager's
cache. Add a new cb to fix it.
If we didn't set values for @type and @iobase in
XML, libvirt will use the default value.
Currently, virt-manager will display "-" if we don't
set any values.
This patch will use default value for display.
And update test case to cover this scenario.
Signed-off-by: Chen Hanxiao <chenhanxiao@cn.fujitsu.com>
Remove backend_mode from VirtualRNGDevice and allow to directly specify
bind and connect sources.
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
Updated by this script:
find -name '*.py' -exec sed -i "s|^\(#.*[^.?\!]\) \(.*[^#]\)$|\1 \2|g" \{\} \;
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
Ensure that any file touched by a @redhat.com author in 2013 has an
updated copyright header.
The files were updated using the build-aux/update-copyright gnulib
script and manually added where the copyright line wasn't present.
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
This could cause issues for people trying unattended non-graphical
kickstart installs and expecting ttyS0 in the guest to be hooked
up to the default console. So to get back the default behavior, you
can do:
--console pty
The XML we use is:
<clock offset="utc">
<timer name="rtc" tickpolicy="catchup"/>
<timer name="pit" tickpolicy="delay"/>
<timer name="hpet" present="no"/>
</clock>
Which translates to the qemu commands:
-no-hpet -no-kvm-pit-reinjection -rtc driftfix=slew
The latter two bits are already used by openstack and gnome boxes by
default.
On RHEL hpet is compiled out so -no-hpet is the default,
but not everywhere else. Though recently the RH guys confirmed that
for regular usage it should be turned off because a) qemu support
is not that good, b) most users don't need it anyways c) it has
a performance penalty.
This default can be overridden from a virt-install command like
--clock rtc_tickpolicy=delay
When Guest() sees that a timer has already been defined, it won't
set the new defaults, and that setting above is the old implied
default rtc setting.
These bits only apply if virt-manager is running directly on a RHEL6.
Since latest virt-manager can't do that thanks to GTK3 conversion,
just drop it all.
This used to happen:
- create VM with cdrom, cdrom gets hdc
- customize before install
- add new cdrom, gets target hda or hdb (first free target)
- new cdrom now has priority in the boot order
Change the logic to try to append targets first, and if there isn't
any space left, use the first free one. This may cause other issues
but we'll just have to wait and see.
It was required a long long time ago, before qemu supported -drive
and possibly ancient xen. Nowadays it should be pointless, and contributes
to some issues like bz 905439
This is where performance issues hit the most. This changes the
RHEL default to remove cache=None for all file volumes. If anyone
cares they can speak up, but it seems an overly unnecessary deviation
from upstream qemu.
The patches to drop /dev/loop0 meant get_blkdisk was inaccurate and
we weren't testing many things that were intended. Use a block disk
from the test driver.
Show a screenshot in the 'new snapshot' wizard. If we successfully create
that snapshot, save the screenshot in
~/.cache/virt-manager/$connuri/$vmuuid/snap-screenshot-$snapname.$ext
And show it in the snapshot details overview. We don't do any reaping
on snapshot delete, vm delete, etc, but that could be added later.
We sort them separately in the snapshot list, explicitly mention that
they are 'external', and add a UI field listing the memory/disk
details.
In general mixing internal and external snapshots is a recipe for
confusion and disaster, so I think the best thing to do is at least
acknowledge their presence in the UI but not make any attempt to
predict what will or will not work.
No unit tests for any of these, just for UI testing at the moment.
These require some non-upstream libvirt patches, but it doesn't
hurt in the interim while they are reviewed.
This is a standalone test like test_urls, and requires interraction. It
pulls down a bunch of kernels from public URL trees, inject known
kickstarts that induce quick failure. User inspects the output to
ensure initrd inject is working as expected.
And hide docs about old --noacpi/--noacpi options. I don't think
anyone really uses them anyways, but if anyone complains we should
just implement --features for the other CLI commands.
Convert virtManager/host.py, do some cleanups and modernization to the
UI there, add tests.
createnet.py hasn't really been touched yet, but still works because
it was building the XML by hand anyways.
So we unify parsing and building the XML.
Since we already do this for vmmDomain, take the opportunity to move
the shared infrastructure into vmmLibvirtObject
We already had this open coded in several classes with varying methods.
Now all someone has to do is
features = XMLChildProperty(CPUFeature)
The xmlbuilder code will pull the relevant xpath from the CPUFeature class,
handling parsing into instances of that class directly. The only
thing that needs to be manually added is the add_* remove_* functions.
This allows to make use of libvirt network filtering support with virt-install.
With the additional option "filterref" in the --network parameter one can
configure any defined nwfilter per network interface, i.e.:
virt-install ... --network network=mynet,model=virtio,filterref=clean-traffic
(crobinso: add an xmlparse test case)
- Unify looking up osdict values
- Have Guest.py pass in its defaults, not encoding them in osdict
- Explicitly specify every device parameter that osdict handles, rather
than have some psuedo magic dictionary format to handle it
The commit efab27afbf02743a3a2582e9a111eb1b7d985b26 in libvirt uses
/dev/%s for logical pools, change the tests to accommodate it.
(crobinso: additional fix to work with older libvirt)
As pointed out, the tests were configuration-dependent and thus might
have failed in case './setup.py configure' was run with non-default
graphics. The default graphics has changed as well and that lead to
test disaster. Reverting first hunk of commit b8a209ef and making the
tests independent of default graphics configuration makes all the
failures go away.
This will save us some test case churn. As an example, we now
do auto building of disk <target> XML and it doesn't alter things.
Without this bus and target are often swapped.
We separate all the provisioning bits to diskbackend.py. VirtualDisk
users now need to explicitly opt in to storage creation by using
set_create_storage().
validation is no longer done automatically, users must call the
validation() command.
__init__ drops all extra parameters.
This will eventually get us to a point where we can unify the manual
XML building and XML parsing machinery, and get consistent validation
behavior between devices.
This base connection object will be used to simplify the API in various
places, reduce libvirt API calls, and better share code between virtinst
and virt-manager. For now it just centralizes connection opening.
This also exposed various places where our handling for older libvirt
was busted, so raise our minimum host version to 0.6.0, the first
version that supports threaded client requests.
I'm sure the solaris support is totally broken, so just drop this
overcomplicated abstraction.
Also fix a bug where we were using usermode networking in a bunch of
places where it isn't supported.
When there are multiple usb devices with same vendor/product
in the host device list, the bus/addr is going to be used when
attaching one of usb devices.
Currently is_dup flag is only useful to VirtualHostDeviceUSB.
I put get_nodedevs_number() in connection.py, so the startup
hooks can use it.
When there are mutiple USB devices with identical vendorId and
productId, with --host-device <vendorId:productId>, virt-install
fails to specify one.
Currently the virt-install add the first node device with
these vendorId and productId.
This fix will throw out an error when argument to --host-device
can't specify unique node device.
For example:
virt-install ${other args} \
--host-device <vendorId:productId>
ERROR 15e1:2007 corresponds to multiple node devices
This patch adds and fixes test cases for cpu mode attribute.
Signed-off-by: Ken ICHIKAWA <ichikawa.ken@jp.fujitsu.com>
Signed-off-by: Hu Tao <hutao@cn.fujitsu.com>
(crobinso: Minor rebase tweaks)
In order to fix last tests that fail for me, I had to modify the
output xml, both modified machines will fit in the first numa node, so
the cpus from that node should be mentioned in the xml and that lead
me to the fact that the problem was in the test xml and not the code.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
For <hostdev> element, mode="capabilities" is used only for
block/character device passthrough and thus it fails with pci
passthrough.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
After commit 590c8dbebf, the tests got
broken due to the fact that the commit broke something that was
working. I'm changing the error to warning and fixing the tests.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
The purpose of this update is:
1. replace use of IPy with ipaddr since ipaddr has needed
functionality and IPy does not.
2. Update the glade-3 ui (vmm-host.ui) to provide
for IPv6 as well as IPv4. If either IPv4 or IPv6
is not defined, its respective frame is hidden.
3. The DNS Domain Name is added to the basic frame.
4. IPv4 forwarding is moved to the basic frame and ipv6
forwarding is added.
5. If an IPv4 address is not defined then the IPv4
is set to Isolated network, internal rout only.
6. IPv6 network has three values for Forwarding:
Isolated network, Isolated network with internal forwarding,
and Routed network.
7. Add network definitions to tests/testdriver.xml which
include IPv6 and other new parameters now being handled.
This patch includes the fix for getting a KeyError exception
when deleting a network definition.
In this update, createnet.py has been changed to use
ipaddr but it still only handles IPv4 network creation.
.
Signed-off-by: Gene Czarcinski <gene@czarc.net>
(crobinso: Add Gene to AUTHORS)
Despite being a known quantity, autotools is so overkill for our needs,
so let's drop it and replace it with a much simpler and easy to customize
system.
Switch over to use GObject introspection bindings for all python
modules related to GObject/GTK3/etc. It is not possible to mix
and match old pyggtk/pygobject manual bindings with new introspection
based bindings so it must be all changed in one go.
Imports like
import gtk
Change to
from gi.repository import Gtk
The vmmGObject class is changed to always inherit from GObject.GObject
There is no compelling reason to avoid a GObject dep for the
virt-manager TUI & it horribly messed up the code.
Signal declarations are changed from
vmmChooseCD.signal_new(vmmChooseCD, "cdrom-chosen", [object, str])
To
__gsignals__ = {
"cdrom-chosen": (GObject.SignalFlags.RUN_FIRST, None, [object, str])
}
which is required by new GObject bindings
Most of the rest of the change is simply dealing with renamed
constants / classes.
Alot of legacy compat code was removed - ie helpers which
check to see if certain GTK2 methods are available are no
longer required since we're mandating GTK3 only.
The event loop is replaced with LibvirtGLib's event loop.
Still todo
- Rip out all DBus stuff & make vmmEngine class inherit GtkApplication
which provides unique support & DBus method handling
- Switch to use LibvirtGConfig & LibvirtGObject for libvirt interaction
- Possibly switch to Python 3 too ?
- Figure out why GNOME keyring is missing Introspection support
My suggestion is that the standalone GIT repo for virt-install
only live on as a support branch for legacy platforms.
A stable-0.9 branch of virt-manager can be kept for legacy PyGtk2
based virt-manager releases.
The virt-manager master branch should exclusively use GObject
inspection and ideally Python3 and contain both the virt-manager
and virt-install codebases in one since they are intimately
related to each other & using separate GIT repos has needlessly
complicated life for everyone.
crobinso:
Some locking fixes
Misc cleanups and dropping now-useless code
Fix dbus usage
Fix graph cell renderer regression
Fix a couple tooltip issues
Cole Robinson