IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an
email to Administrator. User accounts are meant only to access repo
and report issues and/or generate pull requests.
This is a purpose-specific Git hosting for
BaseALT
projects. Thank you for your understanding!
Только зарегистрированные пользователи имеют доступ к сервису!
Для получения аккаунта, обратитесь к администратору.
Our glib integration had a defect that could cause stalls when streaming
a lot of data (like doing a cat /var/log/messages)
Also, add some tweaks to make the previously stated situation perform
better anyways.
We can get weird keyerrors sometimes, if the global tick thread is
scheduled off while a synchronous tick call is run (like after we create
a vm).
Eventually we will do away with manual tick() invocations but this will
do for now.
VM getinfo returns a system error if we are polling while the guest
is being shutdown (since qemu monitor connection hangs up). Make sure
the conn really dropped before we raise this error, but doing a connection
getinfo call.
Coupled with a recent libvirt patch, try and give more info about
a couple common policykit failure scenarios, like launching virt-manager
over SSH or VNC.
Do this by checking to see if we have a 'session' with ConsoleKit: the
above culprits don't give the user a local session, which causes
PolicyKit to deny.
Causes first run of dialogs to fail with a flicker on F16 KDE at least.
Choosecd also was using 'show' instead of 'present', which combined
with the first run fail meant that the dialog couldn't be displayed on
KDE :(
Also, the progress dialog was unconditionally shown, so this should fix
random flicker for quick operations, even in gnome.
This introduces new attribute wrpolicy with only supported
value as immediate. This helps specify whether to skip the
host page cache.
When wrpolicy is specified, meaning when wrpolicy=immediate
a writeback is explicitly initiated for the dirty pages in
the host page cache as part of the guest file write operation
Signed-off-by: Deepak C Shetty <deepakcs@linux.vnet.ibm.com>
It looks like the previous commit where fs readonly
was introduced caused this to happen. fs readonly
checkbox not appearing as its overlapping with the
target field... this patch fixes that.
Signed-off-by: Deepak C Shetty <deepakcs@linux.vnet.ibm.com>
Otherwise the information:
End of file while reading data: nc: invalid option -- 'U'
nc -h for help: Input/output error
is hidden in the error detail. See Debian bug:
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=614291
This patch adds the machine type combo callback
funcitons on non-x86 architectures.
For non-x86 architectures, user can select machine
type and the details of overview page are refreeshed,
and the machine type value is set.
Signed-off-by: Li Zhang <zhlcindy@linux.vnet.ibm.com>
(crobinso: squashed the two patches)
This patch adds support for the user to select Driver Type
on the addhardware->filesystems page. Currently only 'path'
and 'handle' driver types are supported. When anything but
'path' is selected, the security modes do not apply, hence
they are removed from the page based on user's selected
driver type.
Note: Some changes were needed in libvirt also to ensure
smooth experience while dealing with driver and mode
fields. Related libvirt patches are posted at ...
https://www.redhat.com/archives/libvir-list/2011-December/msg00919.html
Signed-off-by: Deepak C Shetty <deepakcs@linux.vnet.ibm.com>
(crobinso: small fix to handle if dev.driver is None)
Enables exporting filesytem as a readonly mount for guest, by
default read-write access is given (currently only works for
QEMU/KVM driver)
The support for the same in libvirt was added recently at ...
https://www.redhat.com/archives/libvir-list/2011-December/msg00937.html
Signed-off-by: Deepak C Shetty <deepakcs@linux.vnet.ibm.com>
This e.g. happens when the volume was moved away but the pool didn't get
refereshed:
Traceback (most recent call last):
File "/usr/share/virt-manager/virtManager/create.py", line 1229, in browse_import
self._browse_file(set_path, is_media=False)
File "/usr/share/virt-manager/virtManager/create.py", line 2118, in _browse_file
self.storage_browser.show(self.topwin, self.conn)
File "/usr/share/virt-manager/virtManager/storagebrowse.py", line 74, in show
self.reset_state(conn)
File "/usr/share/virt-manager/virtManager/storagebrowse.py", line 183, in reset_state
self.pool_selected()
File "/usr/share/virt-manager/virtManager/storagebrowse.py", line 257, in pool_selected
self.populate_storage_volumes()
File "/usr/share/virt-manager/virtManager/storagebrowse.py", line 322, in populate_storage_volumes
path = vol.get_target_path()
File "/usr/share/virt-manager/virtManager/storagevol.py", line 49, in get_target_path
return util.xpath(self.get_xml(), "/volume/target/path")
File "/usr/share/virt-manager/virtManager/libvirtobject.py", line 93, in get_xml
self.refresh_xml()
File "/usr/share/virt-manager/virtManager/libvirtobject.py", line 105, in refresh_xml
self._xml = self._XMLDesc(self._active_xml_flags)
File "/usr/share/virt-manager/virtManager/storagevol.py", line 35, in _XMLDesc
return self.vol.XMLDesc(flags)
File "/usr/lib/python2.7/dist-packages/libvirt.py", line 2002, in XMLDesc
if ret is None: raise libvirtError ('virStorageVolGetXMLDesc() failed', vol=self)
libvirt.libvirtError: cannot stat file '/var/scratch/vms/squeeze-template.img': No such file or directory
Instead of failing the whole pool simply ignore the one failing volume.
Without this patch selecting "Browse..." when creating a new vm from an
existing disk image won't show the volume dialog if the default pool has
a missing volume.
Mode label in FS page should be associated with the
fs-mode-combo, but its erronesouly is associated with
fs-type-combo, fixing that here.
Signed-off-by: Deepak C Shetty <deepakcs@linux.vnet.ibm.com>
In case of qemu for default fs type (mount) target is not
actually a directory, it is merely a arbitrary string tag
that is exported to the guest as a hint for where to mount.
This tag is unique and is exported as a virito-9p-pci
device and hence this patch adds check to ensure the target
string/tag is never entered duplicate by the user when
adding 2 or more filesystem devices.
Signed-off-by: Deepak C Shetty <deepakcs@linux.vnet.ibm.com>
VNC disconnect signals aren't being delivered, meaning we never cleanup
up SSH tunnels for remote VNC connections. This causes a reconnect
attempt to hang, trying to reuse the existing tunnels (which doesn't
work).
The problem is that we are cleaning up the VNC widget too early,
unregistering callbacks before they get a chance to be invoked.
Not cleaning up only causes a small memory leak for every console we
open, so only a minor problem compared to the original issue. A more
involved fix is required to clean up safely.
Since 'tick' isn't really thread safe, and we sometimes call a manual tick
to refresh resources (say immediately after creating a guest), we
can't depend on lists staying consistent.
If doing p2p migration from a remote host to localhost, we need to
generate a remote libvirt URI for qemu:///system. Try and use the remote
auth credentials of source connection.
