transport.close() (necessary to detach reader from the loop) will close
the FD. In case of stdin, it will prevent another call to this waiting
function. Use dup(2) to register cloned FD, which is safe to close
without side effects.
Since qubesd-side clone_vm function is gone, it doesn't make sense to
test it. For some time already this tested only if manual step-by-step
clone implemented _in the test itself_ was done correctly.
Actual cloning is part of qubes-core-admin-client and is tested there.
- nopat - added when PAT in Linux+Xen was buggy, no longer the case for a
long time.
- iommu=soft swiotlb=8192 - necessary to limit required memory on PV with
PCI devices; since we use HVM now (with proper IOMMU translation), this
is no longer needed. Furthermore, it will free some memory in sys-* vms.
Make an exception in vm.start() actually interrupt the test. The
asyncio.wait() returns list of completed tasks, where exception may be
stored - but is not raised directly. Change to asyncio.gather() that will
propagate the exception by default.
As a side effect, avoid deprecated direct coroutine passing to
asyncio.wait(). This functionality in asyncio.gather() is not
deprecated.
Add an API for VMs to announce support for non-service features. This is
very similar to supported-service.* features, but applies to non-service
features. This may be also used for announcing support for features that
do not use qvm-features framework itself - for example some VM kernel
features, installed drivers, packages etc.
QubesOS/qubes-issues#6030
Instead of unintuitive Value Error now we have dedicated QubesLabelNotFoundError.
Goal: to make qvm-prefs be less strange when one mixes up gray and grey again.
Deduplicate entries when iterating over TestVMsCollection values. Some
tests add given VM multiple times, to have it available under different
kind of keys (name, uuid etc) - similar to the real VMsCollection.
* origin/pr/254:
vm: allow StandaloneVM to be a DVM template
vm: do not allow setting template_for_dispvms=False if there are any DispVMs
vm: move DVM template specific code into separate mixin
* test-fixes20200806:
tests/extra: add vm.run(..., gui=) argument
tests: collect detailed diagnostics on failure
tests: workaround a race in qrexec test
tests: fix audio recording test
tests: make qvm-sync-clock test more reliable
Help debugging test failures by collecting detailed information on
failure. It will be logger to the standard logger, which will end up
either on stderr or in journalctl.
qrexec-client-vm may return earlier than it's child process (it exits
right away, without waiting for its child). Add a small wait before
reading exit code from a file.
* paranoid-restore:
tests: paranoid backup restore
Add policy for paranoid mode backup restore
Add an extension preventing starting a VM while it's being restored
Add support for 'tag-created-vm-with' feature
To calculate frequency it needs to use samples per second (44100), not
samples pre recording lenght. This caused shorter recordings to not fit
into the margin.
Compare the time with the "current" time retrieved from ClockVM just
before comparing, not with the test start time. This should work even if
the test machine is quite slow (test taking more than 30s).
Do not allow starting a VM while the restoring management VM has still
control over it. Specifically, that restoring VM will not be able to
start just restored VM.
QubesOS/qubes-issues#5310
When a VM with 'tag-created-vm-with' feature set creates a VM (using
Admin API), that VM will get all the tags listed in the feature.
Multiple tags can be separated with spaces.
This will be useful to tag VMs created during paranoid mode backup
restore.
QubesOS/qubes-issues#5310
As discussed in the PR, sync code will not be interrupted when run from
async code as long as Qubes OS doesn't run dedicated threads for async
& sync code. So there's simply no issue to be expected and thus no special
caution required.
Added:
post_volume_create & post_volume_import as requested by Marek
Removed:
post_ctor as this wasn't really useful anyway, but required a lot of
sync code. Without it, some refactoring & potential async improvements
became possible.
This reverts commit 287a4a0429.
As Marek correctly pointed out, sync functions cannot be run async against one another even if run inside an async function
(the python interpreter will remain active until the next yield and that's at the end of the sync func / inside the async function).
--> So there's no need for a lock.
I still cannot protect against assumptions made by sync code authors about blocking the Qubes OS main loop. Those will be broken.
Moreover the code of this commit was botched anyway.
- Removed all own class attributes to avoid name clashes with delegated
class attributes.
- Implemented the previously missing Pool.usage_details property.
- Shadowed all class attributes as instance properties. This is required
as the parent classes enforce the class attributes upon the
CallbackPool & CallbackVolume classes, but they need to be delegated to
the class of the _cb_impl object. We also cannot implement them as class
attributes in CallbackVolume & CallbackPool as they need to work for
arbitrary backend drivers and two backend drivers must not interfere with
each other. Possible alternative: One could dynamically create classes.
Unfortunately this appears to be necessary due to
various Qubes OS `assert` checks and to get `__str__()` et al
from the super class. It also means that we have to implement
all methods of the super class (in the future as well).
* devel20200705:
tests: skip gnome-terminal on xfce template flavor
tests: fix FD leak in qrexec test
tests: switch default LVM pool to qubes_dom0/vm-pool
backup: fix error handler for scrypt errors
Adjust code for possibly coroutine Volume.export() and Volume.export_end()
storage: add Volume.export_end() function
backup: add support for calling a function after backing up a file/volume
backup: call volume.export() just before actually extracting it
vm/dispvm: place all volumes in the same pool as DispVM's template
tests: extend TestPool storage driver to make create_on_disk working
storage: pass a copy of volume_config to pool.init_volume
tests: cleanup properly in wait_on_fail decorator
* origin/pr/352:
audio: set sink volume to workaround alsa save/restore issue
audio: increase timeout to match hvm loading
audio: auxiliary pauses should be harmless now, place them back just in case
audio: add silence threshold
audio: unload guest' module-vchan-sink in hvm tests
audio: fix prepare_audio_vm
audio: do not use pacat on copying audio_in.raw
Audio: rework audio tests
Similart to property-reset:xid, emit property-reset:stubdom_xid when
domain is started/stopped. This allows client side of the Admin API
(qubes-core-admin-client) to invalidate the cache when necessary.
Found by audio tests: #352
Now Volume.export() may be a coroutine and also may be accompanied by
Volume.export_end() cleaning up after it.
See previous commits for building blocks for this.
This commit adjusts usage of Volume.export() and adds matching
Volume.export_end() throughout the code base.
FixesQubesOS/qubes-issues#5935
This is a counterpart to Volume.export(). Up until now, no driver needed
any cleanup after exporting data, but it doesn't mean there won't be
any. This is especially relevant because Volume.export() is supposed to
return a path of a snapshot from before VM start - which may be a
different one than currently active one.
QubesOS/qubes-issues#5935
When Volume.export is called late and can be also a coroutine, it may
make sense to also have a cleanup function for changes made by it.
This commit only adjust backup code internals, but doesn't call
appropriate Volume function yet.
QubesOS/qubes-issues#5935
There are two reasons for this:
- call it from a coroutine, allowing export() itself be a coroutine
- avoid calling export() when only collecting preliminary backup
summary
Both needs some more changes in other parts of the codebase to be useful
(see next commits).
This will be especially useful when export() will need to make some
changes (like, create a snapshot, mount something etc).
QubesOS/qubes-issues#5935
Make all volume's pool controlled by DisposableVM Template. This
specifically makes DispVM's volatile volume to be placed directly in the
same pool as its template.
FixesQubesOS/qubes-issues#5933
Add dummy TestVolume with empty create() method. Other core code
requires also TestPool.get_volume implemented, so add that too (naive
version remembering instances returned from TestPool.init_volume).
Avoid local modification in a pool's init_volume influence
vm.volume_config. Currently every pool driver replaces
volume_config['pool'] with a pool object (instead of name) and it leads
to confusing cases where depending on start stage, it is sometimes an
object and sometimes a string.
Additionally, some pool drivers may modify volume_config in unexpected
way - for example test pool driver removes 'pool' entry entirely. Avoid
this fragile interface by giving pool driver a copy of volume_config,
instead of vm.volume_config directly.
Note one side effect is that 'vid' (and other pool-specific parameters)
is not set into vm.volume_config directly after creating a VM, but
possibly only after loading from XML. This should not be an issue in
theory (no core code should expect it), but if some place use
volume_config instead of Volume instance for getting pool-specific
options, it should be fixed.
Close transport used to wait for user input, otherwise all further tests
would fail on cleanup (FD leak detected). This in practice is only
useful when using wait_on_fail decorator without --failfast option.
Don't update _size in the getter, so it can be unlocked (which is
helpful for QubesOS/qubes-issues#5935).
!!! If cherry-picking for release4.0, also adjust import_data() to !!!
!!! use self.size (no underscore) instead of self._get_size() !!!
