Consider the following XML snippet:
<memory model=''>
<target>
<size unit='KiB'>523264</size>
<node>0</node>
</target>
</memory>
Whats wrong you ask? The @model attribute. This should result in
an error thrown into users faces during virDomainDefine phase.
Except it doesn't. The XML validation catches this error, but if
users chose to ignore that, they will end up with invalid XML.
Well, they won't be able to start the machine - that's when error
is produced currently. But it would be nice if we could catch the
error like this earlier.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
The original name 'admin_uri_default' was introduced to our code by commit
dbecb87f. However, at that time we already had a separate config file for
admin library but the commit mentioned above didn't properly adjust the
config's option name. The result is that when we're loading the config, we
check a non-existent config option (there's not much to do with the URIs
anyway, since we only allow local connection). Additionally, virt-admin's man
page documents, that the default URI can be altered by setting
admin_uri_default option. So the fix proposed by this patch leaves the
libvirt-admin.conf as is and adjusts the naming in the code as well as in the
virt-admin's man page.
Signed-off-by: Erik Skultety <eskultet@redhat.com>
https://bugzilla.redhat.com/show_bug.cgi?id=1356436
Commit id '56057900' altered the discovery of iSCSI node targets by
using the "--op nonpersistent". This caused issues for clean environments
or if by chance a "-m node -o delete" was executed.
Since each iSCSI Storage Pool has the required iSCSI target path, use
that and the virISCSINodeNew API in order to generate the iSCSI node record.
https://bugzilla.redhat.com/show_bug.cgi?id=1356436
According to RFC 3721 (https://www.ietf.org/rfc/rfc3721.txt), there are
two ways to "discover" targets in/for the iSCSI environment. Discovery
is the process which allows the initiator to find the targets to which
it has access and at least one address at which each target may be
accessed.
The method currently implemented in libvirt using the virISCSIScanTargets
API is known as "SendTargets" discovery. This method is more useful when
the target IP Address and TCP port information are available, e.g. in
libvirt terms the "portal". It returns a list of targets for the portal.
From that list, the target can be found. This operation can also fill an
iSCSI node table into which iSCSI logins may occur. Commit id '56057900'
altered that filling by adding the "--op nonpersistent" since it was
not necessarily desired to perform that for non libvirt related targets.
The second method is "Static Configuration". This method not only needs
the IP Address and TCP port (e.g. portal), but also the iSCSI target name.
In libvirt terms this would be the device path field from the iSCSI pool
<source> XML. This patch implements the second methodology using that
required device path as the targetname.
The current LUKS support has a "luks" volume type which has
a "luks" encryption format.
This partially makes sense if you consider the QEMU shorthand
syntax only requires you to specify a format=luks, and it'll
automagically uses "raw" as the next level driver. QEMU will
however let you override the "raw" with any other driver it
supports (vmdk, qcow, rbd, iscsi, etc, etc)
IOW the intention though is that the "luks" encryption format
is applied to all disk formats (whether raw, qcow2, rbd, gluster
or whatever). As such it doesn't make much sense for libvirt
to say the volume type is "luks" - we should be saying that it
is a "raw" file, but with "luks" encryption applied.
IOW, when creating a storage volume we should use this XML
<volume>
<name>demo.raw</name>
<capacity>5368709120</capacity>
<target>
<format type='raw'/>
<encryption format='luks'>
<secret type='passphrase' uuid='0a81f5b2-8403-7b23-c8d6-21ccd2f80d6f'/>
</encryption>
</target>
</volume>
and when configuring a guest disk we should use
<disk type='file' device='disk'>
<driver name='qemu' type='raw'/>
<source file='/home/berrange/VirtualMachines/demo.raw'/>
<target dev='sda' bus='scsi'/>
<encryption format='luks'>
<secret type='passphrase' uuid='0a81f5b2-8403-7b23-c8d6-21ccd2f80d6f'/>
</encryption>
</disk>
This commit thus removes the "luks" storage volume type added
in
commit 318ebb36f1
Author: John Ferlan <jferlan@redhat.com>
Date: Tue Jun 21 12:59:54 2016 -0400
util: Add 'luks' to the FileTypeInfo
The storage file probing code is modified so that it can probe
the actual encryption formats explicitly, rather than merely
probing existance of encryption and letting the storage driver
guess the format.
The rest of the code is then adapted to deal with
VIR_STORAGE_FILE_RAW w/ VIR_STORAGE_ENCRYPTION_FORMAT_LUKS
instead of just VIR_STORAGE_FILE_LUKS.
The commit mentioned above was included in libvirt v2.0.0.
So when querying volume XML this will be a change in behaviour
vs the 2.0.0 release - it'll report 'raw' instead of 'luks'
for the volume format, but still report 'luks' for encryption
format. I think this change is OK because the storage driver
did not include any support for creating volumes, nor starting
guets with luks volumes in v2.0.0 - that only since then.
Clearly if we change this we must do it before v2.1.0 though.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Refactor the virStorageFileMatchesNNN methods so that
they don't take a struct FileFormatInfo parameter, but
instead get the actual raw dat items they needs. This
will facilitate reuse in other contexts.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
To collect all balloon statistics for all guests it was necessary to make
several libvirt requests. Now it's possible to get all balloon statiscs via
single connectGetAllDomainStats call.
Signed-off-by: Derbyshev Dmitry <dderbyshev@virtuozzo.com>
To allow using failover with gluster it's necessary to specify multiple
volume hosts. Add support for starting qemu with such configurations.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
To allow richer definitions of disk sources add infrastructure that will
allow to register functionst generating a JSON object based definition.
This infrastructure will then convert the definition to the proper
command line syntax and use it in cases where it's necessary. This will
allow to keep legacy definitions for back-compat when possible and use
the new definitions for the configurations requiring them.
Add support for converting objects nested in arrays with a numbering
discriminator on the command line. This syntax is used for the
object-based specification of disk source properties.
As gluster natively supports multiple hosts for failover reasons we can
easily add the support to the storage driver code in libvirt.
Extract the code setting an individual host into a separate function and
call them in a loop. The new code also tries to keep the debug log
entries sane.
Extract the code so that it can be called from multiple places. This
also removes a tricky fallthrough in the large switch in
qemuBuildNetworkDriveStr.
Add a modular parser that will allow to parse 'json' backing definitions
that are supported by qemu. The initial implementation adds support for
the 'file' driver.
Due to the approach qemu took to implement the JSON backing strings it's
possible to specify them in two approaches.
The object approach:
json:{ "file" : { "driver":"file",
"filename":"/path/to/file"
}
}
And a partially flattened approach:
json:{"file.driver":"file"
"file.filename":"/path/to/file"
}
Both of the above are supported by qemu and by the code added in this
commit. The current implementation de-flattens the first level ('file.')
if possible and required. Other handling may be added later but
currently only one level was possible anyways.
The cur_balloon also increases/decreases with dimm hotplug/unplug.
To be consistent, adjust the value for coldplug too. This was inconsistently
taken care when cur_ballon != memory to begin with. The patch fixes it
irrespective of that.
Signed-off-by: Shivaprasad G Bhat <sbhat@linux.vnet.ibm.com>
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Since commit c4bdff19, the path to the configuration file has been constructed
in the following manner:
- if no config filename was passed to virConfLoadConfigPath, libvirt.conf was
used as default
- otherwise the filename was concatenated with
"<config_dir>/libvirt/libvirt%s%s.conf" which in admin case resulted in
"libvirt-libvirt-admin.conf.conf". Obviously, this non-existent config led to
ignoring all user settings in libvirt-admin.conf. This patch requires the
config filename to be always provided as an argument with the concatenation
being simplified.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1357364
Signed-off-by: Erik Skultety <eskultet@redhat.com>
For use with memory hotplug virQEMUBuildCommandLineJSONRecurse attempted
to format JSON arrays as bitmap on the command line. Make the formatter
function configurable so that it can be reused with different syntaxes
of arrays such as numbered arrays for use with disk sources.
This patch extracts the code and adds a parameter for the function that
will allow to plug in different formatters.
Until now the JSON->commandline convertor was used only for objects
created by qemu. To allow reusing it with disk formatter we'll need to
escape ',' as usual in qemu commandlines.
Refactor the command line generator by adding a wrapper (with
documentation) that will handle the outermost object iteration.
This patch also renames the functions and tweaks the error message for
nested arrays to be more universal.
The new function is then reused to simplify qemucommandutiltest.
As we already test that the extraction of the backing store string works
well additional tests for the backing store string parser can be made
simpler.
Export virStorageSourceNewFromBackingAbsolute and use it to parse the
backing store strings, format them using virDomainDiskSourceFormat and
match them against expected XMLs.
Nothing in the code path after the removed call has needs/uses the alias
anyway (as would be the case for command line building or talking to monitor).
The alias is VIR_FREE'd in virDomainDeviceInfoClear which is called for any
device that needs/uses an alias via virDomainDeviceDefFree or virDomainDefFree
as well as during virDomainDeviceInfoFree for host devices.
For persistent domains, the domain definition (including aliases) gets
freed a few screens later when it's replaced with newDef.
For transient domains, the definition is freed/unref'd along with the
virDomainObj a few moments later.
Introduce initial support for domainBlockStats API call that
allow us to query block device statistics. OpenStack nova
uses this API call to query block statistics, alongside
virDomainMemoryStats and virDomainInterfaceStats. Note that
this patch only introduces it for VBD for starters. QDisk
would come in a separate patch series.
A new statistics data structure is introduced to fit common
statistics among others specific to the underlying block
backends. For the VBD statistics on linux these are exported
via sysfs on the path:
"/sys/bus/xen-backend/devices/vbd-<domid>-<devid>/statistics"
To calculate the block devno libxlDiskPathToID is introduced.
Each backend implements its own function to extract statistics,
allowing support for multiple backends and different platforms.
VBD stats are exposed in reqs and number of sectors from
blkback, and it's up to us to convert it to sector sizes.
The sector size is gathered through xenstore in the device
backend entry "physical-sector-size".
BlockStatsFlags variant is also implemented which has the
added benefit of getting the number of flush requests.
Signed-off-by: Joao Martins <joao.m.martins@oracle.com>
QEMU reports timestamp along with other memory statistics, but this information is not saved into domain statistics.
It could be useful to determine if the data reported is fresh or not.
Balloon statistics are not reported in hrf, so no modifications are made in qemu_monitor_text.c.
Signed-off-by: Derbyshev Dmitry <dderbyshev@virtuozzo.com>
'memtotal' in virtio drivers and qemu corresponds to 'available' in libvirt.
Because of that, 'stat-available-memory' is renamed into 'usable'.
Balloon statistics are not reported in hrf, so no modifications are made in qemu_monitor_text.c.
Signed-off-by: Derbyshev Dmitry <dderbyshev@virtuozzo.com>
Dropping the caching of ccw address set.
The cached set is not required anymore, because the set is now being
recalculated from the domain definition on demand, so the cache
can be deleted.
The address sets (pci, ccw, virtio serial) are currently cached
in qemu private data, but all the information required to recreate
these sets is in the domain definition. Therefore I am removing
the redundant data and adding a way to recalculate these sets.
Add a function that calculates the ccw address set
from the domain definition.
Dropping the caching of virtio serial address set.
The cached set is not required anymore, because the set is now being
recalculated from the domain definition on demand, so the cache
can be deleted.
Credit goes to Cole Robinson.
Dropping the caching of virtio serial address set.
Instead of using the cached address set, a function in qemu_hotplug.c
now recalculates it on demand.
Credit goes to Cole Robinson.
The address sets (pci, ccw, virtio serial) are currently cached
in qemu private data, but all the information required to recreate
these sets is in the domain definition. Therefore I am removing
the redundant data and adding a way to recalculate these sets.
Add a function that calculates the virtio serial address set
from the domain definition.
Credit goes to Cole Robinson.
The symbol being missing has been reported as causing build
failures on OS X. If it's not already defined, define it to
zero so that it won't have any effect.
Commit 4a585a88 introduced searching QOM device path by alias, let's use it for
memballoon too. This may speedup the search because in most cases we will find
the correct QOM device path directly by using alias without the need for the
recursion code.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Commit ce745914 introduced detection of actual video ram sizes to fix migration
if QEMU decide to modify the values provided by libvirt. This works perfectly
for domains with number of video devices up to two.
If there are more than two video devices in the guest all the secondary devices
in the XML will have the same memory values. This is because our current code
search for QOM device path only by the device type name and all the secondary
video devices has the same name "qxl".
This patch introduces a new search function that will try to search a QOM device
path using also device's alias if the alias is available. After that it will
fallback to the old recursive code if the alias search found no results.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1358728
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Previously, qemuDomainAttachDeviceFlags was doing two things:
handling the job and attaching devices. Now the second part is
in a new function.
This change is required to make it possible to test more complex
device attachment situations, like attaching a device to both
config and live at once.
We want to be able to pass a NULL instead of the connection
and use this function in tests. To achieve this, the virConnectPtr
is passed instead of virDomainPtr, and the driver is a new separate
parameter.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
So commit 306b3a8504 tried mimicking behaviour of commit 540c339a25, but
added a virObjectRef(vm) only after virDomainObjListAdd() in
lxcDomainDefineXMLFlags() and not in lxcDomainCreateXMLWithFiles().
That way undefining a domain that was started with different XML than
defined will leave the domain object in a state with not enough
references to then remove it. Hence any lxcDomainDestroyFlags() called
afterwards crashes the daemon.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1351057
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Since return code is checked globally at the end of the function, let's
make sure that we set it correctly at any point.
This fixes a regression introduced in commit 0aa19f35 where the first
command to eject changeable media would fail unconditionally.
Signed-off-by: Bjoern Walk <bwalk@linux.vnet.ibm.com>
Reviewed-by: Boris Fiuczynski <fiuczy@linux.vnet.ibm.com>
This patch forces container's init process, to become a session leader,
that is its session ID is made the same as its process ID.
That might seem unnecessary in general, but if we want to checkpoint a
container with CRIU, which is needed for container migration,
we must ensure that the SID of each process inside the container points
to a process that lives in the same PID namespace as the container.
Therefore, we force that the session leader is the init.
Signed-off-by: Katerina Koukiou <k.koukiou@gmail.com>
When parsing a command line with USB devices that have
no address specified, QEMU automatically adds a USB hub
if the device would fill up all the available USB ports.
To help most of the users, add one hub if there are more
USB devices than available ports. For wilder configurations,
expect the user to provide us with more hubs and/or controllers.
Walk through all the usb hubs in the domain definition
that have a USB address specified, create the
corresponding structures in the virDomainUSBAddressSet
and mark the port it occupies as used.
A new type to track USB addresses.
Every <controller type='usb' index='i'/> is represented by an
object of type virDomainUSBAddressHub located at buses[i].
Each of these hubs has up to 'nports' ports.
If a port is occupied, it has the corresponding bit set in
the 'ports' bitmap, e.g. port 1 would have the 0th bit set.
If there is a hub on this port, then hubs[i] will point
to this hub.
Undefine procedure drops domain lock while waiting for detaching
disks vz sdk call. Meanwhile vz sdk event domain-config-changed
arrives, its handler finds domain and is blocked waiting for job
condition. After undefine API call finishes event processing procedes
and tries to refreshes domain config thru existing vz sdk domain handle.
Domain does not exists anymore and event processing fails. Everything
is fine we just don't want to see error message in log for this
particular case.
Fortunately domain has flag that domain is removed from list. This
also imply that vz sdk domain is also undefined. Thus if we check
for this flag right after domain is locked again on accuiring
job condition we gracefully handle this situation.
Actually the race can happen in other situations too. Any
time we wait for job condition in mutualy exclusive job in
time when we acquire it vz sdk domain can cease to exist.
So instead of general internal error we can return domain
not found which is easier to handle. We don't need to patch
other places in mutually exclusive jobs where domain lock
is dropped as if job is started domain can't be undefine
by mutually exclusive undefine job.
The code of this patch is quite similar to qemu driver checks
for is domain is active after acquiring a job. The difference
only while qemu domain is operational while process is active
vz domain is operational while domain exists.
Current vz driver implementation is not usable when it comes to
long runnig operations. Migration or saving a domain blocks all
other operations even query ones which are expecteted to be available.
This patch addresses this problem.
All vz driver API calls fall into next 3 groups:
1. only query domain cache (virDomainObj, vz cache statistic)
examples are vzDomainGetState, vzDomainGetXMLDesc etc.
2. use thread shared sdkdom object
examples are vzDomainSetMemoryFlags, vzDomainAttachDevice etc.
3. use no thread shared sdkdom object nor domain cache
examples are vzDomainSnapshotListNames, vzDomainSnapshotGetXMLDesc etc
API calls from group 1 don't need to be changed as they hold domain lock only
for short period of time. These calls [1] are easily distinguished. They query
domain object thru libvirt common code or query vz sdk statistics handle thru
vz sdk sync operations.
vzDomainInterfaceStats is the only exception. It uses sdkdom object to
convert interface name to its vz sdk stack index which could not be saved in
domain cache. Interface statistics is available thru this stack index as a key
rather than name. As a result we can have accidental 'not known interface'
errors on quering intrerface stats. The reason is that in the process of
updating domain configuration we drop all devices and then recreate them again
in sdkdom object and domain lock can be dropped meanwhile (to remove networks
for existing bridged interfaces and(or) (re)create new ones). We can fix this
by changing the way we support bridged interfaces or by reordering operations
and changing bridged networks beforehand. Anyway this is better than moving
this API call into 2 group and making it an exclusive job.
As to API calls from group 2, first thread shared sdkdom object needs to be
explained. vz sdk has only one handle for a given domain, thus threads need
exclusive access to operate on it. These calls are fixed to drop and reacquire
domain lock on any lengthy operations - namely waiting the result of async vz
sdk operation. As lock is dropped we need to take extra reference to domain
object if it is not taken already as domain object can be deleted from list
while lock is dropped. As this operations use thread shared sdkdom object, the
simplest way to make calls from group 2 be consistent to each other is to make
them mutually exclusive. This is done by taking/releasing job condition thru
calling correspondent job routine. This approach makes group 1 and group
2 calls consistent to each other too. Not all calls of group 2 change the
domain cache but those that do update it thru prlsdkUpdateDomain which holds
the lock thoughout the update.
API calls from group [2] are easily distinguished too. They use
beginEdit/commit to change domain configuration (vzDomainSetMemoryFlags) or/and
update domain cache from sdkdom at the end of operation (vzDomainSuspend).
There is a known issue however. Frankly speaking it was introduced by ealier
patch '[PATCH 6/9] vz: cleanup loading domain code' from a different series.
The patch significantly reduced amount of time when the driver lock is held when
creating domain from API call or as a result of domain added event from vz sdk.
The problem is these two paths race on using thread shared sdkdom as we don't
have libvirt domain object and can not lock on it. However this don't
invalidates the patch as we can't use the former approach of preadding domain
into the list as we need name at least and name is not given by event. Anyway
i'm against adding half baked object into the list. Eventually this race can be
fixed by extra measures. As to current situation races with different
configurations are unlikely and race when adding domain thru vz driver and
simultaneous event from vz sdk is not dangerous as configuration is the same.
The last group [3] is API calls that need only sdkdom object to make vz sdk
call and don't change thread shared sdkdom object or domain cache in any way.
For now these are mostly domain snapshot API calls. The changes are similar to
those of group 2 - they add extra reference and drop/reacquire the lock on waiting
vz async call result. One can simply take the immutable sdkdom object from the
cache and drop the lock for the rest of operations but the chosen approach
makes implementation of these API calls somewhat similar to those of from group
2 and thus a bit futureproof. As calls of group 3 don't need vz driver
domain/vz sdk cache in any way, they are consistent with respect to API calls from
groups 1 and 3.
There is another exception. Calls to make-snapshot/revert-to-snapshot/migrate
are moved to group 2. That is they are made mutually exclusive. The reason
is that libvirt API supports control/query only for one job per domain and
these are jobs that are likely to be queried/aborted.
Appendix.
[1] API calls that only query domain cache.
(marked [*] are included for a different reason)
.domainLookupByID = vzDomainLookupByID, /* 0.10.0 */
.domainLookupByUUID = vzDomainLookupByUUID, /* 0.10.0 */
.domainLookupByName = vzDomainLookupByName, /* 0.10.0 */
.domainGetOSType = vzDomainGetOSType, /* 0.10.0 */
.domainGetInfo = vzDomainGetInfo, /* 0.10.0 */
.domainGetState = vzDomainGetState, /* 0.10.0 */
.domainGetXMLDesc = vzDomainGetXMLDesc, /* 0.10.0 */
.domainIsPersistent = vzDomainIsPersistent, /* 0.10.0 */
.domainGetAutostart = vzDomainGetAutostart, /* 0.10.0 */
.domainGetVcpus = vzDomainGetVcpus, /* 1.2.6 */
.domainIsActive = vzDomainIsActive, /* 1.2.10 */
.domainIsUpdated = vzDomainIsUpdated, /* 1.2.21 */
.domainGetVcpusFlags = vzDomainGetVcpusFlags, /* 1.2.21 */
.domainGetMaxVcpus = vzDomainGetMaxVcpus, /* 1.2.21 */
.domainHasManagedSaveImage = vzDomainHasManagedSaveImage, /* 1.2.13 */
.domainGetMaxMemory = vzDomainGetMaxMemory, /* 1.2.15 */
.domainBlockStats = vzDomainBlockStats, /* 1.2.17 */
.domainBlockStatsFlags = vzDomainBlockStatsFlags, /* 1.2.17 */
.domainInterfaceStats = vzDomainInterfaceStats, /* 1.2.17 */ [*]
.domainMemoryStats = vzDomainMemoryStats, /* 1.2.17 */
.domainMigrateBegin3Params = vzDomainMigrateBegin3Params, /* 1.3.5 */
.domainMigrateConfirm3Params = vzDomainMigrateConfirm3Params, /* 1.3.5 */
[2] API calls that use thread shared sdkdom object
(marked [*] are included for a different reason)
.domainSuspend = vzDomainSuspend, /* 0.10.0 */
.domainResume = vzDomainResume, /* 0.10.0 */
.domainDestroy = vzDomainDestroy, /* 0.10.0 */
.domainShutdown = vzDomainShutdown, /* 0.10.0 */
.domainCreate = vzDomainCreate, /* 0.10.0 */
.domainCreateWithFlags = vzDomainCreateWithFlags, /* 1.2.10 */
.domainReboot = vzDomainReboot, /* 1.3.0 */
.domainDefineXML = vzDomainDefineXML, /* 0.10.0 */
.domainDefineXMLFlags = vzDomainDefineXMLFlags, /* 1.2.12 */ (update part)
.domainUndefine = vzDomainUndefine, /* 1.2.10 */
.domainAttachDevice = vzDomainAttachDevice, /* 1.2.15 */
.domainAttachDeviceFlags = vzDomainAttachDeviceFlags, /* 1.2.15 */
.domainDetachDevice = vzDomainDetachDevice, /* 1.2.15 */
.domainDetachDeviceFlags = vzDomainDetachDeviceFlags, /* 1.2.15 */
.domainSetUserPassword = vzDomainSetUserPassword, /* 1.3.6 */
.domainManagedSave = vzDomainManagedSave, /* 1.2.14 */
.domainSetMemoryFlags = vzDomainSetMemoryFlags, /* 1.3.4 */
.domainSetMemory = vzDomainSetMemory, /* 1.3.4 */
.domainRevertToSnapshot = vzDomainRevertToSnapshot, /* 1.3.5 */ [*]
.domainSnapshotCreateXML = vzDomainSnapshotCreateXML, /* 1.3.5 */ [*]
.domainMigratePerform3Params = vzDomainMigratePerform3Params, /* 1.3.5 */ [*]
.domainUpdateDeviceFlags = vzDomainUpdateDeviceFlags, /* 2.0.0 */
prlsdkHandleVmConfigEvent
[3] API calls that do not use thread shared sdkdom object
.domainManagedSaveRemove = vzDomainManagedSaveRemove, /* 1.2.14 */
.domainSnapshotNum = vzDomainSnapshotNum, /* 1.3.5 */
.domainSnapshotListNames = vzDomainSnapshotListNames, /* 1.3.5 */
.domainListAllSnapshots = vzDomainListAllSnapshots, /* 1.3.5 */
.domainSnapshotGetXMLDesc = vzDomainSnapshotGetXMLDesc, /* 1.3.5 */
.domainSnapshotNumChildren = vzDomainSnapshotNumChildren, /* 1.3.5 */
.domainSnapshotListChildrenNames = vzDomainSnapshotListChildrenNames, /* 1.3.5 */
.domainSnapshotListAllChildren = vzDomainSnapshotListAllChildren, /* 1.3.5 */
.domainSnapshotLookupByName = vzDomainSnapshotLookupByName, /* 1.3.5 */
.domainHasCurrentSnapshot = vzDomainHasCurrentSnapshot, /* 1.3.5 */
.domainSnapshotGetParent = vzDomainSnapshotGetParent, /* 1.3.5 */
.domainSnapshotCurrent = vzDomainSnapshotCurrent, /* 1.3.5 */
.domainSnapshotIsCurrent = vzDomainSnapshotIsCurrent, /* 1.3.5 */
.domainSnapshotHasMetadata = vzDomainSnapshotHasMetadata, /* 1.3.5 */
.domainSnapshotDelete = vzDomainSnapshotDelete, /* 1.3.5 */
[4] Known issues.
1. accidental errors on getting network statistics
2. race with simultaneous use of thread shared domain object on paths
of adding domain thru API and adding domain on vz sdk domain added event.
sdk domain handle is unique per connection so there is
no sense to query it again if we have it in vzDomObjPtr.
Side effect of prlsdkSdkDomainLookupByUUID is refreshing
domain config is of no use too as PrlVm_BeginEdit do it too.
Commit id '5e46d7d6' did not take into account that usage of a luks
volume will require usage of the master key encrypted passphrase for
a QEMU environment. So rather than allow creation of something that
won't be usable, just fail the creation.
Resolves a CI test integration failure with a RHEL6/Centos6 environment.
In order to use a LUKS encrypted device, the design decision was to
generate an encrypted secret based on the master key. However, commit
id 'da86c6c' missed checking for that specifically.
When qemuDomainSecretSetup was implemented, a design decision was made
to "fall back" to a plain text secret setup if the specific cipher was
not available (e.g. virCryptoHaveCipher(VIR_CRYPTO_CIPHER_AES256CBC))
as well as the QEMU_CAPS_OBJECT_SECRET. For the luks encryption setup
there is no fall back to the plaintext secret, thus if that gets set
up by qemuDomainSecretSetup, then we need to fail.
Also, while the qemuxml2argvtest has set the QEMU_CAPS_OBJECT_SECRET
bit, it didn't take into account the second requirement that the
ability to generate the encrypted secret is possible. So modify the
test to not attempt to run the luks-disk if we know we don't have
the encryption algorithm.
Any error happening after the hand shake in the lxc controller
will not result in a failure as errors are checked during the handshake.
Move the handshake after the last possible error.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1301021
Generate the luks command line using the AES secret key to encrypt the
luks secret. A luks secret object will be in addition to a an AES secret.
For hotplug, check if the encinfo exists and if so, add the AES secret
for the passphrase for the secret object used to decrypt the device.
Modify/augment the fakeSecret* in qemuxml2argvtest in order to handle
find a uuid or a volume usage with a specific path prefix in the XML
(corresponds to the already generated XML tests). Add error message
when the 'usageID' is not 'mycluster_myname'. Commit id '1d632c39'
altered the error message generation to rely on the errors from the
secret_driver (or it's faked replacement).
Add the .args output for adding the LUKS disk to the domain
Signed-off-by: John Ferlan <jferlan@redhat.com>
Soon we will be adding luks encryption support. Since a volume could require
both a luks secret and a secret to give to the server to use of the device,
alter the alias generation to create a slightly different alias so that
we don't have two objects with the same alias.
Signed-off-by: John Ferlan <jferlan@redhat.com>
Commit id 'a1344f70a' added AES secret processing for RBD when starting
up a guest. As such, when the hotplug code calls qemuDomainSecretDiskPrepare
an AES secret could be added to the disk about to be hotplugged. If an AES
secret was added, then the hotplug code would need to generate the secret
object because qemuBuildDriveStr would add the "password-secret=" to the
returned 'driveStr' rather than the base64 encoded password.
Signed-off-by: John Ferlan <jferlan@redhat.com>
Partially resolves:
https://bugzilla.redhat.com/show_bug.cgi?id=1301021
If the volume xml was looking to create a luks volume take the necessary
steps in order to make that happen.
The processing will be:
1. create a temporary file (virStorageBackendCreateQemuImgSecretPath)
1a. use the storage driver state dir path that uses the pool and
volume name as a base.
2. create a secret object (virStorageBackendCreateQemuImgSecretObject)
2a. use an alias combinding the volume name and "_luks0"
2b. add the file to the object
3. create/add luks options to the commandline (virQEMUBuildLuksOpts)
3a. at the very least a "key-secret=%s" using the secret object alias
3b. if found in the XML the various "cipher" and "ivgen" options
Signed-off-by: John Ferlan <jferlan@redhat.com>
When formatting the graphics data for TYPE_SPICE, check if the glisten
is NULL before blindly referencing
Found by Coverity
Signed-off-by: John Ferlan <jferlan@redhat.com>
A recent adjustment to qemuDomainAttachRNGDevice to properly cleanup
the props object after a qemuMonitorAddObject also would affect this
code. Alter the cleanup to be similar to RNG changes.
Based on recent review comment - rather than have a spate of goto failxxxx,
change to a boolean based model. Ensures that the original error can be
preserved and cleanup is a bit more orderly if more objects are added.
Based on recent review comment - rather than have a spate of goto failxxxx,
change to a boolean based model. Ensures that the original error can be
preserved and cleanup is a bit more orderly if more objects are added.
Based on recent review comment - rather than have a spate of goto failxxxx,
change to a boolean based model. Ensures that the original error can be
preserved and cleanup is a bit more orderly if more objects are added.
Based on recent review comment - rather than have a spate of goto failxxxx,
change to a boolean based model. Ensures that the original error can be
preserved and cleanup is a bit more orderly if more objects are added.
Based on recent review comment - rather than have a spate of goto failxxxx,
change to a boolean based model. Ensures that the original error can be
preserved and cleanup is a bit more orderly if more objects are added.
Commit da665fbd introduced the following condition to virLXCProcessEnsureRootFS
and openvzReadFSConf:
if (!(<some_var> = virDomainFSDefNew()) < 0)
which broke the build on fedora with GCC 5.3.1: "logical not is only applied to
the left hand side of comparison".
Signed-off-by: Erik Skultety <eskultet@redhat.com>
The commit da665fbd introduced virStorageSourcePtr inside the structure
_virDomainFSDef. This is causing an error when libvirt is being compiled.
make[3]: Entering directory `/media/julio/8d65c59c-6ade-4740-9cdc-38016a4cb8ae
/home/julio/Desktop/virt/libvirt/src'
CC security/virt_aa_helper-virt-aa-helper.o
security/virt-aa-helper.c: In function 'get_files':
security/virt-aa-helper.c:1087:13: error: passing argument 2 of 'vah_add_path'
from incompatible pointer type [-Werror]
if (vah_add_path(&buf, fs->src, "rw", true) != 0)
^
security/virt-aa-helper.c:732:1: note: expected 'const char *' but argument is
of type 'virStorageSourcePtr'
vah_add_path(virBufferPtr buf, const char *path, const char *perms, bool
recursive)
^
cc1: all warnings being treated as errors
Adding the attribute "path" from virStorageSourcePtr fixes this issue.
Signed-off-by: Julio Faracco <jcfaracco@gmail.com>
vz supports only a subset of tcp and udp parameters.
1. tcp type supports only 'raw' protocol.
2. udp type supports only same parameters of 'host' and 'service'
for 'bind' and 'connect'.
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>
After domain is in the domains list let's keep it there. This
is approach taken by qemu driver and vz vzDomainMigrateFinish3Params too.
It quite reasonable, driver domain object is fully constructed and
can be discovered by client later.
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>
9c14a9ab introduced vzNewDomain function to enlist libvirt domain
object before actually creating vz sdk domain. Fix should fix
race on same vz sdk domain added event where libvirt domain object is
enlisted too. But later eb5e9c1e added locked checks for
adding livirtd domain object to list on vz sdk domain added event.
Thus now approach of 9c14a9ab is unnecessary complicated.
See we have otherwise unuseful prlsdkGetDomainIds function only
to create minimal domain definition to create libvirt domain object.
Also vzNewDomain is difficult to use as it creates partially
constructed domain object.
Let's move back to original approach where prlsdkLoadDomain do
all the necessary job. Another benefit is that we can now
take driver lock for bare minimum and in single place. Reducing
locking time have small disadvatage of double parsing on race
conditions which is typical if domain is added thru vz driver.
Well we have this double parse inevitably with current vz sdk api
on any domain updates so i would not take it here seriously.
Performance events subscribtion is done before locked check and
therefore could be done twice on races but this is not the problem.
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>
Vz containers are able to use ploop volumes from storage pools
to work upon.
To use filesystem type volume, pool name and volume name should be
specifaed in <source> :
<filesystem type='volume' accessmode='passthrough'>
<driver type='ploop' format='ploop'/>
<source pool='guest_images' volume='TEST_POOL_CT'/>
<target dir='/'/>
</filesystem>
The information about pool and volume is stored in ct dom configuration:
<StorageURL>libvirt://localhost/pool_name/vol_name</StorageURL>
and can be easily obtained via PrlVmDevHd_GetStorageURL sdk call.
The only shorcoming: if storage pool is moved somewhere the ct
should be redefined in order to refresh the information aboot path
to root.hdd
Signed-off-by: Olga Krishtal <okrishtal@virtuozzo.com>
We do not need to check domainf fs type there,
because it is done in prlsdkCheckUnsupportedParams.
Signed-off-by: Olga Krishtal <okrishtal@virtuozzo.com>
New type of <devices> <filesystem type= 'volume'> is introduced.
This patch allows to use volumes for storing the filesystem, that is
accessed from the guest e.g. root directory for container.
To take advantage of volumes as a backend of filesystem volume
and pool names should be specified:
<filesystem type= 'volume'>
<source pool='pool name' volume='volume name'/>
Signed-off-by: Olga Krishtal <okrishtal@virtuozzo.com>
Adding domain to domain list on preparation step is not correct.
First domain is not fully constructed - domain definition is
missing. Second we can't use VIR_MIGRATE_PARAM_DEST_XML parameter
to parse definition as vz sdk can patch it by itself. Let's add/remove
domain on finish step. This is for synchronization purpose only so domain
is present/absent on destination after migration completion. Actually
domain object will probably be created right after actual vz sdk
migration start by vz sdk domain defined event.
We can not and should not sync domain cache on error path in finish step
of migration. We can not as we really don't know what is the reason of
cancelling and we should not as user should not make assumptions on
state on error path. What we should do is cleaning up temporary migration
state that is induced on prepare step but we don't have one. Thus
cancellation should be noop.
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>
Signed-off-by: Maxim Nestratov <mnestratov@virtuozzo.com>
libvirt domain defined event is issued only on correspondent vz sdk
event. But in case event delivered before domain is added to
domain list we can mistakenly skip this event if prlsdkNewDomainByHandle
return NULL in case of domain is discovered in the list under
the driver lock. Let's return domain object in this case.
Now prlsdkNewDomainByHandle returns NULL only in case of
error which is more convinient.
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>
The first version of migration cookie was rather dumb resulting
in passing empty or unused fields here and there. Add flags to
specify what to bake to and eat from cookie so we deal only
with meaningful data. However for backwards compatibility
we still need to pass at least some faked fields sometimes.
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>
Older libvirt versions send persistent XML in a migration cookie even
when VIR_MIGRATE_PERSIST_DEST flag is not used, but current libvirt
properly fails if the cookie contains unexpected flags. Thus migration
from old libvirt fails with
internal error: Unsupported migration cookie feature persistent
unless VIR_MIGRATE_PERSIST_DEST flag is set.
https://bugzilla.redhat.com/show_bug.cgi?id=1320500
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
A nodedev device definition like this is required for testing
NodeDeviceCreateXML and NodeDeviceDestroy. So unless it's part
of the stock test:///default set there's no way to actually
invoke those functions for the default URI
Convert the individual XML documents into one big XML document
in the format expected by the non-default test://$PATH URI, and
use the same internal helpers for assembling the driver contents.
virConfGetValueLLong() errors out if the value is too big to
fit into a long long integer, but claims the supported range
to be (0,LLONG_MAX) instead of (LLONG_MIN,LLONG_MAX).
When parsing numeric values, we always store them as unsigned
unless they're negative. We can use this fact to simplify the
logic by removing a bunch of unnecessary checks.
Commit 6381c89f8c changed virConfValue to store long long
integers instead of long integers; however, the temporary variable
used in virConfParseLong() was not updated accordingly, causing
trouble for 32-bit machines.
In preparation to tracking which USB addresses are occupied.
Introduce two helper functions for printing the port path
as a string and appending it to a virBuffer.
We were requiring a USB port path in the schema, but not enforcing it.
Omitting the USB port would lead to libvirt formatting it as (null).
Such domain cannot be started and will disappear after libvirtd restart
(since it cannot parse back the XML).
Only format the port if it has been specified and mark it as optional
in the XML schema.
Migration to an older libvirt (pre v1.3.0-175-g7140807) is broken
because older versions of libvirt generated different channel paths and
they didn't drop the default paths when parsing domain XMLs. We'd get
such a nice error message:
internal error: process exited while connecting to monitor:
2016-07-08T15:28:02.665706Z qemu-kvm: -chardev socket,
id=charchannel0,path=/var/lib/libvirt/qemu/channel/target/
domain-3-nest/org.qemu.guest_agent.0,server,nowait: Failed to bind
socket to /var/lib/libvirt/qemu/channel/target/domain-3-nest/
org.qemu.guest_agent.0: No such file or directory
That said, we should not even format the default paths when generating a
migratable XML.
https://bugzilla.redhat.com/show_bug.cgi?id=1320470
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Playing directly with our live definition, updating it, and reverting it
back once we are done is very nice and it's quite dangerous too. Let's
just make a copy of the domain definition if needed and do all tricks on
the copy.
https://bugzilla.redhat.com/show_bug.cgi?id=1320470
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
If size_t is the same size as long long, then we can skip
some of the range checks. This avoids triggering some
bogus compiler warning messages.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The virConf 'l' field is a 'signed long long', so whenever
the 'type' field is VIR_CONF_ULONG, we should explicitly cast
'l' to a 'unsigned long long' before doing range checks.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
This function tries to get a ssize_t value from a config file.
But before returning it, it checks whether the value would fit in
ssize_t and if not an error is printed out among with the range
for the ssize_t type. However, on some platforms SSIZE_MAX may
actually be a signed long type:
util/virconf.c: In function 'virConfGetValueSSizeT':
util/virconf.c:1268:9: error: format '%zd' expects argument of type 'signed size_t', but argument 9 has type 'long int' [-Werror=format=]
virReportError(VIR_ERR_INTERNAL_ERROR,
^
$ grep -r SSIZE_MAX /usr/include/
/usr/include/bits/posix1_lim.h:#ifndef SSIZE_MAX
/usr/include/bits/posix1_lim.h:# define SSIZE_MAX LONG_MAX
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
MinGW complained that we might be dereferencing a NULL pointer. While
that can't be true, the logic certainly allows for that.
../../src/conf/domain_conf.c: In function 'virDomainDefPostParse':
../../src/conf/domain_conf.c:4224:18: error: potential null pointer dereference [-Werror=null-dereference]
if (!vcpu->online && vcpu->cpumask) {
~~~~^~~~~~~~
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
MinGW complained that we might be dereferencing a NULL pointer. While
that can't be true, the logic certainly allows for that.
src/conf/domain_conf.c: In function 'virDomainDefGetVcpuPinInfoHelper':
src/conf/domain_conf.c:1545:17: error: potential null pointer dereference [-Werror=null-dereference]
if (vcpu->cpumask)
~~~~^~~~~~~~~
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
IPv6 RA always contains an implicit default route via
the link-local address of the source of RA. This forces
the guest to install a route via isolated network, which
may disturb the guest's networking in case of multiple interfaces.
More info in 013427e6e7.
The validity of this route is controlled by "default [route] lifetime"
field of RA. If the lifetime is set to 0 seconds, then no route
is installed by receiver.
dnsmasq 2.67+ supports "ra-param=<interface>,<RA interval>,<default
lifetime>" option. We pass "ra-param=*,0,0"
(here, RA_interval=0 means default) to disable default gateway in RA
for isolated networks.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1354238
So we spend some time and effort constructing perfect file name
for an automatic coredump of a domain, but then just leak it and
use the domain name anyway. This is probably due to a silly
mistake that slipped even through review.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
At least with systemd v210, NOTIFY_SOCKET is abstact, e.g.
@/org/freedesktop/systemd1/notify. sendmsg() fails on such a socket
with "Connection refused". The unix(7) man page contains the following
details wrt abstract socket addresses
abstract: an abstract socket address is distinguished (from a
pathname socket) by the fact that sun_path[0] is a null byte
('\0'). The socket's address in this namespace is given by the
additional bytes in sun_path that are covered by the specified
length of the address structure. (Null bytes in the name have
no special significance.)
So we need to be more precise about the address length, setting it to
the sizeof sa_family_t + length of address copied to sun_path instead
of setting it to the sizeof the entire sockaddr_un struct.
Resolves: https://bugzilla.opensuse.org/show_bug.cgi?id=987668
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
When fetching domains with virConnectListAllDomains() and when filtering
by snapshot existence is requested the ESX driver first lists all the
domains and then check one-by-one for snapshot existence. This process
takes unnecessarily long time.
To significantly improve the time necessary to finish the query we can
request the snapshot related info directly when querying the list of
domains from VMware.
Signed-off-by: Tomáš Golembiovský <tgolembi@redhat.com>
In an unlikely event of execve() failing, the virCommandExec()
function does not report any error, even though checks that are
at the beginning of the function are verbose when failing.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
The modification of .volWipe callback wipes ploop volume using one of
given wiping algorithm: dod, nnsa, etc.
However, in case of ploop volume we need to reinitialize root.hds and DiskDescriptor.xml.
v2:
- added check on ploop tools presens
- virCommandAddArgFormat changed to virCommandAddArg
Signed-off-by: Olga Krishtal <okrishtal@virtuozzo.com>
Check whether QEMU supports -device intel-iommu
Note that the presence of this option does not mean that it's
usable because of a bug in earlier QEMU versions, but it's
better than nothing.
https://bugzilla.redhat.com/show_bug.cgi?id=1235580
Currently many users of virConf APIs are defining the same
macros for calling virConfValue() and then doing type
checking. To remove this repeated code, add a set of
typesafe accessor methods.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
If the config file does not end with a \n, the parser will append
one. When re-allocating the array though, it is mistakenly assuming
that 'len' is the length including the trailing NUL, but it does
not. So we must add 2 to len, when reallocating, not 1.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
This one's a bit more complicated. In qemuProcessPrepareDomain()
a master key for encrypting secret for ciphered disks is created.
This object lives within qemuDomainObjPrivate object. It is freed
in qemuProcessStop(), but if nobody calls it (for instance like
our qemuxml2argvtest does), the key object leaks.
==17078== 32 bytes in 1 blocks are definitely lost in loss record 633 of 707
==17078== at 0x4C2C070: calloc (vg_replace_malloc.c:623)
==17078== by 0xAD924DF: virAllocN (viralloc.c:191)
==17078== by 0x5050BA6: virCryptoGenerateRandom (qemuxml2argvmock.c:166)
==17078== by 0x453DC8: qemuDomainMasterKeyCreate (qemu_domain.c:678)
==17078== by 0x47A36B: qemuProcessPrepareDomain (qemu_process.c:4913)
==17078== by 0x47C728: qemuProcessCreatePretendCmd (qemu_process.c:5542)
==17078== by 0x433698: testCompareXMLToArgvFiles (qemuxml2argvtest.c:332)
==17078== by 0x4339AC: testCompareXMLToArgvHelper (qemuxml2argvtest.c:413)
==17078== by 0x446E7A: virTestRun (testutils.c:179)
==17078== by 0x445BD9: mymain (qemuxml2argvtest.c:2022)
==17078== by 0x44886F: virTestMain (testutils.c:969)
==17078== by 0x445D9B: main (qemuxml2argvtest.c:2036)
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Just like every other qemuBuild*CommandLine() function, this uses
a buffer to hold partial cmd line strings too. However, if
there's an error, the control jumps to 'cleanup' label leaving
the buffer behind and thus leaking it.
==2013== 1,006 bytes in 1 blocks are definitely lost in loss record 701 of 711
==2013== at 0x4C29F80: malloc (vg_replace_malloc.c:296)
==2013== by 0x4C2C32F: realloc (vg_replace_malloc.c:692)
==2013== by 0xAD925A8: virReallocN (viralloc.c:245)
==2013== by 0xAD95EA8: virBufferGrow (virbuffer.c:130)
==2013== by 0xAD95F78: virBufferAdd (virbuffer.c:165)
==2013== by 0x5097F5: qemuBuildCpuModelArgStr (qemu_command.c:6339)
==2013== by 0x509CC3: qemuBuildCpuCommandLine (qemu_command.c:6437)
==2013== by 0x51142C: qemuBuildCommandLine (qemu_command.c:9174)
==2013== by 0x47CA3A: qemuProcessCreatePretendCmd (qemu_process.c:5546)
==2013== by 0x433698: testCompareXMLToArgvFiles (qemuxml2argvtest.c:332)
==2013== by 0x4339AC: testCompareXMLToArgvHelper (qemuxml2argvtest.c:413)
==2013== by 0x446E7A: virTestRun (testutils.c:179)
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
When storage secret is parsed in virStorageEncryptionSecretParse(),
virSecretLookupParseSecret() which allocates some memory. This is
however never freed.
==21711== 134 bytes in 6 blocks are definitely lost in loss record 70 of 85
==21711== at 0x4C29F80: malloc (vg_replace_malloc.c:296)
==21711== by 0xBCA0356: xmlStrndup (in /usr/lib64/libxml2.so.2.9.4)
==21711== by 0xA9F432E: virXMLPropString (virxml.c:479)
==21711== by 0xA9D25B0: virSecretLookupParseSecret (virsecret.c:70)
==21711== by 0xA9D616E: virStorageEncryptionSecretParse (virstorageencryption.c:172)
==21711== by 0xA9D66B2: virStorageEncryptionParseXML (virstorageencryption.c:281)
==21711== by 0xA9D68DF: virStorageEncryptionParseNode (virstorageencryption.c:338)
==21711== by 0xAA12575: virDomainDiskDefParseXML (domain_conf.c:7606)
==21711== by 0xAA2CAC6: virDomainDefParseXML (domain_conf.c:16658)
==21711== by 0xAA2FC75: virDomainDefParseNode (domain_conf.c:17472)
==21711== by 0xAA2FAE4: virDomainDefParse (domain_conf.c:17419)
==21711== by 0xAA2FB72: virDomainDefParseFile (domain_conf.c:17443)
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Setting up cgroups and other things for all kinds of threads (the
emulator thread, vCPU threads, I/O threads) was copy-pasted every time
new thing was added. Over time each one of those functions changed a
bit differently. So create one function that does all that setup and
start using it, starting with I/O thread setup. That will shave some
duplicated code and maybe fix some bugs as well.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
The code in qemuDomainObjPrivateXMLParseVcpu for parsing
the 'idstr' string was comparing the overall boolean
result against 0 which was always true
qemu/qemu_domain.c: In function 'qemuDomainObjPrivateXMLParseVcpu':
qemu/qemu_domain.c:1482:59: error: comparison of constant '0' with boolean expression is always false [-Werror=bool-compare]
if ((idstr && virStrToLong_uip(idstr, NULL, 10, &idx)) < 0 ||
^
It was further performing two distinct error checks in
the same conditional and reporting a single error message,
which was misleading in one of the two cases.
This splits the conditional check into two parts with
distinct error messages and fixes the logic error.
Fixes the bug in
commit 5184f398b4
Author: Peter Krempa <pkrempa@redhat.com>
Date: Fri Jul 1 14:56:14 2016 +0200
qemu: Store vCPU thread ids in vcpu private data objects
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
An error in virHostCPUGetKVMMaxVCPUs() means we've been unable
to access /dev/kvm, or we're running on a platform that doesn't
support KVM in the first place.
If that's the case, we shouldn't ignore the error and report
domcapabilities even though we know the user won't be able to
start any KVM guest.
If we don't HAVE_LINUX_KVM_H, we can't query /dev/kvm to discover
the limits on the number of vCPUs, so we report an error and
return a negative value instead.
Allow to store driver specific data on a per-vcpu basis.
Move of the virDomainDef*Vcpus* functions was necessary as
virDomainXMLOptionPtr was declared below this block and I didn't want to
split the function headers.
Most callers make sure that it's never called with an out of range vCPU.
Every other caller reports a different error explicitly. Drop the error
reporting and clean up some dead code paths.
A simple getopt-based argument parser is added for the /usr/sbin/bhyveload
command, loosely based on its argument parser.
The boot disk is guessed by iterating over all
disks and matching their sources. If any non-default arguments are found,
def->os.bootloaderArgs is set accordingly, and the bootloader is treated as a
custom bootloader.
Custom bootloader are supported by setting the def->os.bootloader and
def->os.bootloaderArgs accordingly
grub-bhyve is also treated as a custom bootloader. Since we don't get the
device map in the native format anyways, we can't reconstruct the complete
boot order. While it is possible to check what type the grub boot disk is by
checking if the --root argument is "cd" or "hd0,msdos1", and then just use the
first disk found, implementing the grub-bhyve argument parser as-is in the
grub-bhyve source would mean adding a dependency to argp or duplicating lots
of the code of argp. Therefore it's not really worth implementing that now.
Signed-off-by: Fabian Freyer <fabian.freyer@physik.tu-berlin.de>
A simpe getopt-based argument parser is added for the /usr/sbin/bhyve command,
loosely based on its argument parser, which reads the following from the bhyve
command line string:
* vm name
* number of vcpus
* memory size
* the time offset (UTC or localtime)
* features:
* acpi
* ioapic: While this flag is deprecated in FreeBSD r257423, keep checking for
it for backwards compatibiility.
* the domain UUID; if not explicitely given, one will be generated.
* lpc devices: for now only the com1 and com2 are supported. It is required for
these to be /dev/nmdm[\d+][AB], and the slave devices are automatically
inferred from these to be the corresponding end of the virtual null-modem
cable: /dev/nmdm<N>A <-> /dev/nmdm<N>B
* PCI devices:
* Disks: these are numbered in the order they are found, for virtio and ahci
disks separately. The destination is set to sdX or vdX with X='a'+index;
therefore only 'z'-'a' disks are supported.
Disks are considered to be block devices if the path
starts with /dev, otherwise they are considered to be files.
* Networks: only tap devices are supported. Since it isn't possible to tell
the type of the network, VIR_DOMAIN_NET_TYPE_ETHERNET is assumed, since it
is the most generic. If no mac is specified, one will be generated.
Signed-off-by: Fabian Freyer <fabian.freyer@physik.tu-berlin.de>
First, remove escaped newlines and split up the string into an argv-list for
the bhyve and loader commands, respectively. This is done by iterating over the
string splitting it by newlines, and then re-iterating over each line,
splitting it by spaces.
Since this code reuses part of the code of qemu_parse_command.c
(in bhyveCommandLine2argv), add the appropriate copyright notices.
Signed-off-by: Fabian Freyer <fabian.freyer@physik.tu-berlin.de>
As there is an explicit constructor for the special case of empty
bitmaps, we should mention that the generic constructors rejects the
creation of empty bitmaps.
Reviewed-by: Boris Fiuczynski <fiuczy@linux.vnet.ibm.com>
Reviewed-by: Sascha Silbe <silbe@linux.vnet.ibm.com>
Reviewed-by: Bjoern Walk <bwalk@linux.vnet.ibm.com>
Signed-off-by: Marc Hartmayer <mhartmay@linux.vnet.ibm.com>
Before the variable 'bits' was initialized with 0 (commit
3470cd860d), the following bug was
possible.
A function call with an empty bitmap leads to undefined
behavior. Because if 'bitmap->map_len == 0' 'unusedBits' will be <= 0
and 'sz == 1'. So the non global and non static variable 'bits' would
have never been set. Consequently the check 'bits == 0' results in
undefined behavior.
This patch clarifies the current version of the function by handling the
empty bitmap explicitly. Also, for an empty bitmap there is obviously no
bit set so we can just return -1 (indicating no bit set) right away. The
explicit check for 'bits == 0' after the loop is unnecessary because we
only get to this point if no set bit was found.
Reviewed-by: Boris Fiuczynski <fiuczy@linux.vnet.ibm.com>
Reviewed-by: Sascha Silbe <silbe@linux.vnet.ibm.com>
Reviewed-by: Bjoern Walk <bwalk@linux.vnet.ibm.com>
Signed-off-by: Marc Hartmayer <mhartmay@linux.vnet.ibm.com>
Otherwise migration during which we didn't send client_migrate_info QMP
command will get stuck waiting for SPICE migration to finish if libvirtd
sent the QMP command in a previous migration attempt.
Broken by bd7c8a69.
https://bugzilla.redhat.com/show_bug.cgi?id=1151723
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
People debugging guest OS boot processes and reported that
the default 128 KB size is too small to capture an entire
boot up sequence. Increase the default size to 2 MB which
should allow capturing a full boot up even with verbose
debugging.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
Currently virtlogd has a hardcoded max file size of 128kb
and max of 3 backups. This adds two new config parameters
to /etc/libvirt/virtlogd.conf to let these be customized.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
After 27726d8c21 a privateData is allocated in
virDomainHostdevDefAlloc(). However, the counter part - freeing
them in Free() is missing which leads to the following memory
leak:
==6489== 24 bytes in 1 blocks are definitely lost in loss record 684 of 1,003
==6489== at 0x4C2C070: calloc (vg_replace_malloc.c:623)
==6489== by 0x54B7C94: virAllocVar (viralloc.c:560)
==6489== by 0x5517BE6: virObjectNew (virobject.c:193)
==6489== by 0x1B400121: qemuDomainHostdevPrivateNew (qemu_domain.c:798)
==6489== by 0x5557B24: virDomainHostdevDefAlloc (domain_conf.c:2152)
==6489== by 0x5575578: virDomainHostdevDefParseXML (domain_conf.c:12709)
==6489== by 0x5582292: virDomainDefParseXML (domain_conf.c:16995)
==6489== by 0x5583C98: virDomainDefParseNode (domain_conf.c:17470)
==6489== by 0x5583B07: virDomainDefParse (domain_conf.c:17417)
==6489== by 0x5583B95: virDomainDefParseFile (domain_conf.c:17441)
==6489== by 0x55A3F24: virDomainObjListLoadConfig (virdomainobjlist.c:465)
==6489== by 0x55A43E6: virDomainObjListLoadAllConfigs (virdomainobjlist.c:596)
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
This is preferrable to -nographic which (in addition to disabling
graphics output) redirects the serial port to stdio and on OpenBIOS
enables the firmware's serial console.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Libxl is the last user and I don't have the toolchain prepared to
compile the libxl driver. Move it to the libxl driver to avoid having to
refactor the code.
This is just a convenience method for discarding a list of filters instead of
using a 'for' loop everywhere. It is safe to pass -1 as the number of elements
in the list as well as passing NULL as list reference.
Signed-off-by: Erik Skultety <eskultet@redhat.com>
Provide a separate method to free a logging filter object. This will come handy
once a method to create an individual logging filter object is introduced.
Signed-off-by: Erik Skultety <eskultet@redhat.com>
This is just a convenience method for discarding a list of outputs instead of
using a 'for' loop everywhere. It is safe to pass -1 as the number of elements
in the list as well as passing NULL as list reference.
Signed-off-by: Erik Skultety <eskultet@redhat.com>
Provide a separate method to free a logging output object. This will come handy
once a method to create an individual logging output object is introduced.
Signed-off-by: Erik Skultety <eskultet@redhat.com>
Same as with outputs; since the operations will be further divided into smaller
tasks, creating a filter will become a separate operation that will return
a reference to a newly created filter.
Signed-off-by: Erik Skultety <eskultet@redhat.com>
Right now, we define outputs one after another. However, the correct flow
should be to define a set of outputs as a whole unit. Therefore each output
should be first created, placed into an array/list and the list will be
defined. Output creation should be a separate operation, so an output will be
returned by a reference. From that perspective, it makes perfect sense to
only store pointers to actual outputs.
Signed-off-by: Erik Skultety <eskultet@redhat.com>
In this particular case, reset is meant as clearing the whole list of
outputs/filters, not resetting it to a predefined default setting. Looking at
it from that perspective, returning the number of records removed doesn't help
the caller in any way (not that any of the callers would actually check for
it). Well, callers could detect an error from the number of successfully
removed records, but the only thing that can fail in virLogReset is force
closing a file descriptor in which case the error isn't propagated back to
virLogReset anyway. Conclusion: there is no practical use for having a return
type of 'int' rather than 'void' in this case.
Signed-off-by: Erik Skultety <eskultet@redhat.com>
Due to the way the hardware works, KVM on ppc64 always requires
memory locking; however, that is not the case for non-KVM ppc64
guests, eg. ppc64 guests that are running on x86_64 with TCG.
Only require memory locking for ppc64 guests if they are using
KVM or, as it's the case for all architectures, they have host
devices assigned using VFIO.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1350772
For type='ethernet' interfaces only.
(This patch had been pushed earlier in
commit 0b4645a7e0, but was reverted in
commit 84d47a3cce because it had been
accidentally pushed during the freeze for release 2.0.0)
(This patch had been pushed earlier in
commit cd5c9f21de, but was reverted in
commit 1549f16832 because it had been
accidentally pushed during the freeze for release 2.0.0)
This will apply to any IP address setting that uses
virNetDevIPInfoAddToDev() (which so far is only the guest-side of LXC
type='ethernet' interfaces).
(This patch had been pushed earlier in
commit cb20f989df, but was reverted in
commit cba06aea8d because it had been
accidentally pushed during the freeze for release 2.0.0)
This is place as a sub-element of <source>, where other aspects of the
host-side connection to the network device are located (network or
bridge name, udp listen port, etc). It's a bit odd that the interface
we're configuring with this info is itself named in <target dev='x'/>,
but that ship sailed long ago:
<interface type='ethernet'>
<mac address='00:16:3e:0f:ef:8a'/>
<source>
<ip address='192.168.122.12' family='ipv4'
prefix='24' peer='192.168.122.1'/>
<ip address='192.168.122.13' family='ipv4' prefix='24'/>
<route family='ipv4' address='0.0.0.0'
gateway='192.168.122.1'/>
<route family='ipv4' address='192.168.124.0' prefix='24'
gateway='192.168.124.1'/>
</source>
</interface>
In practice, this will likely only be useful for type='ethernet', so
its presence in any other type of interface is currently forbidden in
the generic device Validate function (but it's been put into the
general population of virDomainNetDef rather than the
ethernet-specific union member so that 1) we can more easily add the
capability to other types if needed, and 2) we can retain the info
when set to an invalid interface type all the way through to
validation and report a proper error, rather than just ignoring it
(which is currently what happens for many other type-specific
settings).
(NB: The already-existing configuration of IP info for the guest-side
of interfaces is in subelements directly under <interface>, and the
name of the guest-side interface (when configurable) is in <guest
dev='x'/>).
(This patch had been pushed earlier in
commit fe6a77898a, but was reverted in
commit d658456530 because it had been
accidentally pushed during the freeze for release 2.0.0)
The peer attribute is used to set the property of the same name in the
interface IP info:
<interface type='ethernet'>
...
<ip family='ipv4' address='192.168.122.5'
prefix='32' peer='192.168.122.6'/>
...
</interface>
Note that this element is used to set the IP information on the
*guest* side interface, not the host side interface - that will be
supported in an upcoming patch.
(This patch now has quite a history: it was originally pushed in
commit 690969af, which was subsequently reverted in commit 1d14b13f,
then reworked and pushed (along with a lot of other related/supporting
patches) in commit 93135abf1; however *that* commit had been
accidentally pushed during dev. freeze for release 2.0.0, so it was
again reverted in commit f6acf039f0).
Signed-off-by: Vasiliy Tolstov <v.tolstov@selfip.ru>
Signed-off-by: Laine Stump <laine@laine.org>
This patch takes the code out of
lxcContainerRenameAndEnableInterfaces() that adds all IP addresses and
IP routes to the interface, and puts it into a utility function
virNetDevIPInfoAddToDev() in virnetdevip.c so that it can be used by
anyone.
One small change in functionality -
lxcContainerRenameAndEnableInterfaces() previously would add all IP
addresses to the interface while it was still offline, then set the
interface online, and then add the routes. Because I don't want the
utility function to set the interface online, I've moved this up so
the interface is first set online, then IP addresses and routes are
added. This is the same order that the network service from
initscripts (in ifup-ether) does it, so it shouldn't pose any problem
(and hasn't, in the tests that I've run).
(This patch had been pushed earlier in commit
f1e0d0da11, but was reverted in commit
05eab47559 because it had been
accidentally pushed during the freeze for release 2.0.0)
Introduce a helper to help determine if a disk src could be possibly used
for a disk secret... Going to need this for hot unplug.
Signed-off-by: John Ferlan <jferlan@redhat.com>
In order to use more common code and set up for a future type, modify the
encryption secret to allow the "usage" attribute or the "uuid" attribute
to define the secret. The "usage" in the case of a volume secret would be
the path to the volume as dictated by the backwards compatibility brought
on by virStorageGenerateQcowEncryption where it set up the usage field as
the vol->target.path and didn't allow someone to provide it. This carries
into virSecretObjListFindByUsageLocked which takes the secret usage attribute
value from from the domain disk definition and compares it against the
usage type from the secret definition. Since none of the code dealing
with qcow/qcow2 encryption secrets uses usage for lookup, it's a mostly
cosmetic change. The real usage comes in a future path where the encryption
is expanded to be a luks volume and the secret will allow definition of
the usage field.
This code will make use of the virSecretLookup{Parse|Format}Secret common code.
Signed-off-by: John Ferlan <jferlan@redhat.com>
Add a new secret type known as "passphrase" - it will handle adding the
secret objects that need a passphrase without a specific username.
The format is:
<secret ...>
<uuid>...</uuid>
...
<usage type='passphrase'>
<name>mumblyfratz</name>
</usage>
</secret>
Signed-off-by: John Ferlan <jferlan@redhat.com>
Since the virSecretDefParseUsage ensures each of the fields is present,
no need to check during virSecretDefFormatUsage (also virBufferEscapeString
is a no-op with a NULL argument).
Signed-off-by: John Ferlan <jferlan@redhat.com>
A helper that will execute a callback on every USB device
in the domain definition.
With an ability to skip USB hubs, since we will want to treat
them differently in some cases.
I'm not sure why our code claimed "-boot menu=on" cannot be used in
combination with per-device bootindex, but it was proved wrong about
four years ago by commit 8c952908. Let's always use bootindex when QEMU
supports it.
https://bugzilla.redhat.com/show_bug.cgi?id=1323085
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Pretending (partial) support for something we don't understand is risky.
Reporting a failure is much better.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
virTypedParameterAssign steals the string rather than copying it into
the typed parameter and thus freeing it leads to a crash when attempting
to serialize the results.
This was introduced in commit 9f50f6e2 and later made an universal
helper in 32e6339c.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1351473
Some code paths already assume that it is allocated since it was always
allocated by virDomainPerfDefParseXML. Make it member of virDomainDef
directly so that we don't have to allocate it all the time.
This fixes crash when attempting to connect to an existing process via
virDomainQemuAttach since we would not allocate it in that code path.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1350688
Ensure that the given controller and all controllers with a smaller
index exist; there must not be any missing index in between.
Reviewed-by: Boris Fiuczynski <fiuczy@linux.vnet.ibm.com>
Reviewed-by: Bjoern Walk <bwalk@linux.vnet.ibm.com>
Signed-off-by: Marc Hartmayer <mhartmay@linux.vnet.ibm.com>
The commit "qemu: hot-plug: Assume support for -device in
qemuDomainAttachSCSIDisk" dropped the code for the automatic SCSI
controller creation used in SCSI disk hot-plugging. If we are
hot-plugging a SCSI disk to a domain and there is no proper SCSI
controller defined, it results in an "error: internal error: Could not
find scsi controller with index X required for device" error.
For that reason reverting a hunk of the commit
d4d32005d6.
This patch also adds an extra comment to the code to clarify the
loop.
Reviewed-by: Boris Fiuczynski <fiuczy@linux.vnet.ibm.com>
Reviewed-by: Bjoern Walk <bwalk@linux.vnet.ibm.com>
Signed-off-by: Marc Hartmayer <mhartmay@linux.vnet.ibm.com>
CVE-2016-5008
Setting an empty graphics password is documented as a way to disable
VNC/SPICE access, but QEMU does not always behaves like that. VNC would
happily accept the empty password. Let's enforce the behavior by setting
password expiration to "now".
https://bugzilla.redhat.com/show_bug.cgi?id=1180092
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Similarly to what virsh virt-login-shell do, call virAdmInitialize prior to
initializing an event loop and initializing the error handler. Commit 97973ebb7
described and fixed an identical issue for libvirt_lxc.
Since virAdmInitialize becomes a public API after applying this patch,
the symbol is also added to public syms and the doc string of the method is
slightly enhanced analogically to virInitialize.
Signed-off-by: Erik Skultety <eskultet@redhat.com>
virNetServerClientGetInfo returns the client's remote address
as a string, which is a part of the client object.
Use VIR_STRDUP to make a copy which can be freely accessed
even after the virNetServerClient object is unlocked.
To reproduce, put a sleep between virObjectUnlock in
virNetServerClientGetInfo and virTypedParamsAddString in
adminClientGetInfo, then close the queried connection during
that sleep.
https://bugzilla.redhat.com/show_bug.cgi?id=1316370
Consider the following disk for a domain:
<disk type='volume' device='cdrom'>
<driver name='qemu' type='raw'/>
<auth username='libvirt'>
<secret type='iscsi' usage='libvirtiscsi'/>
</auth>
<source pool='iscsi-secret-pool' volume='unit:0:0:0' mode='direct' startupPolicy='optional'/>
<target dev='sda' bus='scsi'/>
<readonly/>
<address type='drive' controller='0' bus='0' target='0' unit='0'/>
</disk>
Now, startupPolicy is currently not allowed for iscsi disks, so
one would expect an error message to be thrown. But what a
surprise is waiting for users if they try to start up such
domain:
==15724== Invalid free() / delete / delete[] / realloc()
==15724== at 0x4C2B1F0: free (vg_replace_malloc.c:473)
==15724== by 0x54B7A69: virFree (viralloc.c:582)
==15724== by 0x552DC90: virStorageAuthDefFree (virstoragefile.c:1549)
==15724== by 0x552F023: virStorageSourceClear (virstoragefile.c:2055)
==15724== by 0x552F054: virStorageSourceFree (virstoragefile.c:2067)
==15724== by 0x55556AA: virDomainDiskDefFree (domain_conf.c:1562)
==15724== by 0x5557ABE: virDomainDefFree (domain_conf.c:2547)
==15724== by 0x1B43CC42: qemuProcessStop (qemu_process.c:5918)
==15724== by 0x1B43BA2E: qemuProcessStart (qemu_process.c:5511)
==15724== by 0x1B48993E: qemuDomainObjStart (qemu_driver.c:7050)
==15724== by 0x1B489B9A: qemuDomainCreateWithFlags (qemu_driver.c:7104)
==15724== by 0x1B489C01: qemuDomainCreate (qemu_driver.c:7122)
==15724== Address 0x21cfbb90 is 0 bytes inside a block of size 48 free'd
==15724== at 0x4C2B1F0: free (vg_replace_malloc.c:473)
==15724== by 0x54B7A69: virFree (viralloc.c:582)
==15724== by 0x552DC90: virStorageAuthDefFree (virstoragefile.c:1549)
==15724== by 0x12D1C8D4: virStorageTranslateDiskSourcePool (storage_driver.c:3475)
==15724== by 0x1B4396E4: qemuProcessPrepareDomain (qemu_process.c:4896)
==15724== by 0x1B43B880: qemuProcessStart (qemu_process.c:5466)
==15724== by 0x1B48993E: qemuDomainObjStart (qemu_driver.c:7050)
==15724== by 0x1B489B9A: qemuDomainCreateWithFlags (qemu_driver.c:7104)
==15724== by 0x1B489C01: qemuDomainCreate (qemu_driver.c:7122)
==15724== by 0x561CA97: virDomainCreate (libvirt-domain.c:6787)
==15724== by 0x12B6FD: remoteDispatchDomainCreate (remote_dispatch.h:4116)
==15724== by 0x12B61A: remoteDispatchDomainCreateHelper (remote_dispatch.h:4092)
The problem is, in virStorageTranslateDiskSourcePool disk
def->src->auth is freed, but the pointer is not set to NULL. So
later, when qemuProcessStop starts to free the domain definition,
virStorageAuthDefFree() tries to free the memory again, instead
of jumping out immediately.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
