| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
| |
Signed-off-by: Roman Rakus <rrakus@redhat.com>
|
|
|
|
| |
Signed-off-by: Roman Rakus <rrakus@redhat.com>
|
|
|
|
|
|
|
|
|
|
| |
New Providers:
* LMI_BaseboardProvider
* LMI_BaseboardContainerProvider
Other Changes:
* Removed useless providers_SRCS from CMakeLists.txt
* Better, human friendly chassis name if type is not avaiable
|
|
|
|
|
|
|
| |
Fixes Coverity issues:
* 10153
* 10154
* 10155
|
|
|
|
|
| |
Previous commit renamed LMI_ConcreteJob to LMI_PowerConcreteJob but
didn't rename some symbols in power.c causing compilation to fail.
|
|
|
|
|
|
| |
There is collision with LMI_ConcreteJob provided by 30_LMI_Jobs.mof.
This is just workaround, proper fix would be to use the jobs properly.
The jobs in power provider were just an attempt and need to be fixed.
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| | |
As discussed, we want to expose a simple API for starters here
with just a few methods and properties. If we need to grow from
there it should be in the future when we have more experience
modeling an appropriate CIM model which integrates well with the
patterns and style of other CIM models.
|
| |\ |
|
| | |
| | |
| | |
| | |
| | |
| | | |
We should cleanly shut down all threads when the provider is unloaded,
just to be nice to CIMOMs.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
KReturnInstance macro calls return directly on failure which is bad
because there is no possibility to do cleanups or other error handling.
ReturnInstance macro returns plain boolean to indicate if the instance
is returned properly.
|
| |/ |
|
|/
|
|
|
| |
New Provider:
* LMI_ChassisProvider
|
| |
|
| |
|
|\ |
|
| |
| |
| |
| |
| | |
passed wrong parameter to function, when handling association to
Linux_ComputerSystem
|
| |
| |
| |
| |
| | |
Undefined variable used in logging statement would generate
error for installation from URI.
|
| |\ |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Added input and output parameters to LMI_SoftwareInstallationJob.
Added PreCallIndication and PostCallIndication properties to
LMI_SoftwareMethodResult containg CIM_InstMethodCall indication
instances - that are rather useless but mandatory by Job Control
Profile.
|
| |\ \ |
|
| | | | |
|
| | | | |
|
| | |/
| |/| |
|
|/ /
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Changes:
* Fixed problem when CPU ID from dmidecode is not unique
* Use cache info from sysfs if dmi cpu info is available, but not dmi cache
* When using sysfs cache, create all caches for every CPU
* Fixed physical memory tag when serial number field is missing in dmidecode
output
* Tiny clean-up
Hardware provider was tested in kvm guest with pegasus and this patch is result
of the test. Now, thanks to the all fallback options, hardware provider works
fine in this environment with and without selinux enabled.
|
| |
| |
| |
| |
| |
| |
| |
| | |
Added information:
* detect NUMA layout
* standard memory page size
* all supported sizes of huge memory pages
* current state of transparent huge pages
|
|/ |
|
|\ |
|
| | |
|
| |
| |
| |
| |
| |
| |
| | |
ssh://git.fedorahosted.org/git/openlmi-providers into software-indications"
This reverts commit 0f9da4ac91a0acb2c53d19ca1ec345c07a5043fb, reversing
changes made to 217acf2c624b99db7335acb0aa50d92efcb368a8.
|
| |
| |
| |
| |
| | |
cimom callbacks for filter enablement have misleading param names and
miss proper doc
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
And moved software static filters to shared JobManager.
Indication filters now reflect the name of indication class, which is
instrumented by particular provider. This ensures, that cimom knows,
which provider to call and does not bother the others. So instead of
selecting indication instances from general class CIM_InstModification,
it will be LMI_SoftwareInstModification. This is a common pattern for
indication queries.
Avoided another shortcoming of sfcbmof parser:
/* comment */ something useful is also ignored :-(
Static filters of JobManager made more generic for any provider
to use. Software jobmanager is now using them. This reduces redundation
of code.
|
| | |
|
| |
| |
| |
| |
| | |
We should return CIMInstance instead of CIMError (which has nothing to do with
CIM_Error).
|
| | |
|
| | |
|
|\| |
|
| | |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Mof files prefixed with digits ensuring their correct order of parsing
upon their installation. This avoids problems with sfcbmof compiler,
which can not handle dependent classes being parsed before their
antecedents.
Convention is specified in mof/README.
Removed inclusion of LMI_Qualifiers in LMI_Jobs. LMI_Qualifiers should
be registered in cimom instead of included in dependent mof files.
This allows it to be used by any other providers simultaneously.
Modified references to mof files in cmake build files.
|
| | |
|
|/
|
|
|
|
|
|
|
|
| |
New Providers:
* LMI_PhysicalMemoryProvider
* LMI_PhysicalMemoryRealizesProvider
Other Changes:
* Fixed InstanceID in hardware to match the rest of the providers
* Removed some unused variables
|
|\ |
|
| |
| |
| |
| |
| |
| | |
DEBUG is too verbose. On the other hand, we'll see no logging until
something sets more verbose level, e.g. after configuration load.
|
| |
| |
| |
| | |
clocktime_of_last_state_change was in local time.
|
| | |
|
| |
| |
| |
| | |
Job expiration timer should not be influenced by admin changing date or time.
|
| | |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Standard Python threading.Timer starts separate thread for each timer.
That has two unpleseant consequences:
- We end up with lot of sleeping threads.
- If we want to use logging in thread callbacks, we must register the thread
at CIMOM using AttachThread (=tedious).
So let's create TimerManager
- Uses only on thread.
- It is possible to log stuff in timer callbacks.
|
| | |
|
|\ \ |
|