Libvirt-CIM: The CIM provider for libvirt

Note: this is the flat content of the web site

libvirt-cim

What is libvirt-cim?

Libvirt-CIM is a CIM provider for managing linux virtualization platforms using libvirt. It is written in C and should work in any CIMOM that supports CMPI 2.0 providers. The intent is to implement the SVPC virtualization class model currently available in the DMTF Experimental 2.16 schema.

Releases

Here is the list of official releases. Developers and people looking for the latest features should check the downloads page for a snapshot of the repository.

livirt-cim-0.1: Jan 14 2008

libcmpiutil-0.1: Nov 30 2007

Introduction

Libvirt-CIM is a CIM provider for managing linux virtualization platforms using libvirt. It is written in C and should work in any CIMOM that supports CMPI 2.0 providers. The intent is to implement the SVPC virtualization class model currently available in the DMTF Experimental 2.16 schema.

The providers are currently under heavy development. Focus is on Xen support right now, which means some of the providers have some "shortcuts" hard-coded to Xen right now. The long-term goal is to support all of the platforms that libvirt supports with minimal differences.

Downloads

Official releases can be found on the libvirt.org FTP server. Alternatively, you can grab a tarball or zip file snapshot of the development repository. These snapshots will have the latest features (and bugs).

The libvirt-cim development tree can be found in the libvirt.org/hg repository.

To get a copy of the development tree, use mercurial's clone feature:

$ hg clone http://libvirt.org/hg/libvirt-cim

Patches

To submit patches to libvirt-cim, you must follow the DCO process, outlined below:

Developer's Certificate of Origin 1.1

By making a contribution to this project, I certify that:

  1. The contribution was created in whole or in part by me and I have the right to submit it under the open source license indicated in the file; or

  2. The contribution is based upon previous work that, to the best of my knowledge, is covered under an appropriate open source license and I have the right under that license to submit that work with modifications, whether created in whole or in part by me, under the same open source license (unless I am permitted to submit under a different license), as indicated in the file; or

  3. The contribution was provided directly to me by some other person who certified (1), (2) or (3) and I have not modified it.

  4. I understand and agree that this project and the contribution are public and that a record of the contribution (including all personal information I submit with it, including my sign-off) is maintained indefinitely and may be redistributed consistent with this project or the open source license(s) involved.

then you just add a line saying

Signed-off-by: Random J Developer <random@developer.example.org>

using your real name (sorry, no pseudonyms or anonymous contributions.)

Guidelines for Submitting Patches.

Patches should be submitted using Mercurial's patchbomb extension, and we recommend using the queues extension as well. On top of that, we have some guidelines you should follow when submitting patches. This makes reviewing patches easier, which in turns improves the chances of your patch being accepted in a timely fashion.

For help on how to use the patchbomb extension, see Section 14.4 of Distributed revision control with Mercurial.

For help on the queues extension, see Chapter 12.

Single Patches:

  1. When you add a patch to the queue you have an idea of where you're going with it, and the commit message should reflect that. Be specific. Avoid just saying something like, "Various fixes to AllocationCapabilities." Add a list of what was actually fixed, like, "Add EnumInstanceNames support," and, "Eliminate duplicate instances."

  2. The first line of your commit message will be the subject of the patch email when you send it out, so write it like a subject. Keep it short and to the point, then start a new line and feel free to be as verbose as you need to be. The entire commit message will be included in the patch.

  3. Stay on task with a patch. If you notice other problems while you are working on a patch, and they are not most definitely specific to your patch, they should be another patch. The same goes for nitpicking. While it might be only a line or two here and there that is off track, this is actually one of the easiest ways to make a patch difficult to review. All the trivial changes hide what is really going on. Make the unrelated changes a new patch or don't make them at all.

  4. If your patch addresses a strange issue or a rare edge case that the reviewers are unlikely to be familiar with, make sure the commit message include some example testcase with results, so the reviewers can verify your patch more quickly.

  5. Before you email, export. If you have a patch called "alloc_fixes", which would be emailed with "hg email alloc_fixes", you should first run "hg export alloc_fixes". This lets you review your patch. Does it have any typos in the comments? Did you accidentally include an irrelevant change? Is your commit message still accurate and useful? This is the single biggest step in ensuring you have a good patch.

  6. If your patch needs to be reworked and resent, prepend a "version number" to the first line of the commit message. For example, "Add EnumInstance to RASD," becomes "#2 Add EnumInstance to RASD." This helps mail readers thread discussions correctly and helps maintainers know they are applying the right version of your patch. At the end of the commit message, explain what is different from one version to the next. Nobody likes having to diff a diff.

  7. If your patch depends on a patch that exists on the mailing list but not in the tree, it is okay to send your patch to the list as long as your commit message mentions the dependency. It is also a good idea to import the patch into your tree before you make your patch. For example, a new patch called "cu_statusf API change" is on the list, and your patch needs the new API. Save the email (no need to trim headers) as api_change.eml, then do "hg qimport api_change.eml" and "hg qpush" so that the patch is applied to your tree. Now write your patch on top of it. You should still indicate the dependency in your commit message.

Patchsets:

  1. When you send a group of patches, Mercurial's email extension will create a "header" email. Make the subject and body of that email meaningful, so we know how the patches relate. It's easy to say, "Each patch has a commit message, it's obvious how they work together," but the rest of the list usually won't agree with that. If the commit messages for each patch are good, you shouldn't need more than a sentence or two to tie them all together, but you do need it.

  2. If any of your patches are rejected and you rework them, resend the entire set. This prevents things like, "So use patch 1 of 4 from the set I sent yesterday, 2 and 3 of 4 from the patch I sent an hour later, and patch 4 of 4 from today."

  3. If you resend a patchset, apply part (6) of the Single Patches guidelines to your "Patch [0 of 3]" header email, for all the same reasons.

Questions/Comments on the Guidelines should be directed to:

Jay Gagnon grendel@linux.vnet.ibm.com
Patch Compliance Officer

Schema

The libvirt-cim provider depends on an installed DMTF CIM v2.16 Experimental schema. The package can be obtained here.

To install the schema in Pegasus:

$ PEGASUS_REPO=/var/lib/Pegasus # adjust this as needed
$ mkdir cim216
$ cd cim216
$ unzip $PATH_TO_ZIPFILE
$ sudo cimmofl -uc -aEV -R$PEGASUS_REPO -n /root/virt cimv216.mof
$ sudo cimmofl -uc -aEV -R$PEGASUS_REPO -n /root/virt qualifiers.mof
$ sudo cimmofl -uc -aEV -R$PEGASUS_REPO -n /root/virt qualifiers_optional.mof
$ sudo cimmofl -uc -aEV -R$PEGASUS_REPO -n /root/interop cimv216-interop.mof

To install the schema in SFCB:

$ SFCB_CIM=/usr/local/share/sfcb/CIM # adjust this as needed
$ mkdir cim216
$ cd cim216
$ unzip $PATH_TO_ZIPFILE
$ mv cimv216.mof CIM_Schema.mof
$ sudo cp * $SFCB_CIM
$ sudo sfcbrepos

Note: in both cases, the CIM v2.16 schema seems to have a few classes that don't register correctly. You may need to disable installation of classes with something like the following:

--- CIM_Schema.mof 2007-10-15 00:15:44.000000000 -0700
+++ cimv216.mof 2007-10-22 10:11:19.000000000 -0700
@@ -507,3 +507,3 @@
#pragma include ("Policy/CIM_SharedSecretAuthentication.mof")
-#pragma include ("Security/CIM_SecurityIndication.mof")
+//#pragma include ("Security/CIM_SecurityIndication.mof")
#pragma include ("Support/PRS_Activity.mof")
@@ -728,4 +728,4 @@
#pragma include ("Policy/CIM_PolicyConditionInPolicyRule.mof")
-#pragma include ("Security/CIM_IPNetworkSecurityIndication.mof")
-#pragma include ("Security/CIM_IPPacketFilterIndication.mof")
+//#pragma include ("Security/CIM_IPNetworkSecurityIndication.mof")
+//#pragma include ("Security/CIM_IPPacketFilterIndication.mof")
#pragma include ("Support/PRS_ActivityContact.mof")

cimv216-interop.mof is not part of the official DMTF CIM v2.16 schema package. Please create with the following content:

#pragma locale ("en_US")
#pragma include ("qualifiers.mof")
#pragma include ("qualifiers_optional.mof")
#pragma include ("Core/CIM_ManagedElement.mof")
#pragma include ("Interop/CIM_RegisteredProfile.mof")
#pragma include ("Interop/CIM_RegisteredSubProfile.mof")
#pragma include ("Core/CIM_Dependency.mof")
#pragma include ("Interop/CIM_ElementConformsToProfile.mof")
#pragma include ("Interop/CIM_ReferencedProfile.mof")
#pragma include ("Interop/CIM_SubProfileRequiresProfile.mof")

Platform Support

Currently, libvirt-cim is targetting Xen as its primary support platform because is has the largest installed user base. The long-term plan is to support many others (hopefully any that libvirt supports). This includes KVM and containers.

The code base currently has many Xen-specific "shortcuts" that need to be resolved and generalized in order to support other platforms. A short list of these may include:

Further, supporting new platforms have some registration and modeling implications:

Architecture

The libvirt-cim provider consists of two major parts:

The provider classes implement the actual CIM class model. Some of the provider libraries implement one CIM class and one providier. However, many of them perform more than one task. For example, the Virt_Device and Virt_DevicePool providers implement the device and device pool classes for each of the major device types: Memory, Processor, Network, and Disk.

The helper library contains common routines that almost all of the providers use, such as libvirt connection type detection and device and system XML parsing.