On 08/29/2013 11:18 AM, Viktor Mihajlovski wrote:
The current libvirt-cim implementation makes some assumptions that are only true for x86 architectures. As we want to enable libvirt-cim for s390 we need to makes sure that valid libvirt guest definitions are being built for that architecture while not breaking the existing implementation.
Patch 1 fixes potential memory access problems.
Patches 2 and 3 introduce two new properties arch and machine, effectively a pass-through of the underlying libvirt properties, for the necessary distinction between x86 and other guests, and suppress the default framebuffer for s390.
Patches 4 and 5 make sure that a minimal SVPC guest definition (VSSD and RASD) will result in a correct libvirt guest definition for the current hypervisor.
Boris Fiuczynski (2): libxkutil: Provide easy access to the libvirt capabilities VSSM: Set default values based on libvirt capabilities on DefineSystem calls
Viktor Mihajlovski (3): libxkutil: Improve domain.os_info cleanup VSSD: Add properties for arch and machine S390: Avoid the generation of default input and graphics
libxkutil/Makefile.am | 2 + libxkutil/capability_parsing.c | 556 +++++++++++++++++++++++++++++ libxkutil/capability_parsing.h | 97 +++++ libxkutil/device_parsing.c | 114 +++--- libxkutil/device_parsing.h | 4 +- libxkutil/xml_parse_test.c | 201 ++++++++++- libxkutil/xmlgen.c | 6 + schema/VSSD.mof | 6 + src/Virt_VSSD.c | 9 + src/Virt_VirtualSystemManagementService.c | 165 +++++---- 10 files changed, 1033 insertions(+), 127 deletions(-) create mode 100644 libxkutil/capability_parsing.c create mode 100644 libxkutil/capability_parsing.h
V2 Changes # Split original patch 1 into 2 patches (fix, new feature), otherwise unchanged # Patch 3 also considers QEMU domains now, was KVM only # Patches 4 and 5 address memory leaks and overwrites # cimtest run completes without regressions on x86
Things look good - cimtest and coverity tests both were happy. Since I had already ACK'd patch 1 which is now patch 1 & 2, consider the ACK still valid. Also the old patch 2 ACK is still valid even with the self find of the missing qemu check. I posted nits/questions separately about patches 4 & 5. Since I believe you still need me to push, feel free to post either just a set of diffs for me to squash into or a v3. Is there a desire to get this into a RHEL release eventually or into the "next" RHEL release? I'd rather let this "soak" a while and then work on getting a complete libvirt-cim into a future 6.n release over what I have to do now which is patch 0.6.1 with 0.6.3 based patches... John