On Thu, Apr 16, 2009 at 04:18:58PM +0200, Pritesh Kothari wrote:
Hi All,
I have attached a patch which when applied on the HEAD as of today would allow virtualbox support in libvirt. It takes cares of all the stuff mentioned on the list earlier. Still if I have missed anything, please do tell me.
I actually just tried out your previous patch from 2 days ago and it worked without trouble, so I reckon we can plan to get this driver in the forthcoming release next week.
+static virCapsPtr vboxCapsInit(void) { + struct utsname utsname; + virCapsPtr caps; + virCapsGuestPtr guest; + + uname(&utsname); + + if ((caps = virCapabilitiesNew(utsname.machine, + 0, 0)) == NULL) + goto no_memory; + + if (virCapsInitNUMA(caps) < 0) + goto no_memory; + + virCapabilitiesSetMacPrefix(caps, (unsigned char[]){ 0x08, 0x00, 0x27 }); + + if ((guest = virCapabilitiesAddGuest(caps, + "hvm", + utsname.machine, + sizeof(int) == 4 ? 32 : 64,
I was wondering why the capabilities said '32' as wordsize even on x86_64, and of course this is because 'int' is still 32 bits on x86_64. I'd switch to 'sizeof(size_t)' instead unless someone has better suggestions for determining the native arch wordsize in a portable manner.
+static const char *vboxGetType(virConnectPtr conn ATTRIBUTE_UNUSED) { + DEBUG("%s: in vboxGetType",conn->driver->name); + return strdup("VBox"); +}
This shouldn't strdup the type - the returned string is const, not to be free'd by caller. Even better just remove this method entirely. I don't know why we have this as a driver method at all. The default impl in src/libvirt.c already does the correct thing, returning the conn->driver->name string. We should remove getType from all our driver impls. Daniel -- |: Red Hat, Engineering, London -o- http://people.redhat.com/berrange/ :| |: http://libvirt.org -o- http://virt-manager.org -o- http://ovirt.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: GnuPG: 7D3B9505 -o- F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|