On 04.09.2015 11:40, Daniel P. Berrange wrote:
On Fri, Sep 04, 2015 at 10:56:52AM +0300, Nikolay Shirokovskiy wrote:
@@ -1396,6 +1585,9 @@ static virHypervisorDriver vzDriver = { .domainBlockStatsFlags = vzDomainBlockStatsFlags, /* 1.2.17 */ .domainInterfaceStats = vzDomainInterfaceStats, /* 1.2.17 */ .domainMemoryStats = vzDomainMemoryStats, /* 1.2.17 */ + .connectSupportsFeature = vzConnectSupportsFeature, /* 1.2.20 */ + .domainMigratePrepare3Params = vzDomainMigratePrepare3Params, /* 1.2.20 */ + .domainMigratePerform3Params = vzDomainMigratePerform3Params, /* 1.2.20 */
Somewhat annoyingly you also need to implement the callbacks for .domainMigratePrepare3 and .domainMigratePerform3, as we don't automatically convert non-params usage to the params based method AFAICT.
Your impl of .domainMigratePerform3 could pack the values into a virTypedParams array and then call .domainMigratePerform3Params, or do the reverse. Yes, without plain(non-params) callbacks we get working only toURI3 API function and I create a patch not included in this patchset to make toURI{1,2} work too. I take this approach of converting parameters and use one common worker function but patch a different place - API implementaion itself. So I'll include this patch in next version of the set.
As in this case I need to patch 2 different sets of API implementation *migrate{N} and *migrateURI{N} I'd rather put direct managed support to a different patchset. Is it ok?
Yeah, that'd be fine as long as we still compile at each step it isn't a problem if the impl is not final. Ok. Then I'll send patch supporting toURI{1, 2} in a different patchset as it will be a new thing that could lead to meaningless ping-ponging of parts already areed upon.
Regards, Daniel