fix live migration

[joyent-automation]

fix live migration

This PR was migrated-from-gerrit, https://cr.joyent.us/#/c/5749/.
The raw archive of this CR is here.
See MANTA-4594 for info on Joyent Eng's migration from Gerrit.

CR discussion

@YanChii commented at 2019-03-18T19:32:27

Uploaded patch set 2: Patch Set 1 was rebased.

@YanChii commented at 2019-04-01T15:04:17

Topic set to live migration

@rmustacc commented at 2019-04-05T14:27:26

Patch Set 2:

(1 comment)

@YanChii commented at 2019-04-06T16:17:00

Uploaded patch set 3: Patch Set 2 was rebased.

@rmustacc commented at 2019-04-08T21:14:25

Patch Set 4:

(1 comment)

@YanChii commented at 2019-04-08T23:13:05

Patch Set 4:

(1 comment)

Patch Set 4 code comments

qemu-kvm-x86.c#517 @rmustacc

Please see my previous comment about how this should be implemented.

qemu-kvm-x86.c#517 @YanChii

Sorry, I missed this. Will fix.
But actually this check is not needed. Correct me if I'm wrong but combination KVM+qemu+SunOS+live_migration will always result in 64-bit OS. This code is not used for any other purpose.

qemu-kvm-x86.c#517 @rmustacc

This, as written, is saying that if qemu is 64-bit, then it's fine. While we don't plan on having a 32-bit qemu, a 32-bit qemu could run on a 64-bit OS. If you want something that's actually the equivalent of the existing Linux check, then we should actually do what I suggested. If we're going to assume that we're only ever 64-bit, then we should set it to 1, and explicitly #error if we're not building 64-bit.

Also, the cstyle of the else statement needs to be fixed here.

qemu-kvm-x86.c#517 @YanChii

Even if never used, I feel the most proper way is to implement SI_ARCHITECTURE_K. Other ways seem too lazy to me.
Pushing the fix.

@rmustacc commented at 2019-04-16T15:09:27

Patch Set 4:

(1 comment)

@YanChii commented at 2019-04-17T11:43:50

Patch Set 5:

(1 comment)