qemu: clear qmp state before wait for qemu process

[bergwolf]

So that if there is any remaining state, we do not let it interfere
with the new one. This should fix the occasional vm factory hang.

[katacontainersbot]

PSS Measurement:
Qemu: 171629 KB
Proxy: 5920 KB
Shim: 9089 KB

Memory inside container:
Total Memory: 2043480 KB
Free Memory: 2003620 KB

[codecov]

Codecov Report

Merging #536 into master will decrease coverage by <.01%.
The diff coverage is 0%.

@@            Coverage Diff             @@
##           master     #536      +/-   ##
==========================================
- Coverage   66.67%   66.66%   -0.01%     
==========================================
  Files          93       93              
  Lines        9580     9614      +34     
==========================================
+ Hits         6387     6409      +22     
- Misses       2506     2517      +11     
- Partials      687      688       +1

[opendev-zuul]

Build succeeded (third-party-check pipeline).

• kata-runsh : SUCCESS in 52m 28s

[jodh-intel]

lgtm

/cc @markdryan.

[markdryan]

As long this change abides by the constraints in the API documentation it should be okay:

https://github.com/intel/govmm/blob/master/qemu/qmp.go#L591

If it doesn't you might get a panic.

[bergwolf]

@markdryan At the point of hypervisor.waitsandbox(), I don't think anyone should connect/disconnect to the qemu process concurrently -- we wait for the qemu process to be up before actually talk to it in any means.

[markdryan]

@bergwolf Can we be sure no one else has called Shutdown on this instance? If so it should be okay. You could also read from the disconnectedChannel which will block until the connection is truly closed. This is what all the test cases do, e.g., https://github.com/intel/govmm/blob/master/qemu/qmp_test.go#L497

[bergwolf]

@markdryan q.qmpShutdown() checks for nil qmp handler before calling shutdown. So if shutdown is called before, we are still safe. The only case we need to exclude is concurrent connect/disconnect.

[markdryan]

@bergwolf, I'd missed that. I thought you were calling Shutdown directly. Reading from the disconnected channel might still be a good idea though.

[bergwolf]

@markdryan Yes, you are right. It still makes sense to wait on the disconnected channel. I've added a new commit to do it. PTAL.

[jodh-intel]

lgtm

[katacontainersbot]

PSS Measurement:
Qemu: 167749 KB
Proxy: 5667 KB
Shim: 9076 KB

Memory inside container:
Total Memory: 2043480 KB
Free Memory: 2003092 KB

[opendev-zuul]

Build failed (third-party-check pipeline) integration testing with
OpenStack. For information on how to proceed, see
http://docs.openstack.org/infra/manual/developers.html#automated-testing

• kata-runsh : TIMED_OUT in 1h 01m 14s

[jodh-intel]

Ignoring 0.01% drop in code cov...