Merge pull request xapi-project#3940 from psafont/py-import

CP-12980: maintain import behaviour on package

CA-325319 Fix host-display script console handling

There were two distinct issues with the host-display script. Firstly, it always
assumed com1 was the serial port in use, when in some cases it may be e.g. com2.
Secondly if you go through a disable/enable cycle on a normal (VGA) setup, the
result would be the loss of `console=hvc0` from the dom0 command line, because
the `set-dom0` command will *replace* all existing entries, not just append one
to the line.

These changes resolve these issues.

Signed-off-by: Alex Brett <[email protected]>
(cherry picked from commit 16266f5)

CA-311625: alarm only if the last attempt of PBD.plug fails

Currently, depending on a host's setup, there are up to 3 attempts in plugging
all the PBDs in the startup process: create_storage_localhost,
wait_management_interface, and maybe_wait_for_clustering_ip. We should only
send out an alert if the last attempt still doesn't succeed (in plugging any
PBD).

The patch also consolidates/refactors some functionalities in the
Create_storage module.

Signed-off-by: Zheng Li <[email protected]>

Merge pull request xapi-project#3928 from robhoes/stub

CP-31117: Remove obsolete VM options

Merge pull request xapi-project#3910 from BenSimsCitrix/private/bensi…

…/CP-31450

CP-31450: Add domid to Datapath.attach

Merge pull request xapi-project#3918 from edwintorok/private/edvint/x…

…si-388

CA-322045: tell XAPI to shut down only once

Merge pull request xapi-project#3913 from edwintorok/private/edvint/t…

…oolstack-restart

CA-322045: make light_fuse idempotent

Merge pull request xapi-project#3915 from lippirk/CP-28368

Cp 28368

Merge pull request xapi-project#3908 from MarkSymsCtx/CA-322146

CA-322146: set NBD device scheduler and max_sectors to more efficient…

CA-316165: Convert some Thread.delay to Delay.wait

Following suggestions on the ticket, here we convert some unconditional wait
(Thread.delay) to conditional wait (Delay.wait).

The pattern is, in these tests we expect to observe something to happen, but it
has to be within a reasonable time frame. From testing point of view, we can
not assume a function will always return (especially if it was implemented
wrong) and we certainly don't want to wait indefinitely.

Previously with Thread.delay, we unconditionally wait for a timeout, then check
the observed value. Not pretty, but it doesn't require any kind of
collaboration from the observed party. The problem is, if we set the timeout
tight, it might time out if the system is extremely stressed. But if we set the
timeout too relax, we end up wasting time every single run.

Now With Delay.wait, we wait conditionally on both a timeout and a
signal/condition. In these case, we can now set the timeout generously (say,
times higher than the usual value, as a real hard timeout), but knowing that in
most cases the wait will end much earlier as usual (upon some
signal/condition) instead of breaking out of the timeout.

Signed-off-by: Zheng Li <[email protected]>