FEC sriov-vf-driver lost after unexpected reboot

[Douglas Henrique Koerich]

Brief Description
-----------------
SRIOV VF driver of FEC disappeared on the host after the node rebooted itself.

Severity
--------
Major.

Steps to Reproduce
------------------
The behavior is similar to resetting manually the VF driver to "null" at the database.

Expected Behavior
------------------
SRIOV VF driver persisted in the system after FEC configuration.

Actual Behavior
----------------
SRIOV VF driver reset to "none" after unexpected reboot.

Reproducibility
---------------
Seen once.

System Configuration
--------------------
Seen in DC subcloud, but could happen in any configuration.

Branch/Pull Time/Commit
-----------------------
r/stx4.0.

Last Pass
---------
This is a new test scenario.

Timestamp/Logs
--------------
After reboot, sysinv got this inventory report:
sysinv 2022-02-16 19:59:30.432 103738 INFO sysinv.conductor.manager [-] attr: {'sriov_numvfs': u'1\n', 'driver': u'igb_uio', 'sriov_vf_driver': None, 'sriov_vf_pdevice_id': u'0d5d', 'psvendor': u'Intel Corporation', 'extra_info': u"{'expected_numvfs': 1}", 'pdevice_id': u'0d5c', 'pclass': u'Processing accelerators', 'psdevice': u'Device 0000', 'sriov_vfs_pci_address': u'0000:4c:00.0', 'pvendor': u'Intel Corporation', 'pvendor_id': u'8086', 'pclass_id': u'120001', 'sriov_totalvfs': u'16\n'}

Test Activity
-------------
Other - normal operation.

Workaround
----------
Needs reconfiguration of VF driver.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.opendev.org/c/starlingx/config/+/835304

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.opendev.org/c/starlingx/config/+/835304
Committed: https://opendev.org/starlingx/config/commit/a7b47f6197597a69c22c052f2c74b0a5fa1e541e
Submitter: "Zuul (22348)"
Branch: master

commit a7b47f6197597a69c22c052f2c74b0a5fa1e541e
Author: Douglas Henrique Koerich <email address hidden>
Date: Fri Mar 25 14:14:41 2022 -0300

    Prevent overwrite of sensible FEC configuration

    Regardless these precedent changes:
    https://review.opendev.org/c/starlingx/config/+/733724,
    https://review.opendev.org/c/starlingx/config/+/761176 and
    https://review.opendev.org/c/starlingx/config/+/801055
    that intended to avoid the overwrite of values in database for FEC
    device configuration with the actual (sometimes empty or temporary)
    values coming from PCI inventory report, it has been still noticed that
    under some unexpected reboot such sensible, stored data could be lost.

    This change implements for FEC driver and VF driver the same approach
    taken for number of configured VFs versus actual VFs introduced by the
    changes:
    https://review.opendev.org/c/starlingx/config/+/791531,
    https://review.opendev.org/c/starlingx/config/+/795850 and
    https://review.opendev.org/c/starlingx/config/+/808756

    From now on, the configured FEC driver and VF driver will be available
    as "extra information" (in the 'extra_info' field, together with
    configured number of VFs) that can be compared with the actual settings
    collected from PCI inventory for the host.

    Test Plan:

    PASS: Fresh install of r/stx7.0 with subsequent FEC configuration;
    PASS: Backup & restore of r/stx7.0 after FEC configuration;
    PASS: Patch of r/stx5.0 with previous FEC configuration;
    PASS: Upgrade from r/stx5.0 with previous FEC configuration to patched
          r/stx6.0.
    PASS: Upgrade from patched r/stx5.0 with previous FEC configuration to
          patched r/stx6.0.

    Closes-Bug: #1966471
    Signed-off-by: Douglas Henrique Koerich <email address hidden>
    Change-Id: I5e2e6f55856d5eef0a02117a10e36dc9c5aa9dda