Q4/2021 multipath-tools update

[mwilck]

Hello Christophe,

here is the next PR. Breakdown of important changes:

@xosevp: multiple hardware table and man page fixes

@lucab: multipathd.socket: add missing conditions from service unit (closes #16, closes #15)

@lixiaokeng: persistent reservation fixes

• remove memory debugging code
• memory handling fixes
• multipathd: fix missing persistent reseravtion for active path

@bmarzins:

• add validity checks to config file parser
• improve warning messages from config file parser
• add "reconfigure all" command
• support for partially initialized path devices
• remove dependency on systemd-udev-settle.service (closes #3)

@mwilck:

• improve delayed reconfigure (and multiple subsequent reconfigure requests)
• uxlsnr cleanup: don't sleep or wait while handling clients
• fix exit status of multipath -T
• add wildcard "%I" for init state
• fixes for coverity-reported defects (multipath-tools has zero defects now)
• improved error-checking in VPD parsing code
• cleanup of wildcard-handling code, make wildcard handling table static const
• fix an asan error on map reval
• various other code cleanups
• added github workflows for determining ABI incompatibility and coverity checks (see below).

Notes on the new github workflows

The coverity workflow runs only on the coverity branch. It requires creating repository secrets to be run:

• COVERITY_SCAN_EMAIL: the email address for coverity/synopsis account
• COVERITY_SCAN_TOKEN: the coverity / synopsis access token
• COVERITY_SCAN_PROJECT: the coverity project, e.g. mwilck/multipath-tools

Without these properly set, the workflow will fail.

The ABI checking workflow has two modes:

1. On the branch named by the secret ABI_BRANCH (default: "master"), it creates the reference ABI. This succeeds always. The ABI representation in XML format will be saved in a github workflow artifact.
2. Other branches will be compared to the reference ABI, and the workflow will fail if differences are detected. Both the ABI and the output of the abidiff command will be saved as artifacts. Failure shouldn't be considered fatal in particular on PRs, but if a PR is merged for which the ABI check fails, the library versions must be bumped in the patch set as appropriate.

~@bmarzins, you'll note that I pushed 3 workflow-related fixes without sending them to the mailing list for review. I suppose this is alright. Bugs in workflows tend to surface very quickly.~

[mwilck]

I've added some patches from @xosevp which I'd miss first-time around. Also, the not-yet-reviewed patch for multipathd.service which I just submitted to dm-devel.

[mwilck]

I apologize.

I had to rebase because 2 patches early in the series had been missing "Reviewed-by:" tags. Also, I decided to move my ABI and workflow related patches up to the top and submit them to dm-devel. I did this because I improved the ABI handling once more (less false positives) and added some tidy-up for libmultipath.version according to the late discussion between @bmarzins and myself.

[mwilck]

I pushed once more to add the missing Reviewed-by: tags from Ben for the "last minute" patches.