multipath-tools 0.9.8

[mwilck]

Hi Christophe,

here's a PR for 0.9.8. As usual, all patches have been reviewed, except the ones marked with (*) below, which are some GitHub action changes and 2 minor documentation fixes.

(Side note: we should start thinking about what the next version after 0.9.9 should be).

News

• Socket activation via multipathd.socket has been disabled by default because it has undesirable side effects (fixes #76, at least partially).
• The restorequeueing CLI command now only enables queueing if disablequeueing had been sent before.
• multipathd now tracks the queueing mode of maps in its internal features string. This is helpful to ensure that maps have the desired queuing status. Without this, it can happen that a map remains in queueing state even after the no_path_retry timeout has expired.
• multipathd's map flushing code has been reworked to avoid hangs if there are no paths but outstanding IO. This allows "multipath -F" to retry a flush using the daemon rather than retrying locally.
• Error messages sent from multipathd to the command line client have been improved. The user will now see messages like "map or partition in use" or "device not found" instead of just "fail".
• Enabled -D_FILE_OFFSET_BITS=64 to fix issues with emulated 32-bit environments in the GitHub CI, so that we can now run our CI in arm/v7.

Bug fixes

• A segfault in the 0.9.7 autoresize code has been fixed.
• Fixed a bug introduced in 0.9.6 that had caused map reloads being ommitted when path priorities changed.
• Fixed compilation with gcc 14. (Fixes #80)
• Minor fixes for issues detected by coverity.
• Spelling fixes and other minor fixes.

CI

• Added the check-spelling GitHub action to avoid spelling errors in user-facing files in the future.
• Various improvements and updates for GitHub actions

Patch breakdown

@bmarzins (26): multipathd: Make sure to disable queueing if recovery has failed. multipathd: don't modify the multipath device on show commands libmultipath: keep track of queueing state in features multipathd: only restore queueing if it has been disabled first multipathd: remove nopath flushing code from flush_map() multipathd: make flush_map() delete maps like the multipath command multipathd: disable queueing when removing unknown maps multipathd: simplify cli_del_map() libmultipath: make _dm_flush_map() return an enum libmultipath: make dm_remove_partmaps() a static function multipathd: always start failure replies with "fail\n" multipathd: print extra default reply msg for busy devices multipathd: handle busy devices in cli_del_maps() libmultipath: print error when find_mp_by_str() fails. multipathd: don't open code find_mp_by_str() in client handers multipathd: make cli_handlers check for paths by devt and dev multipathd: add cli_handler reply message for missing devs libdmmp: handle failures in _process_cmd multipath: get rid of unnecessary retries variable multipath: Don't locally retry deletgated remove failures multipath: if delegation times out mark as not delegated multipathd: sync features on flush_map failure corner case multipathd: fix null pointer dereference in uev_update_path multipathd: fix auto-resize configuration libmultipath: fix displaying auto_resize config setting multipath-tools tests: add void parameter to functions

@brianatpurestorage (2): multipathd: the local path change is not considered libmultipath: stop PURE FlashArray from detecting priority

@jsoref (9): spelling: anymore spelling: cannot spelling: case-insensitive spelling: case-sensitive spelling: configuration spelling: correctly spelling: numerically spelling: preexisting spelling: than

@mwilck (29): libmultipath: avoid temporarily enabling queueing on reload libmultipath: fix ANA prioritizer enablement logic multipathd: init_unwinder: protect pthread_cond_wait() call with a loop libmultipath: is_uevent_busy(): check servicing_uev under lock libmultipath: tur: protect pthread_cond_timedwait with a loop multipathd: make update_prio static, and rename refresh_all param multipathd: don't activate socket activation by default multipath: udev rules: use configured $(bindir) in udev rules multipath-tools: Makefile.inc: set _FILE_OFFSET_BITS=64 GitHub workflows: update workflow events () GitHub workflows: update distros to run for () GitHub workflows: split multiarch workflow in rolling/stable () GitHub Workflows: run expensive workflows only on relevant changes () GitHub workflows: add spell checker () Spelling fixes found by check-spelling action () GitHub Workflows: abi: only run if C code has changed () GitHub workflows: abi: error if reference ABI can't be downloaded () GitHub workflows: run on PRs against "queue", too () multipath.conf.5: fix documentation for find_multipaths (fixes #75) () libmultipath: hwtable: fix fast_io_fail for Infinibox 11-dm-mpath.rules: fix list of imported properties 11-dm-mpath.rules: use import logic like 13-dm-disk.rules 11-dm-mpath.rules: don't import DM_UDEV_DISABLE_OTHER_RULES_FLAG 11-dm-mpath.rules: handle reloads during coldplug events 11-dm-mpath.rules: don't save DM_UDEV_DISABLE_OTHER_RULES_FLAG_OLD 11-dm-mpath.rules: clear DM_DISABLE_OTHER_RULES_FLAG for coldplug events libmultipath: bump version to 0.9.8 () multipath-tools: added NEWS.md ()

@xosevp (3): multipath-tools: update ml multipath-tools: fix an assignment ambiguity multipath-tools: remove extra hyphen from CFLAGS std option

[jsoref]

. multipathd now tracks the queueing mode of maps in ~it's~ its internal features string. This is ~helful~ helpful to ensure that maps have the desired queuing status. Without this, it can happen that a map remains in the queueing state even after the no_path_retry timeout has expired.

[mwilck]

. multipathd now tracks the queueing mode of maps in ~it's~ its internal features string. This is ~helful~ helpful to ensure that maps have the desired queuing status. Without this, it can happen that a map remains in the queueing state even after the no_path_retry timeout has expired.

Sic :grin: You don't need no spell-checker, do you? Thanks for your review!

[mwilck]

Pushed a minor change to the _FILE_OFFSET_BITS=64 patch: use $(OPTFLAGS) in the cpp command (although it's rather pointless) to avoid warning about missing optimization flags when building the unit tests.

[mwilck]

Pushed another patch after review from @bmarzins.

Once [PATCH v3] 11-dm-mpath.rules: handle reloads during coldplug events gets reviewed, I can add the udev rule series, too.

[mwilck]

The CI run for "multiarch test for rolling distros / build-current (debian-sid, arm/v7)" has apparently failed because of an issue with downloading the test container. @cvaroqui please restart the workflow if you have time.

[mwilck]

Pushed the the patch series for 11-dm-mpath.rules now, as the series has been completely reviewed.

[mwilck]

Pushed version bump to 0.9.8

[mwilck]

I added a the file NEWS.md that gives an overview about the changes in our code over the last 2 years.

[mwilck]

@cvaroqui, I believe this should be good for merging now.

[mwilck]

Again, the workflow failures seem to have been caused by problems pulling the test containers.

[mwilck]

@cvaroqui, please also push a 0.9.8 tag.