public inbox for development@lists.ipfire.org
 help / color / mirror / Atom feed
From: Adolf Belka <adolf.belka@ipfire.org>
To: development@lists.ipfire.org
Cc: Adolf Belka <adolf.belka@ipfire.org>
Subject: [PATCH] multipath-tools: Update to version 0.15.1
Date: Fri, 20 Nov 2026 21:11:36 +0100	[thread overview]
Message-ID: <20261120201136.642750-2-adolf.belka@ipfire.org> (raw)
In-Reply-To: <20261120201136.642750-1-adolf.belka@ipfire.org>

- Update from version 0.8.9 (Feb 2022) to 0.15.1
- No change in rootfile
- No CVE ID's given but 7 fixes for vulnerabilities in 0.15.1
- Building kpartx in the kpartx directory no longer works as it is mssing some files to
   be built. Based on some input from chat gpt luna-6 the make was moved to the root of
   the tarball but specifying to only build kpartx. This seems to have worked as the
   existing rootfile was accepted with no additional packages etc included in it.
- Changelog
0.15.1
User-visible changes
    The preferredip=parameter for the iet prioritizer has been generalized.
    The new syntax is
	    preferredip=<CIDR>:<Prio>[,<CIDR>:<Prio>,...]
     where CIDR is an IPv4 address block in CIDR format, e.g. 192.168.1.0/24,
    and <Prio> is the numeric priority to assign to IP addresses in this
    block. See the multipath.conf(5) man page for details. The syntax
    preferredip=<IP> is still supported with the same semantics as before.
Vulnerability fixes
    DoS on mulipathd socket by blocking IPC send operations
    DoS on multipathd socket by exhausting connections
    Heap Out-of-Bounds Read in Custom Format String Parser via Trailing %
    Heap out-of-bounds read in device-mapper-multipath ALUA RTPG parsing
    kpartx: Heap Out-of-Bounds Read in GPT Header Validation
    Path traversal in device-mapper-multipath failed_wwids management
    libmpathpersist PRIN READ FULL STATUS parser — unbounded descriptor rewrite
	causes root heap overflow
Bug fixes
    Fix a use-after-free error introduced by the fix for
     #152 in 0.15.0.
     Fixes openSUSE#24.
    When parsing the device-mapper table of a multipath device, the result for
     path arguments was not checked for NULL. This happens if the kernel
     returns an invalid table with missing path arguments. Fix it by adding
     a NULL check (commit ac9fe05).
     Fixes #155,
    #156.
0.15.0
User-visible changes
    All path checkers run in asynchronous mode now by default, using a new
     generic asynchronous checker framework. This improves the stability of
     multipathd in the presence of non-responsive devices when using path
     checkers other than tur and directio. Use the force_sync
     parameter in multipath.conf to switch back to the previous, synchronous
     behavior, especially if you observe strong spikes of CPU load on systems
     with a lot of path devices. Note that the likelihood of such spikes should
     be strongly reduced since multipath-tools 0.10.0. Commit 6f7daba ff.
    The configuration options rr_min_io, rr_min_io_rq, and rr_weight are now
     deprecated and have no effect. These options have not been supported by the
     kernel since version 4.6. Users should remove them from multipath.conf.
     Commits 91f91c7 ff.
    Fix ALUA asymmetric access state descriptions in multipathd logs, so that
     the same terms are used as by the kernel ("lba-dependent", "transitioning").
     Commit fa2eb12.
    Don't set a hardware handler for bio-based multipath devices. The kernel
     rejects this anyway. Commit b6c7aab.
Bug fixes
    Fix WWID detection for legacy devices that use the older SCSI-2 VPD page
     0x83 format for their device identifier. Commit ed6f6ae.
    kpartx: Fix an integer overflow in the GPT partition table size calculation.
    A crafted partition table with an extremely large number of partition entries
     could trigger the overflow. Commit 8697ea7, 15310a3.
    kpartx: Fix several issues in the DASD partition table reader that could be
     triggered by a maliciously crafted disk image. Commits c616a95 ff.
    Fix duplicate "checker timed out" log messages when log_checker_err is
     set to once. Fixes 0.5.0. Commit 8933b22.
    Avoid potential buffer overflows in the iet and datacore prioritizers.
     Commit 4611f97.
    iet prioritizer: avoid misleading error message with systemd 256 and
     newer, and properly use udev to derive path parameters. Commit c493f1f.
     Fixes #145.
    An overlong partition delimiter (-p option) could cause kpartx to crash.
     Fix it. Commit a2f344a.
    Even with the libudev wrapper code introduced in 0.14.0, multipathd ran into
     use-after-free errors in tests where multipathd was restarted frequently.
     Fix this by preventing thread cancellation during libudev calls.
     Fixes #152.
     Commit 47a654f.
    libmpathpersist: Fix self-preemption without holding a reservation.
     Fixes 0.13.0. Commit 4260774.
Other changes
    Man page improvements
CI
    Removed the check_spelling GitHub Action after a
     security incident
    reported against the action's source repository.
    Added coredump collection in the CI.
    GitHub action updates related to node 20 depreciation on GitHub.
    Fixes for various errors in the CI.
0.14.3
	Fix boot failures with multipath introduced in 0.14.1. Commit 01cc89c.
0.14.2
Bug fixes
    Fix a possible NULL pointer dereference in the path purging code.
     Fixes 0.14.0. Commit 86ff48b.
CI
    Updated the test_kpartx test script, and added it to the
     basic-build-and-ci workflow.
0.14.1
Bug fixes
    kpartx: Fix freeing static buffer when operating on regular files.
     Fixes 0.14.0. Commit fab5d44.
     Fixes #139.
    Fix initialization of paths that were offline during path detection.
     Commit 1942fb1.
    Fix printing the "path offline" log message for offline paths that don't
     have a path checker configured. Commit 1a364a1.
Other changes
    If path devices that are members of multipath maps aren't detected during
     multipathd startup or reconfigure, don't add them to newly created maps
     in the first place. In previous versions, such paths would be added to the
     maps, only to be removed later. Commit a04be55.
    Improve the detection of "busy" state in the show status command, such
     that reading udev events from udevd is also counted as busy.
     Commit 7fdd93b.
0.14.0
User-visible changes
    Automatic removal of disconnected SCSI devices
	This release adds support for automatically purging SCSI devices that become
	 disconnected at the storage target (e.g. because a LUN became unmapped by an
	 administrator). It's a long-standing problem that such path devices become
	 "stale" on the Linux side. When they are re-mapped on the storage side,
	 possibly with different parameters, the kernel may not correctly detect
	 the changes. By removing such stale devices, the kernel will be forced to
	 re-probe them, which fixes this problem.
	The feature is controlled by a new configuration option purge_disconnected
	 that can be set in multipath.conf per multipath device, hardware entry, or
	 globally. The feature is switched off by default.
	When purge_disconnected is enabled, SCSI devices for which the storage array
	 reports a "LUN NOT SUPPORTED" status are removed from the system, preventing
	 stale device entries from accumulating.
	As this is a new feature, users are advised to test it in their environment
	 before enabling it in production.
Bug fixes
    Make sure multipathd registers keys all paths of a multipath map after
     mpathpersist registered a key for a map. multipathd could fail to do so in
     some cases, e.g. if paths become unavailable or available while the
     registration was taking place. Fixes 0.13.0. Commits f7d6cd1, 4f3036b.
    Fix mpathpersist --report-capabilities output. Fixes 0.5.0. Commit
     c8ed5e6.
    Improve error handling when retrying REGISTER AND IGNORE persistent
     reservations commands. Fixes 0.13.0. Commit c971036.
    Fix command descriptions in the multipathd man page. Fixes 0.9.2.
     Commit f3ba2e7.
    Fix an undefined symbol error with the LLVM lld linker. Commit a298603.
     Fixes #132, 0.10.0.
    Fix ISO C23 compatibility issue causing errors with new compilers.
     Commit 9f611e2.
    Fix use-after-free error in free_pgvec(). Commit 7a07023.
     Fixes #128, 0.12.0.
    Fix memory leak caused by not joining the "init unwinder" thread.
     Fixes 0.8.6. Commit 29f262b.
    Fix memory leaks in kpartx. Commits 8c39e60, 02e0933. Fixes any version.
    Fix a particular strict aliasing bug in the bitfield code.
    Fixes 0.8.9. Commit f1299f2. This fix isn't strictly necessary any
     more after adding -fno-strict-aliasing (see below), but it doesn't hurt to
     have it.
    Print the warning "setting scsi timeouts is unsupported for protocol" only
     once per protocol. Commit 958c826. Fixes 0.9.0.
    Make sure multipath-tools is compiled with the compiler flag
     -fno-strict-aliasing. This turns out to be necessary because our code
     uses techniques like container_of() which don't work well with
     strict aliasing rules. Commit 1ef540a.
     Fixes #130.
Other changes
    Add wrapper code for libudev to avoid potential issues with calling libudev
     from a multi-threaded program. Commits 69f6590, 6cc012f.
    Clean up the code for freeing struct path and struct multipath objects.
     Specifically, avoid freeing path objects from code that frees multipath
     maps, and vice versa. Treat path orphaning and freeing separately.
    Hardware table: add Seagate Exos and Nytro series.
    Avoid joining threads twice with liburcu 0.14.0 and newer.
     CI updates (GitHub workflows).
     Fix CI for cmocka 2.0. Commits 2c52668, f427f8f.
     Fixes #129
    Remove the "hotplug" mode of kpartx, which has been obsolete since udev
     replaced legacy hotplug.
    Add the ASAN=1 and OPT= make variables (see README.md).
0.13.0
Major rework of the SCSI Persistent Reservation code
	The mpathpersist code had many areas where it did a poor job of
	 mimicking the behavior of setting up persistent reservations on an
	 individual device, especially if some paths to the multipath device
	 were missing or unusable.
Issues improved in this release:
    Persistent reservation operations in cases where the reservation
     is held held by an unavailable path. For example, when such a key
     is unregistered, libmpathpersist now preempts the old key
     after verifying that the multipath device is holding the reservation.
    Releasing a reservation held by an unavailable path. This has always
     had code to handle it, but it relied on optional Persistent
    Reservation features. The new code preempts the key and then restores
     the removed registered keys while the device is suspended.
    Handling removal of keys from paths that were down when the multipath
     device was unregistered.
    Changing how conflicts on key registration are handled. Instead of
     the old rollback method (which was broken anyway),
     libmpathpersist now retries with REGISTER AND IGNORE as long as the
     REGISTER command completed successfully down some of the paths.
    Changing when the reservation key is set to fix corner cases on
     failure and registration while paths are coming up.
    Retrying to fix corner cases when a path is coming up while doing a
     reserve, preempt or clear.
    Allowing registrations to succeed when there are retryable errors,
     if the paths are actually down.
    Fixing the reservation key validation in some commands.
    Handling a race condition that may occur while changing the reservation key
     of a multipath map.
    Handling preemption of registrations of type "all registrants".
In the course of making these changes, a number of corner cases were fixed,
 too. Please consult the commit messages for details.
Other changes
    Updates to the built-in hardware table:
        add some NVMe storage array (VASTData, Infinidat, HITACHI VSP)
        add QSAN
        add EqualLogic PS
0.11.0
User-visible changes
    Modified the systemd unit multipathd.service such that multipathd will now
     restart after a failure or crash.
     Fixes #100.
    Logging changes for verbosity level 3:
        silenced logging of path status if the status is unchanged
        silenced some unhelpful messages from scanning of existing maps
        added a message when partition mappings are removed.
Other major changes
Rework of the path checking algorithm
	This is a continuation of the checker-related work that went into 0.10.0. For
	 asynchronous checker algorithms (i.e. tur and directio), the start of the
	 check and the retrieval of the results are done at separate points in time,
	 which reduces the time for waiting for the checker results of individual paths
	 and thus strongly improves the performance of the checker algorithm, in
	 particular on systems with a large a amount of paths.
The algorithm has the following additional properties:
    multipath devices get synchronized with the kernel occasionally, even if
     they have not paths
    If multiple paths from a multipath device are checked in the same
     loop, the multipath device will only get synchronized with the kernel once.
    All the paths of a multipath device will converge to being checked at
     the same time (at least as much as their differing checker intervals will
     allow).
    The times for checking the paths of each multipath device will spread
     out as much as possible so multipathd doesn't do all of it's checking in
     a burst.
    path checking is done by multipath device (for initialized paths,
     the uninitialized paths are handled after all the adopted paths are
     handled).
Bug fixes
    Fixed the problem that multipathd wouldn't start on systems with certain types
     of device mapper devices, in particular devices with multiple DM targets.
     The problem was introduced in 0.10.0.
     Fixes #102.
    Fixed a corner case in the udev rules which could cause a device not to be
     activated during boot if a cold plug uevent is processed for a previously
     not configured multipath map while this map was suspended. This problem existed
     since 0.9.8.
    Fixed the problem that devices with no_path_retry fail and no setting
     for dev_loss_tmo might get the dev_loss_tmo set to 0, causing the
     device to be deleted immediately in the event of a transport disruption.
     This bug was introduced in 0.9.6.
    Fixed the problem that, if there were multiple maps with deferred failback
     (failback value > 0 in multipath.conf), some maps might fail back later
     than configured. The problem existed since 0.9.6.
    Removed a warning message that multipathd would print if systemd's WATCHDOG_USEC
	environment variable had the value "0", which means that the watchdog is
	simply disabled. This (minor) problem existed since 0.4.9.
    Fixed handling of empty maps, i.e. multipath maps that have a multipath UUID
     but don't contain a device-mapper table. Such maps can occur in very rare
     cases if some device-mapper operation has failed, or if a tool has been
     killed in the process of map creation. multipathd will now detect such
     cases, and either remove these maps or reload them as appropriate.
    During map creation, fixed the case where a map with different name, but
     matching UUID and matching type was already present. multipathd
     previously failed to set up such maps. Now it will reload them with the
     correct content.
    Fixed the logic for enabling the systemd watchdog (WatchdogSec= in the
     systemd unit file for multipathd).
    Fixed a memory leak in the nvme foreign library. The bug existed since
     0.7.8.
    Fixed a problem in the marginal path detection algorithm that could cause
     the io error check for a recently failed path to be delayed. This bug
     existed since 0.7.4.
Other
    Default settings for hardware_handler have been removed from the
     internal hardware table. These settings have been obsoleted by the Linux
     kernel 4.3 and later, which automatically selects hardware handlers when
     SCSI devices are added. See the notes about SCSI_DH_MODULES_PRELOAD in
     README.md.
    Added a hardware table entry for the SCSI Target Subsystem for Linux (SCST).
    The text of the licenses has been updated to the latest versions from the
     Free Software Foundation.
Internal
    libmp_mapinfo() now fills in the name, uuid, and dmi fields
     if requested by the caller, even if it encounters an error or an empty map.
0.10.0 and 0.9.9 down to 0.9.0 are all just listed as Pull Request from Open SUSE

Signed-off-by: Adolf Belka <adolf.belka@ipfire.org>
---
 lfs/multipath-tools | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/lfs/multipath-tools b/lfs/multipath-tools
index 0ff8d813f..dde5f0eae 100644
--- a/lfs/multipath-tools
+++ b/lfs/multipath-tools
@@ -1,7 +1,7 @@
 ###############################################################################
 #                                                                             #
 # IPFire.org - A linux based firewall                                         #
-# Copyright (C) 2007-2018  IPFire Team  <info@ipfire.org>                     #
+# Copyright (C) 2007-2026  IPFire Team  <info@ipfire.org>                     #
 #                                                                             #
 # This program is free software: you can redistribute it and/or modify        #
 # it under the terms of the GNU General Public License as published by        #
@@ -24,7 +24,7 @@
 
 include Config
 
-VER        = 0.8.9
+VER        = 0.15.1
 
 THISAPP    = multipath-tools-$(VER)
 DL_FILE    = $(THISAPP).tar.gz
@@ -40,7 +40,7 @@ objects = $(DL_FILE)
 
 $(DL_FILE) = $(DL_FROM)/$(DL_FILE)
 
-$(DL_FILE)_BLAKE2 = b5ebf3c393f6b60e85678ac07378ae07056b6777409fc1bc4f4133cdd3f8c75a3d76f6e9342208df7fed8fe7812b089eba8f6b769e47e1dd6c8b7fd321bdbd30
+$(DL_FILE)_BLAKE2 = 184cfb7741601c90c6b33b5722ac77d7df10d02514721e38543f78538737218920bdd534ea3664bef46a5ba1bac594dec8961388518a1f717b1875569130b472
 
 install : $(TARGET)
 
@@ -73,7 +73,7 @@ $(subst %,%_BLAKE2,$(objects)) :
 $(TARGET) : $(patsubst %,$(DIR_DL)/%,$(objects))
 	@$(PREBUILD)
 	@rm -rf $(DIR_APP) && cd $(DIR_SRC) && tar axf $(DIR_DL)/$(DL_FILE)
-	cd $(DIR_APP)/kpartx && make $(MAKETUNING) CPPFLAGS=
+	cd $(DIR_APP) && make $(MAKETUNING) kpartx
 	cd $(DIR_APP)/kpartx && make install PREFIX=/usr
 	@rm -rf $(DIR_APP)
 	@$(POSTBUILD)
-- 
2.56.1



      reply	other threads:[~2026-11-20 20:11 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-20 20:11 [PATCH] core207: Ship multipath-tools Adolf Belka
2026-11-20 20:11 ` Adolf Belka [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20261120201136.642750-2-adolf.belka@ipfire.org \
    --to=adolf.belka@ipfire.org \
    --cc=development@lists.ipfire.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox