From: Benjamin Marzinski <bmarzins@redhat.com>
To: Martin Wilck <martin.wilck@suse.com>
Cc: Christophe Varoqui <christophe.varoqui@opensvc.com>,
dm-devel@lists.linux.dev, Martin Wilck <mwilck@suse.com>
Subject: Re: [PATCH 12/21] libmultipath: free_multipath: fix FREE_PATHS case
Date: Thu, 18 Dec 2025 19:15:04 -0500 [thread overview]
Message-ID: <aUSZCCPwZk_qATDx@redhat.com> (raw)
In-Reply-To: <20251217212113.234959-13-mwilck@suse.com>
On Wed, Dec 17, 2025 at 10:21:04PM +0100, Martin Wilck wrote:
> free_multipath was passing FREE_PATHS to both free_pgvec() and
> free_pathvec(), which was wrong because a path could be referenced
> in both structures. This can't happen any more now, as free_pgvec()
> has lost the ability to free paths.
>
> In general, free_multipath can't make any assumptions about whether
> a given path is referenced in mpp->paths, mpp->pg, or both. Therefore
> look first for paths that are referenced in mpp->pg but not in mpp->paths,
> clean them, and finally loop over mpp->paths.
Like I said in 07/21, I don't think we should free paths here. The paths
linked from the multipath device are also in the pathvec. We should
always be freeing them from there, so we don't have a pathvec with freed
paths.
-Ben
> Signed-off-by: Martin Wilck <mwilck@suse.com>
> ---
> libmultipath/structs.c | 44 +++++++++++++++++++++++++++++-------------
> 1 file changed, 31 insertions(+), 13 deletions(-)
>
> diff --git a/libmultipath/structs.c b/libmultipath/structs.c
> index 1ee29f6..29b62e4 100644
> --- a/libmultipath/structs.c
> +++ b/libmultipath/structs.c
> @@ -298,8 +298,12 @@ void free_multipath_attributes(struct multipath *mpp)
> }
>
> void
> -free_multipath (struct multipath * mpp, enum free_path_mode free_paths)
> +free_multipath (struct multipath *mpp, enum free_path_mode free_paths)
> {
> + struct pathgroup *pgp;
> + struct path *pp;
> + int i, j;
> +
> if (!mpp)
> return;
>
> @@ -310,22 +314,36 @@ free_multipath (struct multipath * mpp, enum free_path_mode free_paths)
> mpp->alias = NULL;
> }
>
> - if (free_paths == KEEP_PATHS && mpp->pg) {
> - struct pathgroup *pgp;
> - struct path *pp;
> - int i, j;
> -
> - /*
> - * Make sure paths carry no reference to this mpp any more
> - */
> - vector_foreach_slot(mpp->pg, pgp, i) {
> - vector_foreach_slot(pgp->paths, pp, j)
> - if (pp->mpp == mpp)
> + /*
> + * If free_paths == FREE_PATHS, free all paths.
> + * Otherwise, make sure no references to this mpp remain.
> + *
> + * First loop over mpp->pg and clean the paths that might
> + * be missing in mpp->paths. Then clean the rest
> + * by looping over mpp->paths.
> + */
> + vector_foreach_slot(mpp->pg, pgp, i) {
> + vector_foreach_slot(pgp->paths, pp, j) {
> + if (find_slot(mpp->paths, pp) == -1) {
> + if (free_paths == FREE_PATHS)
> + free_path(pp);
> + else
> pp->mpp = NULL;
> + }
> }
> }
>
> - free_pathvec(mpp->paths, free_paths);
> + vector_foreach_slot(mpp->paths, pp, i) {
> + if (free_paths == FREE_PATHS)
> + free_path(pp);
> + else
> + pp->mpp = NULL;
> + }
> +
> + /*
> + * We've freed paths above already, use KEEP_PATHS here
> + */
> + free_pathvec(mpp->paths, KEEP_PATHS);
> free_pgvec(mpp->pg);
> if (mpp->hwe) {
> vector_free(mpp->hwe);
> --
> 2.52.0
next prev parent reply other threads:[~2025-12-19 0:15 UTC|newest]
Thread overview: 33+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-12-17 21:20 [PATCH 00/21] Multipath-tools: various bug fixes Martin Wilck
2025-12-17 21:20 ` [PATCH 01/21] libmultipath: drop drop_multipath Martin Wilck
2025-12-17 21:20 ` [PATCH 02/21] libmultipath: don't access path members in free_pgvec() Martin Wilck
2025-12-17 21:20 ` [PATCH 03/21] multipathd: free paths in checker_finished() Martin Wilck
2025-12-18 23:34 ` Benjamin Marzinski
2025-12-18 23:40 ` Martin Wilck
2025-12-19 10:38 ` Martin Wilck
2025-12-17 21:20 ` [PATCH 04/21] libmultipath: don't touch mpvec in remove_map() Martin Wilck
2025-12-17 21:20 ` [PATCH 05/21] libmpathutil: constify find_slot() Martin Wilck
2025-12-17 21:20 ` [PATCH 06/21] libmultipath: don't free paths in orphan_paths() Martin Wilck
2025-12-17 21:20 ` [PATCH 07/21] libmultipath: free orphaned paths in check_removed_paths() Martin Wilck
2025-12-17 21:21 ` [PATCH 08/21] libmultipath: remove free_paths argument from free_pathgroup() Martin Wilck
2025-12-17 21:21 ` [PATCH 09/21] libmultipath: fix numeric value of free_paths in free_multipaths() Martin Wilck
2025-12-19 0:10 ` Benjamin Marzinski
2025-12-17 21:21 ` [PATCH 10/21] libmultipath: remove free_paths argument from free_pgvec() Martin Wilck
2025-12-17 21:21 ` [PATCH 11/21] libmultipath: remove free_paths argument from free_multipathvec() Martin Wilck
2025-12-17 21:21 ` [PATCH 12/21] libmultipath: free_multipath: fix FREE_PATHS case Martin Wilck
2025-12-19 0:15 ` Benjamin Marzinski [this message]
2025-12-17 21:21 ` [PATCH 13/21] multipath-tools: Fix ISO C23 errors with strchr() Martin Wilck
2025-12-17 21:21 ` [PATCH 14/21] libmultipath: simplify sysfs_get_target_nodename() Martin Wilck
2025-12-17 21:21 ` [PATCH 15/21] multipathd: join the init_unwinder dummy thread Martin Wilck
2025-12-17 21:21 ` [PATCH 16/21] kpartx: fix some memory leaks Martin Wilck
2025-12-17 21:21 ` [PATCH 17/21] libmpathutil: use union for bitfield Martin Wilck
2025-12-19 0:17 ` Benjamin Marzinski
2025-12-19 8:08 ` Martin Wilck
2025-12-17 21:21 ` [PATCH 18/21] libmpathutil: add wrapper code for libudev Martin Wilck
2025-12-19 0:38 ` Benjamin Marzinski
2025-12-19 8:06 ` Martin Wilck
2025-12-19 17:27 ` Benjamin Marzinski
2025-12-17 21:21 ` [PATCH 19/21] multipath-tools: use the libudev wrapper functions Martin Wilck
2025-12-17 21:21 ` [PATCH 20/21] Makefile: add functionality to determine cmocka version Martin Wilck
2025-12-17 21:21 ` [PATCH 21/21] multipath-tools tests: adaptations for cmocka 2.0 Martin Wilck
2025-12-19 0:40 ` [PATCH 00/21] Multipath-tools: various bug fixes Benjamin Marzinski
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=aUSZCCPwZk_qATDx@redhat.com \
--to=bmarzins@redhat.com \
--cc=christophe.varoqui@opensvc.com \
--cc=dm-devel@lists.linux.dev \
--cc=martin.wilck@suse.com \
--cc=mwilck@suse.com \
/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