device-mapper-multipath-0.8.7-29

This commit is contained in:
Benjamin Marzinski 2024-05-22 00:09:52 -04:00 committed by root
parent 2d7060ef7c
commit 665015b858
12 changed files with 1160 additions and 1 deletions

View File

@ -0,0 +1,2 @@
067d668de8e3a70b7c176bbf0c0616d5835bbe44 multipath-tools-0.8.7.tgz
5d5e16cccc83fd78cf9b95e5c52cc41dbbeb1da7 multipath.conf

View File

@ -0,0 +1,194 @@
From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
From: Benjamin Marzinski <bmarzins@redhat.com>
Date: Tue, 19 Dec 2023 17:51:50 -0500
Subject: [PATCH] libmultipath: keep track of queueing state in features
Make multipathd update mpp->features when in enables or disables
queuing. This patch handles all the cases except failed removes by
dm_suspend_and_flush_map(), which is never called by multipathd.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Reviewed-by: Martin Wilck <mwilck@suse.com>
---
libmultipath/configure.c | 4 +---
libmultipath/devmapper.c | 23 +++++++++++++++++++----
libmultipath/devmapper.h | 2 +-
libmultipath/structs_vec.c | 10 +++++-----
multipathd/main.c | 8 ++++----
5 files changed, 30 insertions(+), 17 deletions(-)
diff --git a/libmultipath/configure.c b/libmultipath/configure.c
index bbdbb8ca..71acb968 100644
--- a/libmultipath/configure.c
+++ b/libmultipath/configure.c
@@ -1279,9 +1279,7 @@ int coalesce_paths (struct vectors *vecs, vector mpvec, char *refwwid,
mpp->no_path_retry != NO_PATH_RETRY_FAIL)
condlog(3, "%s: multipathd not running, unset "
"queue_if_no_path feature", mpp->alias);
- if (!dm_queue_if_no_path(mpp->alias, 0))
- remove_feature(&mpp->features,
- "queue_if_no_path");
+ dm_queue_if_no_path(mpp, 0);
}
if (!is_daemon && mpp->action != ACT_NOTHING)
diff --git a/libmultipath/devmapper.c b/libmultipath/devmapper.c
index f9de3358..5711f0ee 100644
--- a/libmultipath/devmapper.c
+++ b/libmultipath/devmapper.c
@@ -57,6 +57,7 @@ static int dm_cancel_remove_partmaps(const char * mapname);
static int do_foreach_partmaps(const char * mapname,
int (*partmap_func)(const char *, void *),
void *data);
+static int _dm_queue_if_no_path(const char *mapname, int enable);
#ifndef LIBDM_API_COOKIE
static inline int dm_task_set_cookie(struct dm_task *dmt, uint32_t *c, int a)
@@ -1072,7 +1073,7 @@ int _dm_flush_map (const char * mapname, int need_sync, int deferred_remove,
if (need_suspend &&
dm_get_map(mapname, &mapsize, &params) == DMP_OK &&
strstr(params, "queue_if_no_path")) {
- if (!dm_queue_if_no_path(mapname, 0))
+ if (!_dm_queue_if_no_path(mapname, 0))
queue_if_no_path = 1;
else
/* Leave queue_if_no_path alone if unset failed */
@@ -1121,7 +1122,7 @@ int _dm_flush_map (const char * mapname, int need_sync, int deferred_remove,
} while (retries-- > 0);
if (queue_if_no_path == 1)
- dm_queue_if_no_path(mapname, 1);
+ _dm_queue_if_no_path(mapname, 1);
return 1;
}
@@ -1236,8 +1237,8 @@ dm_reinstate_path(const char * mapname, char * path)
return dm_message(mapname, message);
}
-int
-dm_queue_if_no_path(const char *mapname, int enable)
+static int
+_dm_queue_if_no_path(const char *mapname, int enable)
{
char *message;
@@ -1249,6 +1250,20 @@ dm_queue_if_no_path(const char *mapname, int enable)
return dm_message(mapname, message);
}
+int dm_queue_if_no_path(struct multipath *mpp, int enable)
+{
+ int r;
+ static const char no_path_retry[] = "queue_if_no_path";
+
+ if ((r = _dm_queue_if_no_path(mpp->alias, enable)) == 0) {
+ if (enable)
+ add_feature(&mpp->features, no_path_retry);
+ else
+ remove_feature(&mpp->features, no_path_retry);
+ }
+ return r;
+}
+
static int
dm_groupmsg (const char * msg, const char * mapname, int index)
{
diff --git a/libmultipath/devmapper.h b/libmultipath/devmapper.h
index 808da28d..41b8c31d 100644
--- a/libmultipath/devmapper.h
+++ b/libmultipath/devmapper.h
@@ -58,7 +58,7 @@ int dm_cancel_deferred_remove(struct multipath *mpp);
int dm_flush_maps (int retries);
int dm_fail_path(const char * mapname, char * path);
int dm_reinstate_path(const char * mapname, char * path);
-int dm_queue_if_no_path(const char *mapname, int enable);
+int dm_queue_if_no_path(struct multipath *mpp, int enable);
int dm_switchgroup(const char * mapname, int index);
int dm_enablegroup(const char * mapname, int index);
int dm_disablegroup(const char * mapname, int index);
diff --git a/libmultipath/structs_vec.c b/libmultipath/structs_vec.c
index 86ad89ca..56915e96 100644
--- a/libmultipath/structs_vec.c
+++ b/libmultipath/structs_vec.c
@@ -579,7 +579,7 @@ static void leave_recovery_mode(struct multipath *mpp)
*/
if (recovery && (mpp->no_path_retry == NO_PATH_RETRY_QUEUE ||
mpp->no_path_retry > 0)) {
- dm_queue_if_no_path(mpp->alias, 1);
+ dm_queue_if_no_path(mpp, 1);
condlog(2, "%s: queue_if_no_path enabled", mpp->alias);
condlog(1, "%s: Recovered to normal mode", mpp->alias);
}
@@ -598,11 +598,11 @@ void __set_no_path_retry(struct multipath *mpp, bool check_features)
break;
case NO_PATH_RETRY_FAIL:
if (!check_features || is_queueing)
- dm_queue_if_no_path(mpp->alias, 0);
+ dm_queue_if_no_path(mpp, 0);
break;
case NO_PATH_RETRY_QUEUE:
if (!check_features || !is_queueing)
- dm_queue_if_no_path(mpp->alias, 1);
+ dm_queue_if_no_path(mpp, 1);
break;
default:
if (count_active_paths(mpp) > 0) {
@@ -612,7 +612,7 @@ void __set_no_path_retry(struct multipath *mpp, bool check_features)
*/
if ((!check_features || !is_queueing) &&
!mpp->in_recovery)
- dm_queue_if_no_path(mpp->alias, 1);
+ dm_queue_if_no_path(mpp, 1);
leave_recovery_mode(mpp);
} else {
/*
@@ -623,7 +623,7 @@ void __set_no_path_retry(struct multipath *mpp, bool check_features)
*/
if ((!check_features || is_queueing) &&
mpp->in_recovery && mpp->retry_tick == 0)
- dm_queue_if_no_path(mpp->alias, 0);
+ dm_queue_if_no_path(mpp, 0);
if (pathcount(mpp, PATH_PENDING) == 0)
enter_recovery_mode(mpp);
}
diff --git a/multipathd/main.c b/multipathd/main.c
index 26be6dc3..74f8114c 100644
--- a/multipathd/main.c
+++ b/multipathd/main.c
@@ -502,7 +502,7 @@ flush_map_nopaths(struct multipath *mpp, struct vectors *vecs) {
mpp->no_path_retry = NO_PATH_RETRY_FAIL;
mpp->disable_queueing = 1;
mpp->stat_map_failures++;
- dm_queue_if_no_path(mpp->alias, 0);
+ dm_queue_if_no_path(mpp, 0);
}
r = dm_flush_map_nopaths(mpp->alias, mpp->deferred_remove);
if (r) {
@@ -833,7 +833,7 @@ uev_remove_map (struct uevent * uev, struct vectors * vecs)
goto out;
}
- dm_queue_if_no_path(alias, 0);
+ dm_queue_if_no_path(mpp, 0);
remove_map_and_stop_waiter(mpp, vecs);
out:
lock_cleanup_pop(vecs->lock);
@@ -2015,7 +2015,7 @@ retry_count_tick(vector mpvec)
condlog(4, "%s: Retrying.. No active path", mpp->alias);
if(--mpp->retry_tick == 0) {
mpp->stat_map_failures++;
- dm_queue_if_no_path(mpp->alias, 0);
+ dm_queue_if_no_path(mpp, 0);
condlog(2, "%s: Disable queueing", mpp->alias);
}
}
@@ -3110,7 +3110,7 @@ static void cleanup_maps(struct vectors *vecs)
put_multipath_config(conf);
if (queue_without_daemon == QUE_NO_DAEMON_OFF)
vector_foreach_slot(vecs->mpvec, mpp, i)
- dm_queue_if_no_path(mpp->alias, 0);
+ dm_queue_if_no_path(mpp, 0);
remove_maps_and_stop_waiters(vecs);
vecs->mpvec = NULL;
}

View File

@ -0,0 +1,54 @@
From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
From: Benjamin Marzinski <bmarzins@redhat.com>
Date: Thu, 25 Apr 2024 19:35:13 -0400
Subject: [PATCH] libmultipath: export partmap_in_use
A future commit will make use of this function
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Reviewed-by: Martin Wilck <mwilck@suse.com>
Reviewed-by: Benjamin Marzinski <bmarzins@redhat.com>
---
libmultipath/devmapper.c | 2 +-
libmultipath/devmapper.h | 1 +
libmultipath/libmultipath.version | 5 +++++
3 files changed, 7 insertions(+), 1 deletion(-)
diff --git a/libmultipath/devmapper.c b/libmultipath/devmapper.c
index 5711f0ee..4a66e2c4 100644
--- a/libmultipath/devmapper.c
+++ b/libmultipath/devmapper.c
@@ -1028,7 +1028,7 @@ has_partmap(const char *name __attribute__((unused)),
return 1;
}
-static int
+int
partmap_in_use(const char *name, void *data)
{
int part_count, *ret_count = (int *)data;
diff --git a/libmultipath/devmapper.h b/libmultipath/devmapper.h
index 41b8c31d..88e0b114 100644
--- a/libmultipath/devmapper.h
+++ b/libmultipath/devmapper.h
@@ -48,6 +48,7 @@ int dm_get_map(const char *, unsigned long long *, char **);
int dm_get_status(const char *, char **);
int dm_type(const char *, char *);
int dm_is_mpath(const char *);
+int partmap_in_use(const char *name, void *data);
int _dm_flush_map (const char *, int, int, int, int);
int dm_flush_map_nopaths(const char * mapname, int deferred_remove);
#define dm_flush_map(mapname) _dm_flush_map(mapname, 1, 0, 0, 0)
diff --git a/libmultipath/libmultipath.version b/libmultipath/libmultipath.version
index 1d018eab..40d9246d 100644
--- a/libmultipath/libmultipath.version
+++ b/libmultipath/libmultipath.version
@@ -302,3 +302,8 @@ LIBMULTIPATH_9.1.2 {
global:
cleanup_mutex;
} LIBMULTIPATH_9.1.1;
+
+LIBMULTIPATH_9.1.3 {
+global:
+ partmap_in_use;
+} LIBMULTIPATH_9.1.2;

View File

@ -0,0 +1,346 @@
From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
From: Benjamin Marzinski <bmarzins@redhat.com>
Date: Thu, 25 Apr 2024 19:35:14 -0400
Subject: [PATCH] libmultipath: change flush_on_last_del to fix a multipathd
hang
Commit 9bd3482e ("multipathd: make flush_map() delete maps like the
multipath command") fixed an issue where deleting a queueing multipath
device through multipathd could hang because the multipath device had
outstanding IO, even though the only openers of it at the time of
deletion were the kpartx partition devices. However it is still possible
to hang multipathd, because autoremoving the device when all paths have
been deleted doesn't disable queueing. To reproduce this hang:
1. create a multipath device with a kpartx partition on top of it and
no_path_retry set to either "queue" or something long enough to run all
the commands in the reproducer before it disables queueing.
2. disable all the paths to the device with something like:
# echo offline > /sys/block/<path_dev>/device/state
3. Write directly to the multipath device with something like:
# dd if=/dev/zero of=/dev/mapper/<mpath_dev> bs=4K count=1
4. delete all the paths to the device with something like:
# echo 1 > /sys/block/<path_dev>/device/delete
Multipathd will hang trying to delete the kpartx device because, as the
last opener, it must flush the multipath device before closing it.
Because it hangs holding the vecs_lock, multipathd will never disable
queueing on the device, so it will hang forever, even if no_path_retry
is set to a number.
This hang can occur, even if deferred_remove is set. Since nothing has
the kpartx device opened, device-mapper does an immediate remove, which
will still hang. This means that even if deferred_remove is set,
multipathd still cannot delete a map while queueing is enabled. It must
either disable queueing or skip the autoremove.
Mulitpath can currently be configured to avoid this hang by setting
flush_on_last_del yes
However there are good reasons why users wouldn't want to set that. They
may need to be able to survive having all paths getting temporarily
deleted. I should note that this is a pretty rare corner case, since
multipath automatically sets dev_loss_tmo so that it should not trigger
before queueing is disabled.
This commit avoids the hang by changing the possible values for
flush_on_last_del to "never", "unused", and "always", and sets the
default to "unused". "always" works like "yes" did, "never" will not
disable queueing, and "unused" will only disable queueing if nothing has
the kpartx devices or the multipath device open. In order to be safe, if
the device has queue_if_no_paths set (and in case of "unused", the
device is in-use) the autoremove will be skipped. Also, instead of just
trusting the lack of "queue_if_no_paths" in the current mpp->features,
multipathd will tell the kernel to disable queueing, just to be sure it
actually is.
I chose "unused" as the default because this should generally only cause
multipathd to work differently from the users perspective when nothing
has the multipath device open but it is queueing and there is
outstanding IO. Without this change, it would have hung instead of
failing the outstanding IO. However, I do understand that an argument
could be made that "never" makes more sense as default, even though it
will cause multipathd to skip autoremoves in cases where it wouldn't
before. The change to the behavior of deffered_remove will be
noticeable, but skipping an autoremove is much better than hanging.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Reviewed-by: Martin Wilck <mwilck@suse.com>
Reviewed-by: Benjamin Marzinski <bmarzins@redhat.com>
---
libmultipath/defaults.h | 2 +-
libmultipath/dict.c | 72 +++++++++++++++++++++++++++++++++-----
libmultipath/dict.h | 1 +
libmultipath/hwtable.c | 6 ++--
libmultipath/propsel.c | 4 ++-
libmultipath/structs.h | 7 ++--
multipath/multipath.conf.5 | 20 ++++++++---
multipathd/main.c | 39 ++++++++++++++++-----
8 files changed, 122 insertions(+), 29 deletions(-)
diff --git a/libmultipath/defaults.h b/libmultipath/defaults.h
index c3788bbc..5e77387e 100644
--- a/libmultipath/defaults.h
+++ b/libmultipath/defaults.h
@@ -41,7 +41,7 @@
#define DEFAULT_PRIO PRIO_CONST
#define DEFAULT_PRIO_ARGS ""
#define DEFAULT_CHECKER TUR
-#define DEFAULT_FLUSH FLUSH_DISABLED
+#define DEFAULT_FLUSH FLUSH_UNUSED
#define DEFAULT_USER_FRIENDLY_NAMES USER_FRIENDLY_NAMES_OFF
#define DEFAULT_FORCE_SYNC 0
#define UNSET_PARTITION_DELIM "/UNSET/"
diff --git a/libmultipath/dict.c b/libmultipath/dict.c
index ce1b6c99..3c011ece 100644
--- a/libmultipath/dict.c
+++ b/libmultipath/dict.c
@@ -825,14 +825,70 @@ declare_def_snprint(checker_timeout, print_nonzero)
declare_def_handler(allow_usb_devices, set_yes_no)
declare_def_snprint(allow_usb_devices, print_yes_no)
-declare_def_handler(flush_on_last_del, set_yes_no_undef)
-declare_def_snprint_defint(flush_on_last_del, print_yes_no_undef, DEFAULT_FLUSH)
-declare_ovr_handler(flush_on_last_del, set_yes_no_undef)
-declare_ovr_snprint(flush_on_last_del, print_yes_no_undef)
-declare_hw_handler(flush_on_last_del, set_yes_no_undef)
-declare_hw_snprint(flush_on_last_del, print_yes_no_undef)
-declare_mp_handler(flush_on_last_del, set_yes_no_undef)
-declare_mp_snprint(flush_on_last_del, print_yes_no_undef)
+
+static const char * const flush_on_last_del_optvals[] = {
+ [FLUSH_NEVER] = "never",
+ [FLUSH_ALWAYS] = "always",
+ [FLUSH_UNUSED] = "unused",
+};
+
+static int
+set_flush_on_last_del(vector strvec, void *ptr, const char *file, int line_nr)
+{
+ int i;
+ int *flush_val_ptr = (int *)ptr;
+ char *buff;
+
+ buff = set_value(strvec);
+ if (!buff)
+ return 1;
+
+ for (i = FLUSH_NEVER; i <= FLUSH_UNUSED; i++) {
+ if (flush_on_last_del_optvals[i] != NULL &&
+ !strcmp(buff, flush_on_last_del_optvals[i])) {
+ *flush_val_ptr = i;
+ break;
+ }
+ }
+
+ if (i > FLUSH_UNUSED) {
+ bool deprecated = true;
+ if (strcmp(buff, "no") == 0 || strcmp(buff, "0") == 0)
+ *flush_val_ptr = FLUSH_UNUSED;
+ else if (strcmp(buff, "yes") == 0 || strcmp(buff, "1") == 0)
+ *flush_val_ptr = FLUSH_ALWAYS;
+ else {
+ deprecated = false;
+ condlog(1, "%s line %d, invalid value for flush_on_last_del: \"%s\"",
+ file, line_nr, buff);
+ }
+ if (deprecated)
+ condlog(3, "%s line %d, \"%s\" is a deprecated value for flush_on_last_del and is treated as \"%s\"",
+ file, line_nr, buff,
+ flush_on_last_del_optvals[*flush_val_ptr]);
+ }
+
+ free(buff);
+ return 0;
+}
+
+int
+print_flush_on_last_del(struct strbuf *buff, long v)
+{
+ if (v == FLUSH_UNDEF)
+ return 0;
+ return append_strbuf_quoted(buff, flush_on_last_del_optvals[(int)v]);
+}
+
+declare_def_handler(flush_on_last_del, set_flush_on_last_del)
+declare_def_snprint_defint(flush_on_last_del, print_flush_on_last_del,
+ DEFAULT_FLUSH)
+declare_ovr_handler(flush_on_last_del, set_flush_on_last_del)
+declare_ovr_snprint(flush_on_last_del, print_flush_on_last_del)
+declare_hw_handler(flush_on_last_del, set_flush_on_last_del)
+declare_hw_snprint(flush_on_last_del, print_flush_on_last_del)
+declare_mp_handler(flush_on_last_del, set_flush_on_last_del)
+declare_mp_snprint(flush_on_last_del, print_flush_on_last_del)
declare_def_handler(user_friendly_names, set_yes_no_undef)
declare_def_snprint_defint(user_friendly_names, print_yes_no_undef,
diff --git a/libmultipath/dict.h b/libmultipath/dict.h
index d963b4ad..6b1aae5c 100644
--- a/libmultipath/dict.h
+++ b/libmultipath/dict.h
@@ -20,4 +20,5 @@ int print_reservation_key(struct strbuf *buff,
struct be64 key, uint8_t flags, int source);
int print_off_int_undef(struct strbuf *buff, long v);
int print_auto_resize(struct strbuf *buff, long v);
+int print_flush_on_last_del(struct strbuf *buff, long v);
#endif /* _DICT_H */
diff --git a/libmultipath/hwtable.c b/libmultipath/hwtable.c
index 78ac7988..94012d50 100644
--- a/libmultipath/hwtable.c
+++ b/libmultipath/hwtable.c
@@ -60,7 +60,7 @@
.no_path_retry = NO_PATH_RETRY_UNDEF,
.minio = 1000,
.minio_rq = 1,
- .flush_on_last_del = FLUSH_DISABLED,
+ .flush_on_last_del = FLUSH_UNUSED,
.user_friendly_names = USER_FRIENDLY_NAMES_OFF,
.fast_io_fail = 5,
.dev_loss = 600,
@@ -800,7 +800,7 @@ static struct hwentry default_hw[] = {
.no_path_retry = NO_PATH_RETRY_QUEUE,
.pgpolicy = GROUP_BY_PRIO,
.pgfailback = -FAILBACK_IMMEDIATE,
- .flush_on_last_del = FLUSH_ENABLED,
+ .flush_on_last_del = FLUSH_ALWAYS,
.dev_loss = MAX_DEV_LOSS_TMO,
.prio_name = PRIO_ONTAP,
.user_friendly_names = USER_FRIENDLY_NAMES_OFF,
@@ -1122,7 +1122,7 @@ static struct hwentry default_hw[] = {
.no_path_retry = NO_PATH_RETRY_FAIL,
.minio = 1,
.minio_rq = 1,
- .flush_on_last_del = FLUSH_ENABLED,
+ .flush_on_last_del = FLUSH_ALWAYS,
.fast_io_fail = 15,
.dev_loss = 15,
},
diff --git a/libmultipath/propsel.c b/libmultipath/propsel.c
index 9dea6f92..be781ff7 100644
--- a/libmultipath/propsel.c
+++ b/libmultipath/propsel.c
@@ -902,6 +902,7 @@ out:
int select_flush_on_last_del(struct config *conf, struct multipath *mp)
{
const char *origin;
+ STRBUF_ON_STACK(buff);
mp_set_mpe(flush_on_last_del);
mp_set_ovr(flush_on_last_del);
@@ -909,8 +910,9 @@ int select_flush_on_last_del(struct config *conf, struct multipath *mp)
mp_set_conf(flush_on_last_del);
mp_set_default(flush_on_last_del, DEFAULT_FLUSH);
out:
+ print_flush_on_last_del(&buff, mp->flush_on_last_del);
condlog(3, "%s: flush_on_last_del = %s %s", mp->alias,
- (mp->flush_on_last_del == FLUSH_ENABLED)? "yes" : "no", origin);
+ get_strbuf_str(&buff), origin);
return 0;
}
diff --git a/libmultipath/structs.h b/libmultipath/structs.h
index d2ad4867..4bf8c93a 100644
--- a/libmultipath/structs.h
+++ b/libmultipath/structs.h
@@ -109,9 +109,10 @@ enum marginal_pathgroups_mode {
};
enum flush_states {
- FLUSH_UNDEF = YNU_UNDEF,
- FLUSH_DISABLED = YNU_NO,
- FLUSH_ENABLED = YNU_YES,
+ FLUSH_UNDEF,
+ FLUSH_NEVER,
+ FLUSH_ALWAYS,
+ FLUSH_UNUSED,
};
enum log_checker_err_states {
diff --git a/multipath/multipath.conf.5 b/multipath/multipath.conf.5
index 38eb5c90..10eddc0c 100644
--- a/multipath/multipath.conf.5
+++ b/multipath/multipath.conf.5
@@ -672,12 +672,22 @@ The default is: \fBno\fR
.TP
.B flush_on_last_del
If set to
-.I yes
+.I always
, multipathd will disable queueing when the last path to a device has been
-deleted.
-.RS
-.TP
-The default is: \fBno\fR
+deleted. If set to
+.I never
+, multipathd will not disable queueing when the last path to a device has
+been deleted. Since multipath cannot safely remove a device while queueing
+is enabled, setting this to \fInever\fR means that multipathd will not
+automatically remove an unused multipath device whose paths are all deleted if
+it is currently set to queue_if_no_path. If set to
+.I unused
+, multipathd will only disable queueing when the last path is removed if
+nothing currently has the multipath device or any of the kpartx partition
+devices on top of it open.
+.RS
+.TP
+The default is: \fBunused\fR
.RE
.
.
diff --git a/multipathd/main.c b/multipathd/main.c
index 74f8114c..8ec58f5d 100644
--- a/multipathd/main.c
+++ b/multipathd/main.c
@@ -491,19 +491,42 @@ int update_multipath (struct vectors *vecs, char *mapname, int reset)
static bool
flush_map_nopaths(struct multipath *mpp, struct vectors *vecs) {
int r;
+ bool is_queueing = true;
+ if (mpp->features)
+ is_queueing = strstr(mpp->features, "queue_if_no_path");
+
+ /* It's not safe to do a remove of a map that has "queue_if_no_path"
+ * set, since there could be outstanding IO which will cause
+ * multipathd to hang while attempting the remove */
+ if (mpp->flush_on_last_del == FLUSH_NEVER && is_queueing) {
+ condlog(2, "%s: map is queueing, can't remove", mpp->alias);
+ return false;
+ }
+ if (mpp->flush_on_last_del == FLUSH_UNUSED &&
+ partmap_in_use(mpp->alias, NULL) && is_queueing) {
+ condlog(2, "%s: map in use and queueing, can't remove",
+ mpp->alias);
+ return false;
+ }
/*
- * flush_map will fail if the device is open
+ * This will flush FLUSH_NEVER devices and FLUSH_UNUSED devices
+ * that are in use, but only if they are already marked as not
+ * queueing. That is just to make absolutely certain that they
+ * really are not queueing, like they claim.
*/
- if (mpp->flush_on_last_del == FLUSH_ENABLED) {
- condlog(2, "%s Last path deleted, disabling queueing",
+ condlog(is_queueing ? 2 : 3, "%s Last path deleted, disabling queueing",
+ mpp->alias);
+ mpp->retry_tick = 0;
+ mpp->no_path_retry = NO_PATH_RETRY_FAIL;
+ mpp->disable_queueing = 1;
+ mpp->stat_map_failures++;
+ if (dm_queue_if_no_path(mpp, 0) != 0) {
+ condlog(0, "%s: failed to disable queueing. Not removing",
mpp->alias);
- mpp->retry_tick = 0;
- mpp->no_path_retry = NO_PATH_RETRY_FAIL;
- mpp->disable_queueing = 1;
- mpp->stat_map_failures++;
- dm_queue_if_no_path(mpp, 0);
+ return false;
}
+
r = dm_flush_map_nopaths(mpp->alias, mpp->deferred_remove);
if (r) {
if (r == 1)

View File

@ -0,0 +1,40 @@
From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
From: Benjamin Marzinski <bmarzins@redhat.com>
Date: Thu, 25 Apr 2024 19:35:16 -0400
Subject: [PATCH] libmultipath: pad dev_loss_tmo to avoid race with
no_path_retry
Currently multipath makes sure that dev_loss_tmo is at least as long as
the configured no path queueing time. The goal is to make sure that path
devices aren't removed while the multipath device is still queueing in
hopes that they will become usable again.
This is racy. Multipathd may take longer to check the paths than
configured. If strict_timing isn't set, it will definitely take longer.
To account for this, pad the minimum dev_loss_tmo value by five seconds
(one default checker interval) plus one second per minute of no path
queueing time.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
Reviewed-by: Martin Wilck <mwilck@suse.com>
Reviewed-by: Benjamin Marzinski <bmarzins@redhat.com>
---
libmultipath/discovery.c | 5 +++++
1 file changed, 5 insertions(+)
diff --git a/libmultipath/discovery.c b/libmultipath/discovery.c
index ae7eb7e6..b24594cd 100644
--- a/libmultipath/discovery.c
+++ b/libmultipath/discovery.c
@@ -884,6 +884,11 @@ sysfs_set_scsi_tmo (struct config *conf, struct multipath *mpp)
uint64_t no_path_retry_tmo =
(uint64_t)mpp->no_path_retry * conf->checkint;
+ /* pad no_path_retry_tmo by one standard check interval
+ * plus one second per minute to account for timing
+ * issues with the rechecks */
+ no_path_retry_tmo += no_path_retry_tmo / 60 + DEFAULT_CHECKINT;
+
if (no_path_retry_tmo > MAX_DEV_LOSS_TMO)
min_dev_loss = MAX_DEV_LOSS_TMO;
else

View File

@ -0,0 +1,32 @@
From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
From: Nitin Yewale <nyewale@redhat.com>
Date: Thu, 12 Jan 2023 14:28:49 -0600
Subject: [PATCH] libmultipath: remove pathgroup wildcard options
The multipathd command "multipathd show wildcards" shows the pathgroup
format wildcards, but there is no way to use them in a multipathd
command.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
---
libmultipath/print.c | 7 -------
1 file changed, 7 deletions(-)
diff --git a/libmultipath/print.c b/libmultipath/print.c
index 082e4e30..a6e4c774 100644
--- a/libmultipath/print.c
+++ b/libmultipath/print.c
@@ -803,13 +803,6 @@ int snprint_wildcards(struct strbuf *buff)
pd[i].wildcard, pd[i].header)) < 0)
return rc;
- if ((rc = append_strbuf_str(buff, "\npathgroup format wildcards:\n")) < 0)
- return rc;
- for (i = 0; pgd[i].header; i++)
- if ((rc = print_strbuf(buff, "%%%c %s\n",
- pgd[i].wildcard, pgd[i].header)) < 0)
- return rc;
-
return get_strbuf_len(buff) - initial_len;
}

View File

@ -0,0 +1,32 @@
From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
From: Benjamin Marzinski <bmarzins@redhat.com>
Date: Wed, 8 May 2024 19:02:30 -0400
Subject: [PATCH] libmultipath: print all values in snprint_failback
Add the missing output for manual failback and print the defferral time
for deferred failbacks, if one isn't currently in progress.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
---
libmultipath/print.c | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/libmultipath/print.c b/libmultipath/print.c
index a6e4c774..535e0271 100644
--- a/libmultipath/print.c
+++ b/libmultipath/print.c
@@ -199,9 +199,13 @@ snprint_failback (struct strbuf *buff, const struct multipath * mpp)
return append_strbuf_str(buff, "immediate");
if (mpp->pgfailback == -FAILBACK_FOLLOWOVER)
return append_strbuf_str(buff, "followover");
+ if (mpp->pgfailback == -FAILBACK_MANUAL)
+ return append_strbuf_str(buff, "manual");
+ if (mpp->pgfailback == FAILBACK_UNDEF)
+ return append_strbuf_str(buff, "undef");
if (!mpp->failback_tick)
- return append_strbuf_str(buff, "-");
+ return print_strbuf(buff, "%i", mpp->pgfailback);
else
return snprint_progress(buff, mpp->failback_tick,
mpp->pgfailback);

View File

@ -0,0 +1,34 @@
From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
From: Benjamin Marzinski <bmarzins@redhat.com>
Date: Thu, 9 May 2024 12:58:11 -0400
Subject: [PATCH] multipathd: Stop double counting map failures for
no_path_retry > 0
If no_path_retry was greater than 0, multipathd was counting a map
failure when recovery mode was entered, and again when queueing was
disabled. The first one is incorrect, since the map is still queueing.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
---
libmultipath/structs_vec.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/libmultipath/structs_vec.c b/libmultipath/structs_vec.c
index 56915e96..b8e304e0 100644
--- a/libmultipath/structs_vec.c
+++ b/libmultipath/structs_vec.c
@@ -781,10 +781,13 @@ int verify_paths(struct multipath *mpp)
void update_queue_mode_del_path(struct multipath *mpp)
{
int active = count_active_paths(mpp);
+ bool is_queueing = mpp->features &&
+ strstr(mpp->features, "queue_if_no_path");
if (active == 0) {
enter_recovery_mode(mpp);
- if (mpp->no_path_retry != NO_PATH_RETRY_QUEUE)
+ if (mpp->no_path_retry == NO_PATH_RETRY_FAIL ||
+ (mpp->no_path_retry == NO_PATH_RETRY_UNDEF && !is_queueing))
mpp->stat_map_failures++;
}
condlog(2, "%s: remaining active paths: %d", mpp->alias, active);

View File

@ -0,0 +1,102 @@
From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
From: Benjamin Marzinski <bmarzins@redhat.com>
Date: Thu, 9 May 2024 17:15:34 -0400
Subject: [PATCH] multipath-tools man pages: add missing multipathd commands
Also, the description for "del map $map" was incorrect.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
---
multipathd/multipathd.8 | 42 ++++++++++++++++++++++++++++++++++++-----
1 file changed, 37 insertions(+), 5 deletions(-)
diff --git a/multipathd/multipathd.8 b/multipathd/multipathd.8
index 8bd47a80..40a8dc6d 100644
--- a/multipathd/multipathd.8
+++ b/multipathd/multipathd.8
@@ -100,18 +100,24 @@ The following commands can be used in interactive mode:
Show the paths that multipathd is monitoring, and their state.
.
.TP
-.B list|show paths format $format
+.B list|show paths [raw] format $format
Show the paths that multipathd is monitoring, using a format string with path
-format wildcards.
+format wildcards. Adding \fIraw\fR will remove the headers and alignment
+padding from the ouput.
+.
+.TP
+.B list|show path $path
+Show whether path $path is offline or running.
.
.TP
.B list|show maps|multipaths
Show the multipath devices that the multipathd is monitoring.
.
.TP
-.B list|show maps|multipaths format $format
+.B list|show maps|multipaths [raw] format $format
Show the status of all multipath devices that the multipathd is monitoring,
-using a format string with multipath format wildcards.
+using a format string with multipath format wildcards. Adding \fIraw\fR will
+remove the headers and alignment padding from the output.
.
.TP
.B list|show maps|multipaths status
@@ -124,6 +130,10 @@ Show some statistics of all multipath devices that the multipathd is monitoring.
.TP
.B list|show maps|multipaths topology
Show the current multipath topology. Same as '\fImultipath \-ll\fR'.
+.TP
+.
+.B list|show maps|multipaths json
+Show information about all multipath devices in JSON format.
.
.TP
.B list|show topology
@@ -135,6 +145,16 @@ Show topology of a single multipath device specified by $map, for example
36005076303ffc56200000000000010aa. This map could be obtained from '\fIlist maps\fR'.
.
.TP
+.B list|show map|multipath $map [raw] format $format.
+Show the status of multipath device $map, using a format string with multipath
+format wildcards. Adding \fIraw\fR will remove the headers and alignment
+padding from the output.
+.
+.TP
+.B list|show map|multipath $map json
+Show information about multipath device $map in JSON format.
+.
+.TP
.B list|show wildcards
Show the format wildcards used in interactive commands taking $format.
.
@@ -168,6 +188,14 @@ paths, and whether multipathd is currently handling a uevent.
Show the current state of the multipathd daemon.
.
.TP
+.B reset maps|multipaths stats
+Reset the statistics of all multipath devices.
+.
+.TP
+.B reset map|multipath $map stats
+Reset the statistics of multipath device $map.
+.
+.TP
.B add path $path
Add a path to the list of monitored paths. $path is as listed in /sys/block (e.g. sda).
.
@@ -183,8 +211,12 @@ for the multipath device (e.g. mpath1) or the uid of the multipath device
(e.g. 36005076303ffc56200000000000010aa).
.
.TP
+.B remove|del maps|multipaths
+Remove all multipath devices.
+.
+.TP
.B remove|del map|multipath $map
-Stop monitoring a multipath device.
+Remove the multipath device $map.
.
.TP
.B resize map|multipath $map

View File

@ -0,0 +1,54 @@
From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
From: Benjamin Marzinski <bmarzins@redhat.com>
Date: Fri, 10 May 2024 15:36:10 -0400
Subject: [PATCH] libmultipath: change the vend/prod/rev printing
The %s multipath and path wildcards both say they print the device
vend/prod/rev string, but neither of them do. The multipath wildcards
already provide a way to print the revision string and the %s wildcard
is used in the multipath -l output, so leave the wildcard output alone,
and change the description to only mention the vendor and product. There
is no other way to print the revision by path, and the path %s wildcard
is only used in the verbose multipath output, so make it actually print
the revision. Also check for unset strings, and print "##" instead of
nothing.
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
---
libmultipath/print.c | 9 ++++++---
1 file changed, 6 insertions(+), 3 deletions(-)
diff --git a/libmultipath/print.c b/libmultipath/print.c
index 535e0271..4552fd43 100644
--- a/libmultipath/print.c
+++ b/libmultipath/print.c
@@ -309,7 +309,7 @@ snprint_multipath_uuid (struct strbuf *buff, const struct multipath * mpp)
}
static int
-snprint_multipath_vpr (struct strbuf *buff, const struct multipath * mpp)
+snprint_multipath_vp (struct strbuf *buff, const struct multipath * mpp)
{
struct pathgroup * pgp;
struct path * pp;
@@ -511,7 +511,10 @@ snprint_dm_path_state (struct strbuf *buff, const struct path * pp)
static int
snprint_vpr (struct strbuf *buff, const struct path * pp)
{
- return print_strbuf(buff, "%s,%s", pp->vendor_id, pp->product_id);
+ return print_strbuf(buff, "%s,%s,%s",
+ strlen(pp->vendor_id) ? pp->vendor_id : "##",
+ strlen(pp->product_id) ? pp->product_id : "##",
+ strlen(pp->rev) ? pp->rev : "##");
}
static int
@@ -743,7 +746,7 @@ struct multipath_data mpd[] = {
{'2', "map_loads", 0, snprint_map_loads},
{'3', "total_q_time", 0, snprint_total_q_time},
{'4', "q_timeouts", 0, snprint_q_timeouts},
- {'s', "vend/prod/rev", 0, snprint_multipath_vpr},
+ {'s', "vend/prod", 0, snprint_multipath_vp},
{'v', "vend", 0, snprint_multipath_vend},
{'p', "prod", 0, snprint_multipath_prod},
{'e', "rev", 0, snprint_multipath_rev},

View File

@ -0,0 +1,243 @@
From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
From: Benjamin Marzinski <bmarzins@redhat.com>
Date: Fri, 10 May 2024 20:53:33 -0400
Subject: [PATCH] multipath-tools man pages: Add format wildcard descriptions
Suggested-by: Nitin Yewale <nyewale@redhat.com>
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
---
multipathd/multipathd.8 | 193 +++++++++++++++++++++++++++++++++++++++-
1 file changed, 189 insertions(+), 4 deletions(-)
diff --git a/multipathd/multipathd.8 b/multipathd/multipathd.8
index 40a8dc6d..d834f89e 100644
--- a/multipathd/multipathd.8
+++ b/multipathd/multipathd.8
@@ -103,7 +103,7 @@ Show the paths that multipathd is monitoring, and their state.
.B list|show paths [raw] format $format
Show the paths that multipathd is monitoring, using a format string with path
format wildcards. Adding \fIraw\fR will remove the headers and alignment
-padding from the ouput.
+padding from the output. See "Path format wildcards" below.
.
.TP
.B list|show path $path
@@ -117,7 +117,8 @@ Show the multipath devices that the multipathd is monitoring.
.B list|show maps|multipaths [raw] format $format
Show the status of all multipath devices that the multipathd is monitoring,
using a format string with multipath format wildcards. Adding \fIraw\fR will
-remove the headers and alignment padding from the output.
+remove the headers and alignment padding from the output. See "Multipath
+format wildcards" below.
.
.TP
.B list|show maps|multipaths status
@@ -148,7 +149,7 @@ Show topology of a single multipath device specified by $map, for example
.B list|show map|multipath $map [raw] format $format.
Show the status of multipath device $map, using a format string with multipath
format wildcards. Adding \fIraw\fR will remove the headers and alignment
-padding from the output.
+padding from the output. See "Multipath format wildcards" below.
.
.TP
.B list|show map|multipath $map json
@@ -156,7 +157,8 @@ Show information about multipath device $map in JSON format.
.
.TP
.B list|show wildcards
-Show the format wildcards used in interactive commands taking $format.
+Show the format wildcards used in interactive commands taking $format. See
+"Format Wildcards" below.
.
.TP
.B list|show config
@@ -339,6 +341,189 @@ Stop multipathd.
.
.
.\" ----------------------------------------------------------------------------
+.SH "Format Wildcards"
+.\" ----------------------------------------------------------------------------
+.
+Multipathd commands that take a $format option require a format string. This
+string controls how a device is printed and should include format wildcards.
+When the devices are printed, these wildcards will be replaced by the
+appropriate device information. The following wildcards are supported.
+.TP
+.B Multipath format wildcards
+.RS
+.TP 12
+.B %n
+The device name.
+.TP
+.B %w
+The device WWID (uuid).
+.TP
+.B %d
+The device sysfs name (dm-<minor_nr>).
+.TP
+.B %F
+The device \fBfailback\fR setting. For deferred failbacks, it will either
+print the configured time if a deferred failback is not in progress, or
+it will show the current progress of a deferred failback.
+.TP
+.B %Q
+The device \fBno_path_retry\fR setting. If no_path_retry is set to a
+number of retires, it will either print the configured number of checker
+retries if the device is not in recovery mode, the number of seconds until
+queueing is disabled if the device is queueing in recovery mode, or \fIoff\fR
+if the device has disabled queueing.
+.TP
+.B %N
+The number of active paths for the device.
+.TP
+.B %r
+The device write-protect setting, either \fIro\fR or \fIrw\fR.
+.TP
+.B %t
+The device-mapper state of the device, either \fIsuspend\fR or \fIactive\fR.
+.TP
+.B %S
+The device size.
+.TP
+.B %f
+The device table features string.
+.TP
+.B %x
+The number of times the device has entered a state where it will fail IO.
+This is an alias for the \fB%4\fR wildcard.
+This value can be reset with the '\fIreset map $map stats\fR' command.
+.TP
+.B %h
+The device table hardware handler string.
+.TP
+.B %A
+The last action multipathd took on the device. This wildcard is for debugging
+use, as understanding its meaning requires looking at the code.
+.TP
+.B %0
+The number of times a path in the device has failed.
+This value can be reset with the '\fIreset map $map stats\fR' command.
+.TP
+.B %1
+The number of times multipathd has initiated a pathgroup switch for the device.
+This value can be reset with the '\fIreset map $map stats\fR' command.
+.TP
+.B %2
+The number of times multipathd has loaded a new table for the device.
+This value can be reset with the '\fIreset map $map stats\fR' command.
+.TP
+.B %3
+The approximate number of seconds that multipathd has spent queueing with
+no usable paths. This value can be reset with the '\fIreset map $map stats\fR'
+command.
+.TP
+.B %4
+The number of times the device has entered a state where it will fail IO.
+This is an alias for the \fB%x\fR wildcard.
+This value can be reset with the '\fIreset map $map stats\fR' command.
+.TP
+.B %s
+The vendor/product string for the device.
+.TP
+.B %v
+The array vendor string for the device.
+.TP
+.B %p
+The array product string for the device.
+.TP
+.B %e
+The array firmware revision string for the device.
+.TP
+.B %G
+The foreign library used for the device, or \fB--\fR for native device-mapper
+multipath devices.
+.TP
+.B %g
+Data from vendor specific vpd pages for the device, if any.
+.RE
+.
+.
+.TP
+.B Path format wildcards
+.RS
+.TP 12
+.B %w
+The device WWID (uuid).
+.TP
+.B %i
+The device Host:Channel:Id:Lun
+.TP
+.B %d
+The device sysfs name.
+.TP
+.B %D
+The device major:minor
+.TP
+.B %t
+The device-mapper state of the device, either \fIactive\fR or \fIfailed\fR.
+.TP
+.B %o
+Whether the device is \fIoffline\fR or \fIrunning\fR.
+.TP
+.B %T
+The multipathd path checker state of the device.
+.TP
+.B %s
+The vendor/product/revision string for the device.
+.TP
+.B %c
+The device's path checker name.
+.TP
+.B %C
+The progress towards the next path checker run on the device.
+.TP
+.B %p
+The device priority.
+.TP
+.B %S
+The device size.
+.TP
+.B %z
+The device serial number.
+.TP
+.B %M
+The device marginal state, either \fImarginal\fR or \fInormal\fR.
+.TP
+.B %m
+The multipath device that this device is a path of.
+.TP
+.B %N
+The host World Wide Node Name (WWNN) of the device.
+.TP
+.B %n
+The target World Wide Node Name (WWNN) of the device.
+.TP
+.B %R
+The host World Wide Port Name (WWPN) of the device.
+.TP
+.B %r
+The target World Wide Port Name (WWPN) of the device.
+.TP
+.B %a
+The host adapter name for the device (only SCSI devices).
+.TP
+.B %G
+The foreign library used for the device, or \fB--\fR for paths of native
+device-mapper multipath devices.
+.TP
+.B %g
+Data from vendor specific vpd pages for the device, if any.
+.TP
+.B %0
+The number of times this device has failed.
+.TP
+.B %P
+The device protocol. This output can be used for \fIprotocol\fR blacklist
+entries.
+.RE
+.
+.
+.\" ----------------------------------------------------------------------------
.SH "SYSTEMD INTEGRATION"
.\" ----------------------------------------------------------------------------
.

View File

@ -1,6 +1,6 @@
Name: device-mapper-multipath
Version: 0.8.7
Release: 28%{?dist}
Release: 29%{?dist}
Summary: Tools to manage multipath devices using device-mapper
License: GPLv2
URL: http://christophe.varoqui.free.fr/
@ -119,6 +119,16 @@ Patch0106: 0106-multipathd-fix-auto-resize-configuration.patch
Patch0107: 0107-libmultipath-fix-displaying-auto_resize-config-setti.patch
Patch0108: 0108-libmultipath-actually-truncate-too-large-vpd-page.patch
Patch0109: 0109-kpartx-fix-theoretical-overflow-in-loop-device-name.patch
Patch0110: 0110-libmultipath-keep-track-of-queueing-state-in-feature.patch
Patch0111: 0111-libmultipath-export-partmap_in_use.patch
Patch0112: 0112-libmultipath-change-flush_on_last_del-to-fix-a-multi.patch
Patch0113: 0113-libmultipath-pad-dev_loss_tmo-to-avoid-race-with-no_.patch
Patch0114: 0114-libmultipath-remove-pathgroup-wildcard-options.patch
Patch0115: 0115-libmultipath-print-all-values-in-snprint_failback.patch
Patch0116: 0116-multipathd-Stop-double-counting-map-failures-for-no_.patch
Patch0117: 0117-multipath-tools-man-pages-add-missing-multipathd-com.patch
Patch0118: 0118-libmultipath-change-the-vend-prod-rev-printing.patch
Patch0119: 0119-multipath-tools-man-pages-Add-format-wildcard-descri.patch
# runtime
@ -322,6 +332,22 @@ fi
%{_pkgconfdir}/libdmmp.pc
%changelog
* Tue May 21 2024 Benjamin Marzinski <bmarzins@redhat.com> - 0.8.7-29
- Add 0110-libmultipath-keep-track-of-queueing-state-in-feature.patch
- Add 0111-libmultipath-export-partmap_in_use.patch
- Add 0112-libmultipath-change-flush_on_last_del-to-fix-a-multi.patch
- Add 0113-libmultipath-pad-dev_loss_tmo-to-avoid-race-with-no_.patch
* Fixes RHEL-30272
- Add 0114-libmultipath-remove-pathgroup-wildcard-options.patch
- Add 0115-libmultipath-print-all-values-in-snprint_failback.patch
- Add 0116-multipathd-Stop-double-counting-map-failures-for-no_.patch
- Add 0117-multipath-tools-man-pages-add-missing-multipathd-com.patch
- Add 0118-libmultipath-change-the-vend-prod-rev-printing.patch
- Add 0119-multipath-tools-man-pages-Add-format-wildcard-descri.patch
* Fixes RHEL-8304
- Resolves: RHEL-8304
- Resolves: RHEL-30272
* Tue Apr 9 2024 Benjamin Marzinski <bmarzins@redhat.com> - 0.8.7-28
- Add 0108-libmultipath-actually-truncate-too-large-vpd-page.patch
- Add 0109-kpartx-fix-theoretical-overflow-in-loop-device-name.patch