Blame SOURCES/0066-multipathd-add-recheck_wwid-option-to-verify-the-pat.patch

50e627
From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
50e627
From: Benjamin Marzinski <bmarzins@redhat.com>
50e627
Date: Wed, 24 Feb 2021 00:33:22 -0600
50e627
Subject: [PATCH] multipathd: add recheck_wwid option to verify the path wwid
50e627
50e627
There are cases where the wwid of a path changes due to LUN remapping
50e627
without triggering uevent for the changed path. Multipathd has no method
50e627
for trying to catch these cases, and corruption has resulted because of
50e627
it.
50e627
50e627
In order to have a better chance at catching these cases, multipath now
50e627
has a recheck_wwid option. If this is set to "yes", when a failed path
50e627
has become active again, multipathd will recheck its wwid. If multipathd
50e627
notices that a path's wwid has changed, it will remove and re-add the
50e627
path, just like the existing wwid checking code for change events does.
50e627
In cases where the no uevent occurs, both the udev database entry and
50e627
sysfs will have the old wwid, so the only way to get a current wwid is
50e627
to ask the device directly. Currently multipath only has code to
50e627
directly get the wwid for scsi devices, so this option only effects scsi
50e627
devices, and they must be configured to be able to use the uid_fallback
50e627
methods. To make sure both the sysfs and udev database values are
50e627
updated, multipathd triggers a both a rescan of the device and a udev
50e627
add event.
50e627
50e627
Co-developed-by: Chongyun Wu <wucy11@chinatelecom.cn>
50e627
Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
50e627
Reviewed-by: Martin Wilck <mwilck@suse.com>Reviewed-by: Martin Wilck <mwilck@suse.com>
50e627
---
50e627
 libmultipath/config.c      |  2 +
50e627
 libmultipath/config.h      |  2 +
50e627
 libmultipath/configure.c   |  4 +-
50e627
 libmultipath/configure.h   |  2 +
50e627
 libmultipath/defaults.h    |  1 +
50e627
 libmultipath/dict.c        | 11 ++++++
50e627
 libmultipath/discovery.c   |  3 +-
50e627
 libmultipath/discovery.h   |  1 +
50e627
 libmultipath/propsel.c     | 21 ++++++++++
50e627
 libmultipath/propsel.h     |  1 +
50e627
 libmultipath/structs.h     |  7 ++++
50e627
 multipath/multipath.conf.5 | 14 +++++++
50e627
 multipathd/cli_handlers.c  |  9 +++++
50e627
 multipathd/main.c          | 78 ++++++++++++++++++++++++++++++++++++++
50e627
 multipathd/main.h          |  2 +
50e627
 15 files changed, 155 insertions(+), 3 deletions(-)
50e627
50e627
diff --git a/libmultipath/config.c b/libmultipath/config.c
50e627
index dd645f17..abbddaf1 100644
50e627
--- a/libmultipath/config.c
50e627
+++ b/libmultipath/config.c
50e627
@@ -371,6 +371,7 @@ merge_hwe (struct hwentry * dst, struct hwentry * src)
50e627
 	merge_num(max_sectors_kb);
50e627
 	merge_num(ghost_delay);
50e627
 	merge_num(all_tg_pt);
50e627
+	merge_num(recheck_wwid);
50e627
 	merge_num(vpd_vendor_id);
50e627
 	merge_num(san_path_err_threshold);
50e627
 	merge_num(san_path_err_forget_rate);
50e627
@@ -762,6 +763,7 @@ load_config (char * file)
50e627
 	conf->remove_retries = 0;
50e627
 	conf->ghost_delay = DEFAULT_GHOST_DELAY;
50e627
 	conf->all_tg_pt = DEFAULT_ALL_TG_PT;
50e627
+	conf->recheck_wwid = DEFAULT_RECHECK_WWID;
50e627
 	/*
50e627
 	 * preload default hwtable
50e627
 	 */
50e627
diff --git a/libmultipath/config.h b/libmultipath/config.h
50e627
index a22c1b4e..e2e3f143 100644
50e627
--- a/libmultipath/config.h
50e627
+++ b/libmultipath/config.h
50e627
@@ -84,6 +84,7 @@ struct hwentry {
50e627
 	int ghost_delay;
50e627
 	int all_tg_pt;
50e627
 	int vpd_vendor_id;
50e627
+	int recheck_wwid;
50e627
 	char * bl_product;
50e627
 };
50e627
 
50e627
@@ -188,6 +189,7 @@ struct config {
50e627
 	int skip_delegate;
50e627
 	unsigned int version[3];
50e627
 	unsigned int sequence_nr;
50e627
+	int recheck_wwid;
50e627
 
50e627
 	char * multipath_dir;
50e627
 	char * selector;
50e627
diff --git a/libmultipath/configure.c b/libmultipath/configure.c
50e627
index a6893d8d..f24d9283 100644
50e627
--- a/libmultipath/configure.c
50e627
+++ b/libmultipath/configure.c
50e627
@@ -506,8 +506,8 @@ trigger_udev_change(const struct multipath *mpp)
50e627
 	udev_device_unref(udd);
50e627
 }
50e627
 
50e627
-static void trigger_partitions_udev_change(struct udev_device *dev,
50e627
-					   const char *action, int len)
50e627
+void trigger_partitions_udev_change(struct udev_device *dev,
50e627
+				    const char *action, int len)
50e627
 {
50e627
 	struct udev_enumerate *part_enum;
50e627
 	struct udev_list_entry *item;
50e627
diff --git a/libmultipath/configure.h b/libmultipath/configure.h
50e627
index 0e33bf40..81090dd4 100644
50e627
--- a/libmultipath/configure.h
50e627
+++ b/libmultipath/configure.h
50e627
@@ -57,3 +57,5 @@ int get_refwwid (enum mpath_cmds cmd, char * dev, enum devtypes dev_type,
50e627
 int reload_map(struct vectors *vecs, struct multipath *mpp, int refresh, int is_daemon);
50e627
 struct udev_device *get_udev_device(const char *dev, enum devtypes dev_type);
50e627
 void trigger_paths_udev_change(struct multipath *mpp, bool is_mpath);
50e627
+void trigger_partitions_udev_change(struct udev_device *dev, const char *action,
50e627
+				    int len);
50e627
diff --git a/libmultipath/defaults.h b/libmultipath/defaults.h
50e627
index 52fe05b9..f1cb000d 100644
50e627
--- a/libmultipath/defaults.h
50e627
+++ b/libmultipath/defaults.h
50e627
@@ -50,6 +50,7 @@
50e627
 #define DEFAULT_FIND_MULTIPATHS_TIMEOUT -10
50e627
 #define DEFAULT_UNKNOWN_FIND_MULTIPATHS_TIMEOUT 1
50e627
 #define DEFAULT_ALL_TG_PT ALL_TG_PT_OFF
50e627
+#define DEFAULT_RECHECK_WWID RECHECK_WWID_OFF
50e627
 /* Enable all foreign libraries by default */
50e627
 #define DEFAULT_ENABLE_FOREIGN ""
50e627
 
50e627
diff --git a/libmultipath/dict.c b/libmultipath/dict.c
50e627
index 8fd91d8c..13698b76 100644
50e627
--- a/libmultipath/dict.c
50e627
+++ b/libmultipath/dict.c
50e627
@@ -1413,6 +1413,14 @@ declare_hw_snprint(all_tg_pt, print_yes_no_undef)
50e627
 declare_def_handler(marginal_pathgroups, set_yes_no)
50e627
 declare_def_snprint(marginal_pathgroups, print_yes_no)
50e627
 
50e627
+declare_def_handler(recheck_wwid, set_yes_no_undef)
50e627
+declare_def_snprint_defint(recheck_wwid, print_yes_no_undef, DEFAULT_RECHECK_WWID)
50e627
+declare_ovr_handler(recheck_wwid, set_yes_no_undef)
50e627
+declare_ovr_snprint(recheck_wwid, print_yes_no_undef)
50e627
+declare_hw_handler(recheck_wwid, set_yes_no_undef)
50e627
+declare_hw_snprint(recheck_wwid, print_yes_no_undef)
50e627
+
50e627
+
50e627
 static int
50e627
 def_uxsock_timeout_handler(struct config *conf, vector strvec)
50e627
 {
50e627
@@ -1824,6 +1832,7 @@ init_keywords(vector keywords)
50e627
 	install_keyword("enable_foreign", &def_enable_foreign_handler,
50e627
 			&snprint_def_enable_foreign);
50e627
 	install_keyword("marginal_pathgroups", &def_marginal_pathgroups_handler, &snprint_def_marginal_pathgroups);
50e627
+	install_keyword("recheck_wwid", &def_recheck_wwid_handler, &snprint_def_recheck_wwid);
50e627
 	__deprecated install_keyword("default_selector", &def_selector_handler, NULL);
50e627
 	__deprecated install_keyword("default_path_grouping_policy", &def_pgpolicy_handler, NULL);
50e627
 	__deprecated install_keyword("default_uid_attribute", &def_uid_attribute_handler, NULL);
50e627
@@ -1913,6 +1922,7 @@ init_keywords(vector keywords)
50e627
 	install_keyword("ghost_delay", &hw_ghost_delay_handler, &snprint_hw_ghost_delay);
50e627
 	install_keyword("all_tg_pt", &hw_all_tg_pt_handler, &snprint_hw_all_tg_pt);
50e627
 	install_keyword("vpd_vendor", &hw_vpd_vendor_handler, &snprint_hw_vpd_vendor);
50e627
+	install_keyword("recheck_wwid", &hw_recheck_wwid_handler, &snprint_hw_recheck_wwid);
50e627
 	install_sublevel_end();
50e627
 
50e627
 	install_keyword_root("overrides", &overrides_handler);
50e627
@@ -1954,6 +1964,7 @@ init_keywords(vector keywords)
50e627
 	install_keyword("max_sectors_kb", &ovr_max_sectors_kb_handler, &snprint_ovr_max_sectors_kb);
50e627
 	install_keyword("ghost_delay", &ovr_ghost_delay_handler, &snprint_ovr_ghost_delay);
50e627
 	install_keyword("all_tg_pt", &ovr_all_tg_pt_handler, &snprint_ovr_all_tg_pt);
50e627
+	install_keyword("recheck_wwid", &ovr_recheck_wwid_handler, &snprint_ovr_recheck_wwid);
50e627
 
50e627
 	install_keyword_root("multipaths", &multipaths_handler);
50e627
 	install_keyword_multi("multipath", &multipath_handler, NULL);
50e627
diff --git a/libmultipath/discovery.c b/libmultipath/discovery.c
50e627
index 126a70b3..bc267609 100644
50e627
--- a/libmultipath/discovery.c
50e627
+++ b/libmultipath/discovery.c
50e627
@@ -2040,7 +2040,7 @@ static ssize_t uid_fallback(struct path *pp, int path_state,
50e627
 	return len;
50e627
 }
50e627
 
50e627
-static bool has_uid_fallback(struct path *pp)
50e627
+bool has_uid_fallback(struct path *pp)
50e627
 {
50e627
 	/*
50e627
 	 * Falling back to direct WWID determination is dangerous
50e627
@@ -2072,6 +2072,7 @@ get_uid (struct path * pp, int path_state, struct udev_device *udev,
50e627
 		conf = get_multipath_config();
50e627
 		pthread_cleanup_push(put_multipath_config, conf);
50e627
 		select_getuid(conf, pp);
50e627
+		select_recheck_wwid(conf, pp);
50e627
 		pthread_cleanup_pop(1);
50e627
 	}
50e627
 
50e627
diff --git a/libmultipath/discovery.h b/libmultipath/discovery.h
50e627
index d3193daf..a5446b4d 100644
50e627
--- a/libmultipath/discovery.h
50e627
+++ b/libmultipath/discovery.h
50e627
@@ -54,6 +54,7 @@ ssize_t sysfs_get_inquiry(struct udev_device *udev,
50e627
 			  unsigned char *buff, size_t len);
50e627
 int sysfs_get_asymmetric_access_state(struct path *pp,
50e627
 				      char *buff, int buflen);
50e627
+bool has_uid_fallback(struct path *pp);
50e627
 int get_uid(struct path * pp, int path_state, struct udev_device *udev,
50e627
 	    int allow_fallback);
50e627
 bool is_vpd_page_supported(int fd, int pg);
50e627
diff --git a/libmultipath/propsel.c b/libmultipath/propsel.c
50e627
index 1150cfe8..127b3370 100644
50e627
--- a/libmultipath/propsel.c
50e627
+++ b/libmultipath/propsel.c
50e627
@@ -581,6 +581,27 @@ out:
50e627
 	return 0;
50e627
 }
50e627
 
50e627
+/* must be called after select_getuid */
50e627
+int select_recheck_wwid(struct config *conf, struct path * pp)
50e627
+{
50e627
+	const char *origin;
50e627
+
50e627
+	pp_set_ovr(recheck_wwid);
50e627
+	pp_set_hwe(recheck_wwid);
50e627
+	pp_set_conf(recheck_wwid);
50e627
+	pp_set_default(recheck_wwid, DEFAULT_RECHECK_WWID);
50e627
+out:
50e627
+	if (pp->recheck_wwid == RECHECK_WWID_ON &&
50e627
+	    (pp->bus != SYSFS_BUS_SCSI || pp->getuid != NULL ||
50e627
+	     !has_uid_fallback(pp))) {
50e627
+		pp->recheck_wwid = RECHECK_WWID_OFF;
50e627
+		origin = "(setting: unsupported by device type/config)";
50e627
+	}
50e627
+	condlog(3, "%s: recheck_wwid = %i %s", pp->dev, pp->recheck_wwid,
50e627
+		origin);
50e627
+	return 0;
50e627
+}
50e627
+
50e627
 void
50e627
 detect_prio(struct config *conf, struct path * pp)
50e627
 {
50e627
diff --git a/libmultipath/propsel.h b/libmultipath/propsel.h
50e627
index a68bacf0..72a7e33c 100644
50e627
--- a/libmultipath/propsel.h
50e627
+++ b/libmultipath/propsel.h
50e627
@@ -7,6 +7,7 @@ int select_features (struct config *conf, struct multipath * mp);
50e627
 int select_hwhandler (struct config *conf, struct multipath * mp);
50e627
 int select_checker(struct config *conf, struct path *pp);
50e627
 int select_getuid (struct config *conf, struct path * pp);
50e627
+int select_recheck_wwid(struct config *conf, struct path * pp);
50e627
 int select_prio (struct config *conf, struct path * pp);
50e627
 int select_find_multipaths_timeout(struct config *conf, struct path *pp);
50e627
 int select_no_path_retry(struct config *conf, struct multipath *mp);
50e627
diff --git a/libmultipath/structs.h b/libmultipath/structs.h
50e627
index 65542dea..a5dbad5b 100644
50e627
--- a/libmultipath/structs.h
50e627
+++ b/libmultipath/structs.h
50e627
@@ -252,6 +252,12 @@ enum eh_deadline_states {
50e627
 	EH_DEADLINE_ZERO = UOZ_ZERO,
50e627
 };
50e627
 
50e627
+enum recheck_wwid_states {
50e627
+	RECHECK_WWID_UNDEF = YNU_UNDEF,
50e627
+	RECHECK_WWID_OFF = YNU_NO,
50e627
+	RECHECK_WWID_ON = YNU_YES,
50e627
+};
50e627
+
50e627
 struct vpd_vendor_page {
50e627
 	int pg;
50e627
 	const char *name;
50e627
@@ -326,6 +332,7 @@ struct path {
50e627
 	int find_multipaths_timeout;
50e627
 	int marginal;
50e627
 	int vpd_vendor_id;
50e627
+	int recheck_wwid;
50e627
 	/* configlet pointers */
50e627
 	vector hwe;
50e627
 	struct gen_path generic_path;
50e627
diff --git a/multipath/multipath.conf.5 b/multipath/multipath.conf.5
50e627
index a5686090..6da15aad 100644
50e627
--- a/multipath/multipath.conf.5
50e627
+++ b/multipath/multipath.conf.5
50e627
@@ -1256,6 +1256,20 @@ The default is: \fB\(dq\(dq\fR (the empty regular expression)
50e627
 .RE
50e627
 .
50e627
 .
50e627
+.TP
50e627
+.B recheck_wwid
50e627
+If set to \fIyes\fR, when a failed path is restored, its wwid is rechecked. If
50e627
+the wwid has changed, the path is removed from the current multipath device,
50e627
+and re-added as a new path. Multipathd will also recheck a path's wwid if it is
50e627
+manually re-added. This option only works for SCSI devices that are configured
50e627
+to use the default uid_attribute, \fIID_SERIAL\fR, or sysfs for getting their
50e627
+wwid.
50e627
+.RS
50e627
+.TP
50e627
+The default is \fBno\fR
50e627
+.RE
50e627
+.
50e627
+.
50e627
 
50e627
 .
50e627
 .\" ----------------------------------------------------------------------------
50e627
diff --git a/multipathd/cli_handlers.c b/multipathd/cli_handlers.c
50e627
index 782bb003..8b4bd187 100644
50e627
--- a/multipathd/cli_handlers.c
50e627
+++ b/multipathd/cli_handlers.c
50e627
@@ -715,6 +715,15 @@ cli_add_path (void * v, char ** reply, int * len, void * data)
50e627
 	pp = find_path_by_dev(vecs->pathvec, param);
50e627
 	if (pp) {
50e627
 		condlog(2, "%s: path already in pathvec", param);
50e627
+
50e627
+		if (pp->recheck_wwid == RECHECK_WWID_ON &&
50e627
+		    check_path_wwid_change(pp)) {
50e627
+			condlog(0, "%s: wwid changed. Removing device",
50e627
+				pp->dev);
50e627
+			handle_path_wwid_change(pp, vecs);
50e627
+			return 1;
50e627
+		}
50e627
+
50e627
 		if (pp->mpp)
50e627
 			return 0;
50e627
 	} else {
50e627
diff --git a/multipathd/main.c b/multipathd/main.c
50e627
index cc1aeea2..1fbc31eb 100644
50e627
--- a/multipathd/main.c
50e627
+++ b/multipathd/main.c
50e627
@@ -822,6 +822,73 @@ ev_remove_map (char * devname, char * alias, int minor, struct vectors * vecs)
50e627
 	return flush_map(mpp, vecs, 0);
50e627
 }
50e627
 
50e627
+static void
50e627
+rescan_path(struct udev_device *parent)
50e627
+{
50e627
+	while(parent) {
50e627
+		const char *subsys = udev_device_get_subsystem(parent);
50e627
+		if (subsys && !strncmp(subsys, "scsi", 4))
50e627
+			break;
50e627
+		parent = udev_device_get_parent(parent);
50e627
+	}
50e627
+	if (parent)
50e627
+		sysfs_attr_set_value(parent, "rescan", "1", strlen("1"));
50e627
+}
50e627
+
50e627
+void
50e627
+handle_path_wwid_change(struct path *pp, struct vectors *vecs)
50e627
+{
50e627
+	struct udev_device *udd;
50e627
+
50e627
+	if (!pp || !pp->udev)
50e627
+		return;
50e627
+
50e627
+	udd = udev_device_ref(pp->udev);
50e627
+	if (ev_remove_path(pp, vecs, 1) != 0 && pp->mpp) {
50e627
+		pp->dmstate = PSTATE_FAILED;
50e627
+		dm_fail_path(pp->mpp->alias, pp->dev_t);
50e627
+	}
50e627
+	rescan_path(udd);
50e627
+	sysfs_attr_set_value(udd, "uevent", "add", strlen("add"));
50e627
+	trigger_partitions_udev_change(udd, "add", strlen("add"));
50e627
+	udev_device_unref(udd);
50e627
+}
50e627
+
50e627
+bool
50e627
+check_path_wwid_change(struct path *pp)
50e627
+{
50e627
+	char wwid[WWID_SIZE];
50e627
+	int len = 0;
50e627
+	size_t i;
50e627
+
50e627
+	if (!strlen(pp->wwid))
50e627
+		return false;
50e627
+
50e627
+	/* Get the real fresh device wwid by sgio. sysfs still has old
50e627
+	 * data, so only get_vpd_sgio will work to get the new wwid */
50e627
+	len = get_vpd_sgio(pp->fd, 0x83, 0, wwid, WWID_SIZE);
50e627
+
50e627
+	if (len <= 0) {
50e627
+		condlog(2, "%s: failed to check wwid by sgio: len = %d",
50e627
+			pp->dev, len);
50e627
+		return false;
50e627
+	}
50e627
+
50e627
+	/*Strip any trailing blanks */
50e627
+	for (i = strlen(pp->wwid); i > 0 && pp->wwid[i-1] == ' '; i--);
50e627
+		/* no-op */
50e627
+	pp->wwid[i] = '\0';
50e627
+	condlog(4, "%s: Got wwid %s by sgio", pp->dev, wwid);
50e627
+
50e627
+	if (strncmp(wwid, pp->wwid, WWID_SIZE)) {
50e627
+		condlog(0, "%s: wwid '%s' doesn't match wwid '%s' from device",
50e627
+			pp->dev, pp->wwid, wwid);
50e627
+		return true;
50e627
+	}
50e627
+
50e627
+	return false;
50e627
+}
50e627
+
50e627
 static int
50e627
 uev_add_path (struct uevent *uev, struct vectors * vecs, int need_do_map)
50e627
 {
50e627
@@ -1241,6 +1308,7 @@ uev_update_path (struct uevent *uev, struct vectors * vecs)
50e627
 			condlog(0, "%s: path wwid changed from '%s' to '%s'",
50e627
 				uev->kernel, wwid, pp->wwid);
50e627
 			ev_remove_path(pp, vecs, 1);
50e627
+			rescan_path(uev->udev);
50e627
 			needs_reinit = 1;
50e627
 			goto out;
50e627
 		} else {
50e627
@@ -2101,6 +2169,16 @@ check_path (struct vectors * vecs, struct path * pp, unsigned int ticks)
50e627
 		return 0;
50e627
 	set_no_path_retry(pp->mpp);
50e627
 
50e627
+	if (pp->recheck_wwid == RECHECK_WWID_ON &&
50e627
+	    (newstate == PATH_UP || newstate == PATH_GHOST) &&
50e627
+	    ((pp->state != PATH_UP && pp->state != PATH_GHOST) ||
50e627
+	     pp->dmstate == PSTATE_FAILED) &&
50e627
+	    check_path_wwid_change(pp)) {
50e627
+		condlog(0, "%s: path wwid change detected. Removing", pp->dev);
50e627
+		handle_path_wwid_change(pp, vecs);
50e627
+		return 0;
50e627
+	}
50e627
+
50e627
 	if ((newstate == PATH_UP || newstate == PATH_GHOST) &&
50e627
 	    (san_path_check_enabled(pp->mpp) ||
50e627
 	     marginal_path_check_enabled(pp->mpp))) {
50e627
diff --git a/multipathd/main.h b/multipathd/main.h
50e627
index 5dff17e5..8f0028a9 100644
50e627
--- a/multipathd/main.h
50e627
+++ b/multipathd/main.h
50e627
@@ -49,4 +49,6 @@ int update_multipath (struct vectors *vecs, char *mapname, int reset);
50e627
 int update_path_groups(struct multipath *mpp, struct vectors *vecs,
50e627
 		       int refresh);
50e627
 
50e627
+void handle_path_wwid_change(struct path *pp, struct vectors *vecs);
50e627
+bool check_path_wwid_change(struct path *pp);
50e627
 #endif /* MAIN_H */
50e627
-- 
50e627
2.17.2
50e627