From a710cb740cdcd633f09ab0ede9aa34902059badb Mon Sep 17 00:00:00 2001
From: David Teigland <teigland@redhat.com>
Date: Wed, 7 May 2025 17:51:01 -0500
Subject: [PATCH 2/8] lvmlockd: fix hosts check for vgremove

errors from lock manager were not being considered.
EAGAIN from sanlock should be considered EBUSY.

(cherry picked from commit 53752ef851d3210e52297ebb4744fdd766c060c6)
---
 daemons/lvmlockd/lvmlockd-core.c    | 11 ++++++++---
 daemons/lvmlockd/lvmlockd-dlm.c     |  8 ++++++++
 daemons/lvmlockd/lvmlockd-sanlock.c | 26 ++++++++++++++++++++++----
 3 files changed, 38 insertions(+), 7 deletions(-)

diff --git a/daemons/lvmlockd/lvmlockd-core.c b/daemons/lvmlockd/lvmlockd-core.c
index c65a3cc39..e4aa91216 100644
--- a/daemons/lvmlockd/lvmlockd-core.c
+++ b/daemons/lvmlockd/lvmlockd-core.c
@@ -2708,13 +2708,16 @@ static void *lockspace_thread_main(void *arg_in)
 				rv = lm_hosts(ls, 1);
 				if (rv) {
 					/*
+					 * rv < 0: error (don't remove)
+					 * rv > 0: other hosts in lockspace (cannot remove)
+					 * rv = 0: only local host in lockspace (can remove)
 					 * Checking for hosts here in addition to after the
 					 * main loop allows vgremove to fail and be rerun
 					 * after the ls is stopped on other hosts.
 					 */
 					log_error("S %s lockspace hosts %d", ls->name, rv);
 					list_del(&act->list);
-					act->result = -EBUSY;
+					act->result = (rv < 0) ? rv : -EBUSY;
 					add_client_result(act);
 					continue;
 				}
@@ -2727,7 +2730,9 @@ static void *lockspace_thread_main(void *arg_in)
 			if (act->op == LD_OP_BUSY && act->rt == LD_RT_VG) {
 				log_debug("S %s checking if lockspace is busy", ls->name);
 				rv = lm_hosts(ls, 0);
-				if (rv)
+				if (rv < 0)
+					act->result = rv;
+				else if (rv)
 					act->result = -EBUSY;
 				else
 					act->result = 0;
@@ -2743,7 +2748,7 @@ static void *lockspace_thread_main(void *arg_in)
 				if (rv) {
 					log_error("S %s lockspace hosts %d", ls->name, rv);
 					list_del(&act->list);
-					act->result = -EBUSY;
+					act->result = (rv < 0) ? rv : -EBUSY;
 					add_client_result(act);
 					continue;
 				}
diff --git a/daemons/lvmlockd/lvmlockd-dlm.c b/daemons/lvmlockd/lvmlockd-dlm.c
index 72b139170..7529ad327 100644
--- a/daemons/lvmlockd/lvmlockd-dlm.c
+++ b/daemons/lvmlockd/lvmlockd-dlm.c
@@ -799,6 +799,14 @@ int lm_unlock_dlm(struct lockspace *ls, struct resource *r,
  * the stale lockspaces on the others eventually.)
  */
 
+/*
+ * On error, returns < 0
+ *
+ * On success:
+ * If other hosts are found, returns the number.
+ * If no other hosts are found (only ourself), returns 0.
+ */
+
 int lm_hosts_dlm(struct lockspace *ls, int notify)
 {
 	char ls_nodes_path[PATH_MAX];
diff --git a/daemons/lvmlockd/lvmlockd-sanlock.c b/daemons/lvmlockd/lvmlockd-sanlock.c
index 1a3982071..d50d0ce4b 100644
--- a/daemons/lvmlockd/lvmlockd-sanlock.c
+++ b/daemons/lvmlockd/lvmlockd-sanlock.c
@@ -2296,6 +2296,13 @@ int lm_unlock_sanlock(struct lockspace *ls, struct resource *r,
 	return 0;
 }
 
+/*
+ * On error, returns < 0
+ * Else:
+ * If other hosts are found, returns the number.
+ * If no other hosts are found (only ourself), returns 0.
+ */
+
 int lm_hosts_sanlock(struct lockspace *ls, int notify)
 {
 	struct sanlk_host *hss = NULL;
@@ -2310,14 +2317,25 @@ int lm_hosts_sanlock(struct lockspace *ls, int notify)
 		return 0;
 
 	rv = sanlock_get_hosts(ls->name, 0, &hss, &hss_count, 0);
+
+	if (rv == -EAGAIN) {
+		/*
+		 * No host info is available yet (perhaps lockspace was
+		 * just started so other host state is unknown.)  Pretend
+		 * there is one other host (busy).
+		 */
+		log_debug("S %s hosts_san no info, retry later", ls->name);
+		return 1;
+	}
+
 	if (rv < 0) {
 		log_error("S %s hosts_san get_hosts error %d", ls->name, rv);
-		return 0;
+		return -1;
 	}
 
 	if (!hss || !hss_count) {
 		log_error("S %s hosts_san zero hosts", ls->name);
-		return 0;
+		return -1;
 	}
 
 	hs = hss;
@@ -2336,7 +2354,7 @@ int lm_hosts_sanlock(struct lockspace *ls, int notify)
 		}
 
 		state = hs->flags & SANLK_HOST_MASK;
-		if (state == SANLK_HOST_LIVE)
+		if ((state == SANLK_HOST_LIVE) || (state == SANLK_HOST_UNKNOWN))
 			found_others++;
 		hs++;
 	}
@@ -2358,7 +2376,7 @@ int lm_hosts_sanlock(struct lockspace *ls, int notify)
 
 	if (!found_self) {
 		log_error("S %s hosts_san self not found others %d", ls->name, found_others);
-		return 0;
+		return -1;
 	}
 
 	return found_others;
-- 
2.49.0