From: Mikhail Pershin <mpershin@whamcloud.com>
Date: Fri, 14 Mar 2025 14:11:32 +0000 (+0300)
Subject: LU-18815 mgc: don't fail/lbug on many NIDs
X-Git-Url: https://git.whamcloud.com/?a=commitdiff_plain;h=0922d5907365c3e999a8b2c9f236f59c3b75e8a8;p=fs%2Flustre-release.git

LU-18815 mgc: don't fail/lbug on many NIDs

Keep server starting on node with more that 32 NIDs,
allowing first 32 NIDs per target.
Account '-o network' mount option to don't use other
networks as server import peers

Signed-off-by: Mikhail Pershin <mpershin@whamcloud.com>
Change-Id: If4c997be3480eba8b75888a070fb5a721b71b894
Reviewed-on: https://review.whamcloud.com/c/fs/lustre-release/+/58502
Reviewed-by: Sebastien Buisson <sbuisson@ddn.com>
Reviewed-by: Marc Vef <mvef@whamcloud.com>
Reviewed-by: James Simmons <jsimmons@infradead.org>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
---

diff --git a/lustre/obdclass/lustre_peer.c b/lustre/obdclass/lustre_peer.c
index b1664af..f6d216d 100644
--- a/lustre/obdclass/lustre_peer.c
+++ b/lustre/obdclass/lustre_peer.c
@@ -61,7 +61,7 @@ int class_add_uuid(const char *uuid, struct lnet_nid *nid)
 {
 	struct uuid_nid_data *data, *entry;
 	int found = 0;
-	int rc;
+	int rc = 0;
 
 	LASSERT(nid->nid_type != 0);  /* valid newconfig NID is never zero */
 
@@ -87,7 +87,10 @@ int class_add_uuid(const char *uuid, struct lnet_nid *nid)
 					break;
 
 			if (i == entry->un_nid_count) {
-				LASSERT(entry->un_nid_count < MTI_NIDS_MAX);
+				if (i == MTI_NIDS_MAX) {
+					rc = -EOVERFLOW;
+					break;
+				}
 				entry->un_nids[entry->un_nid_count++] = *nid;
 			}
 			break;
@@ -97,6 +100,12 @@ int class_add_uuid(const char *uuid, struct lnet_nid *nid)
 		list_add(&data->un_list, &g_uuid_list);
 	spin_unlock(&g_uuid_lock);
 
+	if (rc) {
+		CWARN("%s: can't add NID %s: rc = %d\n", uuid,
+		       libcfs_nidstr(nid), rc);
+		/* continue with already added NIDs */
+	}
+
 	if (found) {
 		CDEBUG(D_INFO, "found uuid %s %s cnt=%d\n", uuid,
 		       libcfs_nidstr(nid), entry->un_nid_count);
diff --git a/lustre/obdclass/obd_mount.c b/lustre/obdclass/obd_mount.c
index 35030fc..15995fc 100644
--- a/lustre/obdclass/obd_mount.c
+++ b/lustre/obdclass/obd_mount.c
@@ -402,14 +402,19 @@ int lustre_start_mgc(struct super_block *sb)
 	/* Add the primary NIDs for the MGS */
 	i = 0;
 	if (IS_SERVER(lsi)) {
+		char *nidnet = lsi->lsi_lmd->lmd_nidnet;
+
 		/* All mgsnode are listed in lmd_mgs at this moment */
 		ptr = lsi->lsi_lmd->lmd_mgs;
 		CDEBUG(D_MOUNT, "mgs NIDs %s.\n", ptr);
 		if (IS_MGS(lsi)) {
-			/* Use local NIDs (including LO) */
 			struct lnet_processid id;
 
+			/* Use local NIDs (including LO) */
 			while ((rc = LNetGetId(i++, &id, true)) != -ENOENT) {
+				if (nidnet && libcfs_str2net(nidnet) !=
+					      LNET_NID_NET(&id.nid))
+					continue;
 				rc = do_lcfg_nid(mgcname, &id.nid,
 						LCFG_ADD_UUID, nidstr);
 			}
@@ -425,6 +430,10 @@ int lustre_start_mgc(struct super_block *sb)
 			 * by commas.
 			 */
 			while (class_parse_nid(ptr, &nid, &ptr) == 0) {
+				if (nidnet && libcfs_str2net(nidnet) !=
+					      LNET_NID_NET(&nid))
+					continue;
+
 				rc = do_lcfg_nid(mgcname, &nid,
 						 LCFG_ADD_UUID, nidstr);
 				if (rc == 0)
diff --git a/lustre/tests/conf-sanity.sh b/lustre/tests/conf-sanity.sh
index 04a6fb2..191e8db 100755
--- a/lustre/tests/conf-sanity.sh
+++ b/lustre/tests/conf-sanity.sh
@@ -6575,6 +6575,36 @@ test_73b() {
 }
 run_test 73b "Large failnode NID list in mountdata"
 
+cleanup_73c() {
+	LOAD_MODULES_REMOTE=true cleanup
+}
+
+test_73c() {
+	(( $OST1_VERSION >= $(version_code 2.16.54) )) ||
+		skip "Need OST version at least 2.16.54 to don't LBUG"
+
+	cleanup
+	LOAD_MODULES_REMOTE=true load_modules
+
+	INTERFACES=( $(lnet_if_list) )
+	local inf=${INTERFACES[0]}
+
+	do_facet ost1 "$LNETCTL lnet configure" ||
+		error "unable to configure lnet on ost1"
+
+	stack_trap "cleanup_73c"
+
+	for ((n = 100; n <= 135; n++)); do
+		do_facet ost1 "$LNETCTL net add --net ${NETTYPE}$n --if $inf" ||
+			skip "unable to configure net #$n on ost1"
+	done
+
+	echo "restart with 35 nets"
+	start_mgsmds
+	start_ost || error "unable to start ost1"
+}
+run_test 73c "Server mount doesn't fail with > 32 nets"
+
 test_73d() { #LU-18896
 	(( $OST1_VERSION >= $(version_code 2.16.53) )) ||
 		skip "need OST >= 2.16.53 for LU-18896 fix"