From 242d23efc987151ecd34bc0cae4c0b737494fc40 Mon Sep 17 00:00:00 2001
From: Paulo Alcantara <pc@manguebit.com>
Date: Wed, 18 Sep 2024 02:03:35 -0300
Subject: [PATCH 01/13] smb: client: avoid unnecessary reconnects when
 refreshing referrals

Do not mark tcons for reconnect when current connection matches any of
the targets returned by new referral even when there is no cached
entry.

Signed-off-by: Paulo Alcantara (Red Hat) <pc@manguebit.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
---
 fs/smb/client/dfs_cache.c | 187 ++++++++++++++++++++++++--------------
 1 file changed, 117 insertions(+), 70 deletions(-)

diff --git a/fs/smb/client/dfs_cache.c b/fs/smb/client/dfs_cache.c
index 11c8efecf7aa12..3cf7c88489be45 100644
--- a/fs/smb/client/dfs_cache.c
+++ b/fs/smb/client/dfs_cache.c
@@ -1095,16 +1095,18 @@ int dfs_cache_get_tgt_share(char *path, const struct dfs_cache_tgt_iterator *it,
 	return 0;
 }
 
-static bool target_share_equal(struct TCP_Server_Info *server, const char *s1, const char *s2)
+static bool target_share_equal(struct cifs_tcon *tcon, const char *s1)
 {
-	char unc[sizeof("\\\\") + SERVER_NAME_LENGTH] = {0};
+	struct TCP_Server_Info *server = tcon->ses->server;
+	struct sockaddr_storage ss;
 	const char *host;
+	const char *s2 = &tcon->tree_name[1];
 	size_t hostlen;
-	struct sockaddr_storage ss;
+	char unc[sizeof("\\\\") + SERVER_NAME_LENGTH] = {0};
 	bool match;
 	int rc;
 
-	if (strcasecmp(s1, s2))
+	if (strcasecmp(s2, s1))
 		return false;
 
 	/*
@@ -1128,34 +1130,6 @@ static bool target_share_equal(struct TCP_Server_Info *server, const char *s1, c
 	return match;
 }
 
-/*
- * Mark dfs tcon for reconnecting when the currently connected tcon does not match any of the new
- * target shares in @refs.
- */
-static void mark_for_reconnect_if_needed(struct TCP_Server_Info *server,
-					 const char *path,
-					 struct dfs_cache_tgt_list *old_tl,
-					 struct dfs_cache_tgt_list *new_tl)
-{
-	struct dfs_cache_tgt_iterator *oit, *nit;
-
-	for (oit = dfs_cache_get_tgt_iterator(old_tl); oit;
-	     oit = dfs_cache_get_next_tgt(old_tl, oit)) {
-		for (nit = dfs_cache_get_tgt_iterator(new_tl); nit;
-		     nit = dfs_cache_get_next_tgt(new_tl, nit)) {
-			if (target_share_equal(server,
-					       dfs_cache_get_tgt_name(oit),
-					       dfs_cache_get_tgt_name(nit))) {
-				dfs_cache_noreq_update_tgthint(path, nit);
-				return;
-			}
-		}
-	}
-
-	cifs_dbg(FYI, "%s: no cached or matched targets. mark dfs share for reconnect.\n", __func__);
-	cifs_signal_cifsd_for_reconnect(server, true);
-}
-
 static bool is_ses_good(struct cifs_ses *ses)
 {
 	struct TCP_Server_Info *server = ses->server;
@@ -1172,41 +1146,35 @@ static bool is_ses_good(struct cifs_ses *ses)
 	return ret;
 }
 
-/* Refresh dfs referral of @ses and mark it for reconnect if needed */
-static void __refresh_ses_referral(struct cifs_ses *ses, bool force_refresh)
+static char *get_ses_refpath(struct cifs_ses *ses)
 {
 	struct TCP_Server_Info *server = ses->server;
-	DFS_CACHE_TGT_LIST(old_tl);
-	DFS_CACHE_TGT_LIST(new_tl);
-	bool needs_refresh = false;
-	struct cache_entry *ce;
-	unsigned int xid;
-	char *path = NULL;
-	int rc = 0;
-
-	xid = get_xid();
+	char *path = ERR_PTR(-ENOENT);
 
 	mutex_lock(&server->refpath_lock);
 	if (server->leaf_fullpath) {
 		path = kstrdup(server->leaf_fullpath + 1, GFP_ATOMIC);
 		if (!path)
-			rc = -ENOMEM;
+			path = ERR_PTR(-ENOMEM);
 	}
 	mutex_unlock(&server->refpath_lock);
-	if (!path)
-		goto out;
+	return path;
+}
 
-	down_read(&htable_rw_lock);
-	ce = lookup_cache_entry(path);
-	needs_refresh = force_refresh || IS_ERR(ce) || cache_entry_expired(ce);
-	if (!IS_ERR(ce)) {
-		rc = get_targets(ce, &old_tl);
-		cifs_dbg(FYI, "%s: get_targets: %d\n", __func__, rc);
-	}
-	up_read(&htable_rw_lock);
+/* Refresh dfs referral of @ses */
+static void refresh_ses_referral(struct cifs_ses *ses)
+{
+	struct cache_entry *ce;
+	unsigned int xid;
+	char *path;
+	int rc = 0;
 
-	if (!needs_refresh) {
-		rc = 0;
+	xid = get_xid();
+
+	path = get_ses_refpath(ses);
+	if (IS_ERR(path)) {
+		rc = PTR_ERR(path);
+		path = NULL;
 		goto out;
 	}
 
@@ -1217,29 +1185,106 @@ static void __refresh_ses_referral(struct cifs_ses *ses, bool force_refresh)
 		goto out;
 	}
 
-	ce = cache_refresh_path(xid, ses, path, true);
-	if (!IS_ERR(ce)) {
-		rc = get_targets(ce, &new_tl);
+	ce = cache_refresh_path(xid, ses, path, false);
+	if (!IS_ERR(ce))
 		up_read(&htable_rw_lock);
-		cifs_dbg(FYI, "%s: get_targets: %d\n", __func__, rc);
-		mark_for_reconnect_if_needed(server, path, &old_tl, &new_tl);
-	}
+	else
+		rc = PTR_ERR(ce);
 
 out:
 	free_xid(xid);
-	dfs_cache_free_tgts(&old_tl);
-	dfs_cache_free_tgts(&new_tl);
 	kfree(path);
 }
 
-static inline void refresh_ses_referral(struct cifs_ses *ses)
+static int __refresh_tcon_referral(struct cifs_tcon *tcon,
+				   const char *path,
+				   struct dfs_info3_param *refs,
+				   int numrefs, bool force_refresh)
 {
-	__refresh_ses_referral(ses, false);
+	struct cache_entry *ce;
+	bool reconnect = force_refresh;
+	int rc = 0;
+	int i;
+
+	if (unlikely(!numrefs))
+		return 0;
+
+	if (force_refresh) {
+		for (i = 0; i < numrefs; i++) {
+			/* TODO: include prefix paths in the matching */
+			if (target_share_equal(tcon, refs[i].node_name)) {
+				reconnect = false;
+				break;
+			}
+		}
+	}
+
+	down_write(&htable_rw_lock);
+	ce = lookup_cache_entry(path);
+	if (!IS_ERR(ce)) {
+		if (force_refresh || cache_entry_expired(ce))
+			rc = update_cache_entry_locked(ce, refs, numrefs);
+	} else if (PTR_ERR(ce) == -ENOENT) {
+		ce = add_cache_entry_locked(refs, numrefs);
+	}
+	up_write(&htable_rw_lock);
+
+	if (IS_ERR(ce))
+		rc = PTR_ERR(ce);
+	if (reconnect) {
+		cifs_tcon_dbg(FYI, "%s: mark for reconnect\n", __func__);
+		cifs_signal_cifsd_for_reconnect(tcon->ses->server, true);
+	}
+	return rc;
 }
 
-static inline void force_refresh_ses_referral(struct cifs_ses *ses)
+static void refresh_tcon_referral(struct cifs_tcon *tcon, bool force_refresh)
 {
-	__refresh_ses_referral(ses, true);
+	struct dfs_info3_param *refs = NULL;
+	struct cache_entry *ce;
+	struct cifs_ses *ses;
+	unsigned int xid;
+	bool needs_refresh;
+	char *path;
+	int numrefs = 0;
+	int rc = 0;
+
+	xid = get_xid();
+	ses = tcon->ses;
+
+	path = get_ses_refpath(ses);
+	if (IS_ERR(path)) {
+		rc = PTR_ERR(path);
+		path = NULL;
+		goto out;
+	}
+
+	down_read(&htable_rw_lock);
+	ce = lookup_cache_entry(path);
+	needs_refresh = force_refresh || IS_ERR(ce) || cache_entry_expired(ce);
+	if (!needs_refresh) {
+		up_read(&htable_rw_lock);
+		goto out;
+	}
+	up_read(&htable_rw_lock);
+
+	ses = CIFS_DFS_ROOT_SES(ses);
+	if (!is_ses_good(ses)) {
+		cifs_dbg(FYI, "%s: skip cache refresh due to disconnected ipc\n",
+			 __func__);
+		goto out;
+	}
+
+	rc = get_dfs_referral(xid, ses, path, &refs, &numrefs);
+	if (!rc) {
+		rc = __refresh_tcon_referral(tcon, path, refs,
+					     numrefs, force_refresh);
+	}
+
+out:
+	free_xid(xid);
+	kfree(path);
+	free_dfs_info_array(refs, numrefs);
 }
 
 /**
@@ -1280,7 +1325,7 @@ int dfs_cache_remount_fs(struct cifs_sb_info *cifs_sb)
 	 */
 	cifs_sb->mnt_cifs_flags |= CIFS_MOUNT_USE_PREFIX_PATH;
 
-	force_refresh_ses_referral(tcon->ses);
+	refresh_tcon_referral(tcon, true);
 	return 0;
 }
 
@@ -1291,9 +1336,11 @@ void dfs_cache_refresh(struct work_struct *work)
 	struct cifs_ses *ses;
 
 	tcon = container_of(work, struct cifs_tcon, dfs_cache_work.work);
+	ses = tcon->ses->dfs_root_ses;
 
-	for (ses = tcon->ses; ses; ses = ses->dfs_root_ses)
+	for (; ses; ses = ses->dfs_root_ses)
 		refresh_ses_referral(ses);
+	refresh_tcon_referral(tcon, false);
 
 	queue_delayed_work(dfscache_wq, &tcon->dfs_cache_work,
 			   atomic_read(&dfs_cache_ttl) * HZ);

From 9190cc0c97aafdae06015d468c6ca3991e32a23a Mon Sep 17 00:00:00 2001
From: Paulo Alcantara <pc@manguebit.com>
Date: Wed, 18 Sep 2024 02:03:45 -0300
Subject: [PATCH 02/13] smb: client: improve purging of cached referrals

Purge cached referrals that have a single target when reaching maximum
of cache size as the client won't need them to failover.  Otherwise
remove oldest cache entry.

Signed-off-by: Paulo Alcantara (Red Hat) <pc@manguebit.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
---
 fs/smb/client/dfs_cache.c | 32 +++++++++++++++-----------------
 1 file changed, 15 insertions(+), 17 deletions(-)

diff --git a/fs/smb/client/dfs_cache.c b/fs/smb/client/dfs_cache.c
index 3cf7c88489be45..e496dbf8c6228f 100644
--- a/fs/smb/client/dfs_cache.c
+++ b/fs/smb/client/dfs_cache.c
@@ -126,6 +126,7 @@ static inline void free_tgts(struct cache_entry *ce)
 
 static inline void flush_cache_ent(struct cache_entry *ce)
 {
+	cifs_dbg(FYI, "%s: %s\n", __func__, ce->path);
 	hlist_del_init(&ce->hlist);
 	kfree(ce->path);
 	free_tgts(ce);
@@ -441,34 +442,31 @@ static struct cache_entry *alloc_cache_entry(struct dfs_info3_param *refs, int n
 	return ce;
 }
 
-static void remove_oldest_entry_locked(void)
+/* Remove all referrals that have a single target or oldest entry */
+static void purge_cache(void)
 {
 	int i;
 	struct cache_entry *ce;
-	struct cache_entry *to_del = NULL;
-
-	WARN_ON(!rwsem_is_locked(&htable_rw_lock));
+	struct cache_entry *oldest = NULL;
 
 	for (i = 0; i < CACHE_HTABLE_SIZE; i++) {
 		struct hlist_head *l = &cache_htable[i];
+		struct hlist_node *n;
 
-		hlist_for_each_entry(ce, l, hlist) {
+		hlist_for_each_entry_safe(ce, n, l, hlist) {
 			if (hlist_unhashed(&ce->hlist))
 				continue;
-			if (!to_del || timespec64_compare(&ce->etime,
-							  &to_del->etime) < 0)
-				to_del = ce;
+			if (ce->numtgts == 1)
+				flush_cache_ent(ce);
+			else if (!oldest ||
+				 timespec64_compare(&ce->etime,
+						    &oldest->etime) < 0)
+				oldest = ce;
 		}
 	}
 
-	if (!to_del) {
-		cifs_dbg(FYI, "%s: no entry to remove\n", __func__);
-		return;
-	}
-
-	cifs_dbg(FYI, "%s: removing entry\n", __func__);
-	dump_ce(to_del);
-	flush_cache_ent(to_del);
+	if (atomic_read(&cache_count) >= CACHE_MAX_ENTRIES && oldest)
+		flush_cache_ent(oldest);
 }
 
 /* Add a new DFS cache entry */
@@ -484,7 +482,7 @@ static struct cache_entry *add_cache_entry_locked(struct dfs_info3_param *refs,
 
 	if (atomic_read(&cache_count) >= CACHE_MAX_ENTRIES) {
 		cifs_dbg(FYI, "%s: reached max cache size (%d)\n", __func__, CACHE_MAX_ENTRIES);
-		remove_oldest_entry_locked();
+		purge_cache();
 	}
 
 	rc = cache_entry_hash(refs[0].path_name, strlen(refs[0].path_name), &hash);

From 4f42a8b54b5c6e36519aef3cb1f6210e54abd451 Mon Sep 17 00:00:00 2001
From: Paulo Alcantara <pc@manguebit.com>
Date: Wed, 18 Sep 2024 02:03:55 -0300
Subject: [PATCH 03/13] smb: client: fix DFS interlink failover

The DFS interlinks point to different DFS namespaces so make sure to
use the correct DFS root server to chase any DFS links under it by
storing the SMB session in dfs_ref_walk structure and then using it on
every referral walk.

Signed-off-by: Paulo Alcantara (Red Hat) <pc@manguebit.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
---
 fs/smb/client/cifsglob.h   |  3 ++
 fs/smb/client/cifsproto.h  | 12 ++-----
 fs/smb/client/connect.c    | 41 +++++++++++----------
 fs/smb/client/dfs.c        | 73 ++++++++++++++++++--------------------
 fs/smb/client/dfs.h        | 42 ++++++++++++++--------
 fs/smb/client/dfs_cache.c  |  3 +-
 fs/smb/client/fs_context.h |  1 +
 fs/smb/client/misc.c       |  3 ++
 fs/smb/client/namespace.c  |  2 +-
 9 files changed, 94 insertions(+), 86 deletions(-)

diff --git a/fs/smb/client/cifsglob.h b/fs/smb/client/cifsglob.h
index a71a988a92f909..15571cf0ba632a 100644
--- a/fs/smb/client/cifsglob.h
+++ b/fs/smb/client/cifsglob.h
@@ -821,6 +821,7 @@ struct TCP_Server_Info {
 	 * format: \\HOST\SHARE[\OPTIONAL PATH]
 	 */
 	char *leaf_fullpath;
+	bool dfs_conn:1;
 };
 
 static inline bool is_smb1(struct TCP_Server_Info *server)
@@ -1059,6 +1060,7 @@ struct cifs_ses {
 	struct list_head smb_ses_list;
 	struct list_head rlist; /* reconnect list */
 	struct list_head tcon_list;
+	struct list_head dlist; /* dfs list */
 	struct cifs_tcon *tcon_ipc;
 	spinlock_t ses_lock;  /* protect anything here that is not protected */
 	struct mutex session_mutex;
@@ -1287,6 +1289,7 @@ struct cifs_tcon {
 	/* BB add field for back pointer to sb struct(s)? */
 #ifdef CONFIG_CIFS_DFS_UPCALL
 	struct delayed_work dfs_cache_work;
+	struct list_head dfs_ses_list;
 #endif
 	struct delayed_work	query_interfaces; /* query interfaces workqueue job */
 	char *origin_fullpath; /* canonical copy of smb3_fs_context::source */
diff --git a/fs/smb/client/cifsproto.h b/fs/smb/client/cifsproto.h
index c69e3f48a60c52..68c716e6261b3a 100644
--- a/fs/smb/client/cifsproto.h
+++ b/fs/smb/client/cifsproto.h
@@ -724,15 +724,9 @@ static inline int cifs_create_options(struct cifs_sb_info *cifs_sb, int options)
 
 int cifs_wait_for_server_reconnect(struct TCP_Server_Info *server, bool retry);
 
-/* Put references of @ses and its children */
 static inline void cifs_put_smb_ses(struct cifs_ses *ses)
 {
-	struct cifs_ses *next;
-
-	do {
-		next = ses->dfs_root_ses;
-		__cifs_put_smb_ses(ses);
-	} while ((ses = next));
+	__cifs_put_smb_ses(ses);
 }
 
 /* Get an active reference of @ses and its children.
@@ -746,9 +740,7 @@ static inline void cifs_put_smb_ses(struct cifs_ses *ses)
 static inline void cifs_smb_ses_inc_refcount(struct cifs_ses *ses)
 {
 	lockdep_assert_held(&cifs_tcp_ses_lock);
-
-	for (; ses; ses = ses->dfs_root_ses)
-		ses->ses_count++;
+	ses->ses_count++;
 }
 
 static inline bool dfs_src_pathname_equal(const char *s1, const char *s2)
diff --git a/fs/smb/client/connect.c b/fs/smb/client/connect.c
index 08a41c7aaf723f..76f02739dda5ff 100644
--- a/fs/smb/client/connect.c
+++ b/fs/smb/client/connect.c
@@ -1530,6 +1530,9 @@ static int match_server(struct TCP_Server_Info *server,
 	if (server->nosharesock)
 		return 0;
 
+	if (!match_super && (ctx->dfs_conn || server->dfs_conn))
+		return 0;
+
 	/* If multidialect negotiation see if existing sessions match one */
 	if (strcmp(ctx->vals->version_string, SMB3ANY_VERSION_STRING) == 0) {
 		if (server->vals->protocol_id < SMB30_PROT_ID)
@@ -1723,6 +1726,7 @@ cifs_get_tcp_session(struct smb3_fs_context *ctx,
 
 	if (ctx->nosharesock)
 		tcp_ses->nosharesock = true;
+	tcp_ses->dfs_conn = ctx->dfs_conn;
 
 	tcp_ses->ops = ctx->ops;
 	tcp_ses->vals = ctx->vals;
@@ -1873,13 +1877,15 @@ cifs_get_tcp_session(struct smb3_fs_context *ctx,
 }
 
 /* this function must be called with ses_lock and chan_lock held */
-static int match_session(struct cifs_ses *ses, struct smb3_fs_context *ctx)
+static int match_session(struct cifs_ses *ses,
+			 struct smb3_fs_context *ctx,
+			 bool match_super)
 {
 	if (ctx->sectype != Unspecified &&
 	    ctx->sectype != ses->sectype)
 		return 0;
 
-	if (ctx->dfs_root_ses != ses->dfs_root_ses)
+	if (!match_super && ctx->dfs_root_ses != ses->dfs_root_ses)
 		return 0;
 
 	/*
@@ -1998,7 +2004,7 @@ cifs_find_smb_ses(struct TCP_Server_Info *server, struct smb3_fs_context *ctx)
 			continue;
 		}
 		spin_lock(&ses->chan_lock);
-		if (match_session(ses, ctx)) {
+		if (match_session(ses, ctx, false)) {
 			spin_unlock(&ses->chan_lock);
 			spin_unlock(&ses->ses_lock);
 			ret = ses;
@@ -2382,8 +2388,6 @@ cifs_get_smb_ses(struct TCP_Server_Info *server, struct smb3_fs_context *ctx)
 	 * need to lock before changing something in the session.
 	 */
 	spin_lock(&cifs_tcp_ses_lock);
-	if (ctx->dfs_root_ses)
-		cifs_smb_ses_inc_refcount(ctx->dfs_root_ses);
 	ses->dfs_root_ses = ctx->dfs_root_ses;
 	list_add(&ses->smb_ses_list, &server->smb_ses_list);
 	spin_unlock(&cifs_tcp_ses_lock);
@@ -2458,6 +2462,7 @@ cifs_put_tcon(struct cifs_tcon *tcon, enum smb3_tcon_ref_trace trace)
 {
 	unsigned int xid;
 	struct cifs_ses *ses;
+	LIST_HEAD(ses_list);
 
 	/*
 	 * IPC tcon share the lifetime of their session and are
@@ -2482,6 +2487,9 @@ cifs_put_tcon(struct cifs_tcon *tcon, enum smb3_tcon_ref_trace trace)
 
 	list_del_init(&tcon->tcon_list);
 	tcon->status = TID_EXITING;
+#ifdef CONFIG_CIFS_DFS_UPCALL
+	list_replace_init(&tcon->dfs_ses_list, &ses_list);
+#endif
 	spin_unlock(&tcon->tc_lock);
 	spin_unlock(&cifs_tcp_ses_lock);
 
@@ -2509,6 +2517,9 @@ cifs_put_tcon(struct cifs_tcon *tcon, enum smb3_tcon_ref_trace trace)
 	cifs_fscache_release_super_cookie(tcon);
 	tconInfoFree(tcon, netfs_trace_tcon_ref_free);
 	cifs_put_smb_ses(ses);
+#ifdef CONFIG_CIFS_DFS_UPCALL
+	dfs_put_root_smb_sessions(&ses_list);
+#endif
 }
 
 /**
@@ -2892,7 +2903,7 @@ cifs_match_super(struct super_block *sb, void *data)
 	spin_lock(&ses->chan_lock);
 	spin_lock(&tcon->tc_lock);
 	if (!match_server(tcp_srv, ctx, true) ||
-	    !match_session(ses, ctx) ||
+	    !match_session(ses, ctx, true) ||
 	    !match_tcon(tcon, ctx) ||
 	    !match_prepath(sb, tcon, mnt_data)) {
 		rc = 0;
@@ -3623,13 +3634,12 @@ int cifs_is_path_remote(struct cifs_mount_ctx *mnt_ctx)
 int cifs_mount(struct cifs_sb_info *cifs_sb, struct smb3_fs_context *ctx)
 {
 	struct cifs_mount_ctx mnt_ctx = { .cifs_sb = cifs_sb, .fs_ctx = ctx, };
-	bool isdfs;
 	int rc;
 
-	rc = dfs_mount_share(&mnt_ctx, &isdfs);
+	rc = dfs_mount_share(&mnt_ctx);
 	if (rc)
 		goto error;
-	if (!isdfs)
+	if (!ctx->dfs_conn)
 		goto out;
 
 	/*
@@ -4034,7 +4044,7 @@ cifs_set_vol_auth(struct smb3_fs_context *ctx, struct cifs_ses *ses)
 }
 
 static struct cifs_tcon *
-__cifs_construct_tcon(struct cifs_sb_info *cifs_sb, kuid_t fsuid)
+cifs_construct_tcon(struct cifs_sb_info *cifs_sb, kuid_t fsuid)
 {
 	int rc;
 	struct cifs_tcon *master_tcon = cifs_sb_master_tcon(cifs_sb);
@@ -4132,17 +4142,6 @@ __cifs_construct_tcon(struct cifs_sb_info *cifs_sb, kuid_t fsuid)
 	return tcon;
 }
 
-static struct cifs_tcon *
-cifs_construct_tcon(struct cifs_sb_info *cifs_sb, kuid_t fsuid)
-{
-	struct cifs_tcon *ret;
-
-	cifs_mount_lock();
-	ret = __cifs_construct_tcon(cifs_sb, fsuid);
-	cifs_mount_unlock();
-	return ret;
-}
-
 struct cifs_tcon *
 cifs_sb_master_tcon(struct cifs_sb_info *cifs_sb)
 {
diff --git a/fs/smb/client/dfs.c b/fs/smb/client/dfs.c
index 3ec965547e3d4d..3f6077c68d68aa 100644
--- a/fs/smb/client/dfs.c
+++ b/fs/smb/client/dfs.c
@@ -69,7 +69,7 @@ static int get_session(struct cifs_mount_ctx *mnt_ctx, const char *full_path)
  * Get an active reference of @ses so that next call to cifs_put_tcon() won't
  * release it as any new DFS referrals must go through its IPC tcon.
  */
-static void add_root_smb_session(struct cifs_mount_ctx *mnt_ctx)
+static void set_root_smb_session(struct cifs_mount_ctx *mnt_ctx)
 {
 	struct smb3_fs_context *ctx = mnt_ctx->fs_ctx;
 	struct cifs_ses *ses = mnt_ctx->ses;
@@ -95,7 +95,7 @@ static inline int parse_dfs_target(struct smb3_fs_context *ctx,
 	return rc;
 }
 
-static int set_ref_paths(struct cifs_mount_ctx *mnt_ctx,
+static int setup_dfs_ref(struct cifs_mount_ctx *mnt_ctx,
 			 struct dfs_info3_param *tgt,
 			 struct dfs_ref_walk *rw)
 {
@@ -120,6 +120,7 @@ static int set_ref_paths(struct cifs_mount_ctx *mnt_ctx,
 	}
 	ref_walk_path(rw) = ref_path;
 	ref_walk_fpath(rw) = full_path;
+	ref_walk_ses(rw) = ctx->dfs_root_ses;
 	return 0;
 }
 
@@ -128,11 +129,11 @@ static int __dfs_referral_walk(struct cifs_mount_ctx *mnt_ctx,
 {
 	struct smb3_fs_context *ctx = mnt_ctx->fs_ctx;
 	struct dfs_info3_param tgt = {};
-	bool is_refsrv;
 	int rc = -ENOENT;
 
 again:
 	do {
+		ctx->dfs_root_ses = ref_walk_ses(rw);
 		if (ref_walk_empty(rw)) {
 			rc = dfs_get_referral(mnt_ctx, ref_walk_path(rw) + 1,
 					      NULL, ref_walk_tl(rw));
@@ -158,10 +159,7 @@ static int __dfs_referral_walk(struct cifs_mount_ctx *mnt_ctx,
 			if (rc)
 				continue;
 
-			is_refsrv = tgt.server_type == DFS_TYPE_ROOT ||
-				DFS_INTERLINK(tgt.flags);
 			ref_walk_set_tgt_hint(rw);
-
 			if (tgt.flags & DFSREF_STORAGE_SERVER) {
 				rc = cifs_mount_get_tcon(mnt_ctx);
 				if (!rc)
@@ -172,12 +170,10 @@ static int __dfs_referral_walk(struct cifs_mount_ctx *mnt_ctx,
 					continue;
 			}
 
-			if (is_refsrv)
-				add_root_smb_session(mnt_ctx);
-
+			set_root_smb_session(mnt_ctx);
 			rc = ref_walk_advance(rw);
 			if (!rc) {
-				rc = set_ref_paths(mnt_ctx, &tgt, rw);
+				rc = setup_dfs_ref(mnt_ctx, &tgt, rw);
 				if (!rc) {
 					rc = -EREMOTE;
 					goto again;
@@ -193,20 +189,22 @@ static int __dfs_referral_walk(struct cifs_mount_ctx *mnt_ctx,
 	return rc;
 }
 
-static int dfs_referral_walk(struct cifs_mount_ctx *mnt_ctx)
+static int dfs_referral_walk(struct cifs_mount_ctx *mnt_ctx,
+			     struct dfs_ref_walk **rw)
 {
-	struct dfs_ref_walk *rw;
 	int rc;
 
-	rw = ref_walk_alloc();
-	if (IS_ERR(rw))
-		return PTR_ERR(rw);
+	*rw = ref_walk_alloc();
+	if (IS_ERR(*rw)) {
+		rc = PTR_ERR(*rw);
+		*rw = NULL;
+		return rc;
+	}
 
-	ref_walk_init(rw);
-	rc = set_ref_paths(mnt_ctx, NULL, rw);
+	ref_walk_init(*rw);
+	rc = setup_dfs_ref(mnt_ctx, NULL, *rw);
 	if (!rc)
-		rc = __dfs_referral_walk(mnt_ctx, rw);
-	ref_walk_free(rw);
+		rc = __dfs_referral_walk(mnt_ctx, *rw);
 	return rc;
 }
 
@@ -214,16 +212,16 @@ static int __dfs_mount_share(struct cifs_mount_ctx *mnt_ctx)
 {
 	struct cifs_sb_info *cifs_sb = mnt_ctx->cifs_sb;
 	struct smb3_fs_context *ctx = mnt_ctx->fs_ctx;
+	struct dfs_ref_walk *rw = NULL;
 	struct cifs_tcon *tcon;
 	char *origin_fullpath;
-	bool new_tcon = true;
 	int rc;
 
 	origin_fullpath = dfs_get_path(cifs_sb, ctx->source);
 	if (IS_ERR(origin_fullpath))
 		return PTR_ERR(origin_fullpath);
 
-	rc = dfs_referral_walk(mnt_ctx);
+	rc = dfs_referral_walk(mnt_ctx, &rw);
 	if (!rc) {
 		/*
 		 * Prevent superblock from being created with any missing
@@ -241,21 +239,16 @@ static int __dfs_mount_share(struct cifs_mount_ctx *mnt_ctx)
 
 	tcon = mnt_ctx->tcon;
 	spin_lock(&tcon->tc_lock);
-	if (!tcon->origin_fullpath) {
-		tcon->origin_fullpath = origin_fullpath;
-		origin_fullpath = NULL;
-	} else {
-		new_tcon = false;
-	}
+	tcon->origin_fullpath = origin_fullpath;
+	origin_fullpath = NULL;
+	ref_walk_set_tcon(rw, tcon);
 	spin_unlock(&tcon->tc_lock);
-
-	if (new_tcon) {
-		queue_delayed_work(dfscache_wq, &tcon->dfs_cache_work,
-				   dfs_cache_get_ttl() * HZ);
-	}
+	queue_delayed_work(dfscache_wq, &tcon->dfs_cache_work,
+			   dfs_cache_get_ttl() * HZ);
 
 out:
 	kfree(origin_fullpath);
+	ref_walk_free(rw);
 	return rc;
 }
 
@@ -279,7 +272,7 @@ static int update_fs_context_dstaddr(struct smb3_fs_context *ctx)
 	return rc;
 }
 
-int dfs_mount_share(struct cifs_mount_ctx *mnt_ctx, bool *isdfs)
+int dfs_mount_share(struct cifs_mount_ctx *mnt_ctx)
 {
 	struct smb3_fs_context *ctx = mnt_ctx->fs_ctx;
 	bool nodfs = ctx->nodfs;
@@ -289,7 +282,6 @@ int dfs_mount_share(struct cifs_mount_ctx *mnt_ctx, bool *isdfs)
 	if (rc)
 		return rc;
 
-	*isdfs = false;
 	rc = get_session(mnt_ctx, NULL);
 	if (rc)
 		return rc;
@@ -317,10 +309,15 @@ int dfs_mount_share(struct cifs_mount_ctx *mnt_ctx, bool *isdfs)
 		return rc;
 	}
 
-	*isdfs = true;
-	add_root_smb_session(mnt_ctx);
-	rc = __dfs_mount_share(mnt_ctx);
-	dfs_put_root_smb_sessions(mnt_ctx);
+	if (!ctx->dfs_conn) {
+		ctx->dfs_conn = true;
+		cifs_mount_put_conns(mnt_ctx);
+		rc = get_session(mnt_ctx, NULL);
+	}
+	if (!rc) {
+		set_root_smb_session(mnt_ctx);
+		rc = __dfs_mount_share(mnt_ctx);
+	}
 	return rc;
 }
 
diff --git a/fs/smb/client/dfs.h b/fs/smb/client/dfs.h
index e5c4dcf837503a..1aa2bc65b3bc2c 100644
--- a/fs/smb/client/dfs.h
+++ b/fs/smb/client/dfs.h
@@ -19,6 +19,7 @@
 struct dfs_ref {
 	char *path;
 	char *full_path;
+	struct cifs_ses *ses;
 	struct dfs_cache_tgt_list tl;
 	struct dfs_cache_tgt_iterator *tit;
 };
@@ -38,6 +39,7 @@ struct dfs_ref_walk {
 #define ref_walk_path(w)	(ref_walk_cur(w)->path)
 #define ref_walk_fpath(w)	(ref_walk_cur(w)->full_path)
 #define ref_walk_tl(w)		(&ref_walk_cur(w)->tl)
+#define ref_walk_ses(w)	(ref_walk_cur(w)->ses)
 
 static inline struct dfs_ref_walk *ref_walk_alloc(void)
 {
@@ -60,14 +62,19 @@ static inline void __ref_walk_free(struct dfs_ref *ref)
 	kfree(ref->path);
 	kfree(ref->full_path);
 	dfs_cache_free_tgts(&ref->tl);
+	if (ref->ses)
+		cifs_put_smb_ses(ref->ses);
 	memset(ref, 0, sizeof(*ref));
 }
 
 static inline void ref_walk_free(struct dfs_ref_walk *rw)
 {
-	struct dfs_ref *ref = ref_walk_start(rw);
+	struct dfs_ref *ref;
 
-	for (; ref <= ref_walk_end(rw); ref++)
+	if (!rw)
+		return;
+
+	for (ref = ref_walk_start(rw); ref <= ref_walk_end(rw); ref++)
 		__ref_walk_free(ref);
 	kfree(rw);
 }
@@ -116,9 +123,22 @@ static inline void ref_walk_set_tgt_hint(struct dfs_ref_walk *rw)
 				       ref_walk_tit(rw));
 }
 
+static inline void ref_walk_set_tcon(struct dfs_ref_walk *rw,
+				     struct cifs_tcon *tcon)
+{
+	struct dfs_ref *ref = ref_walk_start(rw);
+
+	for (; ref <= ref_walk_cur(rw); ref++) {
+		if (WARN_ON_ONCE(!ref->ses))
+			continue;
+		list_add(&ref->ses->dlist, &tcon->dfs_ses_list);
+		ref->ses = NULL;
+	}
+}
+
 int dfs_parse_target_referral(const char *full_path, const struct dfs_info3_param *ref,
 			      struct smb3_fs_context *ctx);
-int dfs_mount_share(struct cifs_mount_ctx *mnt_ctx, bool *isdfs);
+int dfs_mount_share(struct cifs_mount_ctx *mnt_ctx);
 
 static inline char *dfs_get_path(struct cifs_sb_info *cifs_sb, const char *path)
 {
@@ -142,20 +162,14 @@ static inline int dfs_get_referral(struct cifs_mount_ctx *mnt_ctx, const char *p
  * references of all DFS root sessions that were used across the mount process
  * in dfs_mount_share().
  */
-static inline void dfs_put_root_smb_sessions(struct cifs_mount_ctx *mnt_ctx)
+static inline void dfs_put_root_smb_sessions(struct list_head *head)
 {
-	const struct smb3_fs_context *ctx = mnt_ctx->fs_ctx;
-	struct cifs_ses *ses = ctx->dfs_root_ses;
-	struct cifs_ses *cur;
-
-	if (!ses)
-		return;
+	struct cifs_ses *ses, *n;
 
-	for (cur = ses; cur; cur = cur->dfs_root_ses) {
-		if (cur->dfs_root_ses)
-			cifs_put_smb_ses(cur->dfs_root_ses);
+	list_for_each_entry_safe(ses, n, head, dlist) {
+		list_del_init(&ses->dlist);
+		cifs_put_smb_ses(ses);
 	}
-	cifs_put_smb_ses(ses);
 }
 
 #endif /* _CIFS_DFS_H */
diff --git a/fs/smb/client/dfs_cache.c b/fs/smb/client/dfs_cache.c
index e496dbf8c6228f..110f03df012a00 100644
--- a/fs/smb/client/dfs_cache.c
+++ b/fs/smb/client/dfs_cache.c
@@ -1334,9 +1334,8 @@ void dfs_cache_refresh(struct work_struct *work)
 	struct cifs_ses *ses;
 
 	tcon = container_of(work, struct cifs_tcon, dfs_cache_work.work);
-	ses = tcon->ses->dfs_root_ses;
 
-	for (; ses; ses = ses->dfs_root_ses)
+	list_for_each_entry(ses, &tcon->dfs_ses_list, dlist)
 		refresh_ses_referral(ses);
 	refresh_tcon_referral(tcon, false);
 
diff --git a/fs/smb/client/fs_context.h b/fs/smb/client/fs_context.h
index cf577ec0dd0ac4..69f9d938b336a7 100644
--- a/fs/smb/client/fs_context.h
+++ b/fs/smb/client/fs_context.h
@@ -284,6 +284,7 @@ struct smb3_fs_context {
 	struct cifs_ses *dfs_root_ses;
 	bool dfs_automount:1; /* set for dfs automount only */
 	enum cifs_reparse_type reparse_type;
+	bool dfs_conn:1; /* set for dfs mounts */
 };
 
 extern const struct fs_parameter_spec smb3_fs_parameters[];
diff --git a/fs/smb/client/misc.c b/fs/smb/client/misc.c
index dab526191b0742..47b861517bed53 100644
--- a/fs/smb/client/misc.c
+++ b/fs/smb/client/misc.c
@@ -145,6 +145,9 @@ tcon_info_alloc(bool dir_leases_enabled, enum smb3_tcon_ref_trace trace)
 	mutex_init(&ret_buf->fscache_lock);
 #endif
 	trace_smb3_tcon_ref(ret_buf->debug_id, ret_buf->tc_count, trace);
+#ifdef CONFIG_CIFS_DFS_UPCALL
+	INIT_LIST_HEAD(&ret_buf->dfs_ses_list);
+#endif
 
 	return ret_buf;
 }
diff --git a/fs/smb/client/namespace.c b/fs/smb/client/namespace.c
index 4a517b280f2b79..0f788031b7405f 100644
--- a/fs/smb/client/namespace.c
+++ b/fs/smb/client/namespace.c
@@ -240,7 +240,7 @@ static struct vfsmount *cifs_do_automount(struct path *path)
 		ctx->source = NULL;
 		goto out;
 	}
-	ctx->dfs_automount = is_dfs_mount(mntpt);
+	ctx->dfs_automount = ctx->dfs_conn = is_dfs_mount(mntpt);
 	cifs_dbg(FYI, "%s: ctx: source=%s UNC=%s prepath=%s dfs_automount=%d\n",
 		 __func__, ctx->source, ctx->UNC, ctx->prepath, ctx->dfs_automount);
 

From 85633c00ad03049019df632f2bdcf5ff7efc7796 Mon Sep 17 00:00:00 2001
From: David Howells <dhowells@redhat.com>
Date: Wed, 18 Sep 2024 20:24:24 +0100
Subject: [PATCH 04/13] cifs: Make the write_{enter,done,err} tracepoints
 display netfs info

Make the write RPC tracepoints use the same trace macro complexes as the
read tracepoints and display the netfs request and subrequest IDs where
available (see commit 519be989717c "cifs: Add a tracepoint to track credits
involved in R/W requests").

Signed-off-by: David Howells <dhowells@redhat.com>
cc: Steve French <stfrench@microsoft.com>
cc: Paulo Alcantara (Red Hat) <pc@manguebit.com>
cc: Jeff Layton <jlayton@kernel.org>
cc: linux-cifs@vger.kernel.org
cc: netfs@lists.linux.dev
cc: linux-fsdevel@vger.kernel.org
Signed-off-by: Steve French <stfrench@microsoft.com>
---
 fs/smb/client/smb2pdu.c | 22 +++++++++++++++-------
 fs/smb/client/trace.h   |  6 +++---
 2 files changed, 18 insertions(+), 10 deletions(-)

diff --git a/fs/smb/client/smb2pdu.c b/fs/smb/client/smb2pdu.c
index 2cb1bf65a172a5..bb225758448a6a 100644
--- a/fs/smb/client/smb2pdu.c
+++ b/fs/smb/client/smb2pdu.c
@@ -4866,7 +4866,9 @@ smb2_writev_callback(struct mid_q_entry *mid)
 #endif
 	if (result) {
 		cifs_stats_fail_inc(tcon, SMB2_WRITE_HE);
-		trace_smb3_write_err(wdata->xid,
+		trace_smb3_write_err(wdata->rreq->debug_id,
+				     wdata->subreq.debug_index,
+				     wdata->xid,
 				     wdata->req->cfile->fid.persistent_fid,
 				     tcon->tid, tcon->ses->Suid, wdata->subreq.start,
 				     wdata->subreq.len, wdata->result);
@@ -4874,7 +4876,9 @@ smb2_writev_callback(struct mid_q_entry *mid)
 			pr_warn_once("Out of space writing to %s\n",
 				     tcon->tree_name);
 	} else
-		trace_smb3_write_done(0 /* no xid */,
+		trace_smb3_write_done(wdata->rreq->debug_id,
+				      wdata->subreq.debug_index,
+				      wdata->xid,
 				      wdata->req->cfile->fid.persistent_fid,
 				      tcon->tid, tcon->ses->Suid,
 				      wdata->subreq.start, wdata->subreq.len);
@@ -4952,7 +4956,9 @@ smb2_async_writev(struct cifs_io_subrequest *wdata)
 				offsetof(struct smb2_write_req, Buffer));
 	req->RemainingBytes = 0;
 
-	trace_smb3_write_enter(wdata->xid,
+	trace_smb3_write_enter(wdata->rreq->debug_id,
+			       wdata->subreq.debug_index,
+			       wdata->xid,
 			       io_parms->persistent_fid,
 			       io_parms->tcon->tid,
 			       io_parms->tcon->ses->Suid,
@@ -5032,7 +5038,9 @@ smb2_async_writev(struct cifs_io_subrequest *wdata)
 			     wdata, flags, &wdata->credits);
 	/* Can't touch wdata if rc == 0 */
 	if (rc) {
-		trace_smb3_write_err(xid,
+		trace_smb3_write_err(wdata->rreq->debug_id,
+				     wdata->subreq.debug_index,
+				     xid,
 				     io_parms->persistent_fid,
 				     io_parms->tcon->tid,
 				     io_parms->tcon->ses->Suid,
@@ -5112,7 +5120,7 @@ SMB2_write(const unsigned int xid, struct cifs_io_parms *io_parms,
 				offsetof(struct smb2_write_req, Buffer));
 	req->RemainingBytes = 0;
 
-	trace_smb3_write_enter(xid, io_parms->persistent_fid,
+	trace_smb3_write_enter(0, 0, xid, io_parms->persistent_fid,
 		io_parms->tcon->tid, io_parms->tcon->ses->Suid,
 		io_parms->offset, io_parms->length);
 
@@ -5133,7 +5141,7 @@ SMB2_write(const unsigned int xid, struct cifs_io_parms *io_parms,
 	rsp = (struct smb2_write_rsp *)rsp_iov.iov_base;
 
 	if (rc) {
-		trace_smb3_write_err(xid,
+		trace_smb3_write_err(0, 0, xid,
 				     req->PersistentFileId,
 				     io_parms->tcon->tid,
 				     io_parms->tcon->ses->Suid,
@@ -5142,7 +5150,7 @@ SMB2_write(const unsigned int xid, struct cifs_io_parms *io_parms,
 		cifs_dbg(VFS, "Send error in write = %d\n", rc);
 	} else {
 		*nbytes = le32_to_cpu(rsp->DataLength);
-		trace_smb3_write_done(xid,
+		trace_smb3_write_done(0, 0, xid,
 				      req->PersistentFileId,
 				      io_parms->tcon->tid,
 				      io_parms->tcon->ses->Suid,
diff --git a/fs/smb/client/trace.h b/fs/smb/client/trace.h
index 8e9964001e2aed..0b52d22a91a0cb 100644
--- a/fs/smb/client/trace.h
+++ b/fs/smb/client/trace.h
@@ -157,6 +157,7 @@ DEFINE_EVENT(smb3_rw_err_class, smb3_##name,    \
 	TP_ARGS(rreq_debug_id, rreq_debug_index, xid, fid, tid, sesid, offset, len, rc))
 
 DEFINE_SMB3_RW_ERR_EVENT(read_err);
+DEFINE_SMB3_RW_ERR_EVENT(write_err);
 
 /* For logging errors in other file I/O ops */
 DECLARE_EVENT_CLASS(smb3_other_err_class,
@@ -202,7 +203,6 @@ DEFINE_EVENT(smb3_other_err_class, smb3_##name, \
 		int	rc),			\
 	TP_ARGS(xid, fid, tid, sesid, offset, len, rc))
 
-DEFINE_SMB3_OTHER_ERR_EVENT(write_err);
 DEFINE_SMB3_OTHER_ERR_EVENT(query_dir_err);
 DEFINE_SMB3_OTHER_ERR_EVENT(zero_err);
 DEFINE_SMB3_OTHER_ERR_EVENT(falloc_err);
@@ -370,6 +370,8 @@ DEFINE_EVENT(smb3_rw_done_class, smb3_##name,   \
 
 DEFINE_SMB3_RW_DONE_EVENT(read_enter);
 DEFINE_SMB3_RW_DONE_EVENT(read_done);
+DEFINE_SMB3_RW_DONE_EVENT(write_enter);
+DEFINE_SMB3_RW_DONE_EVENT(write_done);
 
 /* For logging successful other op */
 DECLARE_EVENT_CLASS(smb3_other_done_class,
@@ -411,11 +413,9 @@ DEFINE_EVENT(smb3_other_done_class, smb3_##name,   \
 		__u32	len),			\
 	TP_ARGS(xid, fid, tid, sesid, offset, len))
 
-DEFINE_SMB3_OTHER_DONE_EVENT(write_enter);
 DEFINE_SMB3_OTHER_DONE_EVENT(query_dir_enter);
 DEFINE_SMB3_OTHER_DONE_EVENT(zero_enter);
 DEFINE_SMB3_OTHER_DONE_EVENT(falloc_enter);
-DEFINE_SMB3_OTHER_DONE_EVENT(write_done);
 DEFINE_SMB3_OTHER_DONE_EVENT(query_dir_done);
 DEFINE_SMB3_OTHER_DONE_EVENT(zero_done);
 DEFINE_SMB3_OTHER_DONE_EVENT(falloc_done);

From 0826b134c0b039db4850fb762e79766a45d847c5 Mon Sep 17 00:00:00 2001
From: Paulo Alcantara <pc@manguebit.com>
Date: Wed, 18 Sep 2024 02:04:23 -0300
Subject: [PATCH 05/13] smb: client: fix DFS failover in multiuser mounts

For sessions and tcons created on behalf of new users accessing a
multiuser mount, matching their sessions in tcon_super_cb() with
master tcon will always lead to false as every new user will have its
own session and tcon.

All multiuser sessions, however, will inherit ->dfs_root_ses from
master tcon, so match it instead.

Signed-off-by: Paulo Alcantara (Red Hat) <pc@manguebit.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
---
 fs/smb/client/misc.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/fs/smb/client/misc.c b/fs/smb/client/misc.c
index 47b861517bed53..054f10ebf65a5e 100644
--- a/fs/smb/client/misc.c
+++ b/fs/smb/client/misc.c
@@ -1111,7 +1111,8 @@ static void tcon_super_cb(struct super_block *sb, void *arg)
 	t2 = cifs_sb_master_tcon(cifs_sb);
 
 	spin_lock(&t2->tc_lock);
-	if (t1->ses == t2->ses &&
+	if ((t1->ses == t2->ses ||
+	     t1->ses->dfs_root_ses == t2->ses->dfs_root_ses) &&
 	    t1->ses->server == t2->ses->server &&
 	    t2->origin_fullpath &&
 	    dfs_src_pathname_equal(t2->origin_fullpath, t1->origin_fullpath))

From 4e3ba580f5ab2d74e1e2210aba869aad235349d5 Mon Sep 17 00:00:00 2001
From: Paulo Alcantara <pc@manguebit.com>
Date: Wed, 18 Sep 2024 02:04:27 -0300
Subject: [PATCH 06/13] smb: client: propagate error from cifs_construct_tcon()

Propagate error from cifs_construct_tcon() in cifs_sb_tlink() instead of
always returning -EACCES.

Signed-off-by: Paulo Alcantara (Red Hat) <pc@manguebit.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
---
 fs/smb/client/connect.c | 16 ++++++++++------
 1 file changed, 10 insertions(+), 6 deletions(-)

diff --git a/fs/smb/client/connect.c b/fs/smb/client/connect.c
index 76f02739dda5ff..188a3a1aafa0b9 100644
--- a/fs/smb/client/connect.c
+++ b/fs/smb/client/connect.c
@@ -4211,9 +4211,9 @@ tlink_rb_insert(struct rb_root *root, struct tcon_link *new_tlink)
 struct tcon_link *
 cifs_sb_tlink(struct cifs_sb_info *cifs_sb)
 {
-	int ret;
-	kuid_t fsuid = current_fsuid();
 	struct tcon_link *tlink, *newtlink;
+	kuid_t fsuid = current_fsuid();
+	int err;
 
 	if (!(cifs_sb->mnt_cifs_flags & CIFS_MOUNT_MULTIUSER))
 		return cifs_get_tlink(cifs_sb_master_tlink(cifs_sb));
@@ -4248,9 +4248,9 @@ cifs_sb_tlink(struct cifs_sb_info *cifs_sb)
 		spin_unlock(&cifs_sb->tlink_tree_lock);
 	} else {
 wait_for_construction:
-		ret = wait_on_bit(&tlink->tl_flags, TCON_LINK_PENDING,
+		err = wait_on_bit(&tlink->tl_flags, TCON_LINK_PENDING,
 				  TASK_INTERRUPTIBLE);
-		if (ret) {
+		if (err) {
 			cifs_put_tlink(tlink);
 			return ERR_PTR(-ERESTARTSYS);
 		}
@@ -4261,8 +4261,9 @@ cifs_sb_tlink(struct cifs_sb_info *cifs_sb)
 
 		/* return error if we tried this already recently */
 		if (time_before(jiffies, tlink->tl_time + TLINK_ERROR_EXPIRE)) {
+			err = PTR_ERR(tlink->tl_tcon);
 			cifs_put_tlink(tlink);
-			return ERR_PTR(-EACCES);
+			return ERR_PTR(err);
 		}
 
 		if (test_and_set_bit(TCON_LINK_PENDING, &tlink->tl_flags))
@@ -4274,8 +4275,11 @@ cifs_sb_tlink(struct cifs_sb_info *cifs_sb)
 	wake_up_bit(&tlink->tl_flags, TCON_LINK_PENDING);
 
 	if (IS_ERR(tlink->tl_tcon)) {
+		err = PTR_ERR(tlink->tl_tcon);
+		if (err == -ENOKEY)
+			err = -EACCES;
 		cifs_put_tlink(tlink);
-		return ERR_PTR(-EACCES);
+		return ERR_PTR(err);
 	}
 
 	return tlink;

From a9de67336a4aa3ff2e706ba023fb5f7ff681a954 Mon Sep 17 00:00:00 2001
From: Paulo Alcantara <pc@manguebit.com>
Date: Wed, 18 Sep 2024 21:53:35 -0300
Subject: [PATCH 07/13] smb: client: set correct device number on nfs reparse
 points

Fix major and minor numbers set on special files created with NFS
reparse points.

Signed-off-by: Paulo Alcantara (Red Hat) <pc@manguebit.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
---
 fs/smb/client/reparse.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/fs/smb/client/reparse.c b/fs/smb/client/reparse.c
index 48c27581ec511c..63984796721a50 100644
--- a/fs/smb/client/reparse.c
+++ b/fs/smb/client/reparse.c
@@ -108,8 +108,8 @@ static int nfs_set_reparse_buf(struct reparse_posix_data *buf,
 	buf->InodeType = cpu_to_le64(type);
 	buf->ReparseDataLength = cpu_to_le16(len + dlen -
 					     sizeof(struct reparse_data_buffer));
-	*(__le64 *)buf->DataBuffer = cpu_to_le64(((u64)MAJOR(dev) << 32) |
-						 MINOR(dev));
+	*(__le64 *)buf->DataBuffer = cpu_to_le64(((u64)MINOR(dev) << 32) |
+						 MAJOR(dev));
 	iov->iov_base = buf;
 	iov->iov_len = len + dlen;
 	return 0;

From 663f295e35594f4c2584fc68c28546b747b637cd Mon Sep 17 00:00:00 2001
From: Paulo Alcantara <pc@manguebit.com>
Date: Wed, 18 Sep 2024 21:57:43 -0300
Subject: [PATCH 08/13] smb: client: fix parsing of device numbers

Report correct major and minor numbers from special files created with
NFS reparse points.

Signed-off-by: Paulo Alcantara (Red Hat) <pc@manguebit.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
---
 fs/smb/client/reparse.c | 6 +++---
 fs/smb/client/reparse.h | 9 +--------
 2 files changed, 4 insertions(+), 11 deletions(-)

diff --git a/fs/smb/client/reparse.c b/fs/smb/client/reparse.c
index 63984796721a50..3b48a093cfb1f8 100644
--- a/fs/smb/client/reparse.c
+++ b/fs/smb/client/reparse.c
@@ -468,7 +468,7 @@ static void wsl_to_fattr(struct cifs_open_info_data *data,
 		else if (!strncmp(name, SMB2_WSL_XATTR_MODE, nlen))
 			fattr->cf_mode = (umode_t)le32_to_cpu(*(__le32 *)v);
 		else if (!strncmp(name, SMB2_WSL_XATTR_DEV, nlen))
-			fattr->cf_rdev = wsl_mkdev(v);
+			fattr->cf_rdev = reparse_mkdev(v);
 	} while (next);
 out:
 	fattr->cf_dtype = S_DT(fattr->cf_mode);
@@ -485,11 +485,11 @@ bool cifs_reparse_point_to_fattr(struct cifs_sb_info *cifs_sb,
 		switch (le64_to_cpu(buf->InodeType)) {
 		case NFS_SPECFILE_CHR:
 			fattr->cf_mode |= S_IFCHR;
-			fattr->cf_rdev = reparse_nfs_mkdev(buf);
+			fattr->cf_rdev = reparse_mkdev(buf->DataBuffer);
 			break;
 		case NFS_SPECFILE_BLK:
 			fattr->cf_mode |= S_IFBLK;
-			fattr->cf_rdev = reparse_nfs_mkdev(buf);
+			fattr->cf_rdev = reparse_mkdev(buf->DataBuffer);
 			break;
 		case NFS_SPECFILE_FIFO:
 			fattr->cf_mode |= S_IFIFO;
diff --git a/fs/smb/client/reparse.h b/fs/smb/client/reparse.h
index 2c0644bc4e65a7..158e7b7aae646c 100644
--- a/fs/smb/client/reparse.h
+++ b/fs/smb/client/reparse.h
@@ -18,14 +18,7 @@
  */
 #define IO_REPARSE_TAG_INTERNAL ((__u32)~0U)
 
-static inline dev_t reparse_nfs_mkdev(struct reparse_posix_data *buf)
-{
-	u64 v = le64_to_cpu(*(__le64 *)buf->DataBuffer);
-
-	return MKDEV(v >> 32, v & 0xffffffff);
-}
-
-static inline dev_t wsl_mkdev(void *ptr)
+static inline dev_t reparse_mkdev(void *ptr)
 {
 	u64 v = le64_to_cpu(*(__le64 *)ptr);
 

From 2f3017e7cc7515e0110a3733d8dca84de2a1d23d Mon Sep 17 00:00:00 2001
From: Steve French <stfrench@microsoft.com>
Date: Sat, 21 Sep 2024 23:28:32 -0500
Subject: [PATCH 09/13] smb3: fix incorrect mode displayed for read-only files

Commands like "chmod 0444" mark a file readonly via the attribute flag
(when mapping of mode bits into the ACL are not set, or POSIX extensions
are not negotiated), but they were not reported correctly for stat of
directories (they were reported ok for files and for "ls").  See example
below:

    root:~# ls /mnt2 -l
    total 12
    drwxr-xr-x 2 root root         0 Sep 21 18:03 normaldir
    -rwxr-xr-x 1 root root         0 Sep 21 23:24 normalfile
    dr-xr-xr-x 2 root root         0 Sep 21 17:55 readonly-dir
    -r-xr-xr-x 1 root root 209716224 Sep 21 18:15 readonly-file
    root:~# stat -c %a /mnt2/readonly-dir
    755
    root:~# stat -c %a /mnt2/readonly-file
    555

This fixes the stat of directories when ATTR_READONLY is set
(in cases where the mode can not be obtained other ways).

    root:~# stat -c %a /mnt2/readonly-dir
    555

Cc: stable@vger.kernel.org
Signed-off-by: Steve French <stfrench@microsoft.com>
---
 fs/smb/client/inode.c | 19 +++++++++++--------
 1 file changed, 11 insertions(+), 8 deletions(-)

diff --git a/fs/smb/client/inode.c b/fs/smb/client/inode.c
index 331a86074ae78c..647f9bedd9fc44 100644
--- a/fs/smb/client/inode.c
+++ b/fs/smb/client/inode.c
@@ -834,10 +834,6 @@ static void cifs_open_info_to_fattr(struct cifs_fattr *fattr,
 		fattr->cf_mode = S_IFREG | cifs_sb->ctx->file_mode;
 		fattr->cf_dtype = DT_REG;
 
-		/* clear write bits if ATTR_READONLY is set */
-		if (fattr->cf_cifsattrs & ATTR_READONLY)
-			fattr->cf_mode &= ~(S_IWUGO);
-
 		/*
 		 * Don't accept zero nlink from non-unix servers unless
 		 * delete is pending.  Instead mark it as unknown.
@@ -850,6 +846,10 @@ static void cifs_open_info_to_fattr(struct cifs_fattr *fattr,
 		}
 	}
 
+	/* clear write bits if ATTR_READONLY is set */
+	if (fattr->cf_cifsattrs & ATTR_READONLY)
+		fattr->cf_mode &= ~(S_IWUGO);
+
 out_reparse:
 	if (S_ISLNK(fattr->cf_mode)) {
 		if (likely(data->symlink_target))
@@ -1267,11 +1267,14 @@ static int cifs_get_fattr(struct cifs_open_info_data *data,
 				 __func__, rc);
 			goto out;
 		}
-	}
-
-	/* fill in remaining high mode bits e.g. SUID, VTX */
-	if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_UNX_EMUL)
+	} else if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_UNX_EMUL)
+		/* fill in remaining high mode bits e.g. SUID, VTX */
 		cifs_sfu_mode(fattr, full_path, cifs_sb, xid);
+	else if (!(tcon->posix_extensions))
+		/* clear write bits if ATTR_READONLY is set */
+		if (fattr->cf_cifsattrs & ATTR_READONLY)
+			fattr->cf_mode &= ~(S_IWUGO);
+
 
 	/* check for Minshall+French symlinks */
 	if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_MF_SYMLINKS) {

From 307f77e7f5855cd42c62fee3f97e4dea5a04a15b Mon Sep 17 00:00:00 2001
From: David Howells <dhowells@redhat.com>
Date: Mon, 23 Sep 2024 16:07:51 +0100
Subject: [PATCH 10/13] cifs: Fix reversion of the iter in
 cifs_readv_receive().

cifs_read_iter_from_socket() copies the iterator that's passed in for the
socket to modify as and if it will, and then advances the original iterator
by the amount sent.  However, both callers revert the advancement (although
receive_encrypted_read() zeros beyond the iterator first).  The problem is,
though, that cifs_readv_receive() reverts by the original length, not the
amount transmitted which can cause an oops in iov_iter_revert().

Fix this by:

 (1) Remove the iov_iter_advance() from cifs_read_iter_from_socket().

 (2) Remove the iov_iter_revert() from both callers.  This fixes the bug in
     cifs_readv_receive().

 (3) In receive_encrypted_read(), if we didn't get back as much data as the
     buffer will hold, copy the iterator, advance the copy and use the copy
     to drive iov_iter_zero().

As a bonus, this gets rid of some unnecessary work.

This was triggered by generic/074 with the "-o sign" mount option.

Fixes: 3ee1a1fc3981 ("cifs: Cut over to using netfslib")
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Steve French <sfrench@samba.org>
cc: Paulo Alcantara <pc@manguebit.com>
cc: Shyam Prasad N <nspmangalore@gmail.com>
cc: Rohith Surabattula <rohiths.msft@gmail.com>
cc: Jeff Layton <jlayton@kernel.org>
cc: linux-cifs@vger.kernel.org
cc: netfs@lists.linux.dev
cc: linux-fsdevel@vger.kernel.org
Signed-off-by: Steve French <stfrench@microsoft.com>
---
 fs/smb/client/connect.c   | 6 +-----
 fs/smb/client/smb2ops.c   | 9 ++++++---
 fs/smb/client/transport.c | 3 ---
 3 files changed, 7 insertions(+), 11 deletions(-)

diff --git a/fs/smb/client/connect.c b/fs/smb/client/connect.c
index 188a3a1aafa0b9..0e00c9846f29fc 100644
--- a/fs/smb/client/connect.c
+++ b/fs/smb/client/connect.c
@@ -811,13 +811,9 @@ cifs_read_iter_from_socket(struct TCP_Server_Info *server, struct iov_iter *iter
 			   unsigned int to_read)
 {
 	struct msghdr smb_msg = { .msg_iter = *iter };
-	int ret;
 
 	iov_iter_truncate(&smb_msg.msg_iter, to_read);
-	ret = cifs_readv_from_socket(server, &smb_msg);
-	if (ret > 0)
-		iov_iter_advance(iter, ret);
-	return ret;
+	return cifs_readv_from_socket(server, &smb_msg);
 }
 
 static bool
diff --git a/fs/smb/client/smb2ops.c b/fs/smb/client/smb2ops.c
index 7381ec333c6d22..1ee2dd4a1cae0a 100644
--- a/fs/smb/client/smb2ops.c
+++ b/fs/smb/client/smb2ops.c
@@ -4869,9 +4869,12 @@ receive_encrypted_read(struct TCP_Server_Info *server, struct mid_q_entry **mid,
 		goto discard_data;
 
 	server->total_read += rc;
-	if (rc < len)
-		iov_iter_zero(len - rc, &iter);
-	iov_iter_revert(&iter, len);
+	if (rc < len) {
+		struct iov_iter tmp = iter;
+
+		iov_iter_advance(&tmp, rc);
+		iov_iter_zero(len - rc, &tmp);
+	}
 	iov_iter_truncate(&iter, dw->len);
 
 	rc = cifs_discard_remaining_data(server);
diff --git a/fs/smb/client/transport.c b/fs/smb/client/transport.c
index fd5a85d437590b..91812150186c01 100644
--- a/fs/smb/client/transport.c
+++ b/fs/smb/client/transport.c
@@ -1817,11 +1817,8 @@ cifs_readv_receive(struct TCP_Server_Info *server, struct mid_q_entry *mid)
 		length = data_len; /* An RDMA read is already done. */
 	else
 #endif
-	{
 		length = cifs_read_iter_from_socket(server, &rdata->subreq.io_iter,
 						    data_len);
-		iov_iter_revert(&rdata->subreq.io_iter, data_len);
-	}
 	if (length > 0)
 		rdata->got_bytes += length;
 	server->total_read += length;

From 6c7f1b994a025a2d7748104ea9fc5e7d5808092a Mon Sep 17 00:00:00 2001
From: Paulo Alcantara <pc@manguebit.com>
Date: Wed, 18 Sep 2024 02:04:06 -0300
Subject: [PATCH 11/13] smb: client: print failed session logoffs with FYI

Do not flood dmesg with failed session logoffs as kerberos tickets
getting expired or passwords being rotated is a very common scenario.

Signed-off-by: Paulo Alcantara (Red Hat) <pc@manguebit.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
---
 fs/smb/client/connect.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/fs/smb/client/connect.c b/fs/smb/client/connect.c
index 0e00c9846f29fc..adf8758847f63f 100644
--- a/fs/smb/client/connect.c
+++ b/fs/smb/client/connect.c
@@ -2060,8 +2060,7 @@ void __cifs_put_smb_ses(struct cifs_ses *ses)
 	if (do_logoff) {
 		xid = get_xid();
 		rc = server->ops->logoff(xid, ses);
-		if (rc)
-			cifs_server_dbg(VFS, "%s: Session Logoff failure rc=%d\n",
+		cifs_server_dbg(FYI, "%s: Session Logoff: rc=%d\n",
 				__func__, rc);
 		_free_xid(xid);
 	}

From 387676fabf15f8e772fd22dd05794639115e4216 Mon Sep 17 00:00:00 2001
From: Steve French <stfrench@microsoft.com>
Date: Fri, 26 Jul 2024 18:44:16 -0500
Subject: [PATCH 12/13] cifs: update internal version number

To 2.51

Signed-off-by: Steve French <stfrench@microsoft.com>
---
 fs/smb/client/cifsfs.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/fs/smb/client/cifsfs.h b/fs/smb/client/cifsfs.h
index 61ded59b858f49..71b720dbb2ce34 100644
--- a/fs/smb/client/cifsfs.h
+++ b/fs/smb/client/cifsfs.h
@@ -146,6 +146,6 @@ extern const struct export_operations cifs_export_ops;
 #endif /* CONFIG_CIFS_NFSD_EXPORT */
 
 /* when changing internal version - update following two lines at same time */
-#define SMB3_PRODUCT_BUILD 50
-#define CIFS_VERSION   "2.50"
+#define SMB3_PRODUCT_BUILD 51
+#define CIFS_VERSION   "2.51"
 #endif				/* _CIFSFS_H */

From 665db14d0712ac27f6a0081510bd811efb3faa3c Mon Sep 17 00:00:00 2001
From: David Howells <dhowells@redhat.com>
Date: Tue, 17 Sep 2024 08:54:28 +0100
Subject: [PATCH 13/13] netfs, cifs: Fix mtime/ctime update for mmapped writes

The cifs flag CIFS_INO_MODIFIED_ATTR, which indicates that the mtime and
ctime need to be written back on close, got taken over by netfs as
NETFS_ICTX_MODIFIED_ATTR to avoid the need to call a function pointer to
set it.

The flag gets set correctly on buffered writes, but doesn't get set by
netfs_page_mkwrite(), leading to occasional failures in generic/080 and
generic/215.

Fix this by setting the flag in netfs_page_mkwrite().

Fixes: 73425800ac94 ("netfs, cifs: Move CIFS_INO_MODIFIED_ATTR to netfs_inode")
Reported-by: kernel test robot <oliver.sang@intel.com>
Closes: https://lore.kernel.org/oe-lkp/202409161629.98887b2-oliver.sang@intel.com
Signed-off-by: David Howells <dhowells@redhat.com>
Reviewed-by: Paulo Alcantara (Red Hat) <pc@manguebit.com>
cc: Jeff Layton <jlayton@kernel.org>
cc: linux-cifs@vger.kernel.org
cc: netfs@lists.linux.dev
cc: linux-fsdevel@vger.kernel.org
Signed-off-by: Steve French <stfrench@microsoft.com>
---
 fs/netfs/buffered_write.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/fs/netfs/buffered_write.c b/fs/netfs/buffered_write.c
index d7eae597e54ded..b3910dfcb56d3e 100644
--- a/fs/netfs/buffered_write.c
+++ b/fs/netfs/buffered_write.c
@@ -552,6 +552,7 @@ vm_fault_t netfs_page_mkwrite(struct vm_fault *vmf, struct netfs_group *netfs_gr
 		trace_netfs_folio(folio, netfs_folio_trace_mkwrite);
 	netfs_set_group(folio, netfs_group);
 	file_update_time(file);
+	set_bit(NETFS_ICTX_MODIFIED_ATTR, &ictx->flags);
 	if (ictx->ops->post_modify)
 		ictx->ops->post_modify(inode);
 	ret = VM_FAULT_LOCKED;