Skip to contentSkip to conversation

cifs: ignore ipc reconnect failures during dfs failover

PATCH In review Paulo Alcantara 3 participants 6 messages Dec 29, 2022 – Jan 4, 2023 2 messages not in the archive

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/2 cifs: ignore ipc reconnect failures during dfs failover

If it failed to reconnect ipc used for getting referrals, we can just ignore it as it is not required for reconnecting the share. The worst case would be not being able to detect or chase nested links as long as dfs root server is unreachable.

Before patch:

  $ mount.cifs //root/dfs/link /mnt -o echo_interval=10,...
    -> target share: /fs0/share

disconnect root & fs0

  $ ls /mnt
  ls: cannot access '/mnt': Host is down

connect fs0

  $ ls /mnt
  ls: cannot access '/mnt': Resource temporarily unavailable

After patch:

  $ mount.cifs //root/dfs/link /mnt -o echo_interval=10,...
    -> target share: /fs0/share

disconnect root & fs0

  $ ls /mnt
  ls: cannot access '/mnt': Host is down

connect fs0

  $ ls /mnt
  bar.rtf  dir1  foo
---
 fs/cifs/dfs.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
fs/cifs/dfs.c +1 -2
index b541e68378f6..30086f2060a1 100644
@@ -401,8 +401,7 @@ static int __tree_connect_dfs_target(const unsigned int xid, struct cifs_tcon *t
 		if (ipc->need_reconnect) {
 			scnprintf(tree, MAX_TREE_SIZE, "\\\\%s\\IPC$", server->hostname);
 			rc = ops->tree_connect(xid, ipc->ses, tree, ipc, cifs_sb->local_nls);
-			if (rc)
-				break;
+			cifs_dbg(FYI, "%s: reconnect ipc: %d\n", __func__, rc);
 		}
 
 		scnprintf(tree, MAX_TREE_SIZE, "\\%s", share);
Signature · 1 line
2.39.0
#1 lore raw

PATCH 2/2 cifs: fix race in assemble_neg_contexts()

Serialise access of TCP_Server_Info::hostname in assemble_neg_contexts() by holding the server's mutex otherwise it might end up accessing an already-freed hostname pointer from cifs_reconnect() or cifs_resolve_server().

---
 fs/cifs/smb2pdu.c | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)
fs/cifs/smb2pdu.c +7 -4
index a5695748a89b..2c484d47c592 100644
@@ -541,9 +541,10 @@ static void
 assemble_neg_contexts(struct smb2_negotiate_req *req,
 		      struct TCP_Server_Info *server, unsigned int *total_len)
 {
-	char *pneg_ctxt;
-	char *hostname = NULL;
 	unsigned int ctxt_len, neg_context_count;
+	struct TCP_Server_Info *pserver;
+	char *pneg_ctxt;
+	char *hostname;
 
 	if (*total_len > 200) {
 		/* In case length corrupted don't want to overrun smb buffer */
@@ -574,8 +575,9 @@ assemble_neg_contexts(struct smb2_negotiate_req *req,
 	 * secondary channels don't have the hostname field populated
 	 * use the hostname field in the primary channel instead
 	 */
-	hostname = CIFS_SERVER_IS_CHAN(server) ?
-		server->primary_server->hostname : server->hostname;
+	pserver = CIFS_SERVER_IS_CHAN(server) ? server->primary_server : server;
+	cifs_server_lock(pserver);
+	hostname = pserver->hostname;
 	if (hostname && (hostname[0] != 0)) {
 		ctxt_len = build_netname_ctxt((struct smb2_netname_neg_context *)pneg_ctxt,
 					      hostname);
@@ -584,6 +586,7 @@ assemble_neg_contexts(struct smb2_negotiate_req *req,
 		neg_context_count = 3;
 	} else
 		neg_context_count = 2;
+	cifs_server_unlock(pserver);
 
 	build_posix_ctxt((struct smb2_posix_neg_context *)pneg_ctxt);
 	*total_len += sizeof(struct smb2_posix_neg_context);
Signature · 1 line
2.39.0
Reviewed-by on v1 2/2
#2 lore raw

Re: [PATCH 2/2] cifs: fix race in assemble_neg_contexts()

Quoted 6 lines · Paulo Alcantara

Serialise access of TCP_Server_Info::hostname in assemble_neg_contexts() by holding the server's mutex otherwise it might end up accessing an already-freed hostname pointer from cifs_reconnect() or cifs_resolve_server().

Couldn't reproduce this one as easy as the other one, but it makes sense anyway.

Quoted 44 lines

---
fs/cifs/smb2pdu.c | 11 +++++++----

1 file changed, 7 insertions(+), 4 deletions(-)
fs/cifs/smb2pdu.c +7 -4
index a5695748a89b..2c484d47c592 100644
@@ -541,9 +541,10 @@ static void
 assemble_neg_contexts(struct smb2_negotiate_req *req,
 		      struct TCP_Server_Info *server, unsigned int *total_len)
 {
-	char *pneg_ctxt;
-	char *hostname = NULL;
 	unsigned int ctxt_len, neg_context_count;
+	struct TCP_Server_Info *pserver;
+	char *pneg_ctxt;
+	char *hostname;
 
 	if (*total_len > 200) {
 		/* In case length corrupted don't want to overrun smb buffer */
@@ -574,8 +575,9 @@ assemble_neg_contexts(struct smb2_negotiate_req *req,
 	 * secondary channels don't have the hostname field populated
 	 * use the hostname field in the primary channel instead
 	 */
-	hostname = CIFS_SERVER_IS_CHAN(server) ?
-		server->primary_server->hostname : server->hostname;
+	pserver = CIFS_SERVER_IS_CHAN(server) ? server->primary_server : server;
+	cifs_server_lock(pserver);
+	hostname = pserver->hostname;
 	if (hostname && (hostname[0] != 0)) {
 		ctxt_len = build_netname_ctxt((struct smb2_netname_neg_context *)pneg_ctxt,
 					      hostname);
@@ -584,6 +586,7 @@ assemble_neg_contexts(struct smb2_negotiate_req *req,
 		neg_context_count = 3;
 	} else
 		neg_context_count = 2;
+	cifs_server_unlock(pserver);
 
 	build_posix_ctxt((struct smb2_posix_neg_context *)pneg_ctxt);
 	*total_len += sizeof(struct smb2_posix_neg_context);
Signature · 1 line
2.39.0

A message that is not in the archive (sent off-list or not yet received): CAH2r5mso6PLpKpVxtcUHW4RQ1jc-Tmj5ALOEba2z+40uSDf0JA@mail.gmail.com

#4 lore raw

Re: [PATCH 2/2] cifs: fix race in assemble_neg_contexts()

Quoted 69 lines · Steve French

merged into cifs-2.6.git for-next

Quoted 53 lines · Enzo Matsumiya
Quoted 6 lines · Paulo Alcantara

Serialise access of TCP_Server_Info::hostname in assemble_neg_contexts() by holding the server's mutex otherwise it might end up accessing an already-freed hostname pointer from cifs_reconnect() or cifs_resolve_server().

Couldn't reproduce this one as easy as the other one, but it makes sense anyway.

Quoted 39 lines

---
fs/cifs/smb2pdu.c | 11 +++++++----

1 file changed, 7 insertions(+), 4 deletions(-)
fs/cifs/smb2pdu.c +7 -4
index a5695748a89b..2c484d47c592 100644
@@ -541,9 +541,10 @@ static void
 assemble_neg_contexts(struct smb2_negotiate_req *req,
                    struct TCP_Server_Info *server, unsigned int *total_len)
 {
-      char *pneg_ctxt;
-      char *hostname = NULL;
      unsigned int ctxt_len, neg_context_count;
+      struct TCP_Server_Info *pserver;
+      char *pneg_ctxt;
+      char *hostname;
 
      if (*total_len > 200) {
              /* In case length corrupted don't want to overrun smb buffer */
@@ -574,8 +575,9 @@ assemble_neg_contexts(struct smb2_negotiate_req *req,
       * secondary channels don't have the hostname field populated
       * use the hostname field in the primary channel instead
       */
-      hostname = CIFS_SERVER_IS_CHAN(server) ?
-              server->primary_server->hostname : server->hostname;
+      pserver = CIFS_SERVER_IS_CHAN(server) ? server->primary_server : server;
+      cifs_server_lock(pserver);
+      hostname = pserver->hostname;
      if (hostname && (hostname[0] != 0)) {
              ctxt_len = build_netname_ctxt((struct smb2_netname_neg_context *)pneg_ctxt,
                                            hostname);
@@ -584,6 +586,7 @@ assemble_neg_contexts(struct smb2_negotiate_req *req,
              neg_context_count = 3;
      } else
              neg_context_count = 2;
+      cifs_server_unlock(pserver);
 
Quoted 5 lines
Quoted 5 lines
fs/cifs/smb2pdu.c +0 -1
      build_posix_ctxt((struct smb2_posix_neg_context *)pneg_ctxt);
      *total_len += sizeof(struct smb2_posix_neg_context);
--

2.39.0

--
Thanks,

Steve

Signature · 3 lines
Thanks,

Steve
Reviewed-by on v1 1/2
#5 lore raw
In reply to Paulo Alcantara
Quoted 30 lines · Paulo Alcantara

If it failed to reconnect ipc used for getting referrals, we can just ignore it as it is not required for reconnecting the share. The worst case would be not being able to detect or chase nested links as long as dfs root server is unreachable.

Before patch:

 $ mount.cifs //root/dfs/link /mnt -o echo_interval=10,...
   -> target share: /fs0/share

disconnect root & fs0

 $ ls /mnt
 ls: cannot access '/mnt': Host is down

connect fs0

 $ ls /mnt
 ls: cannot access '/mnt': Resource temporarily unavailable

After patch:

 $ mount.cifs //root/dfs/link /mnt -o echo_interval=10,...
   -> target share: /fs0/share

disconnect root & fs0

 $ ls /mnt
 ls: cannot access '/mnt': Host is down
Quoted 6 lines

connect fs0

 $ ls /mnt
 bar.rtf  dir1  foo
Quoted 21 lines

---
fs/cifs/dfs.c | 3 +--

1 file changed, 1 insertion(+), 2 deletions(-)
fs/cifs/dfs.c +1 -2
index b541e68378f6..30086f2060a1 100644
@@ -401,8 +401,7 @@ static int __tree_connect_dfs_target(const unsigned int xid, struct cifs_tcon *t
 		if (ipc->need_reconnect) {
 			scnprintf(tree, MAX_TREE_SIZE, "\\\\%s\\IPC$", server->hostname);
 			rc = ops->tree_connect(xid, ipc->ses, tree, ipc, cifs_sb->local_nls);
-			if (rc)
-				break;
+			cifs_dbg(FYI, "%s: reconnect ipc: %d\n", __func__, rc);
 		}
 
 		scnprintf(tree, MAX_TREE_SIZE, "\\%s", share);
Signature · 1 line
2.39.0

A message that is not in the archive (sent off-list or not yet received): CAH2r5mv-2MPzd-zJSxDXh5avC4Bhp-BJG9nmr2f=1FR5m6B3Zg@mail.gmail.com

#7 lore raw
Quoted 73 lines · Steve French

merged into cifs-2.6.git for-next

Quoted 58 lines · Enzo Matsumiya
Quoted 30 lines · Paulo Alcantara

If it failed to reconnect ipc used for getting referrals, we can just ignore it as it is not required for reconnecting the share. The worst case would be not being able to detect or chase nested links as long as dfs root server is unreachable.

Before patch:

 $ mount.cifs //root/dfs/link /mnt -o echo_interval=10,...
   -> target share: /fs0/share

disconnect root & fs0

 $ ls /mnt
 ls: cannot access '/mnt': Host is down

connect fs0

 $ ls /mnt
 ls: cannot access '/mnt': Resource temporarily unavailable

After patch:

 $ mount.cifs //root/dfs/link /mnt -o echo_interval=10,...
   -> target share: /fs0/share

disconnect root & fs0

 $ ls /mnt
 ls: cannot access '/mnt': Host is down
Quoted 6 lines

connect fs0

 $ ls /mnt
 bar.rtf  dir1  foo
Quoted 17 lines

---
fs/cifs/dfs.c | 3 +--

1 file changed, 1 insertion(+), 2 deletions(-)
fs/cifs/dfs.c +1 -2
index b541e68378f6..30086f2060a1 100644
@@ -401,8 +401,7 @@ static int __tree_connect_dfs_target(const unsigned int xid, struct cifs_tcon *t
              if (ipc->need_reconnect) {
                      scnprintf(tree, MAX_TREE_SIZE, "\\\\%s\\IPC$", server->hostname);
                      rc = ops->tree_connect(xid, ipc->ses, tree, ipc, cifs_sb->local_nls);
-                      if (rc)
-                              break;
+                      cifs_dbg(FYI, "%s: reconnect ipc: %d\n", __func__, rc);
              }
 
Quoted 4 lines
Quoted 4 lines
fs/cifs/dfs.c +0 -1
              scnprintf(tree, MAX_TREE_SIZE, "\\%s", share);
--

2.39.0

--
Thanks,

Steve

Signature · 3 lines
Thanks,

Steve