CIFS: Properly process SMB3 lease breaks

[ Upstream commit 9bd45408 ]

Currenly we doesn't assume that a server may break a lease
from RWH to RW which causes us setting a wrong lease state
on a file and thus mistakenly flushing data and byte-range
locks and purging cached data on the client. This leads to
performance degradation because subsequent IOs go directly
to the server.

Fix this by propagating new lease state and epoch values
to the oplock break handler through cifsFileInfo structure
and removing the use of cifsInodeInfo flags for that. It
allows to avoid some races of several lease/oplock breaks
using those flags in parallel.
Signed-off-by: Pavel Shilovsky <pshilov@microsoft.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

fs/cifs/cifsglob.h

@@ -242,8 +242,9 @@ struct smb_version_operations {
 	int (*check_message)(char *, unsigned int, struct TCP_Server_Info *);
 	bool (*is_oplock_break)(char *, struct TCP_Server_Info *);
 	int (*handle_cancelled_mid)(char *, struct TCP_Server_Info *);
-	void (*downgrade_oplock)(struct TCP_Server_Info *,
-					struct cifsInodeInfo *, bool);
+	void (*downgrade_oplock)(struct TCP_Server_Info *server,
+				 struct cifsInodeInfo *cinode, __u32 oplock,
+				 unsigned int epoch, bool *purge_cache);
 	/* process transaction2 response */
 	bool (*check_trans2)(struct mid_q_entry *, struct TCP_Server_Info *,
 			     char *, int);
@@ -1080,6 +1081,8 @@ struct cifsFileInfo {
 	unsigned int f_flags;
 	bool invalidHandle:1;	/* file closed via session abend */
 	bool oplock_break_cancelled:1;
+	unsigned int oplock_epoch; /* epoch from the lease break */
+	__u32 oplock_level; /* oplock/lease level from the lease break */
 	int count;
 	spinlock_t file_info_lock; /* protects four flag/count fields above */
 	struct mutex fh_mutex; /* prevents reopen race after dead ses*/
@@ -1191,7 +1194,7 @@ struct cifsInodeInfo {
 	unsigned int epoch;		/* used to track lease state changes */
 #define CIFS_INODE_PENDING_OPLOCK_BREAK   (0) /* oplock break in progress */
 #define CIFS_INODE_PENDING_WRITERS	  (1) /* Writes in progress */
-#define CIFS_INODE_DOWNGRADE_OPLOCK_TO_L2 (2) /* Downgrade oplock to L2 */
+#define CIFS_INODE_FLAG_UNUSED		  (2) /* Unused flag */
 #define CIFS_INO_DELETE_PENDING		  (3) /* delete pending on server */
 #define CIFS_INO_INVALID_MAPPING	  (4) /* pagecache is invalid */
 #define CIFS_INO_LOCK			  (5) /* lock bit for synchronization */

fs/cifs/file.c

@@ -3912,12 +3912,13 @@ void cifs_oplock_break(struct work_struct *work)
 	struct cifs_tcon *tcon = tlink_tcon(cfile->tlink);
 	struct TCP_Server_Info *server = tcon->ses->server;
 	int rc = 0;
+	bool purge_cache = false;
 
 	wait_on_bit(&cinode->flags, CIFS_INODE_PENDING_WRITERS,
 			TASK_UNINTERRUPTIBLE);
 
-	server->ops->downgrade_oplock(server, cinode,
-		test_bit(CIFS_INODE_DOWNGRADE_OPLOCK_TO_L2, &cinode->flags));
+	server->ops->downgrade_oplock(server, cinode, cfile->oplock_level,
+				      cfile->oplock_epoch, &purge_cache);
 
 	if (!CIFS_CACHE_WRITE(cinode) && CIFS_CACHE_READ(cinode) &&
 						cifs_has_mand_locks(cinode)) {
@@ -3932,18 +3933,21 @@ void cifs_oplock_break(struct work_struct *work)
 		else
 			break_lease(inode, O_WRONLY);
 		rc = filemap_fdatawrite(inode->i_mapping);
-		if (!CIFS_CACHE_READ(cinode)) {
+		if (!CIFS_CACHE_READ(cinode) || purge_cache) {
 			rc = filemap_fdatawait(inode->i_mapping);
 			mapping_set_error(inode->i_mapping, rc);
 			cifs_zap_mapping(inode);
 		}
 		cifs_dbg(FYI, "Oplock flush inode %p rc %d\n", inode, rc);
+		if (CIFS_CACHE_WRITE(cinode))
+			goto oplock_break_ack;
 	}
 
 	rc = cifs_push_locks(cfile);
 	if (rc)
 		cifs_dbg(VFS, "Push locks rc = %d\n", rc);
 
+oplock_break_ack:
 	/*
 	 * releasing stale oplock after recent reconnect of smb session using
 	 * a now incorrect file handle is not a data integrity issue but do

fs/cifs/misc.c

@@ -481,21 +481,10 @@ is_valid_oplock_break(char *buffer, struct TCP_Server_Info *srv)
 				set_bit(CIFS_INODE_PENDING_OPLOCK_BREAK,
 					&pCifsInode->flags);
 
-				/*
-				 * Set flag if the server downgrades the oplock
-				 * to L2 else clear.
-				 */
-				if (pSMB->OplockLevel)
-					set_bit(
-					   CIFS_INODE_DOWNGRADE_OPLOCK_TO_L2,
-					   &pCifsInode->flags);
-				else
-					clear_bit(
-					   CIFS_INODE_DOWNGRADE_OPLOCK_TO_L2,
-					   &pCifsInode->flags);
-
-				cifs_queue_oplock_break(netfile);
+				netfile->oplock_epoch = 0;
+				netfile->oplock_level = pSMB->OplockLevel;
 				netfile->oplock_break_cancelled = false;
+				cifs_queue_oplock_break(netfile);
 
 				spin_unlock(&tcon->open_file_lock);
 				spin_unlock(&cifs_tcp_ses_lock);

fs/cifs/smb1ops.c

@@ -378,12 +378,10 @@ coalesce_t2(char *second_buf, struct smb_hdr *target_hdr)
 
 static void
 cifs_downgrade_oplock(struct TCP_Server_Info *server,
-			struct cifsInodeInfo *cinode, bool set_level2)
+		      struct cifsInodeInfo *cinode, __u32 oplock,
+		      unsigned int epoch, bool *purge_cache)
 {
-	if (set_level2)
-		cifs_set_oplock_level(cinode, OPLOCK_READ);
-	else
-		cifs_set_oplock_level(cinode, 0);
+	cifs_set_oplock_level(cinode, oplock);
 }
 
 static bool

fs/cifs/smb2misc.c

@@ -491,7 +491,7 @@ smb2_tcon_has_lease(struct cifs_tcon *tcon, struct smb2_lease_break *rsp,
 
 		cifs_dbg(FYI, "found in the open list\n");
 		cifs_dbg(FYI, "lease key match, lease break 0x%x\n",
-			 le32_to_cpu(rsp->NewLeaseState));
+			 lease_state);
 
 		if (ack_req)
 			cfile->oplock_break_cancelled = false;
@@ -500,17 +500,8 @@ smb2_tcon_has_lease(struct cifs_tcon *tcon, struct smb2_lease_break *rsp,
 
 		set_bit(CIFS_INODE_PENDING_OPLOCK_BREAK, &cinode->flags);
 
-		/*
-		 * Set or clear flags depending on the lease state being READ.
-		 * HANDLE caching flag should be added when the client starts
-		 * to defer closing remote file handles with HANDLE leases.
-		 */
-		if (lease_state & SMB2_LEASE_READ_CACHING_HE)
-			set_bit(CIFS_INODE_DOWNGRADE_OPLOCK_TO_L2,
-				&cinode->flags);
-		else
-			clear_bit(CIFS_INODE_DOWNGRADE_OPLOCK_TO_L2,
-				  &cinode->flags);
+		cfile->oplock_epoch = le16_to_cpu(rsp->Epoch);
+		cfile->oplock_level = lease_state;
 
 		cifs_queue_oplock_break(cfile);
 		kfree(lw);
@@ -533,7 +524,7 @@ smb2_tcon_has_lease(struct cifs_tcon *tcon, struct smb2_lease_break *rsp,
 
 		cifs_dbg(FYI, "found in the pending open list\n");
 		cifs_dbg(FYI, "lease key match, lease break 0x%x\n",
-			 le32_to_cpu(rsp->NewLeaseState));
+			 lease_state);
 
 		open->oplock = lease_state;
 	}
@@ -645,18 +636,9 @@ smb2_is_valid_oplock_break(char *buffer, struct TCP_Server_Info *server)
 				set_bit(CIFS_INODE_PENDING_OPLOCK_BREAK,
 					&cinode->flags);
 
-				/*
-				 * Set flag if the server downgrades the oplock
-				 * to L2 else clear.
-				 */
-				if (rsp->OplockLevel)
-					set_bit(
-					   CIFS_INODE_DOWNGRADE_OPLOCK_TO_L2,
-					   &cinode->flags);
-				else
-					clear_bit(
-					   CIFS_INODE_DOWNGRADE_OPLOCK_TO_L2,
-					   &cinode->flags);
+				cfile->oplock_epoch = 0;
+				cfile->oplock_level = rsp->OplockLevel;
+
 				spin_unlock(&cfile->file_info_lock);
 
 				cifs_queue_oplock_break(cfile);

fs/cifs/smb2ops.c

@@ -1379,22 +1379,38 @@ static long smb3_fallocate(struct file *file, struct cifs_tcon *tcon, int mode,
 
 static void
 smb2_downgrade_oplock(struct TCP_Server_Info *server,
-			struct cifsInodeInfo *cinode, bool set_level2)
+		      struct cifsInodeInfo *cinode, __u32 oplock,
+		      unsigned int epoch, bool *purge_cache)
 {
-	if (set_level2)
-		server->ops->set_oplock_level(cinode, SMB2_OPLOCK_LEVEL_II,
-						0, NULL);
-	else
-		server->ops->set_oplock_level(cinode, 0, 0, NULL);
+	server->ops->set_oplock_level(cinode, oplock, 0, NULL);
 }
 
 static void
-smb21_downgrade_oplock(struct TCP_Server_Info *server,
-		       struct cifsInodeInfo *cinode, bool set_level2)
+smb21_set_oplock_level(struct cifsInodeInfo *cinode, __u32 oplock,
+		       unsigned int epoch, bool *purge_cache);
+
+static void
+smb3_downgrade_oplock(struct TCP_Server_Info *server,
+		       struct cifsInodeInfo *cinode, __u32 oplock,
+		       unsigned int epoch, bool *purge_cache)
 {
-	server->ops->set_oplock_level(cinode,
-				      set_level2 ? SMB2_LEASE_READ_CACHING_HE :
-				      0, 0, NULL);
+	unsigned int old_state = cinode->oplock;
+	unsigned int old_epoch = cinode->epoch;
+	unsigned int new_state;
+
+	if (epoch > old_epoch) {
+		smb21_set_oplock_level(cinode, oplock, 0, NULL);
+		cinode->epoch = epoch;
+	}
+
+	new_state = cinode->oplock;
+	*purge_cache = false;
+
+	if ((old_state & CIFS_CACHE_READ_FLG) != 0 &&
+	    (new_state & CIFS_CACHE_READ_FLG) == 0)
+		*purge_cache = true;
+	else if (old_state == new_state && (epoch - old_epoch > 1))
+		*purge_cache = true;
 }
 
 static void
@@ -1709,7 +1725,7 @@ struct smb_version_operations smb21_operations = {
 	.print_stats = smb2_print_stats,
 	.is_oplock_break = smb2_is_valid_oplock_break,
 	.handle_cancelled_mid = smb2_handle_cancelled_mid,
-	.downgrade_oplock = smb21_downgrade_oplock,
+	.downgrade_oplock = smb2_downgrade_oplock,
 	.need_neg = smb2_need_neg,
 	.negotiate = smb2_negotiate,
 	.negotiate_wsize = smb2_negotiate_wsize,
@@ -1793,7 +1809,7 @@ struct smb_version_operations smb30_operations = {
 	.dump_share_caps = smb2_dump_share_caps,
 	.is_oplock_break = smb2_is_valid_oplock_break,
 	.handle_cancelled_mid = smb2_handle_cancelled_mid,
-	.downgrade_oplock = smb21_downgrade_oplock,
+	.downgrade_oplock = smb3_downgrade_oplock,
 	.need_neg = smb2_need_neg,
 	.negotiate = smb2_negotiate,
 	.negotiate_wsize = smb2_negotiate_wsize,
@@ -1883,7 +1899,7 @@ struct smb_version_operations smb311_operations = {
 	.dump_share_caps = smb2_dump_share_caps,
 	.is_oplock_break = smb2_is_valid_oplock_break,
 	.handle_cancelled_mid = smb2_handle_cancelled_mid,
-	.downgrade_oplock = smb21_downgrade_oplock,
+	.downgrade_oplock = smb3_downgrade_oplock,
 	.need_neg = smb2_need_neg,
 	.negotiate = smb2_negotiate,
 	.negotiate_wsize = smb2_negotiate_wsize,

fs/cifs/smb2pdu.h

@@ -1025,7 +1025,7 @@ struct smb2_oplock_break {
 struct smb2_lease_break {
 	struct smb2_hdr hdr;
 	__le16 StructureSize; /* Must be 44 */
-	__le16 Reserved;
+	__le16 Epoch;
 	__le32 Flags;
 	__u8   LeaseKey[16];
 	__le32 CurrentLeaseState;