Merge branch 'v4-0-test' of ssh://git.samba.org/data/git/samba into v4-0-trivial

(This used to be commit d96fcc52a70fd1236b767b1d88a39d78c4e8c48c)

16 files changed, 1166 insertions, 293 deletions

diff --git a/source4/ntvfs/common/opendb.c b/source4/ntvfs/common/opendb.c
index 4826ca5c26..36144d0406 100644
--- a/source4/ntvfs/common/opendb.c
+++ b/source4/ntvfs/common/opendb.c
@@ -81,6 +81,10 @@ _PUBLIC_ struct odb_lock *odb_lock(TALLOC_CTX *mem_ctx,
 	return ops->odb_lock(mem_ctx, odb, file_key);
 }
 
+_PUBLIC_ DATA_BLOB odb_get_key(TALLOC_CTX *mem_ctx, struct odb_lock *lck)
+{
+	return ops->odb_get_key(mem_ctx, lck);
+}
 
 /*
   register an open file in the open files database. This implements the share_access
@@ -89,15 +93,16 @@ _PUBLIC_ struct odb_lock *odb_lock(TALLOC_CTX *mem_ctx,
   Note that the path is only used by the delete on close logic, not
   for comparing with other filenames
 */
-_PUBLIC_ NTSTATUS odb_open_file(struct odb_lock *lck, void *file_handle,
-				uint32_t stream_id, uint32_t share_access, 
+_PUBLIC_ NTSTATUS odb_open_file(struct odb_lock *lck,
+				void *file_handle, const char *path,
+				uint32_t stream_id, uint32_t share_access,
 				uint32_t access_mask, bool delete_on_close,
-				const char *path, 
+				uint32_t open_disposition, bool break_to_none,
 				uint32_t oplock_level, uint32_t *oplock_granted)
 {
-	return ops->odb_open_file(lck, file_handle, stream_id, share_access,
-				  access_mask, delete_on_close, path, oplock_level,
-				  oplock_granted);
+	return ops->odb_open_file(lck, file_handle, path, stream_id, share_access,
+				  access_mask, delete_on_close, open_disposition,
+				  break_to_none, oplock_level, oplock_granted);
 }
 
 
@@ -161,8 +166,21 @@ _PUBLIC_ NTSTATUS odb_get_delete_on_close(struct odb_context *odb,
   create_options and access_mask
 */
 _PUBLIC_ NTSTATUS odb_can_open(struct odb_lock *lck,
-			       uint32_t share_access, uint32_t create_options, 
-			       uint32_t access_mask)
+			       uint32_t stream_id, uint32_t share_access,
+			       uint32_t access_mask, bool delete_on_close,
+			       uint32_t open_disposition, bool break_to_none)
+{
+	return ops->odb_can_open(lck, stream_id, share_access, access_mask,
+				 delete_on_close, open_disposition, break_to_none);
+}
+
+_PUBLIC_ NTSTATUS odb_update_oplock(struct odb_lock *lck, void *file_handle,
+				    uint32_t oplock_level)
+{
+	return ops->odb_update_oplock(lck, file_handle, oplock_level);
+}
+
+_PUBLIC_ NTSTATUS odb_break_oplocks(struct odb_lock *lck)
 {
-	return ops->odb_can_open(lck, share_access, create_options, access_mask);
+	return ops->odb_break_oplocks(lck);
 }
diff --git a/source4/ntvfs/common/opendb.h b/source4/ntvfs/common/opendb.h
index 231ae3d7de..9591bcf6b9 100644
--- a/source4/ntvfs/common/opendb.h
+++ b/source4/ntvfs/common/opendb.h
@@ -24,10 +24,12 @@ struct opendb_ops {
 					struct ntvfs_context *ntvfs_ctx);
 	struct odb_lock *(*odb_lock)(TALLOC_CTX *mem_ctx,
 				     struct odb_context *odb, DATA_BLOB *file_key);
-	NTSTATUS (*odb_open_file)(struct odb_lock *lck, void *file_handle,
-				  uint32_t stream_id, uint32_t share_access, 
+	DATA_BLOB (*odb_get_key)(TALLOC_CTX *mem_ctx, struct odb_lock *lck);
+	NTSTATUS (*odb_open_file)(struct odb_lock *lck,
+				  void *file_handle, const char *path,
+				  uint32_t stream_id, uint32_t share_access,
 				  uint32_t access_mask, bool delete_on_close,
-				  const char *path, 
+				  uint32_t open_disposition, bool break_to_none,
 				  uint32_t oplock_level, uint32_t *oplock_granted);
 	NTSTATUS (*odb_open_file_pending)(struct odb_lock *lck, void *private);
 	NTSTATUS (*odb_close_file)(struct odb_lock *lck, void *file_handle);
@@ -38,10 +40,18 @@ struct opendb_ops {
 					    DATA_BLOB *key, bool *del_on_close, 
 					    int *open_count, char **path);
 	NTSTATUS (*odb_can_open)(struct odb_lock *lck,
-				 uint32_t share_access, uint32_t create_options, 
-				 uint32_t access_mask);
+				 uint32_t stream_id, uint32_t share_access,
+				 uint32_t access_mask, bool delete_on_close,
+				 uint32_t open_disposition, bool break_to_none);
+	NTSTATUS (*odb_update_oplock)(struct odb_lock *lck, void *file_handle,
+				      uint32_t oplock_level);
+	NTSTATUS (*odb_break_oplocks)(struct odb_lock *lck);
 };
 
+struct opendb_oplock_break {
+	void *file_handle;
+	uint8_t level;
+};
 
 void odb_set_ops(const struct opendb_ops *new_ops);
 void odb_tdb_init_ops(void);
diff --git a/source4/ntvfs/common/opendb_tdb.c b/source4/ntvfs/common/opendb_tdb.c
index abd9ca708b..fe5a0a8864 100644
--- a/source4/ntvfs/common/opendb_tdb.c
+++ b/source4/ntvfs/common/opendb_tdb.c
@@ -2,7 +2,8 @@
    Unix SMB/CIFS implementation.
 
    Copyright (C) Andrew Tridgell 2004
-   
+   Copyright (C) Stefan Metzmacher 2008
+
    This program is free software; you can redistribute it and/or modify
    it under the terms of the GNU General Public License as published by
    the Free Software Foundation; either version 3 of the License, or
@@ -134,12 +135,21 @@ static struct odb_lock *odb_tdb_lock(TALLOC_CTX *mem_ctx,
 	return lck;
 }
 
+static DATA_BLOB odb_tdb_get_key(TALLOC_CTX *mem_ctx, struct odb_lock *lck)
+{
+	return data_blob_talloc(mem_ctx, lck->key.dptr, lck->key.dsize);
+}
+
+
 /*
   determine if two odb_entry structures conflict
 
   return NT_STATUS_OK on no conflict
 */
-static NTSTATUS share_conflict(struct opendb_entry *e1, struct opendb_entry *e2)
+static NTSTATUS share_conflict(struct opendb_entry *e1,
+			       uint32_t stream_id,
+			       uint32_t share_access,
+			       uint32_t access_mask)
 {
 	/* if either open involves no read.write or delete access then
 	   it can't conflict */
@@ -150,18 +160,18 @@ static NTSTATUS share_conflict(struct opendb_entry *e1, struct opendb_entry *e2)
 				 SEC_STD_DELETE))) {
 		return NT_STATUS_OK;
 	}
-	if (!(e2->access_mask & (SEC_FILE_WRITE_DATA |
-				 SEC_FILE_APPEND_DATA |
-				 SEC_FILE_READ_DATA |
-				 SEC_FILE_EXECUTE |
-				 SEC_STD_DELETE))) {
+	if (!(access_mask & (SEC_FILE_WRITE_DATA |
+			     SEC_FILE_APPEND_DATA |
+			     SEC_FILE_READ_DATA |
+			     SEC_FILE_EXECUTE |
+			     SEC_STD_DELETE))) {
 		return NT_STATUS_OK;
 	}
 
 	/* data IO access masks. This is skipped if the two open handles
 	   are on different streams (as in that case the masks don't
 	   interact) */
-	if (e1->stream_id != e2->stream_id) {
+	if (e1->stream_id != stream_id) {
 		return NT_STATUS_OK;
 	}
 
@@ -169,20 +179,20 @@ static NTSTATUS share_conflict(struct opendb_entry *e1, struct opendb_entry *e2)
 	if (((am) & (right)) && !((sa) & (share))) return NT_STATUS_SHARING_VIOLATION
 
 	CHECK_MASK(e1->access_mask, SEC_FILE_WRITE_DATA | SEC_FILE_APPEND_DATA,
-		   e2->share_access, NTCREATEX_SHARE_ACCESS_WRITE);
-	CHECK_MASK(e2->access_mask, SEC_FILE_WRITE_DATA | SEC_FILE_APPEND_DATA,
+		   share_access, NTCREATEX_SHARE_ACCESS_WRITE);
+	CHECK_MASK(access_mask, SEC_FILE_WRITE_DATA | SEC_FILE_APPEND_DATA,
 		   e1->share_access, NTCREATEX_SHARE_ACCESS_WRITE);
 	
 	CHECK_MASK(e1->access_mask, SEC_FILE_READ_DATA | SEC_FILE_EXECUTE,
-		   e2->share_access, NTCREATEX_SHARE_ACCESS_READ);
-	CHECK_MASK(e2->access_mask, SEC_FILE_READ_DATA | SEC_FILE_EXECUTE,
+		   share_access, NTCREATEX_SHARE_ACCESS_READ);
+	CHECK_MASK(access_mask, SEC_FILE_READ_DATA | SEC_FILE_EXECUTE,
 		   e1->share_access, NTCREATEX_SHARE_ACCESS_READ);
 
 	CHECK_MASK(e1->access_mask, SEC_STD_DELETE,
-		   e2->share_access, NTCREATEX_SHARE_ACCESS_DELETE);
-	CHECK_MASK(e2->access_mask, SEC_STD_DELETE,
+		   share_access, NTCREATEX_SHARE_ACCESS_DELETE);
+	CHECK_MASK(access_mask, SEC_STD_DELETE,
 		   e1->share_access, NTCREATEX_SHARE_ACCESS_DELETE);
-
+#undef CHECK_MASK
 	return NT_STATUS_OK;
 }
 
@@ -253,12 +263,129 @@ static NTSTATUS odb_push_record(struct odb_lock *lck, struct opendb_file *file)
 /*
   send an oplock break to a client
 */
-static NTSTATUS odb_oplock_break_send(struct odb_context *odb, struct opendb_entry *e)
+static NTSTATUS odb_oplock_break_send(struct odb_context *odb,
+				      struct opendb_entry *e,
+				      uint8_t level)
 {
+	NTSTATUS status;
+	struct opendb_oplock_break op_break;
+	DATA_BLOB blob;
+
+	ZERO_STRUCT(op_break);
+
 	/* tell the server handling this open file about the need to send the client
 	   a break */
-	return messaging_send_ptr(odb->ntvfs_ctx->msg_ctx, e->server, 
-				  MSG_NTVFS_OPLOCK_BREAK, e->file_handle);
+	op_break.file_handle	= e->file_handle;
+	op_break.level		= level;
+
+	blob = data_blob_const(&op_break, sizeof(op_break));
+
+	status = messaging_send(odb->ntvfs_ctx->msg_ctx, e->server,
+				MSG_NTVFS_OPLOCK_BREAK, &blob);
+	NT_STATUS_NOT_OK_RETURN(status);
+
+	return NT_STATUS_OK;
+}
+
+static bool access_attributes_only(uint32_t access_mask,
+				   uint32_t open_disposition,
+				   bool break_to_none)
+{
+	switch (open_disposition) {
+	case NTCREATEX_DISP_SUPERSEDE:
+	case NTCREATEX_DISP_OVERWRITE_IF:
+	case NTCREATEX_DISP_OVERWRITE:
+		return false;
+	default:
+		break;
+	}
+
+	if (break_to_none) {
+		return false;
+	}
+
+#define CHECK_MASK(m,g) ((m) && (((m) & ~(g))==0) && (((m) & (g)) != 0))
+	return CHECK_MASK(access_mask,
+			  SEC_STD_SYNCHRONIZE |
+			  SEC_FILE_READ_ATTRIBUTE |
+			  SEC_FILE_WRITE_ATTRIBUTE);
+#undef CHECK_MASK
+}
+
+static NTSTATUS odb_tdb_open_can_internal(struct odb_context *odb,
+					  const struct opendb_file *file,
+					  uint32_t stream_id, uint32_t share_access,
+					  uint32_t access_mask, bool delete_on_close,
+					  uint32_t open_disposition, bool break_to_none,
+					  bool *_attrs_only)
+{
+	NTSTATUS status;
+	uint32_t i;
+	bool attrs_only = false;
+
+	/* see if anyone has an oplock, which we need to break */
+	for (i=0;i<file->num_entries;i++) {
+		if (file->entries[i].oplock_level == OPLOCK_BATCH) {
+			bool oplock_return = OPLOCK_BREAK_TO_LEVEL_II;
+			/* if this is an attribute only access
+			 * it doesn't conflict with a BACTCH oplock
+			 * but we'll not grant the oplock below
+			 */
+			attrs_only = access_attributes_only(access_mask,
+							    open_disposition,
+							    break_to_none);
+			if (attrs_only) {
+				break;
+			}
+			/* a batch oplock caches close calls, which
+			   means the client application might have
+			   already closed the file. We have to allow
+			   this close to propogate by sending a oplock
+			   break request and suspending this call
+			   until the break is acknowledged or the file
+			   is closed */
+			if (break_to_none) {
+				oplock_return = OPLOCK_BREAK_TO_NONE;
+			}
+			odb_oplock_break_send(odb, &file->entries[i],
+					      oplock_return);
+			return NT_STATUS_OPLOCK_NOT_GRANTED;
+		}
+	}
+
+	if (file->delete_on_close) {
+		/* while delete on close is set, no new opens are allowed */
+		return NT_STATUS_DELETE_PENDING;
+	}
+
+	if (file->num_entries != 0 && delete_on_close) {
+		return NT_STATUS_SHARING_VIOLATION;
+	}
+
+	/* check for sharing violations */
+	for (i=0;i<file->num_entries;i++) {
+		status = share_conflict(&file->entries[i], stream_id,
+					share_access, access_mask);
+		NT_STATUS_NOT_OK_RETURN(status);
+	}
+
+	/* we now know the open could succeed, but we need to check
+	   for any exclusive oplocks. We can't grant a second open
+	   till these are broken. Note that we check for batch oplocks
+	   before checking for sharing violations, and check for
+	   exclusive oplocks afterwards. */
+	for (i=0;i<file->num_entries;i++) {
+		if (file->entries[i].oplock_level == OPLOCK_EXCLUSIVE) {
+			odb_oplock_break_send(odb, &file->entries[i],
+					      OPLOCK_BREAK_TO_NONE);
+			return NT_STATUS_OPLOCK_NOT_GRANTED;
+		}
+	}
+
+	if (_attrs_only) {
+		*_attrs_only = attrs_only;
+	}
+	return NT_STATUS_OK;
 }
 
 /*
@@ -268,17 +395,18 @@ static NTSTATUS odb_oplock_break_send(struct odb_context *odb, struct opendb_ent
   Note that the path is only used by the delete on close logic, not
   for comparing with other filenames
 */
-static NTSTATUS odb_tdb_open_file(struct odb_lock *lck, void *file_handle,
-				  uint32_t stream_id, uint32_t share_access, 
+static NTSTATUS odb_tdb_open_file(struct odb_lock *lck,
+				  void *file_handle, const char *path,
+				  uint32_t stream_id, uint32_t share_access,
 				  uint32_t access_mask, bool delete_on_close,
-				  const char *path, 
+				  uint32_t open_disposition, bool break_to_none,
 				  uint32_t oplock_level, uint32_t *oplock_granted)
 {
 	struct odb_context *odb = lck->odb;
 	struct opendb_entry e;
-	int i;
 	struct opendb_file file;
 	NTSTATUS status;
+	bool attrs_only = false;
 
 	if (odb->oplocks == false) {
 		oplock_level = OPLOCK_NONE;
@@ -294,6 +422,13 @@ static NTSTATUS odb_tdb_open_file(struct odb_lock *lck, void *file_handle,
 	}
 
 	/* see if it conflicts */
+	status = odb_tdb_open_can_internal(odb, &file, stream_id,
+					   share_access, access_mask,
+					   delete_on_close, open_disposition,
+					   break_to_none, &attrs_only);
+	NT_STATUS_NOT_OK_RETURN(status);
+
+	/* see if it conflicts */
 	e.server          = odb->ntvfs_ctx->server_id;
 	e.file_handle     = file_handle;
 	e.stream_id       = stream_id;
@@ -301,59 +436,37 @@ static NTSTATUS odb_tdb_open_file(struct odb_lock *lck, void *file_handle,
 	e.access_mask     = access_mask;
 	e.delete_on_close = delete_on_close;
 	e.oplock_level    = OPLOCK_NONE;
-		
-	/* see if anyone has an oplock, which we need to break */
-	for (i=0;i<file.num_entries;i++) {
-		if (file.entries[i].oplock_level == OPLOCK_BATCH) {
-			/* a batch oplock caches close calls, which
-			   means the client application might have
-			   already closed the file. We have to allow
-			   this close to propogate by sending a oplock
-			   break request and suspending this call
-			   until the break is acknowledged or the file
-			   is closed */
-			odb_oplock_break_send(odb, &file.entries[i]);
-			return NT_STATUS_OPLOCK_NOT_GRANTED;
-		}
-	}
-
-	if (file.delete_on_close || 
-	    (file.num_entries != 0 && delete_on_close)) {
-		/* while delete on close is set, no new opens are allowed */
-		return NT_STATUS_DELETE_PENDING;
-	}
-
-	/* check for sharing violations */
-	for (i=0;i<file.num_entries;i++) {
-		status = share_conflict(&file.entries[i], &e);
-		NT_STATUS_NOT_OK_RETURN(status);
-	}
-
-	/* we now know the open could succeed, but we need to check
-	   for any exclusive oplocks. We can't grant a second open
-	   till these are broken. Note that we check for batch oplocks
-	   before checking for sharing violations, and check for
-	   exclusive oplocks afterwards. */
-	for (i=0;i<file.num_entries;i++) {
-		if (file.entries[i].oplock_level == OPLOCK_EXCLUSIVE) {
-			odb_oplock_break_send(odb, &file.entries[i]);
-			return NT_STATUS_OPLOCK_NOT_GRANTED;
-		}
-	}
 
 	/*
-	  possibly grant an exclusive or batch oplock if this is the only client
-	  with the file open. We don't yet grant levelII oplocks.
+	  possibly grant an exclusive, batch or level2 oplock
 	*/
-	if (oplock_granted != NULL) {
-		if ((oplock_level == OPLOCK_BATCH ||
-		     oplock_level == OPLOCK_EXCLUSIVE) &&
-		    file.num_entries == 0) {
-			(*oplock_granted) = oplock_level;
+	if (oplock_granted) {
+		if (attrs_only) {
+			e.oplock_level	= OPLOCK_NONE;
+			*oplock_granted	= NO_OPLOCK_RETURN;
+		} else if (oplock_level == OPLOCK_EXCLUSIVE) {
+			if (file.num_entries == 0) {
+				e.oplock_level	= OPLOCK_EXCLUSIVE;
+				*oplock_granted	= EXCLUSIVE_OPLOCK_RETURN;
+			} else {
+				e.oplock_level	= OPLOCK_NONE;
+				*oplock_granted	= NO_OPLOCK_RETURN;
+			}
+		} else if (oplock_level == OPLOCK_BATCH) {
+			if (file.num_entries == 0) {
+				e.oplock_level	= OPLOCK_BATCH;
+				*oplock_granted	= BATCH_OPLOCK_RETURN;
+			} else {
+				e.oplock_level	= OPLOCK_LEVEL_II;
+				*oplock_granted	= LEVEL_II_OPLOCK_RETURN;
+			}
+		} else if (oplock_level == OPLOCK_LEVEL_II) {
+			e.oplock_level	= OPLOCK_LEVEL_II;
+			*oplock_granted	= LEVEL_II_OPLOCK_RETURN;
 		} else {
-			(*oplock_granted) = OPLOCK_NONE;
+			e.oplock_level	= OPLOCK_NONE;
+			*oplock_granted	= NO_OPLOCK_RETURN;
 		}
-		e.oplock_level = (*oplock_granted);
 	}
 
 	/* it doesn't conflict, so add it to the end */
@@ -439,6 +552,82 @@ static NTSTATUS odb_tdb_close_file(struct odb_lock *lck, void *file_handle)
 	return odb_push_record(lck, &file);
 }
 
+/*
+  update the oplock level of the client
+*/
+static NTSTATUS odb_tdb_update_oplock(struct odb_lock *lck, void *file_handle,
+				      uint32_t oplock_level)
+{
+	struct odb_context *odb = lck->odb;
+	struct opendb_file file;
+	int i;
+	NTSTATUS status;
+
+	status = odb_pull_record(lck, &file);
+	NT_STATUS_NOT_OK_RETURN(status);
+
+	/* find the entry, and update it */
+	for (i=0;i<file.num_entries;i++) {
+		if (file_handle == file.entries[i].file_handle &&
+		    cluster_id_equal(&odb->ntvfs_ctx->server_id, &file.entries[i].server)) {
+			file.entries[i].oplock_level = oplock_level;
+			break;
+		}
+	}
+
+	if (i == file.num_entries) {
+		return NT_STATUS_UNSUCCESSFUL;
+	}
+
+	/* send any pending notifications, removing them once sent */
+	for (i=0;i<file.num_pending;i++) {
+		messaging_send_ptr(odb->ntvfs_ctx->msg_ctx,
+				   file.pending[i].server,
+				   MSG_PVFS_RETRY_OPEN,
+				   file.pending[i].notify_ptr);
+	}
+	file.num_pending = 0;
+
+	return odb_push_record(lck, &file);
+}
+
+/*
+  send oplocks breaks to none to all level2 holders
+*/
+static NTSTATUS odb_tdb_break_oplocks(struct odb_lock *lck)
+{
+	struct odb_context *odb = lck->odb;
+	NTSTATUS status;
+	struct opendb_file file;
+	int i;
+	bool modified = true;
+
+	status = odb_pull_record(lck, &file);
+	if (NT_STATUS_EQUAL(status, NT_STATUS_OBJECT_NAME_NOT_FOUND)) {
+		return NT_STATUS_OK;
+	}
+	NT_STATUS_NOT_OK_RETURN(status);
+
+	/* see if anyone has an oplock, which we need to break */
+	for (i=0;i<file.num_entries;i++) {
+		if (file.entries[i].oplock_level == OPLOCK_LEVEL_II) {
+			/*
+			 * there could be multiple level2 oplocks
+			 * and we just send a break to none to all of them
+			 * without waiting for a release
+			 */
+			odb_oplock_break_send(odb, &file.entries[i],
+					      OPLOCK_BREAK_TO_NONE);
+			file.entries[i].oplock_level = OPLOCK_NONE;
+			modified = true;
+		}
+	}
+
+	if (modified) {
+		return odb_push_record(lck, &file);
+	}
+	return NT_STATUS_OK;
+}
 
 /*
   remove a pending opendb entry
@@ -560,14 +749,14 @@ static NTSTATUS odb_tdb_get_delete_on_close(struct odb_context *odb,
   create_options and access_mask
 */
 static NTSTATUS odb_tdb_can_open(struct odb_lock *lck,
-				 uint32_t share_access, uint32_t create_options, 
-				 uint32_t access_mask)
+				 uint32_t stream_id, uint32_t share_access,
+				 uint32_t access_mask, bool delete_on_close,
+				 uint32_t open_disposition, bool break_to_none)
 {
 	struct odb_context *odb = lck->odb;
 	NTSTATUS status;
 	struct opendb_file file;
-	struct opendb_entry e;
-	int i;
+	bool attrs_only = false;
 
 	status = odb_pull_record(lck, &file);
 	if (NT_STATUS_EQUAL(status, NT_STATUS_OBJECT_NAME_NOT_FOUND)) {
@@ -575,32 +764,11 @@ static NTSTATUS odb_tdb_can_open(struct odb_lock *lck,
 	}
 	NT_STATUS_NOT_OK_RETURN(status);
 
-	if ((create_options & NTCREATEX_OPTIONS_DELETE_ON_CLOSE) && 
-	    file.num_entries != 0) {
-		return NT_STATUS_SHARING_VIOLATION;
-	}
-
-	if (file.delete_on_close) {
-		return NT_STATUS_DELETE_PENDING;
-	}
-
-	e.server       = odb->ntvfs_ctx->server_id;
-	e.file_handle  = NULL;
-	e.stream_id    = 0;
-	e.share_access = share_access;
-	e.access_mask  = access_mask;
-		
-	for (i=0;i<file.num_entries;i++) {
-		status = share_conflict(&file.entries[i], &e);
-		if (!NT_STATUS_IS_OK(status)) {
-			/* note that we discard the error code
-			   here. We do this as unless we are actually
-			   doing an open (which comes via a different
-			   function), we need to return a sharing
-			   violation */
-			return NT_STATUS_SHARING_VIOLATION;
-		}
-	}
+	status = odb_tdb_open_can_internal(odb, &file, stream_id,
+					   share_access, access_mask,
+					   delete_on_close, open_disposition,
+					   break_to_none, &attrs_only);
+	NT_STATUS_NOT_OK_RETURN(status);
 
 	return NT_STATUS_OK;
 }
@@ -609,6 +777,7 @@ static NTSTATUS odb_tdb_can_open(struct odb_lock *lck,
 static const struct opendb_ops opendb_tdb_ops = {
 	.odb_init                = odb_tdb_init,
 	.odb_lock                = odb_tdb_lock,
+	.odb_get_key             = odb_tdb_get_key,
 	.odb_open_file           = odb_tdb_open_file,
 	.odb_open_file_pending   = odb_tdb_open_file_pending,
 	.odb_close_file          = odb_tdb_close_file,
@@ -616,7 +785,9 @@ static const struct opendb_ops opendb_tdb_ops = {
 	.odb_rename              = odb_tdb_rename,
 	.odb_set_delete_on_close = odb_tdb_set_delete_on_close,
 	.odb_get_delete_on_close = odb_tdb_get_delete_on_close,
-	.odb_can_open            = odb_tdb_can_open
+	.odb_can_open            = odb_tdb_can_open,
+	.odb_update_oplock       = odb_tdb_update_oplock,
+	.odb_break_oplocks       = odb_tdb_break_oplocks
 };
 
 
diff --git a/source4/ntvfs/posix/config.mk b/source4/ntvfs/posix/config.mk
index 6879940337..88048c2af7 100644
--- a/source4/ntvfs/posix/config.mk
+++ b/source4/ntvfs/posix/config.mk
@@ -53,6 +53,7 @@ OBJ_FILES = \
 		pvfs_resolve.o \
 		pvfs_shortname.o \
 		pvfs_lock.o \
+		pvfs_oplock.o \
 		pvfs_wait.o \
 		pvfs_seek.o \
 		pvfs_ioctl.o \
diff --git a/source4/ntvfs/posix/pvfs_lock.c b/source4/ntvfs/posix/pvfs_lock.c
index b9bb58c71d..baa92880f1 100644
--- a/source4/ntvfs/posix/pvfs_lock.c
+++ b/source4/ntvfs/posix/pvfs_lock.c
@@ -53,7 +53,7 @@ struct pvfs_pending_lock {
 	struct pvfs_file *f;
 	struct ntvfs_request *req;
 	int pending_lock;
-	void *wait_handle;
+	struct pvfs_wait *wait_handle;
 	struct timeval end_time;
 };
 
@@ -294,6 +294,10 @@ NTSTATUS pvfs_lock(struct ntvfs_module_context *ntvfs,
 		return ntvfs_map_lock(ntvfs, req, lck);
 	}
 
+	if (lck->lockx.in.mode & LOCKING_ANDX_OPLOCK_RELEASE) {
+		return pvfs_oplock_release(ntvfs, req, lck);
+	}
+
 	f = pvfs_find_fd(pvfs, req, lck->lockx.in.file.ntvfs);
 	if (!f) {
 		return NT_STATUS_INVALID_HANDLE;
@@ -303,6 +307,9 @@ NTSTATUS pvfs_lock(struct ntvfs_module_context *ntvfs,
 		return NT_STATUS_FILE_IS_A_DIRECTORY;
 	}
 
+	status = pvfs_break_level2_oplocks(f);
+	NT_STATUS_NOT_OK_RETURN(status);
+
 	if (lck->lockx.in.timeout != 0 && 
 	    (req->async_states->state & NTVFS_ASYNC_STATE_MAY_ASYNC)) {
 		pending = talloc(f, struct pvfs_pending_lock);
@@ -338,13 +345,6 @@ NTSTATUS pvfs_lock(struct ntvfs_module_context *ntvfs,
 		return NT_STATUS_DOS(ERRDOS, ERRnoatomiclocks);
 	}
 
-	if (lck->lockx.in.mode & LOCKING_ANDX_OPLOCK_RELEASE) {
-		DEBUG(0,("received unexpected oplock break\n"));
-		talloc_free(pending);
-		return NT_STATUS_NOT_IMPLEMENTED;
-	}
-
-
 	/* the unlocks happen first */
 	locks = lck->lockx.in.locks;
 
diff --git a/source4/ntvfs/posix/pvfs_notify.c b/source4/ntvfs/posix/pvfs_notify.c
index 210f949395..06d2bc8e0c 100644
--- a/source4/ntvfs/posix/pvfs_notify.c
+++ b/source4/ntvfs/posix/pvfs_notify.c
@@ -268,9 +268,11 @@ NTSTATUS pvfs_notify(struct ntvfs_module_context *ntvfs,
 
 	/* if the buffer is empty then start waiting */
 	if (f->notify_buffer->num_changes == 0) {
-		void *wait_handle =
-			pvfs_wait_message(pvfs, req, -1, timeval_zero(), 
-					  pvfs_notify_end, f->notify_buffer);
+		struct pvfs_wait *wait_handle;
+		wait_handle = pvfs_wait_message(pvfs, req, -1,
+						timeval_zero(),
+						pvfs_notify_end,
+						f->notify_buffer);
 		NT_STATUS_HAVE_NO_MEMORY(wait_handle);
 		talloc_steal(req, wait_handle);
 		return NT_STATUS_OK;
diff --git a/source4/ntvfs/posix/pvfs_open.c b/source4/ntvfs/posix/pvfs_open.c
index 8558f0476a..a01352f60c 100644
--- a/source4/ntvfs/posix/pvfs_open.c
+++ b/source4/ntvfs/posix/pvfs_open.c
@@ -268,6 +268,7 @@ static NTSTATUS pvfs_open_directory(struct pvfs_state *pvfs,
 	f->handle->seek_offset       = 0;
 	f->handle->position          = 0;
 	f->handle->mode              = 0;
+	f->handle->oplock            = NULL;
 	f->handle->sticky_write_time = false;
 	f->handle->open_completed    = false;
 
@@ -296,9 +297,10 @@ static NTSTATUS pvfs_open_directory(struct pvfs_state *pvfs,
 		}
 		
 		/* see if we are allowed to open at the same time as existing opens */
-		status = odb_open_file(lck, f->handle, f->handle->name->stream_id,
+		status = odb_open_file(lck, f->handle, name->full_name, name->stream_id,
 				       share_access, access_mask, del_on_close, 
-				       name->full_name, OPLOCK_NONE, NULL);
+				       io->generic.in.open_disposition,
+				       false, OPLOCK_NONE, NULL);
 
 		if (!NT_STATUS_IS_OK(status)) {
 			talloc_free(lck);
@@ -349,9 +351,10 @@ static NTSTATUS pvfs_open_directory(struct pvfs_state *pvfs,
 			return NT_STATUS_INTERNAL_DB_CORRUPTION;
 		}
 
-		status = odb_open_file(lck, f->handle, f->handle->name->stream_id,
+		status = odb_open_file(lck, f->handle, name->full_name, name->stream_id,
 				       share_access, access_mask, del_on_close, 
-				       name->full_name, OPLOCK_NONE, NULL);
+				       io->generic.in.open_disposition,
+				       false, OPLOCK_NONE, NULL);
 
 		if (!NT_STATUS_IS_OK(status)) {
 			goto cleanup_delete;
@@ -669,9 +672,10 @@ static NTSTATUS pvfs_create_file(struct pvfs_state *pvfs,
 		oplock_level = OPLOCK_EXCLUSIVE;
 	}
 
-	status = odb_open_file(lck, f->handle, name->stream_id,
+	status = odb_open_file(lck, f->handle, name->full_name, name->stream_id,
 			       share_access, access_mask, del_on_close, 
-			       name->full_name, oplock_level, &oplock_granted);
+			       io->generic.in.open_disposition,
+			       false, oplock_level, &oplock_granted);
 	talloc_free(lck);
 	if (!NT_STATUS_IS_OK(status)) {
 		/* bad news, we must have hit a race - we don't delete the file
@@ -681,9 +685,6 @@ static NTSTATUS pvfs_create_file(struct pvfs_state *pvfs,
 		return status;
 	}
 
-	if (pvfs->flags & PVFS_FLAG_FAKE_OPLOCKS) {
-		oplock_granted = OPLOCK_BATCH;
-	}
 
 	f->ntvfs             = h;
 	f->pvfs              = pvfs;
@@ -702,12 +703,23 @@ static NTSTATUS pvfs_create_file(struct pvfs_state *pvfs,
 	f->handle->seek_offset       = 0;
 	f->handle->position          = 0;
 	f->handle->mode              = 0;
+	f->handle->oplock            = NULL;
 	f->handle->have_opendb_entry = true;
 	f->handle->sticky_write_time = false;
 	f->handle->open_completed    = false;
 
 	DLIST_ADD(pvfs->files.list, f);
 
+	if (pvfs->flags & PVFS_FLAG_FAKE_OPLOCKS) {
+		oplock_granted = OPLOCK_BATCH;
+	} else if (oplock_granted != OPLOCK_NONE) {
+		status = pvfs_setup_oplock(f, oplock_granted);
+		if (!NT_STATUS_IS_OK(status)) {
+			talloc_free(lck);
+			return status;
+		}
+	}
+
 	/* setup a destructor to avoid file descriptor leaks on
 	   abnormal termination */
 	talloc_set_destructor(f, pvfs_fnum_destructor);
@@ -748,20 +760,25 @@ cleanup_delete:
 	return status;
 }
 
-
 /*
-  state of a pending open retry
+  state of a pending retry
 */
-struct pvfs_open_retry {
+struct pvfs_odb_retry {
 	struct ntvfs_module_context *ntvfs;
 	struct ntvfs_request *req;
-	union smb_open *io;
-	void *wait_handle;
 	DATA_BLOB odb_locking_key;
+	void *io;
+	void *private_data;
+	void (*callback)(struct pvfs_odb_retry *r,
+			 struct ntvfs_module_context *ntvfs,
+			 struct ntvfs_request *req,
+			 void *io,
+			 void *private_data,
+			 enum pvfs_wait_notice reason);
 };
 
-/* destroy a pending open request */
-static int pvfs_retry_destructor(struct pvfs_open_retry *r)
+/* destroy a pending request */
+static int pvfs_odb_retry_destructor(struct pvfs_odb_retry *r)
 {
 	struct pvfs_state *pvfs = r->ntvfs->private_data;
 	if (r->odb_locking_key.data) {
@@ -775,15 +792,92 @@ static int pvfs_retry_destructor(struct pvfs_open_retry *r)
 	return 0;
 }
 
+static void pvfs_odb_retry_callback(void *_r, enum pvfs_wait_notice reason)
+{
+	struct pvfs_odb_retry *r = talloc_get_type(_r, struct pvfs_odb_retry);
+
+	if (reason == PVFS_WAIT_EVENT) {
+		/*
+		 * The pending odb entry is already removed.
+		 * We use a null locking key to indicate this
+		 * to the destructor.
+		 */
+		data_blob_free(&r->odb_locking_key);
+	}
+
+	r->callback(r, r->ntvfs, r->req, r->io, r->private_data, reason);
+}
+
 /*
-  retry an open
+  setup for a retry of a request that was rejected
+  by odb_open_file() or odb_can_open()
 */
-static void pvfs_open_retry(void *private, enum pvfs_wait_notice reason)
+NTSTATUS pvfs_odb_retry_setup(struct ntvfs_module_context *ntvfs,
+			      struct ntvfs_request *req,
+			      struct odb_lock *lck,
+			      struct timeval end_time,
+			      void *io,
+			      void *private_data,
+			      void (*callback)(struct pvfs_odb_retry *r,
+					       struct ntvfs_module_context *ntvfs,
+					       struct ntvfs_request *req,
+					       void *io,
+					       void *private_data,
+					       enum pvfs_wait_notice reason))
 {
-	struct pvfs_open_retry *r = private;
-	struct ntvfs_module_context *ntvfs = r->ntvfs;
-	struct ntvfs_request *req = r->req;
-	union smb_open *io = r->io;
+	struct pvfs_state *pvfs = ntvfs->private_data;
+	struct pvfs_odb_retry *r;
+	struct pvfs_wait *wait_handle;
+	NTSTATUS status;
+
+	r = talloc(req, struct pvfs_odb_retry);
+	NT_STATUS_HAVE_NO_MEMORY(r);
+
+	r->ntvfs = ntvfs;
+	r->req = req;
+	r->io = io;
+	r->private_data = private_data;
+	r->callback = callback;
+	r->odb_locking_key = odb_get_key(r, lck);
+	if (r->odb_locking_key.data == NULL) {
+		return NT_STATUS_NO_MEMORY;
+	}
+
+	/* setup a pending lock */
+	status = odb_open_file_pending(lck, r);
+	if (!NT_STATUS_IS_OK(status)) {
+		return status;
+	}
+
+	talloc_free(lck);
+
+	talloc_set_destructor(r, pvfs_odb_retry_destructor);
+
+	wait_handle = pvfs_wait_message(pvfs, req,
+					MSG_PVFS_RETRY_OPEN, end_time,
+					pvfs_odb_retry_callback, r);
+	if (wait_handle == NULL) {
+		return NT_STATUS_NO_MEMORY;
+	}
+
+	talloc_steal(r, wait_handle);
+
+	talloc_steal(pvfs, r);
+
+	return NT_STATUS_OK;
+}
+
+/*
+  retry an open after a sharing violation
+*/
+static void pvfs_retry_open_sharing(struct pvfs_odb_retry *r,
+				    struct ntvfs_module_context *ntvfs,
+				    struct ntvfs_request *req,
+				    void *_io,
+				    void *private_data,
+				    enum pvfs_wait_notice reason)
+{
+	union smb_open *io = talloc_get_type(_io, union smb_open);
 	NTSTATUS status;
 
 	/* w2k3 ignores SMBntcancel for outstanding open requests. It's probably
@@ -792,8 +886,6 @@ static void pvfs_open_retry(void *private, enum pvfs_wait_notice reason)
 		return;
 	}
 
-	talloc_free(r->wait_handle);
-
 	if (reason == PVFS_WAIT_TIMEOUT) {
 		/* if it timed out, then give the failure
 		   immediately */
@@ -803,9 +895,6 @@ static void pvfs_open_retry(void *private, enum pvfs_wait_notice reason)
 		return;
 	}
 
-	/* the pending odb entry is already removed. We use a null locking
-	   key to indicate this */
-	data_blob_free(&r->odb_locking_key);
 	talloc_free(r);
 
 	/* try the open again, which could trigger another retry setup
@@ -919,10 +1008,10 @@ static NTSTATUS pvfs_open_setup_retry(struct ntvfs_module_context *ntvfs,
 				      struct ntvfs_request *req, 
 				      union smb_open *io,
 				      struct pvfs_file *f,
-				      struct odb_lock *lck)
+				      struct odb_lock *lck,
+				      NTSTATUS parent_status)
 {
 	struct pvfs_state *pvfs = ntvfs->private_data;
-	struct pvfs_open_retry *r;
 	NTSTATUS status;
 	struct timeval end_time;
 
@@ -936,40 +1025,21 @@ static NTSTATUS pvfs_open_setup_retry(struct ntvfs_module_context *ntvfs,
 		}
 	}
 
-	r = talloc(req, struct pvfs_open_retry);
-	if (r == NULL) {
-		return NT_STATUS_NO_MEMORY;
-	}
-
-	r->ntvfs = ntvfs;
-	r->req = req;
-	r->io = io;
-	r->odb_locking_key = data_blob_talloc(r, 
-					      f->handle->odb_locking_key.data, 
-					      f->handle->odb_locking_key.length);
-
-	end_time = timeval_add(&req->statistics.request_time, 0, pvfs->sharing_violation_delay);
-
-	/* setup a pending lock */
-	status = odb_open_file_pending(lck, r);
-	if (!NT_STATUS_IS_OK(status)) {
-		return status;
-	}
-
-	talloc_free(lck);
+	/* the retry should allocate a new file handle */
 	talloc_free(f);
 
-	talloc_set_destructor(r, pvfs_retry_destructor);
-
-	r->wait_handle = pvfs_wait_message(pvfs, req, MSG_PVFS_RETRY_OPEN, end_time, 
-					   pvfs_open_retry, r);
-	if (r->wait_handle == NULL) {
-		return NT_STATUS_NO_MEMORY;
+	if (NT_STATUS_EQUAL(parent_status, NT_STATUS_SHARING_VIOLATION)) {
+		end_time = timeval_add(&req->statistics.request_time,
+				       0, pvfs->sharing_violation_delay);
+	} else if (NT_STATUS_EQUAL(parent_status, NT_STATUS_OPLOCK_NOT_GRANTED)) {
+		end_time = timeval_add(&req->statistics.request_time,
+				       pvfs->oplock_break_timeout, 0);
+	} else {
+		return NT_STATUS_INTERNAL_ERROR;
 	}
 
-	talloc_steal(pvfs, r);
-
-	return NT_STATUS_OK;
+	return pvfs_odb_retry_setup(ntvfs, req, lck, end_time, io, NULL,
+				    pvfs_retry_open_sharing);
 }
 
 /*
@@ -1133,6 +1203,7 @@ NTSTATUS pvfs_open(struct ntvfs_module_context *ntvfs,
 	f->handle->seek_offset       = 0;
 	f->handle->position          = 0;
 	f->handle->mode              = 0;
+	f->handle->oplock            = NULL;
 	f->handle->have_opendb_entry = false;
 	f->handle->sticky_write_time = false;
 	f->handle->open_completed    = false;
@@ -1186,15 +1257,21 @@ NTSTATUS pvfs_open(struct ntvfs_module_context *ntvfs,
 	}
 
 	/* see if we are allowed to open at the same time as existing opens */
-	status = odb_open_file(lck, f->handle, f->handle->name->stream_id,
+	status = odb_open_file(lck, f->handle, name->full_name, name->stream_id,
 			       share_access, access_mask, del_on_close,
-			       name->full_name, oplock_level, &oplock_granted);
+			       io->generic.in.open_disposition,
+			       false, oplock_level, &oplock_granted);
 
-	/* on a sharing violation we need to retry when the file is closed by 
-	   the other user, or after 1 second */
-	if (NT_STATUS_EQUAL(status, NT_STATUS_SHARING_VIOLATION) &&
+	/*
+	 * on a sharing violation we need to retry when the file is closed by
+	 * the other user, or after 1 second
+	 * on a non granted oplock we need to retry when the file is closed by
+	 * the other user, or after 30 seconds
+	*/
+	if ((NT_STATUS_EQUAL(status, NT_STATUS_SHARING_VIOLATION) ||
+	     NT_STATUS_EQUAL(status, NT_STATUS_OPLOCK_NOT_GRANTED)) &&
 	    (req->async_states->state & NTVFS_ASYNC_STATE_MAY_ASYNC)) {
-		return pvfs_open_setup_retry(ntvfs, req, io, f, lck);
+		return pvfs_open_setup_retry(ntvfs, req, io, f, lck, status);
 	}
 
 	if (!NT_STATUS_IS_OK(status)) {
@@ -1204,6 +1281,12 @@ NTSTATUS pvfs_open(struct ntvfs_module_context *ntvfs,
 
 	if (pvfs->flags & PVFS_FLAG_FAKE_OPLOCKS) {
 		oplock_granted = OPLOCK_BATCH;
+	} else if (oplock_granted != OPLOCK_NONE) {
+		status = pvfs_setup_oplock(f, oplock_granted);
+		if (!NT_STATUS_IS_OK(status)) {
+			talloc_free(lck);
+			return status;
+		}
 	}
 
 	f->handle->have_opendb_entry = true;
@@ -1424,6 +1507,9 @@ NTSTATUS pvfs_can_delete(struct pvfs_state *pvfs,
 	NTSTATUS status;
 	DATA_BLOB key;
 	struct odb_lock *lck;
+	uint32_t share_access;
+	uint32_t access_mask;
+	bool delete_on_close;
 
 	status = pvfs_locking_key(name, name, &key);
 	if (!NT_STATUS_IS_OK(status)) {
@@ -1436,21 +1522,38 @@ NTSTATUS pvfs_can_delete(struct pvfs_state *pvfs,
 		return NT_STATUS_INTERNAL_DB_CORRUPTION;
 	}
 
-	status = odb_can_open(lck,
-			      NTCREATEX_SHARE_ACCESS_READ |
-			      NTCREATEX_SHARE_ACCESS_WRITE | 
-			      NTCREATEX_SHARE_ACCESS_DELETE, 
-			      NTCREATEX_OPTIONS_DELETE_ON_CLOSE, 
-			      SEC_STD_DELETE);
+	share_access	= NTCREATEX_SHARE_ACCESS_READ |
+			  NTCREATEX_SHARE_ACCESS_WRITE |
+			  NTCREATEX_SHARE_ACCESS_DELETE;
+	access_mask	= SEC_STD_DELETE;
+	delete_on_close	= true;
+
+	status = odb_can_open(lck, name->stream_id,
+			      share_access, access_mask, delete_on_close,
+			      NTCREATEX_DISP_OPEN, false);
 
 	if (NT_STATUS_IS_OK(status)) {
-		status = pvfs_access_check_simple(pvfs, req, name, SEC_STD_DELETE);
+		status = pvfs_access_check_simple(pvfs, req, name, access_mask);
 	}
 
-	if (!NT_STATUS_IS_OK(status)) {
+	/*
+	 * if it's a sharing violation or we got no oplock
+	 * only keep the lock if the caller requested access
+	 * to the lock
+	 */
+	if (NT_STATUS_EQUAL(status, NT_STATUS_SHARING_VIOLATION) ||
+	    NT_STATUS_EQUAL(status, NT_STATUS_OPLOCK_NOT_GRANTED)) {
+		if (lckp) {
+			*lckp = lck;
+		} else {
+			talloc_free(lck);
+		}
+	} else if (!NT_STATUS_IS_OK(status)) {
 		talloc_free(lck);
-		*lckp = lck;
-	} else if (lckp != NULL) {
+		if (lckp) {
+			*lckp = NULL;
+		}
+	} else if (lckp) {
 		*lckp = lck;
 	}
 
@@ -1469,6 +1572,9 @@ NTSTATUS pvfs_can_rename(struct pvfs_state *pvfs,
 	NTSTATUS status;
 	DATA_BLOB key;
 	struct odb_lock *lck;
+	uint32_t share_access;
+	uint32_t access_mask;
+	bool delete_on_close;
 
 	status = pvfs_locking_key(name, name, &key);
 	if (!NT_STATUS_IS_OK(status)) {
@@ -1481,16 +1587,105 @@ NTSTATUS pvfs_can_rename(struct pvfs_state *pvfs,
 		return NT_STATUS_INTERNAL_DB_CORRUPTION;
 	}
 
-	status = odb_can_open(lck,
-			      NTCREATEX_SHARE_ACCESS_READ |
-			      NTCREATEX_SHARE_ACCESS_WRITE,
-			      0,
-			      SEC_STD_DELETE);
+	share_access	= NTCREATEX_SHARE_ACCESS_READ |
+			  NTCREATEX_SHARE_ACCESS_WRITE;
+	access_mask	= SEC_STD_DELETE;
+	delete_on_close	= false;
 
-	if (!NT_STATUS_IS_OK(status)) {
+	status = odb_can_open(lck, name->stream_id,
+			      share_access, access_mask, delete_on_close,
+			      NTCREATEX_DISP_OPEN, false);
+
+	/*
+	 * if it's a sharing violation or we got no oplock
+	 * only keep the lock if the caller requested access
+	 * to the lock
+	 */
+	if (NT_STATUS_EQUAL(status, NT_STATUS_SHARING_VIOLATION) ||
+	    NT_STATUS_EQUAL(status, NT_STATUS_OPLOCK_NOT_GRANTED)) {
+		if (lckp) {
+			*lckp = lck;
+		} else {
+			talloc_free(lck);
+		}
+	} else if (!NT_STATUS_IS_OK(status)) {
 		talloc_free(lck);
+		if (lckp) {
+			*lckp = NULL;
+		}
+	} else if (lckp) {
 		*lckp = lck;
-	} else if (lckp != NULL) {
+	}
+
+	return status;
+}
+
+/*
+  determine if the file size of a file can be changed,
+  or if it is prevented by an already open file
+*/
+NTSTATUS pvfs_can_update_file_size(struct pvfs_state *pvfs,
+				   struct ntvfs_request *req,
+				   struct pvfs_filename *name,
+				   struct odb_lock **lckp)
+{
+	NTSTATUS status;
+	DATA_BLOB key;
+	struct odb_lock *lck;
+	uint32_t share_access;
+	uint32_t access_mask;
+	bool break_to_none;
+	bool delete_on_close;
+
+	status = pvfs_locking_key(name, name, &key);
+	if (!NT_STATUS_IS_OK(status)) {
+		return NT_STATUS_NO_MEMORY;
+	}
+
+	lck = odb_lock(req, pvfs->odb_context, &key);
+	if (lck == NULL) {
+		DEBUG(0,("Unable to lock opendb for can_stat\n"));
+		return NT_STATUS_INTERNAL_DB_CORRUPTION;
+	}
+
+	share_access	= NTCREATEX_SHARE_ACCESS_READ |
+			  NTCREATEX_SHARE_ACCESS_WRITE |
+			  NTCREATEX_SHARE_ACCESS_DELETE;
+	/*
+	 * I would have thought that we would need to pass
+	 * SEC_FILE_WRITE_DATA | SEC_FILE_APPEND_DATA here too
+	 *
+	 * But you only need SEC_FILE_WRITE_ATTRIBUTE permissions
+	 * to set the filesize.
+	 *
+	 * --metze
+	 */
+	access_mask	= SEC_FILE_WRITE_ATTRIBUTE;
+	delete_on_close	= false;
+	break_to_none	= true;
+
+	status = odb_can_open(lck, name->stream_id,
+			      share_access, access_mask, delete_on_close,
+			      NTCREATEX_DISP_OPEN, break_to_none);
+
+	/*
+	 * if it's a sharing violation or we got no oplock
+	 * only keep the lock if the caller requested access
+	 * to the lock
+	 */
+	if (NT_STATUS_EQUAL(status, NT_STATUS_SHARING_VIOLATION) ||
+	    NT_STATUS_EQUAL(status, NT_STATUS_OPLOCK_NOT_GRANTED)) {
+		if (lckp) {
+			*lckp = lck;
+		} else {
+			talloc_free(lck);
+		}
+	} else if (!NT_STATUS_IS_OK(status)) {
+		talloc_free(lck);
+		if (lckp) {
+			*lckp = NULL;
+		}
+	} else if (lckp) {
 		*lckp = lck;
 	}
 
@@ -1508,6 +1703,9 @@ NTSTATUS pvfs_can_stat(struct pvfs_state *pvfs,
 	NTSTATUS status;
 	DATA_BLOB key;
 	struct odb_lock *lck;
+	uint32_t share_access;
+	uint32_t access_mask;
+	bool delete_on_close;
 
 	status = pvfs_locking_key(name, name, &key);
 	if (!NT_STATUS_IS_OK(status)) {
@@ -1520,10 +1718,18 @@ NTSTATUS pvfs_can_stat(struct pvfs_state *pvfs,
 		return NT_STATUS_INTERNAL_DB_CORRUPTION;
 	}
 
-	status = odb_can_open(lck,
-			      NTCREATEX_SHARE_ACCESS_READ |
-			      NTCREATEX_SHARE_ACCESS_WRITE,
-			      0, 0);
+	share_access	= NTCREATEX_SHARE_ACCESS_READ |
+			  NTCREATEX_SHARE_ACCESS_WRITE;
+	access_mask	= SEC_FILE_READ_ATTRIBUTE;
+	delete_on_close	= false;
+
+	status = odb_can_open(lck, name->stream_id,
+			      share_access, access_mask, delete_on_close,
+			      NTCREATEX_DISP_OPEN, false);
+
+	if (!NT_STATUS_IS_OK(status)) {
+		talloc_free(lck);
+	}
 
 	return status;
 }
diff --git a/source4/ntvfs/posix/pvfs_oplock.c b/source4/ntvfs/posix/pvfs_oplock.c
new file mode 100644
index 0000000000..cf30ddbc59
--- /dev/null
+++ b/source4/ntvfs/posix/pvfs_oplock.c
@@ -0,0 +1,239 @@
+/*
+   Unix SMB/CIFS implementation.
+
+   POSIX NTVFS backend - oplock handling
+
+   Copyright (C) Stefan Metzmacher 2008
+
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.
+
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "includes.h"
+#include "lib/messaging/messaging.h"
+#include "lib/messaging/irpc.h"
+#include "vfs_posix.h"
+
+
+struct pvfs_oplock {
+	struct pvfs_file_handle *handle;
+	struct pvfs_file *file;
+	uint32_t level;
+	struct messaging_context *msg_ctx;
+};
+
+/*
+  receive oplock breaks and forward them to the client
+*/
+static void pvfs_oplock_break(struct pvfs_oplock *opl, uint8_t level)
+{
+	NTSTATUS status;
+	struct pvfs_file *f = opl->file;
+	struct pvfs_file_handle *h = opl->handle;
+	struct pvfs_state *pvfs = h->pvfs;
+
+	DEBUG(10,("pvfs_oplock_break: sending oplock break level %d for '%s' %p\n",
+		level, h->name->original_name, h));
+	status = ntvfs_send_oplock_break(pvfs->ntvfs, f->ntvfs, level);
+	if (!NT_STATUS_IS_OK(status)) {
+		DEBUG(0,("pvfs_oplock_break: sending oplock break failed: %s\n",
+			nt_errstr(status)));
+	}
+}
+
+static void pvfs_oplock_break_dispatch(struct messaging_context *msg,
+				       void *private_data, uint32_t msg_type,
+				       struct server_id src, DATA_BLOB *data)
+{
+	struct pvfs_oplock *opl = talloc_get_type(private_data,
+						  struct pvfs_oplock);
+	struct opendb_oplock_break opb;
+
+	ZERO_STRUCT(opb);
+
+	/* we need to check that this one is for us. See
+	   messaging_send_ptr() for the other side of this.
+	 */
+	if (data->length == sizeof(struct opendb_oplock_break)) {
+		struct opendb_oplock_break *p;
+		p = (struct opendb_oplock_break *)data->data;
+		opb = *p;
+	} else {
+		DEBUG(0,("%s: ignore oplock break with length[%u]\n",
+			__location__, data->length));
+		return;
+	}
+	if (opb.file_handle != opl->handle) {
+		return;
+	}
+
+	/*
+	 * maybe we should use ntvfs_setup_async()
+	 */
+	pvfs_oplock_break(opl, opb.level);
+}
+
+static int pvfs_oplock_destructor(struct pvfs_oplock *opl)
+{
+	messaging_deregister(opl->msg_ctx, MSG_NTVFS_OPLOCK_BREAK, opl);
+	return 0;
+}
+
+NTSTATUS pvfs_setup_oplock(struct pvfs_file *f, uint32_t oplock_granted)
+{
+	NTSTATUS status;
+	struct pvfs_oplock *opl;
+	uint32_t level = OPLOCK_NONE;
+
+	f->handle->oplock = NULL;
+
+	switch (oplock_granted) {
+	case EXCLUSIVE_OPLOCK_RETURN:
+		level = OPLOCK_EXCLUSIVE;
+		break;
+	case BATCH_OPLOCK_RETURN:
+		level = OPLOCK_BATCH;
+		break;
+	case LEVEL_II_OPLOCK_RETURN:
+		level = OPLOCK_LEVEL_II;
+		break;
+	}
+
+	if (level == OPLOCK_NONE) {
+		return NT_STATUS_OK;
+	}
+
+	opl = talloc(f->handle, struct pvfs_oplock);
+	NT_STATUS_HAVE_NO_MEMORY(opl);
+
+	opl->handle	= f->handle;
+	opl->file	= f;
+	opl->level	= level;
+	opl->msg_ctx	= f->pvfs->ntvfs->ctx->msg_ctx;
+
+	status = messaging_register(opl->msg_ctx,
+				    opl,
+				    MSG_NTVFS_OPLOCK_BREAK,
+				    pvfs_oplock_break_dispatch);
+	NT_STATUS_NOT_OK_RETURN(status);
+
+	/* destructor */
+	talloc_set_destructor(opl, pvfs_oplock_destructor);
+
+	f->handle->oplock = opl;
+
+	return NT_STATUS_OK;
+}
+
+NTSTATUS pvfs_oplock_release(struct ntvfs_module_context *ntvfs,
+			     struct ntvfs_request *req, union smb_lock *lck)
+{
+	struct pvfs_state *pvfs = ntvfs->private_data;
+	struct pvfs_file *f;
+	struct pvfs_file_handle *h;
+	struct odb_lock *olck;
+	uint8_t oplock_break;
+	NTSTATUS status;
+
+	f = pvfs_find_fd(pvfs, req, lck->lockx.in.file.ntvfs);
+	if (!f) {
+		return NT_STATUS_INVALID_HANDLE;
+	}
+
+	h = f->handle;
+
+	if (h->fd == -1) {
+		return NT_STATUS_FILE_IS_A_DIRECTORY;
+	}
+
+	if (!h->have_opendb_entry) {
+		return NT_STATUS_FOOBAR;
+	}
+
+	if (!h->oplock) {
+		return NT_STATUS_FOOBAR;
+	}
+
+	olck = odb_lock(h, h->pvfs->odb_context, &h->odb_locking_key);
+	if (olck == NULL) {
+		DEBUG(0,("Unable to lock opendb for oplock update\n"));
+		return NT_STATUS_FOOBAR;
+	}
+
+	oplock_break = (lck->lockx.in.mode >> 8) & 0xFF;
+	if (oplock_break == OPLOCK_BREAK_TO_NONE) {
+		h->oplock->level = OPLOCK_NONE;
+	} else if (oplock_break == OPLOCK_BREAK_TO_LEVEL_II) {
+		h->oplock->level = OPLOCK_LEVEL_II;
+	} else {
+		/* fallback to level II in case of a invalid value */
+		DEBUG(1,("unexpected oplock break level[0x%02X]\n", oplock_break));
+		h->oplock->level = OPLOCK_LEVEL_II;
+	}
+	status = odb_update_oplock(olck, h, h->oplock->level);
+	if (!NT_STATUS_IS_OK(status)) {
+		DEBUG(0,("Unable to update oplock level for '%s' - %s\n",
+			 h->name->full_name, nt_errstr(status)));
+		talloc_free(olck);
+		return status;
+	}
+
+	talloc_free(olck);
+
+	/* after a break to none, we no longer have an oplock attached */
+	if (h->oplock->level == OPLOCK_NONE) {
+		talloc_free(h->oplock);
+		h->oplock = NULL;
+	}
+
+	return NT_STATUS_OK;
+}
+
+NTSTATUS pvfs_break_level2_oplocks(struct pvfs_file *f)
+{
+	struct pvfs_file_handle *h = f->handle;
+	struct odb_lock *olck;
+	NTSTATUS status;
+
+	if (h->oplock && h->oplock->level == OPLOCK_EXCLUSIVE) {
+		return NT_STATUS_OK;
+	}
+
+	olck = odb_lock(h, h->pvfs->odb_context, &h->odb_locking_key);
+	if (olck == NULL) {
+		DEBUG(0,("Unable to lock opendb for oplock update\n"));
+		return NT_STATUS_FOOBAR;
+	}
+
+	if (h->oplock && h->oplock->level == OPLOCK_BATCH) {
+		status = odb_update_oplock(olck, h, OPLOCK_LEVEL_II);
+		if (!NT_STATUS_IS_OK(status)) {
+			DEBUG(0,("Unable to update oplock level for '%s' - %s\n",
+				 h->name->full_name, nt_errstr(status)));
+			talloc_free(olck);
+			return status;
+		}
+	}
+
+	status = odb_break_oplocks(olck);
+	if (!NT_STATUS_IS_OK(status)) {
+		DEBUG(0,("Unable to break level2 oplocks to none for '%s' - %s\n",
+			 h->name->full_name, nt_errstr(status)));
+		talloc_free(olck);
+		return status;
+	}
+
+	talloc_free(olck);
+
+	return NT_STATUS_OK;
+}
diff --git a/source4/ntvfs/posix/pvfs_qfileinfo.c b/source4/ntvfs/posix/pvfs_qfileinfo.c
index 2f01c08fb0..8d23d707a4 100644
--- a/source4/ntvfs/posix/pvfs_qfileinfo.c
+++ b/source4/ntvfs/posix/pvfs_qfileinfo.c
@@ -330,7 +330,7 @@ NTSTATUS pvfs_qpathinfo(struct ntvfs_module_context *ntvfs,
 
 	status = pvfs_can_stat(pvfs, req, name);
 	if (!NT_STATUS_IS_OK(status)) {
-		return NT_STATUS_DELETE_PENDING;
+		return status;
 	}
 
 	status = pvfs_access_check_simple(pvfs, req, name, 
diff --git a/source4/ntvfs/posix/pvfs_rename.c b/source4/ntvfs/posix/pvfs_rename.c
index 65755e05cb..17c073ccab 100644
--- a/source4/ntvfs/posix/pvfs_rename.c
+++ b/source4/ntvfs/posix/pvfs_rename.c
@@ -192,8 +192,8 @@ static NTSTATUS pvfs_rename_one(struct pvfs_state *pvfs,
 {
 	struct pvfs_filename *name1, *name2;
 	TALLOC_CTX *mem_ctx = talloc_new(req);
+	struct odb_lock *lck = NULL;
 	NTSTATUS status;
-	struct odb_lock *lck, *lck2;
 
 	/* resolve the wildcard pattern for this name */
 	fname2 = pvfs_resolve_wildcard(mem_ctx, lp_iconv_convenience(pvfs->ntvfs->ctx->lp_ctx), fname1, fname2);
@@ -216,6 +216,7 @@ static NTSTATUS pvfs_rename_one(struct pvfs_state *pvfs,
 
 	status = pvfs_can_rename(pvfs, req, name1, &lck);
 	if (!NT_STATUS_IS_OK(status)) {
+		talloc_free(lck);
 		goto failed;
 	}
 
@@ -223,7 +224,7 @@ static NTSTATUS pvfs_rename_one(struct pvfs_state *pvfs,
 	status = pvfs_resolve_partial(pvfs, mem_ctx, 
 				      dir_path, fname2, &name2);
 	if (NT_STATUS_IS_OK(status)) {
-		status = pvfs_can_delete(pvfs, req, name2, &lck2);
+		status = pvfs_can_delete(pvfs, req, name2, NULL);
 		if (!NT_STATUS_IS_OK(status)) {
 			goto failed;
 		}
@@ -311,7 +312,7 @@ static NTSTATUS pvfs_rename_mv(struct ntvfs_module_context *ntvfs,
 	struct pvfs_state *pvfs = ntvfs->private_data;
 	NTSTATUS status;
 	struct pvfs_filename *name1, *name2;
-	struct odb_lock *lck;
+	struct odb_lock *lck = NULL;
 
 	/* resolve the cifs name to a posix name */
 	status = pvfs_resolve_name(pvfs, req, ren->rename.in.pattern1, 
@@ -354,6 +355,7 @@ static NTSTATUS pvfs_rename_mv(struct ntvfs_module_context *ntvfs,
 
 	status = pvfs_can_rename(pvfs, req, name1, &lck);
 	if (!NT_STATUS_IS_OK(status)) {
+		talloc_free(lck);
 		return status;
 	}
 
@@ -375,7 +377,6 @@ static NTSTATUS pvfs_rename_nt(struct ntvfs_module_context *ntvfs,
 	struct pvfs_state *pvfs = ntvfs->private_data;
 	NTSTATUS status;
 	struct pvfs_filename *name1, *name2;
-	struct odb_lock *lck;
 
 	switch (ren->ntrename.in.flags) {
 	case RENAME_FLAG_RENAME:
@@ -421,7 +422,7 @@ static NTSTATUS pvfs_rename_nt(struct ntvfs_module_context *ntvfs,
 		return status;
 	}
 
-	status = pvfs_can_rename(pvfs, req, name1, &lck);
+	status = pvfs_can_rename(pvfs, req, name1, NULL);
 	if (!NT_STATUS_IS_OK(status)) {
 		return status;
 	}
diff --git a/source4/ntvfs/posix/pvfs_setfileinfo.c b/source4/ntvfs/posix/pvfs_setfileinfo.c
index 9c78699edb..da1e31e639 100644
--- a/source4/ntvfs/posix/pvfs_setfileinfo.c
+++ b/source4/ntvfs/posix/pvfs_setfileinfo.c
@@ -142,8 +142,6 @@ static NTSTATUS pvfs_setfileinfo_rename(struct pvfs_state *pvfs,
 
 	/* if the destination exists, then check the rename is allowed */
 	if (name2->exists) {
-		struct odb_lock *lck;
-
 		if (strcmp(name2->full_name, name->full_name) == 0) {
 			/* rename to same name is null-op */
 			return NT_STATUS_OK;
@@ -153,7 +151,10 @@ static NTSTATUS pvfs_setfileinfo_rename(struct pvfs_state *pvfs,
 			return NT_STATUS_OBJECT_NAME_COLLISION;
 		}
 
-		status = pvfs_can_delete(pvfs, req, name2, &lck);
+		status = pvfs_can_delete(pvfs, req, name2, NULL);
+		if (NT_STATUS_EQUAL(status, NT_STATUS_DELETE_PENDING)) {
+			return NT_STATUS_ACCESS_DENIED;
+		}
 		if (NT_STATUS_EQUAL(status, NT_STATUS_SHARING_VIOLATION)) {
 			return NT_STATUS_ACCESS_DENIED;
 		}
@@ -354,6 +355,9 @@ NTSTATUS pvfs_setfileinfo(struct ntvfs_module_context *ntvfs,
 
 	case RAW_SFILEINFO_ALLOCATION_INFO:
 	case RAW_SFILEINFO_ALLOCATION_INFORMATION:
+		status = pvfs_break_level2_oplocks(f);
+		NT_STATUS_NOT_OK_RETURN(status);
+
 		newstats.dos.alloc_size = info->allocation_info.in.alloc_size;
 		if (newstats.dos.alloc_size < newstats.st.st_size) {
 			newstats.st.st_size = newstats.dos.alloc_size;
@@ -364,6 +368,9 @@ NTSTATUS pvfs_setfileinfo(struct ntvfs_module_context *ntvfs,
 
 	case RAW_SFILEINFO_END_OF_FILE_INFO:
 	case RAW_SFILEINFO_END_OF_FILE_INFORMATION:
+		status = pvfs_break_level2_oplocks(f);
+		NT_STATUS_NOT_OK_RETURN(status);
+
 		newstats.st.st_size = info->end_of_file_info.in.size;
 		break;
 
@@ -465,6 +472,84 @@ NTSTATUS pvfs_setfileinfo(struct ntvfs_module_context *ntvfs,
 	return pvfs_dosattrib_save(pvfs, h->name, h->fd);
 }
 
+/*
+  retry an open after a sharing violation
+*/
+static void pvfs_retry_setpathinfo(struct pvfs_odb_retry *r,
+				   struct ntvfs_module_context *ntvfs,
+				   struct ntvfs_request *req,
+				   void *_info,
+				   void *private_data,
+				   enum pvfs_wait_notice reason)
+{
+	union smb_setfileinfo *info = talloc_get_type(_info,
+				      union smb_setfileinfo);
+	NTSTATUS status = NT_STATUS_INTERNAL_ERROR;
+
+	talloc_free(r);
+
+	switch (reason) {
+	case PVFS_WAIT_CANCEL:
+/*TODO*/
+		status = NT_STATUS_CANCELLED;
+		break;
+	case PVFS_WAIT_TIMEOUT:
+		/* if it timed out, then give the failure
+		   immediately */
+/*TODO*/
+		status = NT_STATUS_SHARING_VIOLATION;
+		break;
+	case PVFS_WAIT_EVENT:
+
+		/* try the open again, which could trigger another retry setup
+		   if it wants to, so we have to unmark the async flag so we
+		   will know if it does a second async reply */
+		req->async_states->state &= ~NTVFS_ASYNC_STATE_ASYNC;
+
+		status = pvfs_setpathinfo(ntvfs, req, info);
+		if (req->async_states->state & NTVFS_ASYNC_STATE_ASYNC) {
+			/* the 2nd try also replied async, so we don't send
+			   the reply yet */
+			return;
+		}
+
+		/* re-mark it async, just in case someone up the chain does
+		   paranoid checking */
+		req->async_states->state |= NTVFS_ASYNC_STATE_ASYNC;
+		break;
+	}
+
+	/* send the reply up the chain */
+	req->async_states->status = status;
+	req->async_states->send_fn(req);
+}
+
+/*
+  setup for a unlink retry after a sharing violation
+  or a non granted oplock
+*/
+static NTSTATUS pvfs_setpathinfo_setup_retry(struct ntvfs_module_context *ntvfs,
+					     struct ntvfs_request *req,
+					     union smb_setfileinfo *info,
+					     struct odb_lock *lck,
+					     NTSTATUS status)
+{
+	struct pvfs_state *pvfs = ntvfs->private_data;
+	struct timeval end_time;
+
+	if (NT_STATUS_EQUAL(status, NT_STATUS_SHARING_VIOLATION)) {
+		end_time = timeval_add(&req->statistics.request_time,
+				       0, pvfs->sharing_violation_delay);
+	} else if (NT_STATUS_EQUAL(status, NT_STATUS_OPLOCK_NOT_GRANTED)) {
+		end_time = timeval_add(&req->statistics.request_time,
+				       pvfs->oplock_break_timeout, 0);
+	} else {
+		return NT_STATUS_INTERNAL_ERROR;
+	}
+
+	return pvfs_odb_retry_setup(ntvfs, req, lck, end_time, info, NULL,
+				    pvfs_retry_setpathinfo);
+}
 
 /*
   set info on a pathname
@@ -479,6 +564,7 @@ NTSTATUS pvfs_setpathinfo(struct ntvfs_module_context *ntvfs,
 	struct utimbuf unix_times;
 	uint32_t access_needed;
 	uint32_t change_mask = 0;
+	struct odb_lock *lck = NULL;
 
 	/* resolve the cifs name to a posix name */
 	status = pvfs_resolve_name(pvfs, req, info->generic.in.file.path, 
@@ -553,6 +639,20 @@ NTSTATUS pvfs_setpathinfo(struct ntvfs_module_context *ntvfs,
 
 	case RAW_SFILEINFO_ALLOCATION_INFO:
 	case RAW_SFILEINFO_ALLOCATION_INFORMATION:
+		status = pvfs_can_update_file_size(pvfs, req, name, &lck);
+		/*
+		 * on a sharing violation we need to retry when the file is closed by
+		 * the other user, or after 1 second
+		 * on a non granted oplock we need to retry when the file is closed by
+		 * the other user, or after 30 seconds
+		*/
+		if ((NT_STATUS_EQUAL(status, NT_STATUS_SHARING_VIOLATION) ||
+		     NT_STATUS_EQUAL(status, NT_STATUS_OPLOCK_NOT_GRANTED)) &&
+		    (req->async_states->state & NTVFS_ASYNC_STATE_MAY_ASYNC)) {
+			return pvfs_setpathinfo_setup_retry(pvfs->ntvfs, req, info, lck, status);
+		}
+		NT_STATUS_NOT_OK_RETURN(status);
+
 		if (info->allocation_info.in.alloc_size > newstats.dos.alloc_size) {
 			/* strange. Increasing the allocation size via setpathinfo 
 			   should be silently ignored */
@@ -568,6 +668,20 @@ NTSTATUS pvfs_setpathinfo(struct ntvfs_module_context *ntvfs,
 
 	case RAW_SFILEINFO_END_OF_FILE_INFO:
 	case RAW_SFILEINFO_END_OF_FILE_INFORMATION:
+		status = pvfs_can_update_file_size(pvfs, req, name, &lck);
+		/*
+		 * on a sharing violation we need to retry when the file is closed by
+		 * the other user, or after 1 second
+		 * on a non granted oplock we need to retry when the file is closed by
+		 * the other user, or after 30 seconds
+		*/
+		if ((NT_STATUS_EQUAL(status, NT_STATUS_SHARING_VIOLATION) ||
+		     NT_STATUS_EQUAL(status, NT_STATUS_OPLOCK_NOT_GRANTED)) &&
+		    (req->async_states->state & NTVFS_ASYNC_STATE_MAY_ASYNC)) {
+			return pvfs_setpathinfo_setup_retry(pvfs->ntvfs, req, info, lck, status);
+		}
+		NT_STATUS_NOT_OK_RETURN(status);
+
 		newstats.st.st_size = info->end_of_file_info.in.size;
 		break;
 
diff --git a/source4/ntvfs/posix/pvfs_unlink.c b/source4/ntvfs/posix/pvfs_unlink.c
index ef56d99fb5..7a2d964b9d 100644
--- a/source4/ntvfs/posix/pvfs_unlink.c
+++ b/source4/ntvfs/posix/pvfs_unlink.c
@@ -23,71 +23,94 @@
 #include "vfs_posix.h"
 #include "system/dir.h"
 
-
 /*
-  unlink a stream
- */
-static NTSTATUS pvfs_unlink_stream(struct pvfs_state *pvfs, 
-				   struct ntvfs_request *req,
-				   struct pvfs_filename *name, 
-				   uint16_t attrib)
+  retry an open after a sharing violation
+*/
+static void pvfs_retry_unlink(struct pvfs_odb_retry *r,
+			      struct ntvfs_module_context *ntvfs,
+			      struct ntvfs_request *req,
+			      void *_io,
+			      void *private_data,
+			      enum pvfs_wait_notice reason)
 {
-	NTSTATUS status;
-	struct odb_lock *lck;
+	union smb_unlink *io = talloc_get_type(_io, union smb_unlink);
+	NTSTATUS status = NT_STATUS_INTERNAL_ERROR;
+
+	talloc_free(r);
+
+	switch (reason) {
+	case PVFS_WAIT_CANCEL:
+/*TODO*/
+		status = NT_STATUS_CANCELLED;
+		break;
+	case PVFS_WAIT_TIMEOUT:
+		/* if it timed out, then give the failure
+		   immediately */
+/*TODO*/
+		status = NT_STATUS_SHARING_VIOLATION;
+		break;
+	case PVFS_WAIT_EVENT:
+
+		/* try the open again, which could trigger another retry setup
+		   if it wants to, so we have to unmark the async flag so we
+		   will know if it does a second async reply */
+		req->async_states->state &= ~NTVFS_ASYNC_STATE_ASYNC;
+
+		status = pvfs_unlink(ntvfs, req, io);
+		if (req->async_states->state & NTVFS_ASYNC_STATE_ASYNC) {
+			/* the 2nd try also replied async, so we don't send
+			   the reply yet */
+			return;
+		}
 
-	if (!name->stream_exists) {
-		return NT_STATUS_OBJECT_NAME_NOT_FOUND;
+		/* re-mark it async, just in case someone up the chain does
+		   paranoid checking */
+		req->async_states->state |= NTVFS_ASYNC_STATE_ASYNC;
+		break;
 	}
 
-	/* make sure its matches the given attributes */
-	status = pvfs_match_attrib(pvfs, name, attrib, 0);
-	if (!NT_STATUS_IS_OK(status)) {
-		return status;
-	}
+	/* send the reply up the chain */
+	req->async_states->status = status;
+	req->async_states->send_fn(req);
+}
 
-	status = pvfs_can_delete(pvfs, req, name, &lck);
-	if (!NT_STATUS_IS_OK(status)) {
-		return status;
+/*
+  setup for a unlink retry after a sharing violation
+  or a non granted oplock
+*/
+static NTSTATUS pvfs_unlink_setup_retry(struct ntvfs_module_context *ntvfs,
+					struct ntvfs_request *req,
+					union smb_unlink *io,
+					struct odb_lock *lck,
+					NTSTATUS status)
+{
+	struct pvfs_state *pvfs = ntvfs->private_data;
+	struct timeval end_time;
+
+	if (NT_STATUS_EQUAL(status, NT_STATUS_SHARING_VIOLATION)) {
+		end_time = timeval_add(&req->statistics.request_time,
+				       0, pvfs->sharing_violation_delay);
+	} else if (NT_STATUS_EQUAL(status, NT_STATUS_OPLOCK_NOT_GRANTED)) {
+		end_time = timeval_add(&req->statistics.request_time,
+				       pvfs->oplock_break_timeout, 0);
+	} else {
+		return NT_STATUS_INTERNAL_ERROR;
 	}
 
-	return pvfs_stream_delete(pvfs, name, -1);
+	return pvfs_odb_retry_setup(ntvfs, req, lck, end_time, io, NULL,
+				    pvfs_retry_unlink);
 }
 
 
 /*
-  unlink one file
+  unlink a file
 */
-static NTSTATUS pvfs_unlink_one(struct pvfs_state *pvfs, 
-				struct ntvfs_request *req,
-				const char *unix_path, 
-				const char *fname, uint32_t attrib)
+static NTSTATUS pvfs_unlink_file(struct pvfs_state *pvfs,
+				 struct pvfs_filename *name)
 {
-	struct pvfs_filename *name;
 	NTSTATUS status;
-	struct odb_lock *lck;
-
-	/* get a pvfs_filename object */
-	status = pvfs_resolve_partial(pvfs, req, 
-				      unix_path, fname, &name);
-	if (!NT_STATUS_IS_OK(status)) {
-		return status;
-	}
-
-	/* make sure its matches the given attributes */
-	status = pvfs_match_attrib(pvfs, name, attrib, 0);
-	if (!NT_STATUS_IS_OK(status)) {
-		talloc_free(name);
-		return status;
-	}
-
-	status = pvfs_can_delete(pvfs, req, name, &lck);
-	if (!NT_STATUS_IS_OK(status)) {
-		talloc_free(name);
-		return status;
-	}
 
 	if (name->dos.attrib & FILE_ATTRIBUTE_DIRECTORY) {
-		talloc_free(name);
 		return NT_STATUS_FILE_IS_A_DIRECTORY;
 	}
 
@@ -110,12 +133,57 @@ static NTSTATUS pvfs_unlink_one(struct pvfs_state *pvfs,
 			       name->full_name);
 	}
 
-	talloc_free(name);
-
 	return status;
 }
 
 /*
+  unlink one file
+*/
+static NTSTATUS pvfs_unlink_one(struct pvfs_state *pvfs,
+				struct ntvfs_request *req,
+				union smb_unlink *unl,
+				struct pvfs_filename *name)
+{
+	NTSTATUS status;
+	struct odb_lock *lck = NULL;
+
+	/* make sure its matches the given attributes */
+	status = pvfs_match_attrib(pvfs, name,
+				   unl->unlink.in.attrib, 0);
+	if (!NT_STATUS_IS_OK(status)) {
+		return status;
+	}
+
+	status = pvfs_can_delete(pvfs, req, name, &lck);
+
+	/*
+	 * on a sharing violation we need to retry when the file is closed by
+	 * the other user, or after 1 second
+	 * on a non granted oplock we need to retry when the file is closed by
+	 * the other user, or after 30 seconds
+	 */
+	if ((NT_STATUS_EQUAL(status, NT_STATUS_SHARING_VIOLATION) ||
+	     NT_STATUS_EQUAL(status, NT_STATUS_OPLOCK_NOT_GRANTED)) &&
+	    (req->async_states->state & NTVFS_ASYNC_STATE_MAY_ASYNC)) {
+		return pvfs_unlink_setup_retry(pvfs->ntvfs, req, unl, lck, status);
+	}
+
+	if (!NT_STATUS_IS_OK(status)) {
+		return status;
+	}
+
+	if (name->stream_name) {
+		if (!name->stream_exists) {
+			return NT_STATUS_OBJECT_NAME_NOT_FOUND;
+		}
+
+		return pvfs_stream_delete(pvfs, name, -1);
+	}
+
+	return pvfs_unlink_file(pvfs, name);
+}
+
+/*
   delete a file - the dirtype specifies the file types to include in the search. 
   The name can contain CIFS wildcards, but rarely does (except with OS/2 clients)
 */
@@ -147,8 +215,8 @@ NTSTATUS pvfs_unlink(struct ntvfs_module_context *ntvfs,
 		return NT_STATUS_FILE_IS_A_DIRECTORY;
 	}
 
-	if (name->stream_name) {
-		return pvfs_unlink_stream(pvfs, req, name, unl->unlink.in.attrib);
+	if (!name->has_wildcard) {
+		return pvfs_unlink_one(pvfs, req, unl, name);
 	}
 
 	/* get list of matching files */
@@ -158,6 +226,7 @@ NTSTATUS pvfs_unlink(struct ntvfs_module_context *ntvfs,
 	}
 
 	status = NT_STATUS_NO_SUCH_FILE;
+	talloc_free(name);
 
 	ofs = 0;
 
@@ -168,10 +237,20 @@ NTSTATUS pvfs_unlink(struct ntvfs_module_context *ntvfs,
 			return NT_STATUS_OBJECT_NAME_INVALID;
 		}
 
-		status = pvfs_unlink_one(pvfs, req, pvfs_list_unix_path(dir), fname, unl->unlink.in.attrib);
+		/* get a pvfs_filename object */
+		status = pvfs_resolve_partial(pvfs, req,
+					      pvfs_list_unix_path(dir),
+					      fname, &name);
+		if (!NT_STATUS_IS_OK(status)) {
+			return status;
+		}
+
+		status = pvfs_unlink_one(pvfs, req, unl, name);
 		if (NT_STATUS_IS_OK(status)) {
 			total_deleted++;
 		}
+
+		talloc_free(name);
 	}
 
 	if (total_deleted > 0) {
diff --git a/source4/ntvfs/posix/pvfs_wait.c b/source4/ntvfs/posix/pvfs_wait.c
index 989985a033..291250befd 100644
--- a/source4/ntvfs/posix/pvfs_wait.c
+++ b/source4/ntvfs/posix/pvfs_wait.c
@@ -31,7 +31,7 @@ struct pvfs_wait {
 	struct pvfs_wait *next, *prev;
 	struct pvfs_state *pvfs;
 	void (*handler)(void *, enum pvfs_wait_notice);
-	void *private;
+	void *private_data;
 	int msg_type;
 	struct messaging_context *msg_ctx;
 	struct event_context *ev;
@@ -45,20 +45,23 @@ struct pvfs_wait {
   previous ntvfs handlers in the chain (such as security context)
 */
 NTSTATUS pvfs_async_setup(struct ntvfs_module_context *ntvfs,
-			  struct ntvfs_request *req, void *private)
+			  struct ntvfs_request *req, void *private_data)
 {
-	struct pvfs_wait *pwait = private;
-	pwait->handler(pwait->private, pwait->reason);
+	struct pvfs_wait *pwait = talloc_get_type(private_data,
+						  struct pvfs_wait);
+	pwait->handler(pwait->private_data, pwait->reason);
 	return NT_STATUS_OK;
 }
 
 /*
   receive a completion message for a wait
 */
-static void pvfs_wait_dispatch(struct messaging_context *msg, void *private, uint32_t msg_type, 
+static void pvfs_wait_dispatch(struct messaging_context *msg,
+			       void *private_data, uint32_t msg_type,
 			       struct server_id src, DATA_BLOB *data)
 {
-	struct pvfs_wait *pwait = private;
+	struct pvfs_wait *pwait = talloc_get_type(private_data,
+						  struct pvfs_wait);
 	struct ntvfs_request *req;
 	void *p = NULL;
 
@@ -70,7 +73,7 @@ static void pvfs_wait_dispatch(struct messaging_context *msg, void *private, uin
 		pp = (void **)data->data;
 		p = *pp;
 	}
-	if (p == NULL || p != pwait->private) {
+	if (p == NULL || p != pwait->private_data) {
 		return;
 	}
 
@@ -91,9 +94,11 @@ static void pvfs_wait_dispatch(struct messaging_context *msg, void *private, uin
   receive a timeout on a message wait
 */
 static void pvfs_wait_timeout(struct event_context *ev, 
-			      struct timed_event *te, struct timeval t, void *private)
+			      struct timed_event *te, struct timeval t,
+			      void *private_data)
 {
-	struct pvfs_wait *pwait = talloc_get_type(private, struct pvfs_wait);
+	struct pvfs_wait *pwait = talloc_get_type(private_data,
+						  struct pvfs_wait);
 	struct ntvfs_request *req = pwait->req;
 
 	pwait->reason = PVFS_WAIT_TIMEOUT;
@@ -126,12 +131,12 @@ static int pvfs_wait_destructor(struct pvfs_wait *pwait)
   if msg_type == -1 then no message is registered, and it is assumed
   that the caller handles any messaging setup needed
 */
-void *pvfs_wait_message(struct pvfs_state *pvfs, 
-			struct ntvfs_request *req, 
-			int msg_type, 
-			struct timeval end_time,
-			void (*fn)(void *, enum pvfs_wait_notice),
-			void *private)
+struct pvfs_wait *pvfs_wait_message(struct pvfs_state *pvfs,
+				    struct ntvfs_request *req,
+				    int msg_type,
+				    struct timeval end_time,
+				    void (*fn)(void *, enum pvfs_wait_notice),
+				    void *private_data)
 {
 	struct pvfs_wait *pwait;
 
@@ -140,7 +145,7 @@ void *pvfs_wait_message(struct pvfs_state *pvfs,
 		return NULL;
 	}
 
-	pwait->private = private;
+	pwait->private_data = private_data;
 	pwait->handler = fn;
 	pwait->msg_ctx = pvfs->ntvfs->ctx->msg_ctx;
 	pwait->ev = pvfs->ntvfs->ctx->event_ctx;
diff --git a/source4/ntvfs/posix/pvfs_write.c b/source4/ntvfs/posix/pvfs_write.c
index 5a11f01952..dda8c83407 100644
--- a/source4/ntvfs/posix/pvfs_write.c
+++ b/source4/ntvfs/posix/pvfs_write.c
@@ -57,7 +57,10 @@ NTSTATUS pvfs_write(struct ntvfs_module_context *ntvfs,
 				 wr->writex.in.count,
 				 WRITE_LOCK);
 	NT_STATUS_NOT_OK_RETURN(status);
-	
+
+	status = pvfs_break_level2_oplocks(f);
+	NT_STATUS_NOT_OK_RETURN(status);
+
 	if (f->handle->name->stream_name) {
 		ret = pvfs_stream_write(pvfs,
 					f->handle,
diff --git a/source4/ntvfs/posix/vfs_posix.c b/source4/ntvfs/posix/vfs_posix.c
index e058e6f546..ca874d1db1 100644
--- a/source4/ntvfs/posix/vfs_posix.c
+++ b/source4/ntvfs/posix/vfs_posix.c
@@ -91,6 +91,10 @@ static void pvfs_setup_options(struct pvfs_state *pvfs)
 							PVFS_SHARE_DELAY,
 							PVFS_SHARE_DELAY_DEFAULT);
 
+	pvfs->oplock_break_timeout = share_int_option(scfg,
+						      PVFS_OPLOCK_TIMEOUT,
+						      PVFS_OPLOCK_TIMEOUT_DEFAULT);
+
 	pvfs->share_name = talloc_strdup(pvfs, scfg->name);
 
 	pvfs->fs_attribs = 
diff --git a/source4/ntvfs/posix/vfs_posix.h b/source4/ntvfs/posix/vfs_posix.h
index a660da329a..4d22a91714 100644
--- a/source4/ntvfs/posix/vfs_posix.h
+++ b/source4/ntvfs/posix/vfs_posix.h
@@ -28,6 +28,9 @@
 #include "ntvfs/common/ntvfs_common.h"
 #include "dsdb/samdb/samdb.h"
 
+struct pvfs_wait;
+struct pvfs_oplock;
+
 /* this is the private structure for the posix vfs backend. It is used
    to hold per-connection (per tree connect) state information */
 struct pvfs_state {
@@ -49,9 +52,12 @@ struct pvfs_state {
 	   ntcancel */
 	struct pvfs_wait *wait_list;
 
-	/* the sharing violation timeout */
+	/* the sharing violation timeout (nsecs) */
 	uint_t sharing_violation_delay;
 
+	/* the oplock break timeout (secs) */
+	uint_t oplock_break_timeout;
+
 	/* filesystem attributes (see FS_ATTR_*) */
 	uint32_t fs_attribs;
 
@@ -152,6 +158,13 @@ struct pvfs_file_handle {
 
 	bool have_opendb_entry;
 
+	/*
+	 * we need to wait for oplock break requests from other processes,
+	 * and we need to remember the pvfs_file so we can correctly
+	 * forward the oplock break to the client
+	 */
+	struct pvfs_oplock *oplock;
+
 	/* we need this hook back to our parent for lock destruction */
 	struct pvfs_state *pvfs;
 
@@ -228,10 +241,16 @@ struct pvfs_dir;
 /* types of notification for pvfs wait events */
 enum pvfs_wait_notice {PVFS_WAIT_EVENT, PVFS_WAIT_TIMEOUT, PVFS_WAIT_CANCEL};
 
+/*
+  state of a pending retry
+*/
+struct pvfs_odb_retry;
+
 #define PVFS_EADB			"posix:eadb"
 #define PVFS_XATTR			"posix:xattr"
 #define PVFS_FAKE_OPLOCKS		"posix:fakeoplocks"
 #define PVFS_SHARE_DELAY		"posix:sharedelay"
+#define PVFS_OPLOCK_TIMEOUT		"posix:oplocktimeout"
 #define PVFS_ALLOCATION_ROUNDING	"posix:allocationrounding"
 #define PVFS_SEARCH_INACTIVITY		"posix:searchinactivity"
 #define PVFS_ACL			"posix:acl"
@@ -239,7 +258,8 @@ enum pvfs_wait_notice {PVFS_WAIT_EVENT, PVFS_WAIT_TIMEOUT, PVFS_WAIT_CANCEL};
 
 #define PVFS_XATTR_DEFAULT			true
 #define PVFS_FAKE_OPLOCKS_DEFAULT		false
-#define PVFS_SHARE_DELAY_DEFAULT		1000000
+#define PVFS_SHARE_DELAY_DEFAULT		1000000 /* nsecs */
+#define PVFS_OPLOCK_TIMEOUT_DEFAULT		30 /* secs */
 #define PVFS_ALLOCATION_ROUNDING_DEFAULT	512
 #define PVFS_SEARCH_INACTIVITY_DEFAULT		300