r3012: added initial support for byte range locking in the posix vfs. This is

enough for us to pass locktest, but does not yet support lock timeouts
and some of the other esoteric features.
(This used to be commit 58a92abd88f190bc60894a68e0528e95ae33fe39)

16 files changed, 671 insertions, 31 deletions

diff --git a/source4/include/smb.h b/source4/include/smb.h
index 46b2dd03ca..745e90e4ff 100644
--- a/source4/include/smb.h
+++ b/source4/include/smb.h
@@ -611,6 +611,6 @@ typedef struct nt_user_token {
 #define REQ_CONTROL_ASYNC     (1<<2) /* the backend will answer this one later */
 
 /* passed to br lock code */
-enum brl_type {READ_LOCK, WRITE_LOCK, PENDING_LOCK};
+enum brl_type {READ_LOCK, WRITE_LOCK};
 
 #endif /* _SMB_H */
diff --git a/source4/include/smb_interfaces.h b/source4/include/smb_interfaces.h
index 4acdf51b56..e9f51ba4d1 100644
--- a/source4/include/smb_interfaces.h
+++ b/source4/include/smb_interfaces.h
@@ -1450,7 +1450,9 @@ union smb_lock {
 	/* generic interface */
 	struct {
 		enum smb_lock_level level;
-
+		struct {
+			uint16_t fnum;
+		} in;
 	} generic;
 
 	/* SMBlock interface */
diff --git a/source4/ntvfs/common/brlock.c b/source4/ntvfs/common/brlock.c
new file mode 100644
index 0000000000..0eb644e943
--- /dev/null
+++ b/source4/ntvfs/common/brlock.c
@@ -0,0 +1,418 @@
+/* 
+   Unix SMB/CIFS implementation.
+
+   generic byte range locking code
+
+   Copyright (C) Andrew Tridgell 1992-2004
+   Copyright (C) Jeremy Allison 1992-2000
+   
+   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 2 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, write to the Free Software
+   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+*/
+
+/* This module implements a tdb based byte range locking service,
+   replacing the fcntl() based byte range locking previously
+   used. This allows us to provide the same semantics as NT */
+
+#include "includes.h"
+
+struct brl_context {
+	struct tdb_wrap *w;
+	servid_t server;
+	uint16_t tid;
+};
+
+/*
+  in this module a "DATA_BLOB *file_key" is a blob that uniquely identifies
+  a file. For a local posix filesystem this will usually be a combination
+  of the device and inode numbers of the file, but it can be anything 
+  that uniquely idetifies a file for locking purposes, as long
+  as it is applied consistently.
+*/
+
+/*
+  the lock context contains the elements that define whether one
+  lock is the same as another lock
+*/
+struct lock_context {
+	servid_t server;
+	uint16_t smbpid;
+	uint16_t tid;
+};
+
+/* The data in brlock records is an unsorted linear array of these
+   records.  It is unnecessary to store the count as tdb provides the
+   size of the record */
+struct lock_struct {
+	struct lock_context context;
+	uint64_t start;
+	uint64_t size;
+	uint16_t fnum;
+	enum brl_type lock_type;
+};
+
+/*
+  Open up the brlock.tdb database. Close it down using
+  talloc_free()
+*/
+void *brl_init(TALLOC_CTX *mem_ctx, servid_t server, uint16_t tid)
+{
+	char *path;
+	struct brl_context *brl;
+
+	brl = talloc_p(mem_ctx, struct brl_context);
+	if (brl == NULL) {
+		return NULL;
+	}
+
+	path = lock_path(brl, "brlock.tdb");
+	brl->w = tdb_wrap_open(brl, path, 0,  
+			       TDB_DEFAULT|TDB_CLEAR_IF_FIRST,
+			       O_RDWR|O_CREAT, 0600);
+	talloc_free(path);
+	if (brl->w == NULL) {
+		talloc_free(brl);
+		return NULL;
+	}
+
+	brl->server = server;
+	brl->tid = tid;
+
+	return (void *)brl;
+}
+
+
+/*
+  see if two locking contexts are equal
+*/
+static BOOL brl_same_context(struct lock_context *ctx1, struct lock_context *ctx2)
+{
+	return (ctx1->server == ctx2->server &&
+		ctx1->smbpid == ctx2->smbpid &&
+		ctx1->tid == ctx2->tid);
+}
+
+/*
+ See if lock2 can be added when lock1 is in place.
+*/
+static BOOL brl_conflict(struct lock_struct *lck1, 
+			 struct lock_struct *lck2)
+{
+	if (lck1->lock_type == READ_LOCK && lck2->lock_type == READ_LOCK) {
+		return False;
+	}
+
+	if (brl_same_context(&lck1->context, &lck2->context) &&
+	    lck2->lock_type == READ_LOCK && lck1->fnum == lck2->fnum) {
+		return False;
+	}
+
+	if (lck1->start >= (lck2->start + lck2->size) ||
+	    lck2->start >= (lck1->start + lck1->size)) {
+		return False;
+	}
+	    
+	return True;
+} 
+
+
+/*
+ Check to see if this lock conflicts, but ignore our own locks on the
+ same fnum only.
+*/
+static BOOL brl_conflict_other(struct lock_struct *lck1, struct lock_struct *lck2)
+{
+	if (lck1->lock_type == READ_LOCK && lck2->lock_type == READ_LOCK) 
+		return False;
+
+	if (brl_same_context(&lck1->context, &lck2->context) &&
+	    lck1->fnum == lck2->fnum) {
+		return False;
+	}
+
+	if (lck1->start >= (lck2->start + lck2->size) ||
+	    lck2->start >= (lck1->start + lck1->size))
+		return False;
+	    
+	return True;
+} 
+
+
+
+/*
+ Lock a range of bytes.
+*/
+NTSTATUS brl_lock(void *brl_ctx,
+		  DATA_BLOB *file_key, 
+		  uint16_t smbpid,
+		  uint16_t fnum, 
+		  uint64_t start, uint64_t size, 
+		  enum brl_type lock_type)
+{
+	struct brl_context *brl = brl_ctx;
+	TDB_DATA kbuf, dbuf;
+	int count, i;
+	struct lock_struct lock, *locks;
+	char *tp;
+	NTSTATUS status;
+
+	kbuf.dptr = file_key->data;
+	kbuf.dsize = file_key->length;
+
+	if (tdb_chainlock(brl->w->tdb, kbuf) != 0) {
+		return NT_STATUS_INTERNAL_DB_CORRUPTION;
+	}
+
+	dbuf = tdb_fetch(brl->w->tdb, kbuf);
+
+	lock.context.smbpid = smbpid;
+	lock.context.server = brl->server;
+	lock.context.tid = brl->tid;
+	lock.start = start;
+	lock.size = size;
+	lock.fnum = fnum;
+	lock.lock_type = lock_type;
+
+	if (dbuf.dptr) {
+		/* there are existing locks - make sure they don't conflict */
+		locks = (struct lock_struct *)dbuf.dptr;
+		count = dbuf.dsize / sizeof(*locks);
+		for (i=0; i<count; i++) {
+			if (brl_conflict(&locks[i], &lock)) {
+				status = NT_STATUS_LOCK_NOT_GRANTED;
+				goto fail;
+			}
+		}
+	}
+
+	/* no conflicts - add it to the list of locks */
+	tp = Realloc(dbuf.dptr, dbuf.dsize + sizeof(*locks));
+	if (!tp) {
+		status = NT_STATUS_NO_MEMORY;
+		goto fail;
+	} else {
+		dbuf.dptr = tp;
+	}
+	memcpy(dbuf.dptr + dbuf.dsize, &lock, sizeof(lock));
+	dbuf.dsize += sizeof(lock);
+
+	if (tdb_store(brl->w->tdb, kbuf, dbuf, TDB_REPLACE) != 0) {
+		status = NT_STATUS_INTERNAL_DB_CORRUPTION;
+		goto fail;
+	}
+
+	free(dbuf.dptr);
+	tdb_chainunlock(brl->w->tdb, kbuf);
+	return NT_STATUS_OK;
+
+ fail:
+
+	free(dbuf.dptr);
+	tdb_chainunlock(brl->w->tdb, kbuf);
+	return status;
+}
+
+
+/*
+ Unlock a range of bytes.
+*/
+NTSTATUS brl_unlock(void *brl_ctx,
+		    DATA_BLOB *file_key, 
+		    uint16_t smbpid,
+		    uint16_t fnum,
+		    uint64_t start, uint64_t size)
+{
+	struct brl_context *brl = brl_ctx;
+	TDB_DATA kbuf, dbuf;
+	int count, i;
+	struct lock_struct *locks;
+	struct lock_context context;
+	NTSTATUS status;
+
+	kbuf.dptr = file_key->data;
+	kbuf.dsize = file_key->length;
+
+	if (tdb_chainlock(brl->w->tdb, kbuf) != 0) {
+		return NT_STATUS_INTERNAL_DB_CORRUPTION;
+	}
+
+	dbuf = tdb_fetch(brl->w->tdb, kbuf);
+	if (!dbuf.dptr) {
+		tdb_chainunlock(brl->w->tdb, kbuf);
+		return NT_STATUS_RANGE_NOT_LOCKED;
+	}
+
+	context.smbpid = smbpid;
+	context.server = brl->server;
+	context.tid = brl->tid;
+
+	/* there are existing locks - find a match */
+	locks = (struct lock_struct *)dbuf.dptr;
+	count = dbuf.dsize / sizeof(*locks);
+
+	locks = (struct lock_struct *)dbuf.dptr;
+	count = dbuf.dsize / sizeof(*locks);
+	for (i=0; i<count; i++) {
+		struct lock_struct *lock = &locks[i];
+		
+		if (brl_same_context(&lock->context, &context) &&
+		    lock->fnum == fnum &&
+		    lock->start == start &&
+		    lock->size == size) {
+			/* found it - delete it */
+			if (count == 1) {
+				if (tdb_delete(brl->w->tdb, kbuf) != 0) {
+					status = NT_STATUS_INTERNAL_DB_CORRUPTION;
+					goto fail;
+				}
+			} else {
+				if (i < count-1) {
+					memmove(&locks[i], &locks[i+1], 
+						sizeof(*locks)*((count-1) - i));
+				}
+				dbuf.dsize -= sizeof(*locks);
+				if (tdb_store(brl->w->tdb, kbuf, dbuf, TDB_REPLACE) != 0) {
+					status = NT_STATUS_INTERNAL_DB_CORRUPTION;
+					goto fail;
+				}
+			}
+			
+			free(dbuf.dptr);
+			tdb_chainunlock(brl->w->tdb, kbuf);
+			return NT_STATUS_OK;
+		}
+	}
+	
+	/* we didn't find it */
+	status = NT_STATUS_RANGE_NOT_LOCKED;
+
+ fail:
+	free(dbuf.dptr);
+	tdb_chainunlock(brl->w->tdb, kbuf);
+	return status;
+}
+
+
+/*
+  Test if we are allowed to perform IO on a region of an open file
+*/
+NTSTATUS brl_locktest(void *brl_ctx,
+		      DATA_BLOB *file_key, 
+		      uint16_t fnum,
+		      uint16 smbpid, 
+		      uint64_t start, uint64_t size, 
+		      enum brl_type lock_type)
+{
+	struct brl_context *brl = brl_ctx;
+	TDB_DATA kbuf, dbuf;
+	int count, i;
+	struct lock_struct lock, *locks;
+
+	kbuf.dptr = file_key->data;
+	kbuf.dsize = file_key->length;
+
+	dbuf = tdb_fetch(brl->w->tdb, kbuf);
+	if (dbuf.dptr == NULL) {
+		return NT_STATUS_OK;
+	}
+
+	lock.context.smbpid = smbpid;
+	lock.context.server = brl->server;
+	lock.context.tid = brl->tid;
+	lock.start = start;
+	lock.size = size;
+	lock.fnum = fnum;
+	lock.lock_type = lock_type;
+
+	/* there are existing locks - make sure they don't conflict */
+	locks = (struct lock_struct *)dbuf.dptr;
+	count = dbuf.dsize / sizeof(*locks);
+
+	for (i=0; i<count; i++) {
+		if (brl_conflict_other(&locks[i], &lock)) {
+			free(dbuf.dptr);
+			return NT_STATUS_FILE_LOCK_CONFLICT;
+		}
+	}
+
+	free(dbuf.dptr);
+	return NT_STATUS_OK;
+}
+
+
+/*
+ Remove any locks associated with a open file.
+*/
+NTSTATUS brl_close(void *brl_ctx,
+		   DATA_BLOB *file_key, int fnum)
+{
+	struct brl_context *brl = brl_ctx;
+	TDB_DATA kbuf, dbuf;
+	int count, i, dcount=0;
+	struct lock_struct *locks;
+	NTSTATUS status;
+
+	kbuf.dptr = file_key->data;
+	kbuf.dsize = file_key->length;
+
+	if (tdb_chainlock(brl->w->tdb, kbuf) != 0) {
+		return NT_STATUS_INTERNAL_DB_CORRUPTION;
+	}
+
+	dbuf = tdb_fetch(brl->w->tdb, kbuf);
+	if (!dbuf.dptr) {
+		tdb_chainunlock(brl->w->tdb, kbuf);
+		return NT_STATUS_OK;
+	}
+
+	/* there are existing locks - remove any for this fnum */
+	locks = (struct lock_struct *)dbuf.dptr;
+	count = dbuf.dsize / sizeof(*locks);
+
+	for (i=0; i<count; i++) {
+		struct lock_struct *lock = &locks[i];
+
+		if (lock->context.tid == brl->tid &&
+		    lock->context.server == brl->server &&
+		    lock->fnum == fnum) {
+			/* found it - delete it */
+			if (count > 1 && i < count-1) {
+				memmove(&locks[i], &locks[i+1], 
+					sizeof(*locks)*((count-1) - i));
+			}
+			count--;
+			i--;
+			dcount++;
+		}
+	}
+
+	status = NT_STATUS_OK;
+
+	if (count == 0) {
+		if (tdb_delete(brl->w->tdb, kbuf) != 0) {
+			status = NT_STATUS_INTERNAL_DB_CORRUPTION;
+		}
+	} else if (dcount != 0) {
+		dbuf.dsize -= dcount * sizeof(*locks);
+		if (tdb_store(brl->w->tdb, kbuf, dbuf, TDB_REPLACE) != 0) {
+			status = NT_STATUS_INTERNAL_DB_CORRUPTION;
+		}
+	}
+
+	free(dbuf.dptr);
+	tdb_chainunlock(brl->w->tdb, kbuf);
+
+	return status;
+}
+
diff --git a/source4/ntvfs/posix/config.mk b/source4/ntvfs/posix/config.mk
index 019288faaa..8ca5ad7b0b 100644
--- a/source4/ntvfs/posix/config.mk
+++ b/source4/ntvfs/posix/config.mk
@@ -19,6 +19,8 @@ ADD_OBJ_FILES = \
 		ntvfs/posix/pvfs_setfileinfo.o \
 		ntvfs/posix/pvfs_rename.o \
 		ntvfs/posix/pvfs_resolve.o \
-		ntvfs/posix/pvfs_shortname.o
+		ntvfs/posix/pvfs_shortname.o \
+		ntvfs/posix/pvfs_lock.o \
+		ntvfs/common/brlock.o
 # End MODULE ntvfs_posix
 ################################################
diff --git a/source4/ntvfs/posix/pvfs_lock.c b/source4/ntvfs/posix/pvfs_lock.c
new file mode 100644
index 0000000000..d7aca9df8b
--- /dev/null
+++ b/source4/ntvfs/posix/pvfs_lock.c
@@ -0,0 +1,151 @@
+/* 
+   Unix SMB/CIFS implementation.
+
+   POSIX NTVFS backend - locking
+
+   Copyright (C) Andrew Tridgell 2004
+
+   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 2 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, write to the Free Software
+   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+*/
+
+#include "include/includes.h"
+#include "vfs_posix.h"
+
+
+/*
+  check if we can perform IO on a range that might be locked
+*/
+NTSTATUS pvfs_check_lock(struct pvfs_state *pvfs,
+			 struct pvfs_file *f,
+			 uint16_t smbpid,
+			 uint64_t offset, uint64_t count,
+			 enum brl_type rw)
+{
+	if (!(pvfs->flags & PVFS_FLAG_STRICT_LOCKING)) {
+		return NT_STATUS_OK;
+	}
+
+	return brl_locktest(pvfs->brl_context,
+			    &f->locking_key,
+			    f->fnum,
+			    smbpid,
+			    offset, count, rw);
+}
+
+/*
+  lock or unlock a byte range
+*/
+NTSTATUS pvfs_lock(struct ntvfs_module_context *ntvfs,
+		   struct smbsrv_request *req, union smb_lock *lck)
+{
+	struct pvfs_state *pvfs = ntvfs->private_data;
+	struct pvfs_file *f;
+	struct smb_lock_entry *locks;
+	int i;
+	enum brl_type rw;
+
+	f = pvfs_find_fd(pvfs, req, lck->generic.in.fnum);
+	if (!f) {
+		return NT_STATUS_INVALID_HANDLE;
+	}
+
+	switch (lck->generic.level) {
+	case RAW_LOCK_LOCK:
+		return brl_lock(pvfs->brl_context,
+				&f->locking_key,
+				req->smbpid,
+				f->fnum,
+				lck->lock.in.offset,
+				lck->lock.in.count,
+				WRITE_LOCK);
+				
+	case RAW_LOCK_UNLOCK:
+		return brl_unlock(pvfs->brl_context,
+				  &f->locking_key,
+				  req->smbpid,
+				  f->fnum,
+				  lck->lock.in.offset,
+				  lck->lock.in.count);
+
+	case RAW_LOCK_GENERIC:
+		return NT_STATUS_INVALID_LEVEL;
+
+	case RAW_LOCK_LOCKX:
+		/* fall through to the most complex case */
+		break;
+	}
+
+	/* now the lockingX case, most common and also most complex */
+
+	if (lck->lockx.in.mode & LOCKING_ANDX_SHARED_LOCK) {
+		rw = READ_LOCK;
+	} else {
+		rw = WRITE_LOCK;
+	}
+
+	if (lck->lockx.in.mode & 
+	    (LOCKING_ANDX_OPLOCK_RELEASE |
+	     LOCKING_ANDX_CHANGE_LOCKTYPE |
+	     LOCKING_ANDX_CANCEL_LOCK)) {
+		/* todo: need to add support for these */
+		return NT_STATUS_NOT_IMPLEMENTED;
+	}
+
+
+	/* the unlocks happen first */
+	locks = lck->lockx.in.locks;
+
+	for (i=0;i<lck->lockx.in.ulock_cnt;i++) {
+		NTSTATUS status;
+		status = brl_unlock(pvfs->brl_context,
+				    &f->locking_key,
+				    locks[i].pid,
+				    f->fnum,
+				    locks[i].offset,
+				    locks[i].count);
+		if (!NT_STATUS_IS_OK(status)) {
+			return status;
+		}
+	}
+
+	locks += i;
+
+	for (i=0;i<lck->lockx.in.lock_cnt;i++) {
+		NTSTATUS status;
+
+		status = brl_lock(pvfs->brl_context,
+				  &f->locking_key,
+				  locks[i].pid,
+				  f->fnum,
+				  locks[i].offset,
+				  locks[i].count,
+				  rw);
+		if (!NT_STATUS_IS_OK(status)) {
+			/* undo the locks we just did */
+			for (i=i-1;i>=0;i--) {
+				brl_unlock(pvfs->brl_context,
+					   &f->locking_key,
+					   locks[i].pid,
+					   f->fnum,
+					   locks[i].offset,
+					   locks[i].count);
+			}
+			return status;
+		}
+	}
+
+	return NT_STATUS_OK;
+}
+
diff --git a/source4/ntvfs/posix/pvfs_open.c b/source4/ntvfs/posix/pvfs_open.c
index 29e57e5a08..51526461e0 100644
--- a/source4/ntvfs/posix/pvfs_open.c
+++ b/source4/ntvfs/posix/pvfs_open.c
@@ -51,6 +51,9 @@ struct pvfs_file *pvfs_find_fd(struct pvfs_state *pvfs,
 static int pvfs_fd_destructor(void *p)
 {
 	struct pvfs_file *f = p;
+
+	brl_close(f->pvfs->brl_context, &f->locking_key, f->fnum);
+
 	if (f->fd != -1) {
 		close(f->fd);
 		f->fd = -1;
@@ -71,6 +74,10 @@ NTSTATUS pvfs_open(struct ntvfs_module_context *ntvfs,
 	struct pvfs_filename *name;
 	struct pvfs_file *f;
 	NTSTATUS status;
+	struct {
+		dev_t device;
+		ino_t inode;
+	} lock_context;
 
 	if (io->generic.level != RAW_OPEN_GENERIC) {
 		return ntvfs_map_open(req, io, ntvfs);
@@ -161,6 +168,13 @@ do_open:
 	f->name = talloc_steal(f, name);
 	f->session = req->session;
 	f->smbpid = req->smbpid;
+	f->pvfs = pvfs;
+
+	/* we must zero here to take account of padding */
+	ZERO_STRUCT(lock_context);
+	lock_context.device = name->st.st_dev;
+	lock_context.inode = name->st.st_ino;
+	f->locking_key = data_blob_talloc(f, &lock_context, sizeof(lock_context));
 
 	/* setup a destructor to avoid file descriptor leaks on
 	   abnormal termination */
@@ -204,6 +218,11 @@ NTSTATUS pvfs_close(struct ntvfs_module_context *ntvfs,
 		return NT_STATUS_INVALID_HANDLE;
 	}
 
+	status = brl_close(pvfs->brl_context, &f->locking_key, f->fnum);
+	if (!NT_STATUS_IS_OK(status)) {
+		return status;
+	}
+	
 	if (close(f->fd) != 0) {
 		status = pvfs_map_errno(pvfs, errno);
 	} else {
diff --git a/source4/ntvfs/posix/pvfs_read.c b/source4/ntvfs/posix/pvfs_read.c
index e0c7e0b1f2..24142a81ee 100644
--- a/source4/ntvfs/posix/pvfs_read.c
+++ b/source4/ntvfs/posix/pvfs_read.c
@@ -32,6 +32,7 @@ NTSTATUS pvfs_read(struct ntvfs_module_context *ntvfs,
 	struct pvfs_state *pvfs = ntvfs->private_data;
 	ssize_t ret;
 	struct pvfs_file *f;
+	NTSTATUS status;
 
 	if (rd->generic.level != RAW_READ_READX) {
 		return NT_STATUS_NOT_SUPPORTED;
@@ -43,6 +44,14 @@ NTSTATUS pvfs_read(struct ntvfs_module_context *ntvfs,
 		return NT_STATUS_INVALID_HANDLE;
 	}
 
+	status = pvfs_check_lock(pvfs, f, req->smbpid, 
+				 rd->readx.in.offset,
+				 rd->readx.in.maxcnt,
+				 READ_LOCK);
+	if (!NT_STATUS_IS_OK(status)) {
+		return status;
+	}
+
 	ret = pread(f->fd, 
 		    rd->readx.out.data, 
 		    rd->readx.in.maxcnt,
diff --git a/source4/ntvfs/posix/pvfs_write.c b/source4/ntvfs/posix/pvfs_write.c
index a49f4fe947..80a3dae3a7 100644
--- a/source4/ntvfs/posix/pvfs_write.c
+++ b/source4/ntvfs/posix/pvfs_write.c
@@ -33,6 +33,7 @@ NTSTATUS pvfs_write(struct ntvfs_module_context *ntvfs,
 	struct pvfs_state *pvfs = ntvfs->private_data;
 	ssize_t ret;
 	struct pvfs_file *f;
+	NTSTATUS status;
 
 	switch (wr->generic.level) {
 	case RAW_WRITE_WRITEX:
@@ -40,6 +41,14 @@ NTSTATUS pvfs_write(struct ntvfs_module_context *ntvfs,
 		if (!f) {
 			return NT_STATUS_INVALID_HANDLE;
 		}
+		status = pvfs_check_lock(pvfs, f, req->smbpid, 
+					 wr->writex.in.offset,
+					 wr->writex.in.count,
+					 WRITE_LOCK);
+		if (!NT_STATUS_IS_OK(status)) {
+			return status;
+		}
+
 		ret = pwrite(f->fd, 
 			     wr->writex.in.data, 
 			     wr->writex.in.count,
@@ -62,6 +71,14 @@ NTSTATUS pvfs_write(struct ntvfs_module_context *ntvfs,
 			/* a truncate! */
 			ret = ftruncate(f->fd, wr->write.in.offset);
 		} else {
+			status = pvfs_check_lock(pvfs, f, req->smbpid, 
+						 wr->write.in.offset,
+						 wr->write.in.count,
+						 WRITE_LOCK);
+			if (!NT_STATUS_IS_OK(status)) {
+				return status;
+			}
+
 			ret = pwrite(f->fd,
 				     wr->write.in.data, 
 				     wr->write.in.count,
diff --git a/source4/ntvfs/posix/vfs_posix.c b/source4/ntvfs/posix/vfs_posix.c
index a17b90e369..5e7a605c9f 100644
--- a/source4/ntvfs/posix/vfs_posix.c
+++ b/source4/ntvfs/posix/vfs_posix.c
@@ -35,11 +35,12 @@ static void pvfs_setup_options(struct pvfs_state *pvfs)
 {
 	int snum = pvfs->tcon->service;
 
-	if (lp_map_hidden(snum)) pvfs->flags |= PVFS_FLAG_MAP_HIDDEN;
-	if (lp_map_archive(snum)) pvfs->flags |= PVFS_FLAG_MAP_ARCHIVE;
-	if (lp_map_system(snum)) pvfs->flags |= PVFS_FLAG_MAP_SYSTEM;
-	if (lp_readonly(snum)) pvfs->flags |= PVFS_FLAG_READONLY;
-	if (lp_strict_sync(snum)) pvfs->flags |= PVFS_FLAG_STRICT_SYNC;
+	if (lp_map_hidden(snum))     pvfs->flags |= PVFS_FLAG_MAP_HIDDEN;
+	if (lp_map_archive(snum))    pvfs->flags |= PVFS_FLAG_MAP_ARCHIVE;
+	if (lp_map_system(snum))     pvfs->flags |= PVFS_FLAG_MAP_SYSTEM;
+	if (lp_readonly(snum))       pvfs->flags |= PVFS_FLAG_READONLY;
+	if (lp_strict_sync(snum))    pvfs->flags |= PVFS_FLAG_STRICT_SYNC;
+	if (lp_strict_locking(snum)) pvfs->flags |= PVFS_FLAG_STRICT_LOCKING;
 
 	pvfs->share_name = talloc_strdup(pvfs, lp_servicename(snum));
 }
@@ -86,6 +87,13 @@ static NTSTATUS pvfs_connect(struct ntvfs_module_context *ntvfs,
 
 	ntvfs->private_data = pvfs;
 
+	pvfs->brl_context = brl_init(pvfs, 
+				     pvfs->tcon->smb_conn->connection->server_id,  
+				     pvfs->tcon->service);
+	if (pvfs->brl_context == NULL) {
+		return NT_STATUS_INTERNAL_DB_CORRUPTION;
+	}
+
 	status = pvfs_mangle_init(pvfs);
 	if (!NT_STATUS_IS_OK(status)) {
 		return status;
@@ -162,16 +170,6 @@ static NTSTATUS pvfs_seek(struct ntvfs_module_context *ntvfs,
 }
 
 /*
-  lock a byte range
-*/
-static NTSTATUS pvfs_lock(struct ntvfs_module_context *ntvfs,
-			  struct smbsrv_request *req, union smb_lock *lck)
-{
-	DEBUG(0,("pvfs_lock not implemented\n"));
-	return NT_STATUS_NOT_IMPLEMENTED;
-}
-
-/*
   set info on a pathname
 */
 static NTSTATUS pvfs_setpathinfo(struct ntvfs_module_context *ntvfs,
diff --git a/source4/ntvfs/posix/vfs_posix.h b/source4/ntvfs/posix/vfs_posix.h
index 6819d01529..e83f0479a9 100644
--- a/source4/ntvfs/posix/vfs_posix.h
+++ b/source4/ntvfs/posix/vfs_posix.h
@@ -51,6 +51,8 @@ struct pvfs_state {
 	struct pvfs_file *open_files;
 
 	struct pvfs_mangle_context *mangle_ctx;
+
+	void *brl_context;
 };
 
 
@@ -115,6 +117,12 @@ struct pvfs_file {
 	/* we need to remember the client pid that 
 	   opened the file so SMBexit works */
 	uint16_t smbpid;
+
+	/* a unique file key to be used for file locking */
+	DATA_BLOB locking_key;
+
+	/* we need this hook back to our parent for lock destruction */
+	struct pvfs_state *pvfs;
 };
 
 struct pvfs_mangle_context {
@@ -147,11 +155,12 @@ struct pvfs_mangle_context {
 #define PVFS_RESOLVE_STREAMS     (1<<1)
 
 /* flags in pvfs->flags */
-#define PVFS_FLAG_CI_FILESYSTEM (1<<0) /* the filesystem is case insensitive */
-#define PVFS_FLAG_MAP_ARCHIVE   (1<<1)
-#define PVFS_FLAG_MAP_SYSTEM    (1<<2)
-#define PVFS_FLAG_MAP_HIDDEN    (1<<3)
-#define PVFS_FLAG_READONLY      (1<<4)
-#define PVFS_FLAG_STRICT_SYNC   (1<<5)
+#define PVFS_FLAG_CI_FILESYSTEM  (1<<0) /* the filesystem is case insensitive */
+#define PVFS_FLAG_MAP_ARCHIVE    (1<<1)
+#define PVFS_FLAG_MAP_SYSTEM     (1<<2)
+#define PVFS_FLAG_MAP_HIDDEN     (1<<3)
+#define PVFS_FLAG_READONLY       (1<<4)
+#define PVFS_FLAG_STRICT_SYNC    (1<<5)
+#define PVFS_FLAG_STRICT_LOCKING (1<<6)
 
 #endif /* _VFS_POSIX_H_ */
diff --git a/source4/smbd/process_model.h b/source4/smbd/process_model.h
index 376b9a8ef8..79373d8a39 100644
--- a/source4/smbd/process_model.h
+++ b/source4/smbd/process_model.h
@@ -23,8 +23,6 @@
 #ifndef SAMBA_PROCESS_MODEL_H
 #define SAMBA_PROCESS_MODEL_H
 
-struct server_service_connection;
-
 /* modules can use the following to determine if the interface has changed
  * please increment the version number after each interface change
  * with a comment and maybe update struct process_model_critical_sizes.
diff --git a/source4/smbd/process_single.c b/source4/smbd/process_single.c
index 12a265b62f..b96a1d0b2c 100644
--- a/source4/smbd/process_single.c
+++ b/source4/smbd/process_single.c
@@ -49,7 +49,7 @@ static void single_accept_connection(struct event_context *ev, struct fd_event *
 		return;
 	}
 
-	conn = server_setup_connection(ev, server_socket, sock, t);
+	conn = server_setup_connection(ev, server_socket, sock, t, socket_get_fd(sock));
 	if (!conn) {
 		DEBUG(0,("server_setup_connection(ev, server_socket, sock, t) failed\n"));
 		return;
diff --git a/source4/smbd/process_standard.c b/source4/smbd/process_standard.c
index 122c6581b0..d70cfa676b 100644
--- a/source4/smbd/process_standard.c
+++ b/source4/smbd/process_standard.c
@@ -74,7 +74,7 @@ static void standard_accept_connection(struct event_context *ev, struct fd_event
 
 	set_need_random_reseed();
 
-	conn = server_setup_connection(ev, server_socket, sock, t);
+	conn = server_setup_connection(ev, server_socket, sock, t, getpid());
 	if (!conn) {
 		DEBUG(0,("server_setup_connection(ev, server_socket, sock, t) failed\n"));
 		return;
diff --git a/source4/smbd/process_thread.c b/source4/smbd/process_thread.c
index 55688f85e8..85f30c9ddd 100644
--- a/source4/smbd/process_thread.c
+++ b/source4/smbd/process_thread.c
@@ -79,7 +79,7 @@ static void thread_accept_connection(struct event_context *ev, struct fd_event *
 		return; 
 	}
 
-	conn = server_setup_connection(ev, server_socket, sock, t);
+	conn = server_setup_connection(ev, server_socket, sock, t, pthread_self());
 	if (!conn) {
 		DEBUG(0,("server_setup_connection(ev, server_socket, sock, t) failed\n"));
 		event_context_destroy(ev);
diff --git a/source4/smbd/service.c b/source4/smbd/service.c
index 5aae84b2c9..9a7ac73559 100644
--- a/source4/smbd/service.c
+++ b/source4/smbd/service.c
@@ -198,7 +198,8 @@ static int server_destructor(void *ptr)
 struct server_connection *server_setup_connection(struct event_context *ev, 
 						  struct server_socket *server_socket, 
 						  struct socket_context *sock, 
-						  time_t t)
+						  time_t t,
+						  servid_t server_id)
 {
 	struct fd_event fde;
 	struct timed_event idle;
@@ -206,7 +207,7 @@ struct server_connection *server_setup_connection(struct event_context *ev,
 
 	srv_conn = talloc_p(server_socket, struct server_connection);
 	if (!srv_conn) {
-		DEBUG(0,("talloc_p(mem_ctx, struct server_service_connection) failed\n"));
+		DEBUG(0,("talloc_p(mem_ctx, struct server_connection) failed\n"));
 		return NULL;
 	}
 
@@ -229,6 +230,7 @@ struct server_connection *server_setup_connection(struct event_context *ev,
 	srv_conn->server_socket		= server_socket;
 	srv_conn->service		= server_socket->service;
 	srv_conn->socket		= sock;
+	srv_conn->server_id		= server_id;
 
 	/* create a smb server context and add it to out event
 	   handling */
diff --git a/source4/smbd/service.h b/source4/smbd/service.h
index 88618964ce..e9ef0bff06 100644
--- a/source4/smbd/service.h
+++ b/source4/smbd/service.h
@@ -93,6 +93,19 @@ struct server_service {
 	struct server_context *srv_ctx;
 };
 
+/* the concept of whether two operations are on the same server
+   connection or different connections is an important one in SMB, especially
+   for locking and share modes. We will use a servid_t to distinguish different
+   connections 
+
+   this means that (for example) a unique open file is distinguished by the triple
+   of 
+      servid_t server;
+      uint16   tid;
+      uint16   fnum;
+*/
+typedef uint32_t servid_t;
+
 struct server_connection {
 	struct server_connection *next,*prev;
 	void *private_data;
@@ -104,6 +117,8 @@ struct server_connection {
 		time_t idle_time;
 	} event;
 
+	servid_t server_id;
+
 	struct socket_context *socket;
 
 	struct server_socket *server_socket;