Split off of POSIX locking into separate unit as Andrew requested.
authorJeremy Allison <jra@samba.org>
Fri, 28 Apr 2000 19:38:38 +0000 (19:38 +0000)
committerJeremy Allison <jra@samba.org>
Fri, 28 Apr 2000 19:38:38 +0000 (19:38 +0000)
Done so I don't lose my edits... :-).
Jeremy.
(This used to be commit 31a57be658f3fa3569a552e9c5d30174f5a51649)

source3/locking/posix.c [new file with mode: 0644]

diff --git a/source3/locking/posix.c b/source3/locking/posix.c
new file mode 100644 (file)
index 0000000..f05a0a2
--- /dev/null
@@ -0,0 +1,1208 @@
+/* 
+   Unix SMB/Netbios implementation.
+   Version 3.0
+   Locking functions
+   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.
+
+   Revision History:
+
+   POSIX locking support. Jeremy Allison (jeremy@valinux.com), Apr. 2000.
+*/
+
+#include "includes.h"
+extern int DEBUGLEVEL;
+extern int global_smbpid;
+
+/*
+ * The POSIX locking database handle.
+ */
+
+static TDB_CONTEXT *posix_lock_tdb;
+
+/*
+ * The pending close database handle.
+ */
+
+static TDB_CONTEXT *posix_pending_close_tdb;
+
+/*
+ * The data in POSIX lock 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 posix_lock {
+       int fd;
+       SMB_OFF_T start;
+       SMB_OFF_T size;
+       int lock_type;
+};
+
+/*
+ * The data in POSIX pending close records is an unsorted linear array of ints
+ * records.  It is unnecessary to store the count as tdb provides the
+ * size of the record.
+ */
+
+/* The key used in both the POSIX databases. */
+
+struct posix_lock_key {
+       SMB_DEV_T device;
+       SMB_INO_T inode;
+}; 
+
+/*******************************************************************
+ Form a static locking key for a dev/inode pair.
+******************************************************************/
+
+static TDB_DATA locking_key(SMB_DEV_T dev, SMB_INO_T inode)
+{
+       static struct posix_lock_key key;
+       TDB_DATA kbuf;
+       key.dev = dev;
+       key.inode = inode;
+       kbuf.dptr = (char *)&key;
+       kbuf.dsize = sizeof(key);
+       return kbuf;
+}
+
+/*******************************************************************
+ Convenience function to get a key from an fsp.
+******************************************************************/
+
+static TDB_DATA locking_key_fsp(files_struct *fsp)
+{
+       return locking_key(fsp->dev, fsp->inode);
+}
+
+/****************************************************************************
+ Add an fd to the pending close tdb.
+****************************************************************************/
+
+static BOOL add_fd_to_close_entry(files_struct *fsp)
+{
+       struct posix_lock_key = locking_key_fsp(fsp);
+       TDB_DATA kbuf, dbuf;
+       size_t count = 0;
+       int *fd_array = NULL;
+
+       dbuf.dptr = NULL;
+
+       tdb_lockchain(posix_pending_close_tdb, kbuf);
+       dbuf = tdb_fetch(posix_pending_close_tdb, kbuf);
+
+       dbuf.dptr = Realloc(dbuf.dptr, dbuf.dsize + sizeof(int));
+       if (!dbuf.dptr) {
+               DEBUG(0,("add_fd_to_close_entry: Realloc fail !\n"));
+               tdb_unlockchain(posix_pending_close_tdb, kbuf);
+               return False;
+       }
+       memcpy(dbuf.dptr + dbuf.dsize, &fsp->fd, sizeof(int));
+       dbuf.dsize += sizeof(int);
+
+       if (tdb_store(posix_pending_close_tdb, kbuf, dbuf, TDB_REPLACE) == -1) {
+               DEBUG(0,("add_fd_to_close_entry: tdb_store fail !\n"));
+       }
+
+       free(dbuf.dptr);
+       tdb_unlockchain(posix_pending_close_tdb, kbuf);
+       return True;
+}
+
+/****************************************************************************
+ Remove all fd entries for a specific dev/inode pair from the tdb.
+****************************************************************************/
+
+static void delete_close_entries(files_struct *fsp)
+{
+    struct posix_lock_key = locking_key_fsp(fsp);
+    TDB_DATA kbuf, dbuf;
+
+       tdb_lockchain(posix_pending_close_tdb, kbuf);
+       if (tdb_delete(posix_pending_close_tdb, kbuf) == -1)
+               DEBUG(0,("delete_close_entries: tdb_delete fail !\n"));
+       tdb_unlockchain(posix_pending_close_tdb, kbuf);
+}
+
+/****************************************************************************
+ Get the array of POSIX pending close records for an open fsp. Caller must
+ free. Returns number of entries.
+****************************************************************************/
+
+static size_t get_posix_pending_close_entries(files_struct *fsp, int **entries)
+{
+       struct posix_lock_key = locking_key_fsp(fsp);
+       TDB_DATA kbuf, dbuf;
+       size_t count = 0;
+
+       *entries = NULL;
+       dbuf.dptr = NULL;
+
+       tdb_lockchain(posix_pending_close_tdb, kbuf);
+       dbuf = tdb_fetch(posix_pending_close_tdb, kbuf);
+
+    if (!dbuf.dptr) {
+               tdb_unlockchain(posix_pending_close_tdb, kbuf);
+               return 0;
+       }
+
+       *entries = (int *)dbuf.dptr;
+       count = (size_t)(dbuf.dsize / sizeof(int));
+
+       tdb_unlockchain(posix_pending_close_tdb, kbuf);
+
+       return count;
+}
+
+/****************************************************************************
+ Get the array of POSIX locks for an fsp. Caller must free. Returns
+ number of entries.
+****************************************************************************/
+
+static size_t get_posix_lock_entries(files_struct *fsp, struct posix_lock **entries)
+{
+       struct posix_lock_key = locking_key_fsp(fsp);
+       TDB_DATA kbuf, dbuf;
+       size_t count = 0;
+
+       *entries = NULL;
+
+       dbuf.dptr = NULL;
+
+       tdb_lockchain(posix_lock_tdb, kbuf);
+       dbuf = tdb_fetch(posix_lock_tdb, kbuf);
+
+    if (!dbuf.dptr) {
+               tdb_unlockchain(posix_lock_tdb, kbuf);
+               return 0;
+       }
+
+       *entries = (struct posix_lock_struct *)dbuf.dptr;
+       count = (size_t)(dbuf.dsize / sizeof(struct posix_lock_struct));
+
+    tdb_unlockchain(posix_lock_tdb, kbuf);
+
+       return count;
+}
+
+/****************************************************************************
+ Deal with pending closes needed by POSIX locking support.
+****************************************************************************/
+
+int fd_close_posix(struct connection_struct *conn, files_struct *fsp)
+{
+       int saved_errno = 0;
+       int ret;
+       size_t count, i;
+       struct posix_lock *entries = NULL;
+       int *fd_array = NULL;
+
+       if (!lp_posix_locking(SNUM(conn))) {
+               /*
+                * No POSIX to worry about, just close.
+                */
+               ret = conn->vfs_ops.close(fsp->fd);
+               fsp->fd = -1;
+               return ret;
+       }
+
+       /*
+        * Get the number of outstanding POSIX locks on this dev/inode pair.
+        */
+
+       count = get_posix_lock_entries(fsp, &entries);
+       
+       if (count) {
+
+               /*
+                * There are outstanding locks on this dev/inode pair on other fds.
+                * Add our fd to the pending close tdb and set fsp->fd to -1.
+                */
+
+               if (!add_fd_to_close_entry(fsp)) {
+                       free((char *)entries);
+                       return False;
+               }
+
+               free((char *)entries);
+               fsp->fd = -1;
+               return 0;
+       }
+
+       if(entries)
+               free((char *)entries);
+
+       /*
+        * No outstanding POSIX locks. Get the pending close fd's
+        * from the tdb and close them all.
+        */
+
+       count = get_posix_pending_close_entries(fsp, &fd_array)
+
+       if (count) {
+               DEBUG(10,("fd_close_posix: doing close on %u fd's.\n", (unsigned int)count ));
+
+               for(i = 0; i < count; i++) {
+                       if (conn->vfs_ops.close(fd_array[i]) == -1) {
+                               saved_errno = errno;
+                       }
+               }
+
+               if (fd_array)
+                       free((char *)fd_array);
+
+               /*
+                * Delete all fd's stored in the tdb
+                * for this dev/inode pair.
+                */
+
+               delete_close_entries(fsp);
+       }
+
+       if (fd_array)
+               free((char *)fd_array);
+
+       /*
+        * Finally close the fd associated with this fsp.
+        */
+
+       ret = conn->vfs_ops.close(fsp->fd);
+
+       if (saved_errno != 0) {
+        errno = saved_errno;
+               ret = -1;
+    } 
+
+       fsp->fd = -1;
+
+       return ret;
+}
+
+/****************************************************************************
+ Debugging aid :-).
+****************************************************************************/
+
+static const char *posix_lock_type_name(int lock_type)
+{
+       return (lock_type == F_RDLCK) ? "READ" : "WRITE";
+}
+
+/****************************************************************************
+ Add an entry into the POSIX locking tdb.
+****************************************************************************/
+
+static BOOL add_posix_lock_entry(files_struct *fsp, SMB_OFF_T start, SMB_OFF_T size, int lock_type)
+{
+       struct posix_lock_key = locking_key_fsp(fsp);
+       TDB_DATA kbuf, dbuf;
+       struct posix_lock pl;
+
+       /*
+        * Now setup the new record.
+        */
+
+       pl.fd = fsp->fd;
+       pl.start = start;
+       pl.size = size;
+       pl.lock_type = lock_type;
+
+       dbuf.dptr = NULL;
+
+       tdb_lockchain(posix_lock_tdb, kbuf);
+       dbuf = tdb_fetch(posix_lock_tdb, kbuf);
+
+       dbuf.dptr = Realloc(dbuf.dptr, dbuf.dsize + sizeof(*pl));
+       if (!dbuf.dptr) {
+               DEBUG(0,("add_posix_lock_entry: Realloc fail !\n"));
+               goto fail;
+       }
+
+       memcpy(dbuf.dptr + dbuf.dsize, rec, sizeof(*pl));
+       dbuf.dsize += sizeof(*pl);
+
+       if (tdb_store(posix_lock_tdb, kbuf, dbuf, TDB_REPLACE) == -1) {
+               DEBUG(0,("add_posix_lock: Failed to add lock entry on file %s\n", fsp->fsp_name));
+               goto fail;
+       }
+
+    free(dbuf.dptr);
+    tdb_unlockchain(posix_lock_tdb, kbuf);
+
+       DEBUG(10,("add_posix_lock: File %s: type = %s: start=%.0f size=%.0f:dev=%.0f inode=%.0f\n",
+                       fsp->fsp_name, posix_lock_type_name(lock_type), (double)start, (double)size,
+                       (double)fsp->dev, (double)fsp->inode ));
+
+    return True;
+
+ fail:
+    if (dbuf.dptr)
+               free(dbuf.dptr);
+    tdb_unlockchain(tdb, kbuf);
+    return False;
+}
+
+/****************************************************************************
+ Delete an entry from the POSIX locking tdb.
+****************************************************************************/
+
+static BOOL delete_posix_lock_entry(files_struct *fsp, SMB_OFF_T start, SMB_OFF_T size)
+{
+       struct posix_lock_key = locking_key_fsp(fsp);
+       TDB_DATA kbuf, dbuf;
+       struct posix_lock *locks;
+       size_t i, count;
+
+       dbuf.dptr = NULL;
+
+       tdb_lockchain(posix_lock_tdb, kbuf);
+       dbuf = tdb_fetch(posix_lock_tdb, kbuf);
+
+       if (!dbuf.dptr) {
+               DEBUG(10,("delete_posix_lock_entry: tdb_fetch failed !\n"));
+               goto fail;
+       }
+
+       /* There are existing locks - find a match. */
+       locks = (struct lock_struct *)dbuf.dptr;
+       count = (size_t(dbuf.dsize / sizeof(*locks));
+
+       for (i=0; i<count; i++) { 
+               struct posix_lock *pl = &locks[i];
+
+               if (pl->fd == fd &&
+                       pl->start == start &&
+                       pl->size == size) {
+                       /* Found it - delete it. */
+                       if (count == 1) {
+                               tdb_delete(posix_lock_tdb, kbuf);
+                       } else {
+                               if (i < count-1) {
+                                       memmove(&locks[i], &locks[i+1], sizeof(*locks)*((count-1) - i));
+                               }
+                               dbuf.dsize -= sizeof(*locks);
+                               tdb_store(tdb, kbuf, dbuf, TDB_REPLACE);
+                       }
+
+                       free(dbuf.dptr);
+                       tdb_unlockchain(tdb, kbuf);
+                       return True;
+               }
+       }
+
+       /* We didn't find it. */
+
+ fail:
+    if (dbuf.dptr)
+               free(dbuf.dptr);
+    tdb_unlockchain(tdb, kbuf);
+    return False;
+}
+
+/****************************************************************************
+ Utility function to map a lock type correctly depending on the open
+ mode of a file.
+****************************************************************************/
+
+static int map_posix_lock_type( files_struct *fsp, enum brl_type lock_type)
+{
+       if((lock_type == WRITE_LOCK) && !fsp->can_write) {
+               /*
+                * Many UNIX's cannot get a write lock on a file opened read-only.
+                * Win32 locking semantics allow this.
+                * Do the best we can and attempt a read-only lock.
+                */
+               DEBUG(10,("map_posix_lock_type: Downgrading write lock to read due to read-only file.\n"));
+               return F_RDLCK;
+       } else if((lock_type == READ_LOCK) && !fsp->can_read) {
+               /*
+                * Ditto for read locks on write only files.
+                */
+               DEBUG(10,("map_posix_lock_type: Changing read lock to write due to write-only file.\n"));
+               return F_WRLCK;
+       }
+
+  /*
+   * This return should be the most normal, as we attempt
+   * to always open files read/write.
+   */
+
+  return (lock_type == READ_LOCK) ? F_RDLCK : F_WRLCK;
+}
+
+/****************************************************************************
+ Check to see if the given unsigned lock range is within the possible POSIX
+ range. Modifies the given args to be in range if possible, just returns
+ False if not.
+****************************************************************************/
+
+static BOOL posix_lock_in_range(SMB_OFF_T *offset_out, SMB_OFF_T *count_out,
+                                                               SMB_BIG_UINT u_offset, SMB_BIG_UINT u_count)
+{
+       SMB_OFF_T offset;
+       SMB_OFF_T count;
+
+#if defined(LARGE_SMB_OFF_T) && !defined(HAVE_BROKEN_FCNTL64_LOCKS)
+
+    SMB_OFF_T mask2 = ((SMB_OFF_T)0x4) << (SMB_OFF_T_BITS-4);
+    SMB_OFF_T mask = (mask2<<1);
+    SMB_OFF_T neg_mask = ~mask;
+
+       /*
+        * In this case SMB_OFF_T is 64 bits,
+        * and the underlying system can handle 64 bit signed locks.
+        * Cast to signed type.
+        */
+
+       offset = (SMB_OFF_T)u_offset;
+       count = (SMB_OFF_T)u_count;
+
+       /*
+        * Deal with a very common case of count of all ones.
+        * (lock entire file).
+        */
+
+       if(count == (SMB_OFF_T)-1)
+               count &= ~mask;
+
+       /*
+        * POSIX lock ranges cannot be negative.
+        * Fail if any combination becomes negative.
+        */
+
+       if(offset < 0 || count < 0 || (offset + count < 0)) {
+               DEBUG(10,("posix_lock_in_range: negative range: offset = %.0f, count = %.0f. Ignoring lock.\n",
+                               (double)offset, (double)count ));
+               return False;
+       }
+
+       /*
+        * In this case SMB_OFF_T is 64 bits, the offset and count
+        * fit within the positive range, and the underlying
+        * system can handle 64 bit locks. Just return as the
+        * cast values are ok.
+        */
+
+#else /* !LARGE_SMB_OFF_T || HAVE_BROKEN_FCNTL64_LOCKS */
+
+       /*
+        * In this case either SMB_OFF_T is 32 bits,
+        * or the underlying system cannot handle 64 bit signed locks.
+        * Either way we have to try and mangle to fit within 31 bits.
+        * This is difficult.
+        */
+
+#if defined(HAVE_BROKEN_FCNTL64_LOCKS)
+
+       /*
+        * SMB_OFF_T is 64 bits, but we need to use 31 bits due to
+        * broken large locking.
+        */
+
+       /*
+        * Deal with a very common case of count of all ones.
+        * (lock entire file).
+        */
+
+       if(u_count == (SMB_BIG_UINT)-1)
+               count = 0x7FFFFFFF;
+
+       if(((u_offset >> 32) & 0xFFFFFFFF) || ((u_count >> 32) & 0xFFFFFFFF)) {
+               DEBUG(10,("posix_lock_in_range: top 32 bits not zero. offset = %.0f, count = %.0f. Ignoring lock.\n",
+                               (double)u_offset, (double)u_count ));
+               /* Top 32 bits of offset or count were not zero. */
+               return False;
+       }
+
+       /* Cast from 64 bits unsigned to 64 bits signed. */
+       offset = (SMB_OFF_T)u_offset;
+       count = (SMB_OFF_T)u_count;
+
+       /*
+        * Check if we are within the 2^31 range.
+        */
+
+       {
+               int32 low_offset = (int32)offset;
+               int32 low_count = (int32)count;
+
+               if(low_offset < 0 || low_count < 0 || (low_offset + low_count < 0)) {
+                       DEBUG(10,("posix_lock_in_range: not within 2^31 range. low_offset = %d, low_count = %d. Ignoring lock.\n",
+                                       low_offset, low_count ));
+                       return False;
+               }
+       }
+
+       /*
+        * Ok - we can map from a 64 bit number to a 31 bit lock.
+        */
+
+#else /* HAVE_BROKEN_FCNTL64_LOCKS */
+
+       /*
+        * SMB_OFF_T is 32 bits.
+        */
+
+#if defined(HAVE_LONGLONG)
+
+       /*
+        * SMB_BIG_UINT is 64 bits, we can do a 32 bit shift.
+        */
+
+       /*
+        * Deal with a very common case of count of all ones.
+        * (lock entire file).
+        */
+
+       if(u_count == (SMB_BIG_UINT)-1)
+               count = 0x7FFFFFFF;
+
+       if(((u_offset >> 32) & 0xFFFFFFFF) || ((u_count >> 32) & 0xFFFFFFFF)) {
+               DEBUG(10,("posix_lock_in_range: top 32 bits not zero. u_offset = %.0f, u_count = %.0f. Ignoring lock.\n",
+                               (double)u_offset, (double)u_count ));
+               return False;
+       }
+
+       /* Cast from 64 bits unsigned to 32 bits signed. */
+       offset = (SMB_OFF_T)u_offset;
+       count = (SMB_OFF_T)u_count;
+
+       /*
+        * Check if we are within the 2^31 range.
+        */
+
+       if(offset < 0 || count < 0 || (offset + count < 0)) {
+               DEBUG(10,("posix_lock_in_range: not within 2^31 range. offset = %d, count = %d. Ignoring lock.\n",
+                               (int)offset, (int)count ));
+               return False;
+       }
+
+#else /* HAVE_LONGLONG */
+
+       /*
+        * SMB_BIG_UINT and SMB_OFF_T are both 32 bits,
+        * just cast.
+        */
+
+       /*
+        * Deal with a very common case of count of all ones.
+        * (lock entire file).
+        */
+
+       if(u_count == (SMB_BIG_UINT)-1)
+               count = 0x7FFFFFFF;
+
+       /* Cast from 32 bits unsigned to 32 bits signed. */
+       offset = (SMB_OFF_T)u_offset;
+       count = (SMB_OFF_T)u_count;
+
+       /*
+        * Check if we are within the 2^31 range.
+        */
+
+       if(offset < 0 || count < 0 || (offset + count < 0)) {
+               DEBUG(10,("posix_lock_in_range: not within 2^31 range. offset = %d, count = %d. Ignoring lock.\n",
+                               (int)offset, (int)count ));
+               return False;
+       }
+
+#endif /* HAVE_LONGLONG */
+#endif /* LARGE_SMB_OFF_T */
+#endif /* !LARGE_SMB_OFF_T || HAVE_BROKEN_FCNTL64_LOCKS */
+
+       /*
+        * The mapping was successful.
+        */
+
+       DEBUG(10,("posix_lock_in_range: offset_out = %.0f, count_out = %.0f\n",
+                       (double)offset, (double)count ));
+
+       *offset_out = offset;
+       *count_out = count;
+       
+       return True;
+}
+
+/****************************************************************************
+ POSIX function to see if a file region is locked. Returns True if the
+ region is locked, False otherwise.
+****************************************************************************/
+
+BOOL is_posix_locked(files_struct *fsp, SMB_BIG_UINT u_offset, SMB_BIG_UINT u_count, enum brl_type lock_type)
+{
+       SMB_OFF_T offset;
+       SMB_OFF_T count;
+       int posix_lock_type = map_posix_lock_type(fsp,lock_type);
+
+       DEBUG(10,("is_posix_locked: File %s, offset = %.0f, count = %.0f, type = %s\n",
+                       fsp->fsp_name, (double)u_offset, (double)u_count, lock_type_name(lock_type) ));
+
+       /*
+        * If the requested lock won't fit in the POSIX range, we will
+        * never set it, so presume it is not locked.
+        */
+
+       if(!posix_lock_in_range(&offset, &count, u_offset, u_count))
+               return False;
+
+       /*
+        * Note that most UNIX's can *test* for a write lock on
+        * a read-only fd, just not *set* a write lock on a read-only
+        * fd. So we don't need to use map_lock_type here.
+        */ 
+
+       return fcntl_lock(fsp->fd,SMB_F_GETLK,offset,count,posix_lock_type);
+}
+
+/****************************************************************************
+ POSIX function to acquire a lock. Returns True if the
+ lock could be granted, False if not.
+****************************************************************************/
+
+BOOL set_posix_lock(files_struct *fsp, SMB_BIG_UINT u_offset, SMB_BIG_UINT u_count, enum brl_type lock_type)
+{
+       SMB_OFF_T offset;
+       SMB_OFF_T count;
+       BOOL ret = True;
+       int posix_lock_type = map_posix_lock_type(fsp,lock_type);
+
+       DEBUG(5,("set_posix_lock: File %s, offset = %.0f, count = %.0f, type = %s\n",
+                       fsp->fsp_name, (double)u_offset, (double)u_count, lock_type_name(lock_type) ));
+
+       /*
+        * If the requested lock won't fit in the POSIX range, we will
+        * pretend it was successful.
+        */
+
+       if(!posix_lock_in_range(&offset, &count, u_offset, u_count))
+               return True;
+
+       /*
+        * Note that setting multiple overlapping read locks on different
+        * file descriptors will not be held separately by the kernel (POSIX
+        * braindamage), but will be merged into one continuous read lock
+        * range. We cope with this case in the release_posix_lock code
+        * below. JRA.
+        */
+
+    ret = fcntl_lock(fsp->fd,SMB_F_SETLK,offset,count,posix_lock_type);
+
+       if (ret)
+               add_posix_lock_entry(fsp,offset,count,posix_lock_type);
+
+       return ret;
+}
+
+/*
+ * Structure used when splitting a lock range
+ * into a POSIX lock range. Doubly linked list.
+ */
+
+struct unlock_list {
+    struct unlock_list *next;
+    struct unlock_list *prev;
+    SMB_OFF_T start;
+    SMB_OFF_T size;
+       int fd;
+};
+
+/****************************************************************************
+ Create a list of lock ranges that don't overlap a given range. Used in calculating
+ POSIX lock unlocks. This is a difficult function that requires ASCII art to
+ understand it :-).
+****************************************************************************/
+
+static struct unlock_list *posix_unlock_list(TALLOC_CTX *ctx, struct unlock_list *ulhead, files_struct *fsp)
+{
+       struct lock_key key;
+       TDB_DATA kbuf, dbuf;
+       struct lock_struct *locks;
+       int num_locks, i;
+
+       /*
+        * Setup the key for this fetch.
+        */
+       key.device = dev;
+       key.inode = ino;
+       kbuf.dptr = (char *)&key;
+       kbuf.dsize = sizeof(key);
+
+       dbuf.dptr = NULL;
+
+       tdb_lockchain(tdb, kbuf);
+       dbuf = tdb_fetch(tdb, kbuf);
+
+       if (!dbuf.dptr) {
+               tdb_unlockchain(tdb, kbuf);
+               return ulhead;
+       }
+       
+       locks = (struct lock_struct *)dbuf.dptr;
+       num_locks = dbuf.dsize / sizeof(*locks);
+
+       /*
+        * Check the current lock list on this dev/inode pair.
+        * Quit if the list is deleted.
+        */
+
+       DEBUG(10,("brl_unlock_list: curr: start=%.0f,size=%.0f\n",
+               (double)ulhead->start, (double)ulhead->size ));
+
+       for (i=0; i<num_locks && ulhead; i++) {
+
+               struct lock_struct *lock = &locks[i];
+               struct unlock_list *ul_curr;
+
+               /* If it's not this process, ignore it. */
+               if (lock->context.pid != pid)
+                       continue;
+
+               /*
+                * Walk the unlock list, checking for overlaps. Note that
+                * the unlock list can expand within this loop if the current
+                * range being examined needs to be split.
+                */
+
+               for (ul_curr = ulhead; ul_curr;) {
+
+                       DEBUG(10,("brl_unlock_list: lock: start=%.0f,size=%.0f:",
+                               (double)lock->start, (double)lock->size ));
+
+                       if ( (ul_curr->start >= (lock->start + lock->size)) ||
+                                (lock->start > (ul_curr->start + ul_curr->size))) {
+
+                               /* No overlap with this lock - leave this range alone. */
+/*********************************************
+                                             +---------+
+                                             | ul_curr |
+                                             +---------+
+                                +-------+
+                                | lock  |
+                                +-------+
+OR....
+             +---------+
+             | ul_curr |
+             +---------+
+**********************************************/
+
+                               DEBUG(10,("no overlap case.\n" ));
+
+                               ul_curr = ul_curr->next;
+
+                       } else if ( (ul_curr->start >= lock->start) &&
+                                               (ul_curr->start + ul_curr->size <= lock->start + lock->size) ) {
+
+                               /*
+                                * This unlock is completely overlapped by this existing lock range
+                                * and thus should have no effect (not be unlocked). Delete it from the list.
+                                */
+/*********************************************
+                +---------+
+                | ul_curr |
+                +---------+
+        +---------------------------+
+        |       lock                |
+        +---------------------------+
+**********************************************/
+                               /* Save the next pointer */
+                               struct unlock_list *ul_next = ul_curr->next;
+
+                               DEBUG(10,("delete case.\n" ));
+
+                               DLIST_REMOVE(ulhead, ul_curr);
+                               if(ulhead == NULL)
+                                       break; /* No more list... */
+
+                               ul_curr = ul_next;
+                               
+                       } else if ( (ul_curr->start >= lock->start) &&
+                                               (ul_curr->start < lock->start + lock->size) &&
+                                               (ul_curr->start + ul_curr->size > lock->start + lock->size) ) {
+
+                               /*
+                                * This unlock overlaps the existing lock range at the high end.
+                                * Truncate by moving start to existing range end and reducing size.
+                                */
+/*********************************************
+                +---------------+
+                | ul_curr       |
+                +---------------+
+        +---------------+
+        |    lock       |
+        +---------------+
+BECOMES....
+                        +-------+
+                        |ul_curr|
+                        +-------+
+**********************************************/
+
+                               ul_curr->size = (ul_curr->start + ul_curr->size) - (lock->start + lock->size);
+                               ul_curr->start = lock->start + lock->size;
+
+                               DEBUG(10,("truncate high case: start=%.0f,size=%.0f\n",
+                                                               (double)ul_curr->start, (double)ul_curr->size ));
+
+                               ul_curr = ul_curr->next;
+
+                       } else if ( (ul_curr->start < lock->start) &&
+                                               (ul_curr->start + ul_curr->size > lock->start) ) {
+
+                               /*
+                                * This unlock overlaps the existing lock range at the low end.
+                                * Truncate by reducing size.
+                                */
+/*********************************************
+   +---------------+
+   | ul_curr       |
+   +---------------+
+           +---------------+
+           |    lock       |
+           +---------------+
+BECOMES....
+   +-------+
+   |ul_curr|
+   +-------+
+**********************************************/
+
+                               ul_curr->size = lock->start - ul_curr->start;
+
+                               DEBUG(10,("truncate low case: start=%.0f,size=%.0f\n",
+                                                               (double)ul_curr->start, (double)ul_curr->size ));
+
+                               ul_curr = ul_curr->next;
+               
+                       } else if ( (ul_curr->start < lock->start) &&
+                                               (ul_curr->start + ul_curr->size > lock->start + lock->size) ) {
+                               /*
+                                * Worst case scenario. Unlock request completely overlaps an existing
+                                * lock range. Split the request into two, push the new (upper) request
+                                * into the dlink list, and continue with the entry after ul_new (as we
+                                * know that ul_new will not overlap with this lock).
+                                */
+/*********************************************
+        +---------------------------+
+        |       ul_curr             |
+        +---------------------------+
+                +---------+
+                | lock    |
+                +---------+
+BECOMES.....
+        +-------+         +---------+
+        |ul_curr|         |ul_new   |
+        +-------+         +---------+
+**********************************************/
+                               struct unlock_list *ul_new = (struct unlock_list *)talloc(ctx,
+                                                                                                       sizeof(struct unlock_list));
+
+                               if(ul_new == NULL) {
+                                       DEBUG(0,("brl_unlock_list: talloc fail.\n"));
+                                       return NULL; /* The talloc_destroy takes care of cleanup. */
+                               }
+
+                               ZERO_STRUCTP(ul_new);
+                               ul_new->start = lock->start + lock->size;
+                               ul_new->size = ul_curr->start + ul_curr->size - ul_new->start;
+                               ul_new->smbpid = ul_curr->smbpid;
+
+                               /* Add into the dlink list after the ul_curr point - NOT at ulhead. */
+                               DLIST_ADD(ul_curr, ul_new);
+
+                               /* Truncate the ul_curr. */
+                               ul_curr->size = lock->start - ul_curr->start;
+
+                               DEBUG(10,("split case: curr: start=%.0f,size=%.0f \
+new: start=%.0f,size=%.0f\n", (double)ul_curr->start, (double)ul_curr->size,
+                                                               (double)ul_new->start, (double)ul_new->size ));
+
+                               ul_curr = ul_new->next;
+
+                       } else {
+
+                               /*
+                                * This logic case should never happen. Ensure this is the
+                                * case by forcing an abort.... Remove in production.
+                                */
+
+                               smb_panic("logic flaw in cases...\n");
+                       }
+               } /* end for ( ul_curr = ulhead; ul_curr;) */
+       } /* end for (i=0; i<num_locks && ul_head; i++) */
+
+       tdb_unlockchain(tdb, kbuf);
+
+       if (dbuf.dptr)
+               free(dbuf.dptr);
+       
+       return ulhead;
+}
+
+/****************************************************************************
+ POSIX function to release a lock given a list. Returns True if the
+ lock could be released, False if not.
+****************************************************************************/
+
+static BOOL release_posix_lock(files_struct *fsp, SMB_BIG_UINT u_offset, SMB_BIG_UINT u_count)
+{
+       SMB_OFF_T offset;
+       SMB_OFF_T count;
+       BOOL ret = True;
+       TALLOC_CTX *ul_ctx = NULL;
+       struct unlock_list *ulist = NULL;
+       struct unlock_list *ul = NULL;
+
+       DEBUG(5,("release_posix_lock: File %s, offset = %.0f, count = %.0f\n",
+               fsp->fsp_name, (double)offset, (double)count ));
+
+       /*
+        * If the requested lock won't fit in the POSIX range, we will
+        * pretend it was successful.
+        */
+
+       if(!posix_lock_in_range(&offset, &count, u_offset, u_count))
+               return True;
+
+       if ((ul_ctx = talloc_init()) == NULL) {
+        DEBUG(0,("release_posix_lock: unable to init talloc context.\n"));
+               return True; /* Not a fatal error. */
+       }
+
+       if ((ul = (struct unlock_list *)talloc(ul_ctx, sizeof(struct unlock_list))) == NULL) {
+               DEBUG(0,("release_posix_lock: unable to talloc unlock list.\n"));
+               talloc_destroy(ul_ctx);
+               return True; /* Not a fatal error. */
+       }
+
+       /*
+        * Create the initial list entry containing the
+        * lock we want to remove.
+        */
+
+       ZERO_STRUCTP(ul);
+       ul->start = offset;
+       ul->size = count;
+       ul->fd = fsp->fd;
+
+       DLIST_ADD(ulist, ul);
+
+       /*
+        * The following call calculates if there are any
+        * overlapping read locks held by this process on
+        * other fd's open on the same file and creates a
+        * list of unlock ranges that will allow other
+        * POSIX lock ranges to remain on the file whilst the
+        * unlocks are performed.
+        */
+
+       ulist = posix_unlock_list(ul_ctx, ulist, fsp);
+
+       /*
+        * Release the POSIX locks on the list of ranges returned.
+        */
+
+       for(; ulist; ulist = ulist->next) {
+               SMB_OFF_T offset = ulist->start;
+               SMB_OFF_T count = ulist->size;
+
+               DEBUG(5,("release_posix_lock: Real unlock: offset = %.0f, count = %.0f\n",
+                       (double)offset, (double)count ));
+
+               if(u_count == 0) {
+
+                       /*
+                        * This lock must overlap with an existing read-only lock
+                        * held by another fd. Don't do any POSIX call.
+                        */
+
+                       continue;
+               }
+
+               /*
+                * If the requested lock won't fit in the POSIX range, we will
+                * pretend it was successful.
+                */
+
+               if(!posix_lock_in_range(&offset, &count, offset, count))
+                       continue;
+
+               DEBUG(5,("release_posix_lock: Real unlock: offset = %.0f, count = %.0f\n",
+                       (double)offset, (double)count ));
+
+               ret = fcntl_lock(fsp->fd,SMB_F_SETLK,offset,count,F_UNLCK);
+       }
+
+    talloc_destroy(ul_ctx);
+
+       /*
+        * We treat this as one unlock request for POSIX accounting purposes even
+        * if it may have been split into multiple smaller POSIX unlock ranges.
+        */ 
+
+       delete_posix_lock_entry(fsp->
+
+       return ret;
+}
+
+/****************************************************************************
+ Return a lock list associated with an open file.
+****************************************************************************/
+
+struct unlock_list *brl_getlocklist( TALLOC_CTX *ctx, SMB_DEV_T dev, SMB_INO_T ino, pid_t pid, int tid, int fnum)
+{
+       struct lock_key key;
+       TDB_DATA kbuf, dbuf;
+       int i, count;
+       struct lock_struct *locks;
+       struct unlock_list *ulist = NULL;
+
+       key.device = dev;
+       key.inode = ino;
+       kbuf.dptr = (char *)&key;
+       kbuf.dsize = sizeof(key);
+
+       dbuf.dptr = NULL;
+
+       tdb_lockchain(tdb, kbuf);
+       dbuf = tdb_fetch(tdb, kbuf);
+
+       if (!dbuf.dptr) {
+               tdb_unlockchain(tdb, kbuf);
+               return NULL;
+       }
+
+       /* There are existing locks - allocate an entry for each one. */
+       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 == tid &&
+                   lock->context.pid == pid &&
+                   lock->fnum == fnum) {
+
+                               struct unlock_list *ul_new = (struct unlock_list *)talloc(ctx,
+                                                                                                       sizeof(struct unlock_list));
+
+                               if(ul_new == NULL) {
+                                       DEBUG(0,("brl_getlocklist: talloc fail.\n"));
+                                       return NULL; /* The talloc_destroy takes care of cleanup. */
+                               }
+
+                               ZERO_STRUCTP(ul_new);
+                               ul_new->start = lock->start;
+                               ul_new->size = lock->size;
+                               ul_new->smbpid = lock->context.smbpid;
+
+                               DLIST_ADD(ulist, ul_new);
+               }
+       }
+
+       if (dbuf.dptr)
+               free(dbuf.dptr);
+       tdb_unlockchain(tdb, kbuf);
+
+       return ulist;
+}
+
+/****************************************************************************
+ Remove any locks on this fd. Called from file_close().
+****************************************************************************/
+
+void posix_locking_close_file(files_struct *fsp)
+{
+       TALLOC_CTX *ul_ctx = NULL;
+       struct unlock_list *ul = NULL;
+       int eclass;
+       uint32 ecode;
+       struct pending_closes *pc;
+
+               /*
+                * Optimization for the common case where we are the only
+                * opener of a file. If all fd entries are our own, we don't
+                * need to explicitly release all the locks via the POSIX functions,
+                * we can just release all the brl locks, as in the no POSIX locking case.
+                */
+
+               if ((pc = find_pending_close_entry(fsp->dev, fsp->inode)) != NULL) {
+
+                       if (pc->fd_array_size == 1 && pc->fd_array[0] == fsp->fd ) {
+                               /*
+                                * Just release all the brl locks, no need to release individually.
+                                */
+
+                               brl_close(fsp->dev, fsp->inode, pid, fsp->conn->cnum, fsp->fnum);
+                               return;
+                       }
+               }
+
+               if ((ul_ctx = talloc_init()) == NULL) {
+                       DEBUG(0,("locking_close_file: unable to init talloc context.\n"));
+                       return;
+               }
+
+               /*
+                * We need to release all POSIX locks we have on this
+                * fd. Get all our existing locks from the tdb locking database.
+                */
+
+               ul = brl_getlocklist(ul_ctx, fsp->dev, fsp->inode, pid, fsp->conn->cnum, fsp->fnum);
+
+               /*
+                * Now unlock all of them. This will remove the brl entry also
+                * for each lock. Note we need to make sure the global_smbpid matches
+                * the one associated with each lock in case the client plays games
+                * with smbpids (like smbtorture does :-).
+                */
+
+               for(; ul; ul = ul->next) {
+                       global_smbpid = ul->smbpid;
+                       do_unlock(fsp,fsp->conn,ul->size,ul->start,&eclass,&ecode);
+               }
+       
+               talloc_destroy(ul_ctx);
+
+       } else {
+
+               /*
+                * Just release all the brl locks, no need to release individually.
+                */
+
+               brl_close(fsp->dev, fsp->inode, pid, fsp->conn->cnum, fsp->fnum);
+       }
+}
+
+/*******************************************************************
+ Create the in-memory POSIX lock databases.
+********************************************************************/
+
+void posix_lock_init(void)
+{
+       if (posix_lock_tdb && posix_pending_close_tdb)
+               return;
+
+       if (!posix_lock_tdb)
+               posix_lock_tdb = tdb_open(NULL, 0, TDB_CLEAR_IF_FIRST,
+                   O_RDWR|O_CREAT, 0644);
+    if (!posix_lock_tdb) {
+        DEBUG(0,("Failed to open POSIX byte range locking database.\n"));
+    }
+       if (!posix_pending_close_tdb)
+               posix_pending_close_tdb = tdb_open(NULL, 0, TDB_CLEAR_IF_FIRST,
+                   O_RDWR|O_CREAT, 0644);
+    if (!posix_pending_close_tdb) {
+        DEBUG(0,("Failed to open POSIX pending close database.\n"));
+    }
+}