a first pass at Linux kernel oplocks support

(This used to be commit 3253085d9883a181c04b9c9ecf7d0ccdfbcee88d)

6 files changed, 212 insertions, 26 deletions

diff --git a/source3/Makefile.in b/source3/Makefile.in
index a70d094f8a..ba6424e3d2 100644
--- a/source3/Makefile.in
+++ b/source3/Makefile.in
@@ -157,7 +157,7 @@ PASSDB_OBJ = passdb/passdb.o passdb/smbpassfile.o passdb/smbpass.o passdb/secret
 
 PROFILE_OBJ = profile/profile.o
 
-OPLOCK_OBJ = smbd/oplock.o smbd/oplock_irix.o
+OPLOCK_OBJ = smbd/oplock.o smbd/oplock_irix.o smbd/oplock_linux.o
 
 SMBD_OBJ1 = smbd/server.o smbd/files.o smbd/chgpasswd.o smbd/connection.o \
             smbd/dfree.o smbd/dir.o smbd/password.o smbd/conn.o smbd/fileio.o \
diff --git a/source3/include/smb.h b/source3/include/smb.h
index d2c79e3b53..68df3f250c 100644
--- a/source3/include/smb.h
+++ b/source3/include/smb.h
@@ -1607,9 +1607,8 @@ extern int chain_size;
 
 #define KERNEL_OPLOCK_CAPABILITY 0x1
 
-#if defined(HAVE_KERNEL_OPLOCKS_IRIX)
 /*
- * Oplock break command code sent via the kernel interface.
+ * Oplock break command code sent via the kernel interface (if it exists).
  *
  * Form of this is :
  *
@@ -1618,12 +1617,10 @@ extern int chain_size;
  *  | cmd| dev    |  inode |
  *  +----+--------+--------+
  */
-
 #define KERNEL_OPLOCK_BREAK_DEV_OFFSET 2
 #define KERNEL_OPLOCK_BREAK_INODE_OFFSET (KERNEL_OPLOCK_BREAK_DEV_OFFSET + sizeof(SMB_DEV_T))
 #define KERNEL_OPLOCK_BREAK_MSG_LEN (KERNEL_OPLOCK_BREAK_INODE_OFFSET + sizeof(SMB_INO_T))
 
-#endif /* HAVE_KERNEL_OPLOCKS_IRIX */
 
 /* if a kernel does support oplocks then a structure of the following
    typee is used to describe how to interact with the kernel */
diff --git a/source3/param/loadparm.c b/source3/param/loadparm.c
index ef62fc82ca..28dd455e51 100644
--- a/source3/param/loadparm.c
+++ b/source3/param/loadparm.c
@@ -1144,10 +1144,7 @@ static void init_globals(void)
 
 	Globals.bDNSproxy = True;
 
-	/*
-	 * smbd will check at runtime to see if this value
-	 * will really be used or not.
-	 */
+	/* this just means to use them if they exist */
 	Globals.bKernelOplocks = True;
 
 	Globals.bAllowTrustedDomains = True;
@@ -2550,7 +2547,7 @@ static BOOL do_parameter(char *pszParmName, char *pszParmValue)
 	if (!bInGlobalSection && bGlobalOnly)
 		return (True);
 
-	DEBUGADD(3, ("doing parameter %s = %s\n", pszParmName, pszParmValue));
+	DEBUGADD(4, ("doing parameter %s = %s\n", pszParmName, pszParmValue));
 
 	return (lp_do_parameter(bInGlobalSection ? -2 : iServiceIndex,
 				pszParmName, pszParmValue));
@@ -3124,7 +3121,7 @@ BOOL lp_load(char *pszFname, BOOL global_only, BOOL save_defaults,
 	bRetval = pm_process(n2, do_section, do_parameter);
 
 	/* finish up the last section */
-	DEBUG(3, ("pm_process() returned %s\n", BOOLSTR(bRetval)));
+	DEBUG(4, ("pm_process() returned %s\n", BOOLSTR(bRetval)));
 	if (bRetval)
 		if (iServiceIndex >= 0)
 			bRetval = service_ok(iServiceIndex);
diff --git a/source3/smbd/oplock.c b/source3/smbd/oplock.c
index 33d82f3dde..44a8e9b071 100644
--- a/source3/smbd/oplock.c
+++ b/source3/smbd/oplock.c
@@ -1021,22 +1021,9 @@ should be %d\n", (int)pid, share_entry->op_port, global_oplock_port));
     reply_msg_start = &op_break_reply[OPBRK_CMD_HEADER_LEN];
 
 
-#if HAVE_KERNEL_OPLOCKS_IRIX
-    if((reply_msg_len != OPLOCK_BREAK_MSG_LEN) && (reply_msg_len != KERNEL_OPLOCK_BREAK_MSG_LEN))
-#else
-    if(reply_msg_len != OPLOCK_BREAK_MSG_LEN)
-#endif
-    {
-      /* Ignore it. */
-      DEBUG( 0, ( "request_oplock_break: invalid message length (%d) received.", reply_msg_len ) );
-      DEBUGADD( 0, ( "  Ignoring.\n" ) );
-      continue;
-    }
-
     /*
      * Test to see if this is the reply we are awaiting.
      */
-
     if((SVAL(reply_msg_start,OPBRK_MESSAGE_CMD_OFFSET) & CMD_REPLY) &&
        ((SVAL(reply_msg_start,OPBRK_MESSAGE_CMD_OFFSET) & ~CMD_REPLY) == OPLOCK_BREAK_CMD) &&
        (reply_from_port == share_entry->op_port) && 
diff --git a/source3/smbd/oplock_irix.c b/source3/smbd/oplock_irix.c
index 0820c4048d..6eb8ff9191 100644
--- a/source3/smbd/oplock_irix.c
+++ b/source3/smbd/oplock_irix.c
@@ -3,8 +3,8 @@
 
 /* 
    Unix SMB/Netbios implementation.
-   Version 1.9.
-   oplock processing
+   Version 2.x
+   IRIX kernel oplock processing
    Copyright (C) Andrew Tridgell 1992-1998
    
    This program is free software; you can redistribute it and/or modify
diff --git a/source3/smbd/oplock_linux.c b/source3/smbd/oplock_linux.c
new file mode 100644
index 0000000000..d8496ca9ca
--- /dev/null
+++ b/source3/smbd/oplock_linux.c
@@ -0,0 +1,205 @@
+#define OLD_NTDOMAIN 1
+#if HAVE_KERNEL_OPLOCKS_LINUX
+
+/* 
+   Unix SMB/Netbios implementation.
+   Version 3.0
+   kernel oplock processing for Linux
+   Copyright (C) Andrew Tridgell 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.
+*/
+
+#include "includes.h"
+
+extern int DEBUGLEVEL;
+
+static unsigned signals_received;
+static unsigned signals_processed;
+static int fd_pending; /* the fd of the current pending SIGIO */
+
+/****************************************************************************
+handle a SIGIO, incrementing the signals_received and blocking SIGIO
+****************************************************************************/
+static void sigio_handler(int signal, siginfo_t *info, void *unused)
+{
+	fd_pending = info->si_fd;
+	signals_received++;
+	BlockSignals(True, SIGIO);
+}
+
+/****************************************************************************
+ * Deal with the Linux kernel <--> smbd
+ * oplock break protocol.
+****************************************************************************/
+static BOOL linux_oplock_receive_message(fd_set *fds, char *buffer, int buffer_len)
+{
+	SMB_DEV_T dev;
+	SMB_INO_T inode;
+	SMB_STRUCT_STAT sbuf;
+	BOOL ret;
+
+	if (signals_received == signals_processed) return False;
+
+	if (sys_fstat(fd_pending,&sbuf) == -1) {
+		DEBUG(0,("Invalid file descriptor %d in kernel oplock break!\n", fd_pending));
+		ret = False;
+		goto out;
+	}
+
+	dev = sbuf.st_dev;
+	inode = sbuf.st_ino;
+     
+	DEBUG(5,("receive_local_message: kernel oplock break request received for \
+dev = %x, inode = %.0f\n", (unsigned int)dev, (double)inode ));
+     
+	/*
+	 * Create a kernel oplock break message.
+	 */
+     
+	/* Setup the message header */
+	SIVAL(buffer,OPBRK_CMD_LEN_OFFSET,KERNEL_OPLOCK_BREAK_MSG_LEN);
+	SSVAL(buffer,OPBRK_CMD_PORT_OFFSET,0);
+     
+	buffer += OPBRK_CMD_HEADER_LEN;
+     
+	SSVAL(buffer,OPBRK_MESSAGE_CMD_OFFSET,KERNEL_OPLOCK_BREAK_CMD);
+     
+	memcpy(buffer + KERNEL_OPLOCK_BREAK_DEV_OFFSET, (char *)&dev, sizeof(dev));
+	memcpy(buffer + KERNEL_OPLOCK_BREAK_INODE_OFFSET, (char *)&inode, sizeof(inode));	
+
+ out:
+	/* now we can receive more signals */
+	fd_pending = -1;
+	signals_processed++;
+	BlockSignals(False, SIGIO);
+     
+	return True;
+}
+
+
+/****************************************************************************
+ Attempt to set an kernel oplock on a file.
+****************************************************************************/
+static BOOL linux_set_kernel_oplock(files_struct *fsp, int oplock_type)
+{
+	if (fcntl(fsp->fd, F_SETLEASE, F_WRLCK) == -1) {
+		DEBUG(5,("set_file_oplock: Refused oplock on file %s, fd = %d, dev = %x, \
+inode = %.0f.\n",
+			 fsp->fsp_name, fsp->fd, (unsigned int)fsp->dev, (double)fsp->inode));
+		return False;
+	}
+	
+	DEBUG(10,("set_file_oplock: got kernel oplock on file %s, dev = %x, inode = %.0f\n",
+		  fsp->fsp_name, (unsigned int)fsp->dev, (double)fsp->inode));
+
+	return True;
+}
+
+
+/****************************************************************************
+ Release a kernel oplock on a file.
+****************************************************************************/
+static void linux_release_kernel_oplock(files_struct *fsp)
+{
+	if (DEBUGLVL(10)) {
+		/*
+		 * Check and print out the current kernel
+		 * oplock state of this file.
+		 */
+		int state = fcntl(fsp->fd, F_GETLEASE, 0);
+		dbgtext("release_kernel_oplock: file %s, dev = %x, inode = %.0f has kernel \
+oplock state of %x.\n", fsp->fsp_name, (unsigned int)fsp->dev,
+                        (double)fsp->inode, state );
+	}
+
+	/*
+	 * Remove the kernel oplock on this file.
+	 */
+	if (fcntl(fsp->fd, F_SETLEASE, F_UNLCK) == -1) {
+		if (DEBUGLVL(0)) {
+			dbgtext("release_kernel_oplock: Error when removing kernel oplock on file " );
+			dbgtext("%s, dev = %x, inode = %.0f. Error was %s\n",
+				fsp->fsp_name, (unsigned int)fsp->dev, 
+				(double)fsp->inode, strerror(errno) );
+		}
+	}
+}
+
+
+/****************************************************************************
+parse a kernel oplock message
+****************************************************************************/
+static BOOL linux_kernel_oplock_parse(char *msg_start, int msg_len, SMB_INO_T *inode, SMB_DEV_T *dev)
+{
+	/* Ensure that the msg length is correct. */
+	if (msg_len != KERNEL_OPLOCK_BREAK_MSG_LEN) {
+		DEBUG(0,("process_local_message: incorrect length for KERNEL_OPLOCK_BREAK_CMD (was %d, \
+should be %d).\n", msg_len, KERNEL_OPLOCK_BREAK_MSG_LEN));
+		return False;
+	}
+
+        memcpy((char *)inode, msg_start+KERNEL_OPLOCK_BREAK_INODE_OFFSET, sizeof(*inode));
+        memcpy((char *)dev, msg_start+KERNEL_OPLOCK_BREAK_DEV_OFFSET, sizeof(*dev));
+
+        DEBUG(5,("process_local_message: kernel oplock break request for \
+file dev = %x, inode = %.0f\n", (unsigned int)dev, (double)inode));
+
+	return True;
+}
+
+
+/****************************************************************************
+see if a oplock message is waiting
+****************************************************************************/
+static BOOL linux_oplock_msg_waiting(fd_set *fds)
+{
+	return signals_processed != signals_received;
+}
+
+
+/****************************************************************************
+setup kernel oplocks
+****************************************************************************/
+struct kernel_oplocks *linux_init_kernel_oplocks(void) 
+{
+	static struct kernel_oplocks koplocks;
+        struct sigaction act;
+
+        act.sa_handler = NULL;
+        act.sa_sigaction = sigio_handler;
+        act.sa_flags = SA_SIGINFO;
+        if (sigaction(SIGIO, &act, NULL) != 0) {
+		DEBUG(0,("Failed to setup SIGIO handler\n"));
+		return NULL;
+        }
+
+	koplocks.receive_message = linux_oplock_receive_message;
+	koplocks.set_oplock = linux_set_kernel_oplock;
+	koplocks.release_oplock = linux_release_kernel_oplock;
+	koplocks.parse_message = linux_kernel_oplock_parse;
+	koplocks.msg_waiting = linux_oplock_msg_waiting;
+	koplocks.notification_fd = -1;
+
+	return &koplocks;
+}
+
+
+
+#else
+ void oplock_linux_dummy(void) {}
+#endif /* HAVE_KERNEL_OPLOCKS_LINUX */
+
+#undef OLD_NTDOMAIN