source/smbd/oplock_linux.c

   1 /*
   2    Unix SMB/CIFS implementation.
   3    kernel oplock processing for Linux
   4    Copyright (C) Andrew Tridgell 2000
   5
   6    This program is free software; you can redistribute it and/or modify
   7    it under the terms of the GNU General Public License as published by
   8    the Free Software Foundation; either version 2 of the License, or
   9    (at your option) any later version.
  10
  11    This program is distributed in the hope that it will be useful,
  12    but WITHOUT ANY WARRANTY; without even the implied warranty of
  13    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14    GNU General Public License for more details.
  15
  16    You should have received a copy of the GNU General Public License
  17    along with this program; if not, write to the Free Software
  18    Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  19 */
  20
  21 #include "includes.h"
  22
  23 #if HAVE_KERNEL_OPLOCKS_LINUX
  24
  25 static VOLATILE sig_atomic_t signals_received;
  26 static VOLATILE sig_atomic_t signals_processed;
  27 static VOLATILE sig_atomic_t fd_pending; /* the fd of the current pending signal */
  28
  29 #ifndef F_SETLEASE
  30 #define F_SETLEASE      1024
  31 #endif
  32
  33 #ifndef F_GETLEASE
  34 #define F_GETLEASE      1025
  35 #endif
  36
  37 #ifndef CAP_LEASE
  38 #define CAP_LEASE 28
  39 #endif
  40
  41 #ifndef RT_SIGNAL_LEASE
  42 #define RT_SIGNAL_LEASE 33
  43 #endif
  44
  45 #ifndef F_SETSIG
  46 #define F_SETSIG 10
  47 #endif
  48
  49 /****************************************************************************
  50  Handle a LEASE signal, incrementing the signals_received and blocking the signal.
  51 ****************************************************************************/
  52
  53 static void signal_handler(int sig, siginfo_t *info, void *unused)
  54 {
  55         BlockSignals(True, sig);
  56         fd_pending = (sig_atomic_t)info->si_fd;
  57         signals_received++;
  58         sys_select_signal();
  59 }
  60
  61 /****************************************************************************
  62  Try to gain a linux capability.
  63 ****************************************************************************/
  64
  65 static void set_capability(unsigned capability)
  66 {
  67 #ifndef _LINUX_CAPABILITY_VERSION
  68 #define _LINUX_CAPABILITY_VERSION 0x19980330
  69 #endif
  70         /* these can be removed when they are in glibc headers */
  71         struct  {
  72                 uint32 version;
  73                 int pid;
  74         } header;
  75         struct {
  76                 uint32 effective;
  77                 uint32 permitted;
  78                 uint32 inheritable;
  79         } data;
  80
  81         header.version = _LINUX_CAPABILITY_VERSION;
  82         header.pid = 0;
  83
  84         if (capget(&header, &data) == -1) {
  85                 DEBUG(3,("Unable to get kernel capabilities (%s)\n", strerror(errno)));
  86                 return;
  87         }
  88
  89         data.effective |= (1<<capability);
  90
  91         if (capset(&header, &data) == -1) {
  92                 DEBUG(3,("Unable to set %d capability (%s)\n",
  93                          capability, strerror(errno)));
  94         }
  95 }
  96
  97 /****************************************************************************
  98  Call SETLEASE. If we get EACCES then we try setting up the right capability and
  99  try again
 100 ****************************************************************************/
 101
 102 static int linux_setlease(int fd, int leasetype)
 103 {
 104         int ret;
 105
 106         if (fcntl(fd, F_SETSIG, RT_SIGNAL_LEASE) == -1) {
 107                 DEBUG(3,("Failed to set signal handler for kernel lease\n"));
 108                 return -1;
 109         }
 110
 111         ret = fcntl(fd, F_SETLEASE, leasetype);
 112         if (ret == -1 && errno == EACCES) {
 113                 set_capability(CAP_LEASE);
 114                 ret = fcntl(fd, F_SETLEASE, leasetype);
 115         }
 116
 117         return ret;
 118 }
 119
 120 /****************************************************************************
 121  * Deal with the Linux kernel <--> smbd
 122  * oplock break protocol.
 123 ****************************************************************************/
 124
 125 static BOOL linux_oplock_receive_message(fd_set *fds, char *buffer, int buffer_len)
 126 {
 127         BOOL ret = True;
 128         struct files_struct *fsp;
 129
 130         if (signals_received == signals_processed)
 131                 return False;
 132
 133         if ((fsp = file_find_fd(fd_pending)) == NULL) {
 134                 DEBUG(0,("Invalid file descriptor %d in kernel oplock break!\n", (int)fd_pending));
 135                 ret = False;
 136                 goto out;
 137         }
 138
 139         DEBUG(3,("receive_local_message: kernel oplock break request received for \
 140 dev = %x, inode = %.0f\n", (unsigned int)fsp->dev, (double)fsp->inode ));
 141
 142         /*
 143          * Create a kernel oplock break message.
 144          */
 145
 146         /* Setup the message header */
 147         SIVAL(buffer,OPBRK_CMD_LEN_OFFSET,KERNEL_OPLOCK_BREAK_MSG_LEN);
 148         SSVAL(buffer,OPBRK_CMD_PORT_OFFSET,0);
 149
 150         buffer += OPBRK_CMD_HEADER_LEN;
 151
 152         SSVAL(buffer,OPBRK_MESSAGE_CMD_OFFSET,KERNEL_OPLOCK_BREAK_CMD);
 153
 154         memcpy(buffer + KERNEL_OPLOCK_BREAK_DEV_OFFSET, (char *)&fsp->dev, sizeof(fsp->dev));
 155         memcpy(buffer + KERNEL_OPLOCK_BREAK_INODE_OFFSET, (char *)&fsp->inode, sizeof(fsp->inode));
 156         memcpy(buffer + KERNEL_OPLOCK_BREAK_FILEID_OFFSET, (char *)&fsp->file_id, sizeof(fsp->file_id));
 157
 158  out:
 159         /* now we can receive more signals */
 160         fd_pending = (sig_atomic_t)-1;
 161         signals_processed++;
 162         BlockSignals(False, RT_SIGNAL_LEASE);
 163
 164         return ret;
 165 }
 166
 167 /****************************************************************************
 168  Attempt to set an kernel oplock on a file.
 169 ****************************************************************************/
 170
 171 static BOOL linux_set_kernel_oplock(files_struct *fsp, int oplock_type)
 172 {
 173         if (linux_setlease(fsp->fd, F_WRLCK) == -1) {
 174                 DEBUG(3,("set_file_oplock: Refused oplock on file %s, fd = %d, dev = %x, \
 175 inode = %.0f. (%s)\n",
 176                          fsp->fsp_name, fsp->fd,
 177                          (unsigned int)fsp->dev, (double)fsp->inode, strerror(errno)));
 178                 return False;
 179         }
 180
 181         DEBUG(3,("set_file_oplock: got kernel oplock on file %s, dev = %x, inode = %.0f, file_id = %lu\n",
 182                   fsp->fsp_name, (unsigned int)fsp->dev, (double)fsp->inode, fsp->file_id));
 183
 184         return True;
 185 }
 186
 187 /****************************************************************************
 188  Release a kernel oplock on a file.
 189 ****************************************************************************/
 190
 191 static void linux_release_kernel_oplock(files_struct *fsp)
 192 {
 193         if (DEBUGLVL(10)) {
 194                 /*
 195                  * Check and print out the current kernel
 196                  * oplock state of this file.
 197                  */
 198                 int state = fcntl(fsp->fd, F_GETLEASE, 0);
 199                 dbgtext("release_kernel_oplock: file %s, dev = %x, inode = %.0f file_id = %lu has kernel \
 200 oplock state of %x.\n", fsp->fsp_name, (unsigned int)fsp->dev,
 201                         (double)fsp->inode, fsp->file_id, state );
 202         }
 203
 204         /*
 205          * Remove the kernel oplock on this file.
 206          */
 207         if (linux_setlease(fsp->fd, F_UNLCK) == -1) {
 208                 if (DEBUGLVL(0)) {
 209                         dbgtext("release_kernel_oplock: Error when removing kernel oplock on file " );
 210                         dbgtext("%s, dev = %x, inode = %.0f, file_id = %lu. Error was %s\n",
 211                                 fsp->fsp_name, (unsigned int)fsp->dev,
 212                                 (double)fsp->inode, fsp->file_id, strerror(errno) );
 213                 }
 214         }
 215 }
 216
 217 /****************************************************************************
 218  Parse a kernel oplock message.
 219 ****************************************************************************/
 220
 221 static BOOL linux_kernel_oplock_parse(char *msg_start, int msg_len, SMB_INO_T *inode,
 222                 SMB_DEV_T *dev, unsigned long *file_id)
 223 {
 224         /* Ensure that the msg length is correct. */
 225         if (msg_len != KERNEL_OPLOCK_BREAK_MSG_LEN) {
 226                 DEBUG(0,("incorrect length for KERNEL_OPLOCK_BREAK_CMD (was %d, should be %d).\n",
 227                          msg_len, KERNEL_OPLOCK_BREAK_MSG_LEN));
 228                 return False;
 229         }
 230
 231         memcpy((char *)inode, msg_start+KERNEL_OPLOCK_BREAK_INODE_OFFSET, sizeof(*inode));
 232         memcpy((char *)dev, msg_start+KERNEL_OPLOCK_BREAK_DEV_OFFSET, sizeof(*dev));
 233         memcpy((char *)file_id, msg_start+KERNEL_OPLOCK_BREAK_FILEID_OFFSET, sizeof(*file_id));
 234
 235         DEBUG(3,("kernel oplock break request for file dev = %x, inode = %.0f, file_id = %lu\n",
 236                 (unsigned int)*dev, (double)*inode, *file_id));
 237
 238         return True;
 239 }
 240
 241 /****************************************************************************
 242  See if a oplock message is waiting.
 243 ****************************************************************************/
 244
 245 static BOOL linux_oplock_msg_waiting(fd_set *fds)
 246 {
 247         return signals_processed != signals_received;
 248 }
 249
 250 /****************************************************************************
 251  See if the kernel supports oplocks.
 252 ****************************************************************************/
 253
 254 static BOOL linux_oplocks_available(void)
 255 {
 256         int fd, ret;
 257         fd = open("/dev/null", O_RDONLY);
 258         if (fd == -1)
 259                 return False; /* uggh! */
 260         ret = fcntl(fd, F_GETLEASE, 0);
 261         close(fd);
 262         return ret == F_UNLCK;
 263 }
 264
 265 /****************************************************************************
 266  Setup kernel oplocks.
 267 ****************************************************************************/
 268
 269 struct kernel_oplocks *linux_init_kernel_oplocks(void)
 270 {
 271         static struct kernel_oplocks koplocks;
 272         struct sigaction act;
 273
 274         if (!linux_oplocks_available()) {
 275                 DEBUG(3,("Linux kernel oplocks not available\n"));
 276                 return NULL;
 277         }
 278
 279         act.sa_handler = NULL;
 280         act.sa_sigaction = signal_handler;
 281         act.sa_flags = SA_SIGINFO;
 282         if (sigaction(RT_SIGNAL_LEASE, &act, NULL) != 0) {
 283                 DEBUG(0,("Failed to setup RT_SIGNAL_LEASE handler\n"));
 284                 return NULL;
 285         }
 286
 287         koplocks.receive_message = linux_oplock_receive_message;
 288         koplocks.set_oplock = linux_set_kernel_oplock;
 289         koplocks.release_oplock = linux_release_kernel_oplock;
 290         koplocks.parse_message = linux_kernel_oplock_parse;
 291         koplocks.msg_waiting = linux_oplock_msg_waiting;
 292         koplocks.notification_fd = -1;
 293
 294         DEBUG(3,("Linux kernel oplocks enabled\n"));
 295
 296         return &koplocks;
 297 }
 298 #else
 299  void oplock_linux_dummy(void) {}
 300 #endif /* HAVE_KERNEL_OPLOCKS_LINUX */