source/smbd/oplock_linux.c

   1 /*
   2    Unix SMB/Netbios implementation.
   3    Version 3.0
   4    kernel oplock processing for Linux
   5    Copyright (C) Andrew Tridgell 2000
   6
   7    This program is free software; you can redistribute it and/or modify
   8    it under the terms of the GNU General Public License as published by
   9    the Free Software Foundation; either version 2 of the License, or
  10    (at your option) any later version.
  11
  12    This program is distributed in the hope that it will be useful,
  13    but WITHOUT ANY WARRANTY; without even the implied warranty of
  14    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  15    GNU General Public License for more details.
  16
  17    You should have received a copy of the GNU General Public License
  18    along with this program; if not, write to the Free Software
  19    Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  20 */
  21
  22 #include "includes.h"
  23
  24 #if HAVE_KERNEL_OPLOCKS_LINUX
  25
  26 static SIG_ATOMIC_T signals_received;
  27 #define FD_PENDING_SIZE 100
  28 static SIG_ATOMIC_T fd_pending_array[FD_PENDING_SIZE];
  29
  30 #ifndef F_SETLEASE
  31 #define F_SETLEASE      1024
  32 #endif
  33
  34 #ifndef F_GETLEASE
  35 #define F_GETLEASE      1025
  36 #endif
  37
  38 #ifndef CAP_LEASE
  39 #define CAP_LEASE 28
  40 #endif
  41
  42 #ifndef RT_SIGNAL_LEASE
  43 #define RT_SIGNAL_LEASE 33
  44 #endif
  45
  46 #ifndef F_SETSIG
  47 #define F_SETSIG 10
  48 #endif
  49
  50 /****************************************************************************
  51  Handle a LEASE signal, incrementing the signals_received and blocking the signal.
  52 ****************************************************************************/
  53
  54 static void signal_handler(int sig, siginfo_t *info, void *unused)
  55 {
  56         if (signals_received < FD_PENDING_SIZE - 1) {
  57                 fd_pending_array[signals_received] = (SIG_ATOMIC_T)info->si_fd;
  58                 signals_received++;
  59         } /* Else signal is lost. */
  60         sys_select_signal();
  61 }
  62
  63 /****************************************************************************
  64  Try to gain a linux capability.
  65 ****************************************************************************/
  66
  67 static void set_capability(unsigned capability)
  68 {
  69 #ifndef _LINUX_CAPABILITY_VERSION
  70 #define _LINUX_CAPABILITY_VERSION 0x19980330
  71 #endif
  72         /* these can be removed when they are in glibc headers */
  73         struct  {
  74                 uint32 version;
  75                 int pid;
  76         } header;
  77         struct {
  78                 uint32 effective;
  79                 uint32 permitted;
  80                 uint32 inheritable;
  81         } data;
  82
  83         header.version = _LINUX_CAPABILITY_VERSION;
  84         header.pid = 0;
  85
  86         if (capget(&header, &data) == -1) {
  87                 DEBUG(3,("Unable to get kernel capabilities (%s)\n", strerror(errno)));
  88                 return;
  89         }
  90
  91         data.effective |= (1<<capability);
  92
  93         if (capset(&header, &data) == -1) {
  94                 DEBUG(3,("Unable to set %d capability (%s)\n",
  95                          capability, strerror(errno)));
  96         }
  97 }
  98
  99 /****************************************************************************
 100  Call SETLEASE. If we get EACCES then we try setting up the right capability and
 101  try again
 102 ****************************************************************************/
 103
 104 static int linux_setlease(int fd, int leasetype)
 105 {
 106         int ret;
 107
 108         if (fcntl(fd, F_SETSIG, RT_SIGNAL_LEASE) == -1) {
 109                 DEBUG(3,("Failed to set signal handler for kernel lease\n"));
 110                 return -1;
 111         }
 112
 113         ret = fcntl(fd, F_SETLEASE, leasetype);
 114         if (ret == -1 && errno == EACCES) {
 115                 set_capability(CAP_LEASE);
 116                 ret = fcntl(fd, F_SETLEASE, leasetype);
 117         }
 118
 119         return ret;
 120 }
 121
 122 /****************************************************************************
 123  * Deal with the Linux kernel <--> smbd
 124  * oplock break protocol.
 125 ****************************************************************************/
 126
 127 static BOOL linux_oplock_receive_message(fd_set *fds, char *buffer, int buffer_len)
 128 {
 129         int fd;
 130         struct files_struct *fsp;
 131
 132         BlockSignals(True, RT_SIGNAL_LEASE);
 133         fd = fd_pending_array[0];
 134         fsp = file_find_fd(fd);
 135         fd_pending_array[0] = (SIG_ATOMIC_T)-1;
 136         if (signals_received > 1)
 137         memmove((void *)&fd_pending_array[0], (void *)&fd_pending_array[1],
 138                         sizeof(SIG_ATOMIC_T)*(signals_received-1));
 139         signals_received--;
 140         /* now we can receive more signals */
 141         BlockSignals(False, RT_SIGNAL_LEASE);
 142
 143         if (fsp == NULL) {
 144                 DEBUG(0,("Invalid file descriptor %d in kernel oplock break!\n", (int)fd));
 145                 return False;
 146         }
 147
 148         DEBUG(3,("linux_oplock_receive_message: kernel oplock break request received for \
 149 dev = %x, inode = %.0f fd = %d, fileid = %lu \n", (unsigned int)fsp->dev, (double)fsp->inode,
 150                         fd, fsp->file_id));
 151
 152         /*
 153          * Create a kernel oplock break message.
 154          */
 155
 156         /* Setup the message header */
 157         SIVAL(buffer,OPBRK_CMD_LEN_OFFSET,KERNEL_OPLOCK_BREAK_MSG_LEN);
 158         SSVAL(buffer,OPBRK_CMD_PORT_OFFSET,0);
 159
 160         buffer += OPBRK_CMD_HEADER_LEN;
 161
 162         SSVAL(buffer,OPBRK_MESSAGE_CMD_OFFSET,KERNEL_OPLOCK_BREAK_CMD);
 163
 164         memcpy(buffer + KERNEL_OPLOCK_BREAK_DEV_OFFSET, (char *)&fsp->dev, sizeof(fsp->dev));
 165         memcpy(buffer + KERNEL_OPLOCK_BREAK_INODE_OFFSET, (char *)&fsp->inode, sizeof(fsp->inode));
 166         memcpy(buffer + KERNEL_OPLOCK_BREAK_FILEID_OFFSET, (char *)&fsp->file_id, sizeof(fsp->file_id));
 167
 168         return True;
 169 }
 170
 171 /****************************************************************************
 172  Attempt to set an kernel oplock on a file.
 173 ****************************************************************************/
 174
 175 static BOOL linux_set_kernel_oplock(files_struct *fsp, int oplock_type)
 176 {
 177         if (linux_setlease(fsp->fd, F_WRLCK) == -1) {
 178                 DEBUG(3,("linux_set_kernel_oplock: Refused oplock on file %s, fd = %d, dev = %x, \
 179 inode = %.0f. (%s)\n",
 180                          fsp->fsp_name, fsp->fd,
 181                          (unsigned int)fsp->dev, (double)fsp->inode, strerror(errno)));
 182                 return False;
 183         }
 184
 185         DEBUG(3,("linux_set_kernel_oplock: got kernel oplock on file %s, dev = %x, inode = %.0f, file_id = %lu\n",
 186                   fsp->fsp_name, (unsigned int)fsp->dev, (double)fsp->inode, fsp->file_id));
 187
 188         return True;
 189 }
 190
 191 /****************************************************************************
 192  Release a kernel oplock on a file.
 193 ****************************************************************************/
 194
 195 static void linux_release_kernel_oplock(files_struct *fsp)
 196 {
 197         if (DEBUGLVL(10)) {
 198                 /*
 199                  * Check and print out the current kernel
 200                  * oplock state of this file.
 201                  */
 202                 int state = fcntl(fsp->fd, F_GETLEASE, 0);
 203                 dbgtext("linux_release_kernel_oplock: file %s, dev = %x, inode = %.0f file_id = %lu has kernel \
 204 oplock state of %x.\n", fsp->fsp_name, (unsigned int)fsp->dev,
 205                         (double)fsp->inode, fsp->file_id, state );
 206         }
 207
 208         /*
 209          * Remove the kernel oplock on this file.
 210          */
 211         if (linux_setlease(fsp->fd, F_UNLCK) == -1) {
 212                 if (DEBUGLVL(0)) {
 213                         dbgtext("linux_release_kernel_oplock: Error when removing kernel oplock on file " );
 214                         dbgtext("%s, dev = %x, inode = %.0f, file_id = %lu. Error was %s\n",
 215                                 fsp->fsp_name, (unsigned int)fsp->dev,
 216                                 (double)fsp->inode, fsp->file_id, strerror(errno) );
 217                 }
 218         }
 219 }
 220
 221 /****************************************************************************
 222  Parse a kernel oplock message.
 223 ****************************************************************************/
 224
 225 static BOOL linux_kernel_oplock_parse(char *msg_start, int msg_len, SMB_INO_T *inode,
 226                 SMB_DEV_T *dev, unsigned long *file_id)
 227 {
 228         /* Ensure that the msg length is correct. */
 229         if (msg_len != KERNEL_OPLOCK_BREAK_MSG_LEN) {
 230                 DEBUG(0,("incorrect length for KERNEL_OPLOCK_BREAK_CMD (was %d, should be %d).\n",
 231                          msg_len, KERNEL_OPLOCK_BREAK_MSG_LEN));
 232                 return False;
 233         }
 234
 235         memcpy((char *)inode, msg_start+KERNEL_OPLOCK_BREAK_INODE_OFFSET, sizeof(*inode));
 236         memcpy((char *)dev, msg_start+KERNEL_OPLOCK_BREAK_DEV_OFFSET, sizeof(*dev));
 237         memcpy((char *)file_id, msg_start+KERNEL_OPLOCK_BREAK_FILEID_OFFSET, sizeof(*file_id));
 238
 239         DEBUG(3,("kernel oplock break request for file dev = %x, inode = %.0f, file_id = %lu\n",
 240                 (unsigned int)*dev, (double)*inode, *file_id));
 241
 242         return True;
 243 }
 244
 245 /****************************************************************************
 246  See if a oplock message is waiting.
 247 ****************************************************************************/
 248
 249 static BOOL linux_oplock_msg_waiting(fd_set *fds)
 250 {
 251         return signals_received != 0;
 252 }
 253
 254 /****************************************************************************
 255  See if the kernel supports oplocks.
 256 ****************************************************************************/
 257
 258 static BOOL linux_oplocks_available(void)
 259 {
 260         int fd, ret;
 261         fd = open("/dev/null", O_RDONLY);
 262         if (fd == -1)
 263                 return False; /* uggh! */
 264         ret = fcntl(fd, F_GETLEASE, 0);
 265         close(fd);
 266         return ret == F_UNLCK;
 267 }
 268
 269 /****************************************************************************
 270  Setup kernel oplocks.
 271 ****************************************************************************/
 272
 273 struct kernel_oplocks *linux_init_kernel_oplocks(void)
 274 {
 275         static struct kernel_oplocks koplocks;
 276         struct sigaction act;
 277
 278         if (!linux_oplocks_available()) {
 279                 DEBUG(3,("Linux kernel oplocks not available\n"));
 280                 return NULL;
 281         }
 282
 283         act.sa_handler = NULL;
 284         act.sa_sigaction = signal_handler;
 285         act.sa_flags = SA_SIGINFO;
 286         if (sigaction(RT_SIGNAL_LEASE, &act, NULL) != 0) {
 287                 DEBUG(0,("Failed to setup RT_SIGNAL_LEASE handler\n"));
 288                 return NULL;
 289         }
 290
 291         koplocks.receive_message = linux_oplock_receive_message;
 292         koplocks.set_oplock = linux_set_kernel_oplock;
 293         koplocks.release_oplock = linux_release_kernel_oplock;
 294         koplocks.parse_message = linux_kernel_oplock_parse;
 295         koplocks.msg_waiting = linux_oplock_msg_waiting;
 296         koplocks.notification_fd = -1;
 297
 298         DEBUG(3,("Linux kernel oplocks enabled\n"));
 299
 300         return &koplocks;
 301 }
 302 #else
 303  void oplock_linux_dummy(void) {}
 304 #endif /* HAVE_KERNEL_OPLOCKS_LINUX */