usr.sbin/rpc.statd/file.c

   1 /*
   2  * Copyright (c) 1995
   3  *      A.R. Gordon (andrew.gordon@net-tel.co.uk).  All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice, this list of conditions and the following disclaimer.
  10  * 2. Redistributions in binary form must reproduce the above copyright
  11  *    notice, this list of conditions and the following disclaimer in the
  12  *    documentation and/or other materials provided with the distribution.
  13  * 3. All advertising materials mentioning features or use of this software
  14  *    must display the following acknowledgement:
  15  *      This product includes software developed for the FreeBSD project
  16  * 4. Neither the name of the author nor the names of any co-contributors
  17  *    may be used to endorse or promote products derived from this software
  18  *    without specific prior written permission.
  19  *
  20  * THIS SOFTWARE IS PROVIDED BY ANDREW GORDON AND CONTRIBUTORS ``AS IS'' AND
  21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
  24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  30  * SUCH DAMAGE.
  31  *
  32  * $DragonFly: src/usr.sbin/rpc.statd/file.c,v 1.2 2005/11/25 00:32:49 swildner Exp $
  33  */
  34
  35 #include <err.h>
  36 #include <errno.h>
  37 #include <fcntl.h>
  38 #include <stdio.h>
  39 #include <string.h>
  40 #include <unistd.h>
  41 #include <sys/types.h>
  42 #include <sys/mman.h>           /* For mmap()                           */
  43 #include <rpc/rpc.h>
  44 #include <syslog.h>
  45
  46 #include "statd.h"
  47
  48 FileLayout *status_info;        /* Pointer to the mmap()ed status file  */
  49 static int status_fd;           /* File descriptor for the open file    */
  50 static off_t status_file_len;   /* Current on-disc length of file       */
  51
  52 /* sync_file --------------------------------------------------------------- */
  53 /*
  54    Purpose:     Packaged call of msync() to flush changes to mmap()ed file
  55    Returns:     Nothing.  Errors to syslog.
  56 */
  57
  58 void
  59 sync_file(void)
  60 {
  61   if (msync((void *)status_info, 0, 0) < 0)
  62   {
  63     syslog(LOG_ERR, "msync() failed: %s", strerror(errno));
  64   }
  65 }
  66
  67 /* find_host -------------------------------------------------------------- */
  68 /*
  69    Purpose:     Find the entry in the status file for a given host
  70    Returns:     Pointer to that entry in the mmap() region, or NULL.
  71    Notes:       Also creates entries if requested.
  72                 Failure to create also returns NULL.
  73 */
  74
  75 HostInfo *
  76 find_host(char *hostname, int create)
  77 {
  78   HostInfo *hp;
  79   HostInfo *spare_slot = NULL;
  80   HostInfo *result = NULL;
  81   int i;
  82
  83   for (i = 0, hp = status_info->hosts; i < status_info->noOfHosts; i++, hp++)
  84   {
  85     if (!strncasecmp(hostname, hp->hostname, SM_MAXSTRLEN))
  86     {
  87       result = hp;
  88       break;
  89     }
  90     if (!spare_slot && !hp->monList && !hp->notifyReqd)
  91       spare_slot = hp;
  92   }
  93
  94   /* Return if entry found, or if not asked to create one.              */
  95   if (result || !create) return (result);
  96
  97   /* Now create an entry, using the spare slot if one was found or      */
  98   /* adding to the end of the list otherwise, extending file if reqd    */
  99   if (!spare_slot)
 100   {
 101     off_t desired_size;
 102     spare_slot = &status_info->hosts[status_info->noOfHosts];
 103     desired_size = ((char*)spare_slot - (char*)status_info) + sizeof(HostInfo);
 104     if (desired_size > status_file_len)
 105     {
 106       /* Extend file by writing 1 byte of junk at the desired end pos   */
 107       lseek(status_fd, desired_size - 1, SEEK_SET);
 108       i = write(status_fd, &i, 1);
 109       if (i < 1)
 110       {
 111         syslog(LOG_ERR, "Unable to extend status file");
 112         return (NULL);
 113       }
 114       status_file_len = desired_size;
 115     }
 116     status_info->noOfHosts++;
 117   }
 118
 119   /* Initialise the spare slot that has been found/created              */
 120   /* Note that we do not msync(), since the caller is presumed to be    */
 121   /* about to modify the entry further                                  */
 122   memset(spare_slot, 0, sizeof(HostInfo));
 123   strncpy(spare_slot->hostname, hostname, SM_MAXSTRLEN);
 124   return (spare_slot);
 125 }
 126
 127 /* init_file -------------------------------------------------------------- */
 128 /*
 129    Purpose:     Open file, create if necessary, initialise it.
 130    Returns:     Nothing - exits on error
 131    Notes:       Called before process becomes daemon, hence logs to
 132                 stderr rather than syslog.
 133                 Opens the file, then mmap()s it for ease of access.
 134                 Also performs initial clean-up of the file, zeroing
 135                 monitor list pointers, setting the notifyReqd flag in
 136                 all hosts that had a monitor list, and incrementing
 137                 the state number to the next even value.
 138 */
 139
 140 void
 141 init_file(char *filename)
 142 {
 143   int new_file = FALSE;
 144   char buf[HEADER_LEN];
 145   int i;
 146
 147   /* try to open existing file - if not present, create one             */
 148   status_fd = open(filename, O_RDWR);
 149   if ((status_fd < 0) && (errno == ENOENT))
 150   {
 151     status_fd = open(filename, O_RDWR | O_CREAT, 0644);
 152     new_file = TRUE;
 153   }
 154   if (status_fd < 0)
 155     errx(1, "unable to open status file %s", filename);
 156
 157   /* File now open.  mmap() it, with a generous size to allow for       */
 158   /* later growth, where we will extend the file but not re-map it.     */
 159   status_info = (FileLayout *)
 160     mmap(NULL, 0x10000000, PROT_READ | PROT_WRITE, MAP_SHARED, status_fd, 0);
 161
 162   if (status_info == (FileLayout *) MAP_FAILED)
 163     warn("unable to mmap() status file");
 164
 165   status_file_len = lseek(status_fd, 0L, SEEK_END);
 166
 167   /* If the file was not newly created, validate the contents, and if   */
 168   /* defective, re-create from scratch.                                 */
 169   if (!new_file)
 170   {
 171     if ((status_file_len < HEADER_LEN) || (status_file_len
 172       < (HEADER_LEN + sizeof(HostInfo) * status_info->noOfHosts)) )
 173     {
 174       warnx("status file is corrupt");
 175       new_file = TRUE;
 176     }
 177   }
 178
 179   /* Initialisation of a new, empty file.                               */
 180   if (new_file)
 181   {
 182     memset(buf, 0, sizeof(buf));
 183     lseek(status_fd, 0L, SEEK_SET);
 184     write(status_fd, buf, HEADER_LEN);
 185     status_file_len = HEADER_LEN;
 186   }
 187   else
 188   {
 189     /* Clean-up of existing file - monitored hosts will have a pointer  */
 190     /* to a list of clients, which refers to memory in the previous     */
 191     /* incarnation of the program and so are meaningless now.  These    */
 192     /* pointers are zeroed and the fact that the host was previously    */
 193     /* monitored is recorded by setting the notifyReqd flag, which will */
 194     /* in due course cause a SM_NOTIFY to be sent.                      */
 195     /* Note that if we crash twice in quick succession, some hosts may  */
 196     /* already have notifyReqd set, where we didn't manage to notify    */
 197     /* them before the second crash occurred.                           */
 198     for (i = 0; i < status_info->noOfHosts; i++)
 199     {
 200       HostInfo *this_host = &status_info->hosts[i];
 201
 202       if (this_host->monList)
 203       {
 204         this_host->notifyReqd = TRUE;
 205         this_host->monList = NULL;
 206       }
 207     }
 208     /* Select the next higher even number for the state counter         */
 209     status_info->ourState = (status_info->ourState + 2) & 0xfffffffe;
 210 /*???????******/ status_info->ourState++;
 211   }
 212 }
 213
 214 /* xdr_stat_chge ----------------------------------------------------------- */
 215 /*
 216    Purpose:     XDR-encode structure of type stat_chge
 217    Returns:     TRUE if successful
 218    Notes:       This function is missing from librpcsvc, because the
 219                 sm_inter.x distributed by Sun omits the SM_NOTIFY
 220                 procedure used between co-operating statd's
 221 */
 222
 223 bool_t
 224 xdr_stat_chge(XDR *xdrs, stat_chge *objp)
 225 {
 226   if (!xdr_string(xdrs, &objp->mon_name, SM_MAXSTRLEN))
 227   {
 228     return (FALSE);
 229   }
 230   if (!xdr_int(xdrs, &objp->state))
 231   {
 232     return (FALSE);
 233   }
 234   return (TRUE);
 235 }
 236
 237
 238 /* notify_one_host --------------------------------------------------------- */
 239 /*
 240    Purpose:     Perform SM_NOTIFY procedure at specified host
 241    Returns:     TRUE if success, FALSE if failed.
 242 */
 243
 244 static int
 245 notify_one_host(char *hostname)
 246 {
 247   struct timeval timeout = { 20, 0 };   /* 20 secs timeout              */
 248   CLIENT *cli;
 249   char dummy;
 250   stat_chge arg;
 251   char our_hostname[SM_MAXSTRLEN+1];
 252
 253   gethostname(our_hostname, sizeof(our_hostname));
 254   our_hostname[SM_MAXSTRLEN] = '\0';
 255   arg.mon_name = our_hostname;
 256   arg.state = status_info->ourState;
 257
 258   if (debug) syslog (LOG_DEBUG, "Sending SM_NOTIFY to host %s from %s", hostname, our_hostname);
 259
 260   cli = clnt_create(hostname, SM_PROG, SM_VERS, "udp");
 261   if (!cli)
 262   {
 263     syslog(LOG_ERR, "Failed to contact host %s%s", hostname,
 264       clnt_spcreateerror(""));
 265     return (FALSE);
 266   }
 267
 268   if (clnt_call(cli, SM_NOTIFY, xdr_stat_chge, &arg, xdr_void, &dummy, timeout)
 269     != RPC_SUCCESS)
 270   {
 271     syslog(LOG_ERR, "Failed to contact rpc.statd at host %s", hostname);
 272     clnt_destroy(cli);
 273     return (FALSE);
 274   }
 275
 276   clnt_destroy(cli);
 277   return (TRUE);
 278 }
 279
 280 /* notify_hosts ------------------------------------------------------------ */
 281 /*
 282    Purpose:     Send SM_NOTIFY to all hosts marked as requiring it
 283    Returns:     Nothing, immediately - forks a process to do the work.
 284    Notes:       Does nothing if there are no monitored hosts.
 285                 Called after all the initialisation has been done -
 286                 logs to syslog.
 287 */
 288
 289 void
 290 notify_hosts(void)
 291 {
 292   int i;
 293   int attempts;
 294   int work_to_do = FALSE;
 295   HostInfo *hp;
 296   pid_t pid;
 297
 298   /* First check if there is in fact any work to do.                    */
 299   for (i = status_info->noOfHosts, hp = status_info->hosts; i ; i--, hp++)
 300   {
 301     if (hp->notifyReqd)
 302     {
 303       work_to_do = TRUE;
 304       break;
 305     }
 306   }
 307
 308   if (!work_to_do) return;      /* No work found                        */
 309
 310   pid = fork();
 311   if (pid == -1)
 312   {
 313     syslog(LOG_ERR, "Unable to fork notify process - %s", strerror(errno));
 314     return;
 315   }
 316   if (pid) return;
 317
 318   /* Here in the child process.  We continue until all the hosts marked */
 319   /* as requiring notification have been duly notified.                 */
 320   /* If one of the initial attempts fails, we sleep for a while and     */
 321   /* have another go.  This is necessary because when we have crashed,  */
 322   /* (eg. a power outage) it is quite possible that we won't be able to */
 323   /* contact all monitored hosts immediately on restart, either because */
 324   /* they crashed too and take longer to come up (in which case the     */
 325   /* notification isn't really required), or more importantly if some   */
 326   /* router etc. needed to reach the monitored host has not come back   */
 327   /* up yet.  In this case, we will be a bit late in re-establishing    */
 328   /* locks (after the grace period) but that is the best we can do.     */
 329   /* We try 10 times at 5 sec intervals, 10 more times at 1 minute      */
 330   /* intervals, then 24 more times at hourly intervals, finally         */
 331   /* giving up altogether if the host hasn't come back to life after    */
 332   /* 24 hours.                                                          */
 333
 334   for (attempts = 0; attempts < 44; attempts++)
 335   {
 336     work_to_do = FALSE;         /* Unless anything fails                */
 337     for (i = status_info->noOfHosts, hp = status_info->hosts; i ; i--, hp++)
 338     {
 339       if (hp->notifyReqd)
 340       {
 341         if (notify_one_host(hp->hostname))
 342         {
 343           hp->notifyReqd = FALSE;
 344           sync_file();
 345         }
 346         else work_to_do = TRUE;
 347       }
 348     }
 349     if (!work_to_do) break;
 350     if (attempts < 10) sleep(5);
 351     else if (attempts < 20) sleep(60);
 352     else sleep(60*60);
 353   }
 354   exit(0);
 355 }
 356
 357