rpc_statd.tproj/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  * $FreeBSD$
  33  *
  34  */
  35
  36 #include <err.h>
  37 #include <errno.h>
  38 #include <fcntl.h>
  39 #include <stdio.h>
  40 #include <string.h>
  41 #include <unistd.h>
  42 #include <sys/types.h>
  43 #include <sys/mman.h>           /* For mmap()                           */
  44 #include <rpc/rpc.h>
  45 #include <syslog.h>
  46 #include <stdlib.h>
  47
  48 #include "statd.h"
  49
  50 FileLayout *status_info;        /* Pointer to the mmap()ed status file  */
  51 static int status_fd;           /* File descriptor for the open file    */
  52 static off_t status_file_len;   /* Current on-disc length of file       */
  53
  54 /* sync_file --------------------------------------------------------------- */
  55 /*
  56    Purpose:     Packaged call of msync() to flush changes to mmap()ed file
  57    Returns:     Nothing.  Errors to syslog.
  58 */
  59
  60 void sync_file(void)
  61 {
  62   if (msync((void *)status_info, status_file_len, MS_SYNC) < 0 &&
  63       msync((void *)status_info, status_file_len, MS_SYNC) < 0)
  64   {
  65     syslog(LOG_ERR, "msync() failed: %s", strerror(errno));
  66   }
  67 }
  68
  69 /* find_host -------------------------------------------------------------- */
  70 /*
  71    Purpose:     Find the entry in the status file for a given host
  72    Returns:     Pointer to that entry in the mmap() region, or NULL.
  73    Notes:       Also creates entries if requested.
  74                 Failure to create also returns NULL.
  75 */
  76
  77 HostInfo *find_host(char *hostname, int create)
  78 {
  79   HostInfo *hp;
  80   HostInfo *spare_slot = NULL;
  81   HostInfo *result = NULL;
  82   int i;
  83
  84   for (i = 0, hp = status_info->hosts; i < status_info->noOfHosts; i++, hp++)
  85   {
  86     if (!strncasecmp(hostname, hp->hostname, SM_MAXSTRLEN))
  87     {
  88       result = hp;
  89       break;
  90     }
  91     if (!spare_slot && !hp->monList && !hp->notifyReqd)
  92       spare_slot = hp;
  93   }
  94
  95   /* Return if entry found, or if not asked to create one.              */
  96   if (result || !create) return (result);
  97
  98   /* Now create an entry, using the spare slot if one was found or      */
  99   /* adding to the end of the list otherwise, extending file if reqd    */
 100   if (!spare_slot)
 101   {
 102     off_t desired_size;
 103     spare_slot = &status_info->hosts[status_info->noOfHosts];
 104     desired_size = ((char*)spare_slot - (char*)status_info) + sizeof(HostInfo);
 105     if (desired_size > status_file_len)
 106     {
 107       /* Extend file by writing 1 byte of junk at the desired end pos   */
 108       lseek(status_fd, desired_size - 1, SEEK_SET);
 109       i = write(status_fd, &i, 1);
 110       if (i < 1)
 111       {
 112         syslog(LOG_ERR, "Unable to extend status file");
 113         return (NULL);
 114       }
 115       status_file_len = desired_size;
 116     }
 117     status_info->noOfHosts++;
 118   }
 119
 120   /* Initialise the spare slot that has been found/created              */
 121   /* Note that we do not msync(), since the caller is presumed to be    */
 122   /* about to modify the entry further                                  */
 123   memset(spare_slot, 0, sizeof(HostInfo));
 124   strncpy(spare_slot->hostname, hostname, SM_MAXSTRLEN);
 125   return (spare_slot);
 126 }
 127
 128 /* init_file -------------------------------------------------------------- */
 129 /*
 130    Purpose:     Open file, create if necessary, initialise it.
 131    Returns:     Nothing - exits on error
 132    Notes:       Called before process becomes daemon, hence logs to
 133                 stderr rather than syslog.
 134                 Opens the file, then mmap()s it for ease of access.
 135                 Also performs initial clean-up of the file, zeroing
 136                 monitor list pointers, setting the notifyReqd flag in
 137                 all hosts that had a monitor list, and incrementing
 138                 the state number to the next even value.
 139 */
 140
 141 void init_file(const 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   sync_file();
 213 }
 214
 215 /* notify_one_host --------------------------------------------------------- */
 216 /*
 217    Purpose:     Perform SM_NOTIFY procedure at specified host
 218    Returns:     TRUE if success, FALSE if failed.
 219 */
 220
 221 static int notify_one_host(char *hostname)
 222 {
 223   struct timeval timeout = { 20, 0 };   /* 20 secs timeout              */
 224   CLIENT *cli;
 225   char dummy;
 226   stat_chge arg;
 227   char our_hostname[SM_MAXSTRLEN+1];
 228
 229   gethostname(our_hostname, sizeof(our_hostname));
 230   our_hostname[SM_MAXSTRLEN] = '\0';
 231   arg.mon_name = our_hostname;
 232   arg.state = status_info->ourState;
 233
 234   if (debug) syslog (LOG_DEBUG, "Sending SM_NOTIFY to host %s from %s", hostname, our_hostname);
 235
 236   cli = clnt_create(hostname, SM_PROG, SM_VERS, "udp");
 237   if (!cli)
 238   {
 239     syslog(LOG_ERR, "Failed to contact host %s%s", hostname,
 240       clnt_spcreateerror(""));
 241     return (FALSE);
 242   }
 243
 244   if (clnt_call(cli, SM_NOTIFY, xdr_stat_chge, &arg, xdr_void, &dummy, timeout)
 245     != RPC_SUCCESS)
 246   {
 247     syslog(LOG_ERR, "Failed to contact rpc.statd at host %s", hostname);
 248     clnt_destroy(cli);
 249     return (FALSE);
 250   }
 251
 252   clnt_destroy(cli);
 253   return (TRUE);
 254 }
 255
 256 /* notify_hosts ------------------------------------------------------------ */
 257 /*
 258    Purpose:     Send SM_NOTIFY to all hosts marked as requiring it
 259    Returns:     Nothing, immediately - forks a process to do the work.
 260    Notes:       Does nothing if there are no monitored hosts.
 261                 Called after all the initialisation has been done -
 262                 logs to syslog.
 263 */
 264
 265 void notify_hosts(void)
 266 {
 267   int i;
 268   int attempts;
 269   int work_to_do = FALSE;
 270   HostInfo *hp;
 271   pid_t pid;
 272
 273   /* First check if there is in fact any work to do.                    */
 274   for (i = status_info->noOfHosts, hp = status_info->hosts; i ; i--, hp++)
 275   {
 276     if (hp->notifyReqd)
 277     {
 278       work_to_do = TRUE;
 279       break;
 280     }
 281   }
 282
 283   if (!work_to_do) return;      /* No work found                        */
 284
 285   pid = fork();
 286   if (pid == -1)
 287   {
 288     syslog(LOG_ERR, "Unable to fork notify process - %s", strerror(errno));
 289     return;
 290   }
 291   if (pid) return;
 292
 293   if (claim_pid_file("/var/run/statd.notify.pid", 1) < 0)
 294     errx(1, "unable to claim notify pid file");
 295
 296   /* Here in the child process.  We continue until all the hosts marked */
 297   /* as requiring notification have been duly notified.                 */
 298   /* If one of the initial attempts fails, we sleep for a while and     */
 299   /* have another go.  This is necessary because when we have crashed,  */
 300   /* (eg. a power outage) it is quite possible that we won't be able to */
 301   /* contact all monitored hosts immediately on restart, either because */
 302   /* they crashed too and take longer to come up (in which case the     */
 303   /* notification isn't really required), or more importantly if some   */
 304   /* router etc. needed to reach the monitored host has not come back   */
 305   /* up yet.  In this case, we will be a bit late in re-establishing    */
 306   /* locks (after the grace period) but that is the best we can do.     */
 307   /* We try 10 times at 5 sec intervals, 10 more times at 1 minute      */
 308   /* intervals, then 24 more times at hourly intervals, finally         */
 309   /* giving up altogether if the host hasn't come back to life after    */
 310   /* 24 hours.                                                          */
 311
 312   for (attempts = 0; attempts < 44; attempts++)
 313   {
 314     work_to_do = FALSE;         /* Unless anything fails                */
 315     for (i = status_info->noOfHosts, hp = status_info->hosts; i ; i--, hp++)
 316     {
 317       if (hp->notifyReqd)
 318       {
 319         if (notify_one_host(hp->hostname))
 320         {
 321           hp->notifyReqd = FALSE;
 322           sync_file();
 323         }
 324         else work_to_do = TRUE;
 325       }
 326     }
 327     if (!work_to_do) break;
 328     if (attempts < 10) sleep(5);
 329     else if (attempts < 20) sleep(60);
 330     else sleep(60*60);
 331   }
 332   exit(0);
 333 }
 334
 335