utils/astcanary.c

   1 /*
   2  * Asterisk -- An open source telephony toolkit.
   3  *
   4  * Copyright (C) 2007, Digium, Inc.
   5  *
   6  * Tilghman Lesher <tlesher AT digium DOT com>
   7  *
   8  * See http://www.asterisk.org for more information about
   9  * the Asterisk project. Please do not directly contact
  10  * any of the maintainers of this project for assistance;
  11  * the project provides a web site, mailing lists and IRC
  12  * channels for your use.
  13  *
  14  * This program is free software, distributed under the terms of
  15  * the GNU General Public License Version 2. See the LICENSE file
  16  * at the top of the source tree.
  17  */
  18
  19 #include <sys/types.h>
  20 #include <sys/stat.h>
  21 #include <sys/time.h>
  22 #include <sys/resource.h>
  23 #include <utime.h>
  24 #include <fcntl.h>
  25 #include <unistd.h>
  26 #include <stdlib.h>
  27 #include <string.h>
  28
  29 /*!\brief
  30  * At one time, canaries were carried along with coal miners down
  31  * into a mine.  Their purpose was to alert the miners when they
  32  * had drilled into a pocket of methane gas or another noxious
  33  * substance.  The canary, being the most sensitive animal would
  34  * immediately fall over.  Seeing this, the miners could take
  35  * action to escape the mine, seeing an imminent danger.
  36  *
  37  * This process serves a similar purpose, though with the realtime
  38  * priority being the reason.  When a thread starts running away
  39  * with the processor, it is typically difficult to tell what
  40  * thread caused the problem, as the machine acts as if it is
  41  * locked up (in fact, what has happened is that Asterisk runs at
  42  * a higher priority than even the login shell, so the runaway
  43  * thread hogs all available CPU time.
  44  *
  45  * If that happens, this canary process will cease to get any
  46  * process time, which we can monitor with a realtime thread in
  47  * Asterisk.  Should that happen, that monitoring thread may take
  48  * immediate action to slow down Asterisk to regular priority,
  49  * thus allowing an administrator to login to the system and
  50  * restart Asterisk or perhaps take another course of action
  51  * (such as retrieving a backtrace to let the developers know
  52  * what precisely went wrong).
  53  *
  54  * Note that according to POSIX.1, all threads inside a single
  55  * process must share the same priority, so when the monitoring
  56  * thread deprioritizes itself, it deprioritizes all threads at
  57  * the same time.  This is also why this canary must exist as a
  58  * completely separate process and not simply as a thread within
  59  * Asterisk itself.
  60  */
  61
  62 const char explanation[] =
  63 "This file is created when Asterisk is run with a realtime priority (-p).  It\n"
  64 "must continue to exist, and the astcanary process must be allowed to continue\n"
  65 "running, or else the Asterisk process will, within a short period of time,\n"
  66 "slow itself down to regular priority.\n\n"
  67 "The technical explanation for this file is to provide an assurance to Asterisk\n"
  68 "that there are no threads that have gone into runaway mode, thus hogging the\n"
  69 "CPU, and making the Asterisk machine seem to be unresponsive.  When that\n"
  70 "happens, the astcanary process will be unable to update the timestamp on this\n"
  71 "file, and Asterisk will notice within 120 seconds and react.  Slowing the\n"
  72 "Asterisk process down to regular priority will permit an administrator to\n"
  73 "intervene, thus avoiding a need to reboot the entire machine.\n";
  74
  75 int main(int argc, char *argv[])
  76 {
  77         int fd;
  78         /* Run at normal priority */
  79         setpriority(PRIO_PROCESS, 0, 0);
  80         for (;;) {
  81                 /* Update the modification times (checked from Asterisk) */
  82                 if (utime(argv[1], NULL)) {
  83                         /* Recreate the file if it doesn't exist */
  84                         if ((fd = open(argv[1], O_RDWR | O_TRUNC | O_CREAT, 0777)) > -1) {
  85                                 write(fd, explanation, strlen(explanation));
  86                                 close(fd);
  87                         } else {
  88                                 exit(1);
  89                         }
  90                         continue;
  91                 }
  92
  93                 /* Run occasionally */
  94                 sleep(5);
  95         }
  96
  97         /* Never reached */
  98         return 0;
  99 }
 100