/* $NetBSD: xutil.c,v 1.1.1.3 2015/01/17 16:34:18 christos Exp $ */
/*
* Copyright (c) 1997-2014 Erez Zadok
* Copyright (c) 1990 Jan-Simon Pendry
* Copyright (c) 1990 Imperial College of Science, Technology & Medicine
* Copyright (c) 1990 The Regents of the University of California.
* All rights reserved.
*
* This code is derived from software contributed to Berkeley by
* Jan-Simon Pendry at Imperial College, London.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
*
* File: am-utils/libamu/xutil.c
*
*/
/*
* Miscellaneous Utilities: Logging, TTY, timers, signals, RPC, memory, etc.
*/
#ifdef HAVE_CONFIG_H
# include <config.h>
#endif /* HAVE_CONFIG_H */
#include <am_defs.h>
#include <amu.h>
/*
* Logfp is the default logging device, and is initialized to stderr by
* default in dplog/plog below, and in
* amd/amfs_program.c:amfs_program_exec().
*/
FILE *logfp = NULL;
static char *am_progname = "unknown"; /* "amd" */
static char am_hostname[MAXHOSTNAMELEN] = "unknown"; /* Hostname */
pid_t am_mypid = -1; /* process ID */
serv_state amd_state; /* amd's state */
int foreground = 1; /* 1 == this is the top-level server */
u_int debug_flags = D_CONTROL; /* set regardless if compiled with debugging */
#ifdef HAVE_SYSLOG
int syslogging;
#endif /* HAVE_SYSLOG */
static u_int xlog_level = XLOG_DEFAULT;
static u_long amd_program_number = AMQ_PROGRAM;
#ifdef DEBUG_MEM
# if defined(HAVE_MALLINFO) && defined(HAVE_MALLOC_VERIFY)
static int mem_bytes;
static int orig_mem_bytes;
# endif /* not defined(HAVE_MALLINFO) && defined(HAVE_MALLOC_VERIFY) */
#endif /* DEBUG_MEM */
/* forward definitions */
/* for GCC format string auditing */
static void real_plog(int lvl, const char *fmt, va_list vargs)
__attribute__((__format__(__printf__, 2, 0)));
#ifdef DEBUG
/*
* List of debug options.
*/
struct opt_tab dbg_opt[] =
{
{"all", D_ALL}, /* All non-disruptive options */
{"defaults", D_DEFAULT}, /* Default options */
{"test", D_TEST}, /* Full debug - no daemon, no fork, no amq, local mtab */
{"amq", D_AMQ}, /* Register for AMQ program */
{"daemon", D_DAEMON}, /* Enter daemon mode */
{"fork", D_FORK}, /* Fork server (hlfsd only) */
{"full", D_FULL}, /* Program trace */
#ifdef HAVE_CLOCK_GETTIME
{"hrtime", D_HRTIME}, /* Print high resolution time stamps */
#endif /* HAVE_CLOCK_GETTIME */
{"info", D_INFO}, /* info service specific debugging (hesiod, nis, etc) */
{"mem", D_MEM}, /* Trace memory allocations */
{"mtab", D_MTAB}, /* Use local mtab file */
{"readdir", D_READDIR}, /* Check on browsable_dirs progress */
{"str", D_STR}, /* Debug string munging */
{"trace", D_TRACE}, /* Protocol trace */
{"xdrtrace", D_XDRTRACE}, /* Trace xdr routines */
{NULL, 0}
};
#endif /* DEBUG */
/*
* List of log options
*/
struct opt_tab xlog_opt[] =
{
{"all", XLOG_ALL}, /* All messages */
{"defaults", XLOG_DEFAULT}, /* Default messages */
#ifdef DEBUG
{"debug", XLOG_DEBUG}, /* Debug messages */
#endif /* DEBUG */ /* DEBUG */
{"error", XLOG_ERROR}, /* Non-fatal system errors */
{"fatal", XLOG_FATAL}, /* Fatal errors */
{"info", XLOG_INFO}, /* Information */
{"map", XLOG_MAP}, /* Map errors */
{"stats", XLOG_STATS}, /* Additional statistical information */
{"user", XLOG_USER}, /* Non-fatal user errors */
{"warn", XLOG_WARNING}, /* Warnings */
{"warning", XLOG_WARNING}, /* Warnings */
{NULL, 0}
};
void
am_set_progname(char *pn)
{
am_progname = pn;
}
const char *
am_get_progname(void)
{
return am_progname;
}
void
am_set_hostname(char *hn)
{
xstrlcpy(am_hostname, hn, sizeof(am_hostname));
}
const char *
am_get_hostname(void)
{
return am_hostname;
}
pid_t
am_set_mypid(void)
{
am_mypid = getpid();
return am_mypid;
}
long
get_server_pid()
{
return (long) (foreground ? am_mypid : getppid());
}
voidp
xmalloc(int len)
{
voidp p;
int retries = 600;
/*
* Avoid malloc's which return NULL for malloc(0)
*/
if (len == 0)
len = 1;
do {
p = (voidp) malloc((unsigned) len);
if (p) {
if (amuDebug(D_MEM))
plog(XLOG_DEBUG, "Allocated size %d; block %p", len, p);
return p;
}
if (retries > 0) {
plog(XLOG_ERROR, "Retrying memory allocation");
sleep(1);
}
} while (--retries);
plog(XLOG_FATAL, "Out of memory");
going_down(1);
abort();
return 0;
}
/* like xmalloc, but zeros out the bytes */
voidp
xzalloc(int len)
{
voidp p = xmalloc(len);
if (p)
memset(p, 0, len);
return p;
}
voidp
xrealloc(voidp ptr, int len)
{
if (amuDebug(D_MEM))
plog(XLOG_DEBUG, "Reallocated size %d; block %p", len, ptr);
if (len == 0)
len = 1;
if (ptr)
ptr = (voidp) realloc(ptr, (unsigned) len);
else
ptr = (voidp) xmalloc((unsigned) len);
if (!ptr) {
plog(XLOG_FATAL, "Out of memory in realloc");
going_down(1);
abort();
}
return ptr;
}
#ifdef DEBUG_MEM
void
dxfree(char *file, int line, voidp ptr)
{
if (amuDebug(D_MEM))
plog(XLOG_DEBUG, "Free in %s:%d: block %p", file, line, ptr);
/* this is the only place that must NOT use XFREE()!!! */
free(ptr);
ptr = NULL; /* paranoid */
}
# if defined(HAVE_MALLINFO) && defined(HAVE_MALLOC_VERIFY)
static void
checkup_mem(void)
{
struct mallinfo mi = mallinfo();
u_long uordbytes = mi.uordblks * 4096;
if (mem_bytes != uordbytes) {
if (orig_mem_bytes == 0)
mem_bytes = orig_mem_bytes = uordbytes;
else {
fprintf(logfp, "%s[%ld]: ", am_get_progname(), (long) am_mypid);
if (mem_bytes < uordbytes) {
fprintf(logfp, "ALLOC: %ld bytes", uordbytes - mem_bytes);
} else {
fprintf(logfp, "FREE: %ld bytes", mem_bytes - uordbytes);
}
mem_bytes = uordbytes;
fprintf(logfp, ", making %d missing\n", mem_bytes - orig_mem_bytes);
}
}
malloc_verify();
}
# endif /* not defined(HAVE_MALLINFO) && defined(HAVE_MALLOC_VERIFY) */
#endif /* DEBUG_MEM */
/*
* Take a log format string and expand occurrences of %m
* with the current error code taken from errno. Make sure
* 'e' never gets longer than maxlen characters.
*/
static const char *
expand_error(const char *f, char *e, size_t maxlen)
{
const char *p;
char *q;
int error = errno;
size_t len = 0, l;
*e = '\0';
for (p = f, q = e; len < maxlen && (*q = *p); len++, q++, p++) {
if (p[0] == '%' && p[1] == 'm') {
if (len >= maxlen)
break;
xstrlcpy(q, strerror(error), maxlen - len);
l = strlen(q);
if (l != 0)
l--;
len += l;
q += l;
p++;
}
}
e[maxlen - 1] = '\0'; /* null terminate, to be sure */
return e;
}
/*
* Output the time of day and hostname to the logfile
*/
static void
show_time_host_and_name(int lvl)
{
static time_t last_t = 0;
static char *last_ctime = NULL;
time_t t;
#if defined(HAVE_CLOCK_GETTIME) && defined(DEBUG)
struct timespec ts;
#endif /* defined(HAVE_CLOCK_GETTIME) && defined(DEBUG) */
char nsecs[11]; /* '.' + 9 digits + '\0' */
char *sev;
nsecs[0] = '\0';
#if defined(HAVE_CLOCK_GETTIME) && defined(DEBUG)
/*
* Some systems (AIX 4.3) seem to implement clock_gettime() as stub
* returning ENOSYS.
*/
if (clock_gettime(CLOCK_REALTIME, &ts) == 0) {
t = ts.tv_sec;
if (amuDebug(D_HRTIME))
xsnprintf(nsecs, sizeof(nsecs), ".%09ld", ts.tv_nsec);
}
else
#endif /* defined(HAVE_CLOCK_GETTIME) && defined(DEBUG) */
t = clocktime(NULL);
if (t != last_t) {
last_ctime = ctime(&t);
last_t = t;
}
switch (lvl) {
case XLOG_FATAL:
sev = "fatal:";
break;
case XLOG_ERROR:
sev = "error:";
break;
case XLOG_USER:
sev = "user: ";
break;
case XLOG_WARNING:
sev = "warn: ";
break;
case XLOG_INFO:
sev = "info: ";
break;
case XLOG_DEBUG:
sev = "debug:";
break;
case XLOG_MAP:
sev = "map: ";
break;
case XLOG_STATS:
sev = "stats:";
break;
default:
sev = "hmm: ";
break;
}
fprintf(logfp, "%15.15s%s %s %s[%ld]/%s ",
last_ctime + 4, nsecs, am_get_hostname(),
am_get_progname(),
(long) am_mypid,
sev);
}
#ifdef DEBUG
/*
* Switch on/off debug options
*/
int
debug_option(char *opt)
{
u_int dl = debug_flags;
static int initialized_debug_flags = 0;
int rc = cmdoption(opt, dbg_opt, &dl);
if (rc) /* if got any error, don't update debug flags */
return EINVAL;
/*
* If we already initialized the debugging flags once (via amd.conf), then
* don't allow "immutable" flags to be changed again (via amq -D), because
* they could mess Amd's state and only make sense to be set once when Amd
* starts.
*/
if (initialized_debug_flags &&
debug_flags != 0 &&
(dl & D_IMMUTABLE) != (debug_flags & D_IMMUTABLE)) {
plog(XLOG_ERROR, "cannot change immutable debug flags");
/* undo any attempted change to an immutable flag */
dl = (dl & ~D_IMMUTABLE) | (debug_flags & D_IMMUTABLE);
}
initialized_debug_flags = 1;
debug_flags = dl;
return rc;
}
void
dplog(const char *fmt, ...)
{
#ifdef HAVE_SIGACTION
sigset_t old, chld;
#else /* not HAVE_SIGACTION */
int mask;
#endif /* not HAVE_SIGACTION */
va_list ap;
#ifdef HAVE_SIGACTION
sigemptyset(&chld);
sigaddset(&chld, SIGCHLD);
#else /* not HAVE_SIGACTION */
mask = sigblock(sigmask(SIGCHLD));
#endif /* not HAVE_SIGACTION */
sigprocmask(SIG_BLOCK, &chld, &old);
if (!logfp)
logfp = stderr; /* initialize before possible first use */
va_start(ap, fmt);
real_plog(XLOG_DEBUG, fmt, ap);
va_end(ap);
#ifdef HAVE_SIGACTION
sigprocmask(SIG_SETMASK, &old, NULL);
#else /* not HAVE_SIGACTION */
mask = sigblock(sigmask(SIGCHLD));
#endif /* not HAVE_SIGACTION */
}
#endif /* DEBUG */
void
plog(int lvl, const char *fmt, ...)
{
#ifdef HAVE_SIGACTION
sigset_t old, chld;
#else /* not HAVE_SIGACTION */
int mask;
#endif /* not HAVE_SIGACTION */
va_list ap;
#ifdef HAVE_SIGACTION
sigemptyset(&chld);
sigaddset(&chld, SIGCHLD);
sigprocmask(SIG_BLOCK, &chld, &old);
#else /* not HAVE_SIGACTION */
mask = sigblock(sigmask(SIGCHLD));
#endif /* not HAVE_SIGACTION */
if (!logfp)
logfp = stderr; /* initialize before possible first use */
va_start(ap, fmt);
real_plog(lvl, fmt, ap);
va_end(ap);
#ifdef HAVE_SIGACTION
sigprocmask(SIG_SETMASK, &old, NULL);
#else /* not HAVE_SIGACTION */
sigsetmask(mask);
#endif /* not HAVE_SIGACTION */
}
static void
real_plog(int lvl, const char *fmt, va_list vargs)
{
char msg[1024];
char efmt[1024];
char *ptr = msg;
static char last_msg[1024];
static int last_count = 0, last_lvl = 0;
if (!(xlog_level & lvl))
return;
#ifdef DEBUG_MEM
# if defined(HAVE_MALLINFO) && defined(HAVE_MALLOC_VERIFY)
checkup_mem();
# endif /* not defined(HAVE_MALLINFO) && defined(HAVE_MALLOC_VERIFY) */
#endif /* DEBUG_MEM */
/*
* Note: xvsnprintf() may call plog() if a truncation happened, but the
* latter has some code to break out of an infinite loop. See comment in
* xsnprintf() below.
*/
xvsnprintf(ptr, 1023, expand_error(fmt, efmt, 1024), vargs);
ptr += strlen(ptr);
if (*(ptr-1) == '\n')
*--ptr = '\0';
#ifdef HAVE_SYSLOG
if (syslogging) {
switch (lvl) { /* from mike <mcooper@usc.edu> */
case XLOG_FATAL:
lvl = LOG_CRIT;
break;
case XLOG_ERROR:
lvl = LOG_ERR;
break;
case XLOG_USER:
lvl = LOG_WARNING;
break;
case XLOG_WARNING:
lvl = LOG_WARNING;
break;
case XLOG_INFO:
lvl = LOG_INFO;
break;
case XLOG_DEBUG:
lvl = LOG_DEBUG;
break;
case XLOG_MAP:
lvl = LOG_DEBUG;
break;
case XLOG_STATS:
lvl = LOG_INFO;
break;
default:
lvl = LOG_ERR;
break;
}
syslog(lvl, "%s", msg);
return;
}
#endif /* HAVE_SYSLOG */
*ptr++ = '\n';
*ptr = '\0';
/*
* mimic syslog behavior: only write repeated strings if they differ
*/
switch (last_count) {
case 0: /* never printed at all */
last_count = 1;
if (strlcpy(last_msg, msg, sizeof(last_msg)) >= sizeof(last_msg)) /* don't use xstrlcpy here (recursive!) */
fprintf(stderr, "real_plog: string \"%s\" truncated to \"%s\"\n", last_msg, msg);
last_lvl = lvl;
show_time_host_and_name(lvl); /* mimic syslog header */
__IGNORE(fwrite(msg, ptr - msg, 1, logfp));
fflush(logfp);
break;
case 1: /* item printed once, if same, don't repeat */
if (STREQ(last_msg, msg)) {
last_count++;
} else { /* last msg printed once, new one differs */
/* last_count remains at 1 */
if (strlcpy(last_msg, msg, sizeof(last_msg)) >= sizeof(last_msg)) /* don't use xstrlcpy here (recursive!) */
fprintf(stderr, "real_plog: string \"%s\" truncated to \"%s\"\n", last_msg, msg);
last_lvl = lvl;
show_time_host_and_name(lvl); /* mimic syslog header */
__IGNORE(fwrite(msg, ptr - msg, 1, logfp));
fflush(logfp);
}
break;
case 100:
/*
* Don't allow repetitions longer than 100, so you can see when something
* cycles like crazy.
*/
show_time_host_and_name(last_lvl);
xsnprintf(last_msg, sizeof(last_msg),
"last message repeated %d times\n", last_count);
__IGNORE(fwrite(last_msg, strlen(last_msg), 1, logfp));
fflush(logfp);
last_count = 0; /* start from scratch */
break;
default: /* item repeated multiple times */
if (STREQ(last_msg, msg)) {
last_count++;
} else { /* last msg repeated+skipped, new one differs */
show_time_host_and_name(last_lvl);
xsnprintf(last_msg, sizeof(last_msg),
"last message repeated %d times\n", last_count);
__IGNORE(fwrite(last_msg, strlen(last_msg), 1, logfp));
if (strlcpy(last_msg, msg, 1024) >= 1024) /* don't use xstrlcpy here (recursive!) */
fprintf(stderr, "real_plog: string \"%s\" truncated to \"%s\"\n", last_msg, msg);
last_count = 1;
last_lvl = lvl;
show_time_host_and_name(lvl); /* mimic syslog header */
__IGNORE(fwrite(msg, ptr - msg, 1, logfp));
fflush(logfp);
}
break;
}
}
/*
* Display current debug options
*/
void
show_opts(int ch, struct opt_tab *opts)
{
int i;
int s = '{';
fprintf(stderr, "\t[-%c {no}", ch);
for (i = 0; opts[i].opt; i++) {
fprintf(stderr, "%c%s", s, opts[i].opt);
s = ',';
}
fputs("}]\n", stderr);
}
int
cmdoption(char *s, struct opt_tab *optb, u_int *flags)
{
char *p = s;
int errs = 0;
while (p && *p) {
int neg;
char *opt;
struct opt_tab *dp, *dpn = NULL;
s = p;
p = strchr(p, ',');
if (p)
*p = '\0';
/* check for "no" prefix to options */
if (s[0] == 'n' && s[1] == 'o') {
opt = s + 2;
neg = 1;
} else {
opt = s;
neg = 0;
}
/*
* Scan the array of debug options to find the
* corresponding flag value. If it is found
* then set (or clear) the flag (depending on
* whether the option was prefixed with "no").
*/
for (dp = optb; dp->opt; dp++) {
if (STREQ(opt, dp->opt))
break;
if (opt != s && !dpn && STREQ(s, dp->opt))
dpn = dp;
}
if (dp->opt || dpn) {
if (!dp->opt) {
dp = dpn;
neg = !neg;
}
if (neg)
*flags &= ~dp->flag;
else
*flags |= dp->flag;
} else {
/*
* This will log to stderr when parsing the command line
* since any -l option will not yet have taken effect.
*/
plog(XLOG_ERROR, "option \"%s\" not recognized", s);
errs++;
}
/*
* Put the comma back
*/
if (p)
*p++ = ',';
}
return errs;
}
/*
* Switch on/off logging options
*/
int
switch_option(char *opt)
{
u_int xl = xlog_level;
int rc = cmdoption(opt, xlog_opt, &xl);
if (rc) /* if got any error, don't update flags */
return EINVAL;
/*
* Don't allow "mandatory" flags to be turned off, because
* we must always be able to report on flag re/setting errors.
*/
if ((xl & XLOG_MANDATORY) != XLOG_MANDATORY) {
plog(XLOG_ERROR, "cannot turn off mandatory logging options");
xl |= XLOG_MANDATORY;
}
if (xlog_level != xl)
xlog_level = xl; /* set new flags */
return rc;
}
#ifdef LOG_DAEMON
/*
* get syslog facility to use.
* logfile can be "syslog", "syslog:daemon", "syslog:local7", etc.
*/
static int
get_syslog_facility(const char *logfile)
{
char *facstr;
/* parse facility string */
facstr = strchr(logfile, ':');
if (!facstr) /* log file was "syslog" */
return LOG_DAEMON;
facstr++;
if (!facstr || facstr[0] == '\0') { /* log file was "syslog:" */
plog(XLOG_WARNING, "null syslog facility, using LOG_DAEMON");
return LOG_DAEMON;
}
#ifdef LOG_KERN
if (STREQ(facstr, "kern"))
return LOG_KERN;
#endif /* not LOG_KERN */
#ifdef LOG_USER
if (STREQ(facstr, "user"))
return LOG_USER;
#endif /* not LOG_USER */
#ifdef LOG_MAIL
if (STREQ(facstr, "mail"))
return LOG_MAIL;
#endif /* not LOG_MAIL */
if (STREQ(facstr, "daemon"))
return LOG_DAEMON;
#ifdef LOG_AUTH
if (STREQ(facstr, "auth"))
return LOG_AUTH;
#endif /* not LOG_AUTH */
#ifdef LOG_SYSLOG
if (STREQ(facstr, "syslog"))
return LOG_SYSLOG;
#endif /* not LOG_SYSLOG */
#ifdef LOG_LPR
if (STREQ(facstr, "lpr"))
return LOG_LPR;
#endif /* not LOG_LPR */
#ifdef LOG_NEWS
if (STREQ(facstr, "news"))
return LOG_NEWS;
#endif /* not LOG_NEWS */
#ifdef LOG_UUCP
if (STREQ(facstr, "uucp"))
return LOG_UUCP;
#endif /* not LOG_UUCP */
#ifdef LOG_CRON
if (STREQ(facstr, "cron"))
return LOG_CRON;
#endif /* not LOG_CRON */
#ifdef LOG_LOCAL0
if (STREQ(facstr, "local0"))
return LOG_LOCAL0;
#endif /* not LOG_LOCAL0 */
#ifdef LOG_LOCAL1
if (STREQ(facstr, "local1"))
return LOG_LOCAL1;
#endif /* not LOG_LOCAL1 */
#ifdef LOG_LOCAL2
if (STREQ(facstr, "local2"))
return LOG_LOCAL2;
#endif /* not LOG_LOCAL2 */
#ifdef LOG_LOCAL3
if (STREQ(facstr, "local3"))
return LOG_LOCAL3;
#endif /* not LOG_LOCAL3 */
#ifdef LOG_LOCAL4
if (STREQ(facstr, "local4"))
return LOG_LOCAL4;
#endif /* not LOG_LOCAL4 */
#ifdef LOG_LOCAL5
if (STREQ(facstr, "local5"))
return LOG_LOCAL5;
#endif /* not LOG_LOCAL5 */
#ifdef LOG_LOCAL6
if (STREQ(facstr, "local6"))
return LOG_LOCAL6;
#endif /* not LOG_LOCAL6 */
#ifdef LOG_LOCAL7
if (STREQ(facstr, "local7"))
return LOG_LOCAL7;
#endif /* not LOG_LOCAL7 */
/* didn't match anything else */
plog(XLOG_WARNING, "unknown syslog facility \"%s\", using LOG_DAEMON", facstr);
return LOG_DAEMON;
}
#endif /* not LOG_DAEMON */
/*
* Change current logfile
*/
int
switch_to_logfile(char *logfile, int old_umask, int truncate_log)
{
FILE *new_logfp = stderr;
if (logfile) {
#ifdef HAVE_SYSLOG
syslogging = 0;
#endif /* HAVE_SYSLOG */
if (STREQ(logfile, "/dev/stderr"))
new_logfp = stderr;
else if (NSTREQ(logfile, "syslog", strlen("syslog"))) {
#ifdef HAVE_SYSLOG
syslogging = 1;
new_logfp = stderr;
openlog(am_get_progname(),
LOG_PID
# ifdef LOG_NOWAIT
| LOG_NOWAIT
# endif /* LOG_NOWAIT */
# ifdef LOG_DAEMON
, get_syslog_facility(logfile)
# endif /* LOG_DAEMON */
);
#else /* not HAVE_SYSLOG */
plog(XLOG_WARNING, "syslog option not supported, logging unchanged");
#endif /* not HAVE_SYSLOG */
} else { /* regular log file */
(void) umask(old_umask);
if (truncate_log)
__IGNORE(truncate(logfile, 0));
new_logfp = fopen(logfile, "a");
umask(0);
}
}
/*
* If we couldn't open a new file, then continue using the old.
*/
if (!new_logfp && logfile) {
plog(XLOG_USER, "%s: Can't open logfile: %m", logfile);
return 1;
}
/*
* Close the previous file
*/
if (logfp && logfp != stderr)
(void) fclose(logfp);
logfp = new_logfp;
if (logfile)
plog(XLOG_INFO, "switched to logfile \"%s\"", logfile);
else
plog(XLOG_INFO, "no logfile defined; using stderr");
return 0;
}
void
unregister_amq(void)
{
if (amuDebug(D_AMQ)) {
/* find which instance of amd to unregister */
u_long amd_prognum = get_amd_program_number();
if (pmap_unset(amd_prognum, AMQ_VERSION) != 1)
dlog("failed to de-register Amd program %lu, version %lu",
amd_prognum, AMQ_VERSION);
}
}
void
going_down(int rc)
{
if (foreground) {
if (amd_state != Start) {
if (amd_state != Done)
return;
unregister_amq();
}
}
#ifdef MOUNT_TABLE_ON_FILE
/*
* Call unlock_mntlist to free any important resources such as an on-disk
* lock file (/etc/mtab~).
*/
unlock_mntlist();
#endif /* MOUNT_TABLE_ON_FILE */
if (foreground) {
plog(XLOG_INFO, "Finishing with status %d", rc);
} else {
dlog("background process exiting with status %d", rc);
}
/* bye bye... */
exit(rc);
}
/* return the rpc program number under which amd was used */
u_long
get_amd_program_number(void)
{
return amd_program_number;
}
/* set the rpc program number used for amd */
void
set_amd_program_number(u_long program)
{
amd_program_number = program;
}
/*
* Release the controlling tty of the process pid.
*
* Algorithm: try these in order, if available, until one of them
* succeeds: setsid(), ioctl(fd, TIOCNOTTY, 0).
* Do not use setpgid(): on some OSs it may release the controlling tty,
* even if the man page does not mention it, but on other OSs it does not.
* Also avoid setpgrp(): it works on some systems, and on others it is
* identical to setpgid().
*/
void
amu_release_controlling_tty(void)
{
int fd;
/*
* In daemon mode, leaving open file descriptors to terminals or pipes
* can be a really bad idea.
* Case in point: the redhat startup script calls us through their 'initlog'
* program, which exits as soon as the original amd process exits. If,
* at some point, a misbehaved library function decides to print something
* to the screen, we get a SIGPIPE and die.
* And guess what: NIS glibc functions will attempt to print to stderr
* "YPBINDPROC_DOMAIN: Domain not bound" if ypbind is running but can't find
* a ypserver.
*
* So we close all of our "terminal" filedescriptors, i.e. 0, 1 and 2, then
* reopen them as /dev/null.
*
* XXX We should also probably set the SIGPIPE handler to SIG_IGN.
*/
fd = open("/dev/null", O_RDWR);
if (fd < 0) {
plog(XLOG_WARNING, "Could not open /dev/null for rw: %m");
} else {
fflush(stdin); close(0); dup2(fd, 0);
fflush(stdout); close(1); dup2(fd, 1);
fflush(stderr); close(2); dup2(fd, 2);
close(fd);
}
#ifdef HAVE_SETSID
/* XXX: one day maybe use vhangup(2) */
if (setsid() < 0) {
plog(XLOG_WARNING, "Could not release controlling tty using setsid(): %m");
} else {
plog(XLOG_INFO, "released controlling tty using setsid()");
return;
}
#endif /* HAVE_SETSID */
#ifdef TIOCNOTTY
fd = open("/dev/tty", O_RDWR);
if (fd < 0) {
/* not an error if already no controlling tty */
if (errno != ENXIO)
plog(XLOG_WARNING, "Could not open controlling tty: %m");
} else {
if (ioctl(fd, TIOCNOTTY, 0) < 0 && errno != ENOTTY)
plog(XLOG_WARNING, "Could not disassociate tty (TIOCNOTTY): %m");
else
plog(XLOG_INFO, "released controlling tty using ioctl(TIOCNOTTY)");
close(fd);
}
return;
#else
plog(XLOG_ERROR, "unable to release controlling tty");
#endif /* not TIOCNOTTY */
}
/* setup a single signal handler */
void
setup_sighandler(int signum, void (*handler)(int))
{
#ifdef HAVE_SIGACTION
struct sigaction sa;
memset(&sa, 0, sizeof(sa));
sa.sa_flags = 0; /* unnecessary */
sa.sa_handler = handler;
sigemptyset(&(sa.sa_mask)); /* probably unnecessary too */
sigaddset(&(sa.sa_mask), signum);
sigaction(signum, &sa, NULL);
#else /* not HAVE_SIGACTION */
(void) signal(signum, handler);
#endif /* not HAVE_SIGACTION */
}
/*
* Return current time in seconds. If passed a non-null argyument, then
* fill it in with the current time in seconds and microseconds (useful
* for mtime updates).
*/
time_t
clocktime(nfstime *nt)
{
static struct timeval now; /* keep last time, as default */
if (gettimeofday(&now, NULL) < 0) {
plog(XLOG_ERROR, "clocktime: gettimeofday: %m");
/* hack: force time to have incremented by at least 1 second */
now.tv_sec++;
}
/* copy seconds and microseconds. may demote a long to an int */
if (nt) {
nt->nt_seconds = (u_int) now.tv_sec;
nt->nt_useconds = (u_int) now.tv_usec;
}
return (time_t) now.tv_sec;
}
/*
* Make all the directories in the path.
*/
int
mkdirs(char *path, int mode)
{
/*
* take a copy in case path is in readonly store
*/
char *p2 = xstrdup(path);
char *sp = p2;
struct stat stb;
int error_so_far = 0;
/*
* Skip through the string make the directories.
* Mostly ignore errors - the result is tested at the end.
*
* This assumes we are root so that we can do mkdir in a
* mode 555 directory...
*/
while ((sp = strchr(sp + 1, '/'))) {
*sp = '\0';
if (mkdir(p2, mode) < 0) {
error_so_far = errno;
} else {
dlog("mkdir(%s)", p2);
}
*sp = '/';
}
if (mkdir(p2, mode) < 0) {
error_so_far = errno;
} else {
dlog("mkdir(%s)", p2);
}
XFREE(p2);
return stat(path, &stb) == 0 &&
(stb.st_mode & S_IFMT) == S_IFDIR ? 0 : error_so_far;
}
/*
* Remove as many directories in the path as possible.
* Give up if the directory doesn't appear to have
* been created by Amd (not mode dr-x) or an rmdir
* fails for any reason.
*/
void
rmdirs(char *dir)
{
char *xdp = xstrdup(dir);
char *dp;
do {
struct stat stb;
/*
* Try to find out whether this was
* created by amd. Do this by checking
* for owner write permission.
*/
if (stat(xdp, &stb) == 0 && (stb.st_mode & 0200) == 0) {
if (rmdir(xdp) < 0) {
if (errno != ENOTEMPTY &&
errno != EBUSY &&
errno != EEXIST &&
errno != EROFS &&
errno != EINVAL)
plog(XLOG_ERROR, "rmdir(%s): %m", xdp);
break;
} else {
dlog("rmdir(%s)", xdp);
}
} else {
break;
}
dp = strrchr(xdp, '/');
if (dp)
*dp = '\0';
} while (dp && dp > xdp);
XFREE(xdp);
}
/*
* Dup a string
*/
char *
xstrdup(const char *s)
{
size_t len = strlen(s);
char *sp = xmalloc(len + 1);
memcpy(sp, s, len + 1);
return sp;
}