ctdb/common/system_util.c

   1 /*
   2    common system utilities
   3
   4    Copyright (C) Amitay Isaacs  2014
   5    Copyright (C) Martin Schwenke  2014
   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 3 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, see <http://www.gnu.org/licenses/>.
  19 */
  20
  21 #include "includes.h"
  22 #include "system/filesys.h"
  23 #include "system/shmem.h"
  24
  25 #include <libgen.h>
  26
  27 #include "ctdb_private.h"
  28
  29 #if HAVE_SCHED_H
  30 #include <sched.h>
  31 #endif
  32
  33 #if HAVE_PROCINFO_H
  34 #include <procinfo.h>
  35 #endif
  36
  37 /*
  38   if possible, make this task real time
  39  */
  40 bool set_scheduler(void)
  41 {
  42 #ifdef _AIX_
  43 #if HAVE_THREAD_SETSCHED
  44         struct thrdentry64 te;
  45         tid64_t ti;
  46
  47         ti = 0ULL;
  48         if (getthrds64(getpid(), &te, sizeof(te), &ti, 1) != 1) {
  49                 DEBUG(DEBUG_ERR, ("Unable to get thread information\n"));
  50                 return false;
  51         }
  52
  53         if (thread_setsched(te.ti_tid, 0, SCHED_RR) == -1) {
  54                 DEBUG(DEBUG_ERR, ("Unable to set scheduler to SCHED_RR (%s)\n",
  55                                   strerror(errno)));
  56                 return false;
  57         } else {
  58                 return true;
  59         }
  60 #endif
  61 #else /* no AIX */
  62 #if HAVE_SCHED_SETSCHEDULER
  63         struct sched_param p;
  64         int policy = SCHED_FIFO;
  65
  66         p.sched_priority = 1;
  67
  68 #ifdef SCHED_RESET_ON_FORK
  69         policy |= SCHED_RESET_ON_FORK;
  70 #endif
  71         if (sched_setscheduler(0, policy, &p) == -1) {
  72                 DEBUG(DEBUG_CRIT,("Unable to set scheduler to SCHED_FIFO (%s)\n",
  73                          strerror(errno)));
  74                 return false;
  75         } else {
  76                 return true;
  77         }
  78 #endif
  79 #endif
  80         DEBUG(DEBUG_CRIT,("No way to set real-time priority.\n"));
  81         return false;
  82 }
  83
  84 /*
  85   reset scheduler from real-time to normal scheduling
  86  */
  87 void reset_scheduler(void)
  88 {
  89 #ifdef _AIX_
  90 #if HAVE_THREAD_SETSCHED
  91         struct thrdentry64 te;
  92         tid64_t ti;
  93
  94         ti = 0ULL;
  95         if (getthrds64(getpid(), &te, sizeof(te), &ti, 1) != 1) {
  96                 DEBUG(DEBUG_ERR, ("Unable to get thread information\n"));
  97         }
  98         if (thread_setsched(te.ti_tid, 0, SCHED_OTHER) == -1) {
  99                 DEBUG(DEBUG_ERR, ("Unable to set scheduler to SCHED_OTHER\n"));
 100         }
 101 #endif
 102 #else /* no AIX */
 103 #if HAVE_SCHED_SETSCHEDULER
 104 #ifndef SCHED_RESET_ON_FORK
 105         struct sched_param p;
 106
 107         p.sched_priority = 0;
 108         if (sched_setscheduler(0, SCHED_OTHER, &p) == -1) {
 109                 DEBUG(DEBUG_ERR, ("Unable to set scheduler to SCHED_OTHER\n"));
 110         }
 111 #endif
 112 #endif
 113 #endif
 114 }
 115
 116 void set_nonblocking(int fd)
 117 {
 118         int v;
 119
 120         v = fcntl(fd, F_GETFL, 0);
 121         if (v == -1) {
 122                 DEBUG(DEBUG_WARNING, ("Failed to get file status flags - %s\n",
 123                                       strerror(errno)));
 124                 return;
 125         }
 126         if (fcntl(fd, F_SETFL, v | O_NONBLOCK) == -1) {
 127                 DEBUG(DEBUG_WARNING, ("Failed to set non_blocking on fd - %s\n",
 128                                       strerror(errno)));
 129         }
 130 }
 131
 132 void set_close_on_exec(int fd)
 133 {
 134         int v;
 135
 136         v = fcntl(fd, F_GETFD, 0);
 137         if (v == -1) {
 138                 DEBUG(DEBUG_WARNING, ("Failed to get file descriptor flags - %s\n",
 139                                       strerror(errno)));
 140                 return;
 141         }
 142         if (fcntl(fd, F_SETFD, v | FD_CLOEXEC) != 0) {
 143                 DEBUG(DEBUG_WARNING, ("Failed to set close_on_exec on fd - %s\n",
 144                                       strerror(errno)));
 145         }
 146 }
 147
 148
 149 bool parse_ipv4(const char *s, unsigned port, struct sockaddr_in *sin)
 150 {
 151         sin->sin_family = AF_INET;
 152         sin->sin_port   = htons(port);
 153
 154         if (inet_pton(AF_INET, s, &sin->sin_addr) != 1) {
 155                 DEBUG(DEBUG_ERR, (__location__ " Failed to translate %s into sin_addr\n", s));
 156                 return false;
 157         }
 158
 159 #ifdef HAVE_SOCK_SIN_LEN
 160         sin->sin_len = sizeof(*sin);
 161 #endif
 162         return true;
 163 }
 164
 165 static bool parse_ipv6(const char *s, const char *ifaces, unsigned port, ctdb_sock_addr *saddr)
 166 {
 167         saddr->ip6.sin6_family   = AF_INET6;
 168         saddr->ip6.sin6_port     = htons(port);
 169         saddr->ip6.sin6_flowinfo = 0;
 170         saddr->ip6.sin6_scope_id = 0;
 171
 172         if (inet_pton(AF_INET6, s, &saddr->ip6.sin6_addr) != 1) {
 173                 DEBUG(DEBUG_ERR, (__location__ " Failed to translate %s into sin6_addr\n", s));
 174                 return false;
 175         }
 176
 177         if (ifaces && IN6_IS_ADDR_LINKLOCAL(&saddr->ip6.sin6_addr)) {
 178                 if (strchr(ifaces, ',')) {
 179                         DEBUG(DEBUG_ERR, (__location__ " Link local address %s "
 180                                           "is specified for multiple ifaces %s\n",
 181                                           s, ifaces));
 182                         return false;
 183                 }
 184                 saddr->ip6.sin6_scope_id = if_nametoindex(ifaces);
 185         }
 186
 187 #ifdef HAVE_SOCK_SIN_LEN
 188         saddr->ip6.sin6_len = sizeof(*saddr);
 189 #endif
 190         return true;
 191 }
 192
 193 /*
 194   parse an ip
 195  */
 196 bool parse_ip(const char *addr, const char *ifaces, unsigned port, ctdb_sock_addr *saddr)
 197 {
 198         char *p;
 199         bool ret;
 200
 201         ZERO_STRUCTP(saddr); /* valgrind :-) */
 202
 203         /* now is this a ipv4 or ipv6 address ?*/
 204         p = index(addr, ':');
 205         if (p == NULL) {
 206                 ret = parse_ipv4(addr, port, &saddr->ip);
 207         } else {
 208                 ret = parse_ipv6(addr, ifaces, port, saddr);
 209         }
 210
 211         return ret;
 212 }
 213
 214 /*
 215   parse a ip/mask pair
 216  */
 217 bool parse_ip_mask(const char *str, const char *ifaces, ctdb_sock_addr *addr, unsigned *mask)
 218 {
 219         char *p;
 220         char s[64]; /* Much longer than INET6_ADDRSTRLEN */
 221         char *endp = NULL;
 222         ssize_t len;
 223         bool ret;
 224
 225         ZERO_STRUCT(*addr);
 226
 227         len = strlen(str);
 228         if (len >= sizeof(s)) {
 229                 DEBUG(DEBUG_ERR, ("Address %s is unreasonably long\n", str));
 230                 return false;
 231         }
 232
 233         strncpy(s, str, len+1);
 234
 235         p = rindex(s, '/');
 236         if (p == NULL) {
 237                 DEBUG(DEBUG_ERR, (__location__ " This addr: %s does not contain a mask\n", s));
 238                 return false;
 239         }
 240
 241         *mask = strtoul(p+1, &endp, 10);
 242         if (endp == NULL || *endp != 0) {
 243                 /* trailing garbage */
 244                 DEBUG(DEBUG_ERR, (__location__ " Trailing garbage after the mask in %s\n", s));
 245                 return false;
 246         }
 247         *p = 0;
 248
 249
 250         /* now is this a ipv4 or ipv6 address ?*/
 251         ret = parse_ip(s, ifaces, 0, addr);
 252
 253         return ret;
 254 }
 255
 256 /*
 257   parse a ip:port pair
 258  */
 259 bool parse_ip_port(const char *addr, ctdb_sock_addr *saddr)
 260 {
 261         char *p;
 262         char s[64]; /* Much longer than INET6_ADDRSTRLEN */
 263         unsigned port;
 264         char *endp = NULL;
 265         ssize_t len;
 266         bool ret;
 267
 268         len = strlen(addr);
 269         if (len >= sizeof(s)) {
 270                 DEBUG(DEBUG_ERR, ("Address %s is unreasonably long\n", addr));
 271                 return false;
 272         }
 273
 274         strncpy(s, addr, len+1);
 275
 276         p = rindex(s, ':');
 277         if (p == NULL) {
 278                 DEBUG(DEBUG_ERR, (__location__ " This addr: %s does not contain a port number\n", s));
 279                 return false;
 280         }
 281
 282         port = strtoul(p+1, &endp, 10);
 283         if (endp == NULL || *endp != 0) {
 284                 /* trailing garbage */
 285                 DEBUG(DEBUG_ERR, (__location__ " Trailing garbage after the port in %s\n", s));
 286                 return false;
 287         }
 288         *p = 0;
 289
 290         /* now is this a ipv4 or ipv6 address ?*/
 291         ret = parse_ip(s, NULL, port, saddr);
 292
 293         return ret;
 294 }
 295
 296 /* we don't lock future pages here; it would increase the chance that
 297  * we'd fail to mmap later on. */
 298 void lockdown_memory(bool valgrinding)
 299 {
 300 #if defined(HAVE_MLOCKALL) && !defined(_AIX_)
 301         /* Extra stack, please! */
 302         char dummy[10000];
 303         memset(dummy, 0, sizeof(dummy));
 304
 305         if (valgrinding) {
 306                 return;
 307         }
 308
 309         /* Ignore when running in local daemons mode */
 310         if (getuid() != 0) {
 311                 return;
 312         }
 313
 314         /* Avoid compiler optimizing out dummy. */
 315         mlock(dummy, sizeof(dummy));
 316         if (mlockall(MCL_CURRENT) != 0) {
 317                 DEBUG(DEBUG_WARNING,("Failed to lockdown memory: %s'\n",
 318                                      strerror(errno)));
 319         }
 320 #endif
 321 }
 322
 323 int mkdir_p(const char *dir, int mode)
 324 {
 325         char t[PATH_MAX];
 326         ssize_t len;
 327         int ret;
 328
 329         if (strcmp(dir, "/") == 0) {
 330                 return 0;
 331         }
 332
 333         if (strcmp(dir, ".") == 0) {
 334                 return 0;
 335         }
 336
 337         /* Try to create directory */
 338         ret = mkdir(dir, mode);
 339         /* Succeed if that worked or if it already existed */
 340         if (ret == 0 || errno == EEXIST) {
 341                 return 0;
 342         }
 343         /* Fail on anything else except ENOENT */
 344         if (errno != ENOENT) {
 345                 return ret;
 346         }
 347
 348         /* Create ancestors */
 349         len = strlen(dir);
 350         if (len >= PATH_MAX) {
 351                 errno = ENAMETOOLONG;
 352                 return -1;
 353         }
 354         strncpy(t, dir, len+1);
 355
 356         ret = mkdir_p(dirname(t), mode);
 357         if (ret != 0) {
 358                 return ret;
 359         }
 360
 361         /* Create directory */
 362         ret = mkdir(dir, mode);
 363         if ((ret == -1) && (errno == EEXIST)) {
 364                 ret = 0;
 365         }
 366
 367         return ret;
 368 }
 369
 370 void mkdir_p_or_die(const char *dir, int mode)
 371 {
 372         int ret;
 373
 374         ret = mkdir_p(dir, mode);
 375         if (ret != 0) {
 376                 DEBUG(DEBUG_ALERT,
 377                       ("ctdb exiting with error: "
 378                        "failed to create directory \"%s\" (%s)\n",
 379                        dir, strerror(errno)));
 380                 exit(1);
 381         }
 382 }
 383
 384 /* A read wrapper that will deal with EINTR.  For now, copied from
 385  * source3/lib/system.c
 386  */
 387 ssize_t sys_read(int fd, void *buf, size_t count)
 388 {
 389         ssize_t ret;
 390
 391         do {
 392                 ret = read(fd, buf, count);
 393 #if defined(EWOULDBLOCK)
 394         } while (ret == -1 && (errno == EINTR || errno == EAGAIN || errno == EWOULDBLOCK));
 395 #else
 396         } while (ret == -1 && (errno == EINTR || errno == EAGAIN));
 397 #endif
 398         return ret;
 399 }
 400
 401 /* A write wrapper that will deal with EINTR.  For now, copied from
 402  * source3/lib/system.c
 403  */
 404 ssize_t sys_write(int fd, const void *buf, size_t count)
 405 {
 406         ssize_t ret;
 407
 408         do {
 409                 ret = write(fd, buf, count);
 410 #if defined(EWOULDBLOCK)
 411         } while (ret == -1 && (errno == EINTR || errno == EAGAIN || errno == EWOULDBLOCK));
 412 #else
 413         } while (ret == -1 && (errno == EINTR || errno == EAGAIN));
 414 #endif
 415         return ret;
 416 }