1 /* 2 * linux/fs/proc/array.c 3 * 4 * Copyright (C) 1992 by Linus Torvalds 5 * based on ideas by Darren Senn 6 * 7 * Fixes: 8 * Michael. K. Johnson: stat,statm extensions. 9 * <johnsonm@stolaf.edu> 10 * 11 * Pauline Middelink : Made cmdline,envline only break at '\0's, to 12 * make sure SET_PROCTITLE works. Also removed 13 * bad '!' which forced address recalculation for 14 * EVERY character on the current page. 15 * <middelin@polyware.iaf.nl> 16 * 17 * Danny ter Haar : added cpuinfo 18 * <dth@cistron.nl> 19 * 20 * Alessandro Rubini : profile extension. 21 * <rubini@ipvvis.unipv.it> 22 * 23 * Jeff Tranter : added BogoMips field to cpuinfo 24 * <Jeff_Tranter@Mitel.COM> 25 * 26 * Bruno Haible : remove 4K limit for the maps file 27 * <haible@ma2s2.mathematik.uni-karlsruhe.de> 28 * 29 * Yves Arrouye : remove removal of trailing spaces in get_array. 30 * <Yves.Arrouye@marin.fdn.fr> 31 * 32 * Jerome Forissier : added per-CPU time information to /proc/stat 33 * and /proc/<pid>/cpu extension 34 * <forissier@isia.cma.fr> 35 * - Incorporation and non-SMP safe operation 36 * of forissier patch in 2.1.78 by 37 * Hans Marcus <crowbar@concepts.nl> 38 * 39 * aeb@cwi.nl : /proc/partitions 40 * 41 * 42 * Alan Cox : security fixes. 43 * <Alan.Cox@linux.org> 44 * 45 * Al Viro : safe handling of mm_struct 46 * 47 * Gerhard Wichert : added BIGMEM support 48 * Siemens AG <Gerhard.Wichert@pdb.siemens.de> 49 * 50 * Al Viro & Jeff Garzik : moved most of the thing into base.c and 51 * : proc_misc.c. The rest may eventually go into 52 * : base.c too. 53 */ 54 55 #include <linux/types.h> 56 #include <linux/errno.h> 57 #include <linux/time.h> 58 #include <linux/kernel.h> 59 #include <linux/kernel_stat.h> 60 #include <linux/tty.h> 61 #include <linux/string.h> 62 #include <linux/mman.h> 63 #include <linux/proc_fs.h> 64 #include <linux/ioport.h> 65 #include <linux/uaccess.h> 66 #include <linux/io.h> 67 #include <linux/mm.h> 68 #include <linux/hugetlb.h> 69 #include <linux/pagemap.h> 70 #include <linux/swap.h> 71 #include <linux/slab.h> 72 #include <linux/smp.h> 73 #include <linux/signal.h> 74 #include <linux/highmem.h> 75 #include <linux/file.h> 76 #include <linux/times.h> 77 #include <linux/cpuset.h> 78 #include <linux/rcupdate.h> 79 #include <linux/delayacct.h> 80 #include <linux/pid_namespace.h> 81 82 #include <asm/pgtable.h> 83 #include <asm/processor.h> 84 #include "internal.h" 85 86 /* Gcc optimizes away "strlen(x)" for constant x */ 87 #define ADDBUF(buffer, string) \ 88 do { memcpy(buffer, string, strlen(string)); \ 89 buffer += strlen(string); } while (0) 90 91 static inline char *task_name(struct task_struct *p, char *buf) 92 { 93 int i; 94 char *name; 95 char tcomm[sizeof(p->comm)]; 96 97 get_task_comm(tcomm, p); 98 99 ADDBUF(buf, "Name:\t"); 100 name = tcomm; 101 i = sizeof(tcomm); 102 do { 103 unsigned char c = *name; 104 name++; 105 i--; 106 *buf = c; 107 if (!c) 108 break; 109 if (c == '\\') { 110 buf[1] = c; 111 buf += 2; 112 continue; 113 } 114 if (c == '\n') { 115 buf[0] = '\\'; 116 buf[1] = 'n'; 117 buf += 2; 118 continue; 119 } 120 buf++; 121 } while (i); 122 *buf = '\n'; 123 return buf+1; 124 } 125 126 /* 127 * The task state array is a strange "bitmap" of 128 * reasons to sleep. Thus "running" is zero, and 129 * you can test for combinations of others with 130 * simple bit tests. 131 */ 132 static const char *task_state_array[] = { 133 "R (running)", /* 0 */ 134 "S (sleeping)", /* 1 */ 135 "D (disk sleep)", /* 2 */ 136 "T (stopped)", /* 4 */ 137 "T (tracing stop)", /* 8 */ 138 "Z (zombie)", /* 16 */ 139 "X (dead)" /* 32 */ 140 }; 141 142 static inline const char *get_task_state(struct task_struct *tsk) 143 { 144 unsigned int state = (tsk->state & (TASK_RUNNING | 145 TASK_INTERRUPTIBLE | 146 TASK_UNINTERRUPTIBLE | 147 TASK_STOPPED | 148 TASK_TRACED)) | 149 tsk->exit_state; 150 const char **p = &task_state_array[0]; 151 152 while (state) { 153 p++; 154 state >>= 1; 155 } 156 return *p; 157 } 158 159 static inline char *task_state(struct task_struct *p, char *buffer) 160 { 161 struct group_info *group_info; 162 int g; 163 struct fdtable *fdt = NULL; 164 struct pid_namespace *ns; 165 pid_t ppid, tpid; 166 167 ns = current->nsproxy->pid_ns; 168 rcu_read_lock(); 169 ppid = pid_alive(p) ? 170 task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0; 171 tpid = pid_alive(p) && p->ptrace ? 172 task_ppid_nr_ns(rcu_dereference(p->parent), ns) : 0; 173 buffer += sprintf(buffer, 174 "State:\t%s\n" 175 "Tgid:\t%d\n" 176 "Pid:\t%d\n" 177 "PPid:\t%d\n" 178 "TracerPid:\t%d\n" 179 "Uid:\t%d\t%d\t%d\t%d\n" 180 "Gid:\t%d\t%d\t%d\t%d\n", 181 get_task_state(p), 182 task_tgid_nr_ns(p, ns), 183 task_pid_nr_ns(p, ns), 184 ppid, tpid, 185 p->uid, p->euid, p->suid, p->fsuid, 186 p->gid, p->egid, p->sgid, p->fsgid); 187 188 task_lock(p); 189 if (p->files) 190 fdt = files_fdtable(p->files); 191 buffer += sprintf(buffer, 192 "FDSize:\t%d\n" 193 "Groups:\t", 194 fdt ? fdt->max_fds : 0); 195 rcu_read_unlock(); 196 197 group_info = p->group_info; 198 get_group_info(group_info); 199 task_unlock(p); 200 201 for (g = 0; g < min(group_info->ngroups, NGROUPS_SMALL); g++) 202 buffer += sprintf(buffer, "%d ", GROUP_AT(group_info, g)); 203 put_group_info(group_info); 204 205 buffer += sprintf(buffer, "\n"); 206 return buffer; 207 } 208 209 static char *render_sigset_t(const char *header, sigset_t *set, char *buffer) 210 { 211 int i, len; 212 213 len = strlen(header); 214 memcpy(buffer, header, len); 215 buffer += len; 216 217 i = _NSIG; 218 do { 219 int x = 0; 220 221 i -= 4; 222 if (sigismember(set, i+1)) x |= 1; 223 if (sigismember(set, i+2)) x |= 2; 224 if (sigismember(set, i+3)) x |= 4; 225 if (sigismember(set, i+4)) x |= 8; 226 *buffer++ = (x < 10 ? '0' : 'a' - 10) + x; 227 } while (i >= 4); 228 229 *buffer++ = '\n'; 230 *buffer = 0; 231 return buffer; 232 } 233 234 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign, 235 sigset_t *catch) 236 { 237 struct k_sigaction *k; 238 int i; 239 240 k = p->sighand->action; 241 for (i = 1; i <= _NSIG; ++i, ++k) { 242 if (k->sa.sa_handler == SIG_IGN) 243 sigaddset(ign, i); 244 else if (k->sa.sa_handler != SIG_DFL) 245 sigaddset(catch, i); 246 } 247 } 248 249 static inline char *task_sig(struct task_struct *p, char *buffer) 250 { 251 unsigned long flags; 252 sigset_t pending, shpending, blocked, ignored, caught; 253 int num_threads = 0; 254 unsigned long qsize = 0; 255 unsigned long qlim = 0; 256 257 sigemptyset(&pending); 258 sigemptyset(&shpending); 259 sigemptyset(&blocked); 260 sigemptyset(&ignored); 261 sigemptyset(&caught); 262 263 rcu_read_lock(); 264 if (lock_task_sighand(p, &flags)) { 265 pending = p->pending.signal; 266 shpending = p->signal->shared_pending.signal; 267 blocked = p->blocked; 268 collect_sigign_sigcatch(p, &ignored, &caught); 269 num_threads = atomic_read(&p->signal->count); 270 qsize = atomic_read(&p->user->sigpending); 271 qlim = p->signal->rlim[RLIMIT_SIGPENDING].rlim_cur; 272 unlock_task_sighand(p, &flags); 273 } 274 rcu_read_unlock(); 275 276 buffer += sprintf(buffer, "Threads:\t%d\n", num_threads); 277 buffer += sprintf(buffer, "SigQ:\t%lu/%lu\n", qsize, qlim); 278 279 /* render them all */ 280 buffer = render_sigset_t("SigPnd:\t", &pending, buffer); 281 buffer = render_sigset_t("ShdPnd:\t", &shpending, buffer); 282 buffer = render_sigset_t("SigBlk:\t", &blocked, buffer); 283 buffer = render_sigset_t("SigIgn:\t", &ignored, buffer); 284 buffer = render_sigset_t("SigCgt:\t", &caught, buffer); 285 286 return buffer; 287 } 288 289 static inline char *task_cap(struct task_struct *p, char *buffer) 290 { 291 return buffer + sprintf(buffer, "CapInh:\t%016x\n" 292 "CapPrm:\t%016x\n" 293 "CapEff:\t%016x\n", 294 cap_t(p->cap_inheritable), 295 cap_t(p->cap_permitted), 296 cap_t(p->cap_effective)); 297 } 298 299 static inline char *task_context_switch_counts(struct task_struct *p, 300 char *buffer) 301 { 302 return buffer + sprintf(buffer, "voluntary_ctxt_switches:\t%lu\n" 303 "nonvoluntary_ctxt_switches:\t%lu\n", 304 p->nvcsw, 305 p->nivcsw); 306 } 307 308 int proc_pid_status(struct task_struct *task, char *buffer) 309 { 310 char *orig = buffer; 311 struct mm_struct *mm = get_task_mm(task); 312 313 buffer = task_name(task, buffer); 314 buffer = task_state(task, buffer); 315 316 if (mm) { 317 buffer = task_mem(mm, buffer); 318 mmput(mm); 319 } 320 buffer = task_sig(task, buffer); 321 buffer = task_cap(task, buffer); 322 buffer = cpuset_task_status_allowed(task, buffer); 323 #if defined(CONFIG_S390) 324 buffer = task_show_regs(task, buffer); 325 #endif 326 buffer = task_context_switch_counts(task, buffer); 327 return buffer - orig; 328 } 329 330 /* 331 * Use precise platform statistics if available: 332 */ 333 #ifdef CONFIG_VIRT_CPU_ACCOUNTING 334 static cputime_t task_utime(struct task_struct *p) 335 { 336 return p->utime; 337 } 338 339 static cputime_t task_stime(struct task_struct *p) 340 { 341 return p->stime; 342 } 343 #else 344 static cputime_t task_utime(struct task_struct *p) 345 { 346 clock_t utime = cputime_to_clock_t(p->utime), 347 total = utime + cputime_to_clock_t(p->stime); 348 u64 temp; 349 350 /* 351 * Use CFS's precise accounting: 352 */ 353 temp = (u64)nsec_to_clock_t(p->se.sum_exec_runtime); 354 355 if (total) { 356 temp *= utime; 357 do_div(temp, total); 358 } 359 utime = (clock_t)temp; 360 361 p->prev_utime = max(p->prev_utime, clock_t_to_cputime(utime)); 362 return p->prev_utime; 363 } 364 365 static cputime_t task_stime(struct task_struct *p) 366 { 367 clock_t stime; 368 369 /* 370 * Use CFS's precise accounting. (we subtract utime from 371 * the total, to make sure the total observed by userspace 372 * grows monotonically - apps rely on that): 373 */ 374 stime = nsec_to_clock_t(p->se.sum_exec_runtime) - 375 cputime_to_clock_t(task_utime(p)); 376 377 p->prev_stime = max(p->prev_stime, clock_t_to_cputime(stime)); 378 return p->prev_stime; 379 } 380 #endif 381 382 static cputime_t task_gtime(struct task_struct *p) 383 { 384 return p->gtime; 385 } 386 387 static int do_task_stat(struct task_struct *task, char *buffer, int whole) 388 { 389 unsigned long vsize, eip, esp, wchan = ~0UL; 390 long priority, nice; 391 int tty_pgrp = -1, tty_nr = 0; 392 sigset_t sigign, sigcatch; 393 char state; 394 int res; 395 pid_t ppid = 0, pgid = -1, sid = -1; 396 int num_threads = 0; 397 struct mm_struct *mm; 398 unsigned long long start_time; 399 unsigned long cmin_flt = 0, cmaj_flt = 0; 400 unsigned long min_flt = 0, maj_flt = 0; 401 cputime_t cutime, cstime, utime, stime; 402 cputime_t cgtime, gtime; 403 unsigned long rsslim = 0; 404 char tcomm[sizeof(task->comm)]; 405 unsigned long flags; 406 struct pid_namespace *ns; 407 408 ns = current->nsproxy->pid_ns; 409 410 state = *get_task_state(task); 411 vsize = eip = esp = 0; 412 mm = get_task_mm(task); 413 if (mm) { 414 vsize = task_vsize(mm); 415 eip = KSTK_EIP(task); 416 esp = KSTK_ESP(task); 417 } 418 419 get_task_comm(tcomm, task); 420 421 sigemptyset(&sigign); 422 sigemptyset(&sigcatch); 423 cutime = cstime = utime = stime = cputime_zero; 424 cgtime = gtime = cputime_zero; 425 426 rcu_read_lock(); 427 if (lock_task_sighand(task, &flags)) { 428 struct signal_struct *sig = task->signal; 429 430 if (sig->tty) { 431 tty_pgrp = pid_nr_ns(sig->tty->pgrp, ns); 432 tty_nr = new_encode_dev(tty_devnum(sig->tty)); 433 } 434 435 num_threads = atomic_read(&sig->count); 436 collect_sigign_sigcatch(task, &sigign, &sigcatch); 437 438 cmin_flt = sig->cmin_flt; 439 cmaj_flt = sig->cmaj_flt; 440 cutime = sig->cutime; 441 cstime = sig->cstime; 442 cgtime = sig->cgtime; 443 rsslim = sig->rlim[RLIMIT_RSS].rlim_cur; 444 445 /* add up live thread stats at the group level */ 446 if (whole) { 447 struct task_struct *t = task; 448 do { 449 min_flt += t->min_flt; 450 maj_flt += t->maj_flt; 451 utime = cputime_add(utime, task_utime(t)); 452 stime = cputime_add(stime, task_stime(t)); 453 gtime = cputime_add(gtime, task_gtime(t)); 454 t = next_thread(t); 455 } while (t != task); 456 457 min_flt += sig->min_flt; 458 maj_flt += sig->maj_flt; 459 utime = cputime_add(utime, sig->utime); 460 stime = cputime_add(stime, sig->stime); 461 gtime = cputime_add(gtime, sig->gtime); 462 } 463 464 sid = task_session_nr_ns(task, ns); 465 pgid = task_pgrp_nr_ns(task, ns); 466 ppid = task_ppid_nr_ns(task, ns); 467 468 unlock_task_sighand(task, &flags); 469 } 470 rcu_read_unlock(); 471 472 if (!whole || num_threads < 2) 473 wchan = get_wchan(task); 474 if (!whole) { 475 min_flt = task->min_flt; 476 maj_flt = task->maj_flt; 477 utime = task_utime(task); 478 stime = task_stime(task); 479 gtime = task_gtime(task); 480 } 481 482 /* scale priority and nice values from timeslices to -20..20 */ 483 /* to make it look like a "normal" Unix priority/nice value */ 484 priority = task_prio(task); 485 nice = task_nice(task); 486 487 /* Temporary variable needed for gcc-2.96 */ 488 /* convert timespec -> nsec*/ 489 start_time = 490 (unsigned long long)task->real_start_time.tv_sec * NSEC_PER_SEC 491 + task->real_start_time.tv_nsec; 492 /* convert nsec -> ticks */ 493 start_time = nsec_to_clock_t(start_time); 494 495 res = sprintf(buffer, "%d (%s) %c %d %d %d %d %d %u %lu \ 496 %lu %lu %lu %lu %lu %ld %ld %ld %ld %d 0 %llu %lu %ld %lu %lu %lu %lu %lu \ 497 %lu %lu %lu %lu %lu %lu %lu %lu %d %d %u %u %llu %lu %ld\n", 498 task_pid_nr_ns(task, ns), 499 tcomm, 500 state, 501 ppid, 502 pgid, 503 sid, 504 tty_nr, 505 tty_pgrp, 506 task->flags, 507 min_flt, 508 cmin_flt, 509 maj_flt, 510 cmaj_flt, 511 cputime_to_clock_t(utime), 512 cputime_to_clock_t(stime), 513 cputime_to_clock_t(cutime), 514 cputime_to_clock_t(cstime), 515 priority, 516 nice, 517 num_threads, 518 start_time, 519 vsize, 520 mm ? get_mm_rss(mm) : 0, 521 rsslim, 522 mm ? mm->start_code : 0, 523 mm ? mm->end_code : 0, 524 mm ? mm->start_stack : 0, 525 esp, 526 eip, 527 /* The signal information here is obsolete. 528 * It must be decimal for Linux 2.0 compatibility. 529 * Use /proc/#/status for real-time signals. 530 */ 531 task->pending.signal.sig[0] & 0x7fffffffUL, 532 task->blocked.sig[0] & 0x7fffffffUL, 533 sigign .sig[0] & 0x7fffffffUL, 534 sigcatch .sig[0] & 0x7fffffffUL, 535 wchan, 536 0UL, 537 0UL, 538 task->exit_signal, 539 task_cpu(task), 540 task->rt_priority, 541 task->policy, 542 (unsigned long long)delayacct_blkio_ticks(task), 543 cputime_to_clock_t(gtime), 544 cputime_to_clock_t(cgtime)); 545 if (mm) 546 mmput(mm); 547 return res; 548 } 549 550 int proc_tid_stat(struct task_struct *task, char *buffer) 551 { 552 return do_task_stat(task, buffer, 0); 553 } 554 555 int proc_tgid_stat(struct task_struct *task, char *buffer) 556 { 557 return do_task_stat(task, buffer, 1); 558 } 559 560 int proc_pid_statm(struct task_struct *task, char *buffer) 561 { 562 int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0; 563 struct mm_struct *mm = get_task_mm(task); 564 565 if (mm) { 566 size = task_statm(mm, &shared, &text, &data, &resident); 567 mmput(mm); 568 } 569 570 return sprintf(buffer, "%d %d %d %d %d %d %d\n", 571 size, resident, shared, text, lib, data, 0); 572 } 573