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@lxorguk.ukuu.org.uk> 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/fdtable.h> 77 #include <linux/times.h> 78 #include <linux/cpuset.h> 79 #include <linux/rcupdate.h> 80 #include <linux/delayacct.h> 81 #include <linux/seq_file.h> 82 #include <linux/pid_namespace.h> 83 #include <linux/ptrace.h> 84 #include <linux/tracehook.h> 85 #include <linux/swapops.h> 86 87 #include <asm/pgtable.h> 88 #include <asm/processor.h> 89 #include "internal.h" 90 91 static inline void task_name(struct seq_file *m, struct task_struct *p) 92 { 93 int i; 94 char *buf, *end; 95 char *name; 96 char tcomm[sizeof(p->comm)]; 97 98 get_task_comm(tcomm, p); 99 100 seq_printf(m, "Name:\t"); 101 end = m->buf + m->size; 102 buf = m->buf + m->count; 103 name = tcomm; 104 i = sizeof(tcomm); 105 while (i && (buf < end)) { 106 unsigned char c = *name; 107 name++; 108 i--; 109 *buf = c; 110 if (!c) 111 break; 112 if (c == '\\') { 113 buf++; 114 if (buf < end) 115 *buf++ = c; 116 continue; 117 } 118 if (c == '\n') { 119 *buf++ = '\\'; 120 if (buf < end) 121 *buf++ = 'n'; 122 continue; 123 } 124 buf++; 125 } 126 m->count = buf - m->buf; 127 seq_printf(m, "\n"); 128 } 129 130 /* 131 * The task state array is a strange "bitmap" of 132 * reasons to sleep. Thus "running" is zero, and 133 * you can test for combinations of others with 134 * simple bit tests. 135 */ 136 static const char *task_state_array[] = { 137 "R (running)", /* 0 */ 138 "S (sleeping)", /* 1 */ 139 "D (disk sleep)", /* 2 */ 140 "T (stopped)", /* 4 */ 141 "t (tracing stop)", /* 8 */ 142 "Z (zombie)", /* 16 */ 143 "X (dead)", /* 32 */ 144 "x (dead)", /* 64 */ 145 "K (wakekill)", /* 128 */ 146 "W (waking)", /* 256 */ 147 }; 148 149 static inline const char *get_task_state(struct task_struct *tsk) 150 { 151 unsigned int state = (tsk->state & TASK_REPORT) | tsk->exit_state; 152 const char **p = &task_state_array[0]; 153 154 BUILD_BUG_ON(1 + ilog2(TASK_STATE_MAX) != ARRAY_SIZE(task_state_array)); 155 156 while (state) { 157 p++; 158 state >>= 1; 159 } 160 return *p; 161 } 162 163 static inline void task_state(struct seq_file *m, struct pid_namespace *ns, 164 struct pid *pid, struct task_struct *p) 165 { 166 struct group_info *group_info; 167 int g; 168 struct fdtable *fdt = NULL; 169 const struct cred *cred; 170 pid_t ppid, tpid; 171 172 rcu_read_lock(); 173 ppid = pid_alive(p) ? 174 task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0; 175 tpid = 0; 176 if (pid_alive(p)) { 177 struct task_struct *tracer = tracehook_tracer_task(p); 178 if (tracer) 179 tpid = task_pid_nr_ns(tracer, ns); 180 } 181 cred = get_cred((struct cred *) __task_cred(p)); 182 seq_printf(m, 183 "State:\t%s\n" 184 "Tgid:\t%d\n" 185 "Pid:\t%d\n" 186 "PPid:\t%d\n" 187 "TracerPid:\t%d\n" 188 "Uid:\t%d\t%d\t%d\t%d\n" 189 "Gid:\t%d\t%d\t%d\t%d\n", 190 get_task_state(p), 191 task_tgid_nr_ns(p, ns), 192 pid_nr_ns(pid, ns), 193 ppid, tpid, 194 cred->uid, cred->euid, cred->suid, cred->fsuid, 195 cred->gid, cred->egid, cred->sgid, cred->fsgid); 196 197 task_lock(p); 198 if (p->files) 199 fdt = files_fdtable(p->files); 200 seq_printf(m, 201 "FDSize:\t%d\n" 202 "Groups:\t", 203 fdt ? fdt->max_fds : 0); 204 rcu_read_unlock(); 205 206 group_info = cred->group_info; 207 task_unlock(p); 208 209 for (g = 0; g < min(group_info->ngroups, NGROUPS_SMALL); g++) 210 seq_printf(m, "%d ", GROUP_AT(group_info, g)); 211 put_cred(cred); 212 213 seq_printf(m, "\n"); 214 } 215 216 static void render_sigset_t(struct seq_file *m, const char *header, 217 sigset_t *set) 218 { 219 int i; 220 221 seq_printf(m, "%s", header); 222 223 i = _NSIG; 224 do { 225 int x = 0; 226 227 i -= 4; 228 if (sigismember(set, i+1)) x |= 1; 229 if (sigismember(set, i+2)) x |= 2; 230 if (sigismember(set, i+3)) x |= 4; 231 if (sigismember(set, i+4)) x |= 8; 232 seq_printf(m, "%x", x); 233 } while (i >= 4); 234 235 seq_printf(m, "\n"); 236 } 237 238 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign, 239 sigset_t *catch) 240 { 241 struct k_sigaction *k; 242 int i; 243 244 k = p->sighand->action; 245 for (i = 1; i <= _NSIG; ++i, ++k) { 246 if (k->sa.sa_handler == SIG_IGN) 247 sigaddset(ign, i); 248 else if (k->sa.sa_handler != SIG_DFL) 249 sigaddset(catch, i); 250 } 251 } 252 253 static inline void task_sig(struct seq_file *m, struct task_struct *p) 254 { 255 unsigned long flags; 256 sigset_t pending, shpending, blocked, ignored, caught; 257 int num_threads = 0; 258 unsigned long qsize = 0; 259 unsigned long qlim = 0; 260 261 sigemptyset(&pending); 262 sigemptyset(&shpending); 263 sigemptyset(&blocked); 264 sigemptyset(&ignored); 265 sigemptyset(&caught); 266 267 if (lock_task_sighand(p, &flags)) { 268 pending = p->pending.signal; 269 shpending = p->signal->shared_pending.signal; 270 blocked = p->blocked; 271 collect_sigign_sigcatch(p, &ignored, &caught); 272 num_threads = atomic_read(&p->signal->count); 273 rcu_read_lock(); /* FIXME: is this correct? */ 274 qsize = atomic_read(&__task_cred(p)->user->sigpending); 275 rcu_read_unlock(); 276 qlim = p->signal->rlim[RLIMIT_SIGPENDING].rlim_cur; 277 unlock_task_sighand(p, &flags); 278 } 279 280 seq_printf(m, "Threads:\t%d\n", num_threads); 281 seq_printf(m, "SigQ:\t%lu/%lu\n", qsize, qlim); 282 283 /* render them all */ 284 render_sigset_t(m, "SigPnd:\t", &pending); 285 render_sigset_t(m, "ShdPnd:\t", &shpending); 286 render_sigset_t(m, "SigBlk:\t", &blocked); 287 render_sigset_t(m, "SigIgn:\t", &ignored); 288 render_sigset_t(m, "SigCgt:\t", &caught); 289 } 290 291 static void render_cap_t(struct seq_file *m, const char *header, 292 kernel_cap_t *a) 293 { 294 unsigned __capi; 295 296 seq_printf(m, "%s", header); 297 CAP_FOR_EACH_U32(__capi) { 298 seq_printf(m, "%08x", 299 a->cap[(_KERNEL_CAPABILITY_U32S-1) - __capi]); 300 } 301 seq_printf(m, "\n"); 302 } 303 304 static inline void task_cap(struct seq_file *m, struct task_struct *p) 305 { 306 const struct cred *cred; 307 kernel_cap_t cap_inheritable, cap_permitted, cap_effective, cap_bset; 308 309 rcu_read_lock(); 310 cred = __task_cred(p); 311 cap_inheritable = cred->cap_inheritable; 312 cap_permitted = cred->cap_permitted; 313 cap_effective = cred->cap_effective; 314 cap_bset = cred->cap_bset; 315 rcu_read_unlock(); 316 317 render_cap_t(m, "CapInh:\t", &cap_inheritable); 318 render_cap_t(m, "CapPrm:\t", &cap_permitted); 319 render_cap_t(m, "CapEff:\t", &cap_effective); 320 render_cap_t(m, "CapBnd:\t", &cap_bset); 321 } 322 323 static inline void task_context_switch_counts(struct seq_file *m, 324 struct task_struct *p) 325 { 326 seq_printf(m, "voluntary_ctxt_switches:\t%lu\n" 327 "nonvoluntary_ctxt_switches:\t%lu\n", 328 p->nvcsw, 329 p->nivcsw); 330 } 331 332 static void task_cpus_allowed(struct seq_file *m, struct task_struct *task) 333 { 334 seq_printf(m, "Cpus_allowed:\t"); 335 seq_cpumask(m, &task->cpus_allowed); 336 seq_printf(m, "\n"); 337 seq_printf(m, "Cpus_allowed_list:\t"); 338 seq_cpumask_list(m, &task->cpus_allowed); 339 seq_printf(m, "\n"); 340 } 341 342 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns, 343 struct pid *pid, struct task_struct *task) 344 { 345 struct mm_struct *mm = get_task_mm(task); 346 347 task_name(m, task); 348 task_state(m, ns, pid, task); 349 350 if (mm) { 351 task_mem(m, mm); 352 mmput(mm); 353 } 354 task_sig(m, task); 355 task_cap(m, task); 356 task_cpus_allowed(m, task); 357 cpuset_task_status_allowed(m, task); 358 #if defined(CONFIG_S390) 359 task_show_regs(m, task); 360 #endif 361 task_context_switch_counts(m, task); 362 return 0; 363 } 364 365 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns, 366 struct pid *pid, struct task_struct *task, int whole) 367 { 368 unsigned long vsize, eip, esp, wchan = ~0UL; 369 long priority, nice; 370 int tty_pgrp = -1, tty_nr = 0; 371 sigset_t sigign, sigcatch; 372 char state; 373 pid_t ppid = 0, pgid = -1, sid = -1; 374 int num_threads = 0; 375 int permitted; 376 struct mm_struct *mm; 377 unsigned long long start_time; 378 unsigned long cmin_flt = 0, cmaj_flt = 0; 379 unsigned long min_flt = 0, maj_flt = 0; 380 cputime_t cutime, cstime, utime, stime; 381 cputime_t cgtime, gtime; 382 unsigned long rsslim = 0; 383 char tcomm[sizeof(task->comm)]; 384 unsigned long flags; 385 386 state = *get_task_state(task); 387 vsize = eip = esp = 0; 388 permitted = ptrace_may_access(task, PTRACE_MODE_READ); 389 mm = get_task_mm(task); 390 if (mm) { 391 vsize = task_vsize(mm); 392 if (permitted) { 393 eip = KSTK_EIP(task); 394 esp = KSTK_ESP(task); 395 } 396 } 397 398 get_task_comm(tcomm, task); 399 400 sigemptyset(&sigign); 401 sigemptyset(&sigcatch); 402 cutime = cstime = utime = stime = cputime_zero; 403 cgtime = gtime = cputime_zero; 404 405 if (lock_task_sighand(task, &flags)) { 406 struct signal_struct *sig = task->signal; 407 408 if (sig->tty) { 409 struct pid *pgrp = tty_get_pgrp(sig->tty); 410 tty_pgrp = pid_nr_ns(pgrp, ns); 411 put_pid(pgrp); 412 tty_nr = new_encode_dev(tty_devnum(sig->tty)); 413 } 414 415 num_threads = atomic_read(&sig->count); 416 collect_sigign_sigcatch(task, &sigign, &sigcatch); 417 418 cmin_flt = sig->cmin_flt; 419 cmaj_flt = sig->cmaj_flt; 420 cutime = sig->cutime; 421 cstime = sig->cstime; 422 cgtime = sig->cgtime; 423 rsslim = sig->rlim[RLIMIT_RSS].rlim_cur; 424 425 /* add up live thread stats at the group level */ 426 if (whole) { 427 struct task_struct *t = task; 428 do { 429 min_flt += t->min_flt; 430 maj_flt += t->maj_flt; 431 gtime = cputime_add(gtime, t->gtime); 432 t = next_thread(t); 433 } while (t != task); 434 435 min_flt += sig->min_flt; 436 maj_flt += sig->maj_flt; 437 thread_group_times(task, &utime, &stime); 438 gtime = cputime_add(gtime, sig->gtime); 439 } 440 441 sid = task_session_nr_ns(task, ns); 442 ppid = task_tgid_nr_ns(task->real_parent, ns); 443 pgid = task_pgrp_nr_ns(task, ns); 444 445 unlock_task_sighand(task, &flags); 446 } 447 448 if (permitted && (!whole || num_threads < 2)) 449 wchan = get_wchan(task); 450 if (!whole) { 451 min_flt = task->min_flt; 452 maj_flt = task->maj_flt; 453 task_times(task, &utime, &stime); 454 gtime = task->gtime; 455 } 456 457 /* scale priority and nice values from timeslices to -20..20 */ 458 /* to make it look like a "normal" Unix priority/nice value */ 459 priority = task_prio(task); 460 nice = task_nice(task); 461 462 /* Temporary variable needed for gcc-2.96 */ 463 /* convert timespec -> nsec*/ 464 start_time = 465 (unsigned long long)task->real_start_time.tv_sec * NSEC_PER_SEC 466 + task->real_start_time.tv_nsec; 467 /* convert nsec -> ticks */ 468 start_time = nsec_to_clock_t(start_time); 469 470 seq_printf(m, "%d (%s) %c %d %d %d %d %d %u %lu \ 471 %lu %lu %lu %lu %lu %ld %ld %ld %ld %d 0 %llu %lu %ld %lu %lu %lu %lu %lu \ 472 %lu %lu %lu %lu %lu %lu %lu %lu %d %d %u %u %llu %lu %ld\n", 473 pid_nr_ns(pid, ns), 474 tcomm, 475 state, 476 ppid, 477 pgid, 478 sid, 479 tty_nr, 480 tty_pgrp, 481 task->flags, 482 min_flt, 483 cmin_flt, 484 maj_flt, 485 cmaj_flt, 486 cputime_to_clock_t(utime), 487 cputime_to_clock_t(stime), 488 cputime_to_clock_t(cutime), 489 cputime_to_clock_t(cstime), 490 priority, 491 nice, 492 num_threads, 493 start_time, 494 vsize, 495 mm ? get_mm_rss(mm) : 0, 496 rsslim, 497 mm ? mm->start_code : 0, 498 mm ? mm->end_code : 0, 499 (permitted && mm) ? task->stack_start : 0, 500 esp, 501 eip, 502 /* The signal information here is obsolete. 503 * It must be decimal for Linux 2.0 compatibility. 504 * Use /proc/#/status for real-time signals. 505 */ 506 task->pending.signal.sig[0] & 0x7fffffffUL, 507 task->blocked.sig[0] & 0x7fffffffUL, 508 sigign .sig[0] & 0x7fffffffUL, 509 sigcatch .sig[0] & 0x7fffffffUL, 510 wchan, 511 0UL, 512 0UL, 513 task->exit_signal, 514 task_cpu(task), 515 task->rt_priority, 516 task->policy, 517 (unsigned long long)delayacct_blkio_ticks(task), 518 cputime_to_clock_t(gtime), 519 cputime_to_clock_t(cgtime)); 520 if (mm) 521 mmput(mm); 522 return 0; 523 } 524 525 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns, 526 struct pid *pid, struct task_struct *task) 527 { 528 return do_task_stat(m, ns, pid, task, 0); 529 } 530 531 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns, 532 struct pid *pid, struct task_struct *task) 533 { 534 return do_task_stat(m, ns, pid, task, 1); 535 } 536 537 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns, 538 struct pid *pid, struct task_struct *task) 539 { 540 int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0; 541 struct mm_struct *mm = get_task_mm(task); 542 543 if (mm) { 544 size = task_statm(mm, &shared, &text, &data, &resident); 545 mmput(mm); 546 } 547 seq_printf(m, "%d %d %d %d %d %d %d\n", 548 size, resident, shared, text, lib, data, 0); 549 550 return 0; 551 } 552