xref: /linux-6.15/fs/proc/array.c (revision c0a9290e)
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  *                      <[email protected]>
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  *                      <[email protected]>
16  *
17  * Danny ter Haar    :	added cpuinfo
18  *			<[email protected]>
19  *
20  * Alessandro Rubini :  profile extension.
21  *                      <[email protected]>
22  *
23  * Jeff Tranter      :  added BogoMips field to cpuinfo
24  *                      <[email protected]>
25  *
26  * Bruno Haible      :  remove 4K limit for the maps file
27  *			<[email protected]>
28  *
29  * Yves Arrouye      :  remove removal of trailing spaces in get_array.
30  *			<[email protected]>
31  *
32  * Jerome Forissier  :  added per-CPU time information to /proc/stat
33  *                      and /proc/<pid>/cpu extension
34  *                      <[email protected]>
35  *			- Incorporation and non-SMP safe operation
36  *			of forissier patch in 2.1.78 by
37  *			Hans Marcus <[email protected]>
38  *
39  * [email protected]        :  /proc/partitions
40  *
41  *
42  * Alan Cox	     :  security fixes.
43  *			<[email protected]>
44  *
45  * Al Viro           :  safe handling of mm_struct
46  *
47  * Gerhard Wichert   :  added BIGMEM support
48  * Siemens AG           <[email protected]>
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_REPORT) | tsk->exit_state;
145 	const char **p = &task_state_array[0];
146 
147 	while (state) {
148 		p++;
149 		state >>= 1;
150 	}
151 	return *p;
152 }
153 
154 static inline char *task_state(struct task_struct *p, char *buffer)
155 {
156 	struct group_info *group_info;
157 	int g;
158 	struct fdtable *fdt = NULL;
159 	struct pid_namespace *ns;
160 	pid_t ppid, tpid;
161 
162 	ns = current->nsproxy->pid_ns;
163 	rcu_read_lock();
164 	ppid = pid_alive(p) ?
165 		task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
166 	tpid = pid_alive(p) && p->ptrace ?
167 		task_pid_nr_ns(rcu_dereference(p->parent), ns) : 0;
168 	buffer += sprintf(buffer,
169 		"State:\t%s\n"
170 		"Tgid:\t%d\n"
171 		"Pid:\t%d\n"
172 		"PPid:\t%d\n"
173 		"TracerPid:\t%d\n"
174 		"Uid:\t%d\t%d\t%d\t%d\n"
175 		"Gid:\t%d\t%d\t%d\t%d\n",
176 		get_task_state(p),
177 		task_tgid_nr_ns(p, ns),
178 		task_pid_nr_ns(p, ns),
179 		ppid, tpid,
180 		p->uid, p->euid, p->suid, p->fsuid,
181 		p->gid, p->egid, p->sgid, p->fsgid);
182 
183 	task_lock(p);
184 	if (p->files)
185 		fdt = files_fdtable(p->files);
186 	buffer += sprintf(buffer,
187 		"FDSize:\t%d\n"
188 		"Groups:\t",
189 		fdt ? fdt->max_fds : 0);
190 	rcu_read_unlock();
191 
192 	group_info = p->group_info;
193 	get_group_info(group_info);
194 	task_unlock(p);
195 
196 	for (g = 0; g < min(group_info->ngroups, NGROUPS_SMALL); g++)
197 		buffer += sprintf(buffer, "%d ", GROUP_AT(group_info, g));
198 	put_group_info(group_info);
199 
200 	buffer += sprintf(buffer, "\n");
201 	return buffer;
202 }
203 
204 static char *render_sigset_t(const char *header, sigset_t *set, char *buffer)
205 {
206 	int i, len;
207 
208 	len = strlen(header);
209 	memcpy(buffer, header, len);
210 	buffer += len;
211 
212 	i = _NSIG;
213 	do {
214 		int x = 0;
215 
216 		i -= 4;
217 		if (sigismember(set, i+1)) x |= 1;
218 		if (sigismember(set, i+2)) x |= 2;
219 		if (sigismember(set, i+3)) x |= 4;
220 		if (sigismember(set, i+4)) x |= 8;
221 		*buffer++ = (x < 10 ? '0' : 'a' - 10) + x;
222 	} while (i >= 4);
223 
224 	*buffer++ = '\n';
225 	*buffer = 0;
226 	return buffer;
227 }
228 
229 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
230 				    sigset_t *catch)
231 {
232 	struct k_sigaction *k;
233 	int i;
234 
235 	k = p->sighand->action;
236 	for (i = 1; i <= _NSIG; ++i, ++k) {
237 		if (k->sa.sa_handler == SIG_IGN)
238 			sigaddset(ign, i);
239 		else if (k->sa.sa_handler != SIG_DFL)
240 			sigaddset(catch, i);
241 	}
242 }
243 
244 static inline char *task_sig(struct task_struct *p, char *buffer)
245 {
246 	unsigned long flags;
247 	sigset_t pending, shpending, blocked, ignored, caught;
248 	int num_threads = 0;
249 	unsigned long qsize = 0;
250 	unsigned long qlim = 0;
251 
252 	sigemptyset(&pending);
253 	sigemptyset(&shpending);
254 	sigemptyset(&blocked);
255 	sigemptyset(&ignored);
256 	sigemptyset(&caught);
257 
258 	rcu_read_lock();
259 	if (lock_task_sighand(p, &flags)) {
260 		pending = p->pending.signal;
261 		shpending = p->signal->shared_pending.signal;
262 		blocked = p->blocked;
263 		collect_sigign_sigcatch(p, &ignored, &caught);
264 		num_threads = atomic_read(&p->signal->count);
265 		qsize = atomic_read(&p->user->sigpending);
266 		qlim = p->signal->rlim[RLIMIT_SIGPENDING].rlim_cur;
267 		unlock_task_sighand(p, &flags);
268 	}
269 	rcu_read_unlock();
270 
271 	buffer += sprintf(buffer, "Threads:\t%d\n", num_threads);
272 	buffer += sprintf(buffer, "SigQ:\t%lu/%lu\n", qsize, qlim);
273 
274 	/* render them all */
275 	buffer = render_sigset_t("SigPnd:\t", &pending, buffer);
276 	buffer = render_sigset_t("ShdPnd:\t", &shpending, buffer);
277 	buffer = render_sigset_t("SigBlk:\t", &blocked, buffer);
278 	buffer = render_sigset_t("SigIgn:\t", &ignored, buffer);
279 	buffer = render_sigset_t("SigCgt:\t", &caught, buffer);
280 
281 	return buffer;
282 }
283 
284 static char *render_cap_t(const char *header, kernel_cap_t *a, char *buffer)
285 {
286 	unsigned __capi;
287 
288 	buffer += sprintf(buffer, "%s", header);
289 	CAP_FOR_EACH_U32(__capi) {
290 		buffer += sprintf(buffer, "%08x",
291 				  a->cap[(_LINUX_CAPABILITY_U32S-1) - __capi]);
292 	}
293 	return buffer + sprintf(buffer, "\n");
294 }
295 
296 static inline char *task_cap(struct task_struct *p, char *buffer)
297 {
298 	buffer = render_cap_t("CapInh:\t", &p->cap_inheritable, buffer);
299 	buffer = render_cap_t("CapPrm:\t", &p->cap_permitted, buffer);
300 	return render_cap_t("CapEff:\t", &p->cap_effective, buffer);
301 }
302 
303 static inline char *task_context_switch_counts(struct task_struct *p,
304 						char *buffer)
305 {
306 	return buffer + sprintf(buffer, "voluntary_ctxt_switches:\t%lu\n"
307 			    "nonvoluntary_ctxt_switches:\t%lu\n",
308 			    p->nvcsw,
309 			    p->nivcsw);
310 }
311 
312 int proc_pid_status(struct task_struct *task, char *buffer)
313 {
314 	char *orig = buffer;
315 	struct mm_struct *mm = get_task_mm(task);
316 
317 	buffer = task_name(task, buffer);
318 	buffer = task_state(task, buffer);
319 
320 	if (mm) {
321 		buffer = task_mem(mm, buffer);
322 		mmput(mm);
323 	}
324 	buffer = task_sig(task, buffer);
325 	buffer = task_cap(task, buffer);
326 	buffer = cpuset_task_status_allowed(task, buffer);
327 #if defined(CONFIG_S390)
328 	buffer = task_show_regs(task, buffer);
329 #endif
330 	buffer = task_context_switch_counts(task, buffer);
331 	return buffer - orig;
332 }
333 
334 /*
335  * Use precise platform statistics if available:
336  */
337 #ifdef CONFIG_VIRT_CPU_ACCOUNTING
338 static cputime_t task_utime(struct task_struct *p)
339 {
340 	return p->utime;
341 }
342 
343 static cputime_t task_stime(struct task_struct *p)
344 {
345 	return p->stime;
346 }
347 #else
348 static cputime_t task_utime(struct task_struct *p)
349 {
350 	clock_t utime = cputime_to_clock_t(p->utime),
351 		total = utime + cputime_to_clock_t(p->stime);
352 	u64 temp;
353 
354 	/*
355 	 * Use CFS's precise accounting:
356 	 */
357 	temp = (u64)nsec_to_clock_t(p->se.sum_exec_runtime);
358 
359 	if (total) {
360 		temp *= utime;
361 		do_div(temp, total);
362 	}
363 	utime = (clock_t)temp;
364 
365 	p->prev_utime = max(p->prev_utime, clock_t_to_cputime(utime));
366 	return p->prev_utime;
367 }
368 
369 static cputime_t task_stime(struct task_struct *p)
370 {
371 	clock_t stime;
372 
373 	/*
374 	 * Use CFS's precise accounting. (we subtract utime from
375 	 * the total, to make sure the total observed by userspace
376 	 * grows monotonically - apps rely on that):
377 	 */
378 	stime = nsec_to_clock_t(p->se.sum_exec_runtime) -
379 			cputime_to_clock_t(task_utime(p));
380 
381 	if (stime >= 0)
382 		p->prev_stime = max(p->prev_stime, clock_t_to_cputime(stime));
383 
384 	return p->prev_stime;
385 }
386 #endif
387 
388 static cputime_t task_gtime(struct task_struct *p)
389 {
390 	return p->gtime;
391 }
392 
393 static int do_task_stat(struct task_struct *task, char *buffer, int whole)
394 {
395 	unsigned long vsize, eip, esp, wchan = ~0UL;
396 	long priority, nice;
397 	int tty_pgrp = -1, tty_nr = 0;
398 	sigset_t sigign, sigcatch;
399 	char state;
400 	int res;
401 	pid_t ppid = 0, pgid = -1, sid = -1;
402 	int num_threads = 0;
403 	struct mm_struct *mm;
404 	unsigned long long start_time;
405 	unsigned long cmin_flt = 0, cmaj_flt = 0;
406 	unsigned long  min_flt = 0,  maj_flt = 0;
407 	cputime_t cutime, cstime, utime, stime;
408 	cputime_t cgtime, gtime;
409 	unsigned long rsslim = 0;
410 	char tcomm[sizeof(task->comm)];
411 	unsigned long flags;
412 	struct pid_namespace *ns;
413 
414 	ns = current->nsproxy->pid_ns;
415 
416 	state = *get_task_state(task);
417 	vsize = eip = esp = 0;
418 	mm = get_task_mm(task);
419 	if (mm) {
420 		vsize = task_vsize(mm);
421 		eip = KSTK_EIP(task);
422 		esp = KSTK_ESP(task);
423 	}
424 
425 	get_task_comm(tcomm, task);
426 
427 	sigemptyset(&sigign);
428 	sigemptyset(&sigcatch);
429 	cutime = cstime = utime = stime = cputime_zero;
430 	cgtime = gtime = cputime_zero;
431 
432 	rcu_read_lock();
433 	if (lock_task_sighand(task, &flags)) {
434 		struct signal_struct *sig = task->signal;
435 
436 		if (sig->tty) {
437 			tty_pgrp = pid_nr_ns(sig->tty->pgrp, ns);
438 			tty_nr = new_encode_dev(tty_devnum(sig->tty));
439 		}
440 
441 		num_threads = atomic_read(&sig->count);
442 		collect_sigign_sigcatch(task, &sigign, &sigcatch);
443 
444 		cmin_flt = sig->cmin_flt;
445 		cmaj_flt = sig->cmaj_flt;
446 		cutime = sig->cutime;
447 		cstime = sig->cstime;
448 		cgtime = sig->cgtime;
449 		rsslim = sig->rlim[RLIMIT_RSS].rlim_cur;
450 
451 		/* add up live thread stats at the group level */
452 		if (whole) {
453 			struct task_struct *t = task;
454 			do {
455 				min_flt += t->min_flt;
456 				maj_flt += t->maj_flt;
457 				utime = cputime_add(utime, task_utime(t));
458 				stime = cputime_add(stime, task_stime(t));
459 				gtime = cputime_add(gtime, task_gtime(t));
460 				t = next_thread(t);
461 			} while (t != task);
462 
463 			min_flt += sig->min_flt;
464 			maj_flt += sig->maj_flt;
465 			utime = cputime_add(utime, sig->utime);
466 			stime = cputime_add(stime, sig->stime);
467 			gtime = cputime_add(gtime, sig->gtime);
468 		}
469 
470 		sid = task_session_nr_ns(task, ns);
471 		ppid = task_tgid_nr_ns(task->real_parent, ns);
472 		pgid = task_pgrp_nr_ns(task, ns);
473 
474 		unlock_task_sighand(task, &flags);
475 	}
476 	rcu_read_unlock();
477 
478 	if (!whole || num_threads < 2)
479 		wchan = get_wchan(task);
480 	if (!whole) {
481 		min_flt = task->min_flt;
482 		maj_flt = task->maj_flt;
483 		utime = task_utime(task);
484 		stime = task_stime(task);
485 		gtime = task_gtime(task);
486 	}
487 
488 	/* scale priority and nice values from timeslices to -20..20 */
489 	/* to make it look like a "normal" Unix priority/nice value  */
490 	priority = task_prio(task);
491 	nice = task_nice(task);
492 
493 	/* Temporary variable needed for gcc-2.96 */
494 	/* convert timespec -> nsec*/
495 	start_time =
496 		(unsigned long long)task->real_start_time.tv_sec * NSEC_PER_SEC
497 				+ task->real_start_time.tv_nsec;
498 	/* convert nsec -> ticks */
499 	start_time = nsec_to_clock_t(start_time);
500 
501 	res = sprintf(buffer, "%d (%s) %c %d %d %d %d %d %u %lu \
502 %lu %lu %lu %lu %lu %ld %ld %ld %ld %d 0 %llu %lu %ld %lu %lu %lu %lu %lu \
503 %lu %lu %lu %lu %lu %lu %lu %lu %d %d %u %u %llu %lu %ld\n",
504 		task_pid_nr_ns(task, ns),
505 		tcomm,
506 		state,
507 		ppid,
508 		pgid,
509 		sid,
510 		tty_nr,
511 		tty_pgrp,
512 		task->flags,
513 		min_flt,
514 		cmin_flt,
515 		maj_flt,
516 		cmaj_flt,
517 		cputime_to_clock_t(utime),
518 		cputime_to_clock_t(stime),
519 		cputime_to_clock_t(cutime),
520 		cputime_to_clock_t(cstime),
521 		priority,
522 		nice,
523 		num_threads,
524 		start_time,
525 		vsize,
526 		mm ? get_mm_rss(mm) : 0,
527 		rsslim,
528 		mm ? mm->start_code : 0,
529 		mm ? mm->end_code : 0,
530 		mm ? mm->start_stack : 0,
531 		esp,
532 		eip,
533 		/* The signal information here is obsolete.
534 		 * It must be decimal for Linux 2.0 compatibility.
535 		 * Use /proc/#/status for real-time signals.
536 		 */
537 		task->pending.signal.sig[0] & 0x7fffffffUL,
538 		task->blocked.sig[0] & 0x7fffffffUL,
539 		sigign      .sig[0] & 0x7fffffffUL,
540 		sigcatch    .sig[0] & 0x7fffffffUL,
541 		wchan,
542 		0UL,
543 		0UL,
544 		task->exit_signal,
545 		task_cpu(task),
546 		task->rt_priority,
547 		task->policy,
548 		(unsigned long long)delayacct_blkio_ticks(task),
549 		cputime_to_clock_t(gtime),
550 		cputime_to_clock_t(cgtime));
551 	if (mm)
552 		mmput(mm);
553 	return res;
554 }
555 
556 int proc_tid_stat(struct task_struct *task, char *buffer)
557 {
558 	return do_task_stat(task, buffer, 0);
559 }
560 
561 int proc_tgid_stat(struct task_struct *task, char *buffer)
562 {
563 	return do_task_stat(task, buffer, 1);
564 }
565 
566 int proc_pid_statm(struct task_struct *task, char *buffer)
567 {
568 	int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0;
569 	struct mm_struct *mm = get_task_mm(task);
570 
571 	if (mm) {
572 		size = task_statm(mm, &shared, &text, &data, &resident);
573 		mmput(mm);
574 	}
575 
576 	return sprintf(buffer, "%d %d %d %d %d %d %d\n",
577 		       size, resident, shared, text, lib, data, 0);
578 }
579