xref: /openbmc/linux/fs/proc/array.c (revision 367b8112)
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/tracehook.h>
84 
85 #include <asm/pgtable.h>
86 #include <asm/processor.h>
87 #include "internal.h"
88 
89 static inline void task_name(struct seq_file *m, struct task_struct *p)
90 {
91 	int i;
92 	char *buf, *end;
93 	char *name;
94 	char tcomm[sizeof(p->comm)];
95 
96 	get_task_comm(tcomm, p);
97 
98 	seq_printf(m, "Name:\t");
99 	end = m->buf + m->size;
100 	buf = m->buf + m->count;
101 	name = tcomm;
102 	i = sizeof(tcomm);
103 	while (i && (buf < end)) {
104 		unsigned char c = *name;
105 		name++;
106 		i--;
107 		*buf = c;
108 		if (!c)
109 			break;
110 		if (c == '\\') {
111 			buf++;
112 			if (buf < end)
113 				*buf++ = c;
114 			continue;
115 		}
116 		if (c == '\n') {
117 			*buf++ = '\\';
118 			if (buf < end)
119 				*buf++ = 'n';
120 			continue;
121 		}
122 		buf++;
123 	}
124 	m->count = buf - m->buf;
125 	seq_printf(m, "\n");
126 }
127 
128 /*
129  * The task state array is a strange "bitmap" of
130  * reasons to sleep. Thus "running" is zero, and
131  * you can test for combinations of others with
132  * simple bit tests.
133  */
134 static const char *task_state_array[] = {
135 	"R (running)",		/*  0 */
136 	"S (sleeping)",		/*  1 */
137 	"D (disk sleep)",	/*  2 */
138 	"T (stopped)",		/*  4 */
139 	"T (tracing stop)",	/*  8 */
140 	"Z (zombie)",		/* 16 */
141 	"X (dead)"		/* 32 */
142 };
143 
144 static inline const char *get_task_state(struct task_struct *tsk)
145 {
146 	unsigned int state = (tsk->state & TASK_REPORT) | tsk->exit_state;
147 	const char **p = &task_state_array[0];
148 
149 	while (state) {
150 		p++;
151 		state >>= 1;
152 	}
153 	return *p;
154 }
155 
156 static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
157 				struct pid *pid, struct task_struct *p)
158 {
159 	struct group_info *group_info;
160 	int g;
161 	struct fdtable *fdt = NULL;
162 	pid_t ppid, tpid;
163 
164 	rcu_read_lock();
165 	ppid = pid_alive(p) ?
166 		task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
167 	tpid = 0;
168 	if (pid_alive(p)) {
169 		struct task_struct *tracer = tracehook_tracer_task(p);
170 		if (tracer)
171 			tpid = task_pid_nr_ns(tracer, ns);
172 	}
173 	seq_printf(m,
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 		pid_nr_ns(pid, 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 	seq_printf(m,
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 		seq_printf(m, "%d ", GROUP_AT(group_info, g));
203 	put_group_info(group_info);
204 
205 	seq_printf(m, "\n");
206 }
207 
208 static void render_sigset_t(struct seq_file *m, const char *header,
209 				sigset_t *set)
210 {
211 	int i;
212 
213 	seq_printf(m, "%s", header);
214 
215 	i = _NSIG;
216 	do {
217 		int x = 0;
218 
219 		i -= 4;
220 		if (sigismember(set, i+1)) x |= 1;
221 		if (sigismember(set, i+2)) x |= 2;
222 		if (sigismember(set, i+3)) x |= 4;
223 		if (sigismember(set, i+4)) x |= 8;
224 		seq_printf(m, "%x", x);
225 	} while (i >= 4);
226 
227 	seq_printf(m, "\n");
228 }
229 
230 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
231 				    sigset_t *catch)
232 {
233 	struct k_sigaction *k;
234 	int i;
235 
236 	k = p->sighand->action;
237 	for (i = 1; i <= _NSIG; ++i, ++k) {
238 		if (k->sa.sa_handler == SIG_IGN)
239 			sigaddset(ign, i);
240 		else if (k->sa.sa_handler != SIG_DFL)
241 			sigaddset(catch, i);
242 	}
243 }
244 
245 static inline void task_sig(struct seq_file *m, struct task_struct *p)
246 {
247 	unsigned long flags;
248 	sigset_t pending, shpending, blocked, ignored, caught;
249 	int num_threads = 0;
250 	unsigned long qsize = 0;
251 	unsigned long qlim = 0;
252 
253 	sigemptyset(&pending);
254 	sigemptyset(&shpending);
255 	sigemptyset(&blocked);
256 	sigemptyset(&ignored);
257 	sigemptyset(&caught);
258 
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 
270 	seq_printf(m, "Threads:\t%d\n", num_threads);
271 	seq_printf(m, "SigQ:\t%lu/%lu\n", qsize, qlim);
272 
273 	/* render them all */
274 	render_sigset_t(m, "SigPnd:\t", &pending);
275 	render_sigset_t(m, "ShdPnd:\t", &shpending);
276 	render_sigset_t(m, "SigBlk:\t", &blocked);
277 	render_sigset_t(m, "SigIgn:\t", &ignored);
278 	render_sigset_t(m, "SigCgt:\t", &caught);
279 }
280 
281 static void render_cap_t(struct seq_file *m, const char *header,
282 			kernel_cap_t *a)
283 {
284 	unsigned __capi;
285 
286 	seq_printf(m, "%s", header);
287 	CAP_FOR_EACH_U32(__capi) {
288 		seq_printf(m, "%08x",
289 			   a->cap[(_KERNEL_CAPABILITY_U32S-1) - __capi]);
290 	}
291 	seq_printf(m, "\n");
292 }
293 
294 static inline void task_cap(struct seq_file *m, struct task_struct *p)
295 {
296 	render_cap_t(m, "CapInh:\t", &p->cap_inheritable);
297 	render_cap_t(m, "CapPrm:\t", &p->cap_permitted);
298 	render_cap_t(m, "CapEff:\t", &p->cap_effective);
299 	render_cap_t(m, "CapBnd:\t", &p->cap_bset);
300 }
301 
302 static inline void task_context_switch_counts(struct seq_file *m,
303 						struct task_struct *p)
304 {
305 	seq_printf(m,	"voluntary_ctxt_switches:\t%lu\n"
306 			"nonvoluntary_ctxt_switches:\t%lu\n",
307 			p->nvcsw,
308 			p->nivcsw);
309 }
310 
311 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
312 			struct pid *pid, struct task_struct *task)
313 {
314 	struct mm_struct *mm = get_task_mm(task);
315 
316 	task_name(m, task);
317 	task_state(m, ns, pid, task);
318 
319 	if (mm) {
320 		task_mem(m, mm);
321 		mmput(mm);
322 	}
323 	task_sig(m, task);
324 	task_cap(m, task);
325 	cpuset_task_status_allowed(m, task);
326 #if defined(CONFIG_S390)
327 	task_show_regs(m, task);
328 #endif
329 	task_context_switch_counts(m, task);
330 	return 0;
331 }
332 
333 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
334 			struct pid *pid, struct task_struct *task, int whole)
335 {
336 	unsigned long vsize, eip, esp, wchan = ~0UL;
337 	long priority, nice;
338 	int tty_pgrp = -1, tty_nr = 0;
339 	sigset_t sigign, sigcatch;
340 	char state;
341 	pid_t ppid = 0, pgid = -1, sid = -1;
342 	int num_threads = 0;
343 	struct mm_struct *mm;
344 	unsigned long long start_time;
345 	unsigned long cmin_flt = 0, cmaj_flt = 0;
346 	unsigned long  min_flt = 0,  maj_flt = 0;
347 	cputime_t cutime, cstime, utime, stime;
348 	cputime_t cgtime, gtime;
349 	unsigned long rsslim = 0;
350 	char tcomm[sizeof(task->comm)];
351 	unsigned long flags;
352 
353 	state = *get_task_state(task);
354 	vsize = eip = esp = 0;
355 	mm = get_task_mm(task);
356 	if (mm) {
357 		vsize = task_vsize(mm);
358 		eip = KSTK_EIP(task);
359 		esp = KSTK_ESP(task);
360 	}
361 
362 	get_task_comm(tcomm, task);
363 
364 	sigemptyset(&sigign);
365 	sigemptyset(&sigcatch);
366 	cutime = cstime = utime = stime = cputime_zero;
367 	cgtime = gtime = cputime_zero;
368 
369 	if (lock_task_sighand(task, &flags)) {
370 		struct signal_struct *sig = task->signal;
371 
372 		if (sig->tty) {
373 			struct pid *pgrp = tty_get_pgrp(sig->tty);
374 			tty_pgrp = pid_nr_ns(pgrp, ns);
375 			put_pid(pgrp);
376 			tty_nr = new_encode_dev(tty_devnum(sig->tty));
377 		}
378 
379 		num_threads = atomic_read(&sig->count);
380 		collect_sigign_sigcatch(task, &sigign, &sigcatch);
381 
382 		cmin_flt = sig->cmin_flt;
383 		cmaj_flt = sig->cmaj_flt;
384 		cutime = sig->cutime;
385 		cstime = sig->cstime;
386 		cgtime = sig->cgtime;
387 		rsslim = sig->rlim[RLIMIT_RSS].rlim_cur;
388 
389 		/* add up live thread stats at the group level */
390 		if (whole) {
391 			struct task_cputime cputime;
392 			struct task_struct *t = task;
393 			do {
394 				min_flt += t->min_flt;
395 				maj_flt += t->maj_flt;
396 				gtime = cputime_add(gtime, task_gtime(t));
397 				t = next_thread(t);
398 			} while (t != task);
399 
400 			min_flt += sig->min_flt;
401 			maj_flt += sig->maj_flt;
402 			thread_group_cputime(task, &cputime);
403 			utime = cputime.utime;
404 			stime = cputime.stime;
405 			gtime = cputime_add(gtime, sig->gtime);
406 		}
407 
408 		sid = task_session_nr_ns(task, ns);
409 		ppid = task_tgid_nr_ns(task->real_parent, ns);
410 		pgid = task_pgrp_nr_ns(task, ns);
411 
412 		unlock_task_sighand(task, &flags);
413 	}
414 
415 	if (!whole || num_threads < 2)
416 		wchan = get_wchan(task);
417 	if (!whole) {
418 		min_flt = task->min_flt;
419 		maj_flt = task->maj_flt;
420 		utime = task_utime(task);
421 		stime = task_stime(task);
422 		gtime = task_gtime(task);
423 	}
424 
425 	/* scale priority and nice values from timeslices to -20..20 */
426 	/* to make it look like a "normal" Unix priority/nice value  */
427 	priority = task_prio(task);
428 	nice = task_nice(task);
429 
430 	/* Temporary variable needed for gcc-2.96 */
431 	/* convert timespec -> nsec*/
432 	start_time =
433 		(unsigned long long)task->real_start_time.tv_sec * NSEC_PER_SEC
434 				+ task->real_start_time.tv_nsec;
435 	/* convert nsec -> ticks */
436 	start_time = nsec_to_clock_t(start_time);
437 
438 	seq_printf(m, "%d (%s) %c %d %d %d %d %d %u %lu \
439 %lu %lu %lu %lu %lu %ld %ld %ld %ld %d 0 %llu %lu %ld %lu %lu %lu %lu %lu \
440 %lu %lu %lu %lu %lu %lu %lu %lu %d %d %u %u %llu %lu %ld\n",
441 		pid_nr_ns(pid, ns),
442 		tcomm,
443 		state,
444 		ppid,
445 		pgid,
446 		sid,
447 		tty_nr,
448 		tty_pgrp,
449 		task->flags,
450 		min_flt,
451 		cmin_flt,
452 		maj_flt,
453 		cmaj_flt,
454 		cputime_to_clock_t(utime),
455 		cputime_to_clock_t(stime),
456 		cputime_to_clock_t(cutime),
457 		cputime_to_clock_t(cstime),
458 		priority,
459 		nice,
460 		num_threads,
461 		start_time,
462 		vsize,
463 		mm ? get_mm_rss(mm) : 0,
464 		rsslim,
465 		mm ? mm->start_code : 0,
466 		mm ? mm->end_code : 0,
467 		mm ? mm->start_stack : 0,
468 		esp,
469 		eip,
470 		/* The signal information here is obsolete.
471 		 * It must be decimal for Linux 2.0 compatibility.
472 		 * Use /proc/#/status for real-time signals.
473 		 */
474 		task->pending.signal.sig[0] & 0x7fffffffUL,
475 		task->blocked.sig[0] & 0x7fffffffUL,
476 		sigign      .sig[0] & 0x7fffffffUL,
477 		sigcatch    .sig[0] & 0x7fffffffUL,
478 		wchan,
479 		0UL,
480 		0UL,
481 		task->exit_signal,
482 		task_cpu(task),
483 		task->rt_priority,
484 		task->policy,
485 		(unsigned long long)delayacct_blkio_ticks(task),
486 		cputime_to_clock_t(gtime),
487 		cputime_to_clock_t(cgtime));
488 	if (mm)
489 		mmput(mm);
490 	return 0;
491 }
492 
493 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
494 			struct pid *pid, struct task_struct *task)
495 {
496 	return do_task_stat(m, ns, pid, task, 0);
497 }
498 
499 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
500 			struct pid *pid, struct task_struct *task)
501 {
502 	return do_task_stat(m, ns, pid, task, 1);
503 }
504 
505 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
506 			struct pid *pid, struct task_struct *task)
507 {
508 	int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0;
509 	struct mm_struct *mm = get_task_mm(task);
510 
511 	if (mm) {
512 		size = task_statm(mm, &shared, &text, &data, &resident);
513 		mmput(mm);
514 	}
515 	seq_printf(m, "%d %d %d %d %d %d %d\n",
516 			size, resident, shared, text, lib, data, 0);
517 
518 	return 0;
519 }
520