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
81#include <asm/pgtable.h>
82#include <asm/processor.h>
83#include "internal.h"
84
85/* Gcc optimizes away "strlen(x)" for constant x */
86#define ADDBUF(buffer, string) \
87do { memcpy(buffer, string, strlen(string)); \
88     buffer += strlen(string); } while (0)
89
90static inline char *task_name(struct task_struct *p, char *buf)
91{
92	int i;
93	char *name;
94	char tcomm[sizeof(p->comm)];
95
96	get_task_comm(tcomm, p);
97
98	ADDBUF(buf, "Name:\t");
99	name = tcomm;
100	i = sizeof(tcomm);
101	do {
102		unsigned char c = *name;
103		name++;
104		i--;
105		*buf = c;
106		if (!c)
107			break;
108		if (c == '\\') {
109			buf[1] = c;
110			buf += 2;
111			continue;
112		}
113		if (c == '\n') {
114			buf[0] = '\\';
115			buf[1] = 'n';
116			buf += 2;
117			continue;
118		}
119		buf++;
120	} while (i);
121	*buf = '\n';
122	return buf+1;
123}
124
125/*
126 * The task state array is a strange "bitmap" of
127 * reasons to sleep. Thus "running" is zero, and
128 * you can test for combinations of others with
129 * simple bit tests.
130 */
131static const char *task_state_array[] = {
132	"R (running)",		/*  0 */
133	"S (sleeping)",		/*  1 */
134	"D (disk sleep)",	/*  2 */
135	"T (stopped)",		/*  4 */
136	"T (tracing stop)",	/*  8 */
137	"Z (zombie)",		/* 16 */
138	"X (dead)"		/* 32 */
139};
140
141static inline const char *get_task_state(struct task_struct *tsk)
142{
143	unsigned int state = (tsk->state & (TASK_RUNNING |
144					    TASK_INTERRUPTIBLE |
145					    TASK_UNINTERRUPTIBLE |
146					    TASK_STOPPED |
147					    TASK_TRACED)) |
148			(tsk->exit_state & (EXIT_ZOMBIE |
149					    EXIT_DEAD));
150	const char **p = &task_state_array[0];
151
152	while (state) {
153		p++;
154		state >>= 1;
155	}
156	return *p;
157}
158
159static 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
165	rcu_read_lock();
166	buffer += sprintf(buffer,
167		"State:\t%s\n"
168		"Tgid:\t%d\n"
169		"Pid:\t%d\n"
170		"PPid:\t%d\n"
171		"TracerPid:\t%d\n"
172		"Uid:\t%d\t%d\t%d\t%d\n"
173		"Gid:\t%d\t%d\t%d\t%d\n",
174		get_task_state(p),
175		p->tgid, p->pid,
176		pid_alive(p) ? rcu_dereference(p->real_parent)->tgid : 0,
177		pid_alive(p) && p->ptrace ? rcu_dereference(p->parent)->pid : 0,
178		p->uid, p->euid, p->suid, p->fsuid,
179		p->gid, p->egid, p->sgid, p->fsgid);
180
181	task_lock(p);
182	if (p->files)
183		fdt = files_fdtable(p->files);
184	buffer += sprintf(buffer,
185		"FDSize:\t%d\n"
186		"Groups:\t",
187		fdt ? fdt->max_fds : 0);
188	rcu_read_unlock();
189
190	group_info = p->group_info;
191	get_group_info(group_info);
192	task_unlock(p);
193
194	for (g = 0; g < min(group_info->ngroups, NGROUPS_SMALL); g++)
195		buffer += sprintf(buffer, "%d ", GROUP_AT(group_info, g));
196	put_group_info(group_info);
197
198	buffer += sprintf(buffer, "\n");
199	return buffer;
200}
201
202static char *render_sigset_t(const char *header, sigset_t *set, char *buffer)
203{
204	int i, len;
205
206	len = strlen(header);
207	memcpy(buffer, header, len);
208	buffer += len;
209
210	i = _NSIG;
211	do {
212		int x = 0;
213
214		i -= 4;
215		if (sigismember(set, i+1)) x |= 1;
216		if (sigismember(set, i+2)) x |= 2;
217		if (sigismember(set, i+3)) x |= 4;
218		if (sigismember(set, i+4)) x |= 8;
219		*buffer++ = (x < 10 ? '0' : 'a' - 10) + x;
220	} while (i >= 4);
221
222	*buffer++ = '\n';
223	*buffer = 0;
224	return buffer;
225}
226
227static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
228				    sigset_t *catch)
229{
230	struct k_sigaction *k;
231	int i;
232
233	k = p->sighand->action;
234	for (i = 1; i <= _NSIG; ++i, ++k) {
235		if (k->sa.sa_handler == SIG_IGN)
236			sigaddset(ign, i);
237		else if (k->sa.sa_handler != SIG_DFL)
238			sigaddset(catch, i);
239	}
240}
241
242static inline char *task_sig(struct task_struct *p, char *buffer)
243{
244	unsigned long flags;
245	sigset_t pending, shpending, blocked, ignored, caught;
246	int num_threads = 0;
247	unsigned long qsize = 0;
248	unsigned long qlim = 0;
249
250	sigemptyset(&pending);
251	sigemptyset(&shpending);
252	sigemptyset(&blocked);
253	sigemptyset(&ignored);
254	sigemptyset(&caught);
255
256	rcu_read_lock();
257	if (lock_task_sighand(p, &flags)) {
258		pending = p->pending.signal;
259		shpending = p->signal->shared_pending.signal;
260		blocked = p->blocked;
261		collect_sigign_sigcatch(p, &ignored, &caught);
262		num_threads = atomic_read(&p->signal->count);
263		qsize = atomic_read(&p->user->sigpending);
264		qlim = p->signal->rlim[RLIMIT_SIGPENDING].rlim_cur;
265		unlock_task_sighand(p, &flags);
266	}
267	rcu_read_unlock();
268
269	buffer += sprintf(buffer, "Threads:\t%d\n", num_threads);
270	buffer += sprintf(buffer, "SigQ:\t%lu/%lu\n", qsize, qlim);
271
272	/* render them all */
273	buffer = render_sigset_t("SigPnd:\t", &pending, buffer);
274	buffer = render_sigset_t("ShdPnd:\t", &shpending, buffer);
275	buffer = render_sigset_t("SigBlk:\t", &blocked, buffer);
276	buffer = render_sigset_t("SigIgn:\t", &ignored, buffer);
277	buffer = render_sigset_t("SigCgt:\t", &caught, buffer);
278
279	return buffer;
280}
281
282static inline char *task_cap(struct task_struct *p, char *buffer)
283{
284    return buffer + sprintf(buffer, "CapInh:\t%016x\n"
285			    "CapPrm:\t%016x\n"
286			    "CapEff:\t%016x\n",
287			    cap_t(p->cap_inheritable),
288			    cap_t(p->cap_permitted),
289			    cap_t(p->cap_effective));
290}
291
292static inline char *task_context_switch_counts(struct task_struct *p,
293						char *buffer)
294{
295	return buffer + sprintf(buffer, "voluntary_ctxt_switches:\t%lu\n"
296			    "nonvoluntary_ctxt_switches:\t%lu\n",
297			    p->nvcsw,
298			    p->nivcsw);
299}
300
301int proc_pid_status(struct task_struct *task, char *buffer)
302{
303	char *orig = buffer;
304	struct mm_struct *mm = get_task_mm(task);
305
306	buffer = task_name(task, buffer);
307	buffer = task_state(task, buffer);
308
309	if (mm) {
310		buffer = task_mem(mm, buffer);
311		mmput(mm);
312	}
313	buffer = task_sig(task, buffer);
314	buffer = task_cap(task, buffer);
315	buffer = cpuset_task_status_allowed(task, buffer);
316#if defined(CONFIG_S390)
317	buffer = task_show_regs(task, buffer);
318#endif
319	buffer = task_context_switch_counts(task, buffer);
320	return buffer - orig;
321}
322
323/*
324 * Use precise platform statistics if available:
325 */
326#ifdef CONFIG_VIRT_CPU_ACCOUNTING
327static cputime_t task_utime(struct task_struct *p)
328{
329	return p->utime;
330}
331
332static cputime_t task_stime(struct task_struct *p)
333{
334	return p->stime;
335}
336#else
337static cputime_t task_utime(struct task_struct *p)
338{
339	clock_t utime = cputime_to_clock_t(p->utime),
340		total = utime + cputime_to_clock_t(p->stime);
341	u64 temp;
342
343	/*
344	 * Use CFS's precise accounting:
345	 */
346	temp = (u64)nsec_to_clock_t(p->se.sum_exec_runtime);
347
348	if (total) {
349		temp *= utime;
350		do_div(temp, total);
351	}
352	utime = (clock_t)temp;
353
354	p->prev_utime = max(p->prev_utime, clock_t_to_cputime(utime));
355	return p->prev_utime;
356}
357
358static cputime_t task_stime(struct task_struct *p)
359{
360	clock_t stime;
361
362	/*
363	 * Use CFS's precise accounting. (we subtract utime from
364	 * the total, to make sure the total observed by userspace
365	 * grows monotonically - apps rely on that):
366	 */
367	stime = nsec_to_clock_t(p->se.sum_exec_runtime) -
368			cputime_to_clock_t(task_utime(p));
369
370	if (stime >= 0)
371		p->prev_stime = max(p->prev_stime, clock_t_to_cputime(stime));
372
373	return p->prev_stime;
374}
375#endif
376
377static cputime_t task_gtime(struct task_struct *p)
378{
379	return p->gtime;
380}
381
382static int do_task_stat(struct task_struct *task, char *buffer, int whole)
383{
384	unsigned long vsize, eip, esp, wchan = ~0UL;
385	long priority, nice;
386	int tty_pgrp = -1, tty_nr = 0;
387	sigset_t sigign, sigcatch;
388	char state;
389	int res;
390	pid_t ppid = 0, pgid = -1, sid = -1;
391	int num_threads = 0;
392	struct mm_struct *mm;
393	unsigned long long start_time;
394	unsigned long cmin_flt = 0, cmaj_flt = 0;
395	unsigned long  min_flt = 0,  maj_flt = 0;
396	cputime_t cutime, cstime, utime, stime;
397	cputime_t cgtime, gtime;
398	unsigned long rsslim = 0;
399	char tcomm[sizeof(task->comm)];
400	unsigned long flags;
401
402	state = *get_task_state(task);
403	vsize = eip = esp = 0;
404	mm = get_task_mm(task);
405	if (mm) {
406		vsize = task_vsize(mm);
407		eip = KSTK_EIP(task);
408		esp = KSTK_ESP(task);
409	}
410
411	get_task_comm(tcomm, task);
412
413	sigemptyset(&sigign);
414	sigemptyset(&sigcatch);
415	cutime = cstime = utime = stime = cputime_zero;
416	cgtime = gtime = cputime_zero;
417
418	rcu_read_lock();
419	if (lock_task_sighand(task, &flags)) {
420		struct signal_struct *sig = task->signal;
421
422		if (sig->tty) {
423			tty_pgrp = pid_nr(sig->tty->pgrp);
424			tty_nr = new_encode_dev(tty_devnum(sig->tty));
425		}
426
427		num_threads = atomic_read(&sig->count);
428		collect_sigign_sigcatch(task, &sigign, &sigcatch);
429
430		cmin_flt = sig->cmin_flt;
431		cmaj_flt = sig->cmaj_flt;
432		cutime = sig->cutime;
433		cstime = sig->cstime;
434		cgtime = sig->cgtime;
435		rsslim = sig->rlim[RLIMIT_RSS].rlim_cur;
436
437		/* add up live thread stats at the group level */
438		if (whole) {
439			struct task_struct *t = task;
440			do {
441				min_flt += t->min_flt;
442				maj_flt += t->maj_flt;
443				utime = cputime_add(utime, task_utime(t));
444				stime = cputime_add(stime, task_stime(t));
445				gtime = cputime_add(gtime, task_gtime(t));
446				t = next_thread(t);
447			} while (t != task);
448
449			min_flt += sig->min_flt;
450			maj_flt += sig->maj_flt;
451			utime = cputime_add(utime, sig->utime);
452			stime = cputime_add(stime, sig->stime);
453			gtime = cputime_add(gtime, sig->gtime);
454		}
455
456		sid = signal_session(sig);
457		pgid = process_group(task);
458		ppid = rcu_dereference(task->real_parent)->tgid;
459
460		unlock_task_sighand(task, &flags);
461	}
462	rcu_read_unlock();
463
464	if (!whole || num_threads < 2)
465		wchan = get_wchan(task);
466	if (!whole) {
467		min_flt = task->min_flt;
468		maj_flt = task->maj_flt;
469		utime = task_utime(task);
470		stime = task_stime(task);
471		gtime = task_gtime(task);
472	}
473
474	/* scale priority and nice values from timeslices to -20..20 */
475	/* to make it look like a "normal" Unix priority/nice value  */
476	priority = task_prio(task);
477	nice = task_nice(task);
478
479	/* Temporary variable needed for gcc-2.96 */
480	/* convert timespec -> nsec*/
481	start_time =
482		(unsigned long long)task->start_time.tv_sec * NSEC_PER_SEC
483				+ task->start_time.tv_nsec;
484	/* convert nsec -> ticks */
485	start_time = nsec_to_clock_t(start_time);
486
487	res = sprintf(buffer, "%d (%s) %c %d %d %d %d %d %u %lu \
488%lu %lu %lu %lu %lu %ld %ld %ld %ld %d 0 %llu %lu %ld %lu %lu %lu %lu %lu \
489%lu %lu %lu %lu %lu %lu %lu %lu %d %d %u %u %llu %lu %ld\n",
490		task->pid,
491		tcomm,
492		state,
493		ppid,
494		pgid,
495		sid,
496		tty_nr,
497		tty_pgrp,
498		task->flags,
499		min_flt,
500		cmin_flt,
501		maj_flt,
502		cmaj_flt,
503		cputime_to_clock_t(utime),
504		cputime_to_clock_t(stime),
505		cputime_to_clock_t(cutime),
506		cputime_to_clock_t(cstime),
507		priority,
508		nice,
509		num_threads,
510		start_time,
511		vsize,
512		mm ? get_mm_rss(mm) : 0,
513		rsslim,
514		mm ? mm->start_code : 0,
515		mm ? mm->end_code : 0,
516		mm ? mm->start_stack : 0,
517		esp,
518		eip,
519		/* The signal information here is obsolete.
520		 * It must be decimal for Linux 2.0 compatibility.
521		 * Use /proc/#/status for real-time signals.
522		 */
523		task->pending.signal.sig[0] & 0x7fffffffUL,
524		task->blocked.sig[0] & 0x7fffffffUL,
525		sigign      .sig[0] & 0x7fffffffUL,
526		sigcatch    .sig[0] & 0x7fffffffUL,
527		wchan,
528		0UL,
529		0UL,
530		task->exit_signal,
531		task_cpu(task),
532		task->rt_priority,
533		task->policy,
534		(unsigned long long)delayacct_blkio_ticks(task),
535		cputime_to_clock_t(gtime),
536		cputime_to_clock_t(cgtime));
537	if (mm)
538		mmput(mm);
539	return res;
540}
541
542int proc_tid_stat(struct task_struct *task, char *buffer)
543{
544	return do_task_stat(task, buffer, 0);
545}
546
547int proc_tgid_stat(struct task_struct *task, char *buffer)
548{
549	return do_task_stat(task, buffer, 1);
550}
551
552int proc_pid_statm(struct task_struct *task, char *buffer)
553{
554	int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0;
555	struct mm_struct *mm = get_task_mm(task);
556
557	if (mm) {
558		size = task_statm(mm, &shared, &text, &data, &resident);
559		mmput(mm);
560	}
561
562	return sprintf(buffer, "%d %d %d %d %d %d %d\n",
563		       size, resident, shared, text, lib, data, 0);
564}
565