Merge ../torvalds-2.6/
[linux-2.6] / fs / proc / array.c
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/config.h>
56 #include <linux/types.h>
57 #include <linux/errno.h>
58 #include <linux/time.h>
59 #include <linux/kernel.h>
60 #include <linux/kernel_stat.h>
61 #include <linux/tty.h>
62 #include <linux/string.h>
63 #include <linux/mman.h>
64 #include <linux/proc_fs.h>
65 #include <linux/ioport.h>
66 #include <linux/mm.h>
67 #include <linux/hugetlb.h>
68 #include <linux/pagemap.h>
69 #include <linux/swap.h>
70 #include <linux/slab.h>
71 #include <linux/smp.h>
72 #include <linux/signal.h>
73 #include <linux/highmem.h>
74 #include <linux/file.h>
75 #include <linux/times.h>
76 #include <linux/cpuset.h>
77
78 #include <asm/uaccess.h>
79 #include <asm/pgtable.h>
80 #include <asm/io.h>
81 #include <asm/processor.h>
82 #include "internal.h"
83
84 /* Gcc optimizes away "strlen(x)" for constant x */
85 #define ADDBUF(buffer, string) \
86 do { memcpy(buffer, string, strlen(string)); \
87      buffer += strlen(string); } while (0)
88
89 static inline char * task_name(struct task_struct *p, char * buf)
90 {
91         int i;
92         char * name;
93         char tcomm[sizeof(p->comm)];
94
95         get_task_comm(tcomm, p);
96
97         ADDBUF(buf, "Name:\t");
98         name = tcomm;
99         i = sizeof(tcomm);
100         do {
101                 unsigned char c = *name;
102                 name++;
103                 i--;
104                 *buf = c;
105                 if (!c)
106                         break;
107                 if (c == '\\') {
108                         buf[1] = c;
109                         buf += 2;
110                         continue;
111                 }
112                 if (c == '\n') {
113                         buf[0] = '\\';
114                         buf[1] = 'n';
115                         buf += 2;
116                         continue;
117                 }
118                 buf++;
119         } while (i);
120         *buf = '\n';
121         return buf+1;
122 }
123
124 /*
125  * The task state array is a strange "bitmap" of
126  * reasons to sleep. Thus "running" is zero, and
127  * you can test for combinations of others with
128  * simple bit tests.
129  */
130 static const char *task_state_array[] = {
131         "R (running)",          /*  0 */
132         "S (sleeping)",         /*  1 */
133         "D (disk sleep)",       /*  2 */
134         "T (stopped)",          /*  4 */
135         "T (tracing stop)",     /*  8 */
136         "Z (zombie)",           /* 16 */
137         "X (dead)"              /* 32 */
138 };
139
140 static inline const char * get_task_state(struct task_struct *tsk)
141 {
142         unsigned int state = (tsk->state & (TASK_RUNNING |
143                                             TASK_INTERRUPTIBLE |
144                                             TASK_UNINTERRUPTIBLE |
145                                             TASK_STOPPED |
146                                             TASK_TRACED)) |
147                         (tsk->exit_state & (EXIT_ZOMBIE |
148                                             EXIT_DEAD));
149         const char **p = &task_state_array[0];
150
151         while (state) {
152                 p++;
153                 state >>= 1;
154         }
155         return *p;
156 }
157
158 static inline char * task_state(struct task_struct *p, char *buffer)
159 {
160         struct group_info *group_info;
161         int g;
162         struct fdtable *fdt = NULL;
163
164         read_lock(&tasklist_lock);
165         buffer += sprintf(buffer,
166                 "State:\t%s\n"
167                 "SleepAVG:\t%lu%%\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->sleep_avg/1024)*100/(1020000000/1024),
176                 p->tgid,
177                 p->pid, pid_alive(p) ? p->group_leader->real_parent->tgid : 0,
178                 pid_alive(p) && p->ptrace ? p->parent->pid : 0,
179                 p->uid, p->euid, p->suid, p->fsuid,
180                 p->gid, p->egid, p->sgid, p->fsgid);
181         read_unlock(&tasklist_lock);
182         task_lock(p);
183         if (p->files)
184                 fdt = files_fdtable(p->files);
185         buffer += sprintf(buffer,
186                 "FDSize:\t%d\n"
187                 "Groups:\t",
188                 fdt ? fdt->max_fds : 0);
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
202 static 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
227 static 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
242 static inline char * task_sig(struct task_struct *p, char *buffer)
243 {
244         sigset_t pending, shpending, blocked, ignored, caught;
245         int num_threads = 0;
246         unsigned long qsize = 0;
247         unsigned long qlim = 0;
248
249         sigemptyset(&pending);
250         sigemptyset(&shpending);
251         sigemptyset(&blocked);
252         sigemptyset(&ignored);
253         sigemptyset(&caught);
254
255         /* Gather all the data with the appropriate locks held */
256         read_lock(&tasklist_lock);
257         if (p->sighand) {
258                 spin_lock_irq(&p->sighand->siglock);
259                 pending = p->pending.signal;
260                 shpending = p->signal->shared_pending.signal;
261                 blocked = p->blocked;
262                 collect_sigign_sigcatch(p, &ignored, &caught);
263                 num_threads = atomic_read(&p->signal->count);
264                 qsize = atomic_read(&p->user->sigpending);
265                 qlim = p->signal->rlim[RLIMIT_SIGPENDING].rlim_cur;
266                 spin_unlock_irq(&p->sighand->siglock);
267         }
268         read_unlock(&tasklist_lock);
269
270         buffer += sprintf(buffer, "Threads:\t%d\n", num_threads);
271         buffer += sprintf(buffer, "SigQ:\t%lu/%lu\n", qsize, qlim);
272
273         /* render them all */
274         buffer = render_sigset_t("SigPnd:\t", &pending, buffer);
275         buffer = render_sigset_t("ShdPnd:\t", &shpending, buffer);
276         buffer = render_sigset_t("SigBlk:\t", &blocked, buffer);
277         buffer = render_sigset_t("SigIgn:\t", &ignored, buffer);
278         buffer = render_sigset_t("SigCgt:\t", &caught, buffer);
279
280         return buffer;
281 }
282
283 static inline char *task_cap(struct task_struct *p, char *buffer)
284 {
285     return buffer + sprintf(buffer, "CapInh:\t%016x\n"
286                             "CapPrm:\t%016x\n"
287                             "CapEff:\t%016x\n",
288                             cap_t(p->cap_inheritable),
289                             cap_t(p->cap_permitted),
290                             cap_t(p->cap_effective));
291 }
292
293 int proc_pid_status(struct task_struct *task, char * buffer)
294 {
295         char * orig = buffer;
296         struct mm_struct *mm = get_task_mm(task);
297
298         buffer = task_name(task, buffer);
299         buffer = task_state(task, buffer);
300  
301         if (mm) {
302                 buffer = task_mem(mm, buffer);
303                 mmput(mm);
304         }
305         buffer = task_sig(task, buffer);
306         buffer = task_cap(task, buffer);
307         buffer = cpuset_task_status_allowed(task, buffer);
308 #if defined(CONFIG_ARCH_S390)
309         buffer = task_show_regs(task, buffer);
310 #endif
311         return buffer - orig;
312 }
313
314 static int do_task_stat(struct task_struct *task, char * buffer, int whole)
315 {
316         unsigned long vsize, eip, esp, wchan = ~0UL;
317         long priority, nice;
318         int tty_pgrp = -1, tty_nr = 0;
319         sigset_t sigign, sigcatch;
320         char state;
321         int res;
322         pid_t ppid, pgid = -1, sid = -1;
323         int num_threads = 0;
324         struct mm_struct *mm;
325         unsigned long long start_time;
326         unsigned long cmin_flt = 0, cmaj_flt = 0;
327         unsigned long  min_flt = 0,  maj_flt = 0;
328         cputime_t cutime, cstime, utime, stime;
329         unsigned long rsslim = 0;
330         unsigned long it_real_value = 0;
331         struct task_struct *t;
332         char tcomm[sizeof(task->comm)];
333
334         state = *get_task_state(task);
335         vsize = eip = esp = 0;
336         mm = get_task_mm(task);
337         if (mm) {
338                 vsize = task_vsize(mm);
339                 eip = KSTK_EIP(task);
340                 esp = KSTK_ESP(task);
341         }
342
343         get_task_comm(tcomm, task);
344
345         sigemptyset(&sigign);
346         sigemptyset(&sigcatch);
347         cutime = cstime = utime = stime = cputime_zero;
348         read_lock(&tasklist_lock);
349         if (task->sighand) {
350                 spin_lock_irq(&task->sighand->siglock);
351                 num_threads = atomic_read(&task->signal->count);
352                 collect_sigign_sigcatch(task, &sigign, &sigcatch);
353
354                 /* add up live thread stats at the group level */
355                 if (whole) {
356                         t = task;
357                         do {
358                                 min_flt += t->min_flt;
359                                 maj_flt += t->maj_flt;
360                                 utime = cputime_add(utime, t->utime);
361                                 stime = cputime_add(stime, t->stime);
362                                 t = next_thread(t);
363                         } while (t != task);
364                 }
365
366                 spin_unlock_irq(&task->sighand->siglock);
367         }
368         if (task->signal) {
369                 if (task->signal->tty) {
370                         tty_pgrp = task->signal->tty->pgrp;
371                         tty_nr = new_encode_dev(tty_devnum(task->signal->tty));
372                 }
373                 pgid = process_group(task);
374                 sid = task->signal->session;
375                 cmin_flt = task->signal->cmin_flt;
376                 cmaj_flt = task->signal->cmaj_flt;
377                 cutime = task->signal->cutime;
378                 cstime = task->signal->cstime;
379                 rsslim = task->signal->rlim[RLIMIT_RSS].rlim_cur;
380                 if (whole) {
381                         min_flt += task->signal->min_flt;
382                         maj_flt += task->signal->maj_flt;
383                         utime = cputime_add(utime, task->signal->utime);
384                         stime = cputime_add(stime, task->signal->stime);
385                 }
386                 it_real_value = task->signal->it_real_value;
387         }
388         ppid = pid_alive(task) ? task->group_leader->real_parent->tgid : 0;
389         read_unlock(&tasklist_lock);
390
391         if (!whole || num_threads<2)
392                 wchan = get_wchan(task);
393         if (!whole) {
394                 min_flt = task->min_flt;
395                 maj_flt = task->maj_flt;
396                 utime = task->utime;
397                 stime = task->stime;
398         }
399
400         /* scale priority and nice values from timeslices to -20..20 */
401         /* to make it look like a "normal" Unix priority/nice value  */
402         priority = task_prio(task);
403         nice = task_nice(task);
404
405         /* Temporary variable needed for gcc-2.96 */
406         /* convert timespec -> nsec*/
407         start_time = (unsigned long long)task->start_time.tv_sec * NSEC_PER_SEC
408                                 + task->start_time.tv_nsec;
409         /* convert nsec -> ticks */
410         start_time = nsec_to_clock_t(start_time);
411
412         res = sprintf(buffer,"%d (%s) %c %d %d %d %d %d %lu %lu \
413 %lu %lu %lu %lu %lu %ld %ld %ld %ld %d %ld %llu %lu %ld %lu %lu %lu %lu %lu \
414 %lu %lu %lu %lu %lu %lu %lu %lu %d %d %lu %lu\n",
415                 task->pid,
416                 tcomm,
417                 state,
418                 ppid,
419                 pgid,
420                 sid,
421                 tty_nr,
422                 tty_pgrp,
423                 task->flags,
424                 min_flt,
425                 cmin_flt,
426                 maj_flt,
427                 cmaj_flt,
428                 cputime_to_clock_t(utime),
429                 cputime_to_clock_t(stime),
430                 cputime_to_clock_t(cutime),
431                 cputime_to_clock_t(cstime),
432                 priority,
433                 nice,
434                 num_threads,
435                 jiffies_to_clock_t(it_real_value),
436                 start_time,
437                 vsize,
438                 mm ? get_mm_counter(mm, rss) : 0, /* you might want to shift this left 3 */
439                 rsslim,
440                 mm ? mm->start_code : 0,
441                 mm ? mm->end_code : 0,
442                 mm ? mm->start_stack : 0,
443                 esp,
444                 eip,
445                 /* The signal information here is obsolete.
446                  * It must be decimal for Linux 2.0 compatibility.
447                  * Use /proc/#/status for real-time signals.
448                  */
449                 task->pending.signal.sig[0] & 0x7fffffffUL,
450                 task->blocked.sig[0] & 0x7fffffffUL,
451                 sigign      .sig[0] & 0x7fffffffUL,
452                 sigcatch    .sig[0] & 0x7fffffffUL,
453                 wchan,
454                 0UL,
455                 0UL,
456                 task->exit_signal,
457                 task_cpu(task),
458                 task->rt_priority,
459                 task->policy);
460         if(mm)
461                 mmput(mm);
462         return res;
463 }
464
465 int proc_tid_stat(struct task_struct *task, char * buffer)
466 {
467         return do_task_stat(task, buffer, 0);
468 }
469
470 int proc_tgid_stat(struct task_struct *task, char * buffer)
471 {
472         return do_task_stat(task, buffer, 1);
473 }
474
475 int proc_pid_statm(struct task_struct *task, char *buffer)
476 {
477         int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0;
478         struct mm_struct *mm = get_task_mm(task);
479         
480         if (mm) {
481                 size = task_statm(mm, &shared, &text, &data, &resident);
482                 mmput(mm);
483         }
484
485         return sprintf(buffer,"%d %d %d %d %d %d %d\n",
486                        size, resident, shared, text, lib, data, 0);
487 }