Linux-2.6.12-rc2
[linux-flexiantxendom0-natty.git] / 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
163         read_lock(&tasklist_lock);
164         buffer += sprintf(buffer,
165                 "State:\t%s\n"
166                 "SleepAVG:\t%lu%%\n"
167                 "Tgid:\t%d\n"
168                 "Pid:\t%d\n"
169                 "PPid:\t%d\n"
170                 "TracerPid:\t%d\n"
171                 "Uid:\t%d\t%d\t%d\t%d\n"
172                 "Gid:\t%d\t%d\t%d\t%d\n",
173                 get_task_state(p),
174                 (p->sleep_avg/1024)*100/(1020000000/1024),
175                 p->tgid,
176                 p->pid, pid_alive(p) ? p->group_leader->real_parent->tgid : 0,
177                 pid_alive(p) && p->ptrace ? p->parent->pid : 0,
178                 p->uid, p->euid, p->suid, p->fsuid,
179                 p->gid, p->egid, p->sgid, p->fsgid);
180         read_unlock(&tasklist_lock);
181         task_lock(p);
182         buffer += sprintf(buffer,
183                 "FDSize:\t%d\n"
184                 "Groups:\t",
185                 p->files ? p->files->max_fds : 0);
186
187         group_info = p->group_info;
188         get_group_info(group_info);
189         task_unlock(p);
190
191         for (g = 0; g < min(group_info->ngroups,NGROUPS_SMALL); g++)
192                 buffer += sprintf(buffer, "%d ", GROUP_AT(group_info,g));
193         put_group_info(group_info);
194
195         buffer += sprintf(buffer, "\n");
196         return buffer;
197 }
198
199 static char * render_sigset_t(const char *header, sigset_t *set, char *buffer)
200 {
201         int i, len;
202
203         len = strlen(header);
204         memcpy(buffer, header, len);
205         buffer += len;
206
207         i = _NSIG;
208         do {
209                 int x = 0;
210
211                 i -= 4;
212                 if (sigismember(set, i+1)) x |= 1;
213                 if (sigismember(set, i+2)) x |= 2;
214                 if (sigismember(set, i+3)) x |= 4;
215                 if (sigismember(set, i+4)) x |= 8;
216                 *buffer++ = (x < 10 ? '0' : 'a' - 10) + x;
217         } while (i >= 4);
218
219         *buffer++ = '\n';
220         *buffer = 0;
221         return buffer;
222 }
223
224 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
225                                     sigset_t *catch)
226 {
227         struct k_sigaction *k;
228         int i;
229
230         k = p->sighand->action;
231         for (i = 1; i <= _NSIG; ++i, ++k) {
232                 if (k->sa.sa_handler == SIG_IGN)
233                         sigaddset(ign, i);
234                 else if (k->sa.sa_handler != SIG_DFL)
235                         sigaddset(catch, i);
236         }
237 }
238
239 static inline char * task_sig(struct task_struct *p, char *buffer)
240 {
241         sigset_t pending, shpending, blocked, ignored, caught;
242         int num_threads = 0;
243         unsigned long qsize = 0;
244         unsigned long qlim = 0;
245
246         sigemptyset(&pending);
247         sigemptyset(&shpending);
248         sigemptyset(&blocked);
249         sigemptyset(&ignored);
250         sigemptyset(&caught);
251
252         /* Gather all the data with the appropriate locks held */
253         read_lock(&tasklist_lock);
254         if (p->sighand) {
255                 spin_lock_irq(&p->sighand->siglock);
256                 pending = p->pending.signal;
257                 shpending = p->signal->shared_pending.signal;
258                 blocked = p->blocked;
259                 collect_sigign_sigcatch(p, &ignored, &caught);
260                 num_threads = atomic_read(&p->signal->count);
261                 qsize = atomic_read(&p->user->sigpending);
262                 qlim = p->signal->rlim[RLIMIT_SIGPENDING].rlim_cur;
263                 spin_unlock_irq(&p->sighand->siglock);
264         }
265         read_unlock(&tasklist_lock);
266
267         buffer += sprintf(buffer, "Threads:\t%d\n", num_threads);
268         buffer += sprintf(buffer, "SigQ:\t%lu/%lu\n", qsize, qlim);
269
270         /* render them all */
271         buffer = render_sigset_t("SigPnd:\t", &pending, buffer);
272         buffer = render_sigset_t("ShdPnd:\t", &shpending, buffer);
273         buffer = render_sigset_t("SigBlk:\t", &blocked, buffer);
274         buffer = render_sigset_t("SigIgn:\t", &ignored, buffer);
275         buffer = render_sigset_t("SigCgt:\t", &caught, buffer);
276
277         return buffer;
278 }
279
280 static inline char *task_cap(struct task_struct *p, char *buffer)
281 {
282     return buffer + sprintf(buffer, "CapInh:\t%016x\n"
283                             "CapPrm:\t%016x\n"
284                             "CapEff:\t%016x\n",
285                             cap_t(p->cap_inheritable),
286                             cap_t(p->cap_permitted),
287                             cap_t(p->cap_effective));
288 }
289
290 int proc_pid_status(struct task_struct *task, char * buffer)
291 {
292         char * orig = buffer;
293         struct mm_struct *mm = get_task_mm(task);
294
295         buffer = task_name(task, buffer);
296         buffer = task_state(task, buffer);
297  
298         if (mm) {
299                 buffer = task_mem(mm, buffer);
300                 mmput(mm);
301         }
302         buffer = task_sig(task, buffer);
303         buffer = task_cap(task, buffer);
304         buffer = cpuset_task_status_allowed(task, buffer);
305 #if defined(CONFIG_ARCH_S390)
306         buffer = task_show_regs(task, buffer);
307 #endif
308         return buffer - orig;
309 }
310
311 static int do_task_stat(struct task_struct *task, char * buffer, int whole)
312 {
313         unsigned long vsize, eip, esp, wchan = ~0UL;
314         long priority, nice;
315         int tty_pgrp = -1, tty_nr = 0;
316         sigset_t sigign, sigcatch;
317         char state;
318         int res;
319         pid_t ppid, pgid = -1, sid = -1;
320         int num_threads = 0;
321         struct mm_struct *mm;
322         unsigned long long start_time;
323         unsigned long cmin_flt = 0, cmaj_flt = 0;
324         unsigned long  min_flt = 0,  maj_flt = 0;
325         cputime_t cutime, cstime, utime, stime;
326         unsigned long rsslim = 0;
327         unsigned long it_real_value = 0;
328         struct task_struct *t;
329         char tcomm[sizeof(task->comm)];
330
331         state = *get_task_state(task);
332         vsize = eip = esp = 0;
333         mm = get_task_mm(task);
334         if (mm) {
335                 vsize = task_vsize(mm);
336                 eip = KSTK_EIP(task);
337                 esp = KSTK_ESP(task);
338         }
339
340         get_task_comm(tcomm, task);
341
342         sigemptyset(&sigign);
343         sigemptyset(&sigcatch);
344         cutime = cstime = utime = stime = cputime_zero;
345         read_lock(&tasklist_lock);
346         if (task->sighand) {
347                 spin_lock_irq(&task->sighand->siglock);
348                 num_threads = atomic_read(&task->signal->count);
349                 collect_sigign_sigcatch(task, &sigign, &sigcatch);
350
351                 /* add up live thread stats at the group level */
352                 if (whole) {
353                         t = task;
354                         do {
355                                 min_flt += t->min_flt;
356                                 maj_flt += t->maj_flt;
357                                 utime = cputime_add(utime, t->utime);
358                                 stime = cputime_add(stime, t->stime);
359                                 t = next_thread(t);
360                         } while (t != task);
361                 }
362
363                 spin_unlock_irq(&task->sighand->siglock);
364         }
365         if (task->signal) {
366                 if (task->signal->tty) {
367                         tty_pgrp = task->signal->tty->pgrp;
368                         tty_nr = new_encode_dev(tty_devnum(task->signal->tty));
369                 }
370                 pgid = process_group(task);
371                 sid = task->signal->session;
372                 cmin_flt = task->signal->cmin_flt;
373                 cmaj_flt = task->signal->cmaj_flt;
374                 cutime = task->signal->cutime;
375                 cstime = task->signal->cstime;
376                 rsslim = task->signal->rlim[RLIMIT_RSS].rlim_cur;
377                 if (whole) {
378                         min_flt += task->signal->min_flt;
379                         maj_flt += task->signal->maj_flt;
380                         utime = cputime_add(utime, task->signal->utime);
381                         stime = cputime_add(stime, task->signal->stime);
382                 }
383                 it_real_value = task->signal->it_real_value;
384         }
385         ppid = pid_alive(task) ? task->group_leader->real_parent->tgid : 0;
386         read_unlock(&tasklist_lock);
387
388         if (!whole || num_threads<2)
389                 wchan = get_wchan(task);
390         if (!whole) {
391                 min_flt = task->min_flt;
392                 maj_flt = task->maj_flt;
393                 utime = task->utime;
394                 stime = task->stime;
395         }
396
397         /* scale priority and nice values from timeslices to -20..20 */
398         /* to make it look like a "normal" Unix priority/nice value  */
399         priority = task_prio(task);
400         nice = task_nice(task);
401
402         /* Temporary variable needed for gcc-2.96 */
403         /* convert timespec -> nsec*/
404         start_time = (unsigned long long)task->start_time.tv_sec * NSEC_PER_SEC
405                                 + task->start_time.tv_nsec;
406         /* convert nsec -> ticks */
407         start_time = nsec_to_clock_t(start_time);
408
409         res = sprintf(buffer,"%d (%s) %c %d %d %d %d %d %lu %lu \
410 %lu %lu %lu %lu %lu %ld %ld %ld %ld %d %ld %llu %lu %ld %lu %lu %lu %lu %lu \
411 %lu %lu %lu %lu %lu %lu %lu %lu %d %d %lu %lu\n",
412                 task->pid,
413                 tcomm,
414                 state,
415                 ppid,
416                 pgid,
417                 sid,
418                 tty_nr,
419                 tty_pgrp,
420                 task->flags,
421                 min_flt,
422                 cmin_flt,
423                 maj_flt,
424                 cmaj_flt,
425                 cputime_to_clock_t(utime),
426                 cputime_to_clock_t(stime),
427                 cputime_to_clock_t(cutime),
428                 cputime_to_clock_t(cstime),
429                 priority,
430                 nice,
431                 num_threads,
432                 jiffies_to_clock_t(it_real_value),
433                 start_time,
434                 vsize,
435                 mm ? get_mm_counter(mm, rss) : 0, /* you might want to shift this left 3 */
436                 rsslim,
437                 mm ? mm->start_code : 0,
438                 mm ? mm->end_code : 0,
439                 mm ? mm->start_stack : 0,
440                 esp,
441                 eip,
442                 /* The signal information here is obsolete.
443                  * It must be decimal for Linux 2.0 compatibility.
444                  * Use /proc/#/status for real-time signals.
445                  */
446                 task->pending.signal.sig[0] & 0x7fffffffUL,
447                 task->blocked.sig[0] & 0x7fffffffUL,
448                 sigign      .sig[0] & 0x7fffffffUL,
449                 sigcatch    .sig[0] & 0x7fffffffUL,
450                 wchan,
451                 0UL,
452                 0UL,
453                 task->exit_signal,
454                 task_cpu(task),
455                 task->rt_priority,
456                 task->policy);
457         if(mm)
458                 mmput(mm);
459         return res;
460 }
461
462 int proc_tid_stat(struct task_struct *task, char * buffer)
463 {
464         return do_task_stat(task, buffer, 0);
465 }
466
467 int proc_tgid_stat(struct task_struct *task, char * buffer)
468 {
469         return do_task_stat(task, buffer, 1);
470 }
471
472 int proc_pid_statm(struct task_struct *task, char *buffer)
473 {
474         int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0;
475         struct mm_struct *mm = get_task_mm(task);
476         
477         if (mm) {
478                 size = task_statm(mm, &shared, &text, &data, &resident);
479                 mmput(mm);
480         }
481
482         return sprintf(buffer,"%d %d %d %d %d %d %d\n",
483                        size, resident, shared, text, lib, data, 0);
484 }