Merge to VServer 1.9.0
[linux-2.6.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/ninline.h>
77
78 #include <asm/uaccess.h>
79 #include <asm/pgtable.h>
80 #include <asm/io.h>
81 #include <asm/processor.h>
82
83 /* Gcc optimizes away "strlen(x)" for constant x */
84 #define ADDBUF(buffer, string) \
85 do { memcpy(buffer, string, strlen(string)); \
86      buffer += strlen(string); } while (0)
87
88 static inline char * task_name(struct task_struct *p, char * buf)
89 {
90         int i;
91         char * name;
92
93         ADDBUF(buf, "Name:\t");
94         name = p->comm;
95         i = sizeof(p->comm);
96         do {
97                 unsigned char c = *name;
98                 name++;
99                 i--;
100                 *buf = c;
101                 if (!c)
102                         break;
103                 if (c == '\\') {
104                         buf[1] = c;
105                         buf += 2;
106                         continue;
107                 }
108                 if (c == '\n') {
109                         buf[0] = '\\';
110                         buf[1] = 'n';
111                         buf += 2;
112                         continue;
113                 }
114                 buf++;
115         } while (i);
116         *buf = '\n';
117         return buf+1;
118 }
119
120 /*
121  * The task state array is a strange "bitmap" of
122  * reasons to sleep. Thus "running" is zero, and
123  * you can test for combinations of others with
124  * simple bit tests.
125  */
126 static const char *task_state_array[] = {
127         "R (running)",          /*  0 */
128         "S (sleeping)",         /*  1 */
129         "D (disk sleep)",       /*  2 */
130         "T (stopped)",          /*  4 */
131         "Z (zombie)",           /*  8 */
132         "X (dead)",             /* 16 */
133         "H (on hold)"           /* 32 */
134 };
135
136 static inline const char * get_task_state(struct task_struct *tsk)
137 {
138         unsigned int state = tsk->state & (TASK_RUNNING |
139                                            TASK_INTERRUPTIBLE |
140                                            TASK_UNINTERRUPTIBLE |
141                                            TASK_ZOMBIE |
142                                            TASK_STOPPED |
143                                            TASK_ONHOLD);
144         const char **p = &task_state_array[0];
145
146         while (state) {
147                 p++;
148                 state >>= 1;
149         }
150         return *p;
151 }
152
153 static inline char * task_state(struct task_struct *p, char *buffer)
154 {
155         struct group_info *group_info;
156         int g;
157         pid_t ppid;
158
159         read_lock(&tasklist_lock);
160         ppid = vx_map_tgid(current->vx_info, p->real_parent->pid);
161         buffer += sprintf(buffer,
162                 "State:\t%s\n"
163                 "SleepAVG:\t%lu%%\n"
164                 "Tgid:\t%d\n"
165                 "Pid:\t%d\n"
166                 "PPid:\t%d\n"
167                 "TracerPid:\t%d\n"
168                 "Uid:\t%d\t%d\t%d\t%d\n"
169                 "Gid:\t%d\t%d\t%d\t%d\n",
170                 get_task_state(p),
171                 (p->sleep_avg/1024)*100/(1020000000/1024),
172                 p->tgid,
173                 p->pid, p->pid ? ppid : 0,
174                 p->pid && p->ptrace ? p->parent->pid : 0,
175                 p->uid, p->euid, p->suid, p->fsuid,
176                 p->gid, p->egid, p->sgid, p->fsgid);
177         read_unlock(&tasklist_lock);
178         task_lock(p);
179         buffer += sprintf(buffer,
180                 "FDSize:\t%d\n"
181                 "Groups:\t",
182                 p->files ? p->files->max_fds : 0);
183
184         group_info = p->group_info;
185         get_group_info(group_info);
186         task_unlock(p);
187
188         for (g = 0; g < min(group_info->ngroups,NGROUPS_SMALL); g++)
189                 buffer += sprintf(buffer, "%d ", GROUP_AT(group_info,g));
190         put_group_info(group_info);
191
192         buffer += sprintf(buffer, "\n");
193         return buffer;
194 }
195
196 static char * render_sigset_t(const char *header, sigset_t *set, char *buffer)
197 {
198         int i, len;
199
200         len = strlen(header);
201         memcpy(buffer, header, len);
202         buffer += len;
203
204         i = _NSIG;
205         do {
206                 int x = 0;
207
208                 i -= 4;
209                 if (sigismember(set, i+1)) x |= 1;
210                 if (sigismember(set, i+2)) x |= 2;
211                 if (sigismember(set, i+3)) x |= 4;
212                 if (sigismember(set, i+4)) x |= 8;
213                 *buffer++ = (x < 10 ? '0' : 'a' - 10) + x;
214         } while (i >= 4);
215
216         *buffer++ = '\n';
217         *buffer = 0;
218         return buffer;
219 }
220
221 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
222                                     sigset_t *catch)
223 {
224         struct k_sigaction *k;
225         int i;
226
227         k = p->sighand->action;
228         for (i = 1; i <= _NSIG; ++i, ++k) {
229                 if (k->sa.sa_handler == SIG_IGN)
230                         sigaddset(ign, i);
231                 else if (k->sa.sa_handler != SIG_DFL)
232                         sigaddset(catch, i);
233         }
234 }
235
236 static inline char * task_sig(struct task_struct *p, char *buffer)
237 {
238         sigset_t pending, shpending, blocked, ignored, caught;
239         int num_threads = 0;
240
241         sigemptyset(&pending);
242         sigemptyset(&shpending);
243         sigemptyset(&blocked);
244         sigemptyset(&ignored);
245         sigemptyset(&caught);
246
247         /* Gather all the data with the appropriate locks held */
248         read_lock(&tasklist_lock);
249         if (p->sighand) {
250                 spin_lock_irq(&p->sighand->siglock);
251                 pending = p->pending.signal;
252                 shpending = p->signal->shared_pending.signal;
253                 blocked = p->blocked;
254                 collect_sigign_sigcatch(p, &ignored, &caught);
255                 num_threads = atomic_read(&p->signal->count);
256                 spin_unlock_irq(&p->sighand->siglock);
257         }
258         read_unlock(&tasklist_lock);
259
260         buffer += sprintf(buffer, "Threads:\t%d\n", num_threads);
261
262         /* render them all */
263         buffer = render_sigset_t("SigPnd:\t", &pending, buffer);
264         buffer = render_sigset_t("ShdPnd:\t", &shpending, buffer);
265         buffer = render_sigset_t("SigBlk:\t", &blocked, buffer);
266         buffer = render_sigset_t("SigIgn:\t", &ignored, buffer);
267         buffer = render_sigset_t("SigCgt:\t", &caught, buffer);
268
269         return buffer;
270 }
271
272 static inline char *task_cap(struct task_struct *p, char *buffer)
273 {
274     return buffer + sprintf(buffer, "CapInh:\t%016x\n"
275                             "CapPrm:\t%016x\n"
276                             "CapEff:\t%016x\n",
277                             cap_t(p->cap_inheritable),
278                             cap_t(p->cap_permitted),
279                             cap_t(p->cap_effective));
280 }
281
282 extern char *task_mem(struct mm_struct *, char *);
283 int proc_pid_status(struct task_struct *task, char * buffer)
284 {
285         char * orig = buffer;
286 #ifdef  CONFIG_VSERVER_LEGACY           
287         struct vx_info *vxi;
288         struct nx_info *nxi;
289 #endif
290         struct mm_struct *mm = get_task_mm(task);
291
292         buffer = task_name(task, buffer);
293         buffer = task_state(task, buffer);
294  
295         if (mm) {
296                 buffer = task_mem(mm, buffer);
297                 mmput(mm);
298         }
299         buffer = task_sig(task, buffer);
300         buffer = task_cap(task, buffer);
301
302 #ifdef  CONFIG_VSERVER_LEGACY           
303         buffer += sprintf (buffer,"s_context: %d\n", vx_task_xid(task));
304         vxi = task_get_vx_info(task);
305         if (vxi) {
306                 buffer += sprintf (buffer,"ctxflags: %08llx\n"
307                         ,vxi->vx_flags);
308                 buffer += sprintf (buffer,"initpid: %d\n"
309                         ,vxi->vx_initpid);
310         } else {
311                 buffer += sprintf (buffer,"ctxflags: none\n");
312                 buffer += sprintf (buffer,"initpid: none\n");
313         }
314         put_vx_info(vxi);
315         nxi = task_get_nx_info(task);
316         if (nxi) {
317                 int i;
318
319                 buffer += sprintf (buffer,"ipv4root:");
320                 for (i=0; i<nxi->nbipv4; i++){
321                         buffer += sprintf (buffer," %08x/%08x"
322                                 ,nxi->ipv4[i]
323                                 ,nxi->mask[i]);
324                 }
325                 *buffer++ = '\n';
326                 buffer += sprintf (buffer,"ipv4root_bcast: %08x\n"
327                         ,nxi->v4_bcast);
328                 buffer += sprintf (buffer,"ipv4root_refcnt: %d\n"
329                         ,atomic_read(&nxi->nx_refcount));
330         } else {
331                 buffer += sprintf (buffer,"ipv4root: 0\n");
332                 buffer += sprintf (buffer,"ipv4root_bcast: 0\n");
333         }
334         put_nx_info(nxi);
335 #endif
336 #if defined(CONFIG_ARCH_S390)
337         buffer = task_show_regs(task, buffer);
338 #endif
339         return buffer - orig;
340 }
341
342 extern unsigned long task_vsize(struct mm_struct *);
343 int proc_pid_stat(struct task_struct *task, char * buffer)
344 {
345         unsigned long vsize, eip, esp, wchan;
346         long priority, nice;
347         unsigned long long bias_jiffies;
348         int tty_pgrp = -1, tty_nr = 0;
349         sigset_t sigign, sigcatch;
350         char state;
351         int res;
352         pid_t ppid, pgid = -1, sid = -1;
353         int num_threads = 0;
354         struct mm_struct *mm;
355         unsigned long long start_time;
356
357         state = *get_task_state(task);
358         vsize = eip = esp = 0;
359         bias_jiffies = INITIAL_JIFFIES;
360
361         task_lock(task);
362         if (__vx_task_flags(task, VXF_VIRT_UPTIME, 0)) {
363                 bias_jiffies = task->vx_info->cvirt.bias_jiffies;
364                 /* hmm, do we need that? */
365                 if (bias_jiffies > task->start_time)
366                         bias_jiffies = task->start_time;
367         }
368
369         mm = task->mm;
370         if(mm)
371                 mm = mmgrab(mm);
372         task_unlock(task);
373         if (mm) {
374                 down_read(&mm->mmap_sem);
375                 vsize = task_vsize(mm);
376                 eip = KSTK_EIP(task);
377                 esp = KSTK_ESP(task);
378                 up_read(&mm->mmap_sem);
379         }
380
381         wchan = 0;
382         if (current->uid == task->uid || current->euid == task->uid ||
383                                                         capable(CAP_SYS_NICE))
384                 wchan = get_wchan(task);
385
386         sigemptyset(&sigign);
387         sigemptyset(&sigcatch);
388         read_lock(&tasklist_lock);
389         if (task->sighand) {
390                 spin_lock_irq(&task->sighand->siglock);
391                 num_threads = atomic_read(&task->signal->count);
392                 collect_sigign_sigcatch(task, &sigign, &sigcatch);
393                 spin_unlock_irq(&task->sighand->siglock);
394         }
395         if (task->signal) {
396                 if (task->signal->tty) {
397                         tty_pgrp = task->signal->tty->pgrp;
398                         tty_nr = new_encode_dev(tty_devnum(task->signal->tty));
399                 }
400                 pgid = process_group(task);
401                 sid = task->signal->session;
402         }
403         read_unlock(&tasklist_lock);
404
405         /* scale priority and nice values from timeslices to -20..20 */
406         /* to make it look like a "normal" Unix priority/nice value  */
407         priority = task_prio(task);
408         nice = task_nice(task);
409
410         read_lock(&tasklist_lock);
411         ppid = task->pid ? task->real_parent->pid : 0;
412         read_unlock(&tasklist_lock);
413
414         /* Temporary variable needed for gcc-2.96 */
415         start_time = jiffies_64_to_clock_t(task->start_time - bias_jiffies);
416
417         res = sprintf(buffer,"%d (%s) %c %d %d %d %d %d %lu %lu \
418 %lu %lu %lu %lu %lu %ld %ld %ld %ld %d %ld %llu %lu %ld %lu %lu %lu %lu %lu \
419 %lu %lu %lu %lu %lu %lu %lu %lu %d %d %lu %lu\n",
420                 task->pid,
421                 task->comm,
422                 state,
423                 ppid,
424                 pgid,
425                 sid,
426                 tty_nr,
427                 tty_pgrp,
428                 task->flags,
429                 task->min_flt,
430                 task->cmin_flt,
431                 task->maj_flt,
432                 task->cmaj_flt,
433                 jiffies_to_clock_t(task->utime),
434                 jiffies_to_clock_t(task->stime),
435                 jiffies_to_clock_t(task->cutime),
436                 jiffies_to_clock_t(task->cstime),
437                 priority,
438                 nice,
439                 num_threads,
440                 jiffies_to_clock_t(task->it_real_value),
441                 start_time,
442                 vsize,
443                 mm ? mm->rss : 0, /* you might want to shift this left 3 */
444                 task->rlim[RLIMIT_RSS].rlim_cur,
445                 mm ? mm->start_code : 0,
446                 mm ? mm->end_code : 0,
447                 mm ? mm->start_stack : 0,
448                 esp,
449                 eip,
450                 /* The signal information here is obsolete.
451                  * It must be decimal for Linux 2.0 compatibility.
452                  * Use /proc/#/status for real-time signals.
453                  */
454                 task->pending.signal.sig[0] & 0x7fffffffUL,
455                 task->blocked.sig[0] & 0x7fffffffUL,
456                 sigign      .sig[0] & 0x7fffffffUL,
457                 sigcatch    .sig[0] & 0x7fffffffUL,
458                 wchan,
459                 0UL,
460                 0UL,
461                 task->exit_signal,
462                 task_cpu(task),
463                 task->rt_priority,
464                 task->policy);
465         if(mm)
466                 mmput(mm);
467         return res;
468 }
469
470 extern int task_statm(struct mm_struct *, int *, int *, int *, int *);
471 int proc_pid_statm(struct task_struct *task, char *buffer)
472 {
473         int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0;
474         struct mm_struct *mm = get_task_mm(task);
475         
476         if (mm) {
477                 down_read(&mm->mmap_sem);
478                 size = task_statm(mm, &shared, &text, &data, &resident);
479                 up_read(&mm->mmap_sem);
480
481                 mmput(mm);
482         }
483
484         return sprintf(buffer,"%d %d %d %d %d %d %d\n",
485                        size, resident, shared, text, lib, data, 0);
486 }