Merge to Fedora Core 2 kernel-2.6.8-1.521
[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/vs_base.h>
77 #include <linux/vs_context.h>
78 #include <linux/vs_network.h>
79 #include <linux/vs_cvirt.h>
80
81 #include <asm/uaccess.h>
82 #include <asm/pgtable.h>
83 #include <asm/io.h>
84 #include <asm/processor.h>
85
86 /* Gcc optimizes away "strlen(x)" for constant x */
87 #define ADDBUF(buffer, string) \
88 do { memcpy(buffer, string, strlen(string)); \
89      buffer += strlen(string); } while (0)
90
91 static inline char * task_name(struct task_struct *p, char * buf)
92 {
93         int i;
94         char * name;
95
96         ADDBUF(buf, "Name:\t");
97         name = p->comm;
98         i = sizeof(p->comm);
99         do {
100                 unsigned char c = *name;
101                 name++;
102                 i--;
103                 *buf = c;
104                 if (!c)
105                         break;
106                 if (c == '\\') {
107                         buf[1] = c;
108                         buf += 2;
109                         continue;
110                 }
111                 if (c == '\n') {
112                         buf[0] = '\\';
113                         buf[1] = 'n';
114                         buf += 2;
115                         continue;
116                 }
117                 buf++;
118         } while (i);
119         *buf = '\n';
120         return buf+1;
121 }
122
123 /*
124  * The task state array is a strange "bitmap" of
125  * reasons to sleep. Thus "running" is zero, and
126  * you can test for combinations of others with
127  * simple bit tests.
128  */
129 static const char *task_state_array[] = {
130         "R (running)",          /*  0 */
131         "S (sleeping)",         /*  1 */
132         "D (disk sleep)",       /*  2 */
133         "T (stopped)",          /*  4 */
134         "Z (zombie)",           /*  8 */
135         "X (dead)",             /* 16 */
136         "H (on hold)"           /* 32 */
137 };
138
139 static inline const char * get_task_state(struct task_struct *tsk)
140 {
141         unsigned int state = tsk->state & (TASK_RUNNING |
142                                            TASK_INTERRUPTIBLE |
143                                            TASK_UNINTERRUPTIBLE |
144                                            TASK_ZOMBIE |
145                                            TASK_DEAD |
146                                            TASK_STOPPED |
147                                            TASK_ONHOLD);
148         const char **p = &task_state_array[0];
149
150         while (state) {
151                 p++;
152                 state >>= 1;
153         }
154         return *p;
155 }
156
157 static inline char * task_state(struct task_struct *p, char *buffer)
158 {
159         struct group_info *group_info;
160         int g;
161         pid_t ppid;
162
163         read_lock(&tasklist_lock);
164         ppid = vx_map_tgid(current->vx_info, p->real_parent->pid);
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, p->pid ? ppid : 0,
178                 p->pid && 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         buffer += sprintf(buffer,
184                 "FDSize:\t%d\n"
185                 "Groups:\t",
186                 p->files ? p->files->max_fds : 0);
187
188         group_info = p->group_info;
189         get_group_info(group_info);
190         task_unlock(p);
191
192         for (g = 0; g < min(group_info->ngroups,NGROUPS_SMALL); g++)
193                 buffer += sprintf(buffer, "%d ", GROUP_AT(group_info,g));
194         put_group_info(group_info);
195
196         buffer += sprintf(buffer, "\n");
197         return buffer;
198 }
199
200 static char * render_sigset_t(const char *header, sigset_t *set, char *buffer)
201 {
202         int i, len;
203
204         len = strlen(header);
205         memcpy(buffer, header, len);
206         buffer += len;
207
208         i = _NSIG;
209         do {
210                 int x = 0;
211
212                 i -= 4;
213                 if (sigismember(set, i+1)) x |= 1;
214                 if (sigismember(set, i+2)) x |= 2;
215                 if (sigismember(set, i+3)) x |= 4;
216                 if (sigismember(set, i+4)) x |= 8;
217                 *buffer++ = (x < 10 ? '0' : 'a' - 10) + x;
218         } while (i >= 4);
219
220         *buffer++ = '\n';
221         *buffer = 0;
222         return buffer;
223 }
224
225 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
226                                     sigset_t *catch)
227 {
228         struct k_sigaction *k;
229         int i;
230
231         k = p->sighand->action;
232         for (i = 1; i <= _NSIG; ++i, ++k) {
233                 if (k->sa.sa_handler == SIG_IGN)
234                         sigaddset(ign, i);
235                 else if (k->sa.sa_handler != SIG_DFL)
236                         sigaddset(catch, i);
237         }
238 }
239
240 static inline char * task_sig(struct task_struct *p, char *buffer)
241 {
242         sigset_t pending, shpending, blocked, ignored, caught;
243         int num_threads = 0;
244
245         sigemptyset(&pending);
246         sigemptyset(&shpending);
247         sigemptyset(&blocked);
248         sigemptyset(&ignored);
249         sigemptyset(&caught);
250
251         /* Gather all the data with the appropriate locks held */
252         read_lock(&tasklist_lock);
253         if (p->sighand) {
254                 spin_lock_irq(&p->sighand->siglock);
255                 pending = p->pending.signal;
256                 shpending = p->signal->shared_pending.signal;
257                 blocked = p->blocked;
258                 collect_sigign_sigcatch(p, &ignored, &caught);
259                 num_threads = atomic_read(&p->signal->count);
260                 spin_unlock_irq(&p->sighand->siglock);
261         }
262         read_unlock(&tasklist_lock);
263
264         buffer += sprintf(buffer, "Threads:\t%d\n", num_threads);
265
266         /* render them all */
267         buffer = render_sigset_t("SigPnd:\t", &pending, buffer);
268         buffer = render_sigset_t("ShdPnd:\t", &shpending, buffer);
269         buffer = render_sigset_t("SigBlk:\t", &blocked, buffer);
270         buffer = render_sigset_t("SigIgn:\t", &ignored, buffer);
271         buffer = render_sigset_t("SigCgt:\t", &caught, buffer);
272
273         return buffer;
274 }
275
276 static inline char *task_cap(struct task_struct *p, char *buffer)
277 {
278     return buffer + sprintf(buffer, "CapInh:\t%016x\n"
279                             "CapPrm:\t%016x\n"
280                             "CapEff:\t%016x\n",
281                             cap_t(p->cap_inheritable),
282                             cap_t(p->cap_permitted),
283                             cap_t(p->cap_effective));
284 }
285
286 extern char *task_mem(struct mm_struct *, char *);
287 int proc_pid_status(struct task_struct *task, char * buffer)
288 {
289         char * orig = buffer;
290 #ifdef  CONFIG_VSERVER_LEGACY           
291         struct vx_info *vxi;
292         struct nx_info *nxi;
293 #endif
294         struct mm_struct *mm = get_task_mm(task);
295
296         buffer = task_name(task, buffer);
297         buffer = task_state(task, buffer);
298  
299         if (mm) {
300                 buffer = task_mem(mm, buffer);
301                 mmput(mm);
302         }
303         buffer = task_sig(task, buffer);
304         buffer = task_cap(task, buffer);
305
306 #ifdef  CONFIG_VSERVER_LEGACY           
307         buffer += sprintf (buffer,"s_context: %d\n", vx_task_xid(task));
308         vxi = task_get_vx_info(task);
309         if (vxi) {
310                 buffer += sprintf (buffer,"ctxflags: %08llx\n"
311                         ,(unsigned long long)vxi->vx_flags);
312                 buffer += sprintf (buffer,"initpid: %d\n"
313                         ,vxi->vx_initpid);
314         } else {
315                 buffer += sprintf (buffer,"ctxflags: none\n");
316                 buffer += sprintf (buffer,"initpid: none\n");
317         }
318         put_vx_info(vxi);
319         nxi = task_get_nx_info(task);
320         if (nxi) {
321                 int i;
322
323                 buffer += sprintf (buffer,"ipv4root:");
324                 for (i=0; i<nxi->nbipv4; i++){
325                         buffer += sprintf (buffer," %08x/%08x"
326                                 ,nxi->ipv4[i]
327                                 ,nxi->mask[i]);
328                 }
329                 *buffer++ = '\n';
330                 buffer += sprintf (buffer,"ipv4root_bcast: %08x\n"
331                         ,nxi->v4_bcast);
332         } else {
333                 buffer += sprintf (buffer,"ipv4root: 0\n");
334                 buffer += sprintf (buffer,"ipv4root_bcast: 0\n");
335         }
336         put_nx_info(nxi);
337 #endif
338 #if defined(CONFIG_ARCH_S390)
339         buffer = task_show_regs(task, buffer);
340 #endif
341         return buffer - orig;
342 }
343
344 extern unsigned long task_vsize(struct mm_struct *);
345 int proc_pid_stat(struct task_struct *task, char * buffer)
346 {
347         unsigned long vsize, eip, esp, wchan;
348         long priority, nice;
349         unsigned long long bias_jiffies;
350         int tty_pgrp = -1, tty_nr = 0;
351         sigset_t sigign, sigcatch;
352         char state;
353         int res;
354         pid_t ppid, pgid = -1, sid = -1;
355         int num_threads = 0;
356         struct mm_struct *mm;
357         unsigned long long start_time;
358
359         state = *get_task_state(task);
360         vsize = eip = esp = 0;
361         bias_jiffies = INITIAL_JIFFIES;
362
363         task_lock(task);
364         if (__vx_task_flags(task, VXF_VIRT_UPTIME, 0)) {
365                 bias_jiffies = task->vx_info->cvirt.bias_jiffies;
366                 /* hmm, do we need that? */
367                 if (bias_jiffies > task->start_time)
368                         bias_jiffies = task->start_time;
369         }
370
371         mm = task->mm;
372         if(mm)
373                 mm = mmgrab(mm);
374         task_unlock(task);
375         if (mm) {
376                 down_read(&mm->mmap_sem);
377                 vsize = task_vsize(mm);
378                 eip = KSTK_EIP(task);
379                 esp = KSTK_ESP(task);
380                 up_read(&mm->mmap_sem);
381         }
382
383         wchan = 0;
384         if (current->uid == task->uid || current->euid == task->uid ||
385                                                         capable(CAP_SYS_NICE))
386                 wchan = get_wchan(task);
387
388         sigemptyset(&sigign);
389         sigemptyset(&sigcatch);
390         read_lock(&tasklist_lock);
391         if (task->sighand) {
392                 spin_lock_irq(&task->sighand->siglock);
393                 num_threads = atomic_read(&task->signal->count);
394                 collect_sigign_sigcatch(task, &sigign, &sigcatch);
395                 spin_unlock_irq(&task->sighand->siglock);
396         }
397         if (task->signal) {
398                 if (task->signal->tty) {
399                         tty_pgrp = task->signal->tty->pgrp;
400                         tty_nr = new_encode_dev(tty_devnum(task->signal->tty));
401                 }
402                 pgid = process_group(task);
403                 sid = task->signal->session;
404         }
405         read_unlock(&tasklist_lock);
406
407         /* scale priority and nice values from timeslices to -20..20 */
408         /* to make it look like a "normal" Unix priority/nice value  */
409         priority = task_prio(task);
410         nice = task_nice(task);
411
412         read_lock(&tasklist_lock);
413         ppid = task->pid ? task->real_parent->pid : 0;
414         read_unlock(&tasklist_lock);
415
416         /* Temporary variable needed for gcc-2.96 */
417         start_time = jiffies_64_to_clock_t(task->start_time - bias_jiffies);
418
419         res = sprintf(buffer,"%d (%s) %c %d %d %d %d %d %lu %lu \
420 %lu %lu %lu %lu %lu %ld %ld %ld %ld %d %ld %llu %lu %ld %lu %lu %lu %lu %lu \
421 %lu %lu %lu %lu %lu %lu %lu %lu %d %d %lu %lu\n",
422                 task->pid,
423                 task->comm,
424                 state,
425                 ppid,
426                 pgid,
427                 sid,
428                 tty_nr,
429                 tty_pgrp,
430                 task->flags,
431                 task->min_flt,
432                 task->cmin_flt,
433                 task->maj_flt,
434                 task->cmaj_flt,
435                 jiffies_to_clock_t(task->utime),
436                 jiffies_to_clock_t(task->stime),
437                 jiffies_to_clock_t(task->cutime),
438                 jiffies_to_clock_t(task->cstime),
439                 priority,
440                 nice,
441                 num_threads,
442                 jiffies_to_clock_t(task->it_real_value),
443                 start_time,
444                 vsize,
445                 mm ? mm->rss : 0, /* you might want to shift this left 3 */
446                 task->rlim[RLIMIT_RSS].rlim_cur,
447                 mm ? mm->start_code : 0,
448                 mm ? mm->end_code : 0,
449                 mm ? mm->start_stack : 0,
450                 esp,
451                 eip,
452                 /* The signal information here is obsolete.
453                  * It must be decimal for Linux 2.0 compatibility.
454                  * Use /proc/#/status for real-time signals.
455                  */
456                 task->pending.signal.sig[0] & 0x7fffffffUL,
457                 task->blocked.sig[0] & 0x7fffffffUL,
458                 sigign      .sig[0] & 0x7fffffffUL,
459                 sigcatch    .sig[0] & 0x7fffffffUL,
460                 wchan,
461                 0UL,
462                 0UL,
463                 task->exit_signal,
464                 task_cpu(task),
465                 task->rt_priority,
466                 task->policy);
467         if(mm)
468                 mmput(mm);
469         return res;
470 }
471
472 extern int task_statm(struct mm_struct *, int *, int *, int *, int *);
473 int proc_pid_statm(struct task_struct *task, char *buffer)
474 {
475         int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0;
476         struct mm_struct *mm = get_task_mm(task);
477         
478         if (mm) {
479                 down_read(&mm->mmap_sem);
480                 size = task_statm(mm, &shared, &text, &data, &resident);
481                 up_read(&mm->mmap_sem);
482
483                 mmput(mm);
484         }
485
486         return sprintf(buffer,"%d %d %d %d %d %d %d\n",
487                        size, resident, shared, text, lib, data, 0);
488 }
489
490
491 int proc_pid_delay(struct task_struct *task, char * buffer)
492 {
493         int res;
494
495         res  = sprintf(buffer,"%u %llu %llu %u %llu %u %llu\n",
496                        get_delay(task,runs),
497                        (unsigned long long)get_delay(task,runcpu_total),
498                        (unsigned long long)get_delay(task,waitcpu_total),
499                        get_delay(task,num_iowaits),
500                        (unsigned long long)get_delay(task,iowait_total),
501                        get_delay(task,num_memwaits),
502                        (unsigned long long)get_delay(task,mem_iowait_total)
503                 );
504         return res;
505 }
506