projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
userns: Convert the audit loginuid to be a kuid
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / fs / proc / base.c
CommitLineData
1da177e4
LT		 1/*
2 * linux/fs/proc/base.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 *
6 * proc base directory handling functions
7 *
8 * 1999, Al Viro. Rewritten. Now it covers the whole per-process part.
9 * Instead of using magical inumbers to determine the kind of object
10 * we allocate and fill in-core inodes upon lookup. They don't even
11 * go into icache. We cache the reference to task_struct upon lookup too.
12 * Eventually it should become a filesystem in its own. We don't use the
13 * rest of procfs anymore.
e070ad49
ML		 14 *
15 *
16 * Changelog:
17 * 17-Jan-2005
18 * Allan Bezerra
19 * Bruna Moreira <bruna.moreira@indt.org.br>
20 * Edjard Mota <edjard.mota@indt.org.br>
21 * Ilias Biris <ilias.biris@indt.org.br>
22 * Mauricio Lin <mauricio.lin@indt.org.br>
23 *
24 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
25 *
26 * A new process specific entry (smaps) included in /proc. It shows the
27 * size of rss for each memory area. The maps entry lacks information
28 * about physical memory size (rss) for each mapped file, i.e.,
29 * rss information for executables and library files.
30 * This additional information is useful for any tools that need to know
31 * about physical memory consumption for a process specific library.
32 *
33 * Changelog:
34 * 21-Feb-2005
35 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
36 * Pud inclusion in the page table walking.
37 *
38 * ChangeLog:
39 * 10-Mar-2005
40 * 10LE Instituto Nokia de Tecnologia - INdT:
41 * A better way to walks through the page table as suggested by Hugh Dickins.
42 *
43 * Simo Piiroinen <simo.piiroinen@nokia.com>:
44 * Smaps information related to shared, private, clean and dirty pages.
45 *
46 * Paul Mundt <paul.mundt@nokia.com>:
47 * Overall revision about smaps.
1da177e4
LT		 48 */
49
50#include <asm/uaccess.h>
51
1da177e4
LT		 52#include <linux/errno.h>
53#include <linux/time.h>
54#include <linux/proc_fs.h>
55#include <linux/stat.h>
5995477a 56#include <linux/task_io_accounting_ops.h>
1da177e4 57#include <linux/init.h>
16f7e0fe 58#include <linux/capability.h>
1da177e4 59#include <linux/file.h>
9f3acc31 60#include <linux/fdtable.h>
1da177e4
LT		 61#include <linux/string.h>
62#include <linux/seq_file.h>
63#include <linux/namei.h>
6b3286ed 64#include <linux/mnt_namespace.h>
1da177e4 65#include <linux/mm.h>
a63d83f4 66#include <linux/swap.h>
b835996f 67#include <linux/rcupdate.h>
1da177e4 68#include <linux/kallsyms.h>
2ec220e2 69#include <linux/stacktrace.h>
d85f50d5 70#include <linux/resource.h>
5096add8 71#include <linux/module.h>
1da177e4
LT		 72#include <linux/mount.h>
73#include <linux/security.h>
74#include <linux/ptrace.h>
0d094efe 75#include <linux/tracehook.h>
a424316c 76#include <linux/cgroup.h>
1da177e4
LT		 77#include <linux/cpuset.h>
78#include <linux/audit.h>
5addc5dd 79#include <linux/poll.h>
1651e14e 80#include <linux/nsproxy.h>
8ac773b4 81#include <linux/oom.h>
3cb4a0bb 82#include <linux/elf.h>
60347f67 83#include <linux/pid_namespace.h>
22d917d8 84#include <linux/user_namespace.h>
5ad4e53b 85#include <linux/fs_struct.h>
5a0e3ad6 86#include <linux/slab.h>
640708a2 87#include <linux/flex_array.h>
f133ecca
CM		 88#ifdef CONFIG_HARDWALL
89#include <asm/hardwall.h>
90#endif
43d2b113 91#include <trace/events/oom.h>
1da177e4
LT		 92#include "internal.h"
93
0f2fe20f
EB		 94/* NOTE:
95 * Implementing inode permission operations in /proc is almost
96 * certainly an error. Permission checks need to happen during
97 * each system call not at open time. The reason is that most of
98 * what we wish to check for permissions in /proc varies at runtime.
99 *
100 * The classic example of a problem is opening file descriptors
101 * in /proc for a task before it execs a suid executable.
102 */
103
1da177e4 104struct pid_entry {
1da177e4 105 char *name;
c5141e6d 106 int len;
d161a13f 107 umode_t mode;
c5ef1c42 108 const struct inode_operations *iop;
00977a59 109 const struct file_operations *fop;
20cdc894 110 union proc_op op;
1da177e4
LT		 111};
112
61a28784 113#define NOD(NAME, MODE, IOP, FOP, OP) { \
20cdc894 114 .name = (NAME), \
c5141e6d 115 .len = sizeof(NAME) - 1, \
20cdc894
EB		 116 .mode = MODE, \
117 .iop = IOP, \
118 .fop = FOP, \
119 .op = OP, \
120}
121
631f9c18
AD		 122#define DIR(NAME, MODE, iops, fops) \
123 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
124#define LNK(NAME, get_link) \
61a28784 125 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
20cdc894 126 &proc_pid_link_inode_operations, NULL, \
631f9c18
AD		 127 { .proc_get_link = get_link } )
128#define REG(NAME, MODE, fops) \
129 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
130#define INF(NAME, MODE, read) \
61a28784 131 NOD(NAME, (S_IFREG|(MODE)), \
20cdc894 132 NULL, &proc_info_file_operations, \
631f9c18
AD		 133 { .proc_read = read } )
134#define ONE(NAME, MODE, show) \
be614086
EB		 135 NOD(NAME, (S_IFREG|(MODE)), \
136 NULL, &proc_single_file_operations, \
631f9c18 137 { .proc_show = show } )
1da177e4 138
640708a2
PE		 139static int proc_fd_permission(struct inode *inode, int mask);
140
aed54175
VN		 141/*
142 * Count the number of hardlinks for the pid_entry table, excluding the .
143 * and .. links.
144 */
145static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
146 unsigned int n)
147{
148 unsigned int i;
149 unsigned int count;
150
151 count = 0;
152 for (i = 0; i < n; ++i) {
153 if (S_ISDIR(entries[i].mode))
154 ++count;
155 }
156
157 return count;
158}
159
f7ad3c6b 160static int get_task_root(struct task_struct *task, struct path *root)
1da177e4 161{
7c2c7d99
HD		 162 int result = -ENOENT;
163
0494f6ec 164 task_lock(task);
f7ad3c6b
MS		 165 if (task->fs) {
166 get_fs_root(task->fs, root);
7c2c7d99
HD		 167 result = 0;
168 }
0494f6ec 169 task_unlock(task);
7c2c7d99 170 return result;
0494f6ec
MS		 171}
172
7773fbc5 173static int proc_cwd_link(struct dentry *dentry, struct path *path)
0494f6ec 174{
7773fbc5 175 struct task_struct *task = get_proc_task(dentry->d_inode);
0494f6ec 176 int result = -ENOENT;
99f89551
EB		 177
178 if (task) {
f7ad3c6b
MS		 179 task_lock(task);
180 if (task->fs) {
181 get_fs_pwd(task->fs, path);
182 result = 0;
183 }
184 task_unlock(task);
99f89551
EB		 185 put_task_struct(task);
186 }
1da177e4
LT		 187 return result;
188}
189
7773fbc5 190static int proc_root_link(struct dentry *dentry, struct path *path)
1da177e4 191{
7773fbc5 192 struct task_struct *task = get_proc_task(dentry->d_inode);
1da177e4 193 int result = -ENOENT;
99f89551
EB		 194
195 if (task) {
f7ad3c6b 196 result = get_task_root(task, path);
99f89551
EB		 197 put_task_struct(task);
198 }
1da177e4
LT		 199 return result;
200}
201
1da177e4
LT		 202static int proc_pid_cmdline(struct task_struct *task, char * buffer)
203{
204 int res = 0;
205 unsigned int len;
206 struct mm_struct *mm = get_task_mm(task);
207 if (!mm)
208 goto out;
209 if (!mm->arg_end)
210 goto out_mm; /* Shh! No looking before we're done */
211
212 len = mm->arg_end - mm->arg_start;
213
214 if (len > PAGE_SIZE)
215 len = PAGE_SIZE;
216
217 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
218
219 // If the nul at the end of args has been overwritten, then
220 // assume application is using setproctitle(3).
221 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
222 len = strnlen(buffer, res);
223 if (len < res) {
224 res = len;
225 } else {
226 len = mm->env_end - mm->env_start;
227 if (len > PAGE_SIZE - res)
228 len = PAGE_SIZE - res;
229 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
230 res = strnlen(buffer, res);
231 }
232 }
233out_mm:
234 mmput(mm);
235out:
236 return res;
237}
238
239static int proc_pid_auxv(struct task_struct *task, char *buffer)
240{
e7dcd999 241 struct mm_struct *mm = mm_access(task, PTRACE_MODE_READ);
2fadaef4
AV		 242 int res = PTR_ERR(mm);
243 if (mm && !IS_ERR(mm)) {
1da177e4 244 unsigned int nwords = 0;
dfe6b7d9 245 do {
1da177e4 246 nwords += 2;
dfe6b7d9 247 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
1da177e4
LT		 248 res = nwords * sizeof(mm->saved_auxv[0]);
249 if (res > PAGE_SIZE)
250 res = PAGE_SIZE;
251 memcpy(buffer, mm->saved_auxv, res);
252 mmput(mm);
253 }
254 return res;
255}
256
257
258#ifdef CONFIG_KALLSYMS
259/*
260 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
261 * Returns the resolved symbol. If that fails, simply return the address.
262 */
263static int proc_pid_wchan(struct task_struct *task, char *buffer)
264{
ffb45122 265 unsigned long wchan;
9281acea 266 char symname[KSYM_NAME_LEN];
1da177e4
LT		 267
268 wchan = get_wchan(task);
269
9d65cb4a 270 if (lookup_symbol_name(wchan, symname) < 0)
f83ce3e6
JE		 271 if (!ptrace_may_access(task, PTRACE_MODE_READ))
272 return 0;
273 else
274 return sprintf(buffer, "%lu", wchan);
9d65cb4a
AD		 275 else
276 return sprintf(buffer, "%s", symname);
1da177e4
LT		 277}
278#endif /* CONFIG_KALLSYMS */
279
a9712bc1
AV		 280static int lock_trace(struct task_struct *task)
281{
282 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
283 if (err)
284 return err;
285 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
286 mutex_unlock(&task->signal->cred_guard_mutex);
287 return -EPERM;
288 }
289 return 0;
290}
291
292static void unlock_trace(struct task_struct *task)
293{
294 mutex_unlock(&task->signal->cred_guard_mutex);
295}
296
2ec220e2
KC		 297#ifdef CONFIG_STACKTRACE
298
299#define MAX_STACK_TRACE_DEPTH 64
300
301static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
302 struct pid *pid, struct task_struct *task)
303{
304 struct stack_trace trace;
305 unsigned long *entries;
a9712bc1 306 int err;
2ec220e2
KC		 307 int i;
308
309 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
310 if (!entries)
311 return -ENOMEM;
312
313 trace.nr_entries = 0;
314 trace.max_entries = MAX_STACK_TRACE_DEPTH;
315 trace.entries = entries;
316 trace.skip = 0;
2ec220e2 317
a9712bc1
AV		 318 err = lock_trace(task);
319 if (!err) {
320 save_stack_trace_tsk(task, &trace);
321
322 for (i = 0; i < trace.nr_entries; i++) {
b81a618d 323 seq_printf(m, "[<%pK>] %pS\n",
a9712bc1
AV		 324 (void *)entries[i], (void *)entries[i]);
325 }
326 unlock_trace(task);
2ec220e2
KC		 327 }
328 kfree(entries);
329
a9712bc1 330 return err;
2ec220e2
KC		 331}
332#endif
333
1da177e4
LT		 334#ifdef CONFIG_SCHEDSTATS
335/*
336 * Provides /proc/PID/schedstat
337 */
338static int proc_pid_schedstat(struct task_struct *task, char *buffer)
339{
172ba844 340 return sprintf(buffer, "%llu %llu %lu\n",
826e08b0
IM		 341 (unsigned long long)task->se.sum_exec_runtime,
342 (unsigned long long)task->sched_info.run_delay,
2d72376b 343 task->sched_info.pcount);
1da177e4
LT		 344}
345#endif
346
9745512c
AV		 347#ifdef CONFIG_LATENCYTOP
348static int lstats_show_proc(struct seq_file *m, void *v)
349{
350 int i;
13d77c37
HS		 351 struct inode *inode = m->private;
352 struct task_struct *task = get_proc_task(inode);
9745512c 353
13d77c37
HS		 354 if (!task)
355 return -ESRCH;
356 seq_puts(m, "Latency Top version : v0.1\n");
9745512c 357 for (i = 0; i < 32; i++) {
34e49d4f
JP		 358 struct latency_record *lr = &task->latency_record[i];
359 if (lr->backtrace[0]) {
9745512c 360 int q;
34e49d4f
JP		 361 seq_printf(m, "%i %li %li",
362 lr->count, lr->time, lr->max);
9745512c 363 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
34e49d4f
JP		 364 unsigned long bt = lr->backtrace[q];
365 if (!bt)
9745512c 366 break;
34e49d4f 367 if (bt == ULONG_MAX)
9745512c 368 break;
34e49d4f 369 seq_printf(m, " %ps", (void *)bt);
9745512c 370 }
9d6de12f 371 seq_putc(m, '\n');
9745512c
AV		 372 }
373
374 }
13d77c37 375 put_task_struct(task);
9745512c
AV		 376 return 0;
377}
378
379static int lstats_open(struct inode *inode, struct file *file)
380{
13d77c37 381 return single_open(file, lstats_show_proc, inode);
d6643d12
HS		 382}
383
9745512c
AV		 384static ssize_t lstats_write(struct file *file, const char __user *buf,
385 size_t count, loff_t *offs)
386{
13d77c37 387 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
9745512c 388
13d77c37
HS		 389 if (!task)
390 return -ESRCH;
9745512c 391 clear_all_latency_tracing(task);
13d77c37 392 put_task_struct(task);
9745512c
AV		 393
394 return count;
395}
396
397static const struct file_operations proc_lstats_operations = {
398 .open = lstats_open,
399 .read = seq_read,
400 .write = lstats_write,
401 .llseek = seq_lseek,
13d77c37 402 .release = single_release,
9745512c
AV		 403};
404
405#endif
406
1da177e4
LT		 407static int proc_oom_score(struct task_struct *task, char *buffer)
408{
a7f638f9 409 unsigned long totalpages = totalram_pages + total_swap_pages;
b95c35e7 410 unsigned long points = 0;
1da177e4 411
19c5d45a 412 read_lock(&tasklist_lock);
b95c35e7 413 if (pid_alive(task))
a7f638f9
DR		 414 points = oom_badness(task, NULL, NULL, totalpages) *
415 1000 / totalpages;
19c5d45a 416 read_unlock(&tasklist_lock);
1da177e4
LT		 417 return sprintf(buffer, "%lu\n", points);
418}
419
d85f50d5
NH		 420struct limit_names {
421 char *name;
422 char *unit;
423};
424
425static const struct limit_names lnames[RLIM_NLIMITS] = {
cff4edb5 426 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
d85f50d5
NH		 427 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
428 [RLIMIT_DATA] = {"Max data size", "bytes"},
429 [RLIMIT_STACK] = {"Max stack size", "bytes"},
430 [RLIMIT_CORE] = {"Max core file size", "bytes"},
431 [RLIMIT_RSS] = {"Max resident set", "bytes"},
432 [RLIMIT_NPROC] = {"Max processes", "processes"},
433 [RLIMIT_NOFILE] = {"Max open files", "files"},
434 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
435 [RLIMIT_AS] = {"Max address space", "bytes"},
436 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
437 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
438 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
439 [RLIMIT_NICE] = {"Max nice priority", NULL},
440 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
8808117c 441 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
d85f50d5
NH		 442};
443
444/* Display limits for a process */
445static int proc_pid_limits(struct task_struct *task, char *buffer)
446{
447 unsigned int i;
448 int count = 0;
449 unsigned long flags;
450 char *bufptr = buffer;
451
452 struct rlimit rlim[RLIM_NLIMITS];
453
a6bebbc8 454 if (!lock_task_sighand(task, &flags))
d85f50d5 455 return 0;
d85f50d5
NH		 456 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
457 unlock_task_sighand(task, &flags);
d85f50d5
NH		 458
459 /*
460 * print the file header
461 */
462 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
463 "Limit", "Soft Limit", "Hard Limit", "Units");
464
465 for (i = 0; i < RLIM_NLIMITS; i++) {
466 if (rlim[i].rlim_cur == RLIM_INFINITY)
467 count += sprintf(&bufptr[count], "%-25s %-20s ",
468 lnames[i].name, "unlimited");
469 else
470 count += sprintf(&bufptr[count], "%-25s %-20lu ",
471 lnames[i].name, rlim[i].rlim_cur);
472
473 if (rlim[i].rlim_max == RLIM_INFINITY)
474 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
475 else
476 count += sprintf(&bufptr[count], "%-20lu ",
477 rlim[i].rlim_max);
478
479 if (lnames[i].unit)
480 count += sprintf(&bufptr[count], "%-10s\n",
481 lnames[i].unit);
482 else
483 count += sprintf(&bufptr[count], "\n");
484 }
485
486 return count;
487}
488
ebcb6734
RM		 489#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
490static int proc_pid_syscall(struct task_struct *task, char *buffer)
491{
492 long nr;
493 unsigned long args[6], sp, pc;
a9712bc1
AV		 494 int res = lock_trace(task);
495 if (res)
496 return res;
ebcb6734
RM		 497
498 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
a9712bc1
AV		 499 res = sprintf(buffer, "running\n");
500 else if (nr < 0)
501 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
502 else
503 res = sprintf(buffer,
ebcb6734
RM		 504 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
505 nr,
506 args[0], args[1], args[2], args[3], args[4], args[5],
507 sp, pc);
a9712bc1
AV		 508 unlock_trace(task);
509 return res;
ebcb6734
RM		 510}
511#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
512
1da177e4
LT		 513/************************************************************************/
514/* Here the fs part begins */
515/************************************************************************/
516
517/* permission checks */
778c1144 518static int proc_fd_access_allowed(struct inode *inode)
1da177e4 519{
778c1144
EB		 520 struct task_struct *task;
521 int allowed = 0;
df26c40e
EB		 522 /* Allow access to a task's file descriptors if it is us or we
523 * may use ptrace attach to the process and find out that
524 * information.
778c1144
EB		 525 */
526 task = get_proc_task(inode);
df26c40e 527 if (task) {
006ebb40 528 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
778c1144 529 put_task_struct(task);
df26c40e 530 }
778c1144 531 return allowed;
1da177e4
LT		 532}
533
6b4e306a 534int proc_setattr(struct dentry *dentry, struct iattr *attr)
6d76fa58
LT		 535{
536 int error;
537 struct inode *inode = dentry->d_inode;
538
539 if (attr->ia_valid & ATTR_MODE)
540 return -EPERM;
541
542 error = inode_change_ok(inode, attr);
1025774c
CH		 543 if (error)
544 return error;
545
546 if ((attr->ia_valid & ATTR_SIZE) &&
547 attr->ia_size != i_size_read(inode)) {
548 error = vmtruncate(inode, attr->ia_size);
549 if (error)
550 return error;
551 }
552
553 setattr_copy(inode, attr);
554 mark_inode_dirty(inode);
555 return 0;
6d76fa58
LT		 556}
557
0499680a
VK		 558/*
559 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
560 * or euid/egid (for hide_pid_min=2)?
561 */
562static bool has_pid_permissions(struct pid_namespace *pid,
563 struct task_struct *task,
564 int hide_pid_min)
565{
566 if (pid->hide_pid < hide_pid_min)
567 return true;
568 if (in_group_p(pid->pid_gid))
569 return true;
570 return ptrace_may_access(task, PTRACE_MODE_READ);
571}
572
573
574static int proc_pid_permission(struct inode *inode, int mask)
575{
576 struct pid_namespace *pid = inode->i_sb->s_fs_info;
577 struct task_struct *task;
578 bool has_perms;
579
580 task = get_proc_task(inode);
a2ef990a
XF		 581 if (!task)
582 return -ESRCH;
0499680a
VK		 583 has_perms = has_pid_permissions(pid, task, 1);
584 put_task_struct(task);
585
586 if (!has_perms) {
587 if (pid->hide_pid == 2) {
588 /*
589 * Let's make getdents(), stat(), and open()
590 * consistent with each other. If a process
591 * may not stat() a file, it shouldn't be seen
592 * in procfs at all.
593 */
594 return -ENOENT;
595 }
596
597 return -EPERM;
598 }
599 return generic_permission(inode, mask);
600}
601
602
603
c5ef1c42 604static const struct inode_operations proc_def_inode_operations = {
6d76fa58
LT		 605 .setattr = proc_setattr,
606};
607
1da177e4
LT		 608#define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
609
610static ssize_t proc_info_read(struct file * file, char __user * buf,
611 size_t count, loff_t *ppos)
612{
2fddfeef 613 struct inode * inode = file->f_path.dentry->d_inode;
1da177e4
LT		 614 unsigned long page;
615 ssize_t length;
99f89551
EB		 616 struct task_struct *task = get_proc_task(inode);
617
618 length = -ESRCH;
619 if (!task)
620 goto out_no_task;
1da177e4
LT		 621
622 if (count > PROC_BLOCK_SIZE)
623 count = PROC_BLOCK_SIZE;
99f89551
EB		 624
625 length = -ENOMEM;
e12ba74d 626 if (!(page = __get_free_page(GFP_TEMPORARY)))
99f89551 627 goto out;
1da177e4
LT		 628
629 length = PROC_I(inode)->op.proc_read(task, (char*)page);
630
631 if (length >= 0)
632 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
633 free_page(page);
99f89551
EB		 634out:
635 put_task_struct(task);
636out_no_task:
1da177e4
LT		 637 return length;
638}
639
00977a59 640static const struct file_operations proc_info_file_operations = {
1da177e4 641 .read = proc_info_read,
87df8424 642 .llseek = generic_file_llseek,
1da177e4
LT		 643};
644
be614086
EB		 645static int proc_single_show(struct seq_file *m, void *v)
646{
647 struct inode *inode = m->private;
648 struct pid_namespace *ns;
649 struct pid *pid;
650 struct task_struct *task;
651 int ret;
652
653 ns = inode->i_sb->s_fs_info;
654 pid = proc_pid(inode);
655 task = get_pid_task(pid, PIDTYPE_PID);
656 if (!task)
657 return -ESRCH;
658
659 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
660
661 put_task_struct(task);
662 return ret;
663}
664
665static int proc_single_open(struct inode *inode, struct file *filp)
666{
c6a34058 667 return single_open(filp, proc_single_show, inode);
be614086
EB		 668}
669
670static const struct file_operations proc_single_file_operations = {
671 .open = proc_single_open,
672 .read = seq_read,
673 .llseek = seq_lseek,
674 .release = single_release,
675};
676
b409e578 677static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
1da177e4 678{
e268337d
LT		 679 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
680 struct mm_struct *mm;
681
682 if (!task)
683 return -ESRCH;
684
b409e578 685 mm = mm_access(task, mode);
e268337d
LT		 686 put_task_struct(task);
687
688 if (IS_ERR(mm))
689 return PTR_ERR(mm);
690
6d08f2c7
ON		 691 if (mm) {
692 /* ensure this mm_struct can't be freed */
693 atomic_inc(&mm->mm_count);
694 /* but do not pin its memory */
695 mmput(mm);
696 }
697
e268337d
LT		 698 file->private_data = mm;
699
1da177e4
LT		 700 return 0;
701}
702
b409e578
CW		 703static int mem_open(struct inode *inode, struct file *file)
704{
bc452b4b
DH		 705 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
706
707 /* OK to pass negative loff_t, we can catch out-of-range */
708 file->f_mode |= FMODE_UNSIGNED_OFFSET;
709
710 return ret;
b409e578
CW		 711}
712
572d34b9
ON		 713static ssize_t mem_rw(struct file *file, char __user *buf,
714 size_t count, loff_t *ppos, int write)
1da177e4 715{
e268337d 716 struct mm_struct *mm = file->private_data;
572d34b9
ON		 717 unsigned long addr = *ppos;
718 ssize_t copied;
1da177e4 719 char *page;
1da177e4 720
e268337d
LT		 721 if (!mm)
722 return 0;
99f89551 723
30cd8903
KM		 724 page = (char *)__get_free_page(GFP_TEMPORARY);
725 if (!page)
e268337d 726 return -ENOMEM;
1da177e4 727
f7ca54f4 728 copied = 0;
6d08f2c7
ON		 729 if (!atomic_inc_not_zero(&mm->mm_users))
730 goto free;
731
1da177e4 732 while (count > 0) {
572d34b9 733 int this_len = min_t(int, count, PAGE_SIZE);
1da177e4 734
572d34b9 735 if (write && copy_from_user(page, buf, this_len)) {
1da177e4
LT		 736 copied = -EFAULT;
737 break;
738 }
572d34b9
ON		 739
740 this_len = access_remote_vm(mm, addr, page, this_len, write);
741 if (!this_len) {
1da177e4
LT		 742 if (!copied)
743 copied = -EIO;
744 break;
745 }
572d34b9
ON		 746
747 if (!write && copy_to_user(buf, page, this_len)) {
748 copied = -EFAULT;
749 break;
750 }
751
752 buf += this_len;
753 addr += this_len;
754 copied += this_len;
755 count -= this_len;
1da177e4 756 }
572d34b9 757 *ppos = addr;
30cd8903 758
6d08f2c7
ON		 759 mmput(mm);
760free:
30cd8903 761 free_page((unsigned long) page);
1da177e4
LT		 762 return copied;
763}
1da177e4 764
572d34b9
ON		 765static ssize_t mem_read(struct file *file, char __user *buf,
766 size_t count, loff_t *ppos)
767{
768 return mem_rw(file, buf, count, ppos, 0);
769}
770
771static ssize_t mem_write(struct file *file, const char __user *buf,
772 size_t count, loff_t *ppos)
773{
774 return mem_rw(file, (char __user*)buf, count, ppos, 1);
775}
776
85863e47 777loff_t mem_lseek(struct file *file, loff_t offset, int orig)
1da177e4
LT		 778{
779 switch (orig) {
780 case 0:
781 file->f_pos = offset;
782 break;
783 case 1:
784 file->f_pos += offset;
785 break;
786 default:
787 return -EINVAL;
788 }
789 force_successful_syscall_return();
790 return file->f_pos;
791}
792
e268337d
LT		 793static int mem_release(struct inode *inode, struct file *file)
794{
795 struct mm_struct *mm = file->private_data;
71879d3c 796 if (mm)
6d08f2c7 797 mmdrop(mm);
e268337d
LT		 798 return 0;
799}
800
00977a59 801static const struct file_operations proc_mem_operations = {
1da177e4
LT		 802 .llseek = mem_lseek,
803 .read = mem_read,
804 .write = mem_write,
805 .open = mem_open,
e268337d 806 .release = mem_release,
1da177e4
LT		 807};
808
b409e578
CW		 809static int environ_open(struct inode *inode, struct file *file)
810{
811 return __mem_open(inode, file, PTRACE_MODE_READ);
812}
813
315e28c8
JP		 814static ssize_t environ_read(struct file *file, char __user *buf,
815 size_t count, loff_t *ppos)
816{
315e28c8
JP		 817 char *page;
818 unsigned long src = *ppos;
b409e578
CW		 819 int ret = 0;
820 struct mm_struct *mm = file->private_data;
315e28c8 821
b409e578
CW		 822 if (!mm)
823 return 0;
315e28c8 824
315e28c8
JP		 825 page = (char *)__get_free_page(GFP_TEMPORARY);
826 if (!page)
b409e578 827 return -ENOMEM;
315e28c8 828
d6f64b89 829 ret = 0;
b409e578
CW		 830 if (!atomic_inc_not_zero(&mm->mm_users))
831 goto free;
315e28c8 832 while (count > 0) {
e8905ec2
DH		 833 size_t this_len, max_len;
834 int retval;
315e28c8 835
e8905ec2 836 if (src >= (mm->env_end - mm->env_start))
315e28c8
JP		 837 break;
838
e8905ec2
DH		 839 this_len = mm->env_end - (mm->env_start + src);
840
841 max_len = min_t(size_t, PAGE_SIZE, count);
842 this_len = min(max_len, this_len);
315e28c8 843
b409e578 844 retval = access_remote_vm(mm, (mm->env_start + src),
315e28c8
JP		 845 page, this_len, 0);
846
847 if (retval <= 0) {
848 ret = retval;
849 break;
850 }
851
852 if (copy_to_user(buf, page, retval)) {
853 ret = -EFAULT;
854 break;
855 }
856
857 ret += retval;
858 src += retval;
859 buf += retval;
860 count -= retval;
861 }
862 *ppos = src;
315e28c8 863 mmput(mm);
b409e578
CW		 864
865free:
315e28c8 866 free_page((unsigned long) page);
315e28c8
JP		 867 return ret;
868}
869
870static const struct file_operations proc_environ_operations = {
b409e578 871 .open = environ_open,
315e28c8 872 .read = environ_read,
87df8424 873 .llseek = generic_file_llseek,
b409e578 874 .release = mem_release,
315e28c8
JP		 875};
876
1da177e4
LT		 877static ssize_t oom_adjust_read(struct file *file, char __user *buf,
878 size_t count, loff_t *ppos)
879{
2fddfeef 880 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
8578cea7 881 char buffer[PROC_NUMBUF];
1da177e4 882 size_t len;
28b83c51
KM		 883 int oom_adjust = OOM_DISABLE;
884 unsigned long flags;
1da177e4 885
99f89551
EB		 886 if (!task)
887 return -ESRCH;
28b83c51
KM		 888
889 if (lock_task_sighand(task, &flags)) {
890 oom_adjust = task->signal->oom_adj;
891 unlock_task_sighand(task, &flags);
892 }
893
99f89551
EB		 894 put_task_struct(task);
895
8578cea7 896 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
0c28f287
AM		 897
898 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1da177e4
LT		 899}
900
901static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
902 size_t count, loff_t *ppos)
903{
99f89551 904 struct task_struct *task;
5d863b89 905 char buffer[PROC_NUMBUF];
0a8cb8e3 906 int oom_adjust;
28b83c51 907 unsigned long flags;
5d863b89 908 int err;
1da177e4 909
8578cea7
EB		 910 memset(buffer, 0, sizeof(buffer));
911 if (count > sizeof(buffer) - 1)
912 count = sizeof(buffer) - 1;
723548bf
DR		 913 if (copy_from_user(buffer, buf, count)) {
914 err = -EFAULT;
915 goto out;
916 }
5d863b89 917
0a8cb8e3 918 err = kstrtoint(strstrip(buffer), 0, &oom_adjust);
5d863b89 919 if (err)
723548bf 920 goto out;
8ac773b4 921 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
723548bf
DR		 922 oom_adjust != OOM_DISABLE) {
923 err = -EINVAL;
924 goto out;
925 }
5d863b89 926
2fddfeef 927 task = get_proc_task(file->f_path.dentry->d_inode);
723548bf
DR		 928 if (!task) {
929 err = -ESRCH;
930 goto out;
931 }
d19d5476
DR		 932
933 task_lock(task);
934 if (!task->mm) {
935 err = -EINVAL;
936 goto err_task_lock;
937 }
938
28b83c51 939 if (!lock_task_sighand(task, &flags)) {
723548bf 940 err = -ESRCH;
d19d5476 941 goto err_task_lock;
28b83c51
KM		 942 }
943
944 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
723548bf
DR		 945 err = -EACCES;
946 goto err_sighand;
8fb4fc68 947 }
28b83c51 948
51b1bd2a
DR		 949 /*
950 * Warn that /proc/pid/oom_adj is deprecated, see
951 * Documentation/feature-removal-schedule.txt.
952 */
c2142704 953 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
be8f684d
DR		 954 current->comm, task_pid_nr(current), task_pid_nr(task),
955 task_pid_nr(task));
28b83c51 956 task->signal->oom_adj = oom_adjust;
a63d83f4
DR		 957 /*
958 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
959 * value is always attainable.
960 */
961 if (task->signal->oom_adj == OOM_ADJUST_MAX)
962 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
963 else
964 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
965 -OOM_DISABLE;
43d2b113 966 trace_oom_score_adj_update(task);
723548bf 967err_sighand:
28b83c51 968 unlock_task_sighand(task, &flags);
d19d5476
DR		 969err_task_lock:
970 task_unlock(task);
99f89551 971 put_task_struct(task);
723548bf
DR		 972out:
973 return err < 0 ? err : count;
1da177e4
LT		 974}
975
00977a59 976static const struct file_operations proc_oom_adjust_operations = {
1da177e4
LT		 977 .read = oom_adjust_read,
978 .write = oom_adjust_write,
87df8424 979 .llseek = generic_file_llseek,
1da177e4
LT		 980};
981
a63d83f4
DR		 982static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
983 size_t count, loff_t *ppos)
984{
985 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
986 char buffer[PROC_NUMBUF];
987 int oom_score_adj = OOM_SCORE_ADJ_MIN;
988 unsigned long flags;
989 size_t len;
990
991 if (!task)
992 return -ESRCH;
993 if (lock_task_sighand(task, &flags)) {
994 oom_score_adj = task->signal->oom_score_adj;
995 unlock_task_sighand(task, &flags);
996 }
997 put_task_struct(task);
998 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
999 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1000}
1001
1002static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1003 size_t count, loff_t *ppos)
1004{
1005 struct task_struct *task;
1006 char buffer[PROC_NUMBUF];
1007 unsigned long flags;
0a8cb8e3 1008 int oom_score_adj;
a63d83f4
DR		 1009 int err;
1010
1011 memset(buffer, 0, sizeof(buffer));
1012 if (count > sizeof(buffer) - 1)
1013 count = sizeof(buffer) - 1;
723548bf
DR		 1014 if (copy_from_user(buffer, buf, count)) {
1015 err = -EFAULT;
1016 goto out;
1017 }
a63d83f4 1018
0a8cb8e3 1019 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
a63d83f4 1020 if (err)
723548bf 1021 goto out;
a63d83f4 1022 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
723548bf
DR		 1023 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1024 err = -EINVAL;
1025 goto out;
1026 }
a63d83f4
DR		 1027
1028 task = get_proc_task(file->f_path.dentry->d_inode);
723548bf
DR		 1029 if (!task) {
1030 err = -ESRCH;
1031 goto out;
1032 }
d19d5476
DR		 1033
1034 task_lock(task);
1035 if (!task->mm) {
1036 err = -EINVAL;
1037 goto err_task_lock;
1038 }
1039
a63d83f4 1040 if (!lock_task_sighand(task, &flags)) {
723548bf 1041 err = -ESRCH;
d19d5476 1042 goto err_task_lock;
a63d83f4 1043 }
d19d5476 1044
dabb16f6 1045 if (oom_score_adj < task->signal->oom_score_adj_min &&
a63d83f4 1046 !capable(CAP_SYS_RESOURCE)) {
723548bf
DR		 1047 err = -EACCES;
1048 goto err_sighand;
a63d83f4
DR		 1049 }
1050
1051 task->signal->oom_score_adj = oom_score_adj;
dabb16f6
MSB		 1052 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1053 task->signal->oom_score_adj_min = oom_score_adj;
43d2b113 1054 trace_oom_score_adj_update(task);
a63d83f4
DR		 1055 /*
1056 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1057 * always attainable.
1058 */
1059 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1060 task->signal->oom_adj = OOM_DISABLE;
1061 else
1062 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1063 OOM_SCORE_ADJ_MAX;
723548bf 1064err_sighand:
a63d83f4 1065 unlock_task_sighand(task, &flags);
d19d5476
DR		 1066err_task_lock:
1067 task_unlock(task);
a63d83f4 1068 put_task_struct(task);
723548bf
DR		 1069out:
1070 return err < 0 ? err : count;
a63d83f4
DR		 1071}
1072
1073static const struct file_operations proc_oom_score_adj_operations = {
1074 .read = oom_score_adj_read,
1075 .write = oom_score_adj_write,
6038f373 1076 .llseek = default_llseek,
a63d83f4
DR		 1077};
1078
1da177e4
LT		 1079#ifdef CONFIG_AUDITSYSCALL
1080#define TMPBUFLEN 21
1081static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1082 size_t count, loff_t *ppos)
1083{
2fddfeef 1084 struct inode * inode = file->f_path.dentry->d_inode;
99f89551 1085 struct task_struct *task = get_proc_task(inode);
1da177e4
LT		 1086 ssize_t length;
1087 char tmpbuf[TMPBUFLEN];
1088
99f89551
EB		 1089 if (!task)
1090 return -ESRCH;
1da177e4 1091 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
e1760bd5
EB		 1092 from_kuid(file->f_cred->user_ns,
1093 audit_get_loginuid(task)));
99f89551 1094 put_task_struct(task);
1da177e4
LT		 1095 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1096}
1097
1098static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1099 size_t count, loff_t *ppos)
1100{
2fddfeef 1101 struct inode * inode = file->f_path.dentry->d_inode;
1da177e4
LT		 1102 char *page, *tmp;
1103 ssize_t length;
1da177e4 1104 uid_t loginuid;
e1760bd5 1105 kuid_t kloginuid;
1da177e4 1106
7dc52157
PM		 1107 rcu_read_lock();
1108 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1109 rcu_read_unlock();
1da177e4 1110 return -EPERM;
7dc52157
PM		 1111 }
1112 rcu_read_unlock();
1da177e4 1113
e0182909
AV		 1114 if (count >= PAGE_SIZE)
1115 count = PAGE_SIZE - 1;
1da177e4
LT		 1116
1117 if (*ppos != 0) {
1118 /* No partial writes. */
1119 return -EINVAL;
1120 }
e12ba74d 1121 page = (char*)__get_free_page(GFP_TEMPORARY);
1da177e4
LT		 1122 if (!page)
1123 return -ENOMEM;
1124 length = -EFAULT;
1125 if (copy_from_user(page, buf, count))
1126 goto out_free_page;
1127
e0182909 1128 page[count] = '\0';
1da177e4
LT		 1129 loginuid = simple_strtoul(page, &tmp, 10);
1130 if (tmp == page) {
1131 length = -EINVAL;
1132 goto out_free_page;
1133
1134 }
e1760bd5
EB		 1135 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1136 if (!uid_valid(kloginuid)) {
1137 length = -EINVAL;
1138 goto out_free_page;
1139 }
1140
1141 length = audit_set_loginuid(kloginuid);
1da177e4
LT		 1142 if (likely(length == 0))
1143 length = count;
1144
1145out_free_page:
1146 free_page((unsigned long) page);
1147 return length;
1148}
1149
00977a59 1150static const struct file_operations proc_loginuid_operations = {
1da177e4
LT		 1151 .read = proc_loginuid_read,
1152 .write = proc_loginuid_write,
87df8424 1153 .llseek = generic_file_llseek,
1da177e4 1154};
1e0bd755
EP		 1155
1156static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1157 size_t count, loff_t *ppos)
1158{
1159 struct inode * inode = file->f_path.dentry->d_inode;
1160 struct task_struct *task = get_proc_task(inode);
1161 ssize_t length;
1162 char tmpbuf[TMPBUFLEN];
1163
1164 if (!task)
1165 return -ESRCH;
1166 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1167 audit_get_sessionid(task));
1168 put_task_struct(task);
1169 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1170}
1171
1172static const struct file_operations proc_sessionid_operations = {
1173 .read = proc_sessionid_read,
87df8424 1174 .llseek = generic_file_llseek,
1e0bd755 1175};
1da177e4
LT		 1176#endif
1177
f4f154fd
AM		 1178#ifdef CONFIG_FAULT_INJECTION
1179static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1180 size_t count, loff_t *ppos)
1181{
1182 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1183 char buffer[PROC_NUMBUF];
1184 size_t len;
1185 int make_it_fail;
f4f154fd
AM		 1186
1187 if (!task)
1188 return -ESRCH;
1189 make_it_fail = task->make_it_fail;
1190 put_task_struct(task);
1191
1192 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
0c28f287
AM		 1193
1194 return simple_read_from_buffer(buf, count, ppos, buffer, len);
f4f154fd
AM		 1195}
1196
1197static ssize_t proc_fault_inject_write(struct file * file,
1198 const char __user * buf, size_t count, loff_t *ppos)
1199{
1200 struct task_struct *task;
1201 char buffer[PROC_NUMBUF], *end;
1202 int make_it_fail;
1203
1204 if (!capable(CAP_SYS_RESOURCE))
1205 return -EPERM;
1206 memset(buffer, 0, sizeof(buffer));
1207 if (count > sizeof(buffer) - 1)
1208 count = sizeof(buffer) - 1;
1209 if (copy_from_user(buffer, buf, count))
1210 return -EFAULT;
cba8aafe
VL		 1211 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1212 if (*end)
1213 return -EINVAL;
f4f154fd
AM		 1214 task = get_proc_task(file->f_dentry->d_inode);
1215 if (!task)
1216 return -ESRCH;
1217 task->make_it_fail = make_it_fail;
1218 put_task_struct(task);
cba8aafe
VL		 1219
1220 return count;
f4f154fd
AM		 1221}
1222
00977a59 1223static const struct file_operations proc_fault_inject_operations = {
f4f154fd
AM		 1224 .read = proc_fault_inject_read,
1225 .write = proc_fault_inject_write,
87df8424 1226 .llseek = generic_file_llseek,
f4f154fd
AM		 1227};
1228#endif
1229
9745512c 1230
43ae34cb
IM		 1231#ifdef CONFIG_SCHED_DEBUG
1232/*
1233 * Print out various scheduling related per-task fields:
1234 */
1235static int sched_show(struct seq_file *m, void *v)
1236{
1237 struct inode *inode = m->private;
1238 struct task_struct *p;
1239
43ae34cb
IM		 1240 p = get_proc_task(inode);
1241 if (!p)
1242 return -ESRCH;
1243 proc_sched_show_task(p, m);
1244
1245 put_task_struct(p);
1246
1247 return 0;
1248}
1249
1250static ssize_t
1251sched_write(struct file *file, const char __user *buf,
1252 size_t count, loff_t *offset)
1253{
1254 struct inode *inode = file->f_path.dentry->d_inode;
1255 struct task_struct *p;
1256
43ae34cb
IM		 1257 p = get_proc_task(inode);
1258 if (!p)
1259 return -ESRCH;
1260 proc_sched_set_task(p);
1261
1262 put_task_struct(p);
1263
1264 return count;
1265}
1266
1267static int sched_open(struct inode *inode, struct file *filp)
1268{
c6a34058 1269 return single_open(filp, sched_show, inode);
43ae34cb
IM		 1270}
1271
1272static const struct file_operations proc_pid_sched_operations = {
1273 .open = sched_open,
1274 .read = seq_read,
1275 .write = sched_write,
1276 .llseek = seq_lseek,
5ea473a1 1277 .release = single_release,
43ae34cb
IM		 1278};
1279
1280#endif
1281
5091faa4
MG		 1282#ifdef CONFIG_SCHED_AUTOGROUP
1283/*
1284 * Print out autogroup related information:
1285 */
1286static int sched_autogroup_show(struct seq_file *m, void *v)
1287{
1288 struct inode *inode = m->private;
1289 struct task_struct *p;
1290
1291 p = get_proc_task(inode);
1292 if (!p)
1293 return -ESRCH;
1294 proc_sched_autogroup_show_task(p, m);
1295
1296 put_task_struct(p);
1297
1298 return 0;
1299}
1300
1301static ssize_t
1302sched_autogroup_write(struct file *file, const char __user *buf,
1303 size_t count, loff_t *offset)
1304{
1305 struct inode *inode = file->f_path.dentry->d_inode;
1306 struct task_struct *p;
1307 char buffer[PROC_NUMBUF];
0a8cb8e3 1308 int nice;
5091faa4
MG		 1309 int err;
1310
1311 memset(buffer, 0, sizeof(buffer));
1312 if (count > sizeof(buffer) - 1)
1313 count = sizeof(buffer) - 1;
1314 if (copy_from_user(buffer, buf, count))
1315 return -EFAULT;
1316
0a8cb8e3
AD		 1317 err = kstrtoint(strstrip(buffer), 0, &nice);
1318 if (err < 0)
1319 return err;
5091faa4
MG		 1320
1321 p = get_proc_task(inode);
1322 if (!p)
1323 return -ESRCH;
1324
2e5b5b3a 1325 err = proc_sched_autogroup_set_nice(p, nice);
5091faa4
MG		 1326 if (err)
1327 count = err;
1328
1329 put_task_struct(p);
1330
1331 return count;
1332}
1333
1334static int sched_autogroup_open(struct inode *inode, struct file *filp)
1335{
1336 int ret;
1337
1338 ret = single_open(filp, sched_autogroup_show, NULL);
1339 if (!ret) {
1340 struct seq_file *m = filp->private_data;
1341
1342 m->private = inode;
1343 }
1344 return ret;
1345}
1346
1347static const struct file_operations proc_pid_sched_autogroup_operations = {
1348 .open = sched_autogroup_open,
1349 .read = seq_read,
1350 .write = sched_autogroup_write,
1351 .llseek = seq_lseek,
1352 .release = single_release,
1353};
1354
1355#endif /* CONFIG_SCHED_AUTOGROUP */
1356
4614a696
JS		 1357static ssize_t comm_write(struct file *file, const char __user *buf,
1358 size_t count, loff_t *offset)
1359{
1360 struct inode *inode = file->f_path.dentry->d_inode;
1361 struct task_struct *p;
1362 char buffer[TASK_COMM_LEN];
1363
1364 memset(buffer, 0, sizeof(buffer));
1365 if (count > sizeof(buffer) - 1)
1366 count = sizeof(buffer) - 1;
1367 if (copy_from_user(buffer, buf, count))
1368 return -EFAULT;
1369
1370 p = get_proc_task(inode);
1371 if (!p)
1372 return -ESRCH;
1373
1374 if (same_thread_group(current, p))
1375 set_task_comm(p, buffer);
1376 else
1377 count = -EINVAL;
1378
1379 put_task_struct(p);
1380
1381 return count;
1382}
1383
1384static int comm_show(struct seq_file *m, void *v)
1385{
1386 struct inode *inode = m->private;
1387 struct task_struct *p;
1388
1389 p = get_proc_task(inode);
1390 if (!p)
1391 return -ESRCH;
1392
1393 task_lock(p);
1394 seq_printf(m, "%s\n", p->comm);
1395 task_unlock(p);
1396
1397 put_task_struct(p);
1398
1399 return 0;
1400}
1401
1402static int comm_open(struct inode *inode, struct file *filp)
1403{
c6a34058 1404 return single_open(filp, comm_show, inode);
4614a696
JS		 1405}
1406
1407static const struct file_operations proc_pid_set_comm_operations = {
1408 .open = comm_open,
1409 .read = seq_read,
1410 .write = comm_write,
1411 .llseek = seq_lseek,
1412 .release = single_release,
1413};
1414
7773fbc5 1415static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
925d1c40
MH		 1416{
1417 struct task_struct *task;
1418 struct mm_struct *mm;
1419 struct file *exe_file;
1420
7773fbc5 1421 task = get_proc_task(dentry->d_inode);
925d1c40
MH		 1422 if (!task)
1423 return -ENOENT;
1424 mm = get_task_mm(task);
1425 put_task_struct(task);
1426 if (!mm)
1427 return -ENOENT;
1428 exe_file = get_mm_exe_file(mm);
1429 mmput(mm);
1430 if (exe_file) {
1431 *exe_path = exe_file->f_path;
1432 path_get(&exe_file->f_path);
1433 fput(exe_file);
1434 return 0;
1435 } else
1436 return -ENOENT;
1437}
1438
008b150a 1439static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1da177e4
LT		 1440{
1441 struct inode *inode = dentry->d_inode;
408ef013 1442 struct path path;
1da177e4
LT		 1443 int error = -EACCES;
1444
778c1144
EB		 1445 /* Are we allowed to snoop on the tasks file descriptors? */
1446 if (!proc_fd_access_allowed(inode))
1da177e4 1447 goto out;
1da177e4 1448
408ef013
CH		 1449 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1450 if (error)
1451 goto out;
1452
b5fb63c1 1453 nd_jump_link(nd, &path);
408ef013 1454 return NULL;
1da177e4 1455out:
008b150a 1456 return ERR_PTR(error);
1da177e4
LT		 1457}
1458
3dcd25f3 1459static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1da177e4 1460{
e12ba74d 1461 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
3dcd25f3 1462 char *pathname;
1da177e4
LT		 1463 int len;
1464
1465 if (!tmp)
1466 return -ENOMEM;
0c28f287 1467
7b2a69ba 1468 pathname = d_path(path, tmp, PAGE_SIZE);
3dcd25f3
JB		 1469 len = PTR_ERR(pathname);
1470 if (IS_ERR(pathname))
1da177e4 1471 goto out;
3dcd25f3 1472 len = tmp + PAGE_SIZE - 1 - pathname;
1da177e4
LT		 1473
1474 if (len > buflen)
1475 len = buflen;
3dcd25f3 1476 if (copy_to_user(buffer, pathname, len))
1da177e4
LT		 1477 len = -EFAULT;
1478 out:
1479 free_page((unsigned long)tmp);
1480 return len;
1481}
1482
1483static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1484{
1485 int error = -EACCES;
1486 struct inode *inode = dentry->d_inode;
3dcd25f3 1487 struct path path;
1da177e4 1488
778c1144
EB		 1489 /* Are we allowed to snoop on the tasks file descriptors? */
1490 if (!proc_fd_access_allowed(inode))
1da177e4 1491 goto out;
1da177e4 1492
7773fbc5 1493 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1da177e4
LT		 1494 if (error)
1495 goto out;
1496
3dcd25f3
JB		 1497 error = do_proc_readlink(&path, buffer, buflen);
1498 path_put(&path);
1da177e4 1499out:
1da177e4
LT		 1500 return error;
1501}
1502
c5ef1c42 1503static const struct inode_operations proc_pid_link_inode_operations = {
1da177e4 1504 .readlink = proc_pid_readlink,
6d76fa58
LT		 1505 .follow_link = proc_pid_follow_link,
1506 .setattr = proc_setattr,
1da177e4
LT		 1507};
1508
28a6d671
EB		 1509
1510/* building an inode */
1511
1512static int task_dumpable(struct task_struct *task)
1da177e4 1513{
28a6d671
EB		 1514 int dumpable = 0;
1515 struct mm_struct *mm;
1da177e4 1516
28a6d671
EB		 1517 task_lock(task);
1518 mm = task->mm;
1519 if (mm)
6c5d5238 1520 dumpable = get_dumpable(mm);
28a6d671
EB		 1521 task_unlock(task);
1522 if(dumpable == 1)
1523 return 1;
1524 return 0;
1525}
1da177e4 1526
6b4e306a 1527struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
28a6d671
EB		 1528{
1529 struct inode * inode;
1530 struct proc_inode *ei;
c69e8d9c 1531 const struct cred *cred;
1da177e4 1532
28a6d671 1533 /* We need a new inode */
1da177e4 1534
28a6d671
EB		 1535 inode = new_inode(sb);
1536 if (!inode)
1537 goto out;
1538
1539 /* Common stuff */
1540 ei = PROC_I(inode);
85fe4025 1541 inode->i_ino = get_next_ino();
28a6d671 1542 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
28a6d671
EB		 1543 inode->i_op = &proc_def_inode_operations;
1544
1545 /*
1546 * grab the reference to task.
1547 */
1a657f78 1548 ei->pid = get_task_pid(task, PIDTYPE_PID);
28a6d671
EB		 1549 if (!ei->pid)
1550 goto out_unlock;
1551
28a6d671 1552 if (task_dumpable(task)) {
c69e8d9c
DH		 1553 rcu_read_lock();
1554 cred = __task_cred(task);
1555 inode->i_uid = cred->euid;
1556 inode->i_gid = cred->egid;
1557 rcu_read_unlock();
1da177e4 1558 }
28a6d671
EB		 1559 security_task_to_inode(task, inode);
1560
1da177e4 1561out:
28a6d671
EB		 1562 return inode;
1563
1564out_unlock:
1565 iput(inode);
1566 return NULL;
1da177e4
LT		 1567}
1568
6b4e306a 1569int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1da177e4 1570{
1da177e4 1571 struct inode *inode = dentry->d_inode;
28a6d671 1572 struct task_struct *task;
c69e8d9c 1573 const struct cred *cred;
0499680a 1574 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
c69e8d9c 1575
28a6d671 1576 generic_fillattr(inode, stat);
1da177e4 1577
28a6d671 1578 rcu_read_lock();
dcb0f222
EB		 1579 stat->uid = GLOBAL_ROOT_UID;
1580 stat->gid = GLOBAL_ROOT_GID;
28a6d671
EB		 1581 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1582 if (task) {
0499680a
VK		 1583 if (!has_pid_permissions(pid, task, 2)) {
1584 rcu_read_unlock();
1585 /*
1586 * This doesn't prevent learning whether PID exists,
1587 * it only makes getattr() consistent with readdir().
1588 */
1589 return -ENOENT;
1590 }
28a6d671
EB		 1591 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1592 task_dumpable(task)) {
c69e8d9c
DH		 1593 cred = __task_cred(task);
1594 stat->uid = cred->euid;
1595 stat->gid = cred->egid;
1da177e4
LT		 1596 }
1597 }
28a6d671 1598 rcu_read_unlock();
d6e71144 1599 return 0;
1da177e4
LT		 1600}
1601
1da177e4
LT		 1602/* dentry stuff */
1603
1604/*
1605 * Exceptional case: normally we are not allowed to unhash a busy
1606 * directory. In this case, however, we can do it - no aliasing problems
1607 * due to the way we treat inodes.
1608 *
1609 * Rewrite the inode's ownerships here because the owning task may have
1610 * performed a setuid(), etc.
99f89551
EB		 1611 *
1612 * Before the /proc/pid/status file was created the only way to read
1613 * the effective uid of a /process was to stat /proc/pid. Reading
1614 * /proc/pid/status is slow enough that procps and other packages
1615 * kept stating /proc/pid. To keep the rules in /proc simple I have
1616 * made this apply to all per process world readable and executable
1617 * directories.
1da177e4 1618 */
0b728e19 1619int pid_revalidate(struct dentry *dentry, unsigned int flags)
1da177e4 1620{
34286d66
NP		 1621 struct inode *inode;
1622 struct task_struct *task;
c69e8d9c
DH		 1623 const struct cred *cred;
1624
0b728e19 1625 if (flags & LOOKUP_RCU)
34286d66
NP		 1626 return -ECHILD;
1627
1628 inode = dentry->d_inode;
1629 task = get_proc_task(inode);
1630
99f89551
EB		 1631 if (task) {
1632 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1633 task_dumpable(task)) {
c69e8d9c
DH		 1634 rcu_read_lock();
1635 cred = __task_cred(task);
1636 inode->i_uid = cred->euid;
1637 inode->i_gid = cred->egid;
1638 rcu_read_unlock();
1da177e4 1639 } else {
dcb0f222
EB		 1640 inode->i_uid = GLOBAL_ROOT_UID;
1641 inode->i_gid = GLOBAL_ROOT_GID;
1da177e4 1642 }
9ee8ab9f 1643 inode->i_mode &= ~(S_ISUID | S_ISGID);
1da177e4 1644 security_task_to_inode(task, inode);
99f89551 1645 put_task_struct(task);
1da177e4
LT		 1646 return 1;
1647 }
1648 d_drop(dentry);
1649 return 0;
1650}
1651
fe15ce44 1652static int pid_delete_dentry(const struct dentry * dentry)
99f89551 1653{
28a6d671
EB		 1654 /* Is the task we represent dead?
1655 * If so, then don't put the dentry on the lru list,
1656 * kill it immediately.
1657 */
1658 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1659}
1660
6b4e306a 1661const struct dentry_operations pid_dentry_operations =
28a6d671
EB		 1662{
1663 .d_revalidate = pid_revalidate,
1664 .d_delete = pid_delete_dentry,
1665};
1666
1667/* Lookups */
1668
1c0d04c9
EB		 1669/*
1670 * Fill a directory entry.
1671 *
1672 * If possible create the dcache entry and derive our inode number and
1673 * file type from dcache entry.
1674 *
1675 * Since all of the proc inode numbers are dynamically generated, the inode
1676 * numbers do not exist until the inode is cache. This means creating the
1677 * the dcache entry in readdir is necessary to keep the inode numbers
1678 * reported by readdir in sync with the inode numbers reported
1679 * by stat.
1680 */
6b4e306a
EB		 1681int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1682 const char *name, int len,
c5141e6d 1683 instantiate_t instantiate, struct task_struct *task, const void *ptr)
61a28784 1684{
2fddfeef 1685 struct dentry *child, *dir = filp->f_path.dentry;
61a28784
EB		 1686 struct inode *inode;
1687 struct qstr qname;
1688 ino_t ino = 0;
1689 unsigned type = DT_UNKNOWN;
1690
1691 qname.name = name;
1692 qname.len = len;
1693 qname.hash = full_name_hash(name, len);
1694
1695 child = d_lookup(dir, &qname);
1696 if (!child) {
1697 struct dentry *new;
1698 new = d_alloc(dir, &qname);
1699 if (new) {
1700 child = instantiate(dir->d_inode, new, task, ptr);
1701 if (child)
1702 dput(new);
1703 else
1704 child = new;
1705 }
1706 }
1707 if (!child || IS_ERR(child) || !child->d_inode)
1708 goto end_instantiate;
1709 inode = child->d_inode;
1710 if (inode) {
1711 ino = inode->i_ino;
1712 type = inode->i_mode >> 12;
1713 }
1714 dput(child);
1715end_instantiate:
1716 if (!ino)
1717 ino = find_inode_number(dir, &qname);
1718 if (!ino)
1719 ino = 1;
1720 return filldir(dirent, name, len, filp->f_pos, ino, type);
1721}
1722
28a6d671
EB		 1723static unsigned name_to_int(struct dentry *dentry)
1724{
1725 const char *name = dentry->d_name.name;
1726 int len = dentry->d_name.len;
1727 unsigned n = 0;
1728
1729 if (len > 1 && *name == '0')
1730 goto out;
1731 while (len-- > 0) {
1732 unsigned c = *name++ - '0';
1733 if (c > 9)
1734 goto out;
1735 if (n >= (~0U-9)/10)
1736 goto out;
1737 n *= 10;
1738 n += c;
1739 }
1740 return n;
1741out:
1742 return ~0U;
1743}
1744
27932742
MS		 1745#define PROC_FDINFO_MAX 64
1746
3dcd25f3 1747static int proc_fd_info(struct inode *inode, struct path *path, char *info)
28a6d671 1748{
5e442a49
LT		 1749 struct task_struct *task = get_proc_task(inode);
1750 struct files_struct *files = NULL;
28a6d671
EB		 1751 struct file *file;
1752 int fd = proc_fd(inode);
aa6afca5 1753
5e442a49
LT		 1754 if (task) {
1755 files = get_files_struct(task);
1756 put_task_struct(task);
1757 }
1758 if (files) {
1759 /*
1760 * We are not taking a ref to the file structure, so we must
1761 * hold ->file_lock.
1762 */
1763 spin_lock(&files->file_lock);
1764 file = fcheck_files(files, fd);
1765 if (file) {
1766 unsigned int f_flags;
1767 struct fdtable *fdt;
1768
1769 fdt = files_fdtable(files);
1770 f_flags = file->f_flags & ~O_CLOEXEC;
1dce27c5 1771 if (close_on_exec(fd, fdt))
5e442a49
LT		 1772 f_flags |= O_CLOEXEC;
1773
1774 if (path) {
1775 *path = file->f_path;
1776 path_get(&file->f_path);
1777 }
1778 if (info)
1779 snprintf(info, PROC_FDINFO_MAX,
1780 "pos:\t%lli\n"
1781 "flags:\t0%o\n",
1782 (long long) file->f_pos,
1783 f_flags);
1784 spin_unlock(&files->file_lock);
1785 put_files_struct(files);
1786 return 0;
99f89551 1787 }
5e442a49
LT		 1788 spin_unlock(&files->file_lock);
1789 put_files_struct(files);
1790 }
1791 return -ENOENT;
99f89551
EB		 1792}
1793
7773fbc5 1794static int proc_fd_link(struct dentry *dentry, struct path *path)
27932742 1795{
7773fbc5 1796 return proc_fd_info(dentry->d_inode, path, NULL);
27932742
MS		 1797}
1798
0b728e19 1799static int tid_fd_revalidate(struct dentry *dentry, unsigned int flags)
1da177e4 1800{
34286d66
NP		 1801 struct inode *inode;
1802 struct task_struct *task;
1803 int fd;
1da177e4 1804 struct files_struct *files;
c69e8d9c 1805 const struct cred *cred;
1da177e4 1806
0b728e19 1807 if (flags & LOOKUP_RCU)
34286d66
NP		 1808 return -ECHILD;
1809
1810 inode = dentry->d_inode;
1811 task = get_proc_task(inode);
1812 fd = proc_fd(inode);
1813
99f89551
EB		 1814 if (task) {
1815 files = get_files_struct(task);
1816 if (files) {
30a08bf2 1817 struct file *file;
99f89551 1818 rcu_read_lock();
30a08bf2
LT		 1819 file = fcheck_files(files, fd);
1820 if (file) {
0640113b 1821 unsigned f_mode = file->f_mode;
30a08bf2 1822
99f89551
EB		 1823 rcu_read_unlock();
1824 put_files_struct(files);
30a08bf2 1825
99f89551 1826 if (task_dumpable(task)) {
c69e8d9c
DH		 1827 rcu_read_lock();
1828 cred = __task_cred(task);
1829 inode->i_uid = cred->euid;
1830 inode->i_gid = cred->egid;
1831 rcu_read_unlock();
99f89551 1832 } else {
dcb0f222
EB		 1833 inode->i_uid = GLOBAL_ROOT_UID;
1834 inode->i_gid = GLOBAL_ROOT_GID;
99f89551 1835 }
30a08bf2 1836
0640113b
LT		 1837 if (S_ISLNK(inode->i_mode)) {
1838 unsigned i_mode = S_IFLNK;
1839 if (f_mode & FMODE_READ)
1840 i_mode |= S_IRUSR | S_IXUSR;
1841 if (f_mode & FMODE_WRITE)
1842 i_mode |= S_IWUSR | S_IXUSR;
1843 inode->i_mode = i_mode;
1844 }
30a08bf2 1845
99f89551
EB		 1846 security_task_to_inode(task, inode);
1847 put_task_struct(task);
1848 return 1;
1849 }
b835996f 1850 rcu_read_unlock();
1da177e4 1851 put_files_struct(files);
1da177e4 1852 }
99f89551 1853 put_task_struct(task);
1da177e4
LT		 1854 }
1855 d_drop(dentry);
1856 return 0;
1857}
1858
d72f71eb 1859static const struct dentry_operations tid_fd_dentry_operations =
1da177e4
LT		 1860{
1861 .d_revalidate = tid_fd_revalidate,
1862 .d_delete = pid_delete_dentry,
1863};
1864
444ceed8 1865static struct dentry *proc_fd_instantiate(struct inode *dir,
c5141e6d 1866 struct dentry *dentry, struct task_struct *task, const void *ptr)
1da177e4 1867{
af5e6171 1868 unsigned fd = (unsigned long)ptr;
444ceed8
EB		 1869 struct inode *inode;
1870 struct proc_inode *ei;
1871 struct dentry *error = ERR_PTR(-ENOENT);
1da177e4 1872
61a28784 1873 inode = proc_pid_make_inode(dir->i_sb, task);
1da177e4
LT		 1874 if (!inode)
1875 goto out;
1876 ei = PROC_I(inode);
aed7a6c4 1877 ei->fd = fd;
444ceed8 1878
0640113b 1879 inode->i_mode = S_IFLNK;
5e442a49 1880 inode->i_op = &proc_pid_link_inode_operations;
1da177e4
LT		 1881 inode->i_size = 64;
1882 ei->op.proc_get_link = proc_fd_link;
fb045adb 1883 d_set_d_op(dentry, &tid_fd_dentry_operations);
1da177e4 1884 d_add(dentry, inode);
cd6a3ce9 1885 /* Close the race of the process dying before we return the dentry */
0b728e19 1886 if (tid_fd_revalidate(dentry, 0))
444ceed8 1887 error = NULL;
1da177e4 1888
444ceed8
EB		 1889 out:
1890 return error;
1da177e4
LT		 1891}
1892
27932742
MS		 1893static struct dentry *proc_lookupfd_common(struct inode *dir,
1894 struct dentry *dentry,
1895 instantiate_t instantiate)
444ceed8
EB		 1896{
1897 struct task_struct *task = get_proc_task(dir);
1898 unsigned fd = name_to_int(dentry);
1899 struct dentry *result = ERR_PTR(-ENOENT);
1900
1901 if (!task)
1902 goto out_no_task;
1903 if (fd == ~0U)
1904 goto out;
1905
af5e6171 1906 result = instantiate(dir, dentry, task, (void *)(unsigned long)fd);
444ceed8
EB		 1907out:
1908 put_task_struct(task);
1909out_no_task:
1910 return result;
1911}
1912
27932742
MS		 1913static int proc_readfd_common(struct file * filp, void * dirent,
1914 filldir_t filldir, instantiate_t instantiate)
28a6d671 1915{
2fddfeef 1916 struct dentry *dentry = filp->f_path.dentry;
28a6d671
EB		 1917 struct inode *inode = dentry->d_inode;
1918 struct task_struct *p = get_proc_task(inode);
457c2510 1919 unsigned int fd, ino;
28a6d671 1920 int retval;
28a6d671 1921 struct files_struct * files;
1da177e4 1922
28a6d671
EB		 1923 retval = -ENOENT;
1924 if (!p)
1925 goto out_no_task;
1926 retval = 0;
28a6d671
EB		 1927
1928 fd = filp->f_pos;
1929 switch (fd) {
1930 case 0:
1931 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
5e442a49 1932 goto out;
28a6d671
EB		 1933 filp->f_pos++;
1934 case 1:
1935 ino = parent_ino(dentry);
1936 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
5e442a49 1937 goto out;
28a6d671
EB		 1938 filp->f_pos++;
1939 default:
1940 files = get_files_struct(p);
1941 if (!files)
5e442a49 1942 goto out;
28a6d671 1943 rcu_read_lock();
28a6d671 1944 for (fd = filp->f_pos-2;
9b4f526c 1945 fd < files_fdtable(files)->max_fds;
28a6d671 1946 fd++, filp->f_pos++) {
27932742
MS		 1947 char name[PROC_NUMBUF];
1948 int len;
f05ed3f1 1949 int rv;
28a6d671
EB		 1950
1951 if (!fcheck_files(files, fd))
1952 continue;
1953 rcu_read_unlock();
1954
27932742 1955 len = snprintf(name, sizeof(name), "%d", fd);
f05ed3f1
AD		 1956 rv = proc_fill_cache(filp, dirent, filldir,
1957 name, len, instantiate, p,
af5e6171 1958 (void *)(unsigned long)fd);
f05ed3f1
AD		 1959 if (rv < 0)
1960 goto out_fd_loop;
28a6d671
EB		 1961 rcu_read_lock();
1962 }
1963 rcu_read_unlock();
f05ed3f1 1964out_fd_loop:
28a6d671
EB		 1965 put_files_struct(files);
1966 }
1967out:
1968 put_task_struct(p);
1969out_no_task:
1970 return retval;
1971}
1972
27932742 1973static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
00cd8dd3 1974 unsigned int flags)
27932742
MS		 1975{
1976 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1977}
1978
1979static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1980{
1981 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1982}
1983
1984static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1985 size_t len, loff_t *ppos)
1986{
1987 char tmp[PROC_FDINFO_MAX];
3dcd25f3 1988 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
27932742
MS		 1989 if (!err)
1990 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1991 return err;
1992}
1993
1994static const struct file_operations proc_fdinfo_file_operations = {
87df8424 1995 .open = nonseekable_open,
27932742 1996 .read = proc_fdinfo_read,
6038f373 1997 .llseek = no_llseek,
27932742
MS		 1998};
1999
00977a59 2000static const struct file_operations proc_fd_operations = {
28a6d671
EB		 2001 .read = generic_read_dir,
2002 .readdir = proc_readfd,
6038f373 2003 .llseek = default_llseek,
1da177e4
LT		 2004};
2005
640708a2
PE		 2006#ifdef CONFIG_CHECKPOINT_RESTORE
2007
2008/*
2009 * dname_to_vma_addr - maps a dentry name into two unsigned longs
2010 * which represent vma start and end addresses.
2011 */
2012static int dname_to_vma_addr(struct dentry *dentry,
2013 unsigned long *start, unsigned long *end)
2014{
2015 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
2016 return -EINVAL;
2017
2018 return 0;
2019}
2020
0b728e19 2021static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
640708a2
PE		 2022{
2023 unsigned long vm_start, vm_end;
2024 bool exact_vma_exists = false;
2025 struct mm_struct *mm = NULL;
2026 struct task_struct *task;
2027 const struct cred *cred;
2028 struct inode *inode;
2029 int status = 0;
2030
0b728e19 2031 if (flags & LOOKUP_RCU)
640708a2
PE		 2032 return -ECHILD;
2033
2034 if (!capable(CAP_SYS_ADMIN)) {
2035 status = -EACCES;
2036 goto out_notask;
2037 }
2038
2039 inode = dentry->d_inode;
2040 task = get_proc_task(inode);
2041 if (!task)
2042 goto out_notask;
2043
2344bec7
CW		 2044 mm = mm_access(task, PTRACE_MODE_READ);
2045 if (IS_ERR_OR_NULL(mm))
640708a2
PE		 2046 goto out;
2047
2048 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
2049 down_read(&mm->mmap_sem);
2050 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
2051 up_read(&mm->mmap_sem);
2052 }
2053
2054 mmput(mm);
2055
2056 if (exact_vma_exists) {
2057 if (task_dumpable(task)) {
2058 rcu_read_lock();
2059 cred = __task_cred(task);
2060 inode->i_uid = cred->euid;
2061 inode->i_gid = cred->egid;
2062 rcu_read_unlock();
2063 } else {
dcb0f222
EB		 2064 inode->i_uid = GLOBAL_ROOT_UID;
2065 inode->i_gid = GLOBAL_ROOT_GID;
640708a2
PE		 2066 }
2067 security_task_to_inode(task, inode);
2068 status = 1;
2069 }
2070
2071out:
2072 put_task_struct(task);
2073
2074out_notask:
2075 if (status <= 0)
2076 d_drop(dentry);
2077
2078 return status;
2079}
2080
2081static const struct dentry_operations tid_map_files_dentry_operations = {
2082 .d_revalidate = map_files_d_revalidate,
2083 .d_delete = pid_delete_dentry,
2084};
2085
2086static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
2087{
2088 unsigned long vm_start, vm_end;
2089 struct vm_area_struct *vma;
2090 struct task_struct *task;
2091 struct mm_struct *mm;
2092 int rc;
2093
2094 rc = -ENOENT;
2095 task = get_proc_task(dentry->d_inode);
2096 if (!task)
2097 goto out;
2098
2099 mm = get_task_mm(task);
2100 put_task_struct(task);
2101 if (!mm)
2102 goto out;
2103
2104 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2105 if (rc)
2106 goto out_mmput;
2107
2108 down_read(&mm->mmap_sem);
2109 vma = find_exact_vma(mm, vm_start, vm_end);
2110 if (vma && vma->vm_file) {
2111 *path = vma->vm_file->f_path;
2112 path_get(path);
2113 rc = 0;
2114 }
2115 up_read(&mm->mmap_sem);
2116
2117out_mmput:
2118 mmput(mm);
2119out:
2120 return rc;
2121}
2122
2123struct map_files_info {
2124 struct file *file;
2125 unsigned long len;
2126 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
2127};
2128
2129static struct dentry *
2130proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
2131 struct task_struct *task, const void *ptr)
2132{
2133 const struct file *file = ptr;
2134 struct proc_inode *ei;
2135 struct inode *inode;
2136
2137 if (!file)
2138 return ERR_PTR(-ENOENT);
2139
2140 inode = proc_pid_make_inode(dir->i_sb, task);
2141 if (!inode)
2142 return ERR_PTR(-ENOENT);
2143
2144 ei = PROC_I(inode);
2145 ei->op.proc_get_link = proc_map_files_get_link;
2146
2147 inode->i_op = &proc_pid_link_inode_operations;
2148 inode->i_size = 64;
2149 inode->i_mode = S_IFLNK;
2150
2151 if (file->f_mode & FMODE_READ)
2152 inode->i_mode |= S_IRUSR;
2153 if (file->f_mode & FMODE_WRITE)
2154 inode->i_mode |= S_IWUSR;
2155
2156 d_set_d_op(dentry, &tid_map_files_dentry_operations);
2157 d_add(dentry, inode);
2158
2159 return NULL;
2160}
2161
2162static struct dentry *proc_map_files_lookup(struct inode *dir,
00cd8dd3 2163 struct dentry *dentry, unsigned int flags)
640708a2
PE		 2164{
2165 unsigned long vm_start, vm_end;
2166 struct vm_area_struct *vma;
2167 struct task_struct *task;
2168 struct dentry *result;
2169 struct mm_struct *mm;
2170
2171 result = ERR_PTR(-EACCES);
2172 if (!capable(CAP_SYS_ADMIN))
2173 goto out;
2174
2175 result = ERR_PTR(-ENOENT);
2176 task = get_proc_task(dir);
2177 if (!task)
2178 goto out;
2179
2180 result = ERR_PTR(-EACCES);
eb94cd96 2181 if (!ptrace_may_access(task, PTRACE_MODE_READ))
640708a2
PE		 2182 goto out_put_task;
2183
2184 result = ERR_PTR(-ENOENT);
2185 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
eb94cd96 2186 goto out_put_task;
640708a2
PE		 2187
2188 mm = get_task_mm(task);
2189 if (!mm)
eb94cd96 2190 goto out_put_task;
640708a2
PE		 2191
2192 down_read(&mm->mmap_sem);
2193 vma = find_exact_vma(mm, vm_start, vm_end);
2194 if (!vma)
2195 goto out_no_vma;
2196
2197 result = proc_map_files_instantiate(dir, dentry, task, vma->vm_file);
2198
2199out_no_vma:
2200 up_read(&mm->mmap_sem);
2201 mmput(mm);
640708a2
PE		 2202out_put_task:
2203 put_task_struct(task);
2204out:
2205 return result;
2206}
2207
2208static const struct inode_operations proc_map_files_inode_operations = {
2209 .lookup = proc_map_files_lookup,
2210 .permission = proc_fd_permission,
2211 .setattr = proc_setattr,
2212};
2213
2214static int
2215proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
2216{
2217 struct dentry *dentry = filp->f_path.dentry;
2218 struct inode *inode = dentry->d_inode;
2219 struct vm_area_struct *vma;
2220 struct task_struct *task;
2221 struct mm_struct *mm;
2222 ino_t ino;
2223 int ret;
2224
2225 ret = -EACCES;
2226 if (!capable(CAP_SYS_ADMIN))
2227 goto out;
2228
2229 ret = -ENOENT;
2230 task = get_proc_task(inode);
2231 if (!task)
2232 goto out;
2233
2234 ret = -EACCES;
eb94cd96 2235 if (!ptrace_may_access(task, PTRACE_MODE_READ))
640708a2
PE		 2236 goto out_put_task;
2237
2238 ret = 0;
2239 switch (filp->f_pos) {
2240 case 0:
2241 ino = inode->i_ino;
2242 if (filldir(dirent, ".", 1, 0, ino, DT_DIR) < 0)
eb94cd96 2243 goto out_put_task;
640708a2
PE		 2244 filp->f_pos++;
2245 case 1:
2246 ino = parent_ino(dentry);
2247 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
eb94cd96 2248 goto out_put_task;
640708a2
PE		 2249 filp->f_pos++;
2250 default:
2251 {
2252 unsigned long nr_files, pos, i;
2253 struct flex_array *fa = NULL;
2254 struct map_files_info info;
2255 struct map_files_info *p;
2256
2257 mm = get_task_mm(task);
2258 if (!mm)
eb94cd96 2259 goto out_put_task;
640708a2
PE		 2260 down_read(&mm->mmap_sem);
2261
2262 nr_files = 0;
2263
2264 /*
2265 * We need two passes here:
2266 *
2267 * 1) Collect vmas of mapped files with mmap_sem taken
2268 * 2) Release mmap_sem and instantiate entries
2269 *
2270 * otherwise we get lockdep complained, since filldir()
2271 * routine might require mmap_sem taken in might_fault().
2272 */
2273
2274 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2275 if (vma->vm_file && ++pos > filp->f_pos)
2276 nr_files++;
2277 }
2278
2279 if (nr_files) {
2280 fa = flex_array_alloc(sizeof(info), nr_files,
2281 GFP_KERNEL);
2282 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2283 GFP_KERNEL)) {
2284 ret = -ENOMEM;
2285 if (fa)
2286 flex_array_free(fa);
2287 up_read(&mm->mmap_sem);
2288 mmput(mm);
eb94cd96 2289 goto out_put_task;
640708a2
PE		 2290 }
2291 for (i = 0, vma = mm->mmap, pos = 2; vma;
2292 vma = vma->vm_next) {
2293 if (!vma->vm_file)
2294 continue;
2295 if (++pos <= filp->f_pos)
2296 continue;
2297
2298 get_file(vma->vm_file);
2299 info.file = vma->vm_file;
2300 info.len = snprintf(info.name,
2301 sizeof(info.name), "%lx-%lx",
2302 vma->vm_start, vma->vm_end);
2303 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2304 BUG();
2305 }
2306 }
2307 up_read(&mm->mmap_sem);
2308
2309 for (i = 0; i < nr_files; i++) {
2310 p = flex_array_get(fa, i);
2311 ret = proc_fill_cache(filp, dirent, filldir,
2312 p->name, p->len,
2313 proc_map_files_instantiate,
2314 task, p->file);
2315 if (ret)
2316 break;
2317 filp->f_pos++;
2318 fput(p->file);
2319 }
2320 for (; i < nr_files; i++) {
2321 /*
2322 * In case of error don't forget
2323 * to put rest of file refs.
2324 */
2325 p = flex_array_get(fa, i);
2326 fput(p->file);
2327 }
2328 if (fa)
2329 flex_array_free(fa);
2330 mmput(mm);
2331 }
2332 }
2333
640708a2
PE		 2334out_put_task:
2335 put_task_struct(task);
2336out:
2337 return ret;
2338}
2339
2340static const struct file_operations proc_map_files_operations = {
2341 .read = generic_read_dir,
2342 .readdir = proc_map_files_readdir,
2343 .llseek = default_llseek,
2344};
2345
2346#endif /* CONFIG_CHECKPOINT_RESTORE */
2347
8948e11f
AD		 2348/*
2349 * /proc/pid/fd needs a special permission handler so that a process can still
2350 * access /proc/self/fd after it has executed a setuid().
2351 */
10556cb2 2352static int proc_fd_permission(struct inode *inode, int mask)
8948e11f 2353{
2830ba7f 2354 int rv = generic_permission(inode, mask);
8948e11f
AD		 2355 if (rv == 0)
2356 return 0;
2357 if (task_pid(current) == proc_pid(inode))
2358 rv = 0;
2359 return rv;
2360}
2361
1da177e4
LT		 2362/*
2363 * proc directories can do almost nothing..
2364 */
c5ef1c42 2365static const struct inode_operations proc_fd_inode_operations = {
1da177e4 2366 .lookup = proc_lookupfd,
8948e11f 2367 .permission = proc_fd_permission,
6d76fa58 2368 .setattr = proc_setattr,
1da177e4
LT		 2369};
2370
27932742
MS		 2371static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2372 struct dentry *dentry, struct task_struct *task, const void *ptr)
2373{
af5e6171 2374 unsigned fd = (unsigned long)ptr;
27932742
MS		 2375 struct inode *inode;
2376 struct proc_inode *ei;
2377 struct dentry *error = ERR_PTR(-ENOENT);
2378
2379 inode = proc_pid_make_inode(dir->i_sb, task);
2380 if (!inode)
2381 goto out;
2382 ei = PROC_I(inode);
2383 ei->fd = fd;
2384 inode->i_mode = S_IFREG | S_IRUSR;
2385 inode->i_fop = &proc_fdinfo_file_operations;
fb045adb 2386 d_set_d_op(dentry, &tid_fd_dentry_operations);
27932742
MS		 2387 d_add(dentry, inode);
2388 /* Close the race of the process dying before we return the dentry */
0b728e19 2389 if (tid_fd_revalidate(dentry, 0))
27932742
MS		 2390 error = NULL;
2391
2392 out:
2393 return error;
2394}
2395
2396static struct dentry *proc_lookupfdinfo(struct inode *dir,
2397 struct dentry *dentry,
00cd8dd3 2398 unsigned int flags)
27932742
MS		 2399{
2400 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2401}
2402
2403static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2404{
2405 return proc_readfd_common(filp, dirent, filldir,
2406 proc_fdinfo_instantiate);
2407}
2408
2409static const struct file_operations proc_fdinfo_operations = {
2410 .read = generic_read_dir,
2411 .readdir = proc_readfdinfo,
6038f373 2412 .llseek = default_llseek,
27932742
MS		 2413};
2414
2415/*
2416 * proc directories can do almost nothing..
2417 */
2418static const struct inode_operations proc_fdinfo_inode_operations = {
2419 .lookup = proc_lookupfdinfo,
2420 .setattr = proc_setattr,
2421};
2422
2423
444ceed8 2424static struct dentry *proc_pident_instantiate(struct inode *dir,
c5141e6d 2425 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8 2426{
c5141e6d 2427 const struct pid_entry *p = ptr;
444ceed8
EB		 2428 struct inode *inode;
2429 struct proc_inode *ei;
bd6daba9 2430 struct dentry *error = ERR_PTR(-ENOENT);
444ceed8 2431
61a28784 2432 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 2433 if (!inode)
2434 goto out;
2435
2436 ei = PROC_I(inode);
2437 inode->i_mode = p->mode;
2438 if (S_ISDIR(inode->i_mode))
bfe86848 2439 set_nlink(inode, 2); /* Use getattr to fix if necessary */
444ceed8
EB		 2440 if (p->iop)
2441 inode->i_op = p->iop;
2442 if (p->fop)
2443 inode->i_fop = p->fop;
2444 ei->op = p->op;
fb045adb 2445 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 2446 d_add(dentry, inode);
2447 /* Close the race of the process dying before we return the dentry */
0b728e19 2448 if (pid_revalidate(dentry, 0))
444ceed8
EB		 2449 error = NULL;
2450out:
2451 return error;
2452}
2453
1da177e4
LT		 2454static struct dentry *proc_pident_lookup(struct inode *dir,
2455 struct dentry *dentry,
c5141e6d 2456 const struct pid_entry *ents,
7bcd6b0e 2457 unsigned int nents)
1da177e4 2458{
cd6a3ce9 2459 struct dentry *error;
99f89551 2460 struct task_struct *task = get_proc_task(dir);
c5141e6d 2461 const struct pid_entry *p, *last;
1da177e4 2462
cd6a3ce9 2463 error = ERR_PTR(-ENOENT);
1da177e4 2464
99f89551
EB		 2465 if (!task)
2466 goto out_no_task;
1da177e4 2467
20cdc894
EB		 2468 /*
2469 * Yes, it does not scale. And it should not. Don't add
2470 * new entries into /proc/<tgid>/ without very good reasons.
2471 */
7bcd6b0e
EB		 2472 last = &ents[nents - 1];
2473 for (p = ents; p <= last; p++) {
1da177e4
LT		 2474 if (p->len != dentry->d_name.len)
2475 continue;
2476 if (!memcmp(dentry->d_name.name, p->name, p->len))
2477 break;
2478 }
7bcd6b0e 2479 if (p > last)
1da177e4
LT		 2480 goto out;
2481
444ceed8 2482 error = proc_pident_instantiate(dir, dentry, task, p);
1da177e4 2483out:
99f89551
EB		 2484 put_task_struct(task);
2485out_no_task:
cd6a3ce9 2486 return error;
1da177e4
LT		 2487}
2488
c5141e6d
ED		 2489static int proc_pident_fill_cache(struct file *filp, void *dirent,
2490 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
61a28784
EB		 2491{
2492 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2493 proc_pident_instantiate, task, p);
2494}
2495
28a6d671
EB		 2496static int proc_pident_readdir(struct file *filp,
2497 void *dirent, filldir_t filldir,
c5141e6d 2498 const struct pid_entry *ents, unsigned int nents)
28a6d671
EB		 2499{
2500 int i;
2fddfeef 2501 struct dentry *dentry = filp->f_path.dentry;
28a6d671
EB		 2502 struct inode *inode = dentry->d_inode;
2503 struct task_struct *task = get_proc_task(inode);
c5141e6d 2504 const struct pid_entry *p, *last;
28a6d671
EB		 2505 ino_t ino;
2506 int ret;
2507
2508 ret = -ENOENT;
2509 if (!task)
61a28784 2510 goto out_no_task;
28a6d671
EB		 2511
2512 ret = 0;
28a6d671
EB		 2513 i = filp->f_pos;
2514 switch (i) {
2515 case 0:
2516 ino = inode->i_ino;
2517 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2518 goto out;
2519 i++;
2520 filp->f_pos++;
2521 /* fall through */
2522 case 1:
2523 ino = parent_ino(dentry);
2524 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2525 goto out;
2526 i++;
2527 filp->f_pos++;
2528 /* fall through */
2529 default:
2530 i -= 2;
2531 if (i >= nents) {
2532 ret = 1;
2533 goto out;
2534 }
2535 p = ents + i;
7bcd6b0e
EB		 2536 last = &ents[nents - 1];
2537 while (p <= last) {
61a28784 2538 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
28a6d671
EB		 2539 goto out;
2540 filp->f_pos++;
2541 p++;
2542 }
2543 }
2544
2545 ret = 1;
2546out:
61a28784
EB		 2547 put_task_struct(task);
2548out_no_task:
28a6d671 2549 return ret;
1da177e4
LT		 2550}
2551
28a6d671
EB		 2552#ifdef CONFIG_SECURITY
2553static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2554 size_t count, loff_t *ppos)
2555{
2fddfeef 2556 struct inode * inode = file->f_path.dentry->d_inode;
04ff9708 2557 char *p = NULL;
28a6d671
EB		 2558 ssize_t length;
2559 struct task_struct *task = get_proc_task(inode);
2560
28a6d671 2561 if (!task)
04ff9708 2562 return -ESRCH;
28a6d671
EB		 2563
2564 length = security_getprocattr(task,
2fddfeef 2565 (char*)file->f_path.dentry->d_name.name,
04ff9708 2566 &p);
28a6d671 2567 put_task_struct(task);
04ff9708
AV		 2568 if (length > 0)
2569 length = simple_read_from_buffer(buf, count, ppos, p, length);
2570 kfree(p);
28a6d671 2571 return length;
1da177e4
LT		 2572}
2573
28a6d671
EB		 2574static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2575 size_t count, loff_t *ppos)
2576{
2fddfeef 2577 struct inode * inode = file->f_path.dentry->d_inode;
28a6d671
EB		 2578 char *page;
2579 ssize_t length;
2580 struct task_struct *task = get_proc_task(inode);
2581
2582 length = -ESRCH;
2583 if (!task)
2584 goto out_no_task;
2585 if (count > PAGE_SIZE)
2586 count = PAGE_SIZE;
2587
2588 /* No partial writes. */
2589 length = -EINVAL;
2590 if (*ppos != 0)
2591 goto out;
2592
2593 length = -ENOMEM;
e12ba74d 2594 page = (char*)__get_free_page(GFP_TEMPORARY);
28a6d671
EB		 2595 if (!page)
2596 goto out;
2597
2598 length = -EFAULT;
2599 if (copy_from_user(page, buf, count))
2600 goto out_free;
2601
107db7c7 2602 /* Guard against adverse ptrace interaction */
9b1bf12d 2603 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
107db7c7
DH		 2604 if (length < 0)
2605 goto out_free;
2606
28a6d671 2607 length = security_setprocattr(task,
2fddfeef 2608 (char*)file->f_path.dentry->d_name.name,
28a6d671 2609 (void*)page, count);
9b1bf12d 2610 mutex_unlock(&task->signal->cred_guard_mutex);
28a6d671
EB		 2611out_free:
2612 free_page((unsigned long) page);
2613out:
2614 put_task_struct(task);
2615out_no_task:
2616 return length;
2617}
2618
00977a59 2619static const struct file_operations proc_pid_attr_operations = {
28a6d671
EB		 2620 .read = proc_pid_attr_read,
2621 .write = proc_pid_attr_write,
87df8424 2622 .llseek = generic_file_llseek,
28a6d671
EB		 2623};
2624
c5141e6d 2625static const struct pid_entry attr_dir_stuff[] = {
631f9c18
AD		 2626 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2627 REG("prev", S_IRUGO, proc_pid_attr_operations),
2628 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2629 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2630 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2631 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
28a6d671
EB		 2632};
2633
72d9dcfc 2634static int proc_attr_dir_readdir(struct file * filp,
28a6d671
EB		 2635 void * dirent, filldir_t filldir)
2636{
2637 return proc_pident_readdir(filp,dirent,filldir,
72d9dcfc 2638 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2639}
2640
00977a59 2641static const struct file_operations proc_attr_dir_operations = {
1da177e4 2642 .read = generic_read_dir,
72d9dcfc 2643 .readdir = proc_attr_dir_readdir,
6038f373 2644 .llseek = default_llseek,
1da177e4
LT		 2645};
2646
72d9dcfc 2647static struct dentry *proc_attr_dir_lookup(struct inode *dir,
00cd8dd3 2648 struct dentry *dentry, unsigned int flags)
28a6d671 2649{
7bcd6b0e
EB		 2650 return proc_pident_lookup(dir, dentry,
2651 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2652}
2653
c5ef1c42 2654static const struct inode_operations proc_attr_dir_inode_operations = {
72d9dcfc 2655 .lookup = proc_attr_dir_lookup,
99f89551 2656 .getattr = pid_getattr,
6d76fa58 2657 .setattr = proc_setattr,
1da177e4
LT		 2658};
2659
28a6d671
EB		 2660#endif
2661
698ba7b5 2662#ifdef CONFIG_ELF_CORE
3cb4a0bb
KH		 2663static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2664 size_t count, loff_t *ppos)
2665{
2666 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2667 struct mm_struct *mm;
2668 char buffer[PROC_NUMBUF];
2669 size_t len;
2670 int ret;
2671
2672 if (!task)
2673 return -ESRCH;
2674
2675 ret = 0;
2676 mm = get_task_mm(task);
2677 if (mm) {
2678 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2679 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2680 MMF_DUMP_FILTER_SHIFT));
2681 mmput(mm);
2682 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2683 }
2684
2685 put_task_struct(task);
2686
2687 return ret;
2688}
2689
2690static ssize_t proc_coredump_filter_write(struct file *file,
2691 const char __user *buf,
2692 size_t count,
2693 loff_t *ppos)
2694{
2695 struct task_struct *task;
2696 struct mm_struct *mm;
2697 char buffer[PROC_NUMBUF], *end;
2698 unsigned int val;
2699 int ret;
2700 int i;
2701 unsigned long mask;
2702
2703 ret = -EFAULT;
2704 memset(buffer, 0, sizeof(buffer));
2705 if (count > sizeof(buffer) - 1)
2706 count = sizeof(buffer) - 1;
2707 if (copy_from_user(buffer, buf, count))
2708 goto out_no_task;
2709
2710 ret = -EINVAL;
2711 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2712 if (*end == '\n')
2713 end++;
2714 if (end - buffer == 0)
2715 goto out_no_task;
2716
2717 ret = -ESRCH;
2718 task = get_proc_task(file->f_dentry->d_inode);
2719 if (!task)
2720 goto out_no_task;
2721
2722 ret = end - buffer;
2723 mm = get_task_mm(task);
2724 if (!mm)
2725 goto out_no_mm;
2726
2727 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2728 if (val & mask)
2729 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2730 else
2731 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2732 }
2733
2734 mmput(mm);
2735 out_no_mm:
2736 put_task_struct(task);
2737 out_no_task:
2738 return ret;
2739}
2740
2741static const struct file_operations proc_coredump_filter_operations = {
2742 .read = proc_coredump_filter_read,
2743 .write = proc_coredump_filter_write,
87df8424 2744 .llseek = generic_file_llseek,
3cb4a0bb
KH		 2745};
2746#endif
2747
28a6d671
EB		 2748/*
2749 * /proc/self:
2750 */
2751static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2752 int buflen)
2753{
488e5bc4 2754 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
b55fcb22 2755 pid_t tgid = task_tgid_nr_ns(current, ns);
28a6d671 2756 char tmp[PROC_NUMBUF];
b55fcb22 2757 if (!tgid)
488e5bc4 2758 return -ENOENT;
b55fcb22 2759 sprintf(tmp, "%d", tgid);
28a6d671
EB		 2760 return vfs_readlink(dentry,buffer,buflen,tmp);
2761}
2762
2763static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2764{
488e5bc4 2765 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
b55fcb22 2766 pid_t tgid = task_tgid_nr_ns(current, ns);
7fee4868
AV		 2767 char *name = ERR_PTR(-ENOENT);
2768 if (tgid) {
2769 name = __getname();
2770 if (!name)
2771 name = ERR_PTR(-ENOMEM);
2772 else
2773 sprintf(name, "%d", tgid);
2774 }
2775 nd_set_link(nd, name);
2776 return NULL;
2777}
2778
2779static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2780 void *cookie)
2781{
2782 char *s = nd_get_link(nd);
2783 if (!IS_ERR(s))
2784 __putname(s);
28a6d671
EB		 2785}
2786
c5ef1c42 2787static const struct inode_operations proc_self_inode_operations = {
28a6d671
EB		 2788 .readlink = proc_self_readlink,
2789 .follow_link = proc_self_follow_link,
7fee4868 2790 .put_link = proc_self_put_link,
28a6d671
EB		 2791};
2792
801199ce
EB		 2793/*
2794 * proc base
2795 *
2796 * These are the directory entries in the root directory of /proc
2797 * that properly belong to the /proc filesystem, as they describe
2798 * describe something that is process related.
2799 */
c5141e6d 2800static const struct pid_entry proc_base_stuff[] = {
61a28784 2801 NOD("self", S_IFLNK|S_IRWXUGO,
801199ce 2802 &proc_self_inode_operations, NULL, {}),
801199ce
EB		 2803};
2804
444ceed8 2805static struct dentry *proc_base_instantiate(struct inode *dir,
c5141e6d 2806 struct dentry *dentry, struct task_struct *task, const void *ptr)
801199ce 2807{
c5141e6d 2808 const struct pid_entry *p = ptr;
801199ce 2809 struct inode *inode;
801199ce 2810 struct proc_inode *ei;
73d36460 2811 struct dentry *error;
801199ce
EB		 2812
2813 /* Allocate the inode */
2814 error = ERR_PTR(-ENOMEM);
2815 inode = new_inode(dir->i_sb);
2816 if (!inode)
2817 goto out;
2818
2819 /* Initialize the inode */
2820 ei = PROC_I(inode);
85fe4025 2821 inode->i_ino = get_next_ino();
801199ce 2822 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
801199ce
EB		 2823
2824 /*
2825 * grab the reference to the task.
2826 */
1a657f78 2827 ei->pid = get_task_pid(task, PIDTYPE_PID);
801199ce
EB		 2828 if (!ei->pid)
2829 goto out_iput;
2830
801199ce
EB		 2831 inode->i_mode = p->mode;
2832 if (S_ISDIR(inode->i_mode))
bfe86848 2833 set_nlink(inode, 2);
801199ce
EB		 2834 if (S_ISLNK(inode->i_mode))
2835 inode->i_size = 64;
2836 if (p->iop)
2837 inode->i_op = p->iop;
2838 if (p->fop)
2839 inode->i_fop = p->fop;
2840 ei->op = p->op;
801199ce
EB		 2841 d_add(dentry, inode);
2842 error = NULL;
2843out:
801199ce
EB		 2844 return error;
2845out_iput:
2846 iput(inode);
2847 goto out;
2848}
2849
444ceed8
EB		 2850static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2851{
2852 struct dentry *error;
2853 struct task_struct *task = get_proc_task(dir);
c5141e6d 2854 const struct pid_entry *p, *last;
444ceed8
EB		 2855
2856 error = ERR_PTR(-ENOENT);
2857
2858 if (!task)
2859 goto out_no_task;
2860
2861 /* Lookup the directory entry */
7bcd6b0e
EB		 2862 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2863 for (p = proc_base_stuff; p <= last; p++) {
444ceed8
EB		 2864 if (p->len != dentry->d_name.len)
2865 continue;
2866 if (!memcmp(dentry->d_name.name, p->name, p->len))
2867 break;
2868 }
7bcd6b0e 2869 if (p > last)
444ceed8
EB		 2870 goto out;
2871
2872 error = proc_base_instantiate(dir, dentry, task, p);
2873
2874out:
2875 put_task_struct(task);
2876out_no_task:
2877 return error;
2878}
2879
c5141e6d
ED		 2880static int proc_base_fill_cache(struct file *filp, void *dirent,
2881 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
61a28784
EB		 2882{
2883 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2884 proc_base_instantiate, task, p);
2885}
2886
aba76fdb 2887#ifdef CONFIG_TASK_IO_ACCOUNTING
297c5d92
AR		 2888static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2889{
940389b8 2890 struct task_io_accounting acct = task->ioac;
5995477a 2891 unsigned long flags;
293eb1e7 2892 int result;
5995477a 2893
293eb1e7
VK		 2894 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2895 if (result)
2896 return result;
2897
2898 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2899 result = -EACCES;
2900 goto out_unlock;
2901 }
1d1221f3 2902
5995477a
AR		 2903 if (whole && lock_task_sighand(task, &flags)) {
2904 struct task_struct *t = task;
2905
2906 task_io_accounting_add(&acct, &task->signal->ioac);
2907 while_each_thread(task, t)
2908 task_io_accounting_add(&acct, &t->ioac);
2909
2910 unlock_task_sighand(task, &flags);
297c5d92 2911 }
293eb1e7 2912 result = sprintf(buffer,
aba76fdb
AM		 2913 "rchar: %llu\n"
2914 "wchar: %llu\n"
2915 "syscr: %llu\n"
2916 "syscw: %llu\n"
2917 "read_bytes: %llu\n"
2918 "write_bytes: %llu\n"
2919 "cancelled_write_bytes: %llu\n",
7c44319d
AB		 2920 (unsigned long long)acct.rchar,
2921 (unsigned long long)acct.wchar,
2922 (unsigned long long)acct.syscr,
2923 (unsigned long long)acct.syscw,
2924 (unsigned long long)acct.read_bytes,
2925 (unsigned long long)acct.write_bytes,
2926 (unsigned long long)acct.cancelled_write_bytes);
293eb1e7
VK		 2927out_unlock:
2928 mutex_unlock(&task->signal->cred_guard_mutex);
2929 return result;
297c5d92
AR		 2930}
2931
2932static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2933{
2934 return do_io_accounting(task, buffer, 0);
aba76fdb 2935}
297c5d92
AR		 2936
2937static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2938{
2939 return do_io_accounting(task, buffer, 1);
2940}
2941#endif /* CONFIG_TASK_IO_ACCOUNTING */
aba76fdb 2942
22d917d8
EB		 2943#ifdef CONFIG_USER_NS
2944static int proc_id_map_open(struct inode *inode, struct file *file,
2945 struct seq_operations *seq_ops)
2946{
2947 struct user_namespace *ns = NULL;
2948 struct task_struct *task;
2949 struct seq_file *seq;
2950 int ret = -EINVAL;
2951
2952 task = get_proc_task(inode);
2953 if (task) {
2954 rcu_read_lock();
2955 ns = get_user_ns(task_cred_xxx(task, user_ns));
2956 rcu_read_unlock();
2957 put_task_struct(task);
2958 }
2959 if (!ns)
2960 goto err;
2961
2962 ret = seq_open(file, seq_ops);
2963 if (ret)
2964 goto err_put_ns;
2965
2966 seq = file->private_data;
2967 seq->private = ns;
2968
2969 return 0;
2970err_put_ns:
2971 put_user_ns(ns);
2972err:
2973 return ret;
2974}
2975
2976static int proc_id_map_release(struct inode *inode, struct file *file)
2977{
2978 struct seq_file *seq = file->private_data;
2979 struct user_namespace *ns = seq->private;
2980 put_user_ns(ns);
2981 return seq_release(inode, file);
2982}
2983
2984static int proc_uid_map_open(struct inode *inode, struct file *file)
2985{
2986 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2987}
2988
2989static int proc_gid_map_open(struct inode *inode, struct file *file)
2990{
2991 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2992}
2993
2994static const struct file_operations proc_uid_map_operations = {
2995 .open = proc_uid_map_open,
2996 .write = proc_uid_map_write,
2997 .read = seq_read,
2998 .llseek = seq_lseek,
2999 .release = proc_id_map_release,
3000};
3001
3002static const struct file_operations proc_gid_map_operations = {
3003 .open = proc_gid_map_open,
3004 .write = proc_gid_map_write,
3005 .read = seq_read,
3006 .llseek = seq_lseek,
3007 .release = proc_id_map_release,
3008};
3009#endif /* CONFIG_USER_NS */
3010
47830723
KC		 3011static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
3012 struct pid *pid, struct task_struct *task)
3013{
a9712bc1
AV		 3014 int err = lock_trace(task);
3015 if (!err) {
3016 seq_printf(m, "%08x\n", task->personality);
3017 unlock_trace(task);
3018 }
3019 return err;
47830723
KC		 3020}
3021
28a6d671
EB		 3022/*
3023 * Thread groups
3024 */
00977a59 3025static const struct file_operations proc_task_operations;
c5ef1c42 3026static const struct inode_operations proc_task_inode_operations;
20cdc894 3027
c5141e6d 3028static const struct pid_entry tgid_base_stuff[] = {
631f9c18
AD		 3029 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
3030 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
640708a2
PE		 3031#ifdef CONFIG_CHECKPOINT_RESTORE
3032 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
3033#endif
631f9c18 3034 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 3035 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
b2211a36 3036#ifdef CONFIG_NET
631f9c18 3037 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
b2211a36 3038#endif
631f9c18
AD		 3039 REG("environ", S_IRUSR, proc_environ_operations),
3040 INF("auxv", S_IRUSR, proc_pid_auxv),
3041 ONE("status", S_IRUGO, proc_pid_status),
a9712bc1 3042 ONE("personality", S_IRUGO, proc_pid_personality),
3036e7b4 3043 INF("limits", S_IRUGO, proc_pid_limits),
43ae34cb 3044#ifdef CONFIG_SCHED_DEBUG
631f9c18 3045 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
5091faa4
MG		 3046#endif
3047#ifdef CONFIG_SCHED_AUTOGROUP
3048 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
ebcb6734 3049#endif
4614a696 3050 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 3051#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
a9712bc1 3052 INF("syscall", S_IRUGO, proc_pid_syscall),
43ae34cb 3053#endif
631f9c18
AD		 3054 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3055 ONE("stat", S_IRUGO, proc_tgid_stat),
3056 ONE("statm", S_IRUGO, proc_pid_statm),
b7643757 3057 REG("maps", S_IRUGO, proc_pid_maps_operations),
28a6d671 3058#ifdef CONFIG_NUMA
b7643757 3059 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
28a6d671 3060#endif
631f9c18
AD		 3061 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3062 LNK("cwd", proc_cwd_link),
3063 LNK("root", proc_root_link),
3064 LNK("exe", proc_exe_link),
3065 REG("mounts", S_IRUGO, proc_mounts_operations),
3066 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3067 REG("mountstats", S_IRUSR, proc_mountstats_operations),
1e883281 3068#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18 3069 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
b7643757 3070 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
ca6b0bf0 3071 REG("pagemap", S_IRUGO, proc_pagemap_operations),
28a6d671
EB		 3072#endif
3073#ifdef CONFIG_SECURITY
631f9c18 3074 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 3075#endif
3076#ifdef CONFIG_KALLSYMS
631f9c18 3077 INF("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 3078#endif
2ec220e2 3079#ifdef CONFIG_STACKTRACE
a9712bc1 3080 ONE("stack", S_IRUGO, proc_pid_stack),
28a6d671
EB		 3081#endif
3082#ifdef CONFIG_SCHEDSTATS
631f9c18 3083 INF("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 3084#endif
9745512c 3085#ifdef CONFIG_LATENCYTOP
631f9c18 3086 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 3087#endif
8793d854 3088#ifdef CONFIG_PROC_PID_CPUSET
631f9c18 3089 REG("cpuset", S_IRUGO, proc_cpuset_operations),
a424316c
PM		 3090#endif
3091#ifdef CONFIG_CGROUPS
631f9c18 3092 REG("cgroup", S_IRUGO, proc_cgroup_operations),
28a6d671 3093#endif
631f9c18
AD		 3094 INF("oom_score", S_IRUGO, proc_oom_score),
3095 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
a63d83f4 3096 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 3097#ifdef CONFIG_AUDITSYSCALL
631f9c18
AD		 3098 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3099 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 3100#endif
f4f154fd 3101#ifdef CONFIG_FAULT_INJECTION
631f9c18 3102 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 3103#endif
698ba7b5 3104#ifdef CONFIG_ELF_CORE
631f9c18 3105 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3cb4a0bb 3106#endif
aba76fdb 3107#ifdef CONFIG_TASK_IO_ACCOUNTING
1d1221f3 3108 INF("io", S_IRUSR, proc_tgid_io_accounting),
aba76fdb 3109#endif
f133ecca
CM		 3110#ifdef CONFIG_HARDWALL
3111 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3112#endif
22d917d8
EB		 3113#ifdef CONFIG_USER_NS
3114 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3115 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3116#endif
28a6d671 3117};
1da177e4 3118
28a6d671 3119static int proc_tgid_base_readdir(struct file * filp,
1da177e4
LT		 3120 void * dirent, filldir_t filldir)
3121{
3122 return proc_pident_readdir(filp,dirent,filldir,
28a6d671 3123 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 3124}
3125
00977a59 3126static const struct file_operations proc_tgid_base_operations = {
1da177e4 3127 .read = generic_read_dir,
28a6d671 3128 .readdir = proc_tgid_base_readdir,
6038f373 3129 .llseek = default_llseek,
1da177e4
LT		 3130};
3131
00cd8dd3
AV		 3132static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3133{
7bcd6b0e
EB		 3134 return proc_pident_lookup(dir, dentry,
3135 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 3136}
3137
c5ef1c42 3138static const struct inode_operations proc_tgid_base_inode_operations = {
28a6d671 3139 .lookup = proc_tgid_base_lookup,
99f89551 3140 .getattr = pid_getattr,
6d76fa58 3141 .setattr = proc_setattr,
0499680a 3142 .permission = proc_pid_permission,
1da177e4 3143};
1da177e4 3144
60347f67 3145static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
1da177e4 3146{
48e6484d 3147 struct dentry *dentry, *leader, *dir;
8578cea7 3148 char buf[PROC_NUMBUF];
48e6484d
EB		 3149 struct qstr name;
3150
3151 name.name = buf;
60347f67
PE		 3152 name.len = snprintf(buf, sizeof(buf), "%d", pid);
3153 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d 3154 if (dentry) {
29f12ca3 3155 shrink_dcache_parent(dentry);
48e6484d
EB		 3156 d_drop(dentry);
3157 dput(dentry);
3158 }
1da177e4 3159
48e6484d 3160 name.name = buf;
60347f67
PE		 3161 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
3162 leader = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d
EB		 3163 if (!leader)
3164 goto out;
1da177e4 3165
48e6484d
EB		 3166 name.name = "task";
3167 name.len = strlen(name.name);
3168 dir = d_hash_and_lookup(leader, &name);
3169 if (!dir)
3170 goto out_put_leader;
3171
3172 name.name = buf;
60347f67 3173 name.len = snprintf(buf, sizeof(buf), "%d", pid);
48e6484d
EB		 3174 dentry = d_hash_and_lookup(dir, &name);
3175 if (dentry) {
3176 shrink_dcache_parent(dentry);
3177 d_drop(dentry);
3178 dput(dentry);
1da177e4 3179 }
48e6484d
EB		 3180
3181 dput(dir);
3182out_put_leader:
3183 dput(leader);
3184out:
3185 return;
1da177e4
LT		 3186}
3187
0895e91d
RD		 3188/**
3189 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3190 * @task: task that should be flushed.
3191 *
3192 * When flushing dentries from proc, one needs to flush them from global
60347f67 3193 * proc (proc_mnt) and from all the namespaces' procs this task was seen
0895e91d
RD		 3194 * in. This call is supposed to do all of this job.
3195 *
3196 * Looks in the dcache for
3197 * /proc/@pid
3198 * /proc/@tgid/task/@pid
3199 * if either directory is present flushes it and all of it'ts children
3200 * from the dcache.
3201 *
3202 * It is safe and reasonable to cache /proc entries for a task until
3203 * that task exits. After that they just clog up the dcache with
3204 * useless entries, possibly causing useful dcache entries to be
3205 * flushed instead. This routine is proved to flush those useless
3206 * dcache entries at process exit time.
3207 *
3208 * NOTE: This routine is just an optimization so it does not guarantee
3209 * that no dcache entries will exist at process exit time it
3210 * just makes it very unlikely that any will persist.
60347f67
PE		 3211 */
3212
3213void proc_flush_task(struct task_struct *task)
3214{
9fcc2d15 3215 int i;
9b4d1cbe 3216 struct pid *pid, *tgid;
130f77ec
PE		 3217 struct upid *upid;
3218
130f77ec 3219 pid = task_pid(task);
9b4d1cbe 3220 tgid = task_tgid(task);
130f77ec 3221
9fcc2d15 3222 for (i = 0; i <= pid->level; i++) {
130f77ec
PE		 3223 upid = &pid->numbers[i];
3224 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
9b4d1cbe 3225 tgid->numbers[i].nr);
130f77ec 3226 }
6f4e6433
PE		 3227
3228 upid = &pid->numbers[pid->level];
3229 if (upid->nr == 1)
3230 pid_ns_release_proc(upid->ns);
60347f67
PE		 3231}
3232
9711ef99
AB		 3233static struct dentry *proc_pid_instantiate(struct inode *dir,
3234 struct dentry * dentry,
c5141e6d 3235 struct task_struct *task, const void *ptr)
444ceed8
EB		 3236{
3237 struct dentry *error = ERR_PTR(-ENOENT);
3238 struct inode *inode;
3239
61a28784 3240 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 3241 if (!inode)
3242 goto out;
3243
3244 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3245 inode->i_op = &proc_tgid_base_inode_operations;
3246 inode->i_fop = &proc_tgid_base_operations;
3247 inode->i_flags|=S_IMMUTABLE;
aed54175 3248
bfe86848
MS		 3249 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
3250 ARRAY_SIZE(tgid_base_stuff)));
444ceed8 3251
fb045adb 3252 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 3253
3254 d_add(dentry, inode);
3255 /* Close the race of the process dying before we return the dentry */
0b728e19 3256 if (pid_revalidate(dentry, 0))
444ceed8
EB		 3257 error = NULL;
3258out:
3259 return error;
3260}
3261
00cd8dd3 3262struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
1da177e4 3263{
73d36460 3264 struct dentry *result;
1da177e4 3265 struct task_struct *task;
1da177e4 3266 unsigned tgid;
b488893a 3267 struct pid_namespace *ns;
1da177e4 3268
801199ce
EB		 3269 result = proc_base_lookup(dir, dentry);
3270 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3271 goto out;
3272
1da177e4
LT		 3273 tgid = name_to_int(dentry);
3274 if (tgid == ~0U)
3275 goto out;
3276
b488893a 3277 ns = dentry->d_sb->s_fs_info;
de758734 3278 rcu_read_lock();
b488893a 3279 task = find_task_by_pid_ns(tgid, ns);
1da177e4
LT		 3280 if (task)
3281 get_task_struct(task);
de758734 3282 rcu_read_unlock();
1da177e4
LT		 3283 if (!task)
3284 goto out;
3285
444ceed8 3286 result = proc_pid_instantiate(dir, dentry, task, NULL);
1da177e4 3287 put_task_struct(task);
1da177e4 3288out:
cd6a3ce9 3289 return result;
1da177e4
LT		 3290}
3291
1da177e4 3292/*
0804ef4b 3293 * Find the first task with tgid >= tgid
0bc58a91 3294 *
1da177e4 3295 */
19fd4bb2
EB		 3296struct tgid_iter {
3297 unsigned int tgid;
0804ef4b 3298 struct task_struct *task;
19fd4bb2
EB		 3299};
3300static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3301{
0804ef4b 3302 struct pid *pid;
1da177e4 3303
19fd4bb2
EB		 3304 if (iter.task)
3305 put_task_struct(iter.task);
454cc105 3306 rcu_read_lock();
0804ef4b 3307retry:
19fd4bb2
EB		 3308 iter.task = NULL;
3309 pid = find_ge_pid(iter.tgid, ns);
0804ef4b 3310 if (pid) {
19fd4bb2
EB		 3311 iter.tgid = pid_nr_ns(pid, ns);
3312 iter.task = pid_task(pid, PIDTYPE_PID);
0804ef4b
EB		 3313 /* What we to know is if the pid we have find is the
3314 * pid of a thread_group_leader. Testing for task
3315 * being a thread_group_leader is the obvious thing
3316 * todo but there is a window when it fails, due to
3317 * the pid transfer logic in de_thread.
3318 *
3319 * So we perform the straight forward test of seeing
3320 * if the pid we have found is the pid of a thread
3321 * group leader, and don't worry if the task we have
3322 * found doesn't happen to be a thread group leader.
3323 * As we don't care in the case of readdir.
3324 */
19fd4bb2
EB		 3325 if (!iter.task || !has_group_leader_pid(iter.task)) {
3326 iter.tgid += 1;
0804ef4b 3327 goto retry;
19fd4bb2
EB		 3328 }
3329 get_task_struct(iter.task);
0bc58a91 3330 }
454cc105 3331 rcu_read_unlock();
19fd4bb2 3332 return iter;
1da177e4
LT		 3333}
3334
7bcd6b0e 3335#define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
0804ef4b 3336
61a28784 3337static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
19fd4bb2 3338 struct tgid_iter iter)
61a28784
EB		 3339{
3340 char name[PROC_NUMBUF];
19fd4bb2 3341 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
61a28784 3342 return proc_fill_cache(filp, dirent, filldir, name, len,
19fd4bb2 3343 proc_pid_instantiate, iter.task, NULL);
61a28784
EB		 3344}
3345
0499680a
VK		 3346static int fake_filldir(void *buf, const char *name, int namelen,
3347 loff_t offset, u64 ino, unsigned d_type)
3348{
3349 return 0;
3350}
3351
1da177e4
LT		 3352/* for the /proc/ directory itself, after non-process stuff has been done */
3353int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3354{
d8bdc59f
LT		 3355 unsigned int nr;
3356 struct task_struct *reaper;
19fd4bb2 3357 struct tgid_iter iter;
b488893a 3358 struct pid_namespace *ns;
0499680a 3359 filldir_t __filldir;
1da177e4 3360
d8bdc59f
LT		 3361 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3362 goto out_no_task;
3363 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3364
3365 reaper = get_proc_task(filp->f_path.dentry->d_inode);
61a28784
EB		 3366 if (!reaper)
3367 goto out_no_task;
3368
7bcd6b0e 3369 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
c5141e6d 3370 const struct pid_entry *p = &proc_base_stuff[nr];
61a28784 3371 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
801199ce 3372 goto out;
1da177e4
LT		 3373 }
3374
b488893a 3375 ns = filp->f_dentry->d_sb->s_fs_info;
19fd4bb2
EB		 3376 iter.task = NULL;
3377 iter.tgid = filp->f_pos - TGID_OFFSET;
3378 for (iter = next_tgid(ns, iter);
3379 iter.task;
3380 iter.tgid += 1, iter = next_tgid(ns, iter)) {
0499680a
VK		 3381 if (has_pid_permissions(ns, iter.task, 2))
3382 __filldir = filldir;
3383 else
3384 __filldir = fake_filldir;
3385
19fd4bb2 3386 filp->f_pos = iter.tgid + TGID_OFFSET;
0499680a 3387 if (proc_pid_fill_cache(filp, dirent, __filldir, iter) < 0) {
19fd4bb2 3388 put_task_struct(iter.task);
0804ef4b 3389 goto out;
1da177e4 3390 }
0bc58a91 3391 }
0804ef4b
EB		 3392 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3393out:
61a28784
EB		 3394 put_task_struct(reaper);
3395out_no_task:
0bc58a91
EB		 3396 return 0;
3397}
1da177e4 3398
28a6d671
EB		 3399/*
3400 * Tasks
3401 */
c5141e6d 3402static const struct pid_entry tid_base_stuff[] = {
631f9c18 3403 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3835541d 3404 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 3405 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
631f9c18
AD		 3406 REG("environ", S_IRUSR, proc_environ_operations),
3407 INF("auxv", S_IRUSR, proc_pid_auxv),
3408 ONE("status", S_IRUGO, proc_pid_status),
a9712bc1 3409 ONE("personality", S_IRUGO, proc_pid_personality),
3036e7b4 3410 INF("limits", S_IRUGO, proc_pid_limits),
43ae34cb 3411#ifdef CONFIG_SCHED_DEBUG
631f9c18 3412 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
ebcb6734 3413#endif
4614a696 3414 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 3415#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
a9712bc1 3416 INF("syscall", S_IRUGO, proc_pid_syscall),
43ae34cb 3417#endif
631f9c18
AD		 3418 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3419 ONE("stat", S_IRUGO, proc_tid_stat),
3420 ONE("statm", S_IRUGO, proc_pid_statm),
b7643757 3421 REG("maps", S_IRUGO, proc_tid_maps_operations),
81841161
CG		 3422#ifdef CONFIG_CHECKPOINT_RESTORE
3423 REG("children", S_IRUGO, proc_tid_children_operations),
3424#endif
28a6d671 3425#ifdef CONFIG_NUMA
b7643757 3426 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
28a6d671 3427#endif
631f9c18
AD		 3428 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3429 LNK("cwd", proc_cwd_link),
3430 LNK("root", proc_root_link),
3431 LNK("exe", proc_exe_link),
3432 REG("mounts", S_IRUGO, proc_mounts_operations),
3433 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
1e883281 3434#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18 3435 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
b7643757 3436 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
ca6b0bf0 3437 REG("pagemap", S_IRUGO, proc_pagemap_operations),
28a6d671
EB		 3438#endif
3439#ifdef CONFIG_SECURITY
631f9c18 3440 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 3441#endif
3442#ifdef CONFIG_KALLSYMS
631f9c18 3443 INF("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 3444#endif
2ec220e2 3445#ifdef CONFIG_STACKTRACE
a9712bc1 3446 ONE("stack", S_IRUGO, proc_pid_stack),
28a6d671
EB		 3447#endif
3448#ifdef CONFIG_SCHEDSTATS
631f9c18 3449 INF("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 3450#endif
9745512c 3451#ifdef CONFIG_LATENCYTOP
631f9c18 3452 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 3453#endif
8793d854 3454#ifdef CONFIG_PROC_PID_CPUSET
631f9c18 3455 REG("cpuset", S_IRUGO, proc_cpuset_operations),
a424316c
PM		 3456#endif
3457#ifdef CONFIG_CGROUPS
631f9c18 3458 REG("cgroup", S_IRUGO, proc_cgroup_operations),
28a6d671 3459#endif
631f9c18
AD		 3460 INF("oom_score", S_IRUGO, proc_oom_score),
3461 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
a63d83f4 3462 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 3463#ifdef CONFIG_AUDITSYSCALL
631f9c18 3464 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
26ec3c64 3465 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 3466#endif
f4f154fd 3467#ifdef CONFIG_FAULT_INJECTION
631f9c18 3468 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 3469#endif
297c5d92 3470#ifdef CONFIG_TASK_IO_ACCOUNTING
1d1221f3 3471 INF("io", S_IRUSR, proc_tid_io_accounting),
297c5d92 3472#endif
f133ecca
CM		 3473#ifdef CONFIG_HARDWALL
3474 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3475#endif
22d917d8
EB		 3476#ifdef CONFIG_USER_NS
3477 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3478 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3479#endif
28a6d671
EB		 3480};
3481
3482static int proc_tid_base_readdir(struct file * filp,
3483 void * dirent, filldir_t filldir)
3484{
3485 return proc_pident_readdir(filp,dirent,filldir,
3486 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3487}
3488
00cd8dd3
AV		 3489static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3490{
7bcd6b0e
EB		 3491 return proc_pident_lookup(dir, dentry,
3492 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
28a6d671
EB		 3493}
3494
00977a59 3495static const struct file_operations proc_tid_base_operations = {
28a6d671
EB		 3496 .read = generic_read_dir,
3497 .readdir = proc_tid_base_readdir,
6038f373 3498 .llseek = default_llseek,
28a6d671
EB		 3499};
3500
c5ef1c42 3501static const struct inode_operations proc_tid_base_inode_operations = {
28a6d671
EB		 3502 .lookup = proc_tid_base_lookup,
3503 .getattr = pid_getattr,
3504 .setattr = proc_setattr,
3505};
3506
444ceed8 3507static struct dentry *proc_task_instantiate(struct inode *dir,
c5141e6d 3508 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8
EB		 3509{
3510 struct dentry *error = ERR_PTR(-ENOENT);
3511 struct inode *inode;
61a28784 3512 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 3513
3514 if (!inode)
3515 goto out;
3516 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3517 inode->i_op = &proc_tid_base_inode_operations;
3518 inode->i_fop = &proc_tid_base_operations;
3519 inode->i_flags|=S_IMMUTABLE;
aed54175 3520
bfe86848
MS		 3521 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3522 ARRAY_SIZE(tid_base_stuff)));
444ceed8 3523
fb045adb 3524 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 3525
3526 d_add(dentry, inode);
3527 /* Close the race of the process dying before we return the dentry */
0b728e19 3528 if (pid_revalidate(dentry, 0))
444ceed8
EB		 3529 error = NULL;
3530out:
3531 return error;
3532}
3533
00cd8dd3 3534static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
28a6d671
EB		 3535{
3536 struct dentry *result = ERR_PTR(-ENOENT);
3537 struct task_struct *task;
3538 struct task_struct *leader = get_proc_task(dir);
28a6d671 3539 unsigned tid;
b488893a 3540 struct pid_namespace *ns;
28a6d671
EB		 3541
3542 if (!leader)
3543 goto out_no_task;
3544
3545 tid = name_to_int(dentry);
3546 if (tid == ~0U)
3547 goto out;
3548
b488893a 3549 ns = dentry->d_sb->s_fs_info;
28a6d671 3550 rcu_read_lock();
b488893a 3551 task = find_task_by_pid_ns(tid, ns);
28a6d671
EB		 3552 if (task)
3553 get_task_struct(task);
3554 rcu_read_unlock();
3555 if (!task)
3556 goto out;
bac0abd6 3557 if (!same_thread_group(leader, task))
28a6d671
EB		 3558 goto out_drop_task;
3559
444ceed8 3560 result = proc_task_instantiate(dir, dentry, task, NULL);
28a6d671
EB		 3561out_drop_task:
3562 put_task_struct(task);
3563out:
3564 put_task_struct(leader);
3565out_no_task:
3566 return result;
3567}
3568
0bc58a91
EB		 3569/*
3570 * Find the first tid of a thread group to return to user space.
3571 *
3572 * Usually this is just the thread group leader, but if the users
3573 * buffer was too small or there was a seek into the middle of the
3574 * directory we have more work todo.
3575 *
3576 * In the case of a short read we start with find_task_by_pid.
3577 *
3578 * In the case of a seek we start with the leader and walk nr
3579 * threads past it.
3580 */
cc288738 3581static struct task_struct *first_tid(struct task_struct *leader,
b488893a 3582 int tid, int nr, struct pid_namespace *ns)
0bc58a91 3583{
a872ff0c 3584 struct task_struct *pos;
1da177e4 3585
cc288738 3586 rcu_read_lock();
0bc58a91
EB		 3587 /* Attempt to start with the pid of a thread */
3588 if (tid && (nr > 0)) {
b488893a 3589 pos = find_task_by_pid_ns(tid, ns);
a872ff0c
ON		 3590 if (pos && (pos->group_leader == leader))
3591 goto found;
0bc58a91 3592 }
1da177e4 3593
0bc58a91 3594 /* If nr exceeds the number of threads there is nothing todo */
a872ff0c
ON		 3595 pos = NULL;
3596 if (nr && nr >= get_nr_threads(leader))
3597 goto out;
1da177e4 3598
a872ff0c
ON		 3599 /* If we haven't found our starting place yet start
3600 * with the leader and walk nr threads forward.
0bc58a91 3601 */
a872ff0c
ON		 3602 for (pos = leader; nr > 0; --nr) {
3603 pos = next_thread(pos);
3604 if (pos == leader) {
3605 pos = NULL;
3606 goto out;
3607 }
1da177e4 3608 }
a872ff0c
ON		 3609found:
3610 get_task_struct(pos);
3611out:
cc288738 3612 rcu_read_unlock();
0bc58a91
EB		 3613 return pos;
3614}
3615
3616/*
3617 * Find the next thread in the thread list.
3618 * Return NULL if there is an error or no next thread.
3619 *
3620 * The reference to the input task_struct is released.
3621 */
3622static struct task_struct *next_tid(struct task_struct *start)
3623{
c1df7fb8 3624 struct task_struct *pos = NULL;
cc288738 3625 rcu_read_lock();
c1df7fb8 3626 if (pid_alive(start)) {
0bc58a91 3627 pos = next_thread(start);
c1df7fb8
ON		 3628 if (thread_group_leader(pos))
3629 pos = NULL;
3630 else
3631 get_task_struct(pos);
3632 }
cc288738 3633 rcu_read_unlock();
0bc58a91
EB		 3634 put_task_struct(start);
3635 return pos;
1da177e4
LT		 3636}
3637
61a28784
EB		 3638static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3639 struct task_struct *task, int tid)
3640{
3641 char name[PROC_NUMBUF];
3642 int len = snprintf(name, sizeof(name), "%d", tid);
3643 return proc_fill_cache(filp, dirent, filldir, name, len,
3644 proc_task_instantiate, task, NULL);
3645}
3646
1da177e4
LT		 3647/* for the /proc/TGID/task/ directories */
3648static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3649{
2fddfeef 3650 struct dentry *dentry = filp->f_path.dentry;
1da177e4 3651 struct inode *inode = dentry->d_inode;
7d895244 3652 struct task_struct *leader = NULL;
0bc58a91 3653 struct task_struct *task;
1da177e4
LT		 3654 int retval = -ENOENT;
3655 ino_t ino;
0bc58a91 3656 int tid;
b488893a 3657 struct pid_namespace *ns;
1da177e4 3658
7d895244
GC		 3659 task = get_proc_task(inode);
3660 if (!task)
3661 goto out_no_task;
3662 rcu_read_lock();
3663 if (pid_alive(task)) {
3664 leader = task->group_leader;
3665 get_task_struct(leader);
3666 }
3667 rcu_read_unlock();
3668 put_task_struct(task);
99f89551
EB		 3669 if (!leader)
3670 goto out_no_task;
1da177e4
LT		 3671 retval = 0;
3672
ee568b25 3673 switch ((unsigned long)filp->f_pos) {
1da177e4
LT		 3674 case 0:
3675 ino = inode->i_ino;
ee6f779b 3676 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
1da177e4 3677 goto out;
ee6f779b 3678 filp->f_pos++;
1da177e4
LT		 3679 /* fall through */
3680 case 1:
3681 ino = parent_ino(dentry);
ee6f779b 3682 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
1da177e4 3683 goto out;
ee6f779b 3684 filp->f_pos++;
1da177e4
LT		 3685 /* fall through */
3686 }
3687
0bc58a91
EB		 3688 /* f_version caches the tgid value that the last readdir call couldn't
3689 * return. lseek aka telldir automagically resets f_version to 0.
3690 */
b488893a 3691 ns = filp->f_dentry->d_sb->s_fs_info;
2b47c361 3692 tid = (int)filp->f_version;
0bc58a91 3693 filp->f_version = 0;
ee6f779b 3694 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
0bc58a91 3695 task;
ee6f779b 3696 task = next_tid(task), filp->f_pos++) {
b488893a 3697 tid = task_pid_nr_ns(task, ns);
61a28784 3698 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
0bc58a91
EB		 3699 /* returning this tgid failed, save it as the first
3700 * pid for the next readir call */
2b47c361 3701 filp->f_version = (u64)tid;
0bc58a91 3702 put_task_struct(task);
1da177e4 3703 break;
0bc58a91 3704 }
1da177e4
LT		 3705 }
3706out:
99f89551
EB		 3707 put_task_struct(leader);
3708out_no_task:
1da177e4
LT		 3709 return retval;
3710}
6e66b52b
EB		 3711
3712static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3713{
3714 struct inode *inode = dentry->d_inode;
99f89551 3715 struct task_struct *p = get_proc_task(inode);
6e66b52b
EB		 3716 generic_fillattr(inode, stat);
3717
99f89551 3718 if (p) {
99f89551 3719 stat->nlink += get_nr_threads(p);
99f89551 3720 put_task_struct(p);
6e66b52b
EB		 3721 }
3722
3723 return 0;
3724}
28a6d671 3725
c5ef1c42 3726static const struct inode_operations proc_task_inode_operations = {
28a6d671
EB		 3727 .lookup = proc_task_lookup,
3728 .getattr = proc_task_getattr,
3729 .setattr = proc_setattr,
0499680a 3730 .permission = proc_pid_permission,
28a6d671
EB		 3731};
3732
00977a59 3733static const struct file_operations proc_task_operations = {
28a6d671
EB		 3734 .read = generic_read_dir,
3735 .readdir = proc_task_readdir,
6038f373 3736 .llseek = default_llseek,
28a6d671 3737};
kernel source for telekom puls (alcatel/mediatek)