projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
switch ->is_visible() to returning umode_t
[linux-2.6-block.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>
5ad4e53b 84#include <linux/fs_struct.h>
5a0e3ad6 85#include <linux/slab.h>
f133ecca
CM		 86#ifdef CONFIG_HARDWALL
87#include <asm/hardwall.h>
88#endif
1da177e4
LT		 89#include "internal.h"
90
0f2fe20f
EB		 91/* NOTE:
92 * Implementing inode permission operations in /proc is almost
93 * certainly an error. Permission checks need to happen during
94 * each system call not at open time. The reason is that most of
95 * what we wish to check for permissions in /proc varies at runtime.
96 *
97 * The classic example of a problem is opening file descriptors
98 * in /proc for a task before it execs a suid executable.
99 */
100
1da177e4 101struct pid_entry {
1da177e4 102 char *name;
c5141e6d 103 int len;
1da177e4 104 mode_t mode;
c5ef1c42 105 const struct inode_operations *iop;
00977a59 106 const struct file_operations *fop;
20cdc894 107 union proc_op op;
1da177e4
LT		 108};
109
61a28784 110#define NOD(NAME, MODE, IOP, FOP, OP) { \
20cdc894 111 .name = (NAME), \
c5141e6d 112 .len = sizeof(NAME) - 1, \
20cdc894
EB		 113 .mode = MODE, \
114 .iop = IOP, \
115 .fop = FOP, \
116 .op = OP, \
117}
118
631f9c18
AD		 119#define DIR(NAME, MODE, iops, fops) \
120 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
121#define LNK(NAME, get_link) \
61a28784 122 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
20cdc894 123 &proc_pid_link_inode_operations, NULL, \
631f9c18
AD		 124 { .proc_get_link = get_link } )
125#define REG(NAME, MODE, fops) \
126 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
127#define INF(NAME, MODE, read) \
61a28784 128 NOD(NAME, (S_IFREG|(MODE)), \
20cdc894 129 NULL, &proc_info_file_operations, \
631f9c18
AD		 130 { .proc_read = read } )
131#define ONE(NAME, MODE, show) \
be614086
EB		 132 NOD(NAME, (S_IFREG|(MODE)), \
133 NULL, &proc_single_file_operations, \
631f9c18 134 { .proc_show = show } )
1da177e4 135
aed54175
VN		 136/*
137 * Count the number of hardlinks for the pid_entry table, excluding the .
138 * and .. links.
139 */
140static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
141 unsigned int n)
142{
143 unsigned int i;
144 unsigned int count;
145
146 count = 0;
147 for (i = 0; i < n; ++i) {
148 if (S_ISDIR(entries[i].mode))
149 ++count;
150 }
151
152 return count;
153}
154
f7ad3c6b 155static int get_task_root(struct task_struct *task, struct path *root)
1da177e4 156{
7c2c7d99
HD		 157 int result = -ENOENT;
158
0494f6ec 159 task_lock(task);
f7ad3c6b
MS		 160 if (task->fs) {
161 get_fs_root(task->fs, root);
7c2c7d99
HD		 162 result = 0;
163 }
0494f6ec 164 task_unlock(task);
7c2c7d99 165 return result;
0494f6ec
MS		 166}
167
3dcd25f3 168static int proc_cwd_link(struct inode *inode, struct path *path)
0494f6ec 169{
99f89551 170 struct task_struct *task = get_proc_task(inode);
0494f6ec 171 int result = -ENOENT;
99f89551
EB		 172
173 if (task) {
f7ad3c6b
MS		 174 task_lock(task);
175 if (task->fs) {
176 get_fs_pwd(task->fs, path);
177 result = 0;
178 }
179 task_unlock(task);
99f89551
EB		 180 put_task_struct(task);
181 }
1da177e4
LT		 182 return result;
183}
184
3dcd25f3 185static int proc_root_link(struct inode *inode, struct path *path)
1da177e4 186{
99f89551 187 struct task_struct *task = get_proc_task(inode);
1da177e4 188 int result = -ENOENT;
99f89551
EB		 189
190 if (task) {
f7ad3c6b 191 result = get_task_root(task, path);
99f89551
EB		 192 put_task_struct(task);
193 }
1da177e4
LT		 194 return result;
195}
196
8b0db9db 197static struct mm_struct *__check_mem_permission(struct task_struct *task)
638fa202 198{
8b0db9db
SW		 199 struct mm_struct *mm;
200
201 mm = get_task_mm(task);
202 if (!mm)
203 return ERR_PTR(-EINVAL);
204
638fa202
RM		 205 /*
206 * A task can always look at itself, in case it chooses
207 * to use system calls instead of load instructions.
208 */
209 if (task == current)
8b0db9db 210 return mm;
638fa202
RM		 211
212 /*
213 * If current is actively ptrace'ing, and would also be
214 * permitted to freshly attach with ptrace now, permit it.
215 */
0d094efe
RM		 216 if (task_is_stopped_or_traced(task)) {
217 int match;
218 rcu_read_lock();
06d98473 219 match = (ptrace_parent(task) == current);
0d094efe
RM		 220 rcu_read_unlock();
221 if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH))
8b0db9db 222 return mm;
0d094efe 223 }
638fa202
RM		 224
225 /*
25985edc 226 * No one else is allowed.
638fa202 227 */
8b0db9db
SW		 228 mmput(mm);
229 return ERR_PTR(-EPERM);
638fa202 230}
1da177e4 231
18f661bc 232/*
8b0db9db
SW		 233 * If current may access user memory in @task return a reference to the
234 * corresponding mm, otherwise ERR_PTR.
18f661bc 235 */
8b0db9db 236static struct mm_struct *check_mem_permission(struct task_struct *task)
18f661bc 237{
8b0db9db 238 struct mm_struct *mm;
18f661bc
SW		 239 int err;
240
241 /*
242 * Avoid racing if task exec's as we might get a new mm but validate
243 * against old credentials.
244 */
245 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
246 if (err)
8b0db9db 247 return ERR_PTR(err);
18f661bc 248
8b0db9db 249 mm = __check_mem_permission(task);
18f661bc
SW		 250 mutex_unlock(&task->signal->cred_guard_mutex);
251
8b0db9db 252 return mm;
638fa202 253}
1da177e4 254
831830b5
AV		 255struct mm_struct *mm_for_maps(struct task_struct *task)
256{
704b836c 257 struct mm_struct *mm;
ec6fd8a4 258 int err;
704b836c 259
ec6fd8a4
AV		 260 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
261 if (err)
262 return ERR_PTR(err);
00f89d21 263
704b836c
ON		 264 mm = get_task_mm(task);
265 if (mm && mm != current->mm &&
266 !ptrace_may_access(task, PTRACE_MODE_READ)) {
267 mmput(mm);
ec6fd8a4 268 mm = ERR_PTR(-EACCES);
13f0feaf 269 }
9b1bf12d 270 mutex_unlock(&task->signal->cred_guard_mutex);
704b836c 271
831830b5 272 return mm;
831830b5
AV		 273}
274
1da177e4
LT		 275static int proc_pid_cmdline(struct task_struct *task, char * buffer)
276{
277 int res = 0;
278 unsigned int len;
279 struct mm_struct *mm = get_task_mm(task);
280 if (!mm)
281 goto out;
282 if (!mm->arg_end)
283 goto out_mm; /* Shh! No looking before we're done */
284
285 len = mm->arg_end - mm->arg_start;
286
287 if (len > PAGE_SIZE)
288 len = PAGE_SIZE;
289
290 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
291
292 // If the nul at the end of args has been overwritten, then
293 // assume application is using setproctitle(3).
294 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
295 len = strnlen(buffer, res);
296 if (len < res) {
297 res = len;
298 } else {
299 len = mm->env_end - mm->env_start;
300 if (len > PAGE_SIZE - res)
301 len = PAGE_SIZE - res;
302 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
303 res = strnlen(buffer, res);
304 }
305 }
306out_mm:
307 mmput(mm);
308out:
309 return res;
310}
311
312static int proc_pid_auxv(struct task_struct *task, char *buffer)
313{
2fadaef4
AV		 314 struct mm_struct *mm = mm_for_maps(task);
315 int res = PTR_ERR(mm);
316 if (mm && !IS_ERR(mm)) {
1da177e4 317 unsigned int nwords = 0;
dfe6b7d9 318 do {
1da177e4 319 nwords += 2;
dfe6b7d9 320 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
1da177e4
LT		 321 res = nwords * sizeof(mm->saved_auxv[0]);
322 if (res > PAGE_SIZE)
323 res = PAGE_SIZE;
324 memcpy(buffer, mm->saved_auxv, res);
325 mmput(mm);
326 }
327 return res;
328}
329
330
331#ifdef CONFIG_KALLSYMS
332/*
333 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
334 * Returns the resolved symbol. If that fails, simply return the address.
335 */
336static int proc_pid_wchan(struct task_struct *task, char *buffer)
337{
ffb45122 338 unsigned long wchan;
9281acea 339 char symname[KSYM_NAME_LEN];
1da177e4
LT		 340
341 wchan = get_wchan(task);
342
9d65cb4a 343 if (lookup_symbol_name(wchan, symname) < 0)
f83ce3e6
JE		 344 if (!ptrace_may_access(task, PTRACE_MODE_READ))
345 return 0;
346 else
347 return sprintf(buffer, "%lu", wchan);
9d65cb4a
AD		 348 else
349 return sprintf(buffer, "%s", symname);
1da177e4
LT		 350}
351#endif /* CONFIG_KALLSYMS */
352
a9712bc1
AV		 353static int lock_trace(struct task_struct *task)
354{
355 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
356 if (err)
357 return err;
358 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
359 mutex_unlock(&task->signal->cred_guard_mutex);
360 return -EPERM;
361 }
362 return 0;
363}
364
365static void unlock_trace(struct task_struct *task)
366{
367 mutex_unlock(&task->signal->cred_guard_mutex);
368}
369
2ec220e2
KC		 370#ifdef CONFIG_STACKTRACE
371
372#define MAX_STACK_TRACE_DEPTH 64
373
374static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
375 struct pid *pid, struct task_struct *task)
376{
377 struct stack_trace trace;
378 unsigned long *entries;
a9712bc1 379 int err;
2ec220e2
KC		 380 int i;
381
382 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
383 if (!entries)
384 return -ENOMEM;
385
386 trace.nr_entries = 0;
387 trace.max_entries = MAX_STACK_TRACE_DEPTH;
388 trace.entries = entries;
389 trace.skip = 0;
2ec220e2 390
a9712bc1
AV		 391 err = lock_trace(task);
392 if (!err) {
393 save_stack_trace_tsk(task, &trace);
394
395 for (i = 0; i < trace.nr_entries; i++) {
b81a618d 396 seq_printf(m, "[<%pK>] %pS\n",
a9712bc1
AV		 397 (void *)entries[i], (void *)entries[i]);
398 }
399 unlock_trace(task);
2ec220e2
KC		 400 }
401 kfree(entries);
402
a9712bc1 403 return err;
2ec220e2
KC		 404}
405#endif
406
1da177e4
LT		 407#ifdef CONFIG_SCHEDSTATS
408/*
409 * Provides /proc/PID/schedstat
410 */
411static int proc_pid_schedstat(struct task_struct *task, char *buffer)
412{
172ba844 413 return sprintf(buffer, "%llu %llu %lu\n",
826e08b0
IM		 414 (unsigned long long)task->se.sum_exec_runtime,
415 (unsigned long long)task->sched_info.run_delay,
2d72376b 416 task->sched_info.pcount);
1da177e4
LT		 417}
418#endif
419
9745512c
AV		 420#ifdef CONFIG_LATENCYTOP
421static int lstats_show_proc(struct seq_file *m, void *v)
422{
423 int i;
13d77c37
HS		 424 struct inode *inode = m->private;
425 struct task_struct *task = get_proc_task(inode);
9745512c 426
13d77c37
HS		 427 if (!task)
428 return -ESRCH;
429 seq_puts(m, "Latency Top version : v0.1\n");
9745512c 430 for (i = 0; i < 32; i++) {
34e49d4f
JP		 431 struct latency_record *lr = &task->latency_record[i];
432 if (lr->backtrace[0]) {
9745512c 433 int q;
34e49d4f
JP		 434 seq_printf(m, "%i %li %li",
435 lr->count, lr->time, lr->max);
9745512c 436 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
34e49d4f
JP		 437 unsigned long bt = lr->backtrace[q];
438 if (!bt)
9745512c 439 break;
34e49d4f 440 if (bt == ULONG_MAX)
9745512c 441 break;
34e49d4f 442 seq_printf(m, " %ps", (void *)bt);
9745512c 443 }
9d6de12f 444 seq_putc(m, '\n');
9745512c
AV		 445 }
446
447 }
13d77c37 448 put_task_struct(task);
9745512c
AV		 449 return 0;
450}
451
452static int lstats_open(struct inode *inode, struct file *file)
453{
13d77c37 454 return single_open(file, lstats_show_proc, inode);
d6643d12
HS		 455}
456
9745512c
AV		 457static ssize_t lstats_write(struct file *file, const char __user *buf,
458 size_t count, loff_t *offs)
459{
13d77c37 460 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
9745512c 461
13d77c37
HS		 462 if (!task)
463 return -ESRCH;
9745512c 464 clear_all_latency_tracing(task);
13d77c37 465 put_task_struct(task);
9745512c
AV		 466
467 return count;
468}
469
470static const struct file_operations proc_lstats_operations = {
471 .open = lstats_open,
472 .read = seq_read,
473 .write = lstats_write,
474 .llseek = seq_lseek,
13d77c37 475 .release = single_release,
9745512c
AV		 476};
477
478#endif
479
1da177e4
LT		 480static int proc_oom_score(struct task_struct *task, char *buffer)
481{
b95c35e7 482 unsigned long points = 0;
1da177e4 483
19c5d45a 484 read_lock(&tasklist_lock);
b95c35e7 485 if (pid_alive(task))
a63d83f4
DR		 486 points = oom_badness(task, NULL, NULL,
487 totalram_pages + total_swap_pages);
19c5d45a 488 read_unlock(&tasklist_lock);
1da177e4
LT		 489 return sprintf(buffer, "%lu\n", points);
490}
491
d85f50d5
NH		 492struct limit_names {
493 char *name;
494 char *unit;
495};
496
497static const struct limit_names lnames[RLIM_NLIMITS] = {
cff4edb5 498 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
d85f50d5
NH		 499 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
500 [RLIMIT_DATA] = {"Max data size", "bytes"},
501 [RLIMIT_STACK] = {"Max stack size", "bytes"},
502 [RLIMIT_CORE] = {"Max core file size", "bytes"},
503 [RLIMIT_RSS] = {"Max resident set", "bytes"},
504 [RLIMIT_NPROC] = {"Max processes", "processes"},
505 [RLIMIT_NOFILE] = {"Max open files", "files"},
506 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
507 [RLIMIT_AS] = {"Max address space", "bytes"},
508 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
509 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
510 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
511 [RLIMIT_NICE] = {"Max nice priority", NULL},
512 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
8808117c 513 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
d85f50d5
NH		 514};
515
516/* Display limits for a process */
517static int proc_pid_limits(struct task_struct *task, char *buffer)
518{
519 unsigned int i;
520 int count = 0;
521 unsigned long flags;
522 char *bufptr = buffer;
523
524 struct rlimit rlim[RLIM_NLIMITS];
525
a6bebbc8 526 if (!lock_task_sighand(task, &flags))
d85f50d5 527 return 0;
d85f50d5
NH		 528 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
529 unlock_task_sighand(task, &flags);
d85f50d5
NH		 530
531 /*
532 * print the file header
533 */
534 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
535 "Limit", "Soft Limit", "Hard Limit", "Units");
536
537 for (i = 0; i < RLIM_NLIMITS; i++) {
538 if (rlim[i].rlim_cur == RLIM_INFINITY)
539 count += sprintf(&bufptr[count], "%-25s %-20s ",
540 lnames[i].name, "unlimited");
541 else
542 count += sprintf(&bufptr[count], "%-25s %-20lu ",
543 lnames[i].name, rlim[i].rlim_cur);
544
545 if (rlim[i].rlim_max == RLIM_INFINITY)
546 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
547 else
548 count += sprintf(&bufptr[count], "%-20lu ",
549 rlim[i].rlim_max);
550
551 if (lnames[i].unit)
552 count += sprintf(&bufptr[count], "%-10s\n",
553 lnames[i].unit);
554 else
555 count += sprintf(&bufptr[count], "\n");
556 }
557
558 return count;
559}
560
ebcb6734
RM		 561#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
562static int proc_pid_syscall(struct task_struct *task, char *buffer)
563{
564 long nr;
565 unsigned long args[6], sp, pc;
a9712bc1
AV		 566 int res = lock_trace(task);
567 if (res)
568 return res;
ebcb6734
RM		 569
570 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
a9712bc1
AV		 571 res = sprintf(buffer, "running\n");
572 else if (nr < 0)
573 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
574 else
575 res = sprintf(buffer,
ebcb6734
RM		 576 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
577 nr,
578 args[0], args[1], args[2], args[3], args[4], args[5],
579 sp, pc);
a9712bc1
AV		 580 unlock_trace(task);
581 return res;
ebcb6734
RM		 582}
583#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
584
1da177e4
LT		 585/************************************************************************/
586/* Here the fs part begins */
587/************************************************************************/
588
589/* permission checks */
778c1144 590static int proc_fd_access_allowed(struct inode *inode)
1da177e4 591{
778c1144
EB		 592 struct task_struct *task;
593 int allowed = 0;
df26c40e
EB		 594 /* Allow access to a task's file descriptors if it is us or we
595 * may use ptrace attach to the process and find out that
596 * information.
778c1144
EB		 597 */
598 task = get_proc_task(inode);
df26c40e 599 if (task) {
006ebb40 600 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
778c1144 601 put_task_struct(task);
df26c40e 602 }
778c1144 603 return allowed;
1da177e4
LT		 604}
605
6b4e306a 606int proc_setattr(struct dentry *dentry, struct iattr *attr)
6d76fa58
LT		 607{
608 int error;
609 struct inode *inode = dentry->d_inode;
610
611 if (attr->ia_valid & ATTR_MODE)
612 return -EPERM;
613
614 error = inode_change_ok(inode, attr);
1025774c
CH		 615 if (error)
616 return error;
617
618 if ((attr->ia_valid & ATTR_SIZE) &&
619 attr->ia_size != i_size_read(inode)) {
620 error = vmtruncate(inode, attr->ia_size);
621 if (error)
622 return error;
623 }
624
625 setattr_copy(inode, attr);
626 mark_inode_dirty(inode);
627 return 0;
6d76fa58
LT		 628}
629
c5ef1c42 630static const struct inode_operations proc_def_inode_operations = {
6d76fa58
LT		 631 .setattr = proc_setattr,
632};
633
a1a2c409
MS		 634static int mounts_open_common(struct inode *inode, struct file *file,
635 const struct seq_operations *op)
1da177e4 636{
99f89551 637 struct task_struct *task = get_proc_task(inode);
cf7b708c 638 struct nsproxy *nsp;
6b3286ed 639 struct mnt_namespace *ns = NULL;
a1a2c409 640 struct path root;
5addc5dd
AV		 641 struct proc_mounts *p;
642 int ret = -EINVAL;
1da177e4 643
99f89551 644 if (task) {
cf7b708c
PE		 645 rcu_read_lock();
646 nsp = task_nsproxy(task);
647 if (nsp) {
648 ns = nsp->mnt_ns;
863c4702
AD		 649 if (ns)
650 get_mnt_ns(ns);
651 }
cf7b708c 652 rcu_read_unlock();
f7ad3c6b 653 if (ns && get_task_root(task, &root) == 0)
7c2c7d99 654 ret = 0;
99f89551
EB		 655 put_task_struct(task);
656 }
5addc5dd 657
a1a2c409
MS		 658 if (!ns)
659 goto err;
7c2c7d99 660 if (ret)
a1a2c409
MS		 661 goto err_put_ns;
662
a1a2c409
MS		 663 ret = -ENOMEM;
664 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
665 if (!p)
666 goto err_put_path;
667
668 file->private_data = &p->m;
669 ret = seq_open(file, op);
670 if (ret)
671 goto err_free;
672
673 p->m.private = p;
674 p->ns = ns;
675 p->root = root;
f1514638 676 p->m.poll_event = ns->event;
a1a2c409
MS		 677
678 return 0;
679
680 err_free:
681 kfree(p);
682 err_put_path:
683 path_put(&root);
684 err_put_ns:
685 put_mnt_ns(ns);
686 err:
1da177e4
LT		 687 return ret;
688}
689
690static int mounts_release(struct inode *inode, struct file *file)
691{
a1a2c409
MS		 692 struct proc_mounts *p = file->private_data;
693 path_put(&p->root);
694 put_mnt_ns(p->ns);
1da177e4
LT		 695 return seq_release(inode, file);
696}
697
5addc5dd
AV		 698static unsigned mounts_poll(struct file *file, poll_table *wait)
699{
700 struct proc_mounts *p = file->private_data;
31b07093 701 unsigned res = POLLIN | POLLRDNORM;
5addc5dd 702
9f5596af
AV		 703 poll_wait(file, &p->ns->poll, wait);
704 if (mnt_had_events(p))
31b07093 705 res |= POLLERR | POLLPRI;
5addc5dd
AV		 706
707 return res;
708}
709
a1a2c409
MS		 710static int mounts_open(struct inode *inode, struct file *file)
711{
712 return mounts_open_common(inode, file, &mounts_op);
713}
714
00977a59 715static const struct file_operations proc_mounts_operations = {
1da177e4
LT		 716 .open = mounts_open,
717 .read = seq_read,
718 .llseek = seq_lseek,
719 .release = mounts_release,
5addc5dd 720 .poll = mounts_poll,
1da177e4
LT		 721};
722
2d4d4864
RP		 723static int mountinfo_open(struct inode *inode, struct file *file)
724{
725 return mounts_open_common(inode, file, &mountinfo_op);
726}
727
728static const struct file_operations proc_mountinfo_operations = {
729 .open = mountinfo_open,
730 .read = seq_read,
731 .llseek = seq_lseek,
732 .release = mounts_release,
733 .poll = mounts_poll,
734};
735
b4629fe2
CL		 736static int mountstats_open(struct inode *inode, struct file *file)
737{
a1a2c409 738 return mounts_open_common(inode, file, &mountstats_op);
b4629fe2
CL		 739}
740
00977a59 741static const struct file_operations proc_mountstats_operations = {
b4629fe2
CL		 742 .open = mountstats_open,
743 .read = seq_read,
744 .llseek = seq_lseek,
745 .release = mounts_release,
746};
747
1da177e4
LT		 748#define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
749
750static ssize_t proc_info_read(struct file * file, char __user * buf,
751 size_t count, loff_t *ppos)
752{
2fddfeef 753 struct inode * inode = file->f_path.dentry->d_inode;
1da177e4
LT		 754 unsigned long page;
755 ssize_t length;
99f89551
EB		 756 struct task_struct *task = get_proc_task(inode);
757
758 length = -ESRCH;
759 if (!task)
760 goto out_no_task;
1da177e4
LT		 761
762 if (count > PROC_BLOCK_SIZE)
763 count = PROC_BLOCK_SIZE;
99f89551
EB		 764
765 length = -ENOMEM;
e12ba74d 766 if (!(page = __get_free_page(GFP_TEMPORARY)))
99f89551 767 goto out;
1da177e4
LT		 768
769 length = PROC_I(inode)->op.proc_read(task, (char*)page);
770
771 if (length >= 0)
772 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
773 free_page(page);
99f89551
EB		 774out:
775 put_task_struct(task);
776out_no_task:
1da177e4
LT		 777 return length;
778}
779
00977a59 780static const struct file_operations proc_info_file_operations = {
1da177e4 781 .read = proc_info_read,
87df8424 782 .llseek = generic_file_llseek,
1da177e4
LT		 783};
784
be614086
EB		 785static int proc_single_show(struct seq_file *m, void *v)
786{
787 struct inode *inode = m->private;
788 struct pid_namespace *ns;
789 struct pid *pid;
790 struct task_struct *task;
791 int ret;
792
793 ns = inode->i_sb->s_fs_info;
794 pid = proc_pid(inode);
795 task = get_pid_task(pid, PIDTYPE_PID);
796 if (!task)
797 return -ESRCH;
798
799 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
800
801 put_task_struct(task);
802 return ret;
803}
804
805static int proc_single_open(struct inode *inode, struct file *filp)
806{
c6a34058 807 return single_open(filp, proc_single_show, inode);
be614086
EB		 808}
809
810static const struct file_operations proc_single_file_operations = {
811 .open = proc_single_open,
812 .read = seq_read,
813 .llseek = seq_lseek,
814 .release = single_release,
815};
816
1da177e4
LT		 817static int mem_open(struct inode* inode, struct file* file)
818{
819 file->private_data = (void*)((long)current->self_exec_id);
4a3956c7
KH		 820 /* OK to pass negative loff_t, we can catch out-of-range */
821 file->f_mode |= FMODE_UNSIGNED_OFFSET;
1da177e4
LT		 822 return 0;
823}
824
825static ssize_t mem_read(struct file * file, char __user * buf,
826 size_t count, loff_t *ppos)
827{
2fddfeef 828 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1da177e4
LT		 829 char *page;
830 unsigned long src = *ppos;
831 int ret = -ESRCH;
832 struct mm_struct *mm;
833
99f89551
EB		 834 if (!task)
835 goto out_no_task;
836
1da177e4 837 ret = -ENOMEM;
e12ba74d 838 page = (char *)__get_free_page(GFP_TEMPORARY);
1da177e4
LT		 839 if (!page)
840 goto out;
841
8b0db9db
SW		 842 mm = check_mem_permission(task);
843 ret = PTR_ERR(mm);
844 if (IS_ERR(mm))
1da177e4
LT		 845 goto out_free;
846
847 ret = -EIO;
848
849 if (file->private_data != (void*)((long)current->self_exec_id))
850 goto out_put;
851
852 ret = 0;
853
854 while (count > 0) {
855 int this_len, retval;
856
857 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
8b0db9db
SW		 858 retval = access_remote_vm(mm, src, page, this_len, 0);
859 if (!retval) {
1da177e4
LT		 860 if (!ret)
861 ret = -EIO;
862 break;
863 }
864
865 if (copy_to_user(buf, page, retval)) {
866 ret = -EFAULT;
867 break;
868 }
869
870 ret += retval;
871 src += retval;
872 buf += retval;
873 count -= retval;
874 }
875 *ppos = src;
876
877out_put:
878 mmput(mm);
879out_free:
880 free_page((unsigned long) page);
881out:
99f89551
EB		 882 put_task_struct(task);
883out_no_task:
1da177e4
LT		 884 return ret;
885}
886
63967fa9 887static ssize_t mem_write(struct file * file, const char __user *buf,
1da177e4
LT		 888 size_t count, loff_t *ppos)
889{
f7ca54f4 890 int copied;
1da177e4 891 char *page;
2fddfeef 892 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1da177e4 893 unsigned long dst = *ppos;
8b0db9db 894 struct mm_struct *mm;
1da177e4 895
99f89551
EB		 896 copied = -ESRCH;
897 if (!task)
898 goto out_no_task;
899
30cd8903
KM		 900 copied = -ENOMEM;
901 page = (char *)__get_free_page(GFP_TEMPORARY);
902 if (!page)
903 goto out_task;
904
8b0db9db
SW		 905 mm = check_mem_permission(task);
906 copied = PTR_ERR(mm);
907 if (IS_ERR(mm))
30cd8903 908 goto out_free;
1da177e4 909
26947f8c
SW		 910 copied = -EIO;
911 if (file->private_data != (void *)((long)current->self_exec_id))
8b0db9db 912 goto out_mm;
1da177e4 913
f7ca54f4 914 copied = 0;
1da177e4
LT		 915 while (count > 0) {
916 int this_len, retval;
917
918 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
919 if (copy_from_user(page, buf, this_len)) {
920 copied = -EFAULT;
921 break;
922 }
8b0db9db 923 retval = access_remote_vm(mm, dst, page, this_len, 1);
1da177e4
LT		 924 if (!retval) {
925 if (!copied)
926 copied = -EIO;
927 break;
928 }
929 copied += retval;
930 buf += retval;
931 dst += retval;
932 count -= retval;
933 }
934 *ppos = dst;
30cd8903 935
8b0db9db
SW		 936out_mm:
937 mmput(mm);
30cd8903
KM		 938out_free:
939 free_page((unsigned long) page);
8b0db9db 940out_task:
99f89551
EB		 941 put_task_struct(task);
942out_no_task:
1da177e4
LT		 943 return copied;
944}
1da177e4 945
85863e47 946loff_t mem_lseek(struct file *file, loff_t offset, int orig)
1da177e4
LT		 947{
948 switch (orig) {
949 case 0:
950 file->f_pos = offset;
951 break;
952 case 1:
953 file->f_pos += offset;
954 break;
955 default:
956 return -EINVAL;
957 }
958 force_successful_syscall_return();
959 return file->f_pos;
960}
961
00977a59 962static const struct file_operations proc_mem_operations = {
1da177e4
LT		 963 .llseek = mem_lseek,
964 .read = mem_read,
965 .write = mem_write,
966 .open = mem_open,
967};
968
315e28c8
JP		 969static ssize_t environ_read(struct file *file, char __user *buf,
970 size_t count, loff_t *ppos)
971{
972 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
973 char *page;
974 unsigned long src = *ppos;
975 int ret = -ESRCH;
976 struct mm_struct *mm;
977
978 if (!task)
979 goto out_no_task;
980
315e28c8
JP		 981 ret = -ENOMEM;
982 page = (char *)__get_free_page(GFP_TEMPORARY);
983 if (!page)
984 goto out;
985
315e28c8 986
d6f64b89
AV		 987 mm = mm_for_maps(task);
988 ret = PTR_ERR(mm);
989 if (!mm || IS_ERR(mm))
315e28c8
JP		 990 goto out_free;
991
d6f64b89 992 ret = 0;
315e28c8
JP		 993 while (count > 0) {
994 int this_len, retval, max_len;
995
996 this_len = mm->env_end - (mm->env_start + src);
997
998 if (this_len <= 0)
999 break;
1000
1001 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
1002 this_len = (this_len > max_len) ? max_len : this_len;
1003
1004 retval = access_process_vm(task, (mm->env_start + src),
1005 page, this_len, 0);
1006
1007 if (retval <= 0) {
1008 ret = retval;
1009 break;
1010 }
1011
1012 if (copy_to_user(buf, page, retval)) {
1013 ret = -EFAULT;
1014 break;
1015 }
1016
1017 ret += retval;
1018 src += retval;
1019 buf += retval;
1020 count -= retval;
1021 }
1022 *ppos = src;
1023
1024 mmput(mm);
1025out_free:
1026 free_page((unsigned long) page);
1027out:
1028 put_task_struct(task);
1029out_no_task:
1030 return ret;
1031}
1032
1033static const struct file_operations proc_environ_operations = {
1034 .read = environ_read,
87df8424 1035 .llseek = generic_file_llseek,
315e28c8
JP		 1036};
1037
1da177e4
LT		 1038static ssize_t oom_adjust_read(struct file *file, char __user *buf,
1039 size_t count, loff_t *ppos)
1040{
2fddfeef 1041 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
8578cea7 1042 char buffer[PROC_NUMBUF];
1da177e4 1043 size_t len;
28b83c51
KM		 1044 int oom_adjust = OOM_DISABLE;
1045 unsigned long flags;
1da177e4 1046
99f89551
EB		 1047 if (!task)
1048 return -ESRCH;
28b83c51
KM		 1049
1050 if (lock_task_sighand(task, &flags)) {
1051 oom_adjust = task->signal->oom_adj;
1052 unlock_task_sighand(task, &flags);
1053 }
1054
99f89551
EB		 1055 put_task_struct(task);
1056
8578cea7 1057 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
0c28f287
AM		 1058
1059 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1da177e4
LT		 1060}
1061
1062static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1063 size_t count, loff_t *ppos)
1064{
99f89551 1065 struct task_struct *task;
5d863b89 1066 char buffer[PROC_NUMBUF];
0a8cb8e3 1067 int oom_adjust;
28b83c51 1068 unsigned long flags;
5d863b89 1069 int err;
1da177e4 1070
8578cea7
EB		 1071 memset(buffer, 0, sizeof(buffer));
1072 if (count > sizeof(buffer) - 1)
1073 count = sizeof(buffer) - 1;
723548bf
DR		 1074 if (copy_from_user(buffer, buf, count)) {
1075 err = -EFAULT;
1076 goto out;
1077 }
5d863b89 1078
0a8cb8e3 1079 err = kstrtoint(strstrip(buffer), 0, &oom_adjust);
5d863b89 1080 if (err)
723548bf 1081 goto out;
8ac773b4 1082 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
723548bf
DR		 1083 oom_adjust != OOM_DISABLE) {
1084 err = -EINVAL;
1085 goto out;
1086 }
5d863b89 1087
2fddfeef 1088 task = get_proc_task(file->f_path.dentry->d_inode);
723548bf
DR		 1089 if (!task) {
1090 err = -ESRCH;
1091 goto out;
1092 }
d19d5476
DR		 1093
1094 task_lock(task);
1095 if (!task->mm) {
1096 err = -EINVAL;
1097 goto err_task_lock;
1098 }
1099
28b83c51 1100 if (!lock_task_sighand(task, &flags)) {
723548bf 1101 err = -ESRCH;
d19d5476 1102 goto err_task_lock;
28b83c51
KM		 1103 }
1104
1105 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
723548bf
DR		 1106 err = -EACCES;
1107 goto err_sighand;
8fb4fc68 1108 }
28b83c51 1109
51b1bd2a
DR		 1110 /*
1111 * Warn that /proc/pid/oom_adj is deprecated, see
1112 * Documentation/feature-removal-schedule.txt.
1113 */
c2142704 1114 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
be8f684d
DR		 1115 current->comm, task_pid_nr(current), task_pid_nr(task),
1116 task_pid_nr(task));
28b83c51 1117 task->signal->oom_adj = oom_adjust;
a63d83f4
DR		 1118 /*
1119 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1120 * value is always attainable.
1121 */
1122 if (task->signal->oom_adj == OOM_ADJUST_MAX)
1123 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
1124 else
1125 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
1126 -OOM_DISABLE;
723548bf 1127err_sighand:
28b83c51 1128 unlock_task_sighand(task, &flags);
d19d5476
DR		 1129err_task_lock:
1130 task_unlock(task);
99f89551 1131 put_task_struct(task);
723548bf
DR		 1132out:
1133 return err < 0 ? err : count;
1da177e4
LT		 1134}
1135
00977a59 1136static const struct file_operations proc_oom_adjust_operations = {
1da177e4
LT		 1137 .read = oom_adjust_read,
1138 .write = oom_adjust_write,
87df8424 1139 .llseek = generic_file_llseek,
1da177e4
LT		 1140};
1141
a63d83f4
DR		 1142static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1143 size_t count, loff_t *ppos)
1144{
1145 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1146 char buffer[PROC_NUMBUF];
1147 int oom_score_adj = OOM_SCORE_ADJ_MIN;
1148 unsigned long flags;
1149 size_t len;
1150
1151 if (!task)
1152 return -ESRCH;
1153 if (lock_task_sighand(task, &flags)) {
1154 oom_score_adj = task->signal->oom_score_adj;
1155 unlock_task_sighand(task, &flags);
1156 }
1157 put_task_struct(task);
1158 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
1159 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1160}
1161
1162static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1163 size_t count, loff_t *ppos)
1164{
1165 struct task_struct *task;
1166 char buffer[PROC_NUMBUF];
1167 unsigned long flags;
0a8cb8e3 1168 int oom_score_adj;
a63d83f4
DR		 1169 int err;
1170
1171 memset(buffer, 0, sizeof(buffer));
1172 if (count > sizeof(buffer) - 1)
1173 count = sizeof(buffer) - 1;
723548bf
DR		 1174 if (copy_from_user(buffer, buf, count)) {
1175 err = -EFAULT;
1176 goto out;
1177 }
a63d83f4 1178
0a8cb8e3 1179 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
a63d83f4 1180 if (err)
723548bf 1181 goto out;
a63d83f4 1182 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
723548bf
DR		 1183 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1184 err = -EINVAL;
1185 goto out;
1186 }
a63d83f4
DR		 1187
1188 task = get_proc_task(file->f_path.dentry->d_inode);
723548bf
DR		 1189 if (!task) {
1190 err = -ESRCH;
1191 goto out;
1192 }
d19d5476
DR		 1193
1194 task_lock(task);
1195 if (!task->mm) {
1196 err = -EINVAL;
1197 goto err_task_lock;
1198 }
1199
a63d83f4 1200 if (!lock_task_sighand(task, &flags)) {
723548bf 1201 err = -ESRCH;
d19d5476 1202 goto err_task_lock;
a63d83f4 1203 }
d19d5476 1204
dabb16f6 1205 if (oom_score_adj < task->signal->oom_score_adj_min &&
a63d83f4 1206 !capable(CAP_SYS_RESOURCE)) {
723548bf
DR		 1207 err = -EACCES;
1208 goto err_sighand;
a63d83f4
DR		 1209 }
1210
1211 task->signal->oom_score_adj = oom_score_adj;
dabb16f6
MSB		 1212 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1213 task->signal->oom_score_adj_min = oom_score_adj;
a63d83f4
DR		 1214 /*
1215 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1216 * always attainable.
1217 */
1218 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1219 task->signal->oom_adj = OOM_DISABLE;
1220 else
1221 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1222 OOM_SCORE_ADJ_MAX;
723548bf 1223err_sighand:
a63d83f4 1224 unlock_task_sighand(task, &flags);
d19d5476
DR		 1225err_task_lock:
1226 task_unlock(task);
a63d83f4 1227 put_task_struct(task);
723548bf
DR		 1228out:
1229 return err < 0 ? err : count;
a63d83f4
DR		 1230}
1231
1232static const struct file_operations proc_oom_score_adj_operations = {
1233 .read = oom_score_adj_read,
1234 .write = oom_score_adj_write,
6038f373 1235 .llseek = default_llseek,
a63d83f4
DR		 1236};
1237
1da177e4
LT		 1238#ifdef CONFIG_AUDITSYSCALL
1239#define TMPBUFLEN 21
1240static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1241 size_t count, loff_t *ppos)
1242{
2fddfeef 1243 struct inode * inode = file->f_path.dentry->d_inode;
99f89551 1244 struct task_struct *task = get_proc_task(inode);
1da177e4
LT		 1245 ssize_t length;
1246 char tmpbuf[TMPBUFLEN];
1247
99f89551
EB		 1248 if (!task)
1249 return -ESRCH;
1da177e4 1250 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
0c11b942 1251 audit_get_loginuid(task));
99f89551 1252 put_task_struct(task);
1da177e4
LT		 1253 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1254}
1255
1256static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1257 size_t count, loff_t *ppos)
1258{
2fddfeef 1259 struct inode * inode = file->f_path.dentry->d_inode;
1da177e4
LT		 1260 char *page, *tmp;
1261 ssize_t length;
1da177e4
LT		 1262 uid_t loginuid;
1263
1264 if (!capable(CAP_AUDIT_CONTROL))
1265 return -EPERM;
1266
7dc52157
PM		 1267 rcu_read_lock();
1268 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1269 rcu_read_unlock();
1da177e4 1270 return -EPERM;
7dc52157
PM		 1271 }
1272 rcu_read_unlock();
1da177e4 1273
e0182909
AV		 1274 if (count >= PAGE_SIZE)
1275 count = PAGE_SIZE - 1;
1da177e4
LT		 1276
1277 if (*ppos != 0) {
1278 /* No partial writes. */
1279 return -EINVAL;
1280 }
e12ba74d 1281 page = (char*)__get_free_page(GFP_TEMPORARY);
1da177e4
LT		 1282 if (!page)
1283 return -ENOMEM;
1284 length = -EFAULT;
1285 if (copy_from_user(page, buf, count))
1286 goto out_free_page;
1287
e0182909 1288 page[count] = '\0';
1da177e4
LT		 1289 loginuid = simple_strtoul(page, &tmp, 10);
1290 if (tmp == page) {
1291 length = -EINVAL;
1292 goto out_free_page;
1293
1294 }
99f89551 1295 length = audit_set_loginuid(current, loginuid);
1da177e4
LT		 1296 if (likely(length == 0))
1297 length = count;
1298
1299out_free_page:
1300 free_page((unsigned long) page);
1301 return length;
1302}
1303
00977a59 1304static const struct file_operations proc_loginuid_operations = {
1da177e4
LT		 1305 .read = proc_loginuid_read,
1306 .write = proc_loginuid_write,
87df8424 1307 .llseek = generic_file_llseek,
1da177e4 1308};
1e0bd755
EP		 1309
1310static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1311 size_t count, loff_t *ppos)
1312{
1313 struct inode * inode = file->f_path.dentry->d_inode;
1314 struct task_struct *task = get_proc_task(inode);
1315 ssize_t length;
1316 char tmpbuf[TMPBUFLEN];
1317
1318 if (!task)
1319 return -ESRCH;
1320 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1321 audit_get_sessionid(task));
1322 put_task_struct(task);
1323 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1324}
1325
1326static const struct file_operations proc_sessionid_operations = {
1327 .read = proc_sessionid_read,
87df8424 1328 .llseek = generic_file_llseek,
1e0bd755 1329};
1da177e4
LT		 1330#endif
1331
f4f154fd
AM		 1332#ifdef CONFIG_FAULT_INJECTION
1333static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1334 size_t count, loff_t *ppos)
1335{
1336 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1337 char buffer[PROC_NUMBUF];
1338 size_t len;
1339 int make_it_fail;
f4f154fd
AM		 1340
1341 if (!task)
1342 return -ESRCH;
1343 make_it_fail = task->make_it_fail;
1344 put_task_struct(task);
1345
1346 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
0c28f287
AM		 1347
1348 return simple_read_from_buffer(buf, count, ppos, buffer, len);
f4f154fd
AM		 1349}
1350
1351static ssize_t proc_fault_inject_write(struct file * file,
1352 const char __user * buf, size_t count, loff_t *ppos)
1353{
1354 struct task_struct *task;
1355 char buffer[PROC_NUMBUF], *end;
1356 int make_it_fail;
1357
1358 if (!capable(CAP_SYS_RESOURCE))
1359 return -EPERM;
1360 memset(buffer, 0, sizeof(buffer));
1361 if (count > sizeof(buffer) - 1)
1362 count = sizeof(buffer) - 1;
1363 if (copy_from_user(buffer, buf, count))
1364 return -EFAULT;
cba8aafe
VL		 1365 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1366 if (*end)
1367 return -EINVAL;
f4f154fd
AM		 1368 task = get_proc_task(file->f_dentry->d_inode);
1369 if (!task)
1370 return -ESRCH;
1371 task->make_it_fail = make_it_fail;
1372 put_task_struct(task);
cba8aafe
VL		 1373
1374 return count;
f4f154fd
AM		 1375}
1376
00977a59 1377static const struct file_operations proc_fault_inject_operations = {
f4f154fd
AM		 1378 .read = proc_fault_inject_read,
1379 .write = proc_fault_inject_write,
87df8424 1380 .llseek = generic_file_llseek,
f4f154fd
AM		 1381};
1382#endif
1383
9745512c 1384
43ae34cb
IM		 1385#ifdef CONFIG_SCHED_DEBUG
1386/*
1387 * Print out various scheduling related per-task fields:
1388 */
1389static int sched_show(struct seq_file *m, void *v)
1390{
1391 struct inode *inode = m->private;
1392 struct task_struct *p;
1393
43ae34cb
IM		 1394 p = get_proc_task(inode);
1395 if (!p)
1396 return -ESRCH;
1397 proc_sched_show_task(p, m);
1398
1399 put_task_struct(p);
1400
1401 return 0;
1402}
1403
1404static ssize_t
1405sched_write(struct file *file, const char __user *buf,
1406 size_t count, loff_t *offset)
1407{
1408 struct inode *inode = file->f_path.dentry->d_inode;
1409 struct task_struct *p;
1410
43ae34cb
IM		 1411 p = get_proc_task(inode);
1412 if (!p)
1413 return -ESRCH;
1414 proc_sched_set_task(p);
1415
1416 put_task_struct(p);
1417
1418 return count;
1419}
1420
1421static int sched_open(struct inode *inode, struct file *filp)
1422{
c6a34058 1423 return single_open(filp, sched_show, inode);
43ae34cb
IM		 1424}
1425
1426static const struct file_operations proc_pid_sched_operations = {
1427 .open = sched_open,
1428 .read = seq_read,
1429 .write = sched_write,
1430 .llseek = seq_lseek,
5ea473a1 1431 .release = single_release,
43ae34cb
IM		 1432};
1433
1434#endif
1435
5091faa4
MG		 1436#ifdef CONFIG_SCHED_AUTOGROUP
1437/*
1438 * Print out autogroup related information:
1439 */
1440static int sched_autogroup_show(struct seq_file *m, void *v)
1441{
1442 struct inode *inode = m->private;
1443 struct task_struct *p;
1444
1445 p = get_proc_task(inode);
1446 if (!p)
1447 return -ESRCH;
1448 proc_sched_autogroup_show_task(p, m);
1449
1450 put_task_struct(p);
1451
1452 return 0;
1453}
1454
1455static ssize_t
1456sched_autogroup_write(struct file *file, const char __user *buf,
1457 size_t count, loff_t *offset)
1458{
1459 struct inode *inode = file->f_path.dentry->d_inode;
1460 struct task_struct *p;
1461 char buffer[PROC_NUMBUF];
0a8cb8e3 1462 int nice;
5091faa4
MG		 1463 int err;
1464
1465 memset(buffer, 0, sizeof(buffer));
1466 if (count > sizeof(buffer) - 1)
1467 count = sizeof(buffer) - 1;
1468 if (copy_from_user(buffer, buf, count))
1469 return -EFAULT;
1470
0a8cb8e3
AD		 1471 err = kstrtoint(strstrip(buffer), 0, &nice);
1472 if (err < 0)
1473 return err;
5091faa4
MG		 1474
1475 p = get_proc_task(inode);
1476 if (!p)
1477 return -ESRCH;
1478
1479 err = nice;
1480 err = proc_sched_autogroup_set_nice(p, &err);
1481 if (err)
1482 count = err;
1483
1484 put_task_struct(p);
1485
1486 return count;
1487}
1488
1489static int sched_autogroup_open(struct inode *inode, struct file *filp)
1490{
1491 int ret;
1492
1493 ret = single_open(filp, sched_autogroup_show, NULL);
1494 if (!ret) {
1495 struct seq_file *m = filp->private_data;
1496
1497 m->private = inode;
1498 }
1499 return ret;
1500}
1501
1502static const struct file_operations proc_pid_sched_autogroup_operations = {
1503 .open = sched_autogroup_open,
1504 .read = seq_read,
1505 .write = sched_autogroup_write,
1506 .llseek = seq_lseek,
1507 .release = single_release,
1508};
1509
1510#endif /* CONFIG_SCHED_AUTOGROUP */
1511
4614a696 1512static ssize_t comm_write(struct file *file, const char __user *buf,
1513 size_t count, loff_t *offset)
1514{
1515 struct inode *inode = file->f_path.dentry->d_inode;
1516 struct task_struct *p;
1517 char buffer[TASK_COMM_LEN];
1518
1519 memset(buffer, 0, sizeof(buffer));
1520 if (count > sizeof(buffer) - 1)
1521 count = sizeof(buffer) - 1;
1522 if (copy_from_user(buffer, buf, count))
1523 return -EFAULT;
1524
1525 p = get_proc_task(inode);
1526 if (!p)
1527 return -ESRCH;
1528
1529 if (same_thread_group(current, p))
1530 set_task_comm(p, buffer);
1531 else
1532 count = -EINVAL;
1533
1534 put_task_struct(p);
1535
1536 return count;
1537}
1538
1539static int comm_show(struct seq_file *m, void *v)
1540{
1541 struct inode *inode = m->private;
1542 struct task_struct *p;
1543
1544 p = get_proc_task(inode);
1545 if (!p)
1546 return -ESRCH;
1547
1548 task_lock(p);
1549 seq_printf(m, "%s\n", p->comm);
1550 task_unlock(p);
1551
1552 put_task_struct(p);
1553
1554 return 0;
1555}
1556
1557static int comm_open(struct inode *inode, struct file *filp)
1558{
c6a34058 1559 return single_open(filp, comm_show, inode);
4614a696 1560}
1561
1562static const struct file_operations proc_pid_set_comm_operations = {
1563 .open = comm_open,
1564 .read = seq_read,
1565 .write = comm_write,
1566 .llseek = seq_lseek,
1567 .release = single_release,
1568};
1569
925d1c40
MH		 1570static int proc_exe_link(struct inode *inode, struct path *exe_path)
1571{
1572 struct task_struct *task;
1573 struct mm_struct *mm;
1574 struct file *exe_file;
1575
1576 task = get_proc_task(inode);
1577 if (!task)
1578 return -ENOENT;
1579 mm = get_task_mm(task);
1580 put_task_struct(task);
1581 if (!mm)
1582 return -ENOENT;
1583 exe_file = get_mm_exe_file(mm);
1584 mmput(mm);
1585 if (exe_file) {
1586 *exe_path = exe_file->f_path;
1587 path_get(&exe_file->f_path);
1588 fput(exe_file);
1589 return 0;
1590 } else
1591 return -ENOENT;
1592}
1593
008b150a 1594static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1da177e4
LT		 1595{
1596 struct inode *inode = dentry->d_inode;
1597 int error = -EACCES;
1598
1599 /* We don't need a base pointer in the /proc filesystem */
1d957f9b 1600 path_put(&nd->path);
1da177e4 1601
778c1144
EB		 1602 /* Are we allowed to snoop on the tasks file descriptors? */
1603 if (!proc_fd_access_allowed(inode))
1da177e4 1604 goto out;
1da177e4 1605
3dcd25f3 1606 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
1da177e4 1607out:
008b150a 1608 return ERR_PTR(error);
1da177e4
LT		 1609}
1610
3dcd25f3 1611static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1da177e4 1612{
e12ba74d 1613 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
3dcd25f3 1614 char *pathname;
1da177e4
LT		 1615 int len;
1616
1617 if (!tmp)
1618 return -ENOMEM;
0c28f287 1619
7b2a69ba 1620 pathname = d_path(path, tmp, PAGE_SIZE);
3dcd25f3
JB		 1621 len = PTR_ERR(pathname);
1622 if (IS_ERR(pathname))
1da177e4 1623 goto out;
3dcd25f3 1624 len = tmp + PAGE_SIZE - 1 - pathname;
1da177e4
LT		 1625
1626 if (len > buflen)
1627 len = buflen;
3dcd25f3 1628 if (copy_to_user(buffer, pathname, len))
1da177e4
LT		 1629 len = -EFAULT;
1630 out:
1631 free_page((unsigned long)tmp);
1632 return len;
1633}
1634
1635static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1636{
1637 int error = -EACCES;
1638 struct inode *inode = dentry->d_inode;
3dcd25f3 1639 struct path path;
1da177e4 1640
778c1144
EB		 1641 /* Are we allowed to snoop on the tasks file descriptors? */
1642 if (!proc_fd_access_allowed(inode))
1da177e4 1643 goto out;
1da177e4 1644
3dcd25f3 1645 error = PROC_I(inode)->op.proc_get_link(inode, &path);
1da177e4
LT		 1646 if (error)
1647 goto out;
1648
3dcd25f3
JB		 1649 error = do_proc_readlink(&path, buffer, buflen);
1650 path_put(&path);
1da177e4 1651out:
1da177e4
LT		 1652 return error;
1653}
1654
c5ef1c42 1655static const struct inode_operations proc_pid_link_inode_operations = {
1da177e4 1656 .readlink = proc_pid_readlink,
6d76fa58
LT		 1657 .follow_link = proc_pid_follow_link,
1658 .setattr = proc_setattr,
1da177e4
LT		 1659};
1660
28a6d671
EB		 1661
1662/* building an inode */
1663
1664static int task_dumpable(struct task_struct *task)
1da177e4 1665{
28a6d671
EB		 1666 int dumpable = 0;
1667 struct mm_struct *mm;
1da177e4 1668
28a6d671
EB		 1669 task_lock(task);
1670 mm = task->mm;
1671 if (mm)
6c5d5238 1672 dumpable = get_dumpable(mm);
28a6d671
EB		 1673 task_unlock(task);
1674 if(dumpable == 1)
1675 return 1;
1676 return 0;
1677}
1da177e4 1678
6b4e306a 1679struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
28a6d671
EB		 1680{
1681 struct inode * inode;
1682 struct proc_inode *ei;
c69e8d9c 1683 const struct cred *cred;
1da177e4 1684
28a6d671 1685 /* We need a new inode */
1da177e4 1686
28a6d671
EB		 1687 inode = new_inode(sb);
1688 if (!inode)
1689 goto out;
1690
1691 /* Common stuff */
1692 ei = PROC_I(inode);
85fe4025 1693 inode->i_ino = get_next_ino();
28a6d671 1694 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
28a6d671
EB		 1695 inode->i_op = &proc_def_inode_operations;
1696
1697 /*
1698 * grab the reference to task.
1699 */
1a657f78 1700 ei->pid = get_task_pid(task, PIDTYPE_PID);
28a6d671
EB		 1701 if (!ei->pid)
1702 goto out_unlock;
1703
28a6d671 1704 if (task_dumpable(task)) {
c69e8d9c
DH		 1705 rcu_read_lock();
1706 cred = __task_cred(task);
1707 inode->i_uid = cred->euid;
1708 inode->i_gid = cred->egid;
1709 rcu_read_unlock();
1da177e4 1710 }
28a6d671
EB		 1711 security_task_to_inode(task, inode);
1712
1da177e4 1713out:
28a6d671
EB		 1714 return inode;
1715
1716out_unlock:
1717 iput(inode);
1718 return NULL;
1da177e4
LT		 1719}
1720
6b4e306a 1721int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1da177e4 1722{
1da177e4 1723 struct inode *inode = dentry->d_inode;
28a6d671 1724 struct task_struct *task;
c69e8d9c
DH		 1725 const struct cred *cred;
1726
28a6d671 1727 generic_fillattr(inode, stat);
1da177e4 1728
28a6d671
EB		 1729 rcu_read_lock();
1730 stat->uid = 0;
1731 stat->gid = 0;
1732 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1733 if (task) {
1734 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1735 task_dumpable(task)) {
c69e8d9c
DH		 1736 cred = __task_cred(task);
1737 stat->uid = cred->euid;
1738 stat->gid = cred->egid;
1da177e4
LT		 1739 }
1740 }
28a6d671 1741 rcu_read_unlock();
d6e71144 1742 return 0;
1da177e4
LT		 1743}
1744
1da177e4
LT		 1745/* dentry stuff */
1746
1747/*
1748 * Exceptional case: normally we are not allowed to unhash a busy
1749 * directory. In this case, however, we can do it - no aliasing problems
1750 * due to the way we treat inodes.
1751 *
1752 * Rewrite the inode's ownerships here because the owning task may have
1753 * performed a setuid(), etc.
99f89551
EB		 1754 *
1755 * Before the /proc/pid/status file was created the only way to read
1756 * the effective uid of a /process was to stat /proc/pid. Reading
1757 * /proc/pid/status is slow enough that procps and other packages
1758 * kept stating /proc/pid. To keep the rules in /proc simple I have
1759 * made this apply to all per process world readable and executable
1760 * directories.
1da177e4 1761 */
6b4e306a 1762int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1da177e4 1763{
34286d66
NP		 1764 struct inode *inode;
1765 struct task_struct *task;
c69e8d9c
DH		 1766 const struct cred *cred;
1767
34286d66
NP		 1768 if (nd && nd->flags & LOOKUP_RCU)
1769 return -ECHILD;
1770
1771 inode = dentry->d_inode;
1772 task = get_proc_task(inode);
1773
99f89551
EB		 1774 if (task) {
1775 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1776 task_dumpable(task)) {
c69e8d9c
DH		 1777 rcu_read_lock();
1778 cred = __task_cred(task);
1779 inode->i_uid = cred->euid;
1780 inode->i_gid = cred->egid;
1781 rcu_read_unlock();
1da177e4
LT		 1782 } else {
1783 inode->i_uid = 0;
1784 inode->i_gid = 0;
1785 }
9ee8ab9f 1786 inode->i_mode &= ~(S_ISUID | S_ISGID);
1da177e4 1787 security_task_to_inode(task, inode);
99f89551 1788 put_task_struct(task);
1da177e4
LT		 1789 return 1;
1790 }
1791 d_drop(dentry);
1792 return 0;
1793}
1794
fe15ce44 1795static int pid_delete_dentry(const struct dentry * dentry)
99f89551 1796{
28a6d671
EB		 1797 /* Is the task we represent dead?
1798 * If so, then don't put the dentry on the lru list,
1799 * kill it immediately.
1800 */
1801 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1802}
1803
6b4e306a 1804const struct dentry_operations pid_dentry_operations =
28a6d671
EB		 1805{
1806 .d_revalidate = pid_revalidate,
1807 .d_delete = pid_delete_dentry,
1808};
1809
1810/* Lookups */
1811
1c0d04c9
EB		 1812/*
1813 * Fill a directory entry.
1814 *
1815 * If possible create the dcache entry and derive our inode number and
1816 * file type from dcache entry.
1817 *
1818 * Since all of the proc inode numbers are dynamically generated, the inode
1819 * numbers do not exist until the inode is cache. This means creating the
1820 * the dcache entry in readdir is necessary to keep the inode numbers
1821 * reported by readdir in sync with the inode numbers reported
1822 * by stat.
1823 */
6b4e306a
EB		 1824int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1825 const char *name, int len,
c5141e6d 1826 instantiate_t instantiate, struct task_struct *task, const void *ptr)
61a28784 1827{
2fddfeef 1828 struct dentry *child, *dir = filp->f_path.dentry;
61a28784
EB		 1829 struct inode *inode;
1830 struct qstr qname;
1831 ino_t ino = 0;
1832 unsigned type = DT_UNKNOWN;
1833
1834 qname.name = name;
1835 qname.len = len;
1836 qname.hash = full_name_hash(name, len);
1837
1838 child = d_lookup(dir, &qname);
1839 if (!child) {
1840 struct dentry *new;
1841 new = d_alloc(dir, &qname);
1842 if (new) {
1843 child = instantiate(dir->d_inode, new, task, ptr);
1844 if (child)
1845 dput(new);
1846 else
1847 child = new;
1848 }
1849 }
1850 if (!child || IS_ERR(child) || !child->d_inode)
1851 goto end_instantiate;
1852 inode = child->d_inode;
1853 if (inode) {
1854 ino = inode->i_ino;
1855 type = inode->i_mode >> 12;
1856 }
1857 dput(child);
1858end_instantiate:
1859 if (!ino)
1860 ino = find_inode_number(dir, &qname);
1861 if (!ino)
1862 ino = 1;
1863 return filldir(dirent, name, len, filp->f_pos, ino, type);
1864}
1865
28a6d671
EB		 1866static unsigned name_to_int(struct dentry *dentry)
1867{
1868 const char *name = dentry->d_name.name;
1869 int len = dentry->d_name.len;
1870 unsigned n = 0;
1871
1872 if (len > 1 && *name == '0')
1873 goto out;
1874 while (len-- > 0) {
1875 unsigned c = *name++ - '0';
1876 if (c > 9)
1877 goto out;
1878 if (n >= (~0U-9)/10)
1879 goto out;
1880 n *= 10;
1881 n += c;
1882 }
1883 return n;
1884out:
1885 return ~0U;
1886}
1887
27932742
MS		 1888#define PROC_FDINFO_MAX 64
1889
3dcd25f3 1890static int proc_fd_info(struct inode *inode, struct path *path, char *info)
28a6d671 1891{
5e442a49
LT		 1892 struct task_struct *task = get_proc_task(inode);
1893 struct files_struct *files = NULL;
28a6d671
EB		 1894 struct file *file;
1895 int fd = proc_fd(inode);
aa6afca5 1896
5e442a49
LT		 1897 if (task) {
1898 files = get_files_struct(task);
1899 put_task_struct(task);
1900 }
1901 if (files) {
1902 /*
1903 * We are not taking a ref to the file structure, so we must
1904 * hold ->file_lock.
1905 */
1906 spin_lock(&files->file_lock);
1907 file = fcheck_files(files, fd);
1908 if (file) {
1909 unsigned int f_flags;
1910 struct fdtable *fdt;
1911
1912 fdt = files_fdtable(files);
1913 f_flags = file->f_flags & ~O_CLOEXEC;
1914 if (FD_ISSET(fd, fdt->close_on_exec))
1915 f_flags |= O_CLOEXEC;
1916
1917 if (path) {
1918 *path = file->f_path;
1919 path_get(&file->f_path);
1920 }
1921 if (info)
1922 snprintf(info, PROC_FDINFO_MAX,
1923 "pos:\t%lli\n"
1924 "flags:\t0%o\n",
1925 (long long) file->f_pos,
1926 f_flags);
1927 spin_unlock(&files->file_lock);
1928 put_files_struct(files);
1929 return 0;
99f89551 1930 }
5e442a49
LT		 1931 spin_unlock(&files->file_lock);
1932 put_files_struct(files);
1933 }
1934 return -ENOENT;
99f89551
EB		 1935}
1936
3dcd25f3 1937static int proc_fd_link(struct inode *inode, struct path *path)
27932742 1938{
3dcd25f3 1939 return proc_fd_info(inode, path, NULL);
27932742
MS		 1940}
1941
1da177e4
LT		 1942static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1943{
34286d66
NP		 1944 struct inode *inode;
1945 struct task_struct *task;
1946 int fd;
1da177e4 1947 struct files_struct *files;
c69e8d9c 1948 const struct cred *cred;
1da177e4 1949
34286d66
NP		 1950 if (nd && nd->flags & LOOKUP_RCU)
1951 return -ECHILD;
1952
1953 inode = dentry->d_inode;
1954 task = get_proc_task(inode);
1955 fd = proc_fd(inode);
1956
99f89551
EB		 1957 if (task) {
1958 files = get_files_struct(task);
1959 if (files) {
1960 rcu_read_lock();
1961 if (fcheck_files(files, fd)) {
1962 rcu_read_unlock();
1963 put_files_struct(files);
1964 if (task_dumpable(task)) {
c69e8d9c
DH		 1965 rcu_read_lock();
1966 cred = __task_cred(task);
1967 inode->i_uid = cred->euid;
1968 inode->i_gid = cred->egid;
1969 rcu_read_unlock();
99f89551
EB		 1970 } else {
1971 inode->i_uid = 0;
1972 inode->i_gid = 0;
1973 }
9ee8ab9f 1974 inode->i_mode &= ~(S_ISUID | S_ISGID);
99f89551
EB		 1975 security_task_to_inode(task, inode);
1976 put_task_struct(task);
1977 return 1;
1978 }
b835996f 1979 rcu_read_unlock();
1da177e4 1980 put_files_struct(files);
1da177e4 1981 }
99f89551 1982 put_task_struct(task);
1da177e4
LT		 1983 }
1984 d_drop(dentry);
1985 return 0;
1986}
1987
d72f71eb 1988static const struct dentry_operations tid_fd_dentry_operations =
1da177e4
LT		 1989{
1990 .d_revalidate = tid_fd_revalidate,
1991 .d_delete = pid_delete_dentry,
1992};
1993
444ceed8 1994static struct dentry *proc_fd_instantiate(struct inode *dir,
c5141e6d 1995 struct dentry *dentry, struct task_struct *task, const void *ptr)
1da177e4 1996{
c5141e6d 1997 unsigned fd = *(const unsigned *)ptr;
444ceed8
EB		 1998 struct file *file;
1999 struct files_struct *files;
2000 struct inode *inode;
2001 struct proc_inode *ei;
2002 struct dentry *error = ERR_PTR(-ENOENT);
1da177e4 2003
61a28784 2004 inode = proc_pid_make_inode(dir->i_sb, task);
1da177e4
LT		 2005 if (!inode)
2006 goto out;
2007 ei = PROC_I(inode);
aed7a6c4 2008 ei->fd = fd;
1da177e4
LT		 2009 files = get_files_struct(task);
2010 if (!files)
444ceed8 2011 goto out_iput;
1da177e4 2012 inode->i_mode = S_IFLNK;
ca99c1da
DS		 2013
2014 /*
2015 * We are not taking a ref to the file structure, so we must
2016 * hold ->file_lock.
2017 */
2018 spin_lock(&files->file_lock);
1da177e4
LT		 2019 file = fcheck_files(files, fd);
2020 if (!file)
444ceed8 2021 goto out_unlock;
aeb5d727 2022 if (file->f_mode & FMODE_READ)
1da177e4 2023 inode->i_mode |= S_IRUSR | S_IXUSR;
aeb5d727 2024 if (file->f_mode & FMODE_WRITE)
1da177e4 2025 inode->i_mode |= S_IWUSR | S_IXUSR;
ca99c1da 2026 spin_unlock(&files->file_lock);
1da177e4 2027 put_files_struct(files);
444ceed8 2028
5e442a49 2029 inode->i_op = &proc_pid_link_inode_operations;
1da177e4
LT		 2030 inode->i_size = 64;
2031 ei->op.proc_get_link = proc_fd_link;
fb045adb 2032 d_set_d_op(dentry, &tid_fd_dentry_operations);
1da177e4 2033 d_add(dentry, inode);
cd6a3ce9
EB		 2034 /* Close the race of the process dying before we return the dentry */
2035 if (tid_fd_revalidate(dentry, NULL))
444ceed8 2036 error = NULL;
1da177e4 2037
444ceed8
EB		 2038 out:
2039 return error;
2040out_unlock:
ca99c1da 2041 spin_unlock(&files->file_lock);
1da177e4 2042 put_files_struct(files);
444ceed8 2043out_iput:
1da177e4 2044 iput(inode);
cd6a3ce9 2045 goto out;
1da177e4
LT		 2046}
2047
27932742
MS		 2048static struct dentry *proc_lookupfd_common(struct inode *dir,
2049 struct dentry *dentry,
2050 instantiate_t instantiate)
444ceed8
EB		 2051{
2052 struct task_struct *task = get_proc_task(dir);
2053 unsigned fd = name_to_int(dentry);
2054 struct dentry *result = ERR_PTR(-ENOENT);
2055
2056 if (!task)
2057 goto out_no_task;
2058 if (fd == ~0U)
2059 goto out;
2060
27932742 2061 result = instantiate(dir, dentry, task, &fd);
444ceed8
EB		 2062out:
2063 put_task_struct(task);
2064out_no_task:
2065 return result;
2066}
2067
27932742
MS		 2068static int proc_readfd_common(struct file * filp, void * dirent,
2069 filldir_t filldir, instantiate_t instantiate)
28a6d671 2070{
2fddfeef 2071 struct dentry *dentry = filp->f_path.dentry;
28a6d671
EB		 2072 struct inode *inode = dentry->d_inode;
2073 struct task_struct *p = get_proc_task(inode);
457c2510 2074 unsigned int fd, ino;
28a6d671 2075 int retval;
28a6d671 2076 struct files_struct * files;
1da177e4 2077
28a6d671
EB		 2078 retval = -ENOENT;
2079 if (!p)
2080 goto out_no_task;
2081 retval = 0;
28a6d671
EB		 2082
2083 fd = filp->f_pos;
2084 switch (fd) {
2085 case 0:
2086 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
5e442a49 2087 goto out;
28a6d671
EB		 2088 filp->f_pos++;
2089 case 1:
2090 ino = parent_ino(dentry);
2091 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
5e442a49 2092 goto out;
28a6d671
EB		 2093 filp->f_pos++;
2094 default:
2095 files = get_files_struct(p);
2096 if (!files)
5e442a49 2097 goto out;
28a6d671 2098 rcu_read_lock();
28a6d671 2099 for (fd = filp->f_pos-2;
9b4f526c 2100 fd < files_fdtable(files)->max_fds;
28a6d671 2101 fd++, filp->f_pos++) {
27932742
MS		 2102 char name[PROC_NUMBUF];
2103 int len;
28a6d671
EB		 2104
2105 if (!fcheck_files(files, fd))
2106 continue;
2107 rcu_read_unlock();
2108
27932742
MS		 2109 len = snprintf(name, sizeof(name), "%d", fd);
2110 if (proc_fill_cache(filp, dirent, filldir,
2111 name, len, instantiate,
2112 p, &fd) < 0) {
28a6d671
EB		 2113 rcu_read_lock();
2114 break;
2115 }
2116 rcu_read_lock();
2117 }
2118 rcu_read_unlock();
2119 put_files_struct(files);
2120 }
2121out:
2122 put_task_struct(p);
2123out_no_task:
2124 return retval;
2125}
2126
27932742
MS		 2127static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
2128 struct nameidata *nd)
2129{
2130 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
2131}
2132
2133static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
2134{
2135 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
2136}
2137
2138static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
2139 size_t len, loff_t *ppos)
2140{
2141 char tmp[PROC_FDINFO_MAX];
3dcd25f3 2142 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
27932742
MS		 2143 if (!err)
2144 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
2145 return err;
2146}
2147
2148static const struct file_operations proc_fdinfo_file_operations = {
87df8424 2149 .open = nonseekable_open,
27932742 2150 .read = proc_fdinfo_read,
6038f373 2151 .llseek = no_llseek,
27932742
MS		 2152};
2153
00977a59 2154static const struct file_operations proc_fd_operations = {
28a6d671
EB		 2155 .read = generic_read_dir,
2156 .readdir = proc_readfd,
6038f373 2157 .llseek = default_llseek,
1da177e4
LT		 2158};
2159
8948e11f
AD		 2160/*
2161 * /proc/pid/fd needs a special permission handler so that a process can still
2162 * access /proc/self/fd after it has executed a setuid().
2163 */
10556cb2 2164static int proc_fd_permission(struct inode *inode, int mask)
8948e11f 2165{
2830ba7f 2166 int rv = generic_permission(inode, mask);
8948e11f
AD		 2167 if (rv == 0)
2168 return 0;
2169 if (task_pid(current) == proc_pid(inode))
2170 rv = 0;
2171 return rv;
2172}
2173
1da177e4
LT		 2174/*
2175 * proc directories can do almost nothing..
2176 */
c5ef1c42 2177static const struct inode_operations proc_fd_inode_operations = {
1da177e4 2178 .lookup = proc_lookupfd,
8948e11f 2179 .permission = proc_fd_permission,
6d76fa58 2180 .setattr = proc_setattr,
1da177e4
LT		 2181};
2182
27932742
MS		 2183static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2184 struct dentry *dentry, struct task_struct *task, const void *ptr)
2185{
2186 unsigned fd = *(unsigned *)ptr;
2187 struct inode *inode;
2188 struct proc_inode *ei;
2189 struct dentry *error = ERR_PTR(-ENOENT);
2190
2191 inode = proc_pid_make_inode(dir->i_sb, task);
2192 if (!inode)
2193 goto out;
2194 ei = PROC_I(inode);
2195 ei->fd = fd;
2196 inode->i_mode = S_IFREG | S_IRUSR;
2197 inode->i_fop = &proc_fdinfo_file_operations;
fb045adb 2198 d_set_d_op(dentry, &tid_fd_dentry_operations);
27932742
MS		 2199 d_add(dentry, inode);
2200 /* Close the race of the process dying before we return the dentry */
2201 if (tid_fd_revalidate(dentry, NULL))
2202 error = NULL;
2203
2204 out:
2205 return error;
2206}
2207
2208static struct dentry *proc_lookupfdinfo(struct inode *dir,
2209 struct dentry *dentry,
2210 struct nameidata *nd)
2211{
2212 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2213}
2214
2215static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2216{
2217 return proc_readfd_common(filp, dirent, filldir,
2218 proc_fdinfo_instantiate);
2219}
2220
2221static const struct file_operations proc_fdinfo_operations = {
2222 .read = generic_read_dir,
2223 .readdir = proc_readfdinfo,
6038f373 2224 .llseek = default_llseek,
27932742
MS		 2225};
2226
2227/*
2228 * proc directories can do almost nothing..
2229 */
2230static const struct inode_operations proc_fdinfo_inode_operations = {
2231 .lookup = proc_lookupfdinfo,
2232 .setattr = proc_setattr,
2233};
2234
2235
444ceed8 2236static struct dentry *proc_pident_instantiate(struct inode *dir,
c5141e6d 2237 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8 2238{
c5141e6d 2239 const struct pid_entry *p = ptr;
444ceed8
EB		 2240 struct inode *inode;
2241 struct proc_inode *ei;
bd6daba9 2242 struct dentry *error = ERR_PTR(-ENOENT);
444ceed8 2243
61a28784 2244 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 2245 if (!inode)
2246 goto out;
2247
2248 ei = PROC_I(inode);
2249 inode->i_mode = p->mode;
2250 if (S_ISDIR(inode->i_mode))
bfe86848 2251 set_nlink(inode, 2); /* Use getattr to fix if necessary */
444ceed8
EB		 2252 if (p->iop)
2253 inode->i_op = p->iop;
2254 if (p->fop)
2255 inode->i_fop = p->fop;
2256 ei->op = p->op;
fb045adb 2257 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 2258 d_add(dentry, inode);
2259 /* Close the race of the process dying before we return the dentry */
2260 if (pid_revalidate(dentry, NULL))
2261 error = NULL;
2262out:
2263 return error;
2264}
2265
1da177e4
LT		 2266static struct dentry *proc_pident_lookup(struct inode *dir,
2267 struct dentry *dentry,
c5141e6d 2268 const struct pid_entry *ents,
7bcd6b0e 2269 unsigned int nents)
1da177e4 2270{
cd6a3ce9 2271 struct dentry *error;
99f89551 2272 struct task_struct *task = get_proc_task(dir);
c5141e6d 2273 const struct pid_entry *p, *last;
1da177e4 2274
cd6a3ce9 2275 error = ERR_PTR(-ENOENT);
1da177e4 2276
99f89551
EB		 2277 if (!task)
2278 goto out_no_task;
1da177e4 2279
20cdc894
EB		 2280 /*
2281 * Yes, it does not scale. And it should not. Don't add
2282 * new entries into /proc/<tgid>/ without very good reasons.
2283 */
7bcd6b0e
EB		 2284 last = &ents[nents - 1];
2285 for (p = ents; p <= last; p++) {
1da177e4
LT		 2286 if (p->len != dentry->d_name.len)
2287 continue;
2288 if (!memcmp(dentry->d_name.name, p->name, p->len))
2289 break;
2290 }
7bcd6b0e 2291 if (p > last)
1da177e4
LT		 2292 goto out;
2293
444ceed8 2294 error = proc_pident_instantiate(dir, dentry, task, p);
1da177e4 2295out:
99f89551
EB		 2296 put_task_struct(task);
2297out_no_task:
cd6a3ce9 2298 return error;
1da177e4
LT		 2299}
2300
c5141e6d
ED		 2301static int proc_pident_fill_cache(struct file *filp, void *dirent,
2302 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
61a28784
EB		 2303{
2304 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2305 proc_pident_instantiate, task, p);
2306}
2307
28a6d671
EB		 2308static int proc_pident_readdir(struct file *filp,
2309 void *dirent, filldir_t filldir,
c5141e6d 2310 const struct pid_entry *ents, unsigned int nents)
28a6d671
EB		 2311{
2312 int i;
2fddfeef 2313 struct dentry *dentry = filp->f_path.dentry;
28a6d671
EB		 2314 struct inode *inode = dentry->d_inode;
2315 struct task_struct *task = get_proc_task(inode);
c5141e6d 2316 const struct pid_entry *p, *last;
28a6d671
EB		 2317 ino_t ino;
2318 int ret;
2319
2320 ret = -ENOENT;
2321 if (!task)
61a28784 2322 goto out_no_task;
28a6d671
EB		 2323
2324 ret = 0;
28a6d671
EB		 2325 i = filp->f_pos;
2326 switch (i) {
2327 case 0:
2328 ino = inode->i_ino;
2329 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2330 goto out;
2331 i++;
2332 filp->f_pos++;
2333 /* fall through */
2334 case 1:
2335 ino = parent_ino(dentry);
2336 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2337 goto out;
2338 i++;
2339 filp->f_pos++;
2340 /* fall through */
2341 default:
2342 i -= 2;
2343 if (i >= nents) {
2344 ret = 1;
2345 goto out;
2346 }
2347 p = ents + i;
7bcd6b0e
EB		 2348 last = &ents[nents - 1];
2349 while (p <= last) {
61a28784 2350 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
28a6d671
EB		 2351 goto out;
2352 filp->f_pos++;
2353 p++;
2354 }
2355 }
2356
2357 ret = 1;
2358out:
61a28784
EB		 2359 put_task_struct(task);
2360out_no_task:
28a6d671 2361 return ret;
1da177e4
LT		 2362}
2363
28a6d671
EB		 2364#ifdef CONFIG_SECURITY
2365static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2366 size_t count, loff_t *ppos)
2367{
2fddfeef 2368 struct inode * inode = file->f_path.dentry->d_inode;
04ff9708 2369 char *p = NULL;
28a6d671
EB		 2370 ssize_t length;
2371 struct task_struct *task = get_proc_task(inode);
2372
28a6d671 2373 if (!task)
04ff9708 2374 return -ESRCH;
28a6d671
EB		 2375
2376 length = security_getprocattr(task,
2fddfeef 2377 (char*)file->f_path.dentry->d_name.name,
04ff9708 2378 &p);
28a6d671 2379 put_task_struct(task);
04ff9708
AV		 2380 if (length > 0)
2381 length = simple_read_from_buffer(buf, count, ppos, p, length);
2382 kfree(p);
28a6d671 2383 return length;
1da177e4
LT		 2384}
2385
28a6d671
EB		 2386static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2387 size_t count, loff_t *ppos)
2388{
2fddfeef 2389 struct inode * inode = file->f_path.dentry->d_inode;
28a6d671
EB		 2390 char *page;
2391 ssize_t length;
2392 struct task_struct *task = get_proc_task(inode);
2393
2394 length = -ESRCH;
2395 if (!task)
2396 goto out_no_task;
2397 if (count > PAGE_SIZE)
2398 count = PAGE_SIZE;
2399
2400 /* No partial writes. */
2401 length = -EINVAL;
2402 if (*ppos != 0)
2403 goto out;
2404
2405 length = -ENOMEM;
e12ba74d 2406 page = (char*)__get_free_page(GFP_TEMPORARY);
28a6d671
EB		 2407 if (!page)
2408 goto out;
2409
2410 length = -EFAULT;
2411 if (copy_from_user(page, buf, count))
2412 goto out_free;
2413
107db7c7 2414 /* Guard against adverse ptrace interaction */
9b1bf12d 2415 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
107db7c7
DH		 2416 if (length < 0)
2417 goto out_free;
2418
28a6d671 2419 length = security_setprocattr(task,
2fddfeef 2420 (char*)file->f_path.dentry->d_name.name,
28a6d671 2421 (void*)page, count);
9b1bf12d 2422 mutex_unlock(&task->signal->cred_guard_mutex);
28a6d671
EB		 2423out_free:
2424 free_page((unsigned long) page);
2425out:
2426 put_task_struct(task);
2427out_no_task:
2428 return length;
2429}
2430
00977a59 2431static const struct file_operations proc_pid_attr_operations = {
28a6d671
EB		 2432 .read = proc_pid_attr_read,
2433 .write = proc_pid_attr_write,
87df8424 2434 .llseek = generic_file_llseek,
28a6d671
EB		 2435};
2436
c5141e6d 2437static const struct pid_entry attr_dir_stuff[] = {
631f9c18
AD		 2438 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2439 REG("prev", S_IRUGO, proc_pid_attr_operations),
2440 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2441 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2442 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2443 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
28a6d671
EB		 2444};
2445
72d9dcfc 2446static int proc_attr_dir_readdir(struct file * filp,
28a6d671
EB		 2447 void * dirent, filldir_t filldir)
2448{
2449 return proc_pident_readdir(filp,dirent,filldir,
72d9dcfc 2450 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2451}
2452
00977a59 2453static const struct file_operations proc_attr_dir_operations = {
1da177e4 2454 .read = generic_read_dir,
72d9dcfc 2455 .readdir = proc_attr_dir_readdir,
6038f373 2456 .llseek = default_llseek,
1da177e4
LT		 2457};
2458
72d9dcfc 2459static struct dentry *proc_attr_dir_lookup(struct inode *dir,
28a6d671
EB		 2460 struct dentry *dentry, struct nameidata *nd)
2461{
7bcd6b0e
EB		 2462 return proc_pident_lookup(dir, dentry,
2463 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2464}
2465
c5ef1c42 2466static const struct inode_operations proc_attr_dir_inode_operations = {
72d9dcfc 2467 .lookup = proc_attr_dir_lookup,
99f89551 2468 .getattr = pid_getattr,
6d76fa58 2469 .setattr = proc_setattr,
1da177e4
LT		 2470};
2471
28a6d671
EB		 2472#endif
2473
698ba7b5 2474#ifdef CONFIG_ELF_CORE
3cb4a0bb
KH		 2475static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2476 size_t count, loff_t *ppos)
2477{
2478 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2479 struct mm_struct *mm;
2480 char buffer[PROC_NUMBUF];
2481 size_t len;
2482 int ret;
2483
2484 if (!task)
2485 return -ESRCH;
2486
2487 ret = 0;
2488 mm = get_task_mm(task);
2489 if (mm) {
2490 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2491 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2492 MMF_DUMP_FILTER_SHIFT));
2493 mmput(mm);
2494 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2495 }
2496
2497 put_task_struct(task);
2498
2499 return ret;
2500}
2501
2502static ssize_t proc_coredump_filter_write(struct file *file,
2503 const char __user *buf,
2504 size_t count,
2505 loff_t *ppos)
2506{
2507 struct task_struct *task;
2508 struct mm_struct *mm;
2509 char buffer[PROC_NUMBUF], *end;
2510 unsigned int val;
2511 int ret;
2512 int i;
2513 unsigned long mask;
2514
2515 ret = -EFAULT;
2516 memset(buffer, 0, sizeof(buffer));
2517 if (count > sizeof(buffer) - 1)
2518 count = sizeof(buffer) - 1;
2519 if (copy_from_user(buffer, buf, count))
2520 goto out_no_task;
2521
2522 ret = -EINVAL;
2523 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2524 if (*end == '\n')
2525 end++;
2526 if (end - buffer == 0)
2527 goto out_no_task;
2528
2529 ret = -ESRCH;
2530 task = get_proc_task(file->f_dentry->d_inode);
2531 if (!task)
2532 goto out_no_task;
2533
2534 ret = end - buffer;
2535 mm = get_task_mm(task);
2536 if (!mm)
2537 goto out_no_mm;
2538
2539 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2540 if (val & mask)
2541 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2542 else
2543 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2544 }
2545
2546 mmput(mm);
2547 out_no_mm:
2548 put_task_struct(task);
2549 out_no_task:
2550 return ret;
2551}
2552
2553static const struct file_operations proc_coredump_filter_operations = {
2554 .read = proc_coredump_filter_read,
2555 .write = proc_coredump_filter_write,
87df8424 2556 .llseek = generic_file_llseek,
3cb4a0bb
KH		 2557};
2558#endif
2559
28a6d671
EB		 2560/*
2561 * /proc/self:
2562 */
2563static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2564 int buflen)
2565{
488e5bc4 2566 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
b55fcb22 2567 pid_t tgid = task_tgid_nr_ns(current, ns);
28a6d671 2568 char tmp[PROC_NUMBUF];
b55fcb22 2569 if (!tgid)
488e5bc4 2570 return -ENOENT;
b55fcb22 2571 sprintf(tmp, "%d", tgid);
28a6d671
EB		 2572 return vfs_readlink(dentry,buffer,buflen,tmp);
2573}
2574
2575static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2576{
488e5bc4 2577 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
b55fcb22 2578 pid_t tgid = task_tgid_nr_ns(current, ns);
7fee4868
AV		 2579 char *name = ERR_PTR(-ENOENT);
2580 if (tgid) {
2581 name = __getname();
2582 if (!name)
2583 name = ERR_PTR(-ENOMEM);
2584 else
2585 sprintf(name, "%d", tgid);
2586 }
2587 nd_set_link(nd, name);
2588 return NULL;
2589}
2590
2591static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2592 void *cookie)
2593{
2594 char *s = nd_get_link(nd);
2595 if (!IS_ERR(s))
2596 __putname(s);
28a6d671
EB		 2597}
2598
c5ef1c42 2599static const struct inode_operations proc_self_inode_operations = {
28a6d671
EB		 2600 .readlink = proc_self_readlink,
2601 .follow_link = proc_self_follow_link,
7fee4868 2602 .put_link = proc_self_put_link,
28a6d671
EB		 2603};
2604
801199ce
EB		 2605/*
2606 * proc base
2607 *
2608 * These are the directory entries in the root directory of /proc
2609 * that properly belong to the /proc filesystem, as they describe
2610 * describe something that is process related.
2611 */
c5141e6d 2612static const struct pid_entry proc_base_stuff[] = {
61a28784 2613 NOD("self", S_IFLNK|S_IRWXUGO,
801199ce 2614 &proc_self_inode_operations, NULL, {}),
801199ce
EB		 2615};
2616
444ceed8 2617static struct dentry *proc_base_instantiate(struct inode *dir,
c5141e6d 2618 struct dentry *dentry, struct task_struct *task, const void *ptr)
801199ce 2619{
c5141e6d 2620 const struct pid_entry *p = ptr;
801199ce 2621 struct inode *inode;
801199ce 2622 struct proc_inode *ei;
73d36460 2623 struct dentry *error;
801199ce
EB		 2624
2625 /* Allocate the inode */
2626 error = ERR_PTR(-ENOMEM);
2627 inode = new_inode(dir->i_sb);
2628 if (!inode)
2629 goto out;
2630
2631 /* Initialize the inode */
2632 ei = PROC_I(inode);
85fe4025 2633 inode->i_ino = get_next_ino();
801199ce 2634 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
801199ce
EB		 2635
2636 /*
2637 * grab the reference to the task.
2638 */
1a657f78 2639 ei->pid = get_task_pid(task, PIDTYPE_PID);
801199ce
EB		 2640 if (!ei->pid)
2641 goto out_iput;
2642
801199ce
EB		 2643 inode->i_mode = p->mode;
2644 if (S_ISDIR(inode->i_mode))
bfe86848 2645 set_nlink(inode, 2);
801199ce
EB		 2646 if (S_ISLNK(inode->i_mode))
2647 inode->i_size = 64;
2648 if (p->iop)
2649 inode->i_op = p->iop;
2650 if (p->fop)
2651 inode->i_fop = p->fop;
2652 ei->op = p->op;
801199ce
EB		 2653 d_add(dentry, inode);
2654 error = NULL;
2655out:
801199ce
EB		 2656 return error;
2657out_iput:
2658 iput(inode);
2659 goto out;
2660}
2661
444ceed8
EB		 2662static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2663{
2664 struct dentry *error;
2665 struct task_struct *task = get_proc_task(dir);
c5141e6d 2666 const struct pid_entry *p, *last;
444ceed8
EB		 2667
2668 error = ERR_PTR(-ENOENT);
2669
2670 if (!task)
2671 goto out_no_task;
2672
2673 /* Lookup the directory entry */
7bcd6b0e
EB		 2674 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2675 for (p = proc_base_stuff; p <= last; p++) {
444ceed8
EB		 2676 if (p->len != dentry->d_name.len)
2677 continue;
2678 if (!memcmp(dentry->d_name.name, p->name, p->len))
2679 break;
2680 }
7bcd6b0e 2681 if (p > last)
444ceed8
EB		 2682 goto out;
2683
2684 error = proc_base_instantiate(dir, dentry, task, p);
2685
2686out:
2687 put_task_struct(task);
2688out_no_task:
2689 return error;
2690}
2691
c5141e6d
ED		 2692static int proc_base_fill_cache(struct file *filp, void *dirent,
2693 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
61a28784
EB		 2694{
2695 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2696 proc_base_instantiate, task, p);
2697}
2698
aba76fdb 2699#ifdef CONFIG_TASK_IO_ACCOUNTING
297c5d92
AR		 2700static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2701{
940389b8 2702 struct task_io_accounting acct = task->ioac;
5995477a 2703 unsigned long flags;
293eb1e7 2704 int result;
5995477a 2705
293eb1e7
VK		 2706 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2707 if (result)
2708 return result;
2709
2710 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2711 result = -EACCES;
2712 goto out_unlock;
2713 }
1d1221f3 2714
5995477a
AR		 2715 if (whole && lock_task_sighand(task, &flags)) {
2716 struct task_struct *t = task;
2717
2718 task_io_accounting_add(&acct, &task->signal->ioac);
2719 while_each_thread(task, t)
2720 task_io_accounting_add(&acct, &t->ioac);
2721
2722 unlock_task_sighand(task, &flags);
297c5d92 2723 }
293eb1e7 2724 result = sprintf(buffer,
aba76fdb
AM		 2725 "rchar: %llu\n"
2726 "wchar: %llu\n"
2727 "syscr: %llu\n"
2728 "syscw: %llu\n"
2729 "read_bytes: %llu\n"
2730 "write_bytes: %llu\n"
2731 "cancelled_write_bytes: %llu\n",
7c44319d
AB		 2732 (unsigned long long)acct.rchar,
2733 (unsigned long long)acct.wchar,
2734 (unsigned long long)acct.syscr,
2735 (unsigned long long)acct.syscw,
2736 (unsigned long long)acct.read_bytes,
2737 (unsigned long long)acct.write_bytes,
2738 (unsigned long long)acct.cancelled_write_bytes);
293eb1e7
VK		 2739out_unlock:
2740 mutex_unlock(&task->signal->cred_guard_mutex);
2741 return result;
297c5d92
AR		 2742}
2743
2744static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2745{
2746 return do_io_accounting(task, buffer, 0);
aba76fdb 2747}
297c5d92
AR		 2748
2749static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2750{
2751 return do_io_accounting(task, buffer, 1);
2752}
2753#endif /* CONFIG_TASK_IO_ACCOUNTING */
aba76fdb 2754
47830723
KC		 2755static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2756 struct pid *pid, struct task_struct *task)
2757{
a9712bc1
AV		 2758 int err = lock_trace(task);
2759 if (!err) {
2760 seq_printf(m, "%08x\n", task->personality);
2761 unlock_trace(task);
2762 }
2763 return err;
47830723
KC		 2764}
2765
28a6d671
EB		 2766/*
2767 * Thread groups
2768 */
00977a59 2769static const struct file_operations proc_task_operations;
c5ef1c42 2770static const struct inode_operations proc_task_inode_operations;
20cdc894 2771
c5141e6d 2772static const struct pid_entry tgid_base_stuff[] = {
631f9c18
AD		 2773 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2774 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2775 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 2776 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
b2211a36 2777#ifdef CONFIG_NET
631f9c18 2778 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
b2211a36 2779#endif
631f9c18
AD		 2780 REG("environ", S_IRUSR, proc_environ_operations),
2781 INF("auxv", S_IRUSR, proc_pid_auxv),
2782 ONE("status", S_IRUGO, proc_pid_status),
a9712bc1 2783 ONE("personality", S_IRUGO, proc_pid_personality),
3036e7b4 2784 INF("limits", S_IRUGO, proc_pid_limits),
43ae34cb 2785#ifdef CONFIG_SCHED_DEBUG
631f9c18 2786 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
5091faa4
MG		 2787#endif
2788#ifdef CONFIG_SCHED_AUTOGROUP
2789 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
ebcb6734 2790#endif
4614a696 2791 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 2792#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
a9712bc1 2793 INF("syscall", S_IRUGO, proc_pid_syscall),
43ae34cb 2794#endif
631f9c18
AD		 2795 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2796 ONE("stat", S_IRUGO, proc_tgid_stat),
2797 ONE("statm", S_IRUGO, proc_pid_statm),
2798 REG("maps", S_IRUGO, proc_maps_operations),
28a6d671 2799#ifdef CONFIG_NUMA
631f9c18 2800 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
28a6d671 2801#endif
631f9c18
AD		 2802 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2803 LNK("cwd", proc_cwd_link),
2804 LNK("root", proc_root_link),
2805 LNK("exe", proc_exe_link),
2806 REG("mounts", S_IRUGO, proc_mounts_operations),
2807 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2808 REG("mountstats", S_IRUSR, proc_mountstats_operations),
1e883281 2809#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18
AD		 2810 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2811 REG("smaps", S_IRUGO, proc_smaps_operations),
ca6b0bf0 2812 REG("pagemap", S_IRUGO, proc_pagemap_operations),
28a6d671
EB		 2813#endif
2814#ifdef CONFIG_SECURITY
631f9c18 2815 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 2816#endif
2817#ifdef CONFIG_KALLSYMS
631f9c18 2818 INF("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 2819#endif
2ec220e2 2820#ifdef CONFIG_STACKTRACE
a9712bc1 2821 ONE("stack", S_IRUGO, proc_pid_stack),
28a6d671
EB		 2822#endif
2823#ifdef CONFIG_SCHEDSTATS
631f9c18 2824 INF("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 2825#endif
9745512c 2826#ifdef CONFIG_LATENCYTOP
631f9c18 2827 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 2828#endif
8793d854 2829#ifdef CONFIG_PROC_PID_CPUSET
631f9c18 2830 REG("cpuset", S_IRUGO, proc_cpuset_operations),
a424316c
PM		 2831#endif
2832#ifdef CONFIG_CGROUPS
631f9c18 2833 REG("cgroup", S_IRUGO, proc_cgroup_operations),
28a6d671 2834#endif
631f9c18
AD		 2835 INF("oom_score", S_IRUGO, proc_oom_score),
2836 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
a63d83f4 2837 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 2838#ifdef CONFIG_AUDITSYSCALL
631f9c18
AD		 2839 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2840 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 2841#endif
f4f154fd 2842#ifdef CONFIG_FAULT_INJECTION
631f9c18 2843 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 2844#endif
698ba7b5 2845#ifdef CONFIG_ELF_CORE
631f9c18 2846 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3cb4a0bb 2847#endif
aba76fdb 2848#ifdef CONFIG_TASK_IO_ACCOUNTING
1d1221f3 2849 INF("io", S_IRUSR, proc_tgid_io_accounting),
aba76fdb 2850#endif
f133ecca
CM		 2851#ifdef CONFIG_HARDWALL
2852 INF("hardwall", S_IRUGO, proc_pid_hardwall),
2853#endif
28a6d671 2854};
1da177e4 2855
28a6d671 2856static int proc_tgid_base_readdir(struct file * filp,
1da177e4
LT		 2857 void * dirent, filldir_t filldir)
2858{
2859 return proc_pident_readdir(filp,dirent,filldir,
28a6d671 2860 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 2861}
2862
00977a59 2863static const struct file_operations proc_tgid_base_operations = {
1da177e4 2864 .read = generic_read_dir,
28a6d671 2865 .readdir = proc_tgid_base_readdir,
6038f373 2866 .llseek = default_llseek,
1da177e4
LT		 2867};
2868
28a6d671 2869static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
7bcd6b0e
EB		 2870 return proc_pident_lookup(dir, dentry,
2871 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 2872}
2873
c5ef1c42 2874static const struct inode_operations proc_tgid_base_inode_operations = {
28a6d671 2875 .lookup = proc_tgid_base_lookup,
99f89551 2876 .getattr = pid_getattr,
6d76fa58 2877 .setattr = proc_setattr,
1da177e4 2878};
1da177e4 2879
60347f67 2880static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
1da177e4 2881{
48e6484d 2882 struct dentry *dentry, *leader, *dir;
8578cea7 2883 char buf[PROC_NUMBUF];
48e6484d
EB		 2884 struct qstr name;
2885
2886 name.name = buf;
60347f67
PE		 2887 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2888 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d 2889 if (dentry) {
29f12ca3 2890 shrink_dcache_parent(dentry);
48e6484d
EB		 2891 d_drop(dentry);
2892 dput(dentry);
2893 }
1da177e4 2894
48e6484d 2895 name.name = buf;
60347f67
PE		 2896 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2897 leader = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d
EB		 2898 if (!leader)
2899 goto out;
1da177e4 2900
48e6484d
EB		 2901 name.name = "task";
2902 name.len = strlen(name.name);
2903 dir = d_hash_and_lookup(leader, &name);
2904 if (!dir)
2905 goto out_put_leader;
2906
2907 name.name = buf;
60347f67 2908 name.len = snprintf(buf, sizeof(buf), "%d", pid);
48e6484d
EB		 2909 dentry = d_hash_and_lookup(dir, &name);
2910 if (dentry) {
2911 shrink_dcache_parent(dentry);
2912 d_drop(dentry);
2913 dput(dentry);
1da177e4 2914 }
48e6484d
EB		 2915
2916 dput(dir);
2917out_put_leader:
2918 dput(leader);
2919out:
2920 return;
1da177e4
LT		 2921}
2922
0895e91d
RD		 2923/**
2924 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2925 * @task: task that should be flushed.
2926 *
2927 * When flushing dentries from proc, one needs to flush them from global
60347f67 2928 * proc (proc_mnt) and from all the namespaces' procs this task was seen
0895e91d
RD		 2929 * in. This call is supposed to do all of this job.
2930 *
2931 * Looks in the dcache for
2932 * /proc/@pid
2933 * /proc/@tgid/task/@pid
2934 * if either directory is present flushes it and all of it'ts children
2935 * from the dcache.
2936 *
2937 * It is safe and reasonable to cache /proc entries for a task until
2938 * that task exits. After that they just clog up the dcache with
2939 * useless entries, possibly causing useful dcache entries to be
2940 * flushed instead. This routine is proved to flush those useless
2941 * dcache entries at process exit time.
2942 *
2943 * NOTE: This routine is just an optimization so it does not guarantee
2944 * that no dcache entries will exist at process exit time it
2945 * just makes it very unlikely that any will persist.
60347f67
PE		 2946 */
2947
2948void proc_flush_task(struct task_struct *task)
2949{
9fcc2d15 2950 int i;
9b4d1cbe 2951 struct pid *pid, *tgid;
130f77ec
PE		 2952 struct upid *upid;
2953
130f77ec 2954 pid = task_pid(task);
9b4d1cbe 2955 tgid = task_tgid(task);
130f77ec 2956
9fcc2d15 2957 for (i = 0; i <= pid->level; i++) {
130f77ec
PE		 2958 upid = &pid->numbers[i];
2959 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
9b4d1cbe 2960 tgid->numbers[i].nr);
130f77ec 2961 }
6f4e6433
PE		 2962
2963 upid = &pid->numbers[pid->level];
2964 if (upid->nr == 1)
2965 pid_ns_release_proc(upid->ns);
60347f67
PE		 2966}
2967
9711ef99
AB		 2968static struct dentry *proc_pid_instantiate(struct inode *dir,
2969 struct dentry * dentry,
c5141e6d 2970 struct task_struct *task, const void *ptr)
444ceed8
EB		 2971{
2972 struct dentry *error = ERR_PTR(-ENOENT);
2973 struct inode *inode;
2974
61a28784 2975 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 2976 if (!inode)
2977 goto out;
2978
2979 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2980 inode->i_op = &proc_tgid_base_inode_operations;
2981 inode->i_fop = &proc_tgid_base_operations;
2982 inode->i_flags|=S_IMMUTABLE;
aed54175 2983
bfe86848
MS		 2984 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
2985 ARRAY_SIZE(tgid_base_stuff)));
444ceed8 2986
fb045adb 2987 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 2988
2989 d_add(dentry, inode);
2990 /* Close the race of the process dying before we return the dentry */
2991 if (pid_revalidate(dentry, NULL))
2992 error = NULL;
2993out:
2994 return error;
2995}
2996
1da177e4
LT		 2997struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2998{
73d36460 2999 struct dentry *result;
1da177e4 3000 struct task_struct *task;
1da177e4 3001 unsigned tgid;
b488893a 3002 struct pid_namespace *ns;
1da177e4 3003
801199ce
EB		 3004 result = proc_base_lookup(dir, dentry);
3005 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3006 goto out;
3007
1da177e4
LT		 3008 tgid = name_to_int(dentry);
3009 if (tgid == ~0U)
3010 goto out;
3011
b488893a 3012 ns = dentry->d_sb->s_fs_info;
de758734 3013 rcu_read_lock();
b488893a 3014 task = find_task_by_pid_ns(tgid, ns);
1da177e4
LT		 3015 if (task)
3016 get_task_struct(task);
de758734 3017 rcu_read_unlock();
1da177e4
LT		 3018 if (!task)
3019 goto out;
3020
444ceed8 3021 result = proc_pid_instantiate(dir, dentry, task, NULL);
1da177e4 3022 put_task_struct(task);
1da177e4 3023out:
cd6a3ce9 3024 return result;
1da177e4
LT		 3025}
3026
1da177e4 3027/*
0804ef4b 3028 * Find the first task with tgid >= tgid
0bc58a91 3029 *
1da177e4 3030 */
19fd4bb2
EB		 3031struct tgid_iter {
3032 unsigned int tgid;
0804ef4b 3033 struct task_struct *task;
19fd4bb2
EB		 3034};
3035static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3036{
0804ef4b 3037 struct pid *pid;
1da177e4 3038
19fd4bb2
EB		 3039 if (iter.task)
3040 put_task_struct(iter.task);
454cc105 3041 rcu_read_lock();
0804ef4b 3042retry:
19fd4bb2
EB		 3043 iter.task = NULL;
3044 pid = find_ge_pid(iter.tgid, ns);
0804ef4b 3045 if (pid) {
19fd4bb2
EB		 3046 iter.tgid = pid_nr_ns(pid, ns);
3047 iter.task = pid_task(pid, PIDTYPE_PID);
0804ef4b
EB		 3048 /* What we to know is if the pid we have find is the
3049 * pid of a thread_group_leader. Testing for task
3050 * being a thread_group_leader is the obvious thing
3051 * todo but there is a window when it fails, due to
3052 * the pid transfer logic in de_thread.
3053 *
3054 * So we perform the straight forward test of seeing
3055 * if the pid we have found is the pid of a thread
3056 * group leader, and don't worry if the task we have
3057 * found doesn't happen to be a thread group leader.
3058 * As we don't care in the case of readdir.
3059 */
19fd4bb2
EB		 3060 if (!iter.task || !has_group_leader_pid(iter.task)) {
3061 iter.tgid += 1;
0804ef4b 3062 goto retry;
19fd4bb2
EB		 3063 }
3064 get_task_struct(iter.task);
0bc58a91 3065 }
454cc105 3066 rcu_read_unlock();
19fd4bb2 3067 return iter;
1da177e4
LT		 3068}
3069
7bcd6b0e 3070#define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
0804ef4b 3071
61a28784 3072static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
19fd4bb2 3073 struct tgid_iter iter)
61a28784
EB		 3074{
3075 char name[PROC_NUMBUF];
19fd4bb2 3076 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
61a28784 3077 return proc_fill_cache(filp, dirent, filldir, name, len,
19fd4bb2 3078 proc_pid_instantiate, iter.task, NULL);
61a28784
EB		 3079}
3080
1da177e4
LT		 3081/* for the /proc/ directory itself, after non-process stuff has been done */
3082int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3083{
d8bdc59f
LT		 3084 unsigned int nr;
3085 struct task_struct *reaper;
19fd4bb2 3086 struct tgid_iter iter;
b488893a 3087 struct pid_namespace *ns;
1da177e4 3088
d8bdc59f
LT		 3089 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3090 goto out_no_task;
3091 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3092
3093 reaper = get_proc_task(filp->f_path.dentry->d_inode);
61a28784
EB		 3094 if (!reaper)
3095 goto out_no_task;
3096
7bcd6b0e 3097 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
c5141e6d 3098 const struct pid_entry *p = &proc_base_stuff[nr];
61a28784 3099 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
801199ce 3100 goto out;
1da177e4
LT		 3101 }
3102
b488893a 3103 ns = filp->f_dentry->d_sb->s_fs_info;
19fd4bb2
EB		 3104 iter.task = NULL;
3105 iter.tgid = filp->f_pos - TGID_OFFSET;
3106 for (iter = next_tgid(ns, iter);
3107 iter.task;
3108 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3109 filp->f_pos = iter.tgid + TGID_OFFSET;
3110 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
3111 put_task_struct(iter.task);
0804ef4b 3112 goto out;
1da177e4 3113 }
0bc58a91 3114 }
0804ef4b
EB		 3115 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3116out:
61a28784
EB		 3117 put_task_struct(reaper);
3118out_no_task:
0bc58a91
EB		 3119 return 0;
3120}
1da177e4 3121
28a6d671
EB		 3122/*
3123 * Tasks
3124 */
c5141e6d 3125static const struct pid_entry tid_base_stuff[] = {
631f9c18 3126 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3835541d 3127 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 3128 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
631f9c18
AD		 3129 REG("environ", S_IRUSR, proc_environ_operations),
3130 INF("auxv", S_IRUSR, proc_pid_auxv),
3131 ONE("status", S_IRUGO, proc_pid_status),
a9712bc1 3132 ONE("personality", S_IRUGO, proc_pid_personality),
3036e7b4 3133 INF("limits", S_IRUGO, proc_pid_limits),
43ae34cb 3134#ifdef CONFIG_SCHED_DEBUG
631f9c18 3135 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
ebcb6734 3136#endif
4614a696 3137 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 3138#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
a9712bc1 3139 INF("syscall", S_IRUGO, proc_pid_syscall),
43ae34cb 3140#endif
631f9c18
AD		 3141 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3142 ONE("stat", S_IRUGO, proc_tid_stat),
3143 ONE("statm", S_IRUGO, proc_pid_statm),
3144 REG("maps", S_IRUGO, proc_maps_operations),
28a6d671 3145#ifdef CONFIG_NUMA
631f9c18 3146 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
28a6d671 3147#endif
631f9c18
AD		 3148 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3149 LNK("cwd", proc_cwd_link),
3150 LNK("root", proc_root_link),
3151 LNK("exe", proc_exe_link),
3152 REG("mounts", S_IRUGO, proc_mounts_operations),
3153 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
1e883281 3154#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18
AD		 3155 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3156 REG("smaps", S_IRUGO, proc_smaps_operations),
ca6b0bf0 3157 REG("pagemap", S_IRUGO, proc_pagemap_operations),
28a6d671
EB		 3158#endif
3159#ifdef CONFIG_SECURITY
631f9c18 3160 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 3161#endif
3162#ifdef CONFIG_KALLSYMS
631f9c18 3163 INF("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 3164#endif
2ec220e2 3165#ifdef CONFIG_STACKTRACE
a9712bc1 3166 ONE("stack", S_IRUGO, proc_pid_stack),
28a6d671
EB		 3167#endif
3168#ifdef CONFIG_SCHEDSTATS
631f9c18 3169 INF("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 3170#endif
9745512c 3171#ifdef CONFIG_LATENCYTOP
631f9c18 3172 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 3173#endif
8793d854 3174#ifdef CONFIG_PROC_PID_CPUSET
631f9c18 3175 REG("cpuset", S_IRUGO, proc_cpuset_operations),
a424316c
PM		 3176#endif
3177#ifdef CONFIG_CGROUPS
631f9c18 3178 REG("cgroup", S_IRUGO, proc_cgroup_operations),
28a6d671 3179#endif
631f9c18
AD		 3180 INF("oom_score", S_IRUGO, proc_oom_score),
3181 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
a63d83f4 3182 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 3183#ifdef CONFIG_AUDITSYSCALL
631f9c18 3184 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
26ec3c64 3185 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 3186#endif
f4f154fd 3187#ifdef CONFIG_FAULT_INJECTION
631f9c18 3188 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 3189#endif
297c5d92 3190#ifdef CONFIG_TASK_IO_ACCOUNTING
1d1221f3 3191 INF("io", S_IRUSR, proc_tid_io_accounting),
297c5d92 3192#endif
f133ecca
CM		 3193#ifdef CONFIG_HARDWALL
3194 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3195#endif
28a6d671
EB		 3196};
3197
3198static int proc_tid_base_readdir(struct file * filp,
3199 void * dirent, filldir_t filldir)
3200{
3201 return proc_pident_readdir(filp,dirent,filldir,
3202 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3203}
3204
3205static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
7bcd6b0e
EB		 3206 return proc_pident_lookup(dir, dentry,
3207 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
28a6d671
EB		 3208}
3209
00977a59 3210static const struct file_operations proc_tid_base_operations = {
28a6d671
EB		 3211 .read = generic_read_dir,
3212 .readdir = proc_tid_base_readdir,
6038f373 3213 .llseek = default_llseek,
28a6d671
EB		 3214};
3215
c5ef1c42 3216static const struct inode_operations proc_tid_base_inode_operations = {
28a6d671
EB		 3217 .lookup = proc_tid_base_lookup,
3218 .getattr = pid_getattr,
3219 .setattr = proc_setattr,
3220};
3221
444ceed8 3222static struct dentry *proc_task_instantiate(struct inode *dir,
c5141e6d 3223 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8
EB		 3224{
3225 struct dentry *error = ERR_PTR(-ENOENT);
3226 struct inode *inode;
61a28784 3227 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 3228
3229 if (!inode)
3230 goto out;
3231 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3232 inode->i_op = &proc_tid_base_inode_operations;
3233 inode->i_fop = &proc_tid_base_operations;
3234 inode->i_flags|=S_IMMUTABLE;
aed54175 3235
bfe86848
MS		 3236 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3237 ARRAY_SIZE(tid_base_stuff)));
444ceed8 3238
fb045adb 3239 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 3240
3241 d_add(dentry, inode);
3242 /* Close the race of the process dying before we return the dentry */
3243 if (pid_revalidate(dentry, NULL))
3244 error = NULL;
3245out:
3246 return error;
3247}
3248
28a6d671
EB		 3249static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3250{
3251 struct dentry *result = ERR_PTR(-ENOENT);
3252 struct task_struct *task;
3253 struct task_struct *leader = get_proc_task(dir);
28a6d671 3254 unsigned tid;
b488893a 3255 struct pid_namespace *ns;
28a6d671
EB		 3256
3257 if (!leader)
3258 goto out_no_task;
3259
3260 tid = name_to_int(dentry);
3261 if (tid == ~0U)
3262 goto out;
3263
b488893a 3264 ns = dentry->d_sb->s_fs_info;
28a6d671 3265 rcu_read_lock();
b488893a 3266 task = find_task_by_pid_ns(tid, ns);
28a6d671
EB		 3267 if (task)
3268 get_task_struct(task);
3269 rcu_read_unlock();
3270 if (!task)
3271 goto out;
bac0abd6 3272 if (!same_thread_group(leader, task))
28a6d671
EB		 3273 goto out_drop_task;
3274
444ceed8 3275 result = proc_task_instantiate(dir, dentry, task, NULL);
28a6d671
EB		 3276out_drop_task:
3277 put_task_struct(task);
3278out:
3279 put_task_struct(leader);
3280out_no_task:
3281 return result;
3282}
3283
0bc58a91
EB		 3284/*
3285 * Find the first tid of a thread group to return to user space.
3286 *
3287 * Usually this is just the thread group leader, but if the users
3288 * buffer was too small or there was a seek into the middle of the
3289 * directory we have more work todo.
3290 *
3291 * In the case of a short read we start with find_task_by_pid.
3292 *
3293 * In the case of a seek we start with the leader and walk nr
3294 * threads past it.
3295 */
cc288738 3296static struct task_struct *first_tid(struct task_struct *leader,
b488893a 3297 int tid, int nr, struct pid_namespace *ns)
0bc58a91 3298{
a872ff0c 3299 struct task_struct *pos;
1da177e4 3300
cc288738 3301 rcu_read_lock();
0bc58a91
EB		 3302 /* Attempt to start with the pid of a thread */
3303 if (tid && (nr > 0)) {
b488893a 3304 pos = find_task_by_pid_ns(tid, ns);
a872ff0c
ON		 3305 if (pos && (pos->group_leader == leader))
3306 goto found;
0bc58a91 3307 }
1da177e4 3308
0bc58a91 3309 /* If nr exceeds the number of threads there is nothing todo */
a872ff0c
ON		 3310 pos = NULL;
3311 if (nr && nr >= get_nr_threads(leader))
3312 goto out;
1da177e4 3313
a872ff0c
ON		 3314 /* If we haven't found our starting place yet start
3315 * with the leader and walk nr threads forward.
0bc58a91 3316 */
a872ff0c
ON		 3317 for (pos = leader; nr > 0; --nr) {
3318 pos = next_thread(pos);
3319 if (pos == leader) {
3320 pos = NULL;
3321 goto out;
3322 }
1da177e4 3323 }
a872ff0c
ON		 3324found:
3325 get_task_struct(pos);
3326out:
cc288738 3327 rcu_read_unlock();
0bc58a91
EB		 3328 return pos;
3329}
3330
3331/*
3332 * Find the next thread in the thread list.
3333 * Return NULL if there is an error or no next thread.
3334 *
3335 * The reference to the input task_struct is released.
3336 */
3337static struct task_struct *next_tid(struct task_struct *start)
3338{
c1df7fb8 3339 struct task_struct *pos = NULL;
cc288738 3340 rcu_read_lock();
c1df7fb8 3341 if (pid_alive(start)) {
0bc58a91 3342 pos = next_thread(start);
c1df7fb8
ON		 3343 if (thread_group_leader(pos))
3344 pos = NULL;
3345 else
3346 get_task_struct(pos);
3347 }
cc288738 3348 rcu_read_unlock();
0bc58a91
EB		 3349 put_task_struct(start);
3350 return pos;
1da177e4
LT		 3351}
3352
61a28784
EB		 3353static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3354 struct task_struct *task, int tid)
3355{
3356 char name[PROC_NUMBUF];
3357 int len = snprintf(name, sizeof(name), "%d", tid);
3358 return proc_fill_cache(filp, dirent, filldir, name, len,
3359 proc_task_instantiate, task, NULL);
3360}
3361
1da177e4
LT		 3362/* for the /proc/TGID/task/ directories */
3363static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3364{
2fddfeef 3365 struct dentry *dentry = filp->f_path.dentry;
1da177e4 3366 struct inode *inode = dentry->d_inode;
7d895244 3367 struct task_struct *leader = NULL;
0bc58a91 3368 struct task_struct *task;
1da177e4
LT		 3369 int retval = -ENOENT;
3370 ino_t ino;
0bc58a91 3371 int tid;
b488893a 3372 struct pid_namespace *ns;
1da177e4 3373
7d895244
GC		 3374 task = get_proc_task(inode);
3375 if (!task)
3376 goto out_no_task;
3377 rcu_read_lock();
3378 if (pid_alive(task)) {
3379 leader = task->group_leader;
3380 get_task_struct(leader);
3381 }
3382 rcu_read_unlock();
3383 put_task_struct(task);
99f89551
EB		 3384 if (!leader)
3385 goto out_no_task;
1da177e4
LT		 3386 retval = 0;
3387
ee568b25 3388 switch ((unsigned long)filp->f_pos) {
1da177e4
LT		 3389 case 0:
3390 ino = inode->i_ino;
ee6f779b 3391 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
1da177e4 3392 goto out;
ee6f779b 3393 filp->f_pos++;
1da177e4
LT		 3394 /* fall through */
3395 case 1:
3396 ino = parent_ino(dentry);
ee6f779b 3397 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
1da177e4 3398 goto out;
ee6f779b 3399 filp->f_pos++;
1da177e4
LT		 3400 /* fall through */
3401 }
3402
0bc58a91
EB		 3403 /* f_version caches the tgid value that the last readdir call couldn't
3404 * return. lseek aka telldir automagically resets f_version to 0.
3405 */
b488893a 3406 ns = filp->f_dentry->d_sb->s_fs_info;
2b47c361 3407 tid = (int)filp->f_version;
0bc58a91 3408 filp->f_version = 0;
ee6f779b 3409 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
0bc58a91 3410 task;
ee6f779b 3411 task = next_tid(task), filp->f_pos++) {
b488893a 3412 tid = task_pid_nr_ns(task, ns);
61a28784 3413 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
0bc58a91
EB		 3414 /* returning this tgid failed, save it as the first
3415 * pid for the next readir call */
2b47c361 3416 filp->f_version = (u64)tid;
0bc58a91 3417 put_task_struct(task);
1da177e4 3418 break;
0bc58a91 3419 }
1da177e4
LT		 3420 }
3421out:
99f89551
EB		 3422 put_task_struct(leader);
3423out_no_task:
1da177e4
LT		 3424 return retval;
3425}
6e66b52b
EB		 3426
3427static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3428{
3429 struct inode *inode = dentry->d_inode;
99f89551 3430 struct task_struct *p = get_proc_task(inode);
6e66b52b
EB		 3431 generic_fillattr(inode, stat);
3432
99f89551 3433 if (p) {
99f89551 3434 stat->nlink += get_nr_threads(p);
99f89551 3435 put_task_struct(p);
6e66b52b
EB		 3436 }
3437
3438 return 0;
3439}
28a6d671 3440
c5ef1c42 3441static const struct inode_operations proc_task_inode_operations = {
28a6d671
EB		 3442 .lookup = proc_task_lookup,
3443 .getattr = proc_task_getattr,
3444 .setattr = proc_setattr,
3445};
3446
00977a59 3447static const struct file_operations proc_task_operations = {
28a6d671
EB		 3448 .read = generic_read_dir,
3449 .readdir = proc_task_readdir,
6038f373 3450 .llseek = default_llseek,
28a6d671 3451};
Linux 6.x block layer and io_uring tree(s)