projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
proc: do proper range check on readdir offset
[deliverable/linux.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>
1da177e4
LT		 86#include "internal.h"
87
0f2fe20f
EB		 88/* NOTE:
89 * Implementing inode permission operations in /proc is almost
90 * certainly an error. Permission checks need to happen during
91 * each system call not at open time. The reason is that most of
92 * what we wish to check for permissions in /proc varies at runtime.
93 *
94 * The classic example of a problem is opening file descriptors
95 * in /proc for a task before it execs a suid executable.
96 */
97
1da177e4 98struct pid_entry {
1da177e4 99 char *name;
c5141e6d 100 int len;
1da177e4 101 mode_t mode;
c5ef1c42 102 const struct inode_operations *iop;
00977a59 103 const struct file_operations *fop;
20cdc894 104 union proc_op op;
1da177e4
LT		 105};
106
61a28784 107#define NOD(NAME, MODE, IOP, FOP, OP) { \
20cdc894 108 .name = (NAME), \
c5141e6d 109 .len = sizeof(NAME) - 1, \
20cdc894
EB		 110 .mode = MODE, \
111 .iop = IOP, \
112 .fop = FOP, \
113 .op = OP, \
114}
115
631f9c18
AD		 116#define DIR(NAME, MODE, iops, fops) \
117 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
118#define LNK(NAME, get_link) \
61a28784 119 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
20cdc894 120 &proc_pid_link_inode_operations, NULL, \
631f9c18
AD		 121 { .proc_get_link = get_link } )
122#define REG(NAME, MODE, fops) \
123 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
124#define INF(NAME, MODE, read) \
61a28784 125 NOD(NAME, (S_IFREG|(MODE)), \
20cdc894 126 NULL, &proc_info_file_operations, \
631f9c18
AD		 127 { .proc_read = read } )
128#define ONE(NAME, MODE, show) \
be614086
EB		 129 NOD(NAME, (S_IFREG|(MODE)), \
130 NULL, &proc_single_file_operations, \
631f9c18 131 { .proc_show = show } )
1da177e4 132
aed54175
VN		 133/*
134 * Count the number of hardlinks for the pid_entry table, excluding the .
135 * and .. links.
136 */
137static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
138 unsigned int n)
139{
140 unsigned int i;
141 unsigned int count;
142
143 count = 0;
144 for (i = 0; i < n; ++i) {
145 if (S_ISDIR(entries[i].mode))
146 ++count;
147 }
148
149 return count;
150}
151
f7ad3c6b 152static int get_task_root(struct task_struct *task, struct path *root)
1da177e4 153{
7c2c7d99
HD		 154 int result = -ENOENT;
155
0494f6ec 156 task_lock(task);
f7ad3c6b
MS		 157 if (task->fs) {
158 get_fs_root(task->fs, root);
7c2c7d99
HD		 159 result = 0;
160 }
0494f6ec 161 task_unlock(task);
7c2c7d99 162 return result;
0494f6ec
MS		 163}
164
3dcd25f3 165static int proc_cwd_link(struct inode *inode, struct path *path)
0494f6ec 166{
99f89551 167 struct task_struct *task = get_proc_task(inode);
0494f6ec 168 int result = -ENOENT;
99f89551
EB		 169
170 if (task) {
f7ad3c6b
MS		 171 task_lock(task);
172 if (task->fs) {
173 get_fs_pwd(task->fs, path);
174 result = 0;
175 }
176 task_unlock(task);
99f89551
EB		 177 put_task_struct(task);
178 }
1da177e4
LT		 179 return result;
180}
181
3dcd25f3 182static int proc_root_link(struct inode *inode, struct path *path)
1da177e4 183{
99f89551 184 struct task_struct *task = get_proc_task(inode);
1da177e4 185 int result = -ENOENT;
99f89551
EB		 186
187 if (task) {
f7ad3c6b 188 result = get_task_root(task, path);
99f89551
EB		 189 put_task_struct(task);
190 }
1da177e4
LT		 191 return result;
192}
193
8b0db9db 194static struct mm_struct *__check_mem_permission(struct task_struct *task)
638fa202 195{
8b0db9db
SW		 196 struct mm_struct *mm;
197
198 mm = get_task_mm(task);
199 if (!mm)
200 return ERR_PTR(-EINVAL);
201
638fa202
RM		 202 /*
203 * A task can always look at itself, in case it chooses
204 * to use system calls instead of load instructions.
205 */
206 if (task == current)
8b0db9db 207 return mm;
638fa202
RM		 208
209 /*
210 * If current is actively ptrace'ing, and would also be
211 * permitted to freshly attach with ptrace now, permit it.
212 */
0d094efe
RM		 213 if (task_is_stopped_or_traced(task)) {
214 int match;
215 rcu_read_lock();
216 match = (tracehook_tracer_task(task) == current);
217 rcu_read_unlock();
218 if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH))
8b0db9db 219 return mm;
0d094efe 220 }
638fa202
RM		 221
222 /*
25985edc 223 * No one else is allowed.
638fa202 224 */
8b0db9db
SW		 225 mmput(mm);
226 return ERR_PTR(-EPERM);
638fa202 227}
1da177e4 228
18f661bc 229/*
8b0db9db
SW		 230 * If current may access user memory in @task return a reference to the
231 * corresponding mm, otherwise ERR_PTR.
18f661bc 232 */
8b0db9db 233static struct mm_struct *check_mem_permission(struct task_struct *task)
18f661bc 234{
8b0db9db 235 struct mm_struct *mm;
18f661bc
SW		 236 int err;
237
238 /*
239 * Avoid racing if task exec's as we might get a new mm but validate
240 * against old credentials.
241 */
242 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
243 if (err)
8b0db9db 244 return ERR_PTR(err);
18f661bc 245
8b0db9db 246 mm = __check_mem_permission(task);
18f661bc
SW		 247 mutex_unlock(&task->signal->cred_guard_mutex);
248
8b0db9db 249 return mm;
638fa202 250}
1da177e4 251
831830b5
AV		 252struct mm_struct *mm_for_maps(struct task_struct *task)
253{
704b836c 254 struct mm_struct *mm;
ec6fd8a4 255 int err;
704b836c 256
ec6fd8a4
AV		 257 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
258 if (err)
259 return ERR_PTR(err);
00f89d21 260
704b836c
ON		 261 mm = get_task_mm(task);
262 if (mm && mm != current->mm &&
263 !ptrace_may_access(task, PTRACE_MODE_READ)) {
264 mmput(mm);
ec6fd8a4 265 mm = ERR_PTR(-EACCES);
13f0feaf 266 }
9b1bf12d 267 mutex_unlock(&task->signal->cred_guard_mutex);
704b836c 268
831830b5 269 return mm;
831830b5
AV		 270}
271
1da177e4
LT		 272static int proc_pid_cmdline(struct task_struct *task, char * buffer)
273{
274 int res = 0;
275 unsigned int len;
276 struct mm_struct *mm = get_task_mm(task);
277 if (!mm)
278 goto out;
279 if (!mm->arg_end)
280 goto out_mm; /* Shh! No looking before we're done */
281
282 len = mm->arg_end - mm->arg_start;
283
284 if (len > PAGE_SIZE)
285 len = PAGE_SIZE;
286
287 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
288
289 // If the nul at the end of args has been overwritten, then
290 // assume application is using setproctitle(3).
291 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
292 len = strnlen(buffer, res);
293 if (len < res) {
294 res = len;
295 } else {
296 len = mm->env_end - mm->env_start;
297 if (len > PAGE_SIZE - res)
298 len = PAGE_SIZE - res;
299 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
300 res = strnlen(buffer, res);
301 }
302 }
303out_mm:
304 mmput(mm);
305out:
306 return res;
307}
308
309static int proc_pid_auxv(struct task_struct *task, char *buffer)
310{
2fadaef4
AV		 311 struct mm_struct *mm = mm_for_maps(task);
312 int res = PTR_ERR(mm);
313 if (mm && !IS_ERR(mm)) {
1da177e4 314 unsigned int nwords = 0;
dfe6b7d9 315 do {
1da177e4 316 nwords += 2;
dfe6b7d9 317 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
1da177e4
LT		 318 res = nwords * sizeof(mm->saved_auxv[0]);
319 if (res > PAGE_SIZE)
320 res = PAGE_SIZE;
321 memcpy(buffer, mm->saved_auxv, res);
322 mmput(mm);
323 }
324 return res;
325}
326
327
328#ifdef CONFIG_KALLSYMS
329/*
330 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
331 * Returns the resolved symbol. If that fails, simply return the address.
332 */
333static int proc_pid_wchan(struct task_struct *task, char *buffer)
334{
ffb45122 335 unsigned long wchan;
9281acea 336 char symname[KSYM_NAME_LEN];
1da177e4
LT		 337
338 wchan = get_wchan(task);
339
9d65cb4a 340 if (lookup_symbol_name(wchan, symname) < 0)
f83ce3e6
JE		 341 if (!ptrace_may_access(task, PTRACE_MODE_READ))
342 return 0;
343 else
344 return sprintf(buffer, "%lu", wchan);
9d65cb4a
AD		 345 else
346 return sprintf(buffer, "%s", symname);
1da177e4
LT		 347}
348#endif /* CONFIG_KALLSYMS */
349
a9712bc1
AV		 350static int lock_trace(struct task_struct *task)
351{
352 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
353 if (err)
354 return err;
355 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
356 mutex_unlock(&task->signal->cred_guard_mutex);
357 return -EPERM;
358 }
359 return 0;
360}
361
362static void unlock_trace(struct task_struct *task)
363{
364 mutex_unlock(&task->signal->cred_guard_mutex);
365}
366
2ec220e2
KC		 367#ifdef CONFIG_STACKTRACE
368
369#define MAX_STACK_TRACE_DEPTH 64
370
371static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
372 struct pid *pid, struct task_struct *task)
373{
374 struct stack_trace trace;
375 unsigned long *entries;
a9712bc1 376 int err;
2ec220e2
KC		 377 int i;
378
379 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
380 if (!entries)
381 return -ENOMEM;
382
383 trace.nr_entries = 0;
384 trace.max_entries = MAX_STACK_TRACE_DEPTH;
385 trace.entries = entries;
386 trace.skip = 0;
2ec220e2 387
a9712bc1
AV		 388 err = lock_trace(task);
389 if (!err) {
390 save_stack_trace_tsk(task, &trace);
391
392 for (i = 0; i < trace.nr_entries; i++) {
b81a618d 393 seq_printf(m, "[<%pK>] %pS\n",
a9712bc1
AV		 394 (void *)entries[i], (void *)entries[i]);
395 }
396 unlock_trace(task);
2ec220e2
KC		 397 }
398 kfree(entries);
399
a9712bc1 400 return err;
2ec220e2
KC		 401}
402#endif
403
1da177e4
LT		 404#ifdef CONFIG_SCHEDSTATS
405/*
406 * Provides /proc/PID/schedstat
407 */
408static int proc_pid_schedstat(struct task_struct *task, char *buffer)
409{
172ba844 410 return sprintf(buffer, "%llu %llu %lu\n",
826e08b0
IM		 411 (unsigned long long)task->se.sum_exec_runtime,
412 (unsigned long long)task->sched_info.run_delay,
2d72376b 413 task->sched_info.pcount);
1da177e4
LT		 414}
415#endif
416
9745512c
AV		 417#ifdef CONFIG_LATENCYTOP
418static int lstats_show_proc(struct seq_file *m, void *v)
419{
420 int i;
13d77c37
HS		 421 struct inode *inode = m->private;
422 struct task_struct *task = get_proc_task(inode);
9745512c 423
13d77c37
HS		 424 if (!task)
425 return -ESRCH;
426 seq_puts(m, "Latency Top version : v0.1\n");
9745512c 427 for (i = 0; i < 32; i++) {
34e49d4f
JP		 428 struct latency_record *lr = &task->latency_record[i];
429 if (lr->backtrace[0]) {
9745512c 430 int q;
34e49d4f
JP		 431 seq_printf(m, "%i %li %li",
432 lr->count, lr->time, lr->max);
9745512c 433 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
34e49d4f
JP		 434 unsigned long bt = lr->backtrace[q];
435 if (!bt)
9745512c 436 break;
34e49d4f 437 if (bt == ULONG_MAX)
9745512c 438 break;
34e49d4f 439 seq_printf(m, " %ps", (void *)bt);
9745512c 440 }
9d6de12f 441 seq_putc(m, '\n');
9745512c
AV		 442 }
443
444 }
13d77c37 445 put_task_struct(task);
9745512c
AV		 446 return 0;
447}
448
449static int lstats_open(struct inode *inode, struct file *file)
450{
13d77c37 451 return single_open(file, lstats_show_proc, inode);
d6643d12
HS		 452}
453
9745512c
AV		 454static ssize_t lstats_write(struct file *file, const char __user *buf,
455 size_t count, loff_t *offs)
456{
13d77c37 457 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
9745512c 458
13d77c37
HS		 459 if (!task)
460 return -ESRCH;
9745512c 461 clear_all_latency_tracing(task);
13d77c37 462 put_task_struct(task);
9745512c
AV		 463
464 return count;
465}
466
467static const struct file_operations proc_lstats_operations = {
468 .open = lstats_open,
469 .read = seq_read,
470 .write = lstats_write,
471 .llseek = seq_lseek,
13d77c37 472 .release = single_release,
9745512c
AV		 473};
474
475#endif
476
1da177e4
LT		 477static int proc_oom_score(struct task_struct *task, char *buffer)
478{
b95c35e7 479 unsigned long points = 0;
1da177e4 480
19c5d45a 481 read_lock(&tasklist_lock);
b95c35e7 482 if (pid_alive(task))
a63d83f4
DR		 483 points = oom_badness(task, NULL, NULL,
484 totalram_pages + total_swap_pages);
19c5d45a 485 read_unlock(&tasklist_lock);
1da177e4
LT		 486 return sprintf(buffer, "%lu\n", points);
487}
488
d85f50d5
NH		 489struct limit_names {
490 char *name;
491 char *unit;
492};
493
494static const struct limit_names lnames[RLIM_NLIMITS] = {
cff4edb5 495 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
d85f50d5
NH		 496 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
497 [RLIMIT_DATA] = {"Max data size", "bytes"},
498 [RLIMIT_STACK] = {"Max stack size", "bytes"},
499 [RLIMIT_CORE] = {"Max core file size", "bytes"},
500 [RLIMIT_RSS] = {"Max resident set", "bytes"},
501 [RLIMIT_NPROC] = {"Max processes", "processes"},
502 [RLIMIT_NOFILE] = {"Max open files", "files"},
503 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
504 [RLIMIT_AS] = {"Max address space", "bytes"},
505 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
506 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
507 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
508 [RLIMIT_NICE] = {"Max nice priority", NULL},
509 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
8808117c 510 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
d85f50d5
NH		 511};
512
513/* Display limits for a process */
514static int proc_pid_limits(struct task_struct *task, char *buffer)
515{
516 unsigned int i;
517 int count = 0;
518 unsigned long flags;
519 char *bufptr = buffer;
520
521 struct rlimit rlim[RLIM_NLIMITS];
522
a6bebbc8 523 if (!lock_task_sighand(task, &flags))
d85f50d5 524 return 0;
d85f50d5
NH		 525 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
526 unlock_task_sighand(task, &flags);
d85f50d5
NH		 527
528 /*
529 * print the file header
530 */
531 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
532 "Limit", "Soft Limit", "Hard Limit", "Units");
533
534 for (i = 0; i < RLIM_NLIMITS; i++) {
535 if (rlim[i].rlim_cur == RLIM_INFINITY)
536 count += sprintf(&bufptr[count], "%-25s %-20s ",
537 lnames[i].name, "unlimited");
538 else
539 count += sprintf(&bufptr[count], "%-25s %-20lu ",
540 lnames[i].name, rlim[i].rlim_cur);
541
542 if (rlim[i].rlim_max == RLIM_INFINITY)
543 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
544 else
545 count += sprintf(&bufptr[count], "%-20lu ",
546 rlim[i].rlim_max);
547
548 if (lnames[i].unit)
549 count += sprintf(&bufptr[count], "%-10s\n",
550 lnames[i].unit);
551 else
552 count += sprintf(&bufptr[count], "\n");
553 }
554
555 return count;
556}
557
ebcb6734
RM		 558#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
559static int proc_pid_syscall(struct task_struct *task, char *buffer)
560{
561 long nr;
562 unsigned long args[6], sp, pc;
a9712bc1
AV		 563 int res = lock_trace(task);
564 if (res)
565 return res;
ebcb6734
RM		 566
567 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
a9712bc1
AV		 568 res = sprintf(buffer, "running\n");
569 else if (nr < 0)
570 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
571 else
572 res = sprintf(buffer,
ebcb6734
RM		 573 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
574 nr,
575 args[0], args[1], args[2], args[3], args[4], args[5],
576 sp, pc);
a9712bc1
AV		 577 unlock_trace(task);
578 return res;
ebcb6734
RM		 579}
580#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
581
1da177e4
LT		 582/************************************************************************/
583/* Here the fs part begins */
584/************************************************************************/
585
586/* permission checks */
778c1144 587static int proc_fd_access_allowed(struct inode *inode)
1da177e4 588{
778c1144
EB		 589 struct task_struct *task;
590 int allowed = 0;
df26c40e
EB		 591 /* Allow access to a task's file descriptors if it is us or we
592 * may use ptrace attach to the process and find out that
593 * information.
778c1144
EB		 594 */
595 task = get_proc_task(inode);
df26c40e 596 if (task) {
006ebb40 597 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
778c1144 598 put_task_struct(task);
df26c40e 599 }
778c1144 600 return allowed;
1da177e4
LT		 601}
602
6d76fa58
LT		 603static int proc_setattr(struct dentry *dentry, struct iattr *attr)
604{
605 int error;
606 struct inode *inode = dentry->d_inode;
607
608 if (attr->ia_valid & ATTR_MODE)
609 return -EPERM;
610
611 error = inode_change_ok(inode, attr);
1025774c
CH		 612 if (error)
613 return error;
614
615 if ((attr->ia_valid & ATTR_SIZE) &&
616 attr->ia_size != i_size_read(inode)) {
617 error = vmtruncate(inode, attr->ia_size);
618 if (error)
619 return error;
620 }
621
622 setattr_copy(inode, attr);
623 mark_inode_dirty(inode);
624 return 0;
6d76fa58
LT		 625}
626
c5ef1c42 627static const struct inode_operations proc_def_inode_operations = {
6d76fa58
LT		 628 .setattr = proc_setattr,
629};
630
a1a2c409
MS		 631static int mounts_open_common(struct inode *inode, struct file *file,
632 const struct seq_operations *op)
1da177e4 633{
99f89551 634 struct task_struct *task = get_proc_task(inode);
cf7b708c 635 struct nsproxy *nsp;
6b3286ed 636 struct mnt_namespace *ns = NULL;
a1a2c409 637 struct path root;
5addc5dd
AV		 638 struct proc_mounts *p;
639 int ret = -EINVAL;
1da177e4 640
99f89551 641 if (task) {
cf7b708c
PE		 642 rcu_read_lock();
643 nsp = task_nsproxy(task);
644 if (nsp) {
645 ns = nsp->mnt_ns;
863c4702
AD		 646 if (ns)
647 get_mnt_ns(ns);
648 }
cf7b708c 649 rcu_read_unlock();
f7ad3c6b 650 if (ns && get_task_root(task, &root) == 0)
7c2c7d99 651 ret = 0;
99f89551
EB		 652 put_task_struct(task);
653 }
5addc5dd 654
a1a2c409
MS		 655 if (!ns)
656 goto err;
7c2c7d99 657 if (ret)
a1a2c409
MS		 658 goto err_put_ns;
659
a1a2c409
MS		 660 ret = -ENOMEM;
661 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
662 if (!p)
663 goto err_put_path;
664
665 file->private_data = &p->m;
666 ret = seq_open(file, op);
667 if (ret)
668 goto err_free;
669
670 p->m.private = p;
671 p->ns = ns;
672 p->root = root;
673 p->event = ns->event;
674
675 return 0;
676
677 err_free:
678 kfree(p);
679 err_put_path:
680 path_put(&root);
681 err_put_ns:
682 put_mnt_ns(ns);
683 err:
1da177e4
LT		 684 return ret;
685}
686
687static int mounts_release(struct inode *inode, struct file *file)
688{
a1a2c409
MS		 689 struct proc_mounts *p = file->private_data;
690 path_put(&p->root);
691 put_mnt_ns(p->ns);
1da177e4
LT		 692 return seq_release(inode, file);
693}
694
5addc5dd
AV		 695static unsigned mounts_poll(struct file *file, poll_table *wait)
696{
697 struct proc_mounts *p = file->private_data;
31b07093 698 unsigned res = POLLIN | POLLRDNORM;
5addc5dd 699
9f5596af
AV		 700 poll_wait(file, &p->ns->poll, wait);
701 if (mnt_had_events(p))
31b07093 702 res |= POLLERR | POLLPRI;
5addc5dd
AV		 703
704 return res;
705}
706
a1a2c409
MS		 707static int mounts_open(struct inode *inode, struct file *file)
708{
709 return mounts_open_common(inode, file, &mounts_op);
710}
711
00977a59 712static const struct file_operations proc_mounts_operations = {
1da177e4
LT		 713 .open = mounts_open,
714 .read = seq_read,
715 .llseek = seq_lseek,
716 .release = mounts_release,
5addc5dd 717 .poll = mounts_poll,
1da177e4
LT		 718};
719
2d4d4864
RP		 720static int mountinfo_open(struct inode *inode, struct file *file)
721{
722 return mounts_open_common(inode, file, &mountinfo_op);
723}
724
725static const struct file_operations proc_mountinfo_operations = {
726 .open = mountinfo_open,
727 .read = seq_read,
728 .llseek = seq_lseek,
729 .release = mounts_release,
730 .poll = mounts_poll,
731};
732
b4629fe2
CL		 733static int mountstats_open(struct inode *inode, struct file *file)
734{
a1a2c409 735 return mounts_open_common(inode, file, &mountstats_op);
b4629fe2
CL		 736}
737
00977a59 738static const struct file_operations proc_mountstats_operations = {
b4629fe2
CL		 739 .open = mountstats_open,
740 .read = seq_read,
741 .llseek = seq_lseek,
742 .release = mounts_release,
743};
744
1da177e4
LT		 745#define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
746
747static ssize_t proc_info_read(struct file * file, char __user * buf,
748 size_t count, loff_t *ppos)
749{
2fddfeef 750 struct inode * inode = file->f_path.dentry->d_inode;
1da177e4
LT		 751 unsigned long page;
752 ssize_t length;
99f89551
EB		 753 struct task_struct *task = get_proc_task(inode);
754
755 length = -ESRCH;
756 if (!task)
757 goto out_no_task;
1da177e4
LT		 758
759 if (count > PROC_BLOCK_SIZE)
760 count = PROC_BLOCK_SIZE;
99f89551
EB		 761
762 length = -ENOMEM;
e12ba74d 763 if (!(page = __get_free_page(GFP_TEMPORARY)))
99f89551 764 goto out;
1da177e4
LT		 765
766 length = PROC_I(inode)->op.proc_read(task, (char*)page);
767
768 if (length >= 0)
769 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
770 free_page(page);
99f89551
EB		 771out:
772 put_task_struct(task);
773out_no_task:
1da177e4
LT		 774 return length;
775}
776
00977a59 777static const struct file_operations proc_info_file_operations = {
1da177e4 778 .read = proc_info_read,
87df8424 779 .llseek = generic_file_llseek,
1da177e4
LT		 780};
781
be614086
EB		 782static int proc_single_show(struct seq_file *m, void *v)
783{
784 struct inode *inode = m->private;
785 struct pid_namespace *ns;
786 struct pid *pid;
787 struct task_struct *task;
788 int ret;
789
790 ns = inode->i_sb->s_fs_info;
791 pid = proc_pid(inode);
792 task = get_pid_task(pid, PIDTYPE_PID);
793 if (!task)
794 return -ESRCH;
795
796 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
797
798 put_task_struct(task);
799 return ret;
800}
801
802static int proc_single_open(struct inode *inode, struct file *filp)
803{
c6a34058 804 return single_open(filp, proc_single_show, inode);
be614086
EB		 805}
806
807static const struct file_operations proc_single_file_operations = {
808 .open = proc_single_open,
809 .read = seq_read,
810 .llseek = seq_lseek,
811 .release = single_release,
812};
813
1da177e4
LT		 814static int mem_open(struct inode* inode, struct file* file)
815{
816 file->private_data = (void*)((long)current->self_exec_id);
4a3956c7
KH		 817 /* OK to pass negative loff_t, we can catch out-of-range */
818 file->f_mode |= FMODE_UNSIGNED_OFFSET;
1da177e4
LT		 819 return 0;
820}
821
822static ssize_t mem_read(struct file * file, char __user * buf,
823 size_t count, loff_t *ppos)
824{
2fddfeef 825 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1da177e4
LT		 826 char *page;
827 unsigned long src = *ppos;
828 int ret = -ESRCH;
829 struct mm_struct *mm;
830
99f89551
EB		 831 if (!task)
832 goto out_no_task;
833
1da177e4 834 ret = -ENOMEM;
e12ba74d 835 page = (char *)__get_free_page(GFP_TEMPORARY);
1da177e4
LT		 836 if (!page)
837 goto out;
838
8b0db9db
SW		 839 mm = check_mem_permission(task);
840 ret = PTR_ERR(mm);
841 if (IS_ERR(mm))
1da177e4
LT		 842 goto out_free;
843
844 ret = -EIO;
845
846 if (file->private_data != (void*)((long)current->self_exec_id))
847 goto out_put;
848
849 ret = 0;
850
851 while (count > 0) {
852 int this_len, retval;
853
854 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
8b0db9db
SW		 855 retval = access_remote_vm(mm, src, page, this_len, 0);
856 if (!retval) {
1da177e4
LT		 857 if (!ret)
858 ret = -EIO;
859 break;
860 }
861
862 if (copy_to_user(buf, page, retval)) {
863 ret = -EFAULT;
864 break;
865 }
866
867 ret += retval;
868 src += retval;
869 buf += retval;
870 count -= retval;
871 }
872 *ppos = src;
873
874out_put:
875 mmput(mm);
876out_free:
877 free_page((unsigned long) page);
878out:
99f89551
EB		 879 put_task_struct(task);
880out_no_task:
1da177e4
LT		 881 return ret;
882}
883
63967fa9 884static ssize_t mem_write(struct file * file, const char __user *buf,
1da177e4
LT		 885 size_t count, loff_t *ppos)
886{
f7ca54f4 887 int copied;
1da177e4 888 char *page;
2fddfeef 889 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1da177e4 890 unsigned long dst = *ppos;
8b0db9db 891 struct mm_struct *mm;
1da177e4 892
99f89551
EB		 893 copied = -ESRCH;
894 if (!task)
895 goto out_no_task;
896
8b0db9db
SW		 897 mm = check_mem_permission(task);
898 copied = PTR_ERR(mm);
899 if (IS_ERR(mm))
900 goto out_task;
1da177e4 901
26947f8c
SW		 902 copied = -EIO;
903 if (file->private_data != (void *)((long)current->self_exec_id))
8b0db9db 904 goto out_mm;
1da177e4 905
99f89551 906 copied = -ENOMEM;
e12ba74d 907 page = (char *)__get_free_page(GFP_TEMPORARY);
1da177e4 908 if (!page)
8b0db9db 909 goto out_mm;
1da177e4 910
f7ca54f4 911 copied = 0;
1da177e4
LT		 912 while (count > 0) {
913 int this_len, retval;
914
915 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
916 if (copy_from_user(page, buf, this_len)) {
917 copied = -EFAULT;
918 break;
919 }
8b0db9db 920 retval = access_remote_vm(mm, dst, page, this_len, 1);
1da177e4
LT		 921 if (!retval) {
922 if (!copied)
923 copied = -EIO;
924 break;
925 }
926 copied += retval;
927 buf += retval;
928 dst += retval;
929 count -= retval;
930 }
931 *ppos = dst;
932 free_page((unsigned long) page);
8b0db9db
SW		 933out_mm:
934 mmput(mm);
935out_task:
99f89551
EB		 936 put_task_struct(task);
937out_no_task:
1da177e4
LT		 938 return copied;
939}
1da177e4 940
85863e47 941loff_t mem_lseek(struct file *file, loff_t offset, int orig)
1da177e4
LT		 942{
943 switch (orig) {
944 case 0:
945 file->f_pos = offset;
946 break;
947 case 1:
948 file->f_pos += offset;
949 break;
950 default:
951 return -EINVAL;
952 }
953 force_successful_syscall_return();
954 return file->f_pos;
955}
956
00977a59 957static const struct file_operations proc_mem_operations = {
1da177e4
LT		 958 .llseek = mem_lseek,
959 .read = mem_read,
960 .write = mem_write,
961 .open = mem_open,
962};
963
315e28c8
JP		 964static ssize_t environ_read(struct file *file, char __user *buf,
965 size_t count, loff_t *ppos)
966{
967 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
968 char *page;
969 unsigned long src = *ppos;
970 int ret = -ESRCH;
971 struct mm_struct *mm;
972
973 if (!task)
974 goto out_no_task;
975
315e28c8
JP		 976 ret = -ENOMEM;
977 page = (char *)__get_free_page(GFP_TEMPORARY);
978 if (!page)
979 goto out;
980
315e28c8 981
d6f64b89
AV		 982 mm = mm_for_maps(task);
983 ret = PTR_ERR(mm);
984 if (!mm || IS_ERR(mm))
315e28c8
JP		 985 goto out_free;
986
d6f64b89 987 ret = 0;
315e28c8
JP		 988 while (count > 0) {
989 int this_len, retval, max_len;
990
991 this_len = mm->env_end - (mm->env_start + src);
992
993 if (this_len <= 0)
994 break;
995
996 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
997 this_len = (this_len > max_len) ? max_len : this_len;
998
999 retval = access_process_vm(task, (mm->env_start + src),
1000 page, this_len, 0);
1001
1002 if (retval <= 0) {
1003 ret = retval;
1004 break;
1005 }
1006
1007 if (copy_to_user(buf, page, retval)) {
1008 ret = -EFAULT;
1009 break;
1010 }
1011
1012 ret += retval;
1013 src += retval;
1014 buf += retval;
1015 count -= retval;
1016 }
1017 *ppos = src;
1018
1019 mmput(mm);
1020out_free:
1021 free_page((unsigned long) page);
1022out:
1023 put_task_struct(task);
1024out_no_task:
1025 return ret;
1026}
1027
1028static const struct file_operations proc_environ_operations = {
1029 .read = environ_read,
87df8424 1030 .llseek = generic_file_llseek,
315e28c8
JP		 1031};
1032
1da177e4
LT		 1033static ssize_t oom_adjust_read(struct file *file, char __user *buf,
1034 size_t count, loff_t *ppos)
1035{
2fddfeef 1036 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
8578cea7 1037 char buffer[PROC_NUMBUF];
1da177e4 1038 size_t len;
28b83c51
KM		 1039 int oom_adjust = OOM_DISABLE;
1040 unsigned long flags;
1da177e4 1041
99f89551
EB		 1042 if (!task)
1043 return -ESRCH;
28b83c51
KM		 1044
1045 if (lock_task_sighand(task, &flags)) {
1046 oom_adjust = task->signal->oom_adj;
1047 unlock_task_sighand(task, &flags);
1048 }
1049
99f89551
EB		 1050 put_task_struct(task);
1051
8578cea7 1052 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
0c28f287
AM		 1053
1054 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1da177e4
LT		 1055}
1056
1057static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1058 size_t count, loff_t *ppos)
1059{
99f89551 1060 struct task_struct *task;
5d863b89
KM		 1061 char buffer[PROC_NUMBUF];
1062 long oom_adjust;
28b83c51 1063 unsigned long flags;
5d863b89 1064 int err;
1da177e4 1065
8578cea7
EB		 1066 memset(buffer, 0, sizeof(buffer));
1067 if (count > sizeof(buffer) - 1)
1068 count = sizeof(buffer) - 1;
723548bf
DR		 1069 if (copy_from_user(buffer, buf, count)) {
1070 err = -EFAULT;
1071 goto out;
1072 }
5d863b89
KM		 1073
1074 err = strict_strtol(strstrip(buffer), 0, &oom_adjust);
1075 if (err)
723548bf 1076 goto out;
8ac773b4 1077 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
723548bf
DR		 1078 oom_adjust != OOM_DISABLE) {
1079 err = -EINVAL;
1080 goto out;
1081 }
5d863b89 1082
2fddfeef 1083 task = get_proc_task(file->f_path.dentry->d_inode);
723548bf
DR		 1084 if (!task) {
1085 err = -ESRCH;
1086 goto out;
1087 }
d19d5476
DR		 1088
1089 task_lock(task);
1090 if (!task->mm) {
1091 err = -EINVAL;
1092 goto err_task_lock;
1093 }
1094
28b83c51 1095 if (!lock_task_sighand(task, &flags)) {
723548bf 1096 err = -ESRCH;
d19d5476 1097 goto err_task_lock;
28b83c51
KM		 1098 }
1099
1100 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
723548bf
DR		 1101 err = -EACCES;
1102 goto err_sighand;
8fb4fc68 1103 }
28b83c51 1104
3d5992d2
YH		 1105 if (oom_adjust != task->signal->oom_adj) {
1106 if (oom_adjust == OOM_DISABLE)
1107 atomic_inc(&task->mm->oom_disable_count);
1108 if (task->signal->oom_adj == OOM_DISABLE)
1109 atomic_dec(&task->mm->oom_disable_count);
1110 }
1111
51b1bd2a
DR		 1112 /*
1113 * Warn that /proc/pid/oom_adj is deprecated, see
1114 * Documentation/feature-removal-schedule.txt.
1115 */
1116 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, "
1117 "please use /proc/%d/oom_score_adj instead.\n",
1118 current->comm, task_pid_nr(current),
1119 task_pid_nr(task), task_pid_nr(task));
28b83c51 1120 task->signal->oom_adj = oom_adjust;
a63d83f4
DR		 1121 /*
1122 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1123 * value is always attainable.
1124 */
1125 if (task->signal->oom_adj == OOM_ADJUST_MAX)
1126 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
1127 else
1128 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
1129 -OOM_DISABLE;
723548bf 1130err_sighand:
28b83c51 1131 unlock_task_sighand(task, &flags);
d19d5476
DR		 1132err_task_lock:
1133 task_unlock(task);
99f89551 1134 put_task_struct(task);
723548bf
DR		 1135out:
1136 return err < 0 ? err : count;
1da177e4
LT		 1137}
1138
00977a59 1139static const struct file_operations proc_oom_adjust_operations = {
1da177e4
LT		 1140 .read = oom_adjust_read,
1141 .write = oom_adjust_write,
87df8424 1142 .llseek = generic_file_llseek,
1da177e4
LT		 1143};
1144
a63d83f4
DR		 1145static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1146 size_t count, loff_t *ppos)
1147{
1148 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1149 char buffer[PROC_NUMBUF];
1150 int oom_score_adj = OOM_SCORE_ADJ_MIN;
1151 unsigned long flags;
1152 size_t len;
1153
1154 if (!task)
1155 return -ESRCH;
1156 if (lock_task_sighand(task, &flags)) {
1157 oom_score_adj = task->signal->oom_score_adj;
1158 unlock_task_sighand(task, &flags);
1159 }
1160 put_task_struct(task);
1161 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
1162 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1163}
1164
1165static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1166 size_t count, loff_t *ppos)
1167{
1168 struct task_struct *task;
1169 char buffer[PROC_NUMBUF];
1170 unsigned long flags;
1171 long oom_score_adj;
1172 int err;
1173
1174 memset(buffer, 0, sizeof(buffer));
1175 if (count > sizeof(buffer) - 1)
1176 count = sizeof(buffer) - 1;
723548bf
DR		 1177 if (copy_from_user(buffer, buf, count)) {
1178 err = -EFAULT;
1179 goto out;
1180 }
a63d83f4
DR		 1181
1182 err = strict_strtol(strstrip(buffer), 0, &oom_score_adj);
1183 if (err)
723548bf 1184 goto out;
a63d83f4 1185 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
723548bf
DR		 1186 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1187 err = -EINVAL;
1188 goto out;
1189 }
a63d83f4
DR		 1190
1191 task = get_proc_task(file->f_path.dentry->d_inode);
723548bf
DR		 1192 if (!task) {
1193 err = -ESRCH;
1194 goto out;
1195 }
d19d5476
DR		 1196
1197 task_lock(task);
1198 if (!task->mm) {
1199 err = -EINVAL;
1200 goto err_task_lock;
1201 }
1202
a63d83f4 1203 if (!lock_task_sighand(task, &flags)) {
723548bf 1204 err = -ESRCH;
d19d5476 1205 goto err_task_lock;
a63d83f4 1206 }
d19d5476 1207
dabb16f6 1208 if (oom_score_adj < task->signal->oom_score_adj_min &&
a63d83f4 1209 !capable(CAP_SYS_RESOURCE)) {
723548bf
DR		 1210 err = -EACCES;
1211 goto err_sighand;
a63d83f4
DR		 1212 }
1213
3d5992d2
YH		 1214 if (oom_score_adj != task->signal->oom_score_adj) {
1215 if (oom_score_adj == OOM_SCORE_ADJ_MIN)
1216 atomic_inc(&task->mm->oom_disable_count);
1217 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1218 atomic_dec(&task->mm->oom_disable_count);
1219 }
a63d83f4 1220 task->signal->oom_score_adj = oom_score_adj;
dabb16f6
MSB		 1221 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1222 task->signal->oom_score_adj_min = oom_score_adj;
a63d83f4
DR		 1223 /*
1224 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1225 * always attainable.
1226 */
1227 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1228 task->signal->oom_adj = OOM_DISABLE;
1229 else
1230 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1231 OOM_SCORE_ADJ_MAX;
723548bf 1232err_sighand:
a63d83f4 1233 unlock_task_sighand(task, &flags);
d19d5476
DR		 1234err_task_lock:
1235 task_unlock(task);
a63d83f4 1236 put_task_struct(task);
723548bf
DR		 1237out:
1238 return err < 0 ? err : count;
a63d83f4
DR		 1239}
1240
1241static const struct file_operations proc_oom_score_adj_operations = {
1242 .read = oom_score_adj_read,
1243 .write = oom_score_adj_write,
6038f373 1244 .llseek = default_llseek,
a63d83f4
DR		 1245};
1246
1da177e4
LT		 1247#ifdef CONFIG_AUDITSYSCALL
1248#define TMPBUFLEN 21
1249static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1250 size_t count, loff_t *ppos)
1251{
2fddfeef 1252 struct inode * inode = file->f_path.dentry->d_inode;
99f89551 1253 struct task_struct *task = get_proc_task(inode);
1da177e4
LT		 1254 ssize_t length;
1255 char tmpbuf[TMPBUFLEN];
1256
99f89551
EB		 1257 if (!task)
1258 return -ESRCH;
1da177e4 1259 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
0c11b942 1260 audit_get_loginuid(task));
99f89551 1261 put_task_struct(task);
1da177e4
LT		 1262 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1263}
1264
1265static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1266 size_t count, loff_t *ppos)
1267{
2fddfeef 1268 struct inode * inode = file->f_path.dentry->d_inode;
1da177e4
LT		 1269 char *page, *tmp;
1270 ssize_t length;
1da177e4
LT		 1271 uid_t loginuid;
1272
1273 if (!capable(CAP_AUDIT_CONTROL))
1274 return -EPERM;
1275
7dc52157
PM		 1276 rcu_read_lock();
1277 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1278 rcu_read_unlock();
1da177e4 1279 return -EPERM;
7dc52157
PM		 1280 }
1281 rcu_read_unlock();
1da177e4 1282
e0182909
AV		 1283 if (count >= PAGE_SIZE)
1284 count = PAGE_SIZE - 1;
1da177e4
LT		 1285
1286 if (*ppos != 0) {
1287 /* No partial writes. */
1288 return -EINVAL;
1289 }
e12ba74d 1290 page = (char*)__get_free_page(GFP_TEMPORARY);
1da177e4
LT		 1291 if (!page)
1292 return -ENOMEM;
1293 length = -EFAULT;
1294 if (copy_from_user(page, buf, count))
1295 goto out_free_page;
1296
e0182909 1297 page[count] = '\0';
1da177e4
LT		 1298 loginuid = simple_strtoul(page, &tmp, 10);
1299 if (tmp == page) {
1300 length = -EINVAL;
1301 goto out_free_page;
1302
1303 }
99f89551 1304 length = audit_set_loginuid(current, loginuid);
1da177e4
LT		 1305 if (likely(length == 0))
1306 length = count;
1307
1308out_free_page:
1309 free_page((unsigned long) page);
1310 return length;
1311}
1312
00977a59 1313static const struct file_operations proc_loginuid_operations = {
1da177e4
LT		 1314 .read = proc_loginuid_read,
1315 .write = proc_loginuid_write,
87df8424 1316 .llseek = generic_file_llseek,
1da177e4 1317};
1e0bd755
EP		 1318
1319static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1320 size_t count, loff_t *ppos)
1321{
1322 struct inode * inode = file->f_path.dentry->d_inode;
1323 struct task_struct *task = get_proc_task(inode);
1324 ssize_t length;
1325 char tmpbuf[TMPBUFLEN];
1326
1327 if (!task)
1328 return -ESRCH;
1329 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1330 audit_get_sessionid(task));
1331 put_task_struct(task);
1332 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1333}
1334
1335static const struct file_operations proc_sessionid_operations = {
1336 .read = proc_sessionid_read,
87df8424 1337 .llseek = generic_file_llseek,
1e0bd755 1338};
1da177e4
LT		 1339#endif
1340
f4f154fd
AM		 1341#ifdef CONFIG_FAULT_INJECTION
1342static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1343 size_t count, loff_t *ppos)
1344{
1345 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1346 char buffer[PROC_NUMBUF];
1347 size_t len;
1348 int make_it_fail;
f4f154fd
AM		 1349
1350 if (!task)
1351 return -ESRCH;
1352 make_it_fail = task->make_it_fail;
1353 put_task_struct(task);
1354
1355 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
0c28f287
AM		 1356
1357 return simple_read_from_buffer(buf, count, ppos, buffer, len);
f4f154fd
AM		 1358}
1359
1360static ssize_t proc_fault_inject_write(struct file * file,
1361 const char __user * buf, size_t count, loff_t *ppos)
1362{
1363 struct task_struct *task;
1364 char buffer[PROC_NUMBUF], *end;
1365 int make_it_fail;
1366
1367 if (!capable(CAP_SYS_RESOURCE))
1368 return -EPERM;
1369 memset(buffer, 0, sizeof(buffer));
1370 if (count > sizeof(buffer) - 1)
1371 count = sizeof(buffer) - 1;
1372 if (copy_from_user(buffer, buf, count))
1373 return -EFAULT;
cba8aafe
VL		 1374 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1375 if (*end)
1376 return -EINVAL;
f4f154fd
AM		 1377 task = get_proc_task(file->f_dentry->d_inode);
1378 if (!task)
1379 return -ESRCH;
1380 task->make_it_fail = make_it_fail;
1381 put_task_struct(task);
cba8aafe
VL		 1382
1383 return count;
f4f154fd
AM		 1384}
1385
00977a59 1386static const struct file_operations proc_fault_inject_operations = {
f4f154fd
AM		 1387 .read = proc_fault_inject_read,
1388 .write = proc_fault_inject_write,
87df8424 1389 .llseek = generic_file_llseek,
f4f154fd
AM		 1390};
1391#endif
1392
9745512c 1393
43ae34cb
IM		 1394#ifdef CONFIG_SCHED_DEBUG
1395/*
1396 * Print out various scheduling related per-task fields:
1397 */
1398static int sched_show(struct seq_file *m, void *v)
1399{
1400 struct inode *inode = m->private;
1401 struct task_struct *p;
1402
43ae34cb
IM		 1403 p = get_proc_task(inode);
1404 if (!p)
1405 return -ESRCH;
1406 proc_sched_show_task(p, m);
1407
1408 put_task_struct(p);
1409
1410 return 0;
1411}
1412
1413static ssize_t
1414sched_write(struct file *file, const char __user *buf,
1415 size_t count, loff_t *offset)
1416{
1417 struct inode *inode = file->f_path.dentry->d_inode;
1418 struct task_struct *p;
1419
43ae34cb
IM		 1420 p = get_proc_task(inode);
1421 if (!p)
1422 return -ESRCH;
1423 proc_sched_set_task(p);
1424
1425 put_task_struct(p);
1426
1427 return count;
1428}
1429
1430static int sched_open(struct inode *inode, struct file *filp)
1431{
c6a34058 1432 return single_open(filp, sched_show, inode);
43ae34cb
IM		 1433}
1434
1435static const struct file_operations proc_pid_sched_operations = {
1436 .open = sched_open,
1437 .read = seq_read,
1438 .write = sched_write,
1439 .llseek = seq_lseek,
5ea473a1 1440 .release = single_release,
43ae34cb
IM		 1441};
1442
1443#endif
1444
5091faa4
MG		 1445#ifdef CONFIG_SCHED_AUTOGROUP
1446/*
1447 * Print out autogroup related information:
1448 */
1449static int sched_autogroup_show(struct seq_file *m, void *v)
1450{
1451 struct inode *inode = m->private;
1452 struct task_struct *p;
1453
1454 p = get_proc_task(inode);
1455 if (!p)
1456 return -ESRCH;
1457 proc_sched_autogroup_show_task(p, m);
1458
1459 put_task_struct(p);
1460
1461 return 0;
1462}
1463
1464static ssize_t
1465sched_autogroup_write(struct file *file, const char __user *buf,
1466 size_t count, loff_t *offset)
1467{
1468 struct inode *inode = file->f_path.dentry->d_inode;
1469 struct task_struct *p;
1470 char buffer[PROC_NUMBUF];
1471 long nice;
1472 int err;
1473
1474 memset(buffer, 0, sizeof(buffer));
1475 if (count > sizeof(buffer) - 1)
1476 count = sizeof(buffer) - 1;
1477 if (copy_from_user(buffer, buf, count))
1478 return -EFAULT;
1479
1480 err = strict_strtol(strstrip(buffer), 0, &nice);
1481 if (err)
1482 return -EINVAL;
1483
1484 p = get_proc_task(inode);
1485 if (!p)
1486 return -ESRCH;
1487
1488 err = nice;
1489 err = proc_sched_autogroup_set_nice(p, &err);
1490 if (err)
1491 count = err;
1492
1493 put_task_struct(p);
1494
1495 return count;
1496}
1497
1498static int sched_autogroup_open(struct inode *inode, struct file *filp)
1499{
1500 int ret;
1501
1502 ret = single_open(filp, sched_autogroup_show, NULL);
1503 if (!ret) {
1504 struct seq_file *m = filp->private_data;
1505
1506 m->private = inode;
1507 }
1508 return ret;
1509}
1510
1511static const struct file_operations proc_pid_sched_autogroup_operations = {
1512 .open = sched_autogroup_open,
1513 .read = seq_read,
1514 .write = sched_autogroup_write,
1515 .llseek = seq_lseek,
1516 .release = single_release,
1517};
1518
1519#endif /* CONFIG_SCHED_AUTOGROUP */
1520
4614a696 1521static ssize_t comm_write(struct file *file, const char __user *buf,
1522 size_t count, loff_t *offset)
1523{
1524 struct inode *inode = file->f_path.dentry->d_inode;
1525 struct task_struct *p;
1526 char buffer[TASK_COMM_LEN];
1527
1528 memset(buffer, 0, sizeof(buffer));
1529 if (count > sizeof(buffer) - 1)
1530 count = sizeof(buffer) - 1;
1531 if (copy_from_user(buffer, buf, count))
1532 return -EFAULT;
1533
1534 p = get_proc_task(inode);
1535 if (!p)
1536 return -ESRCH;
1537
1538 if (same_thread_group(current, p))
1539 set_task_comm(p, buffer);
1540 else
1541 count = -EINVAL;
1542
1543 put_task_struct(p);
1544
1545 return count;
1546}
1547
1548static int comm_show(struct seq_file *m, void *v)
1549{
1550 struct inode *inode = m->private;
1551 struct task_struct *p;
1552
1553 p = get_proc_task(inode);
1554 if (!p)
1555 return -ESRCH;
1556
1557 task_lock(p);
1558 seq_printf(m, "%s\n", p->comm);
1559 task_unlock(p);
1560
1561 put_task_struct(p);
1562
1563 return 0;
1564}
1565
1566static int comm_open(struct inode *inode, struct file *filp)
1567{
c6a34058 1568 return single_open(filp, comm_show, inode);
4614a696 1569}
1570
1571static const struct file_operations proc_pid_set_comm_operations = {
1572 .open = comm_open,
1573 .read = seq_read,
1574 .write = comm_write,
1575 .llseek = seq_lseek,
1576 .release = single_release,
1577};
1578
925d1c40
MH		 1579/*
1580 * We added or removed a vma mapping the executable. The vmas are only mapped
1581 * during exec and are not mapped with the mmap system call.
1582 * Callers must hold down_write() on the mm's mmap_sem for these
1583 */
1584void added_exe_file_vma(struct mm_struct *mm)
1585{
1586 mm->num_exe_file_vmas++;
1587}
1588
1589void removed_exe_file_vma(struct mm_struct *mm)
1590{
1591 mm->num_exe_file_vmas--;
1592 if ((mm->num_exe_file_vmas == 0) && mm->exe_file){
1593 fput(mm->exe_file);
1594 mm->exe_file = NULL;
1595 }
1596
1597}
1598
1599void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1600{
1601 if (new_exe_file)
1602 get_file(new_exe_file);
1603 if (mm->exe_file)
1604 fput(mm->exe_file);
1605 mm->exe_file = new_exe_file;
1606 mm->num_exe_file_vmas = 0;
1607}
1608
1609struct file *get_mm_exe_file(struct mm_struct *mm)
1610{
1611 struct file *exe_file;
1612
1613 /* We need mmap_sem to protect against races with removal of
1614 * VM_EXECUTABLE vmas */
1615 down_read(&mm->mmap_sem);
1616 exe_file = mm->exe_file;
1617 if (exe_file)
1618 get_file(exe_file);
1619 up_read(&mm->mmap_sem);
1620 return exe_file;
1621}
1622
1623void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
1624{
1625 /* It's safe to write the exe_file pointer without exe_file_lock because
1626 * this is called during fork when the task is not yet in /proc */
1627 newmm->exe_file = get_mm_exe_file(oldmm);
1628}
1629
1630static int proc_exe_link(struct inode *inode, struct path *exe_path)
1631{
1632 struct task_struct *task;
1633 struct mm_struct *mm;
1634 struct file *exe_file;
1635
1636 task = get_proc_task(inode);
1637 if (!task)
1638 return -ENOENT;
1639 mm = get_task_mm(task);
1640 put_task_struct(task);
1641 if (!mm)
1642 return -ENOENT;
1643 exe_file = get_mm_exe_file(mm);
1644 mmput(mm);
1645 if (exe_file) {
1646 *exe_path = exe_file->f_path;
1647 path_get(&exe_file->f_path);
1648 fput(exe_file);
1649 return 0;
1650 } else
1651 return -ENOENT;
1652}
1653
008b150a 1654static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1da177e4
LT		 1655{
1656 struct inode *inode = dentry->d_inode;
1657 int error = -EACCES;
1658
1659 /* We don't need a base pointer in the /proc filesystem */
1d957f9b 1660 path_put(&nd->path);
1da177e4 1661
778c1144
EB		 1662 /* Are we allowed to snoop on the tasks file descriptors? */
1663 if (!proc_fd_access_allowed(inode))
1da177e4 1664 goto out;
1da177e4 1665
3dcd25f3 1666 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
1da177e4 1667out:
008b150a 1668 return ERR_PTR(error);
1da177e4
LT		 1669}
1670
3dcd25f3 1671static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1da177e4 1672{
e12ba74d 1673 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
3dcd25f3 1674 char *pathname;
1da177e4
LT		 1675 int len;
1676
1677 if (!tmp)
1678 return -ENOMEM;
0c28f287 1679
7b2a69ba 1680 pathname = d_path(path, tmp, PAGE_SIZE);
3dcd25f3
JB		 1681 len = PTR_ERR(pathname);
1682 if (IS_ERR(pathname))
1da177e4 1683 goto out;
3dcd25f3 1684 len = tmp + PAGE_SIZE - 1 - pathname;
1da177e4
LT		 1685
1686 if (len > buflen)
1687 len = buflen;
3dcd25f3 1688 if (copy_to_user(buffer, pathname, len))
1da177e4
LT		 1689 len = -EFAULT;
1690 out:
1691 free_page((unsigned long)tmp);
1692 return len;
1693}
1694
1695static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1696{
1697 int error = -EACCES;
1698 struct inode *inode = dentry->d_inode;
3dcd25f3 1699 struct path path;
1da177e4 1700
778c1144
EB		 1701 /* Are we allowed to snoop on the tasks file descriptors? */
1702 if (!proc_fd_access_allowed(inode))
1da177e4 1703 goto out;
1da177e4 1704
3dcd25f3 1705 error = PROC_I(inode)->op.proc_get_link(inode, &path);
1da177e4
LT		 1706 if (error)
1707 goto out;
1708
3dcd25f3
JB		 1709 error = do_proc_readlink(&path, buffer, buflen);
1710 path_put(&path);
1da177e4 1711out:
1da177e4
LT		 1712 return error;
1713}
1714
c5ef1c42 1715static const struct inode_operations proc_pid_link_inode_operations = {
1da177e4 1716 .readlink = proc_pid_readlink,
6d76fa58
LT		 1717 .follow_link = proc_pid_follow_link,
1718 .setattr = proc_setattr,
1da177e4
LT		 1719};
1720
28a6d671
EB		 1721
1722/* building an inode */
1723
1724static int task_dumpable(struct task_struct *task)
1da177e4 1725{
28a6d671
EB		 1726 int dumpable = 0;
1727 struct mm_struct *mm;
1da177e4 1728
28a6d671
EB		 1729 task_lock(task);
1730 mm = task->mm;
1731 if (mm)
6c5d5238 1732 dumpable = get_dumpable(mm);
28a6d671
EB		 1733 task_unlock(task);
1734 if(dumpable == 1)
1735 return 1;
1736 return 0;
1737}
1da177e4 1738
1da177e4 1739
61a28784 1740static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
28a6d671
EB		 1741{
1742 struct inode * inode;
1743 struct proc_inode *ei;
c69e8d9c 1744 const struct cred *cred;
1da177e4 1745
28a6d671 1746 /* We need a new inode */
1da177e4 1747
28a6d671
EB		 1748 inode = new_inode(sb);
1749 if (!inode)
1750 goto out;
1751
1752 /* Common stuff */
1753 ei = PROC_I(inode);
85fe4025 1754 inode->i_ino = get_next_ino();
28a6d671 1755 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
28a6d671
EB		 1756 inode->i_op = &proc_def_inode_operations;
1757
1758 /*
1759 * grab the reference to task.
1760 */
1a657f78 1761 ei->pid = get_task_pid(task, PIDTYPE_PID);
28a6d671
EB		 1762 if (!ei->pid)
1763 goto out_unlock;
1764
28a6d671 1765 if (task_dumpable(task)) {
c69e8d9c
DH		 1766 rcu_read_lock();
1767 cred = __task_cred(task);
1768 inode->i_uid = cred->euid;
1769 inode->i_gid = cred->egid;
1770 rcu_read_unlock();
1da177e4 1771 }
28a6d671
EB		 1772 security_task_to_inode(task, inode);
1773
1da177e4 1774out:
28a6d671
EB		 1775 return inode;
1776
1777out_unlock:
1778 iput(inode);
1779 return NULL;
1da177e4
LT		 1780}
1781
28a6d671 1782static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1da177e4 1783{
1da177e4 1784 struct inode *inode = dentry->d_inode;
28a6d671 1785 struct task_struct *task;
c69e8d9c
DH		 1786 const struct cred *cred;
1787
28a6d671 1788 generic_fillattr(inode, stat);
1da177e4 1789
28a6d671
EB		 1790 rcu_read_lock();
1791 stat->uid = 0;
1792 stat->gid = 0;
1793 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1794 if (task) {
1795 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1796 task_dumpable(task)) {
c69e8d9c
DH		 1797 cred = __task_cred(task);
1798 stat->uid = cred->euid;
1799 stat->gid = cred->egid;
1da177e4
LT		 1800 }
1801 }
28a6d671 1802 rcu_read_unlock();
d6e71144 1803 return 0;
1da177e4
LT		 1804}
1805
1da177e4
LT		 1806/* dentry stuff */
1807
1808/*
1809 * Exceptional case: normally we are not allowed to unhash a busy
1810 * directory. In this case, however, we can do it - no aliasing problems
1811 * due to the way we treat inodes.
1812 *
1813 * Rewrite the inode's ownerships here because the owning task may have
1814 * performed a setuid(), etc.
99f89551
EB		 1815 *
1816 * Before the /proc/pid/status file was created the only way to read
1817 * the effective uid of a /process was to stat /proc/pid. Reading
1818 * /proc/pid/status is slow enough that procps and other packages
1819 * kept stating /proc/pid. To keep the rules in /proc simple I have
1820 * made this apply to all per process world readable and executable
1821 * directories.
1da177e4
LT		 1822 */
1823static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1824{
34286d66
NP		 1825 struct inode *inode;
1826 struct task_struct *task;
c69e8d9c
DH		 1827 const struct cred *cred;
1828
34286d66
NP		 1829 if (nd && nd->flags & LOOKUP_RCU)
1830 return -ECHILD;
1831
1832 inode = dentry->d_inode;
1833 task = get_proc_task(inode);
1834
99f89551
EB		 1835 if (task) {
1836 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1837 task_dumpable(task)) {
c69e8d9c
DH		 1838 rcu_read_lock();
1839 cred = __task_cred(task);
1840 inode->i_uid = cred->euid;
1841 inode->i_gid = cred->egid;
1842 rcu_read_unlock();
1da177e4
LT		 1843 } else {
1844 inode->i_uid = 0;
1845 inode->i_gid = 0;
1846 }
9ee8ab9f 1847 inode->i_mode &= ~(S_ISUID | S_ISGID);
1da177e4 1848 security_task_to_inode(task, inode);
99f89551 1849 put_task_struct(task);
1da177e4
LT		 1850 return 1;
1851 }
1852 d_drop(dentry);
1853 return 0;
1854}
1855
fe15ce44 1856static int pid_delete_dentry(const struct dentry * dentry)
99f89551 1857{
28a6d671
EB		 1858 /* Is the task we represent dead?
1859 * If so, then don't put the dentry on the lru list,
1860 * kill it immediately.
1861 */
1862 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1863}
1864
d72f71eb 1865static const struct dentry_operations pid_dentry_operations =
28a6d671
EB		 1866{
1867 .d_revalidate = pid_revalidate,
1868 .d_delete = pid_delete_dentry,
1869};
1870
1871/* Lookups */
1872
c5141e6d
ED		 1873typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1874 struct task_struct *, const void *);
61a28784 1875
1c0d04c9
EB		 1876/*
1877 * Fill a directory entry.
1878 *
1879 * If possible create the dcache entry and derive our inode number and
1880 * file type from dcache entry.
1881 *
1882 * Since all of the proc inode numbers are dynamically generated, the inode
1883 * numbers do not exist until the inode is cache. This means creating the
1884 * the dcache entry in readdir is necessary to keep the inode numbers
1885 * reported by readdir in sync with the inode numbers reported
1886 * by stat.
1887 */
61a28784
EB		 1888static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1889 char *name, int len,
c5141e6d 1890 instantiate_t instantiate, struct task_struct *task, const void *ptr)
61a28784 1891{
2fddfeef 1892 struct dentry *child, *dir = filp->f_path.dentry;
61a28784
EB		 1893 struct inode *inode;
1894 struct qstr qname;
1895 ino_t ino = 0;
1896 unsigned type = DT_UNKNOWN;
1897
1898 qname.name = name;
1899 qname.len = len;
1900 qname.hash = full_name_hash(name, len);
1901
1902 child = d_lookup(dir, &qname);
1903 if (!child) {
1904 struct dentry *new;
1905 new = d_alloc(dir, &qname);
1906 if (new) {
1907 child = instantiate(dir->d_inode, new, task, ptr);
1908 if (child)
1909 dput(new);
1910 else
1911 child = new;
1912 }
1913 }
1914 if (!child || IS_ERR(child) || !child->d_inode)
1915 goto end_instantiate;
1916 inode = child->d_inode;
1917 if (inode) {
1918 ino = inode->i_ino;
1919 type = inode->i_mode >> 12;
1920 }
1921 dput(child);
1922end_instantiate:
1923 if (!ino)
1924 ino = find_inode_number(dir, &qname);
1925 if (!ino)
1926 ino = 1;
1927 return filldir(dirent, name, len, filp->f_pos, ino, type);
1928}
1929
28a6d671
EB		 1930static unsigned name_to_int(struct dentry *dentry)
1931{
1932 const char *name = dentry->d_name.name;
1933 int len = dentry->d_name.len;
1934 unsigned n = 0;
1935
1936 if (len > 1 && *name == '0')
1937 goto out;
1938 while (len-- > 0) {
1939 unsigned c = *name++ - '0';
1940 if (c > 9)
1941 goto out;
1942 if (n >= (~0U-9)/10)
1943 goto out;
1944 n *= 10;
1945 n += c;
1946 }
1947 return n;
1948out:
1949 return ~0U;
1950}
1951
27932742
MS		 1952#define PROC_FDINFO_MAX 64
1953
3dcd25f3 1954static int proc_fd_info(struct inode *inode, struct path *path, char *info)
28a6d671
EB		 1955{
1956 struct task_struct *task = get_proc_task(inode);
1957 struct files_struct *files = NULL;
1958 struct file *file;
1959 int fd = proc_fd(inode);
99f89551 1960
99f89551 1961 if (task) {
28a6d671
EB		 1962 files = get_files_struct(task);
1963 put_task_struct(task);
1964 }
1965 if (files) {
1966 /*
1967 * We are not taking a ref to the file structure, so we must
1968 * hold ->file_lock.
1969 */
1970 spin_lock(&files->file_lock);
1971 file = fcheck_files(files, fd);
1972 if (file) {
3dcd25f3
JB		 1973 if (path) {
1974 *path = file->f_path;
1975 path_get(&file->f_path);
1976 }
27932742
MS		 1977 if (info)
1978 snprintf(info, PROC_FDINFO_MAX,
1979 "pos:\t%lli\n"
1980 "flags:\t0%o\n",
1981 (long long) file->f_pos,
1982 file->f_flags);
28a6d671
EB		 1983 spin_unlock(&files->file_lock);
1984 put_files_struct(files);
1985 return 0;
99f89551 1986 }
28a6d671
EB		 1987 spin_unlock(&files->file_lock);
1988 put_files_struct(files);
99f89551 1989 }
28a6d671 1990 return -ENOENT;
99f89551
EB		 1991}
1992
3dcd25f3 1993static int proc_fd_link(struct inode *inode, struct path *path)
27932742 1994{
3dcd25f3 1995 return proc_fd_info(inode, path, NULL);
27932742
MS		 1996}
1997
1da177e4
LT		 1998static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1999{
34286d66
NP		 2000 struct inode *inode;
2001 struct task_struct *task;
2002 int fd;
1da177e4 2003 struct files_struct *files;
c69e8d9c 2004 const struct cred *cred;
1da177e4 2005
34286d66
NP		 2006 if (nd && nd->flags & LOOKUP_RCU)
2007 return -ECHILD;
2008
2009 inode = dentry->d_inode;
2010 task = get_proc_task(inode);
2011 fd = proc_fd(inode);
2012
99f89551
EB		 2013 if (task) {
2014 files = get_files_struct(task);
2015 if (files) {
2016 rcu_read_lock();
2017 if (fcheck_files(files, fd)) {
2018 rcu_read_unlock();
2019 put_files_struct(files);
2020 if (task_dumpable(task)) {
c69e8d9c
DH		 2021 rcu_read_lock();
2022 cred = __task_cred(task);
2023 inode->i_uid = cred->euid;
2024 inode->i_gid = cred->egid;
2025 rcu_read_unlock();
99f89551
EB		 2026 } else {
2027 inode->i_uid = 0;
2028 inode->i_gid = 0;
2029 }
9ee8ab9f 2030 inode->i_mode &= ~(S_ISUID | S_ISGID);
99f89551
EB		 2031 security_task_to_inode(task, inode);
2032 put_task_struct(task);
2033 return 1;
2034 }
b835996f 2035 rcu_read_unlock();
1da177e4 2036 put_files_struct(files);
1da177e4 2037 }
99f89551 2038 put_task_struct(task);
1da177e4
LT		 2039 }
2040 d_drop(dentry);
2041 return 0;
2042}
2043
d72f71eb 2044static const struct dentry_operations tid_fd_dentry_operations =
1da177e4
LT		 2045{
2046 .d_revalidate = tid_fd_revalidate,
2047 .d_delete = pid_delete_dentry,
2048};
2049
444ceed8 2050static struct dentry *proc_fd_instantiate(struct inode *dir,
c5141e6d 2051 struct dentry *dentry, struct task_struct *task, const void *ptr)
1da177e4 2052{
c5141e6d 2053 unsigned fd = *(const unsigned *)ptr;
444ceed8
EB		 2054 struct file *file;
2055 struct files_struct *files;
2056 struct inode *inode;
2057 struct proc_inode *ei;
2058 struct dentry *error = ERR_PTR(-ENOENT);
1da177e4 2059
61a28784 2060 inode = proc_pid_make_inode(dir->i_sb, task);
1da177e4
LT		 2061 if (!inode)
2062 goto out;
2063 ei = PROC_I(inode);
aed7a6c4 2064 ei->fd = fd;
1da177e4
LT		 2065 files = get_files_struct(task);
2066 if (!files)
444ceed8 2067 goto out_iput;
1da177e4 2068 inode->i_mode = S_IFLNK;
ca99c1da
DS		 2069
2070 /*
2071 * We are not taking a ref to the file structure, so we must
2072 * hold ->file_lock.
2073 */
2074 spin_lock(&files->file_lock);
1da177e4
LT		 2075 file = fcheck_files(files, fd);
2076 if (!file)
444ceed8 2077 goto out_unlock;
aeb5d727 2078 if (file->f_mode & FMODE_READ)
1da177e4 2079 inode->i_mode |= S_IRUSR | S_IXUSR;
aeb5d727 2080 if (file->f_mode & FMODE_WRITE)
1da177e4 2081 inode->i_mode |= S_IWUSR | S_IXUSR;
ca99c1da 2082 spin_unlock(&files->file_lock);
1da177e4 2083 put_files_struct(files);
444ceed8 2084
1da177e4
LT		 2085 inode->i_op = &proc_pid_link_inode_operations;
2086 inode->i_size = 64;
2087 ei->op.proc_get_link = proc_fd_link;
fb045adb 2088 d_set_d_op(dentry, &tid_fd_dentry_operations);
1da177e4 2089 d_add(dentry, inode);
cd6a3ce9
EB		 2090 /* Close the race of the process dying before we return the dentry */
2091 if (tid_fd_revalidate(dentry, NULL))
444ceed8 2092 error = NULL;
1da177e4 2093
444ceed8
EB		 2094 out:
2095 return error;
2096out_unlock:
ca99c1da 2097 spin_unlock(&files->file_lock);
1da177e4 2098 put_files_struct(files);
444ceed8 2099out_iput:
1da177e4 2100 iput(inode);
cd6a3ce9 2101 goto out;
1da177e4
LT		 2102}
2103
27932742
MS		 2104static struct dentry *proc_lookupfd_common(struct inode *dir,
2105 struct dentry *dentry,
2106 instantiate_t instantiate)
444ceed8
EB		 2107{
2108 struct task_struct *task = get_proc_task(dir);
2109 unsigned fd = name_to_int(dentry);
2110 struct dentry *result = ERR_PTR(-ENOENT);
2111
2112 if (!task)
2113 goto out_no_task;
2114 if (fd == ~0U)
2115 goto out;
2116
27932742 2117 result = instantiate(dir, dentry, task, &fd);
444ceed8
EB		 2118out:
2119 put_task_struct(task);
2120out_no_task:
2121 return result;
2122}
2123
27932742
MS		 2124static int proc_readfd_common(struct file * filp, void * dirent,
2125 filldir_t filldir, instantiate_t instantiate)
28a6d671 2126{
2fddfeef 2127 struct dentry *dentry = filp->f_path.dentry;
28a6d671
EB		 2128 struct inode *inode = dentry->d_inode;
2129 struct task_struct *p = get_proc_task(inode);
457c2510 2130 unsigned int fd, ino;
28a6d671 2131 int retval;
28a6d671 2132 struct files_struct * files;
1da177e4 2133
28a6d671
EB		 2134 retval = -ENOENT;
2135 if (!p)
2136 goto out_no_task;
2137 retval = 0;
28a6d671
EB		 2138
2139 fd = filp->f_pos;
2140 switch (fd) {
2141 case 0:
2142 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
2143 goto out;
2144 filp->f_pos++;
2145 case 1:
2146 ino = parent_ino(dentry);
2147 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2148 goto out;
2149 filp->f_pos++;
2150 default:
2151 files = get_files_struct(p);
2152 if (!files)
2153 goto out;
2154 rcu_read_lock();
28a6d671 2155 for (fd = filp->f_pos-2;
9b4f526c 2156 fd < files_fdtable(files)->max_fds;
28a6d671 2157 fd++, filp->f_pos++) {
27932742
MS		 2158 char name[PROC_NUMBUF];
2159 int len;
28a6d671
EB		 2160
2161 if (!fcheck_files(files, fd))
2162 continue;
2163 rcu_read_unlock();
2164
27932742
MS		 2165 len = snprintf(name, sizeof(name), "%d", fd);
2166 if (proc_fill_cache(filp, dirent, filldir,
2167 name, len, instantiate,
2168 p, &fd) < 0) {
28a6d671
EB		 2169 rcu_read_lock();
2170 break;
2171 }
2172 rcu_read_lock();
2173 }
2174 rcu_read_unlock();
2175 put_files_struct(files);
2176 }
2177out:
2178 put_task_struct(p);
2179out_no_task:
2180 return retval;
2181}
2182
27932742
MS		 2183static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
2184 struct nameidata *nd)
2185{
2186 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
2187}
2188
2189static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
2190{
2191 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
2192}
2193
2194static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
2195 size_t len, loff_t *ppos)
2196{
2197 char tmp[PROC_FDINFO_MAX];
3dcd25f3 2198 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
27932742
MS		 2199 if (!err)
2200 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
2201 return err;
2202}
2203
2204static const struct file_operations proc_fdinfo_file_operations = {
87df8424 2205 .open = nonseekable_open,
27932742 2206 .read = proc_fdinfo_read,
6038f373 2207 .llseek = no_llseek,
27932742
MS		 2208};
2209
00977a59 2210static const struct file_operations proc_fd_operations = {
28a6d671
EB		 2211 .read = generic_read_dir,
2212 .readdir = proc_readfd,
6038f373 2213 .llseek = default_llseek,
1da177e4
LT		 2214};
2215
8948e11f
AD		 2216/*
2217 * /proc/pid/fd needs a special permission handler so that a process can still
2218 * access /proc/self/fd after it has executed a setuid().
2219 */
b74c79e9 2220static int proc_fd_permission(struct inode *inode, int mask, unsigned int flags)
8948e11f
AD		 2221{
2222 int rv;
2223
b74c79e9
NP		 2224 if (flags & IPERM_FLAG_RCU)
2225 return -ECHILD;
2226 rv = generic_permission(inode, mask, flags, NULL);
8948e11f
AD		 2227 if (rv == 0)
2228 return 0;
2229 if (task_pid(current) == proc_pid(inode))
2230 rv = 0;
2231 return rv;
2232}
2233
1da177e4
LT		 2234/*
2235 * proc directories can do almost nothing..
2236 */
c5ef1c42 2237static const struct inode_operations proc_fd_inode_operations = {
1da177e4 2238 .lookup = proc_lookupfd,
8948e11f 2239 .permission = proc_fd_permission,
6d76fa58 2240 .setattr = proc_setattr,
1da177e4
LT		 2241};
2242
27932742
MS		 2243static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2244 struct dentry *dentry, struct task_struct *task, const void *ptr)
2245{
2246 unsigned fd = *(unsigned *)ptr;
2247 struct inode *inode;
2248 struct proc_inode *ei;
2249 struct dentry *error = ERR_PTR(-ENOENT);
2250
2251 inode = proc_pid_make_inode(dir->i_sb, task);
2252 if (!inode)
2253 goto out;
2254 ei = PROC_I(inode);
2255 ei->fd = fd;
2256 inode->i_mode = S_IFREG | S_IRUSR;
2257 inode->i_fop = &proc_fdinfo_file_operations;
fb045adb 2258 d_set_d_op(dentry, &tid_fd_dentry_operations);
27932742
MS		 2259 d_add(dentry, inode);
2260 /* Close the race of the process dying before we return the dentry */
2261 if (tid_fd_revalidate(dentry, NULL))
2262 error = NULL;
2263
2264 out:
2265 return error;
2266}
2267
2268static struct dentry *proc_lookupfdinfo(struct inode *dir,
2269 struct dentry *dentry,
2270 struct nameidata *nd)
2271{
2272 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2273}
2274
2275static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2276{
2277 return proc_readfd_common(filp, dirent, filldir,
2278 proc_fdinfo_instantiate);
2279}
2280
2281static const struct file_operations proc_fdinfo_operations = {
2282 .read = generic_read_dir,
2283 .readdir = proc_readfdinfo,
6038f373 2284 .llseek = default_llseek,
27932742
MS		 2285};
2286
2287/*
2288 * proc directories can do almost nothing..
2289 */
2290static const struct inode_operations proc_fdinfo_inode_operations = {
2291 .lookup = proc_lookupfdinfo,
2292 .setattr = proc_setattr,
2293};
2294
2295
444ceed8 2296static struct dentry *proc_pident_instantiate(struct inode *dir,
c5141e6d 2297 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8 2298{
c5141e6d 2299 const struct pid_entry *p = ptr;
444ceed8
EB		 2300 struct inode *inode;
2301 struct proc_inode *ei;
bd6daba9 2302 struct dentry *error = ERR_PTR(-ENOENT);
444ceed8 2303
61a28784 2304 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 2305 if (!inode)
2306 goto out;
2307
2308 ei = PROC_I(inode);
2309 inode->i_mode = p->mode;
2310 if (S_ISDIR(inode->i_mode))
2311 inode->i_nlink = 2; /* Use getattr to fix if necessary */
2312 if (p->iop)
2313 inode->i_op = p->iop;
2314 if (p->fop)
2315 inode->i_fop = p->fop;
2316 ei->op = p->op;
fb045adb 2317 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 2318 d_add(dentry, inode);
2319 /* Close the race of the process dying before we return the dentry */
2320 if (pid_revalidate(dentry, NULL))
2321 error = NULL;
2322out:
2323 return error;
2324}
2325
1da177e4
LT		 2326static struct dentry *proc_pident_lookup(struct inode *dir,
2327 struct dentry *dentry,
c5141e6d 2328 const struct pid_entry *ents,
7bcd6b0e 2329 unsigned int nents)
1da177e4 2330{
cd6a3ce9 2331 struct dentry *error;
99f89551 2332 struct task_struct *task = get_proc_task(dir);
c5141e6d 2333 const struct pid_entry *p, *last;
1da177e4 2334
cd6a3ce9 2335 error = ERR_PTR(-ENOENT);
1da177e4 2336
99f89551
EB		 2337 if (!task)
2338 goto out_no_task;
1da177e4 2339
20cdc894
EB		 2340 /*
2341 * Yes, it does not scale. And it should not. Don't add
2342 * new entries into /proc/<tgid>/ without very good reasons.
2343 */
7bcd6b0e
EB		 2344 last = &ents[nents - 1];
2345 for (p = ents; p <= last; p++) {
1da177e4
LT		 2346 if (p->len != dentry->d_name.len)
2347 continue;
2348 if (!memcmp(dentry->d_name.name, p->name, p->len))
2349 break;
2350 }
7bcd6b0e 2351 if (p > last)
1da177e4
LT		 2352 goto out;
2353
444ceed8 2354 error = proc_pident_instantiate(dir, dentry, task, p);
1da177e4 2355out:
99f89551
EB		 2356 put_task_struct(task);
2357out_no_task:
cd6a3ce9 2358 return error;
1da177e4
LT		 2359}
2360
c5141e6d
ED		 2361static int proc_pident_fill_cache(struct file *filp, void *dirent,
2362 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
61a28784
EB		 2363{
2364 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2365 proc_pident_instantiate, task, p);
2366}
2367
28a6d671
EB		 2368static int proc_pident_readdir(struct file *filp,
2369 void *dirent, filldir_t filldir,
c5141e6d 2370 const struct pid_entry *ents, unsigned int nents)
28a6d671
EB		 2371{
2372 int i;
2fddfeef 2373 struct dentry *dentry = filp->f_path.dentry;
28a6d671
EB		 2374 struct inode *inode = dentry->d_inode;
2375 struct task_struct *task = get_proc_task(inode);
c5141e6d 2376 const struct pid_entry *p, *last;
28a6d671
EB		 2377 ino_t ino;
2378 int ret;
2379
2380 ret = -ENOENT;
2381 if (!task)
61a28784 2382 goto out_no_task;
28a6d671
EB		 2383
2384 ret = 0;
28a6d671
EB		 2385 i = filp->f_pos;
2386 switch (i) {
2387 case 0:
2388 ino = inode->i_ino;
2389 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2390 goto out;
2391 i++;
2392 filp->f_pos++;
2393 /* fall through */
2394 case 1:
2395 ino = parent_ino(dentry);
2396 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2397 goto out;
2398 i++;
2399 filp->f_pos++;
2400 /* fall through */
2401 default:
2402 i -= 2;
2403 if (i >= nents) {
2404 ret = 1;
2405 goto out;
2406 }
2407 p = ents + i;
7bcd6b0e
EB		 2408 last = &ents[nents - 1];
2409 while (p <= last) {
61a28784 2410 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
28a6d671
EB		 2411 goto out;
2412 filp->f_pos++;
2413 p++;
2414 }
2415 }
2416
2417 ret = 1;
2418out:
61a28784
EB		 2419 put_task_struct(task);
2420out_no_task:
28a6d671 2421 return ret;
1da177e4
LT		 2422}
2423
28a6d671
EB		 2424#ifdef CONFIG_SECURITY
2425static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2426 size_t count, loff_t *ppos)
2427{
2fddfeef 2428 struct inode * inode = file->f_path.dentry->d_inode;
04ff9708 2429 char *p = NULL;
28a6d671
EB		 2430 ssize_t length;
2431 struct task_struct *task = get_proc_task(inode);
2432
28a6d671 2433 if (!task)
04ff9708 2434 return -ESRCH;
28a6d671
EB		 2435
2436 length = security_getprocattr(task,
2fddfeef 2437 (char*)file->f_path.dentry->d_name.name,
04ff9708 2438 &p);
28a6d671 2439 put_task_struct(task);
04ff9708
AV		 2440 if (length > 0)
2441 length = simple_read_from_buffer(buf, count, ppos, p, length);
2442 kfree(p);
28a6d671 2443 return length;
1da177e4
LT		 2444}
2445
28a6d671
EB		 2446static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2447 size_t count, loff_t *ppos)
2448{
2fddfeef 2449 struct inode * inode = file->f_path.dentry->d_inode;
28a6d671
EB		 2450 char *page;
2451 ssize_t length;
2452 struct task_struct *task = get_proc_task(inode);
2453
2454 length = -ESRCH;
2455 if (!task)
2456 goto out_no_task;
2457 if (count > PAGE_SIZE)
2458 count = PAGE_SIZE;
2459
2460 /* No partial writes. */
2461 length = -EINVAL;
2462 if (*ppos != 0)
2463 goto out;
2464
2465 length = -ENOMEM;
e12ba74d 2466 page = (char*)__get_free_page(GFP_TEMPORARY);
28a6d671
EB		 2467 if (!page)
2468 goto out;
2469
2470 length = -EFAULT;
2471 if (copy_from_user(page, buf, count))
2472 goto out_free;
2473
107db7c7 2474 /* Guard against adverse ptrace interaction */
9b1bf12d 2475 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
107db7c7
DH		 2476 if (length < 0)
2477 goto out_free;
2478
28a6d671 2479 length = security_setprocattr(task,
2fddfeef 2480 (char*)file->f_path.dentry->d_name.name,
28a6d671 2481 (void*)page, count);
9b1bf12d 2482 mutex_unlock(&task->signal->cred_guard_mutex);
28a6d671
EB		 2483out_free:
2484 free_page((unsigned long) page);
2485out:
2486 put_task_struct(task);
2487out_no_task:
2488 return length;
2489}
2490
00977a59 2491static const struct file_operations proc_pid_attr_operations = {
28a6d671
EB		 2492 .read = proc_pid_attr_read,
2493 .write = proc_pid_attr_write,
87df8424 2494 .llseek = generic_file_llseek,
28a6d671
EB		 2495};
2496
c5141e6d 2497static const struct pid_entry attr_dir_stuff[] = {
631f9c18
AD		 2498 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2499 REG("prev", S_IRUGO, proc_pid_attr_operations),
2500 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2501 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2502 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2503 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
28a6d671
EB		 2504};
2505
72d9dcfc 2506static int proc_attr_dir_readdir(struct file * filp,
28a6d671
EB		 2507 void * dirent, filldir_t filldir)
2508{
2509 return proc_pident_readdir(filp,dirent,filldir,
72d9dcfc 2510 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2511}
2512
00977a59 2513static const struct file_operations proc_attr_dir_operations = {
1da177e4 2514 .read = generic_read_dir,
72d9dcfc 2515 .readdir = proc_attr_dir_readdir,
6038f373 2516 .llseek = default_llseek,
1da177e4
LT		 2517};
2518
72d9dcfc 2519static struct dentry *proc_attr_dir_lookup(struct inode *dir,
28a6d671
EB		 2520 struct dentry *dentry, struct nameidata *nd)
2521{
7bcd6b0e
EB		 2522 return proc_pident_lookup(dir, dentry,
2523 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2524}
2525
c5ef1c42 2526static const struct inode_operations proc_attr_dir_inode_operations = {
72d9dcfc 2527 .lookup = proc_attr_dir_lookup,
99f89551 2528 .getattr = pid_getattr,
6d76fa58 2529 .setattr = proc_setattr,
1da177e4
LT		 2530};
2531
28a6d671
EB		 2532#endif
2533
698ba7b5 2534#ifdef CONFIG_ELF_CORE
3cb4a0bb
KH		 2535static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2536 size_t count, loff_t *ppos)
2537{
2538 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2539 struct mm_struct *mm;
2540 char buffer[PROC_NUMBUF];
2541 size_t len;
2542 int ret;
2543
2544 if (!task)
2545 return -ESRCH;
2546
2547 ret = 0;
2548 mm = get_task_mm(task);
2549 if (mm) {
2550 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2551 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2552 MMF_DUMP_FILTER_SHIFT));
2553 mmput(mm);
2554 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2555 }
2556
2557 put_task_struct(task);
2558
2559 return ret;
2560}
2561
2562static ssize_t proc_coredump_filter_write(struct file *file,
2563 const char __user *buf,
2564 size_t count,
2565 loff_t *ppos)
2566{
2567 struct task_struct *task;
2568 struct mm_struct *mm;
2569 char buffer[PROC_NUMBUF], *end;
2570 unsigned int val;
2571 int ret;
2572 int i;
2573 unsigned long mask;
2574
2575 ret = -EFAULT;
2576 memset(buffer, 0, sizeof(buffer));
2577 if (count > sizeof(buffer) - 1)
2578 count = sizeof(buffer) - 1;
2579 if (copy_from_user(buffer, buf, count))
2580 goto out_no_task;
2581
2582 ret = -EINVAL;
2583 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2584 if (*end == '\n')
2585 end++;
2586 if (end - buffer == 0)
2587 goto out_no_task;
2588
2589 ret = -ESRCH;
2590 task = get_proc_task(file->f_dentry->d_inode);
2591 if (!task)
2592 goto out_no_task;
2593
2594 ret = end - buffer;
2595 mm = get_task_mm(task);
2596 if (!mm)
2597 goto out_no_mm;
2598
2599 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2600 if (val & mask)
2601 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2602 else
2603 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2604 }
2605
2606 mmput(mm);
2607 out_no_mm:
2608 put_task_struct(task);
2609 out_no_task:
2610 return ret;
2611}
2612
2613static const struct file_operations proc_coredump_filter_operations = {
2614 .read = proc_coredump_filter_read,
2615 .write = proc_coredump_filter_write,
87df8424 2616 .llseek = generic_file_llseek,
3cb4a0bb
KH		 2617};
2618#endif
2619
28a6d671
EB		 2620/*
2621 * /proc/self:
2622 */
2623static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2624 int buflen)
2625{
488e5bc4 2626 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
b55fcb22 2627 pid_t tgid = task_tgid_nr_ns(current, ns);
28a6d671 2628 char tmp[PROC_NUMBUF];
b55fcb22 2629 if (!tgid)
488e5bc4 2630 return -ENOENT;
b55fcb22 2631 sprintf(tmp, "%d", tgid);
28a6d671
EB		 2632 return vfs_readlink(dentry,buffer,buflen,tmp);
2633}
2634
2635static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2636{
488e5bc4 2637 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
b55fcb22 2638 pid_t tgid = task_tgid_nr_ns(current, ns);
7fee4868
AV		 2639 char *name = ERR_PTR(-ENOENT);
2640 if (tgid) {
2641 name = __getname();
2642 if (!name)
2643 name = ERR_PTR(-ENOMEM);
2644 else
2645 sprintf(name, "%d", tgid);
2646 }
2647 nd_set_link(nd, name);
2648 return NULL;
2649}
2650
2651static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2652 void *cookie)
2653{
2654 char *s = nd_get_link(nd);
2655 if (!IS_ERR(s))
2656 __putname(s);
28a6d671
EB		 2657}
2658
c5ef1c42 2659static const struct inode_operations proc_self_inode_operations = {
28a6d671
EB		 2660 .readlink = proc_self_readlink,
2661 .follow_link = proc_self_follow_link,
7fee4868 2662 .put_link = proc_self_put_link,
28a6d671
EB		 2663};
2664
801199ce
EB		 2665/*
2666 * proc base
2667 *
2668 * These are the directory entries in the root directory of /proc
2669 * that properly belong to the /proc filesystem, as they describe
2670 * describe something that is process related.
2671 */
c5141e6d 2672static const struct pid_entry proc_base_stuff[] = {
61a28784 2673 NOD("self", S_IFLNK|S_IRWXUGO,
801199ce 2674 &proc_self_inode_operations, NULL, {}),
801199ce
EB		 2675};
2676
444ceed8 2677static struct dentry *proc_base_instantiate(struct inode *dir,
c5141e6d 2678 struct dentry *dentry, struct task_struct *task, const void *ptr)
801199ce 2679{
c5141e6d 2680 const struct pid_entry *p = ptr;
801199ce 2681 struct inode *inode;
801199ce 2682 struct proc_inode *ei;
73d36460 2683 struct dentry *error;
801199ce
EB		 2684
2685 /* Allocate the inode */
2686 error = ERR_PTR(-ENOMEM);
2687 inode = new_inode(dir->i_sb);
2688 if (!inode)
2689 goto out;
2690
2691 /* Initialize the inode */
2692 ei = PROC_I(inode);
85fe4025 2693 inode->i_ino = get_next_ino();
801199ce 2694 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
801199ce
EB		 2695
2696 /*
2697 * grab the reference to the task.
2698 */
1a657f78 2699 ei->pid = get_task_pid(task, PIDTYPE_PID);
801199ce
EB		 2700 if (!ei->pid)
2701 goto out_iput;
2702
801199ce
EB		 2703 inode->i_mode = p->mode;
2704 if (S_ISDIR(inode->i_mode))
2705 inode->i_nlink = 2;
2706 if (S_ISLNK(inode->i_mode))
2707 inode->i_size = 64;
2708 if (p->iop)
2709 inode->i_op = p->iop;
2710 if (p->fop)
2711 inode->i_fop = p->fop;
2712 ei->op = p->op;
801199ce
EB		 2713 d_add(dentry, inode);
2714 error = NULL;
2715out:
801199ce
EB		 2716 return error;
2717out_iput:
2718 iput(inode);
2719 goto out;
2720}
2721
444ceed8
EB		 2722static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2723{
2724 struct dentry *error;
2725 struct task_struct *task = get_proc_task(dir);
c5141e6d 2726 const struct pid_entry *p, *last;
444ceed8
EB		 2727
2728 error = ERR_PTR(-ENOENT);
2729
2730 if (!task)
2731 goto out_no_task;
2732
2733 /* Lookup the directory entry */
7bcd6b0e
EB		 2734 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2735 for (p = proc_base_stuff; p <= last; p++) {
444ceed8
EB		 2736 if (p->len != dentry->d_name.len)
2737 continue;
2738 if (!memcmp(dentry->d_name.name, p->name, p->len))
2739 break;
2740 }
7bcd6b0e 2741 if (p > last)
444ceed8
EB		 2742 goto out;
2743
2744 error = proc_base_instantiate(dir, dentry, task, p);
2745
2746out:
2747 put_task_struct(task);
2748out_no_task:
2749 return error;
2750}
2751
c5141e6d
ED		 2752static int proc_base_fill_cache(struct file *filp, void *dirent,
2753 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
61a28784
EB		 2754{
2755 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2756 proc_base_instantiate, task, p);
2757}
2758
aba76fdb 2759#ifdef CONFIG_TASK_IO_ACCOUNTING
297c5d92
AR		 2760static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2761{
940389b8 2762 struct task_io_accounting acct = task->ioac;
5995477a
AR		 2763 unsigned long flags;
2764
2765 if (whole && lock_task_sighand(task, &flags)) {
2766 struct task_struct *t = task;
2767
2768 task_io_accounting_add(&acct, &task->signal->ioac);
2769 while_each_thread(task, t)
2770 task_io_accounting_add(&acct, &t->ioac);
2771
2772 unlock_task_sighand(task, &flags);
297c5d92 2773 }
aba76fdb
AM		 2774 return sprintf(buffer,
2775 "rchar: %llu\n"
2776 "wchar: %llu\n"
2777 "syscr: %llu\n"
2778 "syscw: %llu\n"
2779 "read_bytes: %llu\n"
2780 "write_bytes: %llu\n"
2781 "cancelled_write_bytes: %llu\n",
7c44319d
AB		 2782 (unsigned long long)acct.rchar,
2783 (unsigned long long)acct.wchar,
2784 (unsigned long long)acct.syscr,
2785 (unsigned long long)acct.syscw,
2786 (unsigned long long)acct.read_bytes,
2787 (unsigned long long)acct.write_bytes,
2788 (unsigned long long)acct.cancelled_write_bytes);
297c5d92
AR		 2789}
2790
2791static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2792{
2793 return do_io_accounting(task, buffer, 0);
aba76fdb 2794}
297c5d92
AR		 2795
2796static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2797{
2798 return do_io_accounting(task, buffer, 1);
2799}
2800#endif /* CONFIG_TASK_IO_ACCOUNTING */
aba76fdb 2801
47830723
KC		 2802static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2803 struct pid *pid, struct task_struct *task)
2804{
a9712bc1
AV		 2805 int err = lock_trace(task);
2806 if (!err) {
2807 seq_printf(m, "%08x\n", task->personality);
2808 unlock_trace(task);
2809 }
2810 return err;
47830723
KC		 2811}
2812
28a6d671
EB		 2813/*
2814 * Thread groups
2815 */
00977a59 2816static const struct file_operations proc_task_operations;
c5ef1c42 2817static const struct inode_operations proc_task_inode_operations;
20cdc894 2818
c5141e6d 2819static const struct pid_entry tgid_base_stuff[] = {
631f9c18
AD		 2820 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2821 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2822 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
b2211a36 2823#ifdef CONFIG_NET
631f9c18 2824 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
b2211a36 2825#endif
631f9c18
AD		 2826 REG("environ", S_IRUSR, proc_environ_operations),
2827 INF("auxv", S_IRUSR, proc_pid_auxv),
2828 ONE("status", S_IRUGO, proc_pid_status),
a9712bc1 2829 ONE("personality", S_IRUGO, proc_pid_personality),
3036e7b4 2830 INF("limits", S_IRUGO, proc_pid_limits),
43ae34cb 2831#ifdef CONFIG_SCHED_DEBUG
631f9c18 2832 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
5091faa4
MG		 2833#endif
2834#ifdef CONFIG_SCHED_AUTOGROUP
2835 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
ebcb6734 2836#endif
4614a696 2837 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 2838#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
a9712bc1 2839 INF("syscall", S_IRUGO, proc_pid_syscall),
43ae34cb 2840#endif
631f9c18
AD		 2841 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2842 ONE("stat", S_IRUGO, proc_tgid_stat),
2843 ONE("statm", S_IRUGO, proc_pid_statm),
2844 REG("maps", S_IRUGO, proc_maps_operations),
28a6d671 2845#ifdef CONFIG_NUMA
631f9c18 2846 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
28a6d671 2847#endif
631f9c18
AD		 2848 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2849 LNK("cwd", proc_cwd_link),
2850 LNK("root", proc_root_link),
2851 LNK("exe", proc_exe_link),
2852 REG("mounts", S_IRUGO, proc_mounts_operations),
2853 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2854 REG("mountstats", S_IRUSR, proc_mountstats_operations),
1e883281 2855#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18
AD		 2856 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2857 REG("smaps", S_IRUGO, proc_smaps_operations),
ca6b0bf0 2858 REG("pagemap", S_IRUGO, proc_pagemap_operations),
28a6d671
EB		 2859#endif
2860#ifdef CONFIG_SECURITY
631f9c18 2861 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 2862#endif
2863#ifdef CONFIG_KALLSYMS
631f9c18 2864 INF("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 2865#endif
2ec220e2 2866#ifdef CONFIG_STACKTRACE
a9712bc1 2867 ONE("stack", S_IRUGO, proc_pid_stack),
28a6d671
EB		 2868#endif
2869#ifdef CONFIG_SCHEDSTATS
631f9c18 2870 INF("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 2871#endif
9745512c 2872#ifdef CONFIG_LATENCYTOP
631f9c18 2873 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 2874#endif
8793d854 2875#ifdef CONFIG_PROC_PID_CPUSET
631f9c18 2876 REG("cpuset", S_IRUGO, proc_cpuset_operations),
a424316c
PM		 2877#endif
2878#ifdef CONFIG_CGROUPS
631f9c18 2879 REG("cgroup", S_IRUGO, proc_cgroup_operations),
28a6d671 2880#endif
631f9c18
AD		 2881 INF("oom_score", S_IRUGO, proc_oom_score),
2882 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
a63d83f4 2883 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 2884#ifdef CONFIG_AUDITSYSCALL
631f9c18
AD		 2885 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2886 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 2887#endif
f4f154fd 2888#ifdef CONFIG_FAULT_INJECTION
631f9c18 2889 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 2890#endif
698ba7b5 2891#ifdef CONFIG_ELF_CORE
631f9c18 2892 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3cb4a0bb 2893#endif
aba76fdb 2894#ifdef CONFIG_TASK_IO_ACCOUNTING
631f9c18 2895 INF("io", S_IRUGO, proc_tgid_io_accounting),
aba76fdb 2896#endif
28a6d671 2897};
1da177e4 2898
28a6d671 2899static int proc_tgid_base_readdir(struct file * filp,
1da177e4
LT		 2900 void * dirent, filldir_t filldir)
2901{
2902 return proc_pident_readdir(filp,dirent,filldir,
28a6d671 2903 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 2904}
2905
00977a59 2906static const struct file_operations proc_tgid_base_operations = {
1da177e4 2907 .read = generic_read_dir,
28a6d671 2908 .readdir = proc_tgid_base_readdir,
6038f373 2909 .llseek = default_llseek,
1da177e4
LT		 2910};
2911
28a6d671 2912static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
7bcd6b0e
EB		 2913 return proc_pident_lookup(dir, dentry,
2914 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 2915}
2916
c5ef1c42 2917static const struct inode_operations proc_tgid_base_inode_operations = {
28a6d671 2918 .lookup = proc_tgid_base_lookup,
99f89551 2919 .getattr = pid_getattr,
6d76fa58 2920 .setattr = proc_setattr,
1da177e4 2921};
1da177e4 2922
60347f67 2923static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
1da177e4 2924{
48e6484d 2925 struct dentry *dentry, *leader, *dir;
8578cea7 2926 char buf[PROC_NUMBUF];
48e6484d
EB		 2927 struct qstr name;
2928
2929 name.name = buf;
60347f67
PE		 2930 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2931 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d 2932 if (dentry) {
29f12ca3 2933 shrink_dcache_parent(dentry);
48e6484d
EB		 2934 d_drop(dentry);
2935 dput(dentry);
2936 }
1da177e4 2937
48e6484d 2938 name.name = buf;
60347f67
PE		 2939 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2940 leader = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d
EB		 2941 if (!leader)
2942 goto out;
1da177e4 2943
48e6484d
EB		 2944 name.name = "task";
2945 name.len = strlen(name.name);
2946 dir = d_hash_and_lookup(leader, &name);
2947 if (!dir)
2948 goto out_put_leader;
2949
2950 name.name = buf;
60347f67 2951 name.len = snprintf(buf, sizeof(buf), "%d", pid);
48e6484d
EB		 2952 dentry = d_hash_and_lookup(dir, &name);
2953 if (dentry) {
2954 shrink_dcache_parent(dentry);
2955 d_drop(dentry);
2956 dput(dentry);
1da177e4 2957 }
48e6484d
EB		 2958
2959 dput(dir);
2960out_put_leader:
2961 dput(leader);
2962out:
2963 return;
1da177e4
LT		 2964}
2965
0895e91d
RD		 2966/**
2967 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2968 * @task: task that should be flushed.
2969 *
2970 * When flushing dentries from proc, one needs to flush them from global
60347f67 2971 * proc (proc_mnt) and from all the namespaces' procs this task was seen
0895e91d
RD		 2972 * in. This call is supposed to do all of this job.
2973 *
2974 * Looks in the dcache for
2975 * /proc/@pid
2976 * /proc/@tgid/task/@pid
2977 * if either directory is present flushes it and all of it'ts children
2978 * from the dcache.
2979 *
2980 * It is safe and reasonable to cache /proc entries for a task until
2981 * that task exits. After that they just clog up the dcache with
2982 * useless entries, possibly causing useful dcache entries to be
2983 * flushed instead. This routine is proved to flush those useless
2984 * dcache entries at process exit time.
2985 *
2986 * NOTE: This routine is just an optimization so it does not guarantee
2987 * that no dcache entries will exist at process exit time it
2988 * just makes it very unlikely that any will persist.
60347f67
PE		 2989 */
2990
2991void proc_flush_task(struct task_struct *task)
2992{
9fcc2d15 2993 int i;
9b4d1cbe 2994 struct pid *pid, *tgid;
130f77ec
PE		 2995 struct upid *upid;
2996
130f77ec 2997 pid = task_pid(task);
9b4d1cbe 2998 tgid = task_tgid(task);
130f77ec 2999
9fcc2d15 3000 for (i = 0; i <= pid->level; i++) {
130f77ec
PE		 3001 upid = &pid->numbers[i];
3002 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
9b4d1cbe 3003 tgid->numbers[i].nr);
130f77ec 3004 }
6f4e6433
PE		 3005
3006 upid = &pid->numbers[pid->level];
3007 if (upid->nr == 1)
3008 pid_ns_release_proc(upid->ns);
60347f67
PE		 3009}
3010
9711ef99
AB		 3011static struct dentry *proc_pid_instantiate(struct inode *dir,
3012 struct dentry * dentry,
c5141e6d 3013 struct task_struct *task, const void *ptr)
444ceed8
EB		 3014{
3015 struct dentry *error = ERR_PTR(-ENOENT);
3016 struct inode *inode;
3017
61a28784 3018 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 3019 if (!inode)
3020 goto out;
3021
3022 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3023 inode->i_op = &proc_tgid_base_inode_operations;
3024 inode->i_fop = &proc_tgid_base_operations;
3025 inode->i_flags|=S_IMMUTABLE;
aed54175
VN		 3026
3027 inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
3028 ARRAY_SIZE(tgid_base_stuff));
444ceed8 3029
fb045adb 3030 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 3031
3032 d_add(dentry, inode);
3033 /* Close the race of the process dying before we return the dentry */
3034 if (pid_revalidate(dentry, NULL))
3035 error = NULL;
3036out:
3037 return error;
3038}
3039
1da177e4
LT		 3040struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3041{
73d36460 3042 struct dentry *result;
1da177e4 3043 struct task_struct *task;
1da177e4 3044 unsigned tgid;
b488893a 3045 struct pid_namespace *ns;
1da177e4 3046
801199ce
EB		 3047 result = proc_base_lookup(dir, dentry);
3048 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3049 goto out;
3050
1da177e4
LT		 3051 tgid = name_to_int(dentry);
3052 if (tgid == ~0U)
3053 goto out;
3054
b488893a 3055 ns = dentry->d_sb->s_fs_info;
de758734 3056 rcu_read_lock();
b488893a 3057 task = find_task_by_pid_ns(tgid, ns);
1da177e4
LT		 3058 if (task)
3059 get_task_struct(task);
de758734 3060 rcu_read_unlock();
1da177e4
LT		 3061 if (!task)
3062 goto out;
3063
444ceed8 3064 result = proc_pid_instantiate(dir, dentry, task, NULL);
1da177e4 3065 put_task_struct(task);
1da177e4 3066out:
cd6a3ce9 3067 return result;
1da177e4
LT		 3068}
3069
1da177e4 3070/*
0804ef4b 3071 * Find the first task with tgid >= tgid
0bc58a91 3072 *
1da177e4 3073 */
19fd4bb2
EB		 3074struct tgid_iter {
3075 unsigned int tgid;
0804ef4b 3076 struct task_struct *task;
19fd4bb2
EB		 3077};
3078static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3079{
0804ef4b 3080 struct pid *pid;
1da177e4 3081
19fd4bb2
EB		 3082 if (iter.task)
3083 put_task_struct(iter.task);
454cc105 3084 rcu_read_lock();
0804ef4b 3085retry:
19fd4bb2
EB		 3086 iter.task = NULL;
3087 pid = find_ge_pid(iter.tgid, ns);
0804ef4b 3088 if (pid) {
19fd4bb2
EB		 3089 iter.tgid = pid_nr_ns(pid, ns);
3090 iter.task = pid_task(pid, PIDTYPE_PID);
0804ef4b
EB		 3091 /* What we to know is if the pid we have find is the
3092 * pid of a thread_group_leader. Testing for task
3093 * being a thread_group_leader is the obvious thing
3094 * todo but there is a window when it fails, due to
3095 * the pid transfer logic in de_thread.
3096 *
3097 * So we perform the straight forward test of seeing
3098 * if the pid we have found is the pid of a thread
3099 * group leader, and don't worry if the task we have
3100 * found doesn't happen to be a thread group leader.
3101 * As we don't care in the case of readdir.
3102 */
19fd4bb2
EB		 3103 if (!iter.task || !has_group_leader_pid(iter.task)) {
3104 iter.tgid += 1;
0804ef4b 3105 goto retry;
19fd4bb2
EB		 3106 }
3107 get_task_struct(iter.task);
0bc58a91 3108 }
454cc105 3109 rcu_read_unlock();
19fd4bb2 3110 return iter;
1da177e4
LT		 3111}
3112
7bcd6b0e 3113#define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
0804ef4b 3114
61a28784 3115static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
19fd4bb2 3116 struct tgid_iter iter)
61a28784
EB		 3117{
3118 char name[PROC_NUMBUF];
19fd4bb2 3119 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
61a28784 3120 return proc_fill_cache(filp, dirent, filldir, name, len,
19fd4bb2 3121 proc_pid_instantiate, iter.task, NULL);
61a28784
EB		 3122}
3123
1da177e4
LT		 3124/* for the /proc/ directory itself, after non-process stuff has been done */
3125int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3126{
d8bdc59f
LT		 3127 unsigned int nr;
3128 struct task_struct *reaper;
19fd4bb2 3129 struct tgid_iter iter;
b488893a 3130 struct pid_namespace *ns;
1da177e4 3131
d8bdc59f
LT		 3132 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3133 goto out_no_task;
3134 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3135
3136 reaper = get_proc_task(filp->f_path.dentry->d_inode);
61a28784
EB		 3137 if (!reaper)
3138 goto out_no_task;
3139
7bcd6b0e 3140 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
c5141e6d 3141 const struct pid_entry *p = &proc_base_stuff[nr];
61a28784 3142 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
801199ce 3143 goto out;
1da177e4
LT		 3144 }
3145
b488893a 3146 ns = filp->f_dentry->d_sb->s_fs_info;
19fd4bb2
EB		 3147 iter.task = NULL;
3148 iter.tgid = filp->f_pos - TGID_OFFSET;
3149 for (iter = next_tgid(ns, iter);
3150 iter.task;
3151 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3152 filp->f_pos = iter.tgid + TGID_OFFSET;
3153 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
3154 put_task_struct(iter.task);
0804ef4b 3155 goto out;
1da177e4 3156 }
0bc58a91 3157 }
0804ef4b
EB		 3158 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3159out:
61a28784
EB		 3160 put_task_struct(reaper);
3161out_no_task:
0bc58a91
EB		 3162 return 0;
3163}
1da177e4 3164
28a6d671
EB		 3165/*
3166 * Tasks
3167 */
c5141e6d 3168static const struct pid_entry tid_base_stuff[] = {
631f9c18 3169 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3835541d 3170 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
631f9c18
AD		 3171 REG("environ", S_IRUSR, proc_environ_operations),
3172 INF("auxv", S_IRUSR, proc_pid_auxv),
3173 ONE("status", S_IRUGO, proc_pid_status),
a9712bc1 3174 ONE("personality", S_IRUGO, proc_pid_personality),
3036e7b4 3175 INF("limits", S_IRUGO, proc_pid_limits),
43ae34cb 3176#ifdef CONFIG_SCHED_DEBUG
631f9c18 3177 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
ebcb6734 3178#endif
4614a696 3179 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 3180#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
a9712bc1 3181 INF("syscall", S_IRUGO, proc_pid_syscall),
43ae34cb 3182#endif
631f9c18
AD		 3183 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3184 ONE("stat", S_IRUGO, proc_tid_stat),
3185 ONE("statm", S_IRUGO, proc_pid_statm),
3186 REG("maps", S_IRUGO, proc_maps_operations),
28a6d671 3187#ifdef CONFIG_NUMA
631f9c18 3188 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
28a6d671 3189#endif
631f9c18
AD		 3190 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3191 LNK("cwd", proc_cwd_link),
3192 LNK("root", proc_root_link),
3193 LNK("exe", proc_exe_link),
3194 REG("mounts", S_IRUGO, proc_mounts_operations),
3195 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
1e883281 3196#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18
AD		 3197 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3198 REG("smaps", S_IRUGO, proc_smaps_operations),
ca6b0bf0 3199 REG("pagemap", S_IRUGO, proc_pagemap_operations),
28a6d671
EB		 3200#endif
3201#ifdef CONFIG_SECURITY
631f9c18 3202 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 3203#endif
3204#ifdef CONFIG_KALLSYMS
631f9c18 3205 INF("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 3206#endif
2ec220e2 3207#ifdef CONFIG_STACKTRACE
a9712bc1 3208 ONE("stack", S_IRUGO, proc_pid_stack),
28a6d671
EB		 3209#endif
3210#ifdef CONFIG_SCHEDSTATS
631f9c18 3211 INF("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 3212#endif
9745512c 3213#ifdef CONFIG_LATENCYTOP
631f9c18 3214 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 3215#endif
8793d854 3216#ifdef CONFIG_PROC_PID_CPUSET
631f9c18 3217 REG("cpuset", S_IRUGO, proc_cpuset_operations),
a424316c
PM		 3218#endif
3219#ifdef CONFIG_CGROUPS
631f9c18 3220 REG("cgroup", S_IRUGO, proc_cgroup_operations),
28a6d671 3221#endif
631f9c18
AD		 3222 INF("oom_score", S_IRUGO, proc_oom_score),
3223 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
a63d83f4 3224 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 3225#ifdef CONFIG_AUDITSYSCALL
631f9c18 3226 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
26ec3c64 3227 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 3228#endif
f4f154fd 3229#ifdef CONFIG_FAULT_INJECTION
631f9c18 3230 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 3231#endif
297c5d92 3232#ifdef CONFIG_TASK_IO_ACCOUNTING
631f9c18 3233 INF("io", S_IRUGO, proc_tid_io_accounting),
297c5d92 3234#endif
28a6d671
EB		 3235};
3236
3237static int proc_tid_base_readdir(struct file * filp,
3238 void * dirent, filldir_t filldir)
3239{
3240 return proc_pident_readdir(filp,dirent,filldir,
3241 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3242}
3243
3244static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
7bcd6b0e
EB		 3245 return proc_pident_lookup(dir, dentry,
3246 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
28a6d671
EB		 3247}
3248
00977a59 3249static const struct file_operations proc_tid_base_operations = {
28a6d671
EB		 3250 .read = generic_read_dir,
3251 .readdir = proc_tid_base_readdir,
6038f373 3252 .llseek = default_llseek,
28a6d671
EB		 3253};
3254
c5ef1c42 3255static const struct inode_operations proc_tid_base_inode_operations = {
28a6d671
EB		 3256 .lookup = proc_tid_base_lookup,
3257 .getattr = pid_getattr,
3258 .setattr = proc_setattr,
3259};
3260
444ceed8 3261static struct dentry *proc_task_instantiate(struct inode *dir,
c5141e6d 3262 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8
EB		 3263{
3264 struct dentry *error = ERR_PTR(-ENOENT);
3265 struct inode *inode;
61a28784 3266 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 3267
3268 if (!inode)
3269 goto out;
3270 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3271 inode->i_op = &proc_tid_base_inode_operations;
3272 inode->i_fop = &proc_tid_base_operations;
3273 inode->i_flags|=S_IMMUTABLE;
aed54175
VN		 3274
3275 inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
3276 ARRAY_SIZE(tid_base_stuff));
444ceed8 3277
fb045adb 3278 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 3279
3280 d_add(dentry, inode);
3281 /* Close the race of the process dying before we return the dentry */
3282 if (pid_revalidate(dentry, NULL))
3283 error = NULL;
3284out:
3285 return error;
3286}
3287
28a6d671
EB		 3288static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3289{
3290 struct dentry *result = ERR_PTR(-ENOENT);
3291 struct task_struct *task;
3292 struct task_struct *leader = get_proc_task(dir);
28a6d671 3293 unsigned tid;
b488893a 3294 struct pid_namespace *ns;
28a6d671
EB		 3295
3296 if (!leader)
3297 goto out_no_task;
3298
3299 tid = name_to_int(dentry);
3300 if (tid == ~0U)
3301 goto out;
3302
b488893a 3303 ns = dentry->d_sb->s_fs_info;
28a6d671 3304 rcu_read_lock();
b488893a 3305 task = find_task_by_pid_ns(tid, ns);
28a6d671
EB		 3306 if (task)
3307 get_task_struct(task);
3308 rcu_read_unlock();
3309 if (!task)
3310 goto out;
bac0abd6 3311 if (!same_thread_group(leader, task))
28a6d671
EB		 3312 goto out_drop_task;
3313
444ceed8 3314 result = proc_task_instantiate(dir, dentry, task, NULL);
28a6d671
EB		 3315out_drop_task:
3316 put_task_struct(task);
3317out:
3318 put_task_struct(leader);
3319out_no_task:
3320 return result;
3321}
3322
0bc58a91
EB		 3323/*
3324 * Find the first tid of a thread group to return to user space.
3325 *
3326 * Usually this is just the thread group leader, but if the users
3327 * buffer was too small or there was a seek into the middle of the
3328 * directory we have more work todo.
3329 *
3330 * In the case of a short read we start with find_task_by_pid.
3331 *
3332 * In the case of a seek we start with the leader and walk nr
3333 * threads past it.
3334 */
cc288738 3335static struct task_struct *first_tid(struct task_struct *leader,
b488893a 3336 int tid, int nr, struct pid_namespace *ns)
0bc58a91 3337{
a872ff0c 3338 struct task_struct *pos;
1da177e4 3339
cc288738 3340 rcu_read_lock();
0bc58a91
EB		 3341 /* Attempt to start with the pid of a thread */
3342 if (tid && (nr > 0)) {
b488893a 3343 pos = find_task_by_pid_ns(tid, ns);
a872ff0c
ON		 3344 if (pos && (pos->group_leader == leader))
3345 goto found;
0bc58a91 3346 }
1da177e4 3347
0bc58a91 3348 /* If nr exceeds the number of threads there is nothing todo */
a872ff0c
ON		 3349 pos = NULL;
3350 if (nr && nr >= get_nr_threads(leader))
3351 goto out;
1da177e4 3352
a872ff0c
ON		 3353 /* If we haven't found our starting place yet start
3354 * with the leader and walk nr threads forward.
0bc58a91 3355 */
a872ff0c
ON		 3356 for (pos = leader; nr > 0; --nr) {
3357 pos = next_thread(pos);
3358 if (pos == leader) {
3359 pos = NULL;
3360 goto out;
3361 }
1da177e4 3362 }
a872ff0c
ON		 3363found:
3364 get_task_struct(pos);
3365out:
cc288738 3366 rcu_read_unlock();
0bc58a91
EB		 3367 return pos;
3368}
3369
3370/*
3371 * Find the next thread in the thread list.
3372 * Return NULL if there is an error or no next thread.
3373 *
3374 * The reference to the input task_struct is released.
3375 */
3376static struct task_struct *next_tid(struct task_struct *start)
3377{
c1df7fb8 3378 struct task_struct *pos = NULL;
cc288738 3379 rcu_read_lock();
c1df7fb8 3380 if (pid_alive(start)) {
0bc58a91 3381 pos = next_thread(start);
c1df7fb8
ON		 3382 if (thread_group_leader(pos))
3383 pos = NULL;
3384 else
3385 get_task_struct(pos);
3386 }
cc288738 3387 rcu_read_unlock();
0bc58a91
EB		 3388 put_task_struct(start);
3389 return pos;
1da177e4
LT		 3390}
3391
61a28784
EB		 3392static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3393 struct task_struct *task, int tid)
3394{
3395 char name[PROC_NUMBUF];
3396 int len = snprintf(name, sizeof(name), "%d", tid);
3397 return proc_fill_cache(filp, dirent, filldir, name, len,
3398 proc_task_instantiate, task, NULL);
3399}
3400
1da177e4
LT		 3401/* for the /proc/TGID/task/ directories */
3402static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3403{
2fddfeef 3404 struct dentry *dentry = filp->f_path.dentry;
1da177e4 3405 struct inode *inode = dentry->d_inode;
7d895244 3406 struct task_struct *leader = NULL;
0bc58a91 3407 struct task_struct *task;
1da177e4
LT		 3408 int retval = -ENOENT;
3409 ino_t ino;
0bc58a91 3410 int tid;
b488893a 3411 struct pid_namespace *ns;
1da177e4 3412
7d895244
GC		 3413 task = get_proc_task(inode);
3414 if (!task)
3415 goto out_no_task;
3416 rcu_read_lock();
3417 if (pid_alive(task)) {
3418 leader = task->group_leader;
3419 get_task_struct(leader);
3420 }
3421 rcu_read_unlock();
3422 put_task_struct(task);
99f89551
EB		 3423 if (!leader)
3424 goto out_no_task;
1da177e4
LT		 3425 retval = 0;
3426
ee568b25 3427 switch ((unsigned long)filp->f_pos) {
1da177e4
LT		 3428 case 0:
3429 ino = inode->i_ino;
ee6f779b 3430 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
1da177e4 3431 goto out;
ee6f779b 3432 filp->f_pos++;
1da177e4
LT		 3433 /* fall through */
3434 case 1:
3435 ino = parent_ino(dentry);
ee6f779b 3436 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
1da177e4 3437 goto out;
ee6f779b 3438 filp->f_pos++;
1da177e4
LT		 3439 /* fall through */
3440 }
3441
0bc58a91
EB		 3442 /* f_version caches the tgid value that the last readdir call couldn't
3443 * return. lseek aka telldir automagically resets f_version to 0.
3444 */
b488893a 3445 ns = filp->f_dentry->d_sb->s_fs_info;
2b47c361 3446 tid = (int)filp->f_version;
0bc58a91 3447 filp->f_version = 0;
ee6f779b 3448 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
0bc58a91 3449 task;
ee6f779b 3450 task = next_tid(task), filp->f_pos++) {
b488893a 3451 tid = task_pid_nr_ns(task, ns);
61a28784 3452 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
0bc58a91
EB		 3453 /* returning this tgid failed, save it as the first
3454 * pid for the next readir call */
2b47c361 3455 filp->f_version = (u64)tid;
0bc58a91 3456 put_task_struct(task);
1da177e4 3457 break;
0bc58a91 3458 }
1da177e4
LT		 3459 }
3460out:
99f89551
EB		 3461 put_task_struct(leader);
3462out_no_task:
1da177e4
LT		 3463 return retval;
3464}
6e66b52b
EB		 3465
3466static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3467{
3468 struct inode *inode = dentry->d_inode;
99f89551 3469 struct task_struct *p = get_proc_task(inode);
6e66b52b
EB		 3470 generic_fillattr(inode, stat);
3471
99f89551 3472 if (p) {
99f89551 3473 stat->nlink += get_nr_threads(p);
99f89551 3474 put_task_struct(p);
6e66b52b
EB		 3475 }
3476
3477 return 0;
3478}
28a6d671 3479
c5ef1c42 3480static const struct inode_operations proc_task_inode_operations = {
28a6d671
EB		 3481 .lookup = proc_task_lookup,
3482 .getattr = proc_task_getattr,
3483 .setattr = proc_setattr,
3484};
3485
00977a59 3486static const struct file_operations proc_task_operations = {
28a6d671
EB		 3487 .read = generic_read_dir,
3488 .readdir = proc_task_readdir,
6038f373 3489 .llseek = default_llseek,
28a6d671 3490};
Linux kernel - Deliverables
This page took 0.825335 seconds and 5 git commands to generate.