projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
media: staging: media: use relevant lock
[linux-2.6-block.git] / fs / proc / base.c
CommitLineData
b2441318 1// SPDX-License-Identifier: GPL-2.0
1da177e4
LT		 2/*
3 * linux/fs/proc/base.c
4 *
5 * Copyright (C) 1991, 1992 Linus Torvalds
6 *
7 * proc base directory handling functions
8 *
9 * 1999, Al Viro. Rewritten. Now it covers the whole per-process part.
10 * Instead of using magical inumbers to determine the kind of object
11 * we allocate and fill in-core inodes upon lookup. They don't even
12 * go into icache. We cache the reference to task_struct upon lookup too.
13 * Eventually it should become a filesystem in its own. We don't use the
14 * rest of procfs anymore.
e070ad49
ML		 15 *
16 *
17 * Changelog:
18 * 17-Jan-2005
19 * Allan Bezerra
20 * Bruna Moreira <bruna.moreira@indt.org.br>
21 * Edjard Mota <edjard.mota@indt.org.br>
22 * Ilias Biris <ilias.biris@indt.org.br>
23 * Mauricio Lin <mauricio.lin@indt.org.br>
24 *
25 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
26 *
27 * A new process specific entry (smaps) included in /proc. It shows the
28 * size of rss for each memory area. The maps entry lacks information
29 * about physical memory size (rss) for each mapped file, i.e.,
30 * rss information for executables and library files.
31 * This additional information is useful for any tools that need to know
32 * about physical memory consumption for a process specific library.
33 *
34 * Changelog:
35 * 21-Feb-2005
36 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
37 * Pud inclusion in the page table walking.
38 *
39 * ChangeLog:
40 * 10-Mar-2005
41 * 10LE Instituto Nokia de Tecnologia - INdT:
42 * A better way to walks through the page table as suggested by Hugh Dickins.
43 *
44 * Simo Piiroinen <simo.piiroinen@nokia.com>:
45 * Smaps information related to shared, private, clean and dirty pages.
46 *
47 * Paul Mundt <paul.mundt@nokia.com>:
48 * Overall revision about smaps.
1da177e4
LT		 49 */
50
7c0f6ba6 51#include <linux/uaccess.h>
1da177e4 52
1da177e4
LT		 53#include <linux/errno.h>
54#include <linux/time.h>
55#include <linux/proc_fs.h>
56#include <linux/stat.h>
5995477a 57#include <linux/task_io_accounting_ops.h>
1da177e4 58#include <linux/init.h>
16f7e0fe 59#include <linux/capability.h>
1da177e4 60#include <linux/file.h>
9f3acc31 61#include <linux/fdtable.h>
1da177e4
LT		 62#include <linux/string.h>
63#include <linux/seq_file.h>
64#include <linux/namei.h>
6b3286ed 65#include <linux/mnt_namespace.h>
1da177e4 66#include <linux/mm.h>
a63d83f4 67#include <linux/swap.h>
b835996f 68#include <linux/rcupdate.h>
1da177e4 69#include <linux/kallsyms.h>
2ec220e2 70#include <linux/stacktrace.h>
d85f50d5 71#include <linux/resource.h>
5096add8 72#include <linux/module.h>
1da177e4
LT		 73#include <linux/mount.h>
74#include <linux/security.h>
75#include <linux/ptrace.h>
0d094efe 76#include <linux/tracehook.h>
87ebdc00 77#include <linux/printk.h>
efb1a57d 78#include <linux/cache.h>
a424316c 79#include <linux/cgroup.h>
1da177e4
LT		 80#include <linux/cpuset.h>
81#include <linux/audit.h>
5addc5dd 82#include <linux/poll.h>
1651e14e 83#include <linux/nsproxy.h>
8ac773b4 84#include <linux/oom.h>
3cb4a0bb 85#include <linux/elf.h>
60347f67 86#include <linux/pid_namespace.h>
22d917d8 87#include <linux/user_namespace.h>
5ad4e53b 88#include <linux/fs_struct.h>
5a0e3ad6 89#include <linux/slab.h>
4eb5aaa3 90#include <linux/sched/autogroup.h>
6e84f315 91#include <linux/sched/mm.h>
f7ccbae4 92#include <linux/sched/coredump.h>
b17b0153 93#include <linux/sched/debug.h>
3905f9ad 94#include <linux/sched/stat.h>
640708a2 95#include <linux/flex_array.h>
48f6a7a5 96#include <linux/posix-timers.h>
43d2b113 97#include <trace/events/oom.h>
1da177e4 98#include "internal.h"
faf60af1 99#include "fd.h"
1da177e4 100
ac7f1061
AD		 101#include "../../lib/kstrtox.h"
102
0f2fe20f
EB		 103/* NOTE:
104 * Implementing inode permission operations in /proc is almost
105 * certainly an error. Permission checks need to happen during
106 * each system call not at open time. The reason is that most of
107 * what we wish to check for permissions in /proc varies at runtime.
108 *
109 * The classic example of a problem is opening file descriptors
110 * in /proc for a task before it execs a suid executable.
111 */
112
efb1a57d
AD		 113static u8 nlink_tid __ro_after_init;
114static u8 nlink_tgid __ro_after_init;
1270dd8d 115
1da177e4 116struct pid_entry {
cedbccab 117 const char *name;
623f594e 118 unsigned int len;
d161a13f 119 umode_t mode;
c5ef1c42 120 const struct inode_operations *iop;
00977a59 121 const struct file_operations *fop;
20cdc894 122 union proc_op op;
1da177e4
LT		 123};
124
61a28784 125#define NOD(NAME, MODE, IOP, FOP, OP) { \
20cdc894 126 .name = (NAME), \
c5141e6d 127 .len = sizeof(NAME) - 1, \
20cdc894
EB		 128 .mode = MODE, \
129 .iop = IOP, \
130 .fop = FOP, \
131 .op = OP, \
132}
133
631f9c18
AD		 134#define DIR(NAME, MODE, iops, fops) \
135 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
136#define LNK(NAME, get_link) \
61a28784 137 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
20cdc894 138 &proc_pid_link_inode_operations, NULL, \
631f9c18
AD		 139 { .proc_get_link = get_link } )
140#define REG(NAME, MODE, fops) \
141 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
631f9c18 142#define ONE(NAME, MODE, show) \
be614086
EB		 143 NOD(NAME, (S_IFREG|(MODE)), \
144 NULL, &proc_single_file_operations, \
631f9c18 145 { .proc_show = show } )
1da177e4 146
aed54175
VN		 147/*
148 * Count the number of hardlinks for the pid_entry table, excluding the .
149 * and .. links.
150 */
1270dd8d 151static unsigned int __init pid_entry_nlink(const struct pid_entry *entries,
aed54175
VN		 152 unsigned int n)
153{
154 unsigned int i;
155 unsigned int count;
156
1270dd8d 157 count = 2;
aed54175
VN		 158 for (i = 0; i < n; ++i) {
159 if (S_ISDIR(entries[i].mode))
160 ++count;
161 }
162
163 return count;
164}
165
f7ad3c6b 166static int get_task_root(struct task_struct *task, struct path *root)
1da177e4 167{
7c2c7d99
HD		 168 int result = -ENOENT;
169
0494f6ec 170 task_lock(task);
f7ad3c6b
MS		 171 if (task->fs) {
172 get_fs_root(task->fs, root);
7c2c7d99
HD		 173 result = 0;
174 }
0494f6ec 175 task_unlock(task);
7c2c7d99 176 return result;
0494f6ec
MS		 177}
178
7773fbc5 179static int proc_cwd_link(struct dentry *dentry, struct path *path)
0494f6ec 180{
2b0143b5 181 struct task_struct *task = get_proc_task(d_inode(dentry));
0494f6ec 182 int result = -ENOENT;
99f89551
EB		 183
184 if (task) {
f7ad3c6b
MS		 185 task_lock(task);
186 if (task->fs) {
187 get_fs_pwd(task->fs, path);
188 result = 0;
189 }
190 task_unlock(task);
99f89551
EB		 191 put_task_struct(task);
192 }
1da177e4
LT		 193 return result;
194}
195
7773fbc5 196static int proc_root_link(struct dentry *dentry, struct path *path)
1da177e4 197{
2b0143b5 198 struct task_struct *task = get_proc_task(d_inode(dentry));
1da177e4 199 int result = -ENOENT;
99f89551
EB		 200
201 if (task) {
f7ad3c6b 202 result = get_task_root(task, path);
99f89551
EB		 203 put_task_struct(task);
204 }
1da177e4
LT		 205 return result;
206}
207
c2c0bb44
AD		 208static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,
209 size_t _count, loff_t *pos)
1da177e4 210{
c2c0bb44
AD		 211 struct task_struct *tsk;
212 struct mm_struct *mm;
213 char *page;
214 unsigned long count = _count;
215 unsigned long arg_start, arg_end, env_start, env_end;
216 unsigned long len1, len2, len;
217 unsigned long p;
218 char c;
219 ssize_t rv;
220
221 BUG_ON(*pos < 0);
222
223 tsk = get_proc_task(file_inode(file));
224 if (!tsk)
225 return -ESRCH;
226 mm = get_task_mm(tsk);
227 put_task_struct(tsk);
228 if (!mm)
229 return 0;
230 /* Check if process spawned far enough to have cmdline. */
231 if (!mm->env_end) {
232 rv = 0;
233 goto out_mmput;
234 }
235
0ee931c4 236 page = (char *)__get_free_page(GFP_KERNEL);
c2c0bb44
AD		 237 if (!page) {
238 rv = -ENOMEM;
239 goto out_mmput;
240 }
241
242 down_read(&mm->mmap_sem);
243 arg_start = mm->arg_start;
244 arg_end = mm->arg_end;
245 env_start = mm->env_start;
246 env_end = mm->env_end;
247 up_read(&mm->mmap_sem);
248
249 BUG_ON(arg_start > arg_end);
250 BUG_ON(env_start > env_end);
251
252 len1 = arg_end - arg_start;
253 len2 = env_end - env_start;
254
3581d458
AD		 255 /* Empty ARGV. */
256 if (len1 == 0) {
257 rv = 0;
258 goto out_free_page;
259 }
2ca66ff7 260 /*
c2c0bb44
AD		 261 * Inherently racy -- command line shares address space
262 * with code and data.
2ca66ff7 263 */
272ddc8b 264 rv = access_remote_vm(mm, arg_end - 1, &c, 1, 0);
c2c0bb44
AD		 265 if (rv <= 0)
266 goto out_free_page;
267
268 rv = 0;
269
270 if (c == '\0') {
271 /* Command line (set of strings) occupies whole ARGV. */
272 if (len1 <= *pos)
273 goto out_free_page;
274
275 p = arg_start + *pos;
276 len = len1 - *pos;
277 while (count > 0 && len > 0) {
278 unsigned int _count;
279 int nr_read;
280
281 _count = min3(count, len, PAGE_SIZE);
272ddc8b 282 nr_read = access_remote_vm(mm, p, page, _count, 0);
c2c0bb44
AD		 283 if (nr_read < 0)
284 rv = nr_read;
285 if (nr_read <= 0)
286 goto out_free_page;
287
288 if (copy_to_user(buf, page, nr_read)) {
289 rv = -EFAULT;
290 goto out_free_page;
291 }
292
293 p += nr_read;
294 len -= nr_read;
295 buf += nr_read;
296 count -= nr_read;
297 rv += nr_read;
298 }
299 } else {
c2c0bb44
AD		 300 /*
301 * Command line (1 string) occupies ARGV and
302 * extends into ENVP.
303 */
a0a07b87
AD		 304 struct {
305 unsigned long p;
306 unsigned long len;
307 } cmdline[2] = {
308 { .p = arg_start, .len = len1 },
309 { .p = env_start, .len = len2 },
310 };
311 loff_t pos1 = *pos;
312 unsigned int i;
313
314 i = 0;
315 while (i < 2 && pos1 >= cmdline[i].len) {
316 pos1 -= cmdline[i].len;
317 i++;
c2c0bb44 318 }
a0a07b87
AD		 319 while (i < 2) {
320 p = cmdline[i].p + pos1;
321 len = cmdline[i].len - pos1;
322 while (count > 0 && len > 0) {
323 unsigned int _count, l;
324 int nr_read;
325 bool final;
326
327 _count = min3(count, len, PAGE_SIZE);
328 nr_read = access_remote_vm(mm, p, page, _count, 0);
329 if (nr_read < 0)
330 rv = nr_read;
331 if (nr_read <= 0)
332 goto out_free_page;
333
334 /*
335 * Command line can be shorter than whole ARGV
336 * even if last "marker" byte says it is not.
337 */
338 final = false;
339 l = strnlen(page, nr_read);
340 if (l < nr_read) {
341 nr_read = l;
342 final = true;
343 }
344
345 if (copy_to_user(buf, page, nr_read)) {
346 rv = -EFAULT;
347 goto out_free_page;
348 }
349
350 p += nr_read;
351 len -= nr_read;
352 buf += nr_read;
353 count -= nr_read;
354 rv += nr_read;
355
356 if (final)
357 goto out_free_page;
c2c0bb44
AD		 358 }
359
a0a07b87
AD		 360 /* Only first chunk can be read partially. */
361 pos1 = 0;
362 i++;
c2c0bb44 363 }
c2c0bb44
AD		 364 }
365
366out_free_page:
367 free_page((unsigned long)page);
368out_mmput:
369 mmput(mm);
370 if (rv > 0)
371 *pos += rv;
372 return rv;
1da177e4
LT		 373}
374
c2c0bb44
AD		 375static const struct file_operations proc_pid_cmdline_ops = {
376 .read = proc_pid_cmdline_read,
377 .llseek = generic_file_llseek,
378};
379
1da177e4
LT		 380#ifdef CONFIG_KALLSYMS
381/*
382 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
383 * Returns the resolved symbol. If that fails, simply return the address.
384 */
edfcd606
AD		 385static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns,
386 struct pid *pid, struct task_struct *task)
1da177e4 387{
ffb45122 388 unsigned long wchan;
9281acea 389 char symname[KSYM_NAME_LEN];
1da177e4 390
24b2ec21
AD		 391 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
392 goto print0;
1da177e4 393
24b2ec21
AD		 394 wchan = get_wchan(task);
395 if (wchan && !lookup_symbol_name(wchan, symname)) {
21dae0ad 396 seq_puts(m, symname);
24b2ec21
AD		 397 return 0;
398 }
25ce3191 399
24b2ec21
AD		 400print0:
401 seq_putc(m, '0');
25ce3191 402 return 0;
1da177e4
LT		 403}
404#endif /* CONFIG_KALLSYMS */
405
a9712bc1
AV		 406static int lock_trace(struct task_struct *task)
407{
408 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
409 if (err)
410 return err;
caaee623 411 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) {
a9712bc1
AV		 412 mutex_unlock(&task->signal->cred_guard_mutex);
413 return -EPERM;
414 }
415 return 0;
416}
417
418static void unlock_trace(struct task_struct *task)
419{
420 mutex_unlock(&task->signal->cred_guard_mutex);
421}
422
2ec220e2
KC		 423#ifdef CONFIG_STACKTRACE
424
425#define MAX_STACK_TRACE_DEPTH 64
426
427static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
428 struct pid *pid, struct task_struct *task)
429{
430 struct stack_trace trace;
431 unsigned long *entries;
a9712bc1 432 int err;
2ec220e2
KC		 433 int i;
434
435 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
436 if (!entries)
437 return -ENOMEM;
438
439 trace.nr_entries = 0;
440 trace.max_entries = MAX_STACK_TRACE_DEPTH;
441 trace.entries = entries;
442 trace.skip = 0;
2ec220e2 443
a9712bc1
AV		 444 err = lock_trace(task);
445 if (!err) {
446 save_stack_trace_tsk(task, &trace);
447
448 for (i = 0; i < trace.nr_entries; i++) {
8f5abe84 449 seq_printf(m, "[<0>] %pB\n", (void *)entries[i]);
a9712bc1
AV		 450 }
451 unlock_trace(task);
2ec220e2
KC		 452 }
453 kfree(entries);
454
a9712bc1 455 return err;
2ec220e2
KC		 456}
457#endif
458
5968cece 459#ifdef CONFIG_SCHED_INFO
1da177e4
LT		 460/*
461 * Provides /proc/PID/schedstat
462 */
f6e826ca
AD		 463static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
464 struct pid *pid, struct task_struct *task)
1da177e4 465{
5968cece
NR		 466 if (unlikely(!sched_info_on()))
467 seq_printf(m, "0 0 0\n");
468 else
469 seq_printf(m, "%llu %llu %lu\n",
25ce3191
JP		 470 (unsigned long long)task->se.sum_exec_runtime,
471 (unsigned long long)task->sched_info.run_delay,
472 task->sched_info.pcount);
473
474 return 0;
1da177e4
LT		 475}
476#endif
477
9745512c
AV		 478#ifdef CONFIG_LATENCYTOP
479static int lstats_show_proc(struct seq_file *m, void *v)
480{
481 int i;
13d77c37
HS		 482 struct inode *inode = m->private;
483 struct task_struct *task = get_proc_task(inode);
9745512c 484
13d77c37
HS		 485 if (!task)
486 return -ESRCH;
487 seq_puts(m, "Latency Top version : v0.1\n");
9745512c 488 for (i = 0; i < 32; i++) {
34e49d4f
JP		 489 struct latency_record *lr = &task->latency_record[i];
490 if (lr->backtrace[0]) {
9745512c 491 int q;
34e49d4f
JP		 492 seq_printf(m, "%i %li %li",
493 lr->count, lr->time, lr->max);
9745512c 494 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
34e49d4f
JP		 495 unsigned long bt = lr->backtrace[q];
496 if (!bt)
9745512c 497 break;
34e49d4f 498 if (bt == ULONG_MAX)
9745512c 499 break;
34e49d4f 500 seq_printf(m, " %ps", (void *)bt);
9745512c 501 }
9d6de12f 502 seq_putc(m, '\n');
9745512c
AV		 503 }
504
505 }
13d77c37 506 put_task_struct(task);
9745512c
AV		 507 return 0;
508}
509
510static int lstats_open(struct inode *inode, struct file *file)
511{
13d77c37 512 return single_open(file, lstats_show_proc, inode);
d6643d12
HS		 513}
514
9745512c
AV		 515static ssize_t lstats_write(struct file *file, const char __user *buf,
516 size_t count, loff_t *offs)
517{
496ad9aa 518 struct task_struct *task = get_proc_task(file_inode(file));
9745512c 519
13d77c37
HS		 520 if (!task)
521 return -ESRCH;
9745512c 522 clear_all_latency_tracing(task);
13d77c37 523 put_task_struct(task);
9745512c
AV		 524
525 return count;
526}
527
528static const struct file_operations proc_lstats_operations = {
529 .open = lstats_open,
530 .read = seq_read,
531 .write = lstats_write,
532 .llseek = seq_lseek,
13d77c37 533 .release = single_release,
9745512c
AV		 534};
535
536#endif
537
6ba51e37
AD		 538static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
539 struct pid *pid, struct task_struct *task)
1da177e4 540{
a7f638f9 541 unsigned long totalpages = totalram_pages + total_swap_pages;
b95c35e7 542 unsigned long points = 0;
1da177e4 543
ef419398
ON		 544 points = oom_badness(task, NULL, NULL, totalpages) *
545 1000 / totalpages;
25ce3191
JP		 546 seq_printf(m, "%lu\n", points);
547
548 return 0;
1da177e4
LT		 549}
550
d85f50d5 551struct limit_names {
cedbccab
AD		 552 const char *name;
553 const char *unit;
d85f50d5
NH		 554};
555
556static const struct limit_names lnames[RLIM_NLIMITS] = {
cff4edb5 557 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
d85f50d5
NH		 558 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
559 [RLIMIT_DATA] = {"Max data size", "bytes"},
560 [RLIMIT_STACK] = {"Max stack size", "bytes"},
561 [RLIMIT_CORE] = {"Max core file size", "bytes"},
562 [RLIMIT_RSS] = {"Max resident set", "bytes"},
563 [RLIMIT_NPROC] = {"Max processes", "processes"},
564 [RLIMIT_NOFILE] = {"Max open files", "files"},
565 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
566 [RLIMIT_AS] = {"Max address space", "bytes"},
567 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
568 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
569 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
570 [RLIMIT_NICE] = {"Max nice priority", NULL},
571 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
8808117c 572 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
d85f50d5
NH		 573};
574
575/* Display limits for a process */
1c963eb1
AD		 576static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
577 struct pid *pid, struct task_struct *task)
d85f50d5
NH		 578{
579 unsigned int i;
d85f50d5 580 unsigned long flags;
d85f50d5
NH		 581
582 struct rlimit rlim[RLIM_NLIMITS];
583
a6bebbc8 584 if (!lock_task_sighand(task, &flags))
d85f50d5 585 return 0;
d85f50d5
NH		 586 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
587 unlock_task_sighand(task, &flags);
d85f50d5
NH		 588
589 /*
590 * print the file header
591 */
1c963eb1 592 seq_printf(m, "%-25s %-20s %-20s %-10s\n",
25ce3191 593 "Limit", "Soft Limit", "Hard Limit", "Units");
d85f50d5
NH		 594
595 for (i = 0; i < RLIM_NLIMITS; i++) {
596 if (rlim[i].rlim_cur == RLIM_INFINITY)
1c963eb1 597 seq_printf(m, "%-25s %-20s ",
25ce3191 598 lnames[i].name, "unlimited");
d85f50d5 599 else
1c963eb1 600 seq_printf(m, "%-25s %-20lu ",
25ce3191 601 lnames[i].name, rlim[i].rlim_cur);
d85f50d5
NH		 602
603 if (rlim[i].rlim_max == RLIM_INFINITY)
1c963eb1 604 seq_printf(m, "%-20s ", "unlimited");
d85f50d5 605 else
1c963eb1 606 seq_printf(m, "%-20lu ", rlim[i].rlim_max);
d85f50d5
NH		 607
608 if (lnames[i].unit)
1c963eb1 609 seq_printf(m, "%-10s\n", lnames[i].unit);
d85f50d5 610 else
1c963eb1 611 seq_putc(m, '\n');
d85f50d5
NH		 612 }
613
1c963eb1 614 return 0;
d85f50d5
NH		 615}
616
ebcb6734 617#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
09d93bd6
AD		 618static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
619 struct pid *pid, struct task_struct *task)
ebcb6734
RM		 620{
621 long nr;
622 unsigned long args[6], sp, pc;
25ce3191
JP		 623 int res;
624
625 res = lock_trace(task);
a9712bc1
AV		 626 if (res)
627 return res;
ebcb6734
RM		 628
629 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
09d93bd6 630 seq_puts(m, "running\n");
a9712bc1 631 else if (nr < 0)
09d93bd6 632 seq_printf(m, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
a9712bc1 633 else
09d93bd6 634 seq_printf(m,
ebcb6734
RM		 635 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
636 nr,
637 args[0], args[1], args[2], args[3], args[4], args[5],
638 sp, pc);
a9712bc1 639 unlock_trace(task);
25ce3191
JP		 640
641 return 0;
ebcb6734
RM		 642}
643#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
644
1da177e4
LT		 645/************************************************************************/
646/* Here the fs part begins */
647/************************************************************************/
648
649/* permission checks */
778c1144 650static int proc_fd_access_allowed(struct inode *inode)
1da177e4 651{
778c1144
EB		 652 struct task_struct *task;
653 int allowed = 0;
df26c40e
EB		 654 /* Allow access to a task's file descriptors if it is us or we
655 * may use ptrace attach to the process and find out that
656 * information.
778c1144
EB		 657 */
658 task = get_proc_task(inode);
df26c40e 659 if (task) {
caaee623 660 allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
778c1144 661 put_task_struct(task);
df26c40e 662 }
778c1144 663 return allowed;
1da177e4
LT		 664}
665
6b4e306a 666int proc_setattr(struct dentry *dentry, struct iattr *attr)
6d76fa58
LT		 667{
668 int error;
2b0143b5 669 struct inode *inode = d_inode(dentry);
6d76fa58
LT		 670
671 if (attr->ia_valid & ATTR_MODE)
672 return -EPERM;
673
31051c85 674 error = setattr_prepare(dentry, attr);
1025774c
CH		 675 if (error)
676 return error;
677
1025774c
CH		 678 setattr_copy(inode, attr);
679 mark_inode_dirty(inode);
680 return 0;
6d76fa58
LT		 681}
682
0499680a
VK		 683/*
684 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
685 * or euid/egid (for hide_pid_min=2)?
686 */
687static bool has_pid_permissions(struct pid_namespace *pid,
688 struct task_struct *task,
689 int hide_pid_min)
690{
691 if (pid->hide_pid < hide_pid_min)
692 return true;
693 if (in_group_p(pid->pid_gid))
694 return true;
caaee623 695 return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
0499680a
VK		 696}
697
698
699static int proc_pid_permission(struct inode *inode, int mask)
700{
701 struct pid_namespace *pid = inode->i_sb->s_fs_info;
702 struct task_struct *task;
703 bool has_perms;
704
705 task = get_proc_task(inode);
a2ef990a
XF		 706 if (!task)
707 return -ESRCH;
796f571b 708 has_perms = has_pid_permissions(pid, task, HIDEPID_NO_ACCESS);
0499680a
VK		 709 put_task_struct(task);
710
711 if (!has_perms) {
796f571b 712 if (pid->hide_pid == HIDEPID_INVISIBLE) {
0499680a
VK		 713 /*
714 * Let's make getdents(), stat(), and open()
715 * consistent with each other. If a process
716 * may not stat() a file, it shouldn't be seen
717 * in procfs at all.
718 */
719 return -ENOENT;
720 }
721
722 return -EPERM;
723 }
724 return generic_permission(inode, mask);
725}
726
727
728
c5ef1c42 729static const struct inode_operations proc_def_inode_operations = {
6d76fa58
LT		 730 .setattr = proc_setattr,
731};
732
be614086
EB		 733static int proc_single_show(struct seq_file *m, void *v)
734{
735 struct inode *inode = m->private;
736 struct pid_namespace *ns;
737 struct pid *pid;
738 struct task_struct *task;
739 int ret;
740
741 ns = inode->i_sb->s_fs_info;
742 pid = proc_pid(inode);
743 task = get_pid_task(pid, PIDTYPE_PID);
744 if (!task)
745 return -ESRCH;
746
747 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
748
749 put_task_struct(task);
750 return ret;
751}
752
753static int proc_single_open(struct inode *inode, struct file *filp)
754{
c6a34058 755 return single_open(filp, proc_single_show, inode);
be614086
EB		 756}
757
758static const struct file_operations proc_single_file_operations = {
759 .open = proc_single_open,
760 .read = seq_read,
761 .llseek = seq_lseek,
762 .release = single_release,
763};
764
5381e169
ON		 765
766struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
1da177e4 767{
5381e169
ON		 768 struct task_struct *task = get_proc_task(inode);
769 struct mm_struct *mm = ERR_PTR(-ESRCH);
e268337d 770
5381e169 771 if (task) {
caaee623 772 mm = mm_access(task, mode | PTRACE_MODE_FSCREDS);
5381e169 773 put_task_struct(task);
e268337d 774
5381e169
ON		 775 if (!IS_ERR_OR_NULL(mm)) {
776 /* ensure this mm_struct can't be freed */
f1f10076 777 mmgrab(mm);
5381e169
ON		 778 /* but do not pin its memory */
779 mmput(mm);
780 }
781 }
782
783 return mm;
784}
785
786static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
787{
788 struct mm_struct *mm = proc_mem_open(inode, mode);
e268337d
LT		 789
790 if (IS_ERR(mm))
791 return PTR_ERR(mm);
792
e268337d 793 file->private_data = mm;
1da177e4
LT		 794 return 0;
795}
796
b409e578
CW		 797static int mem_open(struct inode *inode, struct file *file)
798{
bc452b4b
DH		 799 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
800
801 /* OK to pass negative loff_t, we can catch out-of-range */
802 file->f_mode |= FMODE_UNSIGNED_OFFSET;
803
804 return ret;
b409e578
CW		 805}
806
572d34b9
ON		 807static ssize_t mem_rw(struct file *file, char __user *buf,
808 size_t count, loff_t *ppos, int write)
1da177e4 809{
e268337d 810 struct mm_struct *mm = file->private_data;
572d34b9
ON		 811 unsigned long addr = *ppos;
812 ssize_t copied;
1da177e4 813 char *page;
272ddc8b 814 unsigned int flags;
1da177e4 815
e268337d
LT		 816 if (!mm)
817 return 0;
99f89551 818
0ee931c4 819 page = (char *)__get_free_page(GFP_KERNEL);
30cd8903 820 if (!page)
e268337d 821 return -ENOMEM;
1da177e4 822
f7ca54f4 823 copied = 0;
388f7934 824 if (!mmget_not_zero(mm))
6d08f2c7
ON		 825 goto free;
826
f511c0b1 827 flags = FOLL_FORCE | (write ? FOLL_WRITE : 0);
6347e8d5 828
1da177e4 829 while (count > 0) {
572d34b9 830 int this_len = min_t(int, count, PAGE_SIZE);
1da177e4 831
572d34b9 832 if (write && copy_from_user(page, buf, this_len)) {
1da177e4
LT		 833 copied = -EFAULT;
834 break;
835 }
572d34b9 836
6347e8d5 837 this_len = access_remote_vm(mm, addr, page, this_len, flags);
572d34b9 838 if (!this_len) {
1da177e4
LT		 839 if (!copied)
840 copied = -EIO;
841 break;
842 }
572d34b9
ON		 843
844 if (!write && copy_to_user(buf, page, this_len)) {
845 copied = -EFAULT;
846 break;
847 }
848
849 buf += this_len;
850 addr += this_len;
851 copied += this_len;
852 count -= this_len;
1da177e4 853 }
572d34b9 854 *ppos = addr;
30cd8903 855
6d08f2c7
ON		 856 mmput(mm);
857free:
30cd8903 858 free_page((unsigned long) page);
1da177e4
LT		 859 return copied;
860}
1da177e4 861
572d34b9
ON		 862static ssize_t mem_read(struct file *file, char __user *buf,
863 size_t count, loff_t *ppos)
864{
865 return mem_rw(file, buf, count, ppos, 0);
866}
867
868static ssize_t mem_write(struct file *file, const char __user *buf,
869 size_t count, loff_t *ppos)
870{
871 return mem_rw(file, (char __user*)buf, count, ppos, 1);
872}
873
85863e47 874loff_t mem_lseek(struct file *file, loff_t offset, int orig)
1da177e4
LT		 875{
876 switch (orig) {
877 case 0:
878 file->f_pos = offset;
879 break;
880 case 1:
881 file->f_pos += offset;
882 break;
883 default:
884 return -EINVAL;
885 }
886 force_successful_syscall_return();
887 return file->f_pos;
888}
889
e268337d
LT		 890static int mem_release(struct inode *inode, struct file *file)
891{
892 struct mm_struct *mm = file->private_data;
71879d3c 893 if (mm)
6d08f2c7 894 mmdrop(mm);
e268337d
LT		 895 return 0;
896}
897
00977a59 898static const struct file_operations proc_mem_operations = {
1da177e4
LT		 899 .llseek = mem_lseek,
900 .read = mem_read,
901 .write = mem_write,
902 .open = mem_open,
e268337d 903 .release = mem_release,
1da177e4
LT		 904};
905
b409e578
CW		 906static int environ_open(struct inode *inode, struct file *file)
907{
908 return __mem_open(inode, file, PTRACE_MODE_READ);
909}
910
315e28c8
JP		 911static ssize_t environ_read(struct file *file, char __user *buf,
912 size_t count, loff_t *ppos)
913{
315e28c8
JP		 914 char *page;
915 unsigned long src = *ppos;
b409e578
CW		 916 int ret = 0;
917 struct mm_struct *mm = file->private_data;
a3b609ef 918 unsigned long env_start, env_end;
315e28c8 919
8148a73c
MK		 920 /* Ensure the process spawned far enough to have an environment. */
921 if (!mm || !mm->env_end)
b409e578 922 return 0;
315e28c8 923
0ee931c4 924 page = (char *)__get_free_page(GFP_KERNEL);
315e28c8 925 if (!page)
b409e578 926 return -ENOMEM;
315e28c8 927
d6f64b89 928 ret = 0;
388f7934 929 if (!mmget_not_zero(mm))
b409e578 930 goto free;
a3b609ef
MG		 931
932 down_read(&mm->mmap_sem);
933 env_start = mm->env_start;
934 env_end = mm->env_end;
935 up_read(&mm->mmap_sem);
936
315e28c8 937 while (count > 0) {
e8905ec2
DH		 938 size_t this_len, max_len;
939 int retval;
315e28c8 940
a3b609ef 941 if (src >= (env_end - env_start))
315e28c8
JP		 942 break;
943
a3b609ef 944 this_len = env_end - (env_start + src);
e8905ec2
DH		 945
946 max_len = min_t(size_t, PAGE_SIZE, count);
947 this_len = min(max_len, this_len);
315e28c8 948
272ddc8b 949 retval = access_remote_vm(mm, (env_start + src), page, this_len, 0);
315e28c8
JP		 950
951 if (retval <= 0) {
952 ret = retval;
953 break;
954 }
955
956 if (copy_to_user(buf, page, retval)) {
957 ret = -EFAULT;
958 break;
959 }
960
961 ret += retval;
962 src += retval;
963 buf += retval;
964 count -= retval;
965 }
966 *ppos = src;
315e28c8 967 mmput(mm);
b409e578
CW		 968
969free:
315e28c8 970 free_page((unsigned long) page);
315e28c8
JP		 971 return ret;
972}
973
974static const struct file_operations proc_environ_operations = {
b409e578 975 .open = environ_open,
315e28c8 976 .read = environ_read,
87df8424 977 .llseek = generic_file_llseek,
b409e578 978 .release = mem_release,
315e28c8
JP		 979};
980
c5317167
AV		 981static int auxv_open(struct inode *inode, struct file *file)
982{
983 return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS);
984}
985
986static ssize_t auxv_read(struct file *file, char __user *buf,
987 size_t count, loff_t *ppos)
988{
989 struct mm_struct *mm = file->private_data;
990 unsigned int nwords = 0;
06b2849d
LY		 991
992 if (!mm)
993 return 0;
c5317167
AV		 994 do {
995 nwords += 2;
996 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
997 return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv,
998 nwords * sizeof(mm->saved_auxv[0]));
999}
1000
1001static const struct file_operations proc_auxv_operations = {
1002 .open = auxv_open,
1003 .read = auxv_read,
1004 .llseek = generic_file_llseek,
1005 .release = mem_release,
1006};
1007
fa0cbbf1
DR		 1008static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
1009 loff_t *ppos)
1010{
496ad9aa 1011 struct task_struct *task = get_proc_task(file_inode(file));
fa0cbbf1
DR		 1012 char buffer[PROC_NUMBUF];
1013 int oom_adj = OOM_ADJUST_MIN;
1014 size_t len;
fa0cbbf1
DR		 1015
1016 if (!task)
1017 return -ESRCH;
f913da59
MH		 1018 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
1019 oom_adj = OOM_ADJUST_MAX;
1020 else
1021 oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
1022 OOM_SCORE_ADJ_MAX;
fa0cbbf1
DR		 1023 put_task_struct(task);
1024 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
1025 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1026}
1027
1d5f0acb
MH		 1028static int __set_oom_adj(struct file *file, int oom_adj, bool legacy)
1029{
1030 static DEFINE_MUTEX(oom_adj_mutex);
44a70ade 1031 struct mm_struct *mm = NULL;
1d5f0acb
MH		 1032 struct task_struct *task;
1033 int err = 0;
1034
1035 task = get_proc_task(file_inode(file));
1036 if (!task)
1037 return -ESRCH;
1038
1039 mutex_lock(&oom_adj_mutex);
1040 if (legacy) {
1041 if (oom_adj < task->signal->oom_score_adj &&
1042 !capable(CAP_SYS_RESOURCE)) {
1043 err = -EACCES;
1044 goto err_unlock;
1045 }
1046 /*
1047 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1048 * /proc/pid/oom_score_adj instead.
1049 */
1050 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1051 current->comm, task_pid_nr(current), task_pid_nr(task),
1052 task_pid_nr(task));
1053 } else {
1054 if ((short)oom_adj < task->signal->oom_score_adj_min &&
1055 !capable(CAP_SYS_RESOURCE)) {
1056 err = -EACCES;
1057 goto err_unlock;
1058 }
1059 }
1060
44a70ade
MH		 1061 /*
1062 * Make sure we will check other processes sharing the mm if this is
1063 * not vfrok which wants its own oom_score_adj.
1064 * pin the mm so it doesn't go away and get reused after task_unlock
1065 */
1066 if (!task->vfork_done) {
1067 struct task_struct *p = find_lock_task_mm(task);
1068
1069 if (p) {
1070 if (atomic_read(&p->mm->mm_users) > 1) {
1071 mm = p->mm;
f1f10076 1072 mmgrab(mm);
44a70ade
MH		 1073 }
1074 task_unlock(p);
1075 }
1076 }
1077
1d5f0acb
MH		 1078 task->signal->oom_score_adj = oom_adj;
1079 if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1080 task->signal->oom_score_adj_min = (short)oom_adj;
1081 trace_oom_score_adj_update(task);
44a70ade
MH		 1082
1083 if (mm) {
1084 struct task_struct *p;
1085
1086 rcu_read_lock();
1087 for_each_process(p) {
1088 if (same_thread_group(task, p))
1089 continue;
1090
1091 /* do not touch kernel threads or the global init */
1092 if (p->flags & PF_KTHREAD || is_global_init(p))
1093 continue;
1094
1095 task_lock(p);
1096 if (!p->vfork_done && process_shares_mm(p, mm)) {
1097 pr_info("updating oom_score_adj for %d (%s) from %d to %d because it shares mm with %d (%s). Report if this is unexpected.\n",
1098 task_pid_nr(p), p->comm,
1099 p->signal->oom_score_adj, oom_adj,
1100 task_pid_nr(task), task->comm);
1101 p->signal->oom_score_adj = oom_adj;
1102 if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1103 p->signal->oom_score_adj_min = (short)oom_adj;
1104 }
1105 task_unlock(p);
1106 }
1107 rcu_read_unlock();
1108 mmdrop(mm);
1109 }
1d5f0acb
MH		 1110err_unlock:
1111 mutex_unlock(&oom_adj_mutex);
1112 put_task_struct(task);
1113 return err;
1114}
f913da59 1115
b72bdfa7
DR		 1116/*
1117 * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1118 * kernels. The effective policy is defined by oom_score_adj, which has a
1119 * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1120 * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1121 * Processes that become oom disabled via oom_adj will still be oom disabled
1122 * with this implementation.
1123 *
1124 * oom_adj cannot be removed since existing userspace binaries use it.
1125 */
fa0cbbf1
DR		 1126static ssize_t oom_adj_write(struct file *file, const char __user *buf,
1127 size_t count, loff_t *ppos)
1128{
fa0cbbf1
DR		 1129 char buffer[PROC_NUMBUF];
1130 int oom_adj;
fa0cbbf1
DR		 1131 int err;
1132
1133 memset(buffer, 0, sizeof(buffer));
1134 if (count > sizeof(buffer) - 1)
1135 count = sizeof(buffer) - 1;
1136 if (copy_from_user(buffer, buf, count)) {
1137 err = -EFAULT;
1138 goto out;
1139 }
1140
1141 err = kstrtoint(strstrip(buffer), 0, &oom_adj);
1142 if (err)
1143 goto out;
1144 if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
1145 oom_adj != OOM_DISABLE) {
1146 err = -EINVAL;
1147 goto out;
1148 }
1149
fa0cbbf1
DR		 1150 /*
1151 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1152 * value is always attainable.
1153 */
1154 if (oom_adj == OOM_ADJUST_MAX)
1155 oom_adj = OOM_SCORE_ADJ_MAX;
1156 else
1157 oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
1158
1d5f0acb 1159 err = __set_oom_adj(file, oom_adj, true);
fa0cbbf1
DR		 1160out:
1161 return err < 0 ? err : count;
1162}
1163
1164static const struct file_operations proc_oom_adj_operations = {
1165 .read = oom_adj_read,
1166 .write = oom_adj_write,
1167 .llseek = generic_file_llseek,
1168};
1169
a63d83f4
DR		 1170static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1171 size_t count, loff_t *ppos)
1172{
496ad9aa 1173 struct task_struct *task = get_proc_task(file_inode(file));
a63d83f4 1174 char buffer[PROC_NUMBUF];
a9c58b90 1175 short oom_score_adj = OOM_SCORE_ADJ_MIN;
a63d83f4
DR		 1176 size_t len;
1177
1178 if (!task)
1179 return -ESRCH;
f913da59 1180 oom_score_adj = task->signal->oom_score_adj;
a63d83f4 1181 put_task_struct(task);
a9c58b90 1182 len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
a63d83f4
DR		 1183 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1184}
1185
1186static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1187 size_t count, loff_t *ppos)
1188{
a63d83f4 1189 char buffer[PROC_NUMBUF];
0a8cb8e3 1190 int oom_score_adj;
a63d83f4
DR		 1191 int err;
1192
1193 memset(buffer, 0, sizeof(buffer));
1194 if (count > sizeof(buffer) - 1)
1195 count = sizeof(buffer) - 1;
723548bf
DR		 1196 if (copy_from_user(buffer, buf, count)) {
1197 err = -EFAULT;
1198 goto out;
1199 }
a63d83f4 1200
0a8cb8e3 1201 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
a63d83f4 1202 if (err)
723548bf 1203 goto out;
a63d83f4 1204 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
723548bf
DR		 1205 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1206 err = -EINVAL;
1207 goto out;
1208 }
a63d83f4 1209
1d5f0acb 1210 err = __set_oom_adj(file, oom_score_adj, false);
723548bf
DR		 1211out:
1212 return err < 0 ? err : count;
a63d83f4
DR		 1213}
1214
1215static const struct file_operations proc_oom_score_adj_operations = {
1216 .read = oom_score_adj_read,
1217 .write = oom_score_adj_write,
6038f373 1218 .llseek = default_llseek,
a63d83f4
DR		 1219};
1220
1da177e4 1221#ifdef CONFIG_AUDITSYSCALL
b4eb4f7f 1222#define TMPBUFLEN 11
1da177e4
LT		 1223static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1224 size_t count, loff_t *ppos)
1225{
496ad9aa 1226 struct inode * inode = file_inode(file);
99f89551 1227 struct task_struct *task = get_proc_task(inode);
1da177e4
LT		 1228 ssize_t length;
1229 char tmpbuf[TMPBUFLEN];
1230
99f89551
EB		 1231 if (!task)
1232 return -ESRCH;
1da177e4 1233 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
e1760bd5
EB		 1234 from_kuid(file->f_cred->user_ns,
1235 audit_get_loginuid(task)));
99f89551 1236 put_task_struct(task);
1da177e4
LT		 1237 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1238}
1239
1240static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1241 size_t count, loff_t *ppos)
1242{
496ad9aa 1243 struct inode * inode = file_inode(file);
1da177e4 1244 uid_t loginuid;
e1760bd5 1245 kuid_t kloginuid;
774636e1 1246 int rv;
1da177e4 1247
7dc52157
PM		 1248 rcu_read_lock();
1249 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1250 rcu_read_unlock();
1da177e4 1251 return -EPERM;
7dc52157
PM		 1252 }
1253 rcu_read_unlock();
1da177e4 1254
1da177e4
LT		 1255 if (*ppos != 0) {
1256 /* No partial writes. */
1257 return -EINVAL;
1258 }
1da177e4 1259
774636e1
AD		 1260 rv = kstrtou32_from_user(buf, count, 10, &loginuid);
1261 if (rv < 0)
1262 return rv;
81407c84
EP		 1263
1264 /* is userspace tring to explicitly UNSET the loginuid? */
1265 if (loginuid == AUDIT_UID_UNSET) {
1266 kloginuid = INVALID_UID;
1267 } else {
1268 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
774636e1
AD		 1269 if (!uid_valid(kloginuid))
1270 return -EINVAL;
e1760bd5
EB		 1271 }
1272
774636e1
AD		 1273 rv = audit_set_loginuid(kloginuid);
1274 if (rv < 0)
1275 return rv;
1276 return count;
1da177e4
LT		 1277}
1278
00977a59 1279static const struct file_operations proc_loginuid_operations = {
1da177e4
LT		 1280 .read = proc_loginuid_read,
1281 .write = proc_loginuid_write,
87df8424 1282 .llseek = generic_file_llseek,
1da177e4 1283};
1e0bd755
EP		 1284
1285static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1286 size_t count, loff_t *ppos)
1287{
496ad9aa 1288 struct inode * inode = file_inode(file);
1e0bd755
EP		 1289 struct task_struct *task = get_proc_task(inode);
1290 ssize_t length;
1291 char tmpbuf[TMPBUFLEN];
1292
1293 if (!task)
1294 return -ESRCH;
1295 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1296 audit_get_sessionid(task));
1297 put_task_struct(task);
1298 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1299}
1300
1301static const struct file_operations proc_sessionid_operations = {
1302 .read = proc_sessionid_read,
87df8424 1303 .llseek = generic_file_llseek,
1e0bd755 1304};
1da177e4
LT		 1305#endif
1306
f4f154fd
AM		 1307#ifdef CONFIG_FAULT_INJECTION
1308static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1309 size_t count, loff_t *ppos)
1310{
496ad9aa 1311 struct task_struct *task = get_proc_task(file_inode(file));
f4f154fd
AM		 1312 char buffer[PROC_NUMBUF];
1313 size_t len;
1314 int make_it_fail;
f4f154fd
AM		 1315
1316 if (!task)
1317 return -ESRCH;
1318 make_it_fail = task->make_it_fail;
1319 put_task_struct(task);
1320
1321 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
0c28f287
AM		 1322
1323 return simple_read_from_buffer(buf, count, ppos, buffer, len);
f4f154fd
AM		 1324}
1325
1326static ssize_t proc_fault_inject_write(struct file * file,
1327 const char __user * buf, size_t count, loff_t *ppos)
1328{
1329 struct task_struct *task;
774636e1 1330 char buffer[PROC_NUMBUF];
f4f154fd 1331 int make_it_fail;
774636e1 1332 int rv;
f4f154fd
AM		 1333
1334 if (!capable(CAP_SYS_RESOURCE))
1335 return -EPERM;
1336 memset(buffer, 0, sizeof(buffer));
1337 if (count > sizeof(buffer) - 1)
1338 count = sizeof(buffer) - 1;
1339 if (copy_from_user(buffer, buf, count))
1340 return -EFAULT;
774636e1
AD		 1341 rv = kstrtoint(strstrip(buffer), 0, &make_it_fail);
1342 if (rv < 0)
1343 return rv;
16caed31
DJ		 1344 if (make_it_fail < 0 || make_it_fail > 1)
1345 return -EINVAL;
1346
496ad9aa 1347 task = get_proc_task(file_inode(file));
f4f154fd
AM		 1348 if (!task)
1349 return -ESRCH;
1350 task->make_it_fail = make_it_fail;
1351 put_task_struct(task);
cba8aafe
VL		 1352
1353 return count;
f4f154fd
AM		 1354}
1355
00977a59 1356static const struct file_operations proc_fault_inject_operations = {
f4f154fd
AM		 1357 .read = proc_fault_inject_read,
1358 .write = proc_fault_inject_write,
87df8424 1359 .llseek = generic_file_llseek,
f4f154fd 1360};
e41d5818
DV		 1361
1362static ssize_t proc_fail_nth_write(struct file *file, const char __user *buf,
1363 size_t count, loff_t *ppos)
1364{
1365 struct task_struct *task;
9049f2f6
AM		 1366 int err;
1367 unsigned int n;
e41d5818 1368
1203c8e6
AM		 1369 err = kstrtouint_from_user(buf, count, 0, &n);
1370 if (err)
1371 return err;
1372
e41d5818
DV		 1373 task = get_proc_task(file_inode(file));
1374 if (!task)
1375 return -ESRCH;
9f7118b2 1376 task->fail_nth = n;
e41d5818 1377 put_task_struct(task);
1203c8e6 1378
e41d5818
DV		 1379 return count;
1380}
1381
1382static ssize_t proc_fail_nth_read(struct file *file, char __user *buf,
1383 size_t count, loff_t *ppos)
1384{
1385 struct task_struct *task;
bfc74093
AM		 1386 char numbuf[PROC_NUMBUF];
1387 ssize_t len;
e41d5818
DV		 1388
1389 task = get_proc_task(file_inode(file));
1390 if (!task)
1391 return -ESRCH;
9f7118b2 1392 len = snprintf(numbuf, sizeof(numbuf), "%u\n", task->fail_nth);
bfc74093 1393 len = simple_read_from_buffer(buf, count, ppos, numbuf, len);
1203c8e6 1394 put_task_struct(task);
bfc74093
AM		 1395
1396 return len;
e41d5818
DV		 1397}
1398
1399static const struct file_operations proc_fail_nth_operations = {
1400 .read = proc_fail_nth_read,
1401 .write = proc_fail_nth_write,
1402};
f4f154fd
AM		 1403#endif
1404
9745512c 1405
43ae34cb
IM		 1406#ifdef CONFIG_SCHED_DEBUG
1407/*
1408 * Print out various scheduling related per-task fields:
1409 */
1410static int sched_show(struct seq_file *m, void *v)
1411{
1412 struct inode *inode = m->private;
74dc3384 1413 struct pid_namespace *ns = inode->i_sb->s_fs_info;
43ae34cb
IM		 1414 struct task_struct *p;
1415
43ae34cb
IM		 1416 p = get_proc_task(inode);
1417 if (!p)
1418 return -ESRCH;
74dc3384 1419 proc_sched_show_task(p, ns, m);
43ae34cb
IM		 1420
1421 put_task_struct(p);
1422
1423 return 0;
1424}
1425
1426static ssize_t
1427sched_write(struct file *file, const char __user *buf,
1428 size_t count, loff_t *offset)
1429{
496ad9aa 1430 struct inode *inode = file_inode(file);
43ae34cb
IM		 1431 struct task_struct *p;
1432
43ae34cb
IM		 1433 p = get_proc_task(inode);
1434 if (!p)
1435 return -ESRCH;
1436 proc_sched_set_task(p);
1437
1438 put_task_struct(p);
1439
1440 return count;
1441}
1442
1443static int sched_open(struct inode *inode, struct file *filp)
1444{
c6a34058 1445 return single_open(filp, sched_show, inode);
43ae34cb
IM		 1446}
1447
1448static const struct file_operations proc_pid_sched_operations = {
1449 .open = sched_open,
1450 .read = seq_read,
1451 .write = sched_write,
1452 .llseek = seq_lseek,
5ea473a1 1453 .release = single_release,
43ae34cb
IM		 1454};
1455
1456#endif
1457
5091faa4
MG		 1458#ifdef CONFIG_SCHED_AUTOGROUP
1459/*
1460 * Print out autogroup related information:
1461 */
1462static int sched_autogroup_show(struct seq_file *m, void *v)
1463{
1464 struct inode *inode = m->private;
1465 struct task_struct *p;
1466
1467 p = get_proc_task(inode);
1468 if (!p)
1469 return -ESRCH;
1470 proc_sched_autogroup_show_task(p, m);
1471
1472 put_task_struct(p);
1473
1474 return 0;
1475}
1476
1477static ssize_t
1478sched_autogroup_write(struct file *file, const char __user *buf,
1479 size_t count, loff_t *offset)
1480{
496ad9aa 1481 struct inode *inode = file_inode(file);
5091faa4
MG		 1482 struct task_struct *p;
1483 char buffer[PROC_NUMBUF];
0a8cb8e3 1484 int nice;
5091faa4
MG		 1485 int err;
1486
1487 memset(buffer, 0, sizeof(buffer));
1488 if (count > sizeof(buffer) - 1)
1489 count = sizeof(buffer) - 1;
1490 if (copy_from_user(buffer, buf, count))
1491 return -EFAULT;
1492
0a8cb8e3
AD		 1493 err = kstrtoint(strstrip(buffer), 0, &nice);
1494 if (err < 0)
1495 return err;
5091faa4
MG		 1496
1497 p = get_proc_task(inode);
1498 if (!p)
1499 return -ESRCH;
1500
2e5b5b3a 1501 err = proc_sched_autogroup_set_nice(p, nice);
5091faa4
MG		 1502 if (err)
1503 count = err;
1504
1505 put_task_struct(p);
1506
1507 return count;
1508}
1509
1510static int sched_autogroup_open(struct inode *inode, struct file *filp)
1511{
1512 int ret;
1513
1514 ret = single_open(filp, sched_autogroup_show, NULL);
1515 if (!ret) {
1516 struct seq_file *m = filp->private_data;
1517
1518 m->private = inode;
1519 }
1520 return ret;
1521}
1522
1523static const struct file_operations proc_pid_sched_autogroup_operations = {
1524 .open = sched_autogroup_open,
1525 .read = seq_read,
1526 .write = sched_autogroup_write,
1527 .llseek = seq_lseek,
1528 .release = single_release,
1529};
1530
1531#endif /* CONFIG_SCHED_AUTOGROUP */
1532
4614a696 1533static ssize_t comm_write(struct file *file, const char __user *buf,
1534 size_t count, loff_t *offset)
1535{
496ad9aa 1536 struct inode *inode = file_inode(file);
4614a696 1537 struct task_struct *p;
1538 char buffer[TASK_COMM_LEN];
830e0fc9 1539 const size_t maxlen = sizeof(buffer) - 1;
4614a696 1540
1541 memset(buffer, 0, sizeof(buffer));
830e0fc9 1542 if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
4614a696 1543 return -EFAULT;
1544
1545 p = get_proc_task(inode);
1546 if (!p)
1547 return -ESRCH;
1548
1549 if (same_thread_group(current, p))
1550 set_task_comm(p, buffer);
1551 else
1552 count = -EINVAL;
1553
1554 put_task_struct(p);
1555
1556 return count;
1557}
1558
1559static int comm_show(struct seq_file *m, void *v)
1560{
1561 struct inode *inode = m->private;
1562 struct task_struct *p;
1563
1564 p = get_proc_task(inode);
1565 if (!p)
1566 return -ESRCH;
1567
1568 task_lock(p);
1569 seq_printf(m, "%s\n", p->comm);
1570 task_unlock(p);
1571
1572 put_task_struct(p);
1573
1574 return 0;
1575}
1576
1577static int comm_open(struct inode *inode, struct file *filp)
1578{
c6a34058 1579 return single_open(filp, comm_show, inode);
4614a696 1580}
1581
1582static const struct file_operations proc_pid_set_comm_operations = {
1583 .open = comm_open,
1584 .read = seq_read,
1585 .write = comm_write,
1586 .llseek = seq_lseek,
1587 .release = single_release,
1588};
1589
7773fbc5 1590static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
925d1c40
MH		 1591{
1592 struct task_struct *task;
925d1c40
MH		 1593 struct file *exe_file;
1594
2b0143b5 1595 task = get_proc_task(d_inode(dentry));
925d1c40
MH		 1596 if (!task)
1597 return -ENOENT;
cd81a917 1598 exe_file = get_task_exe_file(task);
925d1c40 1599 put_task_struct(task);
925d1c40
MH		 1600 if (exe_file) {
1601 *exe_path = exe_file->f_path;
1602 path_get(&exe_file->f_path);
1603 fput(exe_file);
1604 return 0;
1605 } else
1606 return -ENOENT;
1607}
1608
6b255391 1609static const char *proc_pid_get_link(struct dentry *dentry,
fceef393
AV		 1610 struct inode *inode,
1611 struct delayed_call *done)
1da177e4 1612{
408ef013 1613 struct path path;
1da177e4
LT		 1614 int error = -EACCES;
1615
6b255391
AV		 1616 if (!dentry)
1617 return ERR_PTR(-ECHILD);
1618
778c1144
EB		 1619 /* Are we allowed to snoop on the tasks file descriptors? */
1620 if (!proc_fd_access_allowed(inode))
1da177e4 1621 goto out;
1da177e4 1622
408ef013
CH		 1623 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1624 if (error)
1625 goto out;
1626
6e77137b 1627 nd_jump_link(&path);
408ef013 1628 return NULL;
1da177e4 1629out:
008b150a 1630 return ERR_PTR(error);
1da177e4
LT		 1631}
1632
3dcd25f3 1633static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1da177e4 1634{
0ee931c4 1635 char *tmp = (char *)__get_free_page(GFP_KERNEL);
3dcd25f3 1636 char *pathname;
1da177e4
LT		 1637 int len;
1638
1639 if (!tmp)
1640 return -ENOMEM;
0c28f287 1641
7b2a69ba 1642 pathname = d_path(path, tmp, PAGE_SIZE);
3dcd25f3
JB		 1643 len = PTR_ERR(pathname);
1644 if (IS_ERR(pathname))
1da177e4 1645 goto out;
3dcd25f3 1646 len = tmp + PAGE_SIZE - 1 - pathname;
1da177e4
LT		 1647
1648 if (len > buflen)
1649 len = buflen;
3dcd25f3 1650 if (copy_to_user(buffer, pathname, len))
1da177e4
LT		 1651 len = -EFAULT;
1652 out:
1653 free_page((unsigned long)tmp);
1654 return len;
1655}
1656
1657static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1658{
1659 int error = -EACCES;
2b0143b5 1660 struct inode *inode = d_inode(dentry);
3dcd25f3 1661 struct path path;
1da177e4 1662
778c1144
EB		 1663 /* Are we allowed to snoop on the tasks file descriptors? */
1664 if (!proc_fd_access_allowed(inode))
1da177e4 1665 goto out;
1da177e4 1666
7773fbc5 1667 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1da177e4
LT		 1668 if (error)
1669 goto out;
1670
3dcd25f3
JB		 1671 error = do_proc_readlink(&path, buffer, buflen);
1672 path_put(&path);
1da177e4 1673out:
1da177e4
LT		 1674 return error;
1675}
1676
faf60af1 1677const struct inode_operations proc_pid_link_inode_operations = {
1da177e4 1678 .readlink = proc_pid_readlink,
6b255391 1679 .get_link = proc_pid_get_link,
6d76fa58 1680 .setattr = proc_setattr,
1da177e4
LT		 1681};
1682
28a6d671
EB		 1683
1684/* building an inode */
1685
c6eb50d2 1686void task_dump_owner(struct task_struct *task, umode_t mode,
68eb94f1
EB		 1687 kuid_t *ruid, kgid_t *rgid)
1688{
1689 /* Depending on the state of dumpable compute who should own a
1690 * proc file for a task.
1691 */
1692 const struct cred *cred;
1693 kuid_t uid;
1694 kgid_t gid;
1695
2e0ad552
AD		 1696 if (unlikely(task->flags & PF_KTHREAD)) {
1697 *ruid = GLOBAL_ROOT_UID;
1698 *rgid = GLOBAL_ROOT_GID;
1699 return;
1700 }
1701
68eb94f1
EB		 1702 /* Default to the tasks effective ownership */
1703 rcu_read_lock();
1704 cred = __task_cred(task);
1705 uid = cred->euid;
1706 gid = cred->egid;
1707 rcu_read_unlock();
1708
1709 /*
1710 * Before the /proc/pid/status file was created the only way to read
1711 * the effective uid of a /process was to stat /proc/pid. Reading
1712 * /proc/pid/status is slow enough that procps and other packages
1713 * kept stating /proc/pid. To keep the rules in /proc simple I have
1714 * made this apply to all per process world readable and executable
1715 * directories.
1716 */
1717 if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) {
1718 struct mm_struct *mm;
1719 task_lock(task);
1720 mm = task->mm;
1721 /* Make non-dumpable tasks owned by some root */
1722 if (mm) {
1723 if (get_dumpable(mm) != SUID_DUMP_USER) {
1724 struct user_namespace *user_ns = mm->user_ns;
1725
1726 uid = make_kuid(user_ns, 0);
1727 if (!uid_valid(uid))
1728 uid = GLOBAL_ROOT_UID;
1729
1730 gid = make_kgid(user_ns, 0);
1731 if (!gid_valid(gid))
1732 gid = GLOBAL_ROOT_GID;
1733 }
1734 } else {
1735 uid = GLOBAL_ROOT_UID;
1736 gid = GLOBAL_ROOT_GID;
1737 }
1738 task_unlock(task);
1739 }
1740 *ruid = uid;
1741 *rgid = gid;
1742}
1743
db978da8
AG		 1744struct inode *proc_pid_make_inode(struct super_block * sb,
1745 struct task_struct *task, umode_t mode)
28a6d671
EB		 1746{
1747 struct inode * inode;
1748 struct proc_inode *ei;
1da177e4 1749
28a6d671 1750 /* We need a new inode */
1da177e4 1751
28a6d671
EB		 1752 inode = new_inode(sb);
1753 if (!inode)
1754 goto out;
1755
1756 /* Common stuff */
1757 ei = PROC_I(inode);
db978da8 1758 inode->i_mode = mode;
85fe4025 1759 inode->i_ino = get_next_ino();
078cd827 1760 inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode);
28a6d671
EB		 1761 inode->i_op = &proc_def_inode_operations;
1762
1763 /*
1764 * grab the reference to task.
1765 */
1a657f78 1766 ei->pid = get_task_pid(task, PIDTYPE_PID);
28a6d671
EB		 1767 if (!ei->pid)
1768 goto out_unlock;
1769
68eb94f1 1770 task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
28a6d671
EB		 1771 security_task_to_inode(task, inode);
1772
1da177e4 1773out:
28a6d671
EB		 1774 return inode;
1775
1776out_unlock:
1777 iput(inode);
1778 return NULL;
1da177e4
LT		 1779}
1780
a528d35e
DH		 1781int pid_getattr(const struct path *path, struct kstat *stat,
1782 u32 request_mask, unsigned int query_flags)
1da177e4 1783{
a528d35e 1784 struct inode *inode = d_inode(path->dentry);
28a6d671 1785 struct task_struct *task;
a528d35e 1786 struct pid_namespace *pid = path->dentry->d_sb->s_fs_info;
c69e8d9c 1787
28a6d671 1788 generic_fillattr(inode, stat);
1da177e4 1789
28a6d671 1790 rcu_read_lock();
dcb0f222
EB		 1791 stat->uid = GLOBAL_ROOT_UID;
1792 stat->gid = GLOBAL_ROOT_GID;
28a6d671
EB		 1793 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1794 if (task) {
796f571b 1795 if (!has_pid_permissions(pid, task, HIDEPID_INVISIBLE)) {
0499680a
VK		 1796 rcu_read_unlock();
1797 /*
1798 * This doesn't prevent learning whether PID exists,
1799 * it only makes getattr() consistent with readdir().
1800 */
1801 return -ENOENT;
1802 }
68eb94f1 1803 task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid);
1da177e4 1804 }
28a6d671 1805 rcu_read_unlock();
d6e71144 1806 return 0;
1da177e4
LT		 1807}
1808
1da177e4
LT		 1809/* dentry stuff */
1810
1811/*
1812 * Exceptional case: normally we are not allowed to unhash a busy
1813 * directory. In this case, however, we can do it - no aliasing problems
1814 * due to the way we treat inodes.
1815 *
1816 * Rewrite the inode's ownerships here because the owning task may have
1817 * performed a setuid(), etc.
99f89551 1818 *
1da177e4 1819 */
0b728e19 1820int pid_revalidate(struct dentry *dentry, unsigned int flags)
1da177e4 1821{
34286d66
NP		 1822 struct inode *inode;
1823 struct task_struct *task;
c69e8d9c 1824
0b728e19 1825 if (flags & LOOKUP_RCU)
34286d66
NP		 1826 return -ECHILD;
1827
2b0143b5 1828 inode = d_inode(dentry);
34286d66
NP		 1829 task = get_proc_task(inode);
1830
99f89551 1831 if (task) {
68eb94f1
EB		 1832 task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid);
1833
9ee8ab9f 1834 inode->i_mode &= ~(S_ISUID | S_ISGID);
1da177e4 1835 security_task_to_inode(task, inode);
99f89551 1836 put_task_struct(task);
1da177e4
LT		 1837 return 1;
1838 }
1da177e4
LT		 1839 return 0;
1840}
1841
d855a4b7
ON		 1842static inline bool proc_inode_is_dead(struct inode *inode)
1843{
1844 return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
1845}
1846
1dd704b6
DH		 1847int pid_delete_dentry(const struct dentry *dentry)
1848{
1849 /* Is the task we represent dead?
1850 * If so, then don't put the dentry on the lru list,
1851 * kill it immediately.
1852 */
2b0143b5 1853 return proc_inode_is_dead(d_inode(dentry));
1dd704b6
DH		 1854}
1855
6b4e306a 1856const struct dentry_operations pid_dentry_operations =
28a6d671
EB		 1857{
1858 .d_revalidate = pid_revalidate,
1859 .d_delete = pid_delete_dentry,
1860};
1861
1862/* Lookups */
1863
1c0d04c9
EB		 1864/*
1865 * Fill a directory entry.
1866 *
1867 * If possible create the dcache entry and derive our inode number and
1868 * file type from dcache entry.
1869 *
1870 * Since all of the proc inode numbers are dynamically generated, the inode
1871 * numbers do not exist until the inode is cache. This means creating the
1872 * the dcache entry in readdir is necessary to keep the inode numbers
1873 * reported by readdir in sync with the inode numbers reported
1874 * by stat.
1875 */
f0c3b509 1876bool proc_fill_cache(struct file *file, struct dir_context *ctx,
6b4e306a 1877 const char *name, int len,
c5141e6d 1878 instantiate_t instantiate, struct task_struct *task, const void *ptr)
61a28784 1879{
f0c3b509 1880 struct dentry *child, *dir = file->f_path.dentry;
1df98b8b 1881 struct qstr qname = QSTR_INIT(name, len);
61a28784 1882 struct inode *inode;
1df98b8b
AV		 1883 unsigned type;
1884 ino_t ino;
61a28784 1885
1df98b8b 1886 child = d_hash_and_lookup(dir, &qname);
61a28784 1887 if (!child) {
3781764b
AV		 1888 DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq);
1889 child = d_alloc_parallel(dir, &qname, &wq);
1890 if (IS_ERR(child))
1df98b8b 1891 goto end_instantiate;
3781764b
AV		 1892 if (d_in_lookup(child)) {
1893 int err = instantiate(d_inode(dir), child, task, ptr);
1894 d_lookup_done(child);
1895 if (err < 0) {
1896 dput(child);
1897 goto end_instantiate;
1898 }
61a28784
EB		 1899 }
1900 }
2b0143b5 1901 inode = d_inode(child);
147ce699
AV		 1902 ino = inode->i_ino;
1903 type = inode->i_mode >> 12;
61a28784 1904 dput(child);
f0c3b509 1905 return dir_emit(ctx, name, len, ino, type);
1df98b8b
AV		 1906
1907end_instantiate:
1908 return dir_emit(ctx, name, len, 1, DT_UNKNOWN);
61a28784
EB		 1909}
1910
640708a2
PE		 1911/*
1912 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1913 * which represent vma start and end addresses.
1914 */
1915static int dname_to_vma_addr(struct dentry *dentry,
1916 unsigned long *start, unsigned long *end)
1917{
ac7f1061
AD		 1918 const char *str = dentry->d_name.name;
1919 unsigned long long sval, eval;
1920 unsigned int len;
1921
35318db5
AD		 1922 if (str[0] == '0' && str[1] != '-')
1923 return -EINVAL;
ac7f1061
AD		 1924 len = _parse_integer(str, 16, &sval);
1925 if (len & KSTRTOX_OVERFLOW)
1926 return -EINVAL;
1927 if (sval != (unsigned long)sval)
1928 return -EINVAL;
1929 str += len;
1930
1931 if (*str != '-')
640708a2 1932 return -EINVAL;
ac7f1061
AD		 1933 str++;
1934
35318db5
AD		 1935 if (str[0] == '0' && str[1])
1936 return -EINVAL;
ac7f1061
AD		 1937 len = _parse_integer(str, 16, &eval);
1938 if (len & KSTRTOX_OVERFLOW)
1939 return -EINVAL;
1940 if (eval != (unsigned long)eval)
1941 return -EINVAL;
1942 str += len;
1943
1944 if (*str != '\0')
1945 return -EINVAL;
1946
1947 *start = sval;
1948 *end = eval;
640708a2
PE		 1949
1950 return 0;
1951}
1952
0b728e19 1953static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
640708a2
PE		 1954{
1955 unsigned long vm_start, vm_end;
1956 bool exact_vma_exists = false;
1957 struct mm_struct *mm = NULL;
1958 struct task_struct *task;
640708a2
PE		 1959 struct inode *inode;
1960 int status = 0;
1961
0b728e19 1962 if (flags & LOOKUP_RCU)
640708a2
PE		 1963 return -ECHILD;
1964
2b0143b5 1965 inode = d_inode(dentry);
640708a2
PE		 1966 task = get_proc_task(inode);
1967 if (!task)
1968 goto out_notask;
1969
caaee623 1970 mm = mm_access(task, PTRACE_MODE_READ_FSCREDS);
2344bec7 1971 if (IS_ERR_OR_NULL(mm))
640708a2
PE		 1972 goto out;
1973
1974 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1975 down_read(&mm->mmap_sem);
1976 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1977 up_read(&mm->mmap_sem);
1978 }
1979
1980 mmput(mm);
1981
1982 if (exact_vma_exists) {
68eb94f1
EB		 1983 task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
1984
640708a2
PE		 1985 security_task_to_inode(task, inode);
1986 status = 1;
1987 }
1988
1989out:
1990 put_task_struct(task);
1991
1992out_notask:
640708a2
PE		 1993 return status;
1994}
1995
1996static const struct dentry_operations tid_map_files_dentry_operations = {
1997 .d_revalidate = map_files_d_revalidate,
1998 .d_delete = pid_delete_dentry,
1999};
2000
6b255391 2001static int map_files_get_link(struct dentry *dentry, struct path *path)
640708a2
PE		 2002{
2003 unsigned long vm_start, vm_end;
2004 struct vm_area_struct *vma;
2005 struct task_struct *task;
2006 struct mm_struct *mm;
2007 int rc;
2008
2009 rc = -ENOENT;
2b0143b5 2010 task = get_proc_task(d_inode(dentry));
640708a2
PE		 2011 if (!task)
2012 goto out;
2013
2014 mm = get_task_mm(task);
2015 put_task_struct(task);
2016 if (!mm)
2017 goto out;
2018
2019 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2020 if (rc)
2021 goto out_mmput;
2022
70335abb 2023 rc = -ENOENT;
640708a2
PE		 2024 down_read(&mm->mmap_sem);
2025 vma = find_exact_vma(mm, vm_start, vm_end);
2026 if (vma && vma->vm_file) {
2027 *path = vma->vm_file->f_path;
2028 path_get(path);
2029 rc = 0;
2030 }
2031 up_read(&mm->mmap_sem);
2032
2033out_mmput:
2034 mmput(mm);
2035out:
2036 return rc;
2037}
2038
2039struct map_files_info {
20d28cde
AD		 2040 unsigned long start;
2041 unsigned long end;
7b540d06 2042 fmode_t mode;
640708a2
PE		 2043};
2044
bdb4d100
CO		 2045/*
2046 * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the
2047 * symlinks may be used to bypass permissions on ancestor directories in the
2048 * path to the file in question.
2049 */
2050static const char *
6b255391 2051proc_map_files_get_link(struct dentry *dentry,
fceef393
AV		 2052 struct inode *inode,
2053 struct delayed_call *done)
bdb4d100
CO		 2054{
2055 if (!capable(CAP_SYS_ADMIN))
2056 return ERR_PTR(-EPERM);
2057
fceef393 2058 return proc_pid_get_link(dentry, inode, done);
bdb4d100
CO		 2059}
2060
2061/*
6b255391 2062 * Identical to proc_pid_link_inode_operations except for get_link()
bdb4d100
CO		 2063 */
2064static const struct inode_operations proc_map_files_link_inode_operations = {
2065 .readlink = proc_pid_readlink,
6b255391 2066 .get_link = proc_map_files_get_link,
bdb4d100
CO		 2067 .setattr = proc_setattr,
2068};
2069
c52a47ac 2070static int
640708a2
PE		 2071proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
2072 struct task_struct *task, const void *ptr)
2073{
7b540d06 2074 fmode_t mode = (fmode_t)(unsigned long)ptr;
640708a2
PE		 2075 struct proc_inode *ei;
2076 struct inode *inode;
2077
db978da8
AG		 2078 inode = proc_pid_make_inode(dir->i_sb, task, S_IFLNK |
2079 ((mode & FMODE_READ ) ? S_IRUSR : 0) |
2080 ((mode & FMODE_WRITE) ? S_IWUSR : 0));
640708a2 2081 if (!inode)
c52a47ac 2082 return -ENOENT;
640708a2
PE		 2083
2084 ei = PROC_I(inode);
6b255391 2085 ei->op.proc_get_link = map_files_get_link;
640708a2 2086
bdb4d100 2087 inode->i_op = &proc_map_files_link_inode_operations;
640708a2 2088 inode->i_size = 64;
640708a2
PE		 2089
2090 d_set_d_op(dentry, &tid_map_files_dentry_operations);
2091 d_add(dentry, inode);
2092
c52a47ac 2093 return 0;
640708a2
PE		 2094}
2095
2096static struct dentry *proc_map_files_lookup(struct inode *dir,
00cd8dd3 2097 struct dentry *dentry, unsigned int flags)
640708a2
PE		 2098{
2099 unsigned long vm_start, vm_end;
2100 struct vm_area_struct *vma;
2101 struct task_struct *task;
c52a47ac 2102 int result;
640708a2
PE		 2103 struct mm_struct *mm;
2104
c52a47ac 2105 result = -ENOENT;
640708a2
PE		 2106 task = get_proc_task(dir);
2107 if (!task)
2108 goto out;
2109
c52a47ac 2110 result = -EACCES;
caaee623 2111 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
640708a2
PE		 2112 goto out_put_task;
2113
c52a47ac 2114 result = -ENOENT;
640708a2 2115 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
eb94cd96 2116 goto out_put_task;
640708a2
PE		 2117
2118 mm = get_task_mm(task);
2119 if (!mm)
eb94cd96 2120 goto out_put_task;
640708a2
PE		 2121
2122 down_read(&mm->mmap_sem);
2123 vma = find_exact_vma(mm, vm_start, vm_end);
2124 if (!vma)
2125 goto out_no_vma;
2126
05f56484
SK		 2127 if (vma->vm_file)
2128 result = proc_map_files_instantiate(dir, dentry, task,
2129 (void *)(unsigned long)vma->vm_file->f_mode);
640708a2
PE		 2130
2131out_no_vma:
2132 up_read(&mm->mmap_sem);
2133 mmput(mm);
640708a2
PE		 2134out_put_task:
2135 put_task_struct(task);
2136out:
c52a47ac 2137 return ERR_PTR(result);
640708a2
PE		 2138}
2139
2140static const struct inode_operations proc_map_files_inode_operations = {
2141 .lookup = proc_map_files_lookup,
2142 .permission = proc_fd_permission,
2143 .setattr = proc_setattr,
2144};
2145
2146static int
f0c3b509 2147proc_map_files_readdir(struct file *file, struct dir_context *ctx)
640708a2 2148{
640708a2
PE		 2149 struct vm_area_struct *vma;
2150 struct task_struct *task;
2151 struct mm_struct *mm;
f0c3b509
AV		 2152 unsigned long nr_files, pos, i;
2153 struct flex_array *fa = NULL;
2154 struct map_files_info info;
2155 struct map_files_info *p;
640708a2
PE		 2156 int ret;
2157
640708a2 2158 ret = -ENOENT;
f0c3b509 2159 task = get_proc_task(file_inode(file));
640708a2
PE		 2160 if (!task)
2161 goto out;
2162
2163 ret = -EACCES;
caaee623 2164 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
640708a2
PE		 2165 goto out_put_task;
2166
2167 ret = 0;
f0c3b509
AV		 2168 if (!dir_emit_dots(file, ctx))
2169 goto out_put_task;
640708a2 2170
f0c3b509
AV		 2171 mm = get_task_mm(task);
2172 if (!mm)
2173 goto out_put_task;
2174 down_read(&mm->mmap_sem);
640708a2 2175
f0c3b509 2176 nr_files = 0;
640708a2 2177
f0c3b509
AV		 2178 /*
2179 * We need two passes here:
2180 *
2181 * 1) Collect vmas of mapped files with mmap_sem taken
2182 * 2) Release mmap_sem and instantiate entries
2183 *
2184 * otherwise we get lockdep complained, since filldir()
2185 * routine might require mmap_sem taken in might_fault().
2186 */
640708a2 2187
f0c3b509
AV		 2188 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2189 if (vma->vm_file && ++pos > ctx->pos)
2190 nr_files++;
2191 }
2192
2193 if (nr_files) {
2194 fa = flex_array_alloc(sizeof(info), nr_files,
2195 GFP_KERNEL);
2196 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2197 GFP_KERNEL)) {
2198 ret = -ENOMEM;
2199 if (fa)
2200 flex_array_free(fa);
2201 up_read(&mm->mmap_sem);
2202 mmput(mm);
2203 goto out_put_task;
640708a2 2204 }
f0c3b509
AV		 2205 for (i = 0, vma = mm->mmap, pos = 2; vma;
2206 vma = vma->vm_next) {
2207 if (!vma->vm_file)
2208 continue;
2209 if (++pos <= ctx->pos)
2210 continue;
2211
20d28cde
AD		 2212 info.start = vma->vm_start;
2213 info.end = vma->vm_end;
f0c3b509 2214 info.mode = vma->vm_file->f_mode;
f0c3b509
AV		 2215 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2216 BUG();
640708a2 2217 }
640708a2 2218 }
f0c3b509 2219 up_read(&mm->mmap_sem);
fe079a5e 2220 mmput(mm);
f0c3b509
AV		 2221
2222 for (i = 0; i < nr_files; i++) {
20d28cde
AD		 2223 char buf[4 * sizeof(long) + 2]; /* max: %lx-%lx\0 */
2224 unsigned int len;
2225
f0c3b509 2226 p = flex_array_get(fa, i);
20d28cde 2227 len = snprintf(buf, sizeof(buf), "%lx-%lx", p->start, p->end);
f0c3b509 2228 if (!proc_fill_cache(file, ctx,
20d28cde 2229 buf, len,
f0c3b509
AV		 2230 proc_map_files_instantiate,
2231 task,
2232 (void *)(unsigned long)p->mode))
2233 break;
2234 ctx->pos++;
640708a2 2235 }
f0c3b509
AV		 2236 if (fa)
2237 flex_array_free(fa);
640708a2 2238
640708a2
PE		 2239out_put_task:
2240 put_task_struct(task);
2241out:
2242 return ret;
2243}
2244
2245static const struct file_operations proc_map_files_operations = {
2246 .read = generic_read_dir,
f50752ea
AV		 2247 .iterate_shared = proc_map_files_readdir,
2248 .llseek = generic_file_llseek,
640708a2
PE		 2249};
2250
b18b6a9c 2251#if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
48f6a7a5
PE		 2252struct timers_private {
2253 struct pid *pid;
2254 struct task_struct *task;
2255 struct sighand_struct *sighand;
57b8015e 2256 struct pid_namespace *ns;
48f6a7a5
PE		 2257 unsigned long flags;
2258};
2259
2260static void *timers_start(struct seq_file *m, loff_t *pos)
2261{
2262 struct timers_private *tp = m->private;
2263
2264 tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2265 if (!tp->task)
2266 return ERR_PTR(-ESRCH);
2267
2268 tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2269 if (!tp->sighand)
2270 return ERR_PTR(-ESRCH);
2271
2272 return seq_list_start(&tp->task->signal->posix_timers, *pos);
2273}
2274
2275static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2276{
2277 struct timers_private *tp = m->private;
2278 return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2279}
2280
2281static void timers_stop(struct seq_file *m, void *v)
2282{
2283 struct timers_private *tp = m->private;
2284
2285 if (tp->sighand) {
2286 unlock_task_sighand(tp->task, &tp->flags);
2287 tp->sighand = NULL;
2288 }
2289
2290 if (tp->task) {
2291 put_task_struct(tp->task);
2292 tp->task = NULL;
2293 }
2294}
2295
2296static int show_timer(struct seq_file *m, void *v)
2297{
2298 struct k_itimer *timer;
57b8015e
PE		 2299 struct timers_private *tp = m->private;
2300 int notify;
cedbccab 2301 static const char * const nstr[] = {
57b8015e
PE		 2302 [SIGEV_SIGNAL] = "signal",
2303 [SIGEV_NONE] = "none",
2304 [SIGEV_THREAD] = "thread",
2305 };
48f6a7a5
PE		 2306
2307 timer = list_entry((struct list_head *)v, struct k_itimer, list);
57b8015e
PE		 2308 notify = timer->it_sigev_notify;
2309
48f6a7a5 2310 seq_printf(m, "ID: %d\n", timer->it_id);
ba3edf1f 2311 seq_printf(m, "signal: %d/%px\n",
25ce3191
JP		 2312 timer->sigq->info.si_signo,
2313 timer->sigq->info.si_value.sival_ptr);
57b8015e 2314 seq_printf(m, "notify: %s/%s.%d\n",
25ce3191
JP		 2315 nstr[notify & ~SIGEV_THREAD_ID],
2316 (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2317 pid_nr_ns(timer->it_pid, tp->ns));
15ef0298 2318 seq_printf(m, "ClockID: %d\n", timer->it_clock);
48f6a7a5
PE		 2319
2320 return 0;
2321}
2322
2323static const struct seq_operations proc_timers_seq_ops = {
2324 .start = timers_start,
2325 .next = timers_next,
2326 .stop = timers_stop,
2327 .show = show_timer,
2328};
2329
2330static int proc_timers_open(struct inode *inode, struct file *file)
2331{
2332 struct timers_private *tp;
2333
2334 tp = __seq_open_private(file, &proc_timers_seq_ops,
2335 sizeof(struct timers_private));
2336 if (!tp)
2337 return -ENOMEM;
2338
2339 tp->pid = proc_pid(inode);
57b8015e 2340 tp->ns = inode->i_sb->s_fs_info;
48f6a7a5
PE		 2341 return 0;
2342}
2343
2344static const struct file_operations proc_timers_operations = {
2345 .open = proc_timers_open,
2346 .read = seq_read,
2347 .llseek = seq_lseek,
2348 .release = seq_release_private,
2349};
b5946bea 2350#endif
640708a2 2351
5de23d43
JS		 2352static ssize_t timerslack_ns_write(struct file *file, const char __user *buf,
2353 size_t count, loff_t *offset)
2354{
2355 struct inode *inode = file_inode(file);
2356 struct task_struct *p;
2357 u64 slack_ns;
2358 int err;
2359
2360 err = kstrtoull_from_user(buf, count, 10, &slack_ns);
2361 if (err < 0)
2362 return err;
2363
2364 p = get_proc_task(inode);
2365 if (!p)
2366 return -ESRCH;
2367
4b2bd5fe
JS		 2368 if (p != current) {
2369 if (!capable(CAP_SYS_NICE)) {
2370 count = -EPERM;
2371 goto out;
2372 }
5de23d43 2373
4b2bd5fe
JS		 2374 err = security_task_setscheduler(p);
2375 if (err) {
2376 count = err;
2377 goto out;
2378 }
904763e1
JS		 2379 }
2380
7abbaf94
JS		 2381 task_lock(p);
2382 if (slack_ns == 0)
2383 p->timer_slack_ns = p->default_timer_slack_ns;
2384 else
2385 p->timer_slack_ns = slack_ns;
2386 task_unlock(p);
2387
2388out:
5de23d43
JS		 2389 put_task_struct(p);
2390
2391 return count;
2392}
2393
2394static int timerslack_ns_show(struct seq_file *m, void *v)
2395{
2396 struct inode *inode = m->private;
2397 struct task_struct *p;
7abbaf94 2398 int err = 0;
5de23d43
JS		 2399
2400 p = get_proc_task(inode);
2401 if (!p)
2402 return -ESRCH;
2403
4b2bd5fe 2404 if (p != current) {
5de23d43 2405
4b2bd5fe
JS		 2406 if (!capable(CAP_SYS_NICE)) {
2407 err = -EPERM;
2408 goto out;
2409 }
2410 err = security_task_getscheduler(p);
2411 if (err)
2412 goto out;
2413 }
904763e1 2414
7abbaf94
JS		 2415 task_lock(p);
2416 seq_printf(m, "%llu\n", p->timer_slack_ns);
2417 task_unlock(p);
2418
2419out:
5de23d43
JS		 2420 put_task_struct(p);
2421
2422 return err;
2423}
2424
2425static int timerslack_ns_open(struct inode *inode, struct file *filp)
2426{
2427 return single_open(filp, timerslack_ns_show, inode);
2428}
2429
2430static const struct file_operations proc_pid_set_timerslack_ns_operations = {
2431 .open = timerslack_ns_open,
2432 .read = seq_read,
2433 .write = timerslack_ns_write,
2434 .llseek = seq_lseek,
2435 .release = single_release,
2436};
2437
c52a47ac 2438static int proc_pident_instantiate(struct inode *dir,
c5141e6d 2439 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8 2440{
c5141e6d 2441 const struct pid_entry *p = ptr;
444ceed8
EB		 2442 struct inode *inode;
2443 struct proc_inode *ei;
444ceed8 2444
db978da8 2445 inode = proc_pid_make_inode(dir->i_sb, task, p->mode);
444ceed8
EB		 2446 if (!inode)
2447 goto out;
2448
2449 ei = PROC_I(inode);
444ceed8 2450 if (S_ISDIR(inode->i_mode))
bfe86848 2451 set_nlink(inode, 2); /* Use getattr to fix if necessary */
444ceed8
EB		 2452 if (p->iop)
2453 inode->i_op = p->iop;
2454 if (p->fop)
2455 inode->i_fop = p->fop;
2456 ei->op = p->op;
fb045adb 2457 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 2458 d_add(dentry, inode);
2459 /* Close the race of the process dying before we return the dentry */
0b728e19 2460 if (pid_revalidate(dentry, 0))
c52a47ac 2461 return 0;
444ceed8 2462out:
c52a47ac 2463 return -ENOENT;
444ceed8
EB		 2464}
2465
1da177e4
LT		 2466static struct dentry *proc_pident_lookup(struct inode *dir,
2467 struct dentry *dentry,
c5141e6d 2468 const struct pid_entry *ents,
7bcd6b0e 2469 unsigned int nents)
1da177e4 2470{
c52a47ac 2471 int error;
99f89551 2472 struct task_struct *task = get_proc_task(dir);
c5141e6d 2473 const struct pid_entry *p, *last;
1da177e4 2474
c52a47ac 2475 error = -ENOENT;
1da177e4 2476
99f89551
EB		 2477 if (!task)
2478 goto out_no_task;
1da177e4 2479
20cdc894
EB		 2480 /*
2481 * Yes, it does not scale. And it should not. Don't add
2482 * new entries into /proc/<tgid>/ without very good reasons.
2483 */
bac5f5d5
AD		 2484 last = &ents[nents];
2485 for (p = ents; p < last; p++) {
1da177e4
LT		 2486 if (p->len != dentry->d_name.len)
2487 continue;
2488 if (!memcmp(dentry->d_name.name, p->name, p->len))
2489 break;
2490 }
bac5f5d5 2491 if (p >= last)
1da177e4
LT		 2492 goto out;
2493
444ceed8 2494 error = proc_pident_instantiate(dir, dentry, task, p);
1da177e4 2495out:
99f89551
EB		 2496 put_task_struct(task);
2497out_no_task:
c52a47ac 2498 return ERR_PTR(error);
1da177e4
LT		 2499}
2500
f0c3b509 2501static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
c5141e6d 2502 const struct pid_entry *ents, unsigned int nents)
28a6d671 2503{
f0c3b509
AV		 2504 struct task_struct *task = get_proc_task(file_inode(file));
2505 const struct pid_entry *p;
28a6d671 2506
28a6d671 2507 if (!task)
f0c3b509 2508 return -ENOENT;
28a6d671 2509
f0c3b509
AV		 2510 if (!dir_emit_dots(file, ctx))
2511 goto out;
2512
2513 if (ctx->pos >= nents + 2)
2514 goto out;
28a6d671 2515
bac5f5d5 2516 for (p = ents + (ctx->pos - 2); p < ents + nents; p++) {
f0c3b509
AV		 2517 if (!proc_fill_cache(file, ctx, p->name, p->len,
2518 proc_pident_instantiate, task, p))
2519 break;
2520 ctx->pos++;
2521 }
28a6d671 2522out:
61a28784 2523 put_task_struct(task);
f0c3b509 2524 return 0;
1da177e4
LT		 2525}
2526
28a6d671
EB		 2527#ifdef CONFIG_SECURITY
2528static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2529 size_t count, loff_t *ppos)
2530{
496ad9aa 2531 struct inode * inode = file_inode(file);
04ff9708 2532 char *p = NULL;
28a6d671
EB		 2533 ssize_t length;
2534 struct task_struct *task = get_proc_task(inode);
2535
28a6d671 2536 if (!task)
04ff9708 2537 return -ESRCH;
28a6d671
EB		 2538
2539 length = security_getprocattr(task,
2fddfeef 2540 (char*)file->f_path.dentry->d_name.name,
04ff9708 2541 &p);
28a6d671 2542 put_task_struct(task);
04ff9708
AV		 2543 if (length > 0)
2544 length = simple_read_from_buffer(buf, count, ppos, p, length);
2545 kfree(p);
28a6d671 2546 return length;
1da177e4
LT		 2547}
2548
28a6d671
EB		 2549static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2550 size_t count, loff_t *ppos)
2551{
496ad9aa 2552 struct inode * inode = file_inode(file);
bb646cdb 2553 void *page;
28a6d671
EB		 2554 ssize_t length;
2555 struct task_struct *task = get_proc_task(inode);
2556
2557 length = -ESRCH;
2558 if (!task)
2559 goto out_no_task;
b21507e2
SS		 2560
2561 /* A task may only write its own attributes. */
2562 length = -EACCES;
2563 if (current != task)
2564 goto out;
2565
28a6d671
EB		 2566 if (count > PAGE_SIZE)
2567 count = PAGE_SIZE;
2568
2569 /* No partial writes. */
2570 length = -EINVAL;
2571 if (*ppos != 0)
2572 goto out;
2573
bb646cdb
AV		 2574 page = memdup_user(buf, count);
2575 if (IS_ERR(page)) {
2576 length = PTR_ERR(page);
28a6d671 2577 goto out;
bb646cdb 2578 }
28a6d671 2579
107db7c7 2580 /* Guard against adverse ptrace interaction */
b21507e2 2581 length = mutex_lock_interruptible(&current->signal->cred_guard_mutex);
107db7c7
DH		 2582 if (length < 0)
2583 goto out_free;
2584
b21507e2 2585 length = security_setprocattr(file->f_path.dentry->d_name.name,
bb646cdb 2586 page, count);
b21507e2 2587 mutex_unlock(&current->signal->cred_guard_mutex);
28a6d671 2588out_free:
bb646cdb 2589 kfree(page);
28a6d671
EB		 2590out:
2591 put_task_struct(task);
2592out_no_task:
2593 return length;
2594}
2595
00977a59 2596static const struct file_operations proc_pid_attr_operations = {
28a6d671
EB		 2597 .read = proc_pid_attr_read,
2598 .write = proc_pid_attr_write,
87df8424 2599 .llseek = generic_file_llseek,
28a6d671
EB		 2600};
2601
c5141e6d 2602static const struct pid_entry attr_dir_stuff[] = {
631f9c18
AD		 2603 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2604 REG("prev", S_IRUGO, proc_pid_attr_operations),
2605 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2606 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2607 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2608 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
28a6d671
EB		 2609};
2610
f0c3b509 2611static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
28a6d671 2612{
f0c3b509
AV		 2613 return proc_pident_readdir(file, ctx,
2614 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2615}
2616
00977a59 2617static const struct file_operations proc_attr_dir_operations = {
1da177e4 2618 .read = generic_read_dir,
f50752ea
AV		 2619 .iterate_shared = proc_attr_dir_readdir,
2620 .llseek = generic_file_llseek,
1da177e4
LT		 2621};
2622
72d9dcfc 2623static struct dentry *proc_attr_dir_lookup(struct inode *dir,
00cd8dd3 2624 struct dentry *dentry, unsigned int flags)
28a6d671 2625{
7bcd6b0e
EB		 2626 return proc_pident_lookup(dir, dentry,
2627 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2628}
2629
c5ef1c42 2630static const struct inode_operations proc_attr_dir_inode_operations = {
72d9dcfc 2631 .lookup = proc_attr_dir_lookup,
99f89551 2632 .getattr = pid_getattr,
6d76fa58 2633 .setattr = proc_setattr,
1da177e4
LT		 2634};
2635
28a6d671
EB		 2636#endif
2637
698ba7b5 2638#ifdef CONFIG_ELF_CORE
3cb4a0bb
KH		 2639static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2640 size_t count, loff_t *ppos)
2641{
496ad9aa 2642 struct task_struct *task = get_proc_task(file_inode(file));
3cb4a0bb
KH		 2643 struct mm_struct *mm;
2644 char buffer[PROC_NUMBUF];
2645 size_t len;
2646 int ret;
2647
2648 if (!task)
2649 return -ESRCH;
2650
2651 ret = 0;
2652 mm = get_task_mm(task);
2653 if (mm) {
2654 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2655 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2656 MMF_DUMP_FILTER_SHIFT));
2657 mmput(mm);
2658 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2659 }
2660
2661 put_task_struct(task);
2662
2663 return ret;
2664}
2665
2666static ssize_t proc_coredump_filter_write(struct file *file,
2667 const char __user *buf,
2668 size_t count,
2669 loff_t *ppos)
2670{
2671 struct task_struct *task;
2672 struct mm_struct *mm;
3cb4a0bb
KH		 2673 unsigned int val;
2674 int ret;
2675 int i;
2676 unsigned long mask;
2677
774636e1
AD		 2678 ret = kstrtouint_from_user(buf, count, 0, &val);
2679 if (ret < 0)
2680 return ret;
3cb4a0bb
KH		 2681
2682 ret = -ESRCH;
496ad9aa 2683 task = get_proc_task(file_inode(file));
3cb4a0bb
KH		 2684 if (!task)
2685 goto out_no_task;
2686
3cb4a0bb
KH		 2687 mm = get_task_mm(task);
2688 if (!mm)
2689 goto out_no_mm;
41a0c249 2690 ret = 0;
3cb4a0bb
KH		 2691
2692 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2693 if (val & mask)
2694 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2695 else
2696 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2697 }
2698
2699 mmput(mm);
2700 out_no_mm:
2701 put_task_struct(task);
2702 out_no_task:
774636e1
AD		 2703 if (ret < 0)
2704 return ret;
2705 return count;
3cb4a0bb
KH		 2706}
2707
2708static const struct file_operations proc_coredump_filter_operations = {
2709 .read = proc_coredump_filter_read,
2710 .write = proc_coredump_filter_write,
87df8424 2711 .llseek = generic_file_llseek,
3cb4a0bb
KH		 2712};
2713#endif
2714
aba76fdb 2715#ifdef CONFIG_TASK_IO_ACCOUNTING
19aadc98 2716static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
297c5d92 2717{
940389b8 2718 struct task_io_accounting acct = task->ioac;
5995477a 2719 unsigned long flags;
293eb1e7 2720 int result;
5995477a 2721
293eb1e7
VK		 2722 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2723 if (result)
2724 return result;
2725
caaee623 2726 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) {
293eb1e7
VK		 2727 result = -EACCES;
2728 goto out_unlock;
2729 }
1d1221f3 2730
5995477a
AR		 2731 if (whole && lock_task_sighand(task, &flags)) {
2732 struct task_struct *t = task;
2733
2734 task_io_accounting_add(&acct, &task->signal->ioac);
2735 while_each_thread(task, t)
2736 task_io_accounting_add(&acct, &t->ioac);
2737
2738 unlock_task_sighand(task, &flags);
297c5d92 2739 }
25ce3191
JP		 2740 seq_printf(m,
2741 "rchar: %llu\n"
2742 "wchar: %llu\n"
2743 "syscr: %llu\n"
2744 "syscw: %llu\n"
2745 "read_bytes: %llu\n"
2746 "write_bytes: %llu\n"
2747 "cancelled_write_bytes: %llu\n",
2748 (unsigned long long)acct.rchar,
2749 (unsigned long long)acct.wchar,
2750 (unsigned long long)acct.syscr,
2751 (unsigned long long)acct.syscw,
2752 (unsigned long long)acct.read_bytes,
2753 (unsigned long long)acct.write_bytes,
2754 (unsigned long long)acct.cancelled_write_bytes);
2755 result = 0;
2756
293eb1e7
VK		 2757out_unlock:
2758 mutex_unlock(&task->signal->cred_guard_mutex);
2759 return result;
297c5d92
AR		 2760}
2761
19aadc98
AD		 2762static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2763 struct pid *pid, struct task_struct *task)
297c5d92 2764{
19aadc98 2765 return do_io_accounting(task, m, 0);
aba76fdb 2766}
297c5d92 2767
19aadc98
AD		 2768static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2769 struct pid *pid, struct task_struct *task)
297c5d92 2770{
19aadc98 2771 return do_io_accounting(task, m, 1);
297c5d92
AR		 2772}
2773#endif /* CONFIG_TASK_IO_ACCOUNTING */
aba76fdb 2774
22d917d8
EB		 2775#ifdef CONFIG_USER_NS
2776static int proc_id_map_open(struct inode *inode, struct file *file,
ccf94f1b 2777 const struct seq_operations *seq_ops)
22d917d8
EB		 2778{
2779 struct user_namespace *ns = NULL;
2780 struct task_struct *task;
2781 struct seq_file *seq;
2782 int ret = -EINVAL;
2783
2784 task = get_proc_task(inode);
2785 if (task) {
2786 rcu_read_lock();
2787 ns = get_user_ns(task_cred_xxx(task, user_ns));
2788 rcu_read_unlock();
2789 put_task_struct(task);
2790 }
2791 if (!ns)
2792 goto err;
2793
2794 ret = seq_open(file, seq_ops);
2795 if (ret)
2796 goto err_put_ns;
2797
2798 seq = file->private_data;
2799 seq->private = ns;
2800
2801 return 0;
2802err_put_ns:
2803 put_user_ns(ns);
2804err:
2805 return ret;
2806}
2807
2808static int proc_id_map_release(struct inode *inode, struct file *file)
2809{
2810 struct seq_file *seq = file->private_data;
2811 struct user_namespace *ns = seq->private;
2812 put_user_ns(ns);
2813 return seq_release(inode, file);
2814}
2815
2816static int proc_uid_map_open(struct inode *inode, struct file *file)
2817{
2818 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2819}
2820
2821static int proc_gid_map_open(struct inode *inode, struct file *file)
2822{
2823 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2824}
2825
f76d207a
EB		 2826static int proc_projid_map_open(struct inode *inode, struct file *file)
2827{
2828 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2829}
2830
22d917d8
EB		 2831static const struct file_operations proc_uid_map_operations = {
2832 .open = proc_uid_map_open,
2833 .write = proc_uid_map_write,
2834 .read = seq_read,
2835 .llseek = seq_lseek,
2836 .release = proc_id_map_release,
2837};
2838
2839static const struct file_operations proc_gid_map_operations = {
2840 .open = proc_gid_map_open,
2841 .write = proc_gid_map_write,
2842 .read = seq_read,
2843 .llseek = seq_lseek,
2844 .release = proc_id_map_release,
2845};
f76d207a
EB		 2846
2847static const struct file_operations proc_projid_map_operations = {
2848 .open = proc_projid_map_open,
2849 .write = proc_projid_map_write,
2850 .read = seq_read,
2851 .llseek = seq_lseek,
2852 .release = proc_id_map_release,
2853};
9cc46516
EB		 2854
2855static int proc_setgroups_open(struct inode *inode, struct file *file)
2856{
2857 struct user_namespace *ns = NULL;
2858 struct task_struct *task;
2859 int ret;
2860
2861 ret = -ESRCH;
2862 task = get_proc_task(inode);
2863 if (task) {
2864 rcu_read_lock();
2865 ns = get_user_ns(task_cred_xxx(task, user_ns));
2866 rcu_read_unlock();
2867 put_task_struct(task);
2868 }
2869 if (!ns)
2870 goto err;
2871
2872 if (file->f_mode & FMODE_WRITE) {
2873 ret = -EACCES;
2874 if (!ns_capable(ns, CAP_SYS_ADMIN))
2875 goto err_put_ns;
2876 }
2877
2878 ret = single_open(file, &proc_setgroups_show, ns);
2879 if (ret)
2880 goto err_put_ns;
2881
2882 return 0;
2883err_put_ns:
2884 put_user_ns(ns);
2885err:
2886 return ret;
2887}
2888
2889static int proc_setgroups_release(struct inode *inode, struct file *file)
2890{
2891 struct seq_file *seq = file->private_data;
2892 struct user_namespace *ns = seq->private;
2893 int ret = single_release(inode, file);
2894 put_user_ns(ns);
2895 return ret;
2896}
2897
2898static const struct file_operations proc_setgroups_operations = {
2899 .open = proc_setgroups_open,
2900 .write = proc_setgroups_write,
2901 .read = seq_read,
2902 .llseek = seq_lseek,
2903 .release = proc_setgroups_release,
2904};
22d917d8
EB		 2905#endif /* CONFIG_USER_NS */
2906
47830723
KC		 2907static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2908 struct pid *pid, struct task_struct *task)
2909{
a9712bc1
AV		 2910 int err = lock_trace(task);
2911 if (!err) {
2912 seq_printf(m, "%08x\n", task->personality);
2913 unlock_trace(task);
2914 }
2915 return err;
47830723
KC		 2916}
2917
7c23b330
JP		 2918#ifdef CONFIG_LIVEPATCH
2919static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns,
2920 struct pid *pid, struct task_struct *task)
2921{
2922 seq_printf(m, "%d\n", task->patch_state);
2923 return 0;
2924}
2925#endif /* CONFIG_LIVEPATCH */
2926
28a6d671
EB		 2927/*
2928 * Thread groups
2929 */
00977a59 2930static const struct file_operations proc_task_operations;
c5ef1c42 2931static const struct inode_operations proc_task_inode_operations;
20cdc894 2932
c5141e6d 2933static const struct pid_entry tgid_base_stuff[] = {
631f9c18
AD		 2934 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2935 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
640708a2 2936 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
631f9c18 2937 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 2938 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
b2211a36 2939#ifdef CONFIG_NET
631f9c18 2940 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
b2211a36 2941#endif
631f9c18 2942 REG("environ", S_IRUSR, proc_environ_operations),
c5317167 2943 REG("auxv", S_IRUSR, proc_auxv_operations),
631f9c18 2944 ONE("status", S_IRUGO, proc_pid_status),
35a35046 2945 ONE("personality", S_IRUSR, proc_pid_personality),
1c963eb1 2946 ONE("limits", S_IRUGO, proc_pid_limits),
43ae34cb 2947#ifdef CONFIG_SCHED_DEBUG
631f9c18 2948 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
5091faa4
MG		 2949#endif
2950#ifdef CONFIG_SCHED_AUTOGROUP
2951 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
ebcb6734 2952#endif
4614a696 2953 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 2954#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
09d93bd6 2955 ONE("syscall", S_IRUSR, proc_pid_syscall),
43ae34cb 2956#endif
c2c0bb44 2957 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
631f9c18
AD		 2958 ONE("stat", S_IRUGO, proc_tgid_stat),
2959 ONE("statm", S_IRUGO, proc_pid_statm),
b7643757 2960 REG("maps", S_IRUGO, proc_pid_maps_operations),
28a6d671 2961#ifdef CONFIG_NUMA
b7643757 2962 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
28a6d671 2963#endif
631f9c18
AD		 2964 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2965 LNK("cwd", proc_cwd_link),
2966 LNK("root", proc_root_link),
2967 LNK("exe", proc_exe_link),
2968 REG("mounts", S_IRUGO, proc_mounts_operations),
2969 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2970 REG("mountstats", S_IRUSR, proc_mountstats_operations),
1e883281 2971#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18 2972 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
b7643757 2973 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
493b0e9d 2974 REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
32ed74a4 2975 REG("pagemap", S_IRUSR, proc_pagemap_operations),
28a6d671
EB		 2976#endif
2977#ifdef CONFIG_SECURITY
631f9c18 2978 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 2979#endif
2980#ifdef CONFIG_KALLSYMS
edfcd606 2981 ONE("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 2982#endif
2ec220e2 2983#ifdef CONFIG_STACKTRACE
35a35046 2984 ONE("stack", S_IRUSR, proc_pid_stack),
28a6d671 2985#endif
5968cece 2986#ifdef CONFIG_SCHED_INFO
f6e826ca 2987 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 2988#endif
9745512c 2989#ifdef CONFIG_LATENCYTOP
631f9c18 2990 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 2991#endif
8793d854 2992#ifdef CONFIG_PROC_PID_CPUSET
52de4779 2993 ONE("cpuset", S_IRUGO, proc_cpuset_show),
a424316c
PM		 2994#endif
2995#ifdef CONFIG_CGROUPS
006f4ac4 2996 ONE("cgroup", S_IRUGO, proc_cgroup_show),
28a6d671 2997#endif
6ba51e37 2998 ONE("oom_score", S_IRUGO, proc_oom_score),
fa0cbbf1 2999 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
a63d83f4 3000 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 3001#ifdef CONFIG_AUDITSYSCALL
631f9c18
AD		 3002 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3003 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 3004#endif
f4f154fd 3005#ifdef CONFIG_FAULT_INJECTION
631f9c18 3006 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
168c42bc 3007 REG("fail-nth", 0644, proc_fail_nth_operations),
f4f154fd 3008#endif
698ba7b5 3009#ifdef CONFIG_ELF_CORE
631f9c18 3010 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3cb4a0bb 3011#endif
aba76fdb 3012#ifdef CONFIG_TASK_IO_ACCOUNTING
19aadc98 3013 ONE("io", S_IRUSR, proc_tgid_io_accounting),
aba76fdb 3014#endif
22d917d8
EB		 3015#ifdef CONFIG_USER_NS
3016 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3017 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
f76d207a 3018 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
9cc46516 3019 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
22d917d8 3020#endif
b18b6a9c 3021#if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
48f6a7a5
PE		 3022 REG("timers", S_IRUGO, proc_timers_operations),
3023#endif
5de23d43 3024 REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations),
7c23b330
JP		 3025#ifdef CONFIG_LIVEPATCH
3026 ONE("patch_state", S_IRUSR, proc_pid_patch_state),
3027#endif
28a6d671 3028};
1da177e4 3029
f0c3b509 3030static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
1da177e4 3031{
f0c3b509
AV		 3032 return proc_pident_readdir(file, ctx,
3033 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 3034}
3035
00977a59 3036static const struct file_operations proc_tgid_base_operations = {
1da177e4 3037 .read = generic_read_dir,
f50752ea
AV		 3038 .iterate_shared = proc_tgid_base_readdir,
3039 .llseek = generic_file_llseek,
1da177e4
LT		 3040};
3041
00cd8dd3
AV		 3042static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3043{
7bcd6b0e
EB		 3044 return proc_pident_lookup(dir, dentry,
3045 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 3046}
3047
c5ef1c42 3048static const struct inode_operations proc_tgid_base_inode_operations = {
28a6d671 3049 .lookup = proc_tgid_base_lookup,
99f89551 3050 .getattr = pid_getattr,
6d76fa58 3051 .setattr = proc_setattr,
0499680a 3052 .permission = proc_pid_permission,
1da177e4 3053};
1da177e4 3054
60347f67 3055static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
1da177e4 3056{
48e6484d 3057 struct dentry *dentry, *leader, *dir;
e3912ac3 3058 char buf[10 + 1];
48e6484d
EB		 3059 struct qstr name;
3060
3061 name.name = buf;
e3912ac3 3062 name.len = snprintf(buf, sizeof(buf), "%u", pid);
4f522a24 3063 /* no ->d_hash() rejects on procfs */
60347f67 3064 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d 3065 if (dentry) {
bbd51924 3066 d_invalidate(dentry);
48e6484d
EB		 3067 dput(dentry);
3068 }
1da177e4 3069
c35a7f18
ON		 3070 if (pid == tgid)
3071 return;
3072
48e6484d 3073 name.name = buf;
e3912ac3 3074 name.len = snprintf(buf, sizeof(buf), "%u", tgid);
60347f67 3075 leader = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d
EB		 3076 if (!leader)
3077 goto out;
1da177e4 3078
48e6484d
EB		 3079 name.name = "task";
3080 name.len = strlen(name.name);
3081 dir = d_hash_and_lookup(leader, &name);
3082 if (!dir)
3083 goto out_put_leader;
3084
3085 name.name = buf;
e3912ac3 3086 name.len = snprintf(buf, sizeof(buf), "%u", pid);
48e6484d
EB		 3087 dentry = d_hash_and_lookup(dir, &name);
3088 if (dentry) {
bbd51924 3089 d_invalidate(dentry);
48e6484d 3090 dput(dentry);
1da177e4 3091 }
48e6484d
EB		 3092
3093 dput(dir);
3094out_put_leader:
3095 dput(leader);
3096out:
3097 return;
1da177e4
LT		 3098}
3099
0895e91d
RD		 3100/**
3101 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3102 * @task: task that should be flushed.
3103 *
3104 * When flushing dentries from proc, one needs to flush them from global
60347f67 3105 * proc (proc_mnt) and from all the namespaces' procs this task was seen
0895e91d
RD		 3106 * in. This call is supposed to do all of this job.
3107 *
3108 * Looks in the dcache for
3109 * /proc/@pid
3110 * /proc/@tgid/task/@pid
3111 * if either directory is present flushes it and all of it'ts children
3112 * from the dcache.
3113 *
3114 * It is safe and reasonable to cache /proc entries for a task until
3115 * that task exits. After that they just clog up the dcache with
3116 * useless entries, possibly causing useful dcache entries to be
3117 * flushed instead. This routine is proved to flush those useless
3118 * dcache entries at process exit time.
3119 *
3120 * NOTE: This routine is just an optimization so it does not guarantee
3121 * that no dcache entries will exist at process exit time it
3122 * just makes it very unlikely that any will persist.
60347f67
PE		 3123 */
3124
3125void proc_flush_task(struct task_struct *task)
3126{
9fcc2d15 3127 int i;
9b4d1cbe 3128 struct pid *pid, *tgid;
130f77ec
PE		 3129 struct upid *upid;
3130
130f77ec 3131 pid = task_pid(task);
9b4d1cbe 3132 tgid = task_tgid(task);
130f77ec 3133
9fcc2d15 3134 for (i = 0; i <= pid->level; i++) {
130f77ec
PE		 3135 upid = &pid->numbers[i];
3136 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
9b4d1cbe 3137 tgid->numbers[i].nr);
130f77ec 3138 }
60347f67
PE		 3139}
3140
c52a47ac
AV		 3141static int proc_pid_instantiate(struct inode *dir,
3142 struct dentry * dentry,
3143 struct task_struct *task, const void *ptr)
444ceed8 3144{
444ceed8
EB		 3145 struct inode *inode;
3146
db978da8 3147 inode = proc_pid_make_inode(dir->i_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
444ceed8
EB		 3148 if (!inode)
3149 goto out;
3150
444ceed8
EB		 3151 inode->i_op = &proc_tgid_base_inode_operations;
3152 inode->i_fop = &proc_tgid_base_operations;
3153 inode->i_flags|=S_IMMUTABLE;
aed54175 3154
1270dd8d 3155 set_nlink(inode, nlink_tgid);
444ceed8 3156
fb045adb 3157 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 3158
3159 d_add(dentry, inode);
3160 /* Close the race of the process dying before we return the dentry */
0b728e19 3161 if (pid_revalidate(dentry, 0))
c52a47ac 3162 return 0;
444ceed8 3163out:
c52a47ac 3164 return -ENOENT;
444ceed8
EB		 3165}
3166
00cd8dd3 3167struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
1da177e4 3168{
335eb531 3169 int result = -ENOENT;
1da177e4 3170 struct task_struct *task;
1da177e4 3171 unsigned tgid;
b488893a 3172 struct pid_namespace *ns;
1da177e4 3173
dbcdb504 3174 tgid = name_to_int(&dentry->d_name);
1da177e4
LT		 3175 if (tgid == ~0U)
3176 goto out;
3177
b488893a 3178 ns = dentry->d_sb->s_fs_info;
de758734 3179 rcu_read_lock();
b488893a 3180 task = find_task_by_pid_ns(tgid, ns);
1da177e4
LT		 3181 if (task)
3182 get_task_struct(task);
de758734 3183 rcu_read_unlock();
1da177e4
LT		 3184 if (!task)
3185 goto out;
3186
444ceed8 3187 result = proc_pid_instantiate(dir, dentry, task, NULL);
1da177e4 3188 put_task_struct(task);
1da177e4 3189out:
c52a47ac 3190 return ERR_PTR(result);
1da177e4
LT		 3191}
3192
1da177e4 3193/*
0804ef4b 3194 * Find the first task with tgid >= tgid
0bc58a91 3195 *
1da177e4 3196 */
19fd4bb2
EB		 3197struct tgid_iter {
3198 unsigned int tgid;
0804ef4b 3199 struct task_struct *task;
19fd4bb2
EB		 3200};
3201static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3202{
0804ef4b 3203 struct pid *pid;
1da177e4 3204
19fd4bb2
EB		 3205 if (iter.task)
3206 put_task_struct(iter.task);
454cc105 3207 rcu_read_lock();
0804ef4b 3208retry:
19fd4bb2
EB		 3209 iter.task = NULL;
3210 pid = find_ge_pid(iter.tgid, ns);
0804ef4b 3211 if (pid) {
19fd4bb2
EB		 3212 iter.tgid = pid_nr_ns(pid, ns);
3213 iter.task = pid_task(pid, PIDTYPE_PID);
0804ef4b
EB		 3214 /* What we to know is if the pid we have find is the
3215 * pid of a thread_group_leader. Testing for task
3216 * being a thread_group_leader is the obvious thing
3217 * todo but there is a window when it fails, due to
3218 * the pid transfer logic in de_thread.
3219 *
3220 * So we perform the straight forward test of seeing
3221 * if the pid we have found is the pid of a thread
3222 * group leader, and don't worry if the task we have
3223 * found doesn't happen to be a thread group leader.
3224 * As we don't care in the case of readdir.
3225 */
19fd4bb2
EB		 3226 if (!iter.task || !has_group_leader_pid(iter.task)) {
3227 iter.tgid += 1;
0804ef4b 3228 goto retry;
19fd4bb2
EB		 3229 }
3230 get_task_struct(iter.task);
0bc58a91 3231 }
454cc105 3232 rcu_read_unlock();
19fd4bb2 3233 return iter;
1da177e4
LT		 3234}
3235
0097875b 3236#define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
0804ef4b 3237
1da177e4 3238/* for the /proc/ directory itself, after non-process stuff has been done */
f0c3b509 3239int proc_pid_readdir(struct file *file, struct dir_context *ctx)
1da177e4 3240{
19fd4bb2 3241 struct tgid_iter iter;
3aa3377f 3242 struct pid_namespace *ns = file_inode(file)->i_sb->s_fs_info;
f0c3b509 3243 loff_t pos = ctx->pos;
1da177e4 3244
021ada7d 3245 if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
f0c3b509 3246 return 0;
1da177e4 3247
0097875b 3248 if (pos == TGID_OFFSET - 2) {
2b0143b5 3249 struct inode *inode = d_inode(ns->proc_self);
db963164 3250 if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
f0c3b509 3251 return 0;
0097875b
EB		 3252 ctx->pos = pos = pos + 1;
3253 }
3254 if (pos == TGID_OFFSET - 1) {
2b0143b5 3255 struct inode *inode = d_inode(ns->proc_thread_self);
0097875b
EB		 3256 if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3257 return 0;
3258 ctx->pos = pos = pos + 1;
021ada7d 3259 }
0097875b 3260 iter.tgid = pos - TGID_OFFSET;
19fd4bb2 3261 iter.task = NULL;
19fd4bb2
EB		 3262 for (iter = next_tgid(ns, iter);
3263 iter.task;
3264 iter.tgid += 1, iter = next_tgid(ns, iter)) {
e3912ac3 3265 char name[10 + 1];
f0c3b509 3266 int len;
3ba4bcee
ED		 3267
3268 cond_resched();
796f571b 3269 if (!has_pid_permissions(ns, iter.task, HIDEPID_INVISIBLE))
f0c3b509 3270 continue;
0499680a 3271
e3912ac3 3272 len = snprintf(name, sizeof(name), "%u", iter.tgid);
f0c3b509
AV		 3273 ctx->pos = iter.tgid + TGID_OFFSET;
3274 if (!proc_fill_cache(file, ctx, name, len,
3275 proc_pid_instantiate, iter.task, NULL)) {
19fd4bb2 3276 put_task_struct(iter.task);
f0c3b509 3277 return 0;
1da177e4 3278 }
0bc58a91 3279 }
f0c3b509 3280 ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
0bc58a91
EB		 3281 return 0;
3282}
1da177e4 3283
1b3044e3
JD		 3284/*
3285 * proc_tid_comm_permission is a special permission function exclusively
3286 * used for the node /proc/<pid>/task/<tid>/comm.
3287 * It bypasses generic permission checks in the case where a task of the same
3288 * task group attempts to access the node.
3289 * The rationale behind this is that glibc and bionic access this node for
3290 * cross thread naming (pthread_set/getname_np(!self)). However, if
3291 * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3292 * which locks out the cross thread naming implementation.
3293 * This function makes sure that the node is always accessible for members of
3294 * same thread group.
3295 */
3296static int proc_tid_comm_permission(struct inode *inode, int mask)
3297{
3298 bool is_same_tgroup;
3299 struct task_struct *task;
3300
3301 task = get_proc_task(inode);
3302 if (!task)
3303 return -ESRCH;
3304 is_same_tgroup = same_thread_group(current, task);
3305 put_task_struct(task);
3306
3307 if (likely(is_same_tgroup && !(mask & MAY_EXEC))) {
3308 /* This file (/proc/<pid>/task/<tid>/comm) can always be
3309 * read or written by the members of the corresponding
3310 * thread group.
3311 */
3312 return 0;
3313 }
3314
3315 return generic_permission(inode, mask);
3316}
3317
3318static const struct inode_operations proc_tid_comm_inode_operations = {
3319 .permission = proc_tid_comm_permission,
3320};
3321
28a6d671
EB		 3322/*
3323 * Tasks
3324 */
c5141e6d 3325static const struct pid_entry tid_base_stuff[] = {
631f9c18 3326 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3835541d 3327 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 3328 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
6ba8ed79
EB		 3329#ifdef CONFIG_NET
3330 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3331#endif
631f9c18 3332 REG("environ", S_IRUSR, proc_environ_operations),
c5317167 3333 REG("auxv", S_IRUSR, proc_auxv_operations),
631f9c18 3334 ONE("status", S_IRUGO, proc_pid_status),
35a35046 3335 ONE("personality", S_IRUSR, proc_pid_personality),
1c963eb1 3336 ONE("limits", S_IRUGO, proc_pid_limits),
43ae34cb 3337#ifdef CONFIG_SCHED_DEBUG
631f9c18 3338 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
ebcb6734 3339#endif
1b3044e3
JD		 3340 NOD("comm", S_IFREG|S_IRUGO|S_IWUSR,
3341 &proc_tid_comm_inode_operations,
3342 &proc_pid_set_comm_operations, {}),
ebcb6734 3343#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
09d93bd6 3344 ONE("syscall", S_IRUSR, proc_pid_syscall),
43ae34cb 3345#endif
c2c0bb44 3346 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
631f9c18
AD		 3347 ONE("stat", S_IRUGO, proc_tid_stat),
3348 ONE("statm", S_IRUGO, proc_pid_statm),
b7643757 3349 REG("maps", S_IRUGO, proc_tid_maps_operations),
2e13ba54 3350#ifdef CONFIG_PROC_CHILDREN
81841161
CG		 3351 REG("children", S_IRUGO, proc_tid_children_operations),
3352#endif
28a6d671 3353#ifdef CONFIG_NUMA
b7643757 3354 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
28a6d671 3355#endif
631f9c18
AD		 3356 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3357 LNK("cwd", proc_cwd_link),
3358 LNK("root", proc_root_link),
3359 LNK("exe", proc_exe_link),
3360 REG("mounts", S_IRUGO, proc_mounts_operations),
3361 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
1e883281 3362#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18 3363 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
b7643757 3364 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
493b0e9d 3365 REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
32ed74a4 3366 REG("pagemap", S_IRUSR, proc_pagemap_operations),
28a6d671
EB		 3367#endif
3368#ifdef CONFIG_SECURITY
631f9c18 3369 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 3370#endif
3371#ifdef CONFIG_KALLSYMS
edfcd606 3372 ONE("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 3373#endif
2ec220e2 3374#ifdef CONFIG_STACKTRACE
35a35046 3375 ONE("stack", S_IRUSR, proc_pid_stack),
28a6d671 3376#endif
5968cece 3377#ifdef CONFIG_SCHED_INFO
f6e826ca 3378 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 3379#endif
9745512c 3380#ifdef CONFIG_LATENCYTOP
631f9c18 3381 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 3382#endif
8793d854 3383#ifdef CONFIG_PROC_PID_CPUSET
52de4779 3384 ONE("cpuset", S_IRUGO, proc_cpuset_show),
a424316c
PM		 3385#endif
3386#ifdef CONFIG_CGROUPS
006f4ac4 3387 ONE("cgroup", S_IRUGO, proc_cgroup_show),
28a6d671 3388#endif
6ba51e37 3389 ONE("oom_score", S_IRUGO, proc_oom_score),
fa0cbbf1 3390 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
a63d83f4 3391 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 3392#ifdef CONFIG_AUDITSYSCALL
631f9c18 3393 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
26ec3c64 3394 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 3395#endif
f4f154fd 3396#ifdef CONFIG_FAULT_INJECTION
631f9c18 3397 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
1203c8e6 3398 REG("fail-nth", 0644, proc_fail_nth_operations),
f4f154fd 3399#endif
297c5d92 3400#ifdef CONFIG_TASK_IO_ACCOUNTING
19aadc98 3401 ONE("io", S_IRUSR, proc_tid_io_accounting),
297c5d92 3402#endif
22d917d8
EB		 3403#ifdef CONFIG_USER_NS
3404 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3405 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
f76d207a 3406 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
9cc46516 3407 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
22d917d8 3408#endif
7c23b330
JP		 3409#ifdef CONFIG_LIVEPATCH
3410 ONE("patch_state", S_IRUSR, proc_pid_patch_state),
3411#endif
28a6d671
EB		 3412};
3413
f0c3b509 3414static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
28a6d671 3415{
f0c3b509
AV		 3416 return proc_pident_readdir(file, ctx,
3417 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
28a6d671
EB		 3418}
3419
00cd8dd3
AV		 3420static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3421{
7bcd6b0e
EB		 3422 return proc_pident_lookup(dir, dentry,
3423 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
28a6d671
EB		 3424}
3425
00977a59 3426static const struct file_operations proc_tid_base_operations = {
28a6d671 3427 .read = generic_read_dir,
f50752ea
AV		 3428 .iterate_shared = proc_tid_base_readdir,
3429 .llseek = generic_file_llseek,
28a6d671
EB		 3430};
3431
c5ef1c42 3432static const struct inode_operations proc_tid_base_inode_operations = {
28a6d671
EB		 3433 .lookup = proc_tid_base_lookup,
3434 .getattr = pid_getattr,
3435 .setattr = proc_setattr,
3436};
3437
c52a47ac 3438static int proc_task_instantiate(struct inode *dir,
c5141e6d 3439 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8 3440{
444ceed8 3441 struct inode *inode;
db978da8 3442 inode = proc_pid_make_inode(dir->i_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
444ceed8
EB		 3443
3444 if (!inode)
3445 goto out;
444ceed8
EB		 3446 inode->i_op = &proc_tid_base_inode_operations;
3447 inode->i_fop = &proc_tid_base_operations;
3448 inode->i_flags|=S_IMMUTABLE;
aed54175 3449
1270dd8d 3450 set_nlink(inode, nlink_tid);
444ceed8 3451
fb045adb 3452 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 3453
3454 d_add(dentry, inode);
3455 /* Close the race of the process dying before we return the dentry */
0b728e19 3456 if (pid_revalidate(dentry, 0))
c52a47ac 3457 return 0;
444ceed8 3458out:
c52a47ac 3459 return -ENOENT;
444ceed8
EB		 3460}
3461
00cd8dd3 3462static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
28a6d671 3463{
c52a47ac 3464 int result = -ENOENT;
28a6d671
EB		 3465 struct task_struct *task;
3466 struct task_struct *leader = get_proc_task(dir);
28a6d671 3467 unsigned tid;
b488893a 3468 struct pid_namespace *ns;
28a6d671
EB		 3469
3470 if (!leader)
3471 goto out_no_task;
3472
dbcdb504 3473 tid = name_to_int(&dentry->d_name);
28a6d671
EB		 3474 if (tid == ~0U)
3475 goto out;
3476
b488893a 3477 ns = dentry->d_sb->s_fs_info;
28a6d671 3478 rcu_read_lock();
b488893a 3479 task = find_task_by_pid_ns(tid, ns);
28a6d671
EB		 3480 if (task)
3481 get_task_struct(task);
3482 rcu_read_unlock();
3483 if (!task)
3484 goto out;
bac0abd6 3485 if (!same_thread_group(leader, task))
28a6d671
EB		 3486 goto out_drop_task;
3487
444ceed8 3488 result = proc_task_instantiate(dir, dentry, task, NULL);
28a6d671
EB		 3489out_drop_task:
3490 put_task_struct(task);
3491out:
3492 put_task_struct(leader);
3493out_no_task:
c52a47ac 3494 return ERR_PTR(result);
28a6d671
EB		 3495}
3496
0bc58a91
EB		 3497/*
3498 * Find the first tid of a thread group to return to user space.
3499 *
3500 * Usually this is just the thread group leader, but if the users
3501 * buffer was too small or there was a seek into the middle of the
3502 * directory we have more work todo.
3503 *
3504 * In the case of a short read we start with find_task_by_pid.
3505 *
3506 * In the case of a seek we start with the leader and walk nr
3507 * threads past it.
3508 */
9f6e963f
ON		 3509static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3510 struct pid_namespace *ns)
0bc58a91 3511{
d855a4b7 3512 struct task_struct *pos, *task;
9f6e963f
ON		 3513 unsigned long nr = f_pos;
3514
3515 if (nr != f_pos) /* 32bit overflow? */
3516 return NULL;
1da177e4 3517
cc288738 3518 rcu_read_lock();
d855a4b7
ON		 3519 task = pid_task(pid, PIDTYPE_PID);
3520 if (!task)
3521 goto fail;
3522
3523 /* Attempt to start with the tid of a thread */
9f6e963f 3524 if (tid && nr) {
b488893a 3525 pos = find_task_by_pid_ns(tid, ns);
d855a4b7 3526 if (pos && same_thread_group(pos, task))
a872ff0c 3527 goto found;
0bc58a91 3528 }
1da177e4 3529
0bc58a91 3530 /* If nr exceeds the number of threads there is nothing todo */
9f6e963f 3531 if (nr >= get_nr_threads(task))
c986c14a 3532 goto fail;
1da177e4 3533
a872ff0c
ON		 3534 /* If we haven't found our starting place yet start
3535 * with the leader and walk nr threads forward.
0bc58a91 3536 */
d855a4b7 3537 pos = task = task->group_leader;
c986c14a 3538 do {
9f6e963f 3539 if (!nr--)
c986c14a 3540 goto found;
d855a4b7 3541 } while_each_thread(task, pos);
c986c14a
ON		 3542fail:
3543 pos = NULL;
3544 goto out;
a872ff0c
ON		 3545found:
3546 get_task_struct(pos);
3547out:
cc288738 3548 rcu_read_unlock();
0bc58a91
EB		 3549 return pos;
3550}
3551
3552/*
3553 * Find the next thread in the thread list.
3554 * Return NULL if there is an error or no next thread.
3555 *
3556 * The reference to the input task_struct is released.
3557 */
3558static struct task_struct *next_tid(struct task_struct *start)
3559{
c1df7fb8 3560 struct task_struct *pos = NULL;
cc288738 3561 rcu_read_lock();
c1df7fb8 3562 if (pid_alive(start)) {
0bc58a91 3563 pos = next_thread(start);
c1df7fb8
ON		 3564 if (thread_group_leader(pos))
3565 pos = NULL;
3566 else
3567 get_task_struct(pos);
3568 }
cc288738 3569 rcu_read_unlock();
0bc58a91
EB		 3570 put_task_struct(start);
3571 return pos;
1da177e4
LT		 3572}
3573
3574/* for the /proc/TGID/task/ directories */
f0c3b509 3575static int proc_task_readdir(struct file *file, struct dir_context *ctx)
1da177e4 3576{
d855a4b7
ON		 3577 struct inode *inode = file_inode(file);
3578 struct task_struct *task;
b488893a 3579 struct pid_namespace *ns;
f0c3b509 3580 int tid;
1da177e4 3581
d855a4b7 3582 if (proc_inode_is_dead(inode))
f0c3b509 3583 return -ENOENT;
1da177e4 3584
f0c3b509 3585 if (!dir_emit_dots(file, ctx))
d855a4b7 3586 return 0;
1da177e4 3587
0bc58a91
EB		 3588 /* f_version caches the tgid value that the last readdir call couldn't
3589 * return. lseek aka telldir automagically resets f_version to 0.
3590 */
3aa3377f 3591 ns = inode->i_sb->s_fs_info;
f0c3b509
AV		 3592 tid = (int)file->f_version;
3593 file->f_version = 0;
d855a4b7 3594 for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
0bc58a91 3595 task;
f0c3b509 3596 task = next_tid(task), ctx->pos++) {
e3912ac3 3597 char name[10 + 1];
f0c3b509 3598 int len;
b488893a 3599 tid = task_pid_nr_ns(task, ns);
e3912ac3 3600 len = snprintf(name, sizeof(name), "%u", tid);
f0c3b509
AV		 3601 if (!proc_fill_cache(file, ctx, name, len,
3602 proc_task_instantiate, task, NULL)) {
0bc58a91
EB		 3603 /* returning this tgid failed, save it as the first
3604 * pid for the next readir call */
f0c3b509 3605 file->f_version = (u64)tid;
0bc58a91 3606 put_task_struct(task);
1da177e4 3607 break;
0bc58a91 3608 }
1da177e4 3609 }
d855a4b7 3610
f0c3b509 3611 return 0;
1da177e4 3612}
6e66b52b 3613
a528d35e
DH		 3614static int proc_task_getattr(const struct path *path, struct kstat *stat,
3615 u32 request_mask, unsigned int query_flags)
6e66b52b 3616{
a528d35e 3617 struct inode *inode = d_inode(path->dentry);
99f89551 3618 struct task_struct *p = get_proc_task(inode);
6e66b52b
EB		 3619 generic_fillattr(inode, stat);
3620
99f89551 3621 if (p) {
99f89551 3622 stat->nlink += get_nr_threads(p);
99f89551 3623 put_task_struct(p);
6e66b52b
EB		 3624 }
3625
3626 return 0;
3627}
28a6d671 3628
c5ef1c42 3629static const struct inode_operations proc_task_inode_operations = {
28a6d671
EB		 3630 .lookup = proc_task_lookup,
3631 .getattr = proc_task_getattr,
3632 .setattr = proc_setattr,
0499680a 3633 .permission = proc_pid_permission,
28a6d671
EB		 3634};
3635
00977a59 3636static const struct file_operations proc_task_operations = {
28a6d671 3637 .read = generic_read_dir,
f50752ea
AV		 3638 .iterate_shared = proc_task_readdir,
3639 .llseek = generic_file_llseek,
28a6d671 3640};
1270dd8d
AD		 3641
3642void __init set_proc_pid_nlink(void)
3643{
3644 nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3645 nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3646}
Linux 6.x block layer and io_uring tree(s)