projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
proc: faster /proc/cmdline
[linux-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
1696 /* Default to the tasks effective ownership */
1697 rcu_read_lock();
1698 cred = __task_cred(task);
1699 uid = cred->euid;
1700 gid = cred->egid;
1701 rcu_read_unlock();
1702
1703 /*
1704 * Before the /proc/pid/status file was created the only way to read
1705 * the effective uid of a /process was to stat /proc/pid. Reading
1706 * /proc/pid/status is slow enough that procps and other packages
1707 * kept stating /proc/pid. To keep the rules in /proc simple I have
1708 * made this apply to all per process world readable and executable
1709 * directories.
1710 */
1711 if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) {
1712 struct mm_struct *mm;
1713 task_lock(task);
1714 mm = task->mm;
1715 /* Make non-dumpable tasks owned by some root */
1716 if (mm) {
1717 if (get_dumpable(mm) != SUID_DUMP_USER) {
1718 struct user_namespace *user_ns = mm->user_ns;
1719
1720 uid = make_kuid(user_ns, 0);
1721 if (!uid_valid(uid))
1722 uid = GLOBAL_ROOT_UID;
1723
1724 gid = make_kgid(user_ns, 0);
1725 if (!gid_valid(gid))
1726 gid = GLOBAL_ROOT_GID;
1727 }
1728 } else {
1729 uid = GLOBAL_ROOT_UID;
1730 gid = GLOBAL_ROOT_GID;
1731 }
1732 task_unlock(task);
1733 }
1734 *ruid = uid;
1735 *rgid = gid;
1736}
1737
db978da8
AG		 1738struct inode *proc_pid_make_inode(struct super_block * sb,
1739 struct task_struct *task, umode_t mode)
28a6d671
EB		 1740{
1741 struct inode * inode;
1742 struct proc_inode *ei;
1da177e4 1743
28a6d671 1744 /* We need a new inode */
1da177e4 1745
28a6d671
EB		 1746 inode = new_inode(sb);
1747 if (!inode)
1748 goto out;
1749
1750 /* Common stuff */
1751 ei = PROC_I(inode);
db978da8 1752 inode->i_mode = mode;
85fe4025 1753 inode->i_ino = get_next_ino();
078cd827 1754 inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode);
28a6d671
EB		 1755 inode->i_op = &proc_def_inode_operations;
1756
1757 /*
1758 * grab the reference to task.
1759 */
1a657f78 1760 ei->pid = get_task_pid(task, PIDTYPE_PID);
28a6d671
EB		 1761 if (!ei->pid)
1762 goto out_unlock;
1763
68eb94f1 1764 task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
28a6d671
EB		 1765 security_task_to_inode(task, inode);
1766
1da177e4 1767out:
28a6d671
EB		 1768 return inode;
1769
1770out_unlock:
1771 iput(inode);
1772 return NULL;
1da177e4
LT		 1773}
1774
a528d35e
DH		 1775int pid_getattr(const struct path *path, struct kstat *stat,
1776 u32 request_mask, unsigned int query_flags)
1da177e4 1777{
a528d35e 1778 struct inode *inode = d_inode(path->dentry);
28a6d671 1779 struct task_struct *task;
a528d35e 1780 struct pid_namespace *pid = path->dentry->d_sb->s_fs_info;
c69e8d9c 1781
28a6d671 1782 generic_fillattr(inode, stat);
1da177e4 1783
28a6d671 1784 rcu_read_lock();
dcb0f222
EB		 1785 stat->uid = GLOBAL_ROOT_UID;
1786 stat->gid = GLOBAL_ROOT_GID;
28a6d671
EB		 1787 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1788 if (task) {
796f571b 1789 if (!has_pid_permissions(pid, task, HIDEPID_INVISIBLE)) {
0499680a
VK		 1790 rcu_read_unlock();
1791 /*
1792 * This doesn't prevent learning whether PID exists,
1793 * it only makes getattr() consistent with readdir().
1794 */
1795 return -ENOENT;
1796 }
68eb94f1 1797 task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid);
1da177e4 1798 }
28a6d671 1799 rcu_read_unlock();
d6e71144 1800 return 0;
1da177e4
LT		 1801}
1802
1da177e4
LT		 1803/* dentry stuff */
1804
1805/*
1806 * Exceptional case: normally we are not allowed to unhash a busy
1807 * directory. In this case, however, we can do it - no aliasing problems
1808 * due to the way we treat inodes.
1809 *
1810 * Rewrite the inode's ownerships here because the owning task may have
1811 * performed a setuid(), etc.
99f89551 1812 *
1da177e4 1813 */
0b728e19 1814int pid_revalidate(struct dentry *dentry, unsigned int flags)
1da177e4 1815{
34286d66
NP		 1816 struct inode *inode;
1817 struct task_struct *task;
c69e8d9c 1818
0b728e19 1819 if (flags & LOOKUP_RCU)
34286d66
NP		 1820 return -ECHILD;
1821
2b0143b5 1822 inode = d_inode(dentry);
34286d66
NP		 1823 task = get_proc_task(inode);
1824
99f89551 1825 if (task) {
68eb94f1
EB		 1826 task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid);
1827
9ee8ab9f 1828 inode->i_mode &= ~(S_ISUID | S_ISGID);
1da177e4 1829 security_task_to_inode(task, inode);
99f89551 1830 put_task_struct(task);
1da177e4
LT		 1831 return 1;
1832 }
1da177e4
LT		 1833 return 0;
1834}
1835
d855a4b7
ON		 1836static inline bool proc_inode_is_dead(struct inode *inode)
1837{
1838 return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
1839}
1840
1dd704b6
DH		 1841int pid_delete_dentry(const struct dentry *dentry)
1842{
1843 /* Is the task we represent dead?
1844 * If so, then don't put the dentry on the lru list,
1845 * kill it immediately.
1846 */
2b0143b5 1847 return proc_inode_is_dead(d_inode(dentry));
1dd704b6
DH		 1848}
1849
6b4e306a 1850const struct dentry_operations pid_dentry_operations =
28a6d671
EB		 1851{
1852 .d_revalidate = pid_revalidate,
1853 .d_delete = pid_delete_dentry,
1854};
1855
1856/* Lookups */
1857
1c0d04c9
EB		 1858/*
1859 * Fill a directory entry.
1860 *
1861 * If possible create the dcache entry and derive our inode number and
1862 * file type from dcache entry.
1863 *
1864 * Since all of the proc inode numbers are dynamically generated, the inode
1865 * numbers do not exist until the inode is cache. This means creating the
1866 * the dcache entry in readdir is necessary to keep the inode numbers
1867 * reported by readdir in sync with the inode numbers reported
1868 * by stat.
1869 */
f0c3b509 1870bool proc_fill_cache(struct file *file, struct dir_context *ctx,
6b4e306a 1871 const char *name, int len,
c5141e6d 1872 instantiate_t instantiate, struct task_struct *task, const void *ptr)
61a28784 1873{
f0c3b509 1874 struct dentry *child, *dir = file->f_path.dentry;
1df98b8b 1875 struct qstr qname = QSTR_INIT(name, len);
61a28784 1876 struct inode *inode;
1df98b8b
AV		 1877 unsigned type;
1878 ino_t ino;
61a28784 1879
1df98b8b 1880 child = d_hash_and_lookup(dir, &qname);
61a28784 1881 if (!child) {
3781764b
AV		 1882 DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq);
1883 child = d_alloc_parallel(dir, &qname, &wq);
1884 if (IS_ERR(child))
1df98b8b 1885 goto end_instantiate;
3781764b
AV		 1886 if (d_in_lookup(child)) {
1887 int err = instantiate(d_inode(dir), child, task, ptr);
1888 d_lookup_done(child);
1889 if (err < 0) {
1890 dput(child);
1891 goto end_instantiate;
1892 }
61a28784
EB		 1893 }
1894 }
2b0143b5 1895 inode = d_inode(child);
147ce699
AV		 1896 ino = inode->i_ino;
1897 type = inode->i_mode >> 12;
61a28784 1898 dput(child);
f0c3b509 1899 return dir_emit(ctx, name, len, ino, type);
1df98b8b
AV		 1900
1901end_instantiate:
1902 return dir_emit(ctx, name, len, 1, DT_UNKNOWN);
61a28784
EB		 1903}
1904
640708a2
PE		 1905/*
1906 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1907 * which represent vma start and end addresses.
1908 */
1909static int dname_to_vma_addr(struct dentry *dentry,
1910 unsigned long *start, unsigned long *end)
1911{
ac7f1061
AD		 1912 const char *str = dentry->d_name.name;
1913 unsigned long long sval, eval;
1914 unsigned int len;
1915
35318db5
AD		 1916 if (str[0] == '0' && str[1] != '-')
1917 return -EINVAL;
ac7f1061
AD		 1918 len = _parse_integer(str, 16, &sval);
1919 if (len & KSTRTOX_OVERFLOW)
1920 return -EINVAL;
1921 if (sval != (unsigned long)sval)
1922 return -EINVAL;
1923 str += len;
1924
1925 if (*str != '-')
640708a2 1926 return -EINVAL;
ac7f1061
AD		 1927 str++;
1928
35318db5
AD		 1929 if (str[0] == '0' && str[1])
1930 return -EINVAL;
ac7f1061
AD		 1931 len = _parse_integer(str, 16, &eval);
1932 if (len & KSTRTOX_OVERFLOW)
1933 return -EINVAL;
1934 if (eval != (unsigned long)eval)
1935 return -EINVAL;
1936 str += len;
1937
1938 if (*str != '\0')
1939 return -EINVAL;
1940
1941 *start = sval;
1942 *end = eval;
640708a2
PE		 1943
1944 return 0;
1945}
1946
0b728e19 1947static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
640708a2
PE		 1948{
1949 unsigned long vm_start, vm_end;
1950 bool exact_vma_exists = false;
1951 struct mm_struct *mm = NULL;
1952 struct task_struct *task;
640708a2
PE		 1953 struct inode *inode;
1954 int status = 0;
1955
0b728e19 1956 if (flags & LOOKUP_RCU)
640708a2
PE		 1957 return -ECHILD;
1958
2b0143b5 1959 inode = d_inode(dentry);
640708a2
PE		 1960 task = get_proc_task(inode);
1961 if (!task)
1962 goto out_notask;
1963
caaee623 1964 mm = mm_access(task, PTRACE_MODE_READ_FSCREDS);
2344bec7 1965 if (IS_ERR_OR_NULL(mm))
640708a2
PE		 1966 goto out;
1967
1968 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1969 down_read(&mm->mmap_sem);
1970 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1971 up_read(&mm->mmap_sem);
1972 }
1973
1974 mmput(mm);
1975
1976 if (exact_vma_exists) {
68eb94f1
EB		 1977 task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
1978
640708a2
PE		 1979 security_task_to_inode(task, inode);
1980 status = 1;
1981 }
1982
1983out:
1984 put_task_struct(task);
1985
1986out_notask:
640708a2
PE		 1987 return status;
1988}
1989
1990static const struct dentry_operations tid_map_files_dentry_operations = {
1991 .d_revalidate = map_files_d_revalidate,
1992 .d_delete = pid_delete_dentry,
1993};
1994
6b255391 1995static int map_files_get_link(struct dentry *dentry, struct path *path)
640708a2
PE		 1996{
1997 unsigned long vm_start, vm_end;
1998 struct vm_area_struct *vma;
1999 struct task_struct *task;
2000 struct mm_struct *mm;
2001 int rc;
2002
2003 rc = -ENOENT;
2b0143b5 2004 task = get_proc_task(d_inode(dentry));
640708a2
PE		 2005 if (!task)
2006 goto out;
2007
2008 mm = get_task_mm(task);
2009 put_task_struct(task);
2010 if (!mm)
2011 goto out;
2012
2013 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2014 if (rc)
2015 goto out_mmput;
2016
70335abb 2017 rc = -ENOENT;
640708a2
PE		 2018 down_read(&mm->mmap_sem);
2019 vma = find_exact_vma(mm, vm_start, vm_end);
2020 if (vma && vma->vm_file) {
2021 *path = vma->vm_file->f_path;
2022 path_get(path);
2023 rc = 0;
2024 }
2025 up_read(&mm->mmap_sem);
2026
2027out_mmput:
2028 mmput(mm);
2029out:
2030 return rc;
2031}
2032
2033struct map_files_info {
20d28cde
AD		 2034 unsigned long start;
2035 unsigned long end;
7b540d06 2036 fmode_t mode;
640708a2
PE		 2037};
2038
bdb4d100
CO		 2039/*
2040 * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the
2041 * symlinks may be used to bypass permissions on ancestor directories in the
2042 * path to the file in question.
2043 */
2044static const char *
6b255391 2045proc_map_files_get_link(struct dentry *dentry,
fceef393
AV		 2046 struct inode *inode,
2047 struct delayed_call *done)
bdb4d100
CO		 2048{
2049 if (!capable(CAP_SYS_ADMIN))
2050 return ERR_PTR(-EPERM);
2051
fceef393 2052 return proc_pid_get_link(dentry, inode, done);
bdb4d100
CO		 2053}
2054
2055/*
6b255391 2056 * Identical to proc_pid_link_inode_operations except for get_link()
bdb4d100
CO		 2057 */
2058static const struct inode_operations proc_map_files_link_inode_operations = {
2059 .readlink = proc_pid_readlink,
6b255391 2060 .get_link = proc_map_files_get_link,
bdb4d100
CO		 2061 .setattr = proc_setattr,
2062};
2063
c52a47ac 2064static int
640708a2
PE		 2065proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
2066 struct task_struct *task, const void *ptr)
2067{
7b540d06 2068 fmode_t mode = (fmode_t)(unsigned long)ptr;
640708a2
PE		 2069 struct proc_inode *ei;
2070 struct inode *inode;
2071
db978da8
AG		 2072 inode = proc_pid_make_inode(dir->i_sb, task, S_IFLNK |
2073 ((mode & FMODE_READ ) ? S_IRUSR : 0) |
2074 ((mode & FMODE_WRITE) ? S_IWUSR : 0));
640708a2 2075 if (!inode)
c52a47ac 2076 return -ENOENT;
640708a2
PE		 2077
2078 ei = PROC_I(inode);
6b255391 2079 ei->op.proc_get_link = map_files_get_link;
640708a2 2080
bdb4d100 2081 inode->i_op = &proc_map_files_link_inode_operations;
640708a2 2082 inode->i_size = 64;
640708a2
PE		 2083
2084 d_set_d_op(dentry, &tid_map_files_dentry_operations);
2085 d_add(dentry, inode);
2086
c52a47ac 2087 return 0;
640708a2
PE		 2088}
2089
2090static struct dentry *proc_map_files_lookup(struct inode *dir,
00cd8dd3 2091 struct dentry *dentry, unsigned int flags)
640708a2
PE		 2092{
2093 unsigned long vm_start, vm_end;
2094 struct vm_area_struct *vma;
2095 struct task_struct *task;
c52a47ac 2096 int result;
640708a2
PE		 2097 struct mm_struct *mm;
2098
c52a47ac 2099 result = -ENOENT;
640708a2
PE		 2100 task = get_proc_task(dir);
2101 if (!task)
2102 goto out;
2103
c52a47ac 2104 result = -EACCES;
caaee623 2105 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
640708a2
PE		 2106 goto out_put_task;
2107
c52a47ac 2108 result = -ENOENT;
640708a2 2109 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
eb94cd96 2110 goto out_put_task;
640708a2
PE		 2111
2112 mm = get_task_mm(task);
2113 if (!mm)
eb94cd96 2114 goto out_put_task;
640708a2
PE		 2115
2116 down_read(&mm->mmap_sem);
2117 vma = find_exact_vma(mm, vm_start, vm_end);
2118 if (!vma)
2119 goto out_no_vma;
2120
05f56484
SK		 2121 if (vma->vm_file)
2122 result = proc_map_files_instantiate(dir, dentry, task,
2123 (void *)(unsigned long)vma->vm_file->f_mode);
640708a2
PE		 2124
2125out_no_vma:
2126 up_read(&mm->mmap_sem);
2127 mmput(mm);
640708a2
PE		 2128out_put_task:
2129 put_task_struct(task);
2130out:
c52a47ac 2131 return ERR_PTR(result);
640708a2
PE		 2132}
2133
2134static const struct inode_operations proc_map_files_inode_operations = {
2135 .lookup = proc_map_files_lookup,
2136 .permission = proc_fd_permission,
2137 .setattr = proc_setattr,
2138};
2139
2140static int
f0c3b509 2141proc_map_files_readdir(struct file *file, struct dir_context *ctx)
640708a2 2142{
640708a2
PE		 2143 struct vm_area_struct *vma;
2144 struct task_struct *task;
2145 struct mm_struct *mm;
f0c3b509
AV		 2146 unsigned long nr_files, pos, i;
2147 struct flex_array *fa = NULL;
2148 struct map_files_info info;
2149 struct map_files_info *p;
640708a2
PE		 2150 int ret;
2151
640708a2 2152 ret = -ENOENT;
f0c3b509 2153 task = get_proc_task(file_inode(file));
640708a2
PE		 2154 if (!task)
2155 goto out;
2156
2157 ret = -EACCES;
caaee623 2158 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
640708a2
PE		 2159 goto out_put_task;
2160
2161 ret = 0;
f0c3b509
AV		 2162 if (!dir_emit_dots(file, ctx))
2163 goto out_put_task;
640708a2 2164
f0c3b509
AV		 2165 mm = get_task_mm(task);
2166 if (!mm)
2167 goto out_put_task;
2168 down_read(&mm->mmap_sem);
640708a2 2169
f0c3b509 2170 nr_files = 0;
640708a2 2171
f0c3b509
AV		 2172 /*
2173 * We need two passes here:
2174 *
2175 * 1) Collect vmas of mapped files with mmap_sem taken
2176 * 2) Release mmap_sem and instantiate entries
2177 *
2178 * otherwise we get lockdep complained, since filldir()
2179 * routine might require mmap_sem taken in might_fault().
2180 */
640708a2 2181
f0c3b509
AV		 2182 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2183 if (vma->vm_file && ++pos > ctx->pos)
2184 nr_files++;
2185 }
2186
2187 if (nr_files) {
2188 fa = flex_array_alloc(sizeof(info), nr_files,
2189 GFP_KERNEL);
2190 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2191 GFP_KERNEL)) {
2192 ret = -ENOMEM;
2193 if (fa)
2194 flex_array_free(fa);
2195 up_read(&mm->mmap_sem);
2196 mmput(mm);
2197 goto out_put_task;
640708a2 2198 }
f0c3b509
AV		 2199 for (i = 0, vma = mm->mmap, pos = 2; vma;
2200 vma = vma->vm_next) {
2201 if (!vma->vm_file)
2202 continue;
2203 if (++pos <= ctx->pos)
2204 continue;
2205
20d28cde
AD		 2206 info.start = vma->vm_start;
2207 info.end = vma->vm_end;
f0c3b509 2208 info.mode = vma->vm_file->f_mode;
f0c3b509
AV		 2209 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2210 BUG();
640708a2 2211 }
640708a2 2212 }
f0c3b509
AV		 2213 up_read(&mm->mmap_sem);
2214
2215 for (i = 0; i < nr_files; i++) {
20d28cde
AD		 2216 char buf[4 * sizeof(long) + 2]; /* max: %lx-%lx\0 */
2217 unsigned int len;
2218
f0c3b509 2219 p = flex_array_get(fa, i);
20d28cde 2220 len = snprintf(buf, sizeof(buf), "%lx-%lx", p->start, p->end);
f0c3b509 2221 if (!proc_fill_cache(file, ctx,
20d28cde 2222 buf, len,
f0c3b509
AV		 2223 proc_map_files_instantiate,
2224 task,
2225 (void *)(unsigned long)p->mode))
2226 break;
2227 ctx->pos++;
640708a2 2228 }
f0c3b509
AV		 2229 if (fa)
2230 flex_array_free(fa);
2231 mmput(mm);
640708a2 2232
640708a2
PE		 2233out_put_task:
2234 put_task_struct(task);
2235out:
2236 return ret;
2237}
2238
2239static const struct file_operations proc_map_files_operations = {
2240 .read = generic_read_dir,
f50752ea
AV		 2241 .iterate_shared = proc_map_files_readdir,
2242 .llseek = generic_file_llseek,
640708a2
PE		 2243};
2244
b18b6a9c 2245#if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
48f6a7a5
PE		 2246struct timers_private {
2247 struct pid *pid;
2248 struct task_struct *task;
2249 struct sighand_struct *sighand;
57b8015e 2250 struct pid_namespace *ns;
48f6a7a5
PE		 2251 unsigned long flags;
2252};
2253
2254static void *timers_start(struct seq_file *m, loff_t *pos)
2255{
2256 struct timers_private *tp = m->private;
2257
2258 tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2259 if (!tp->task)
2260 return ERR_PTR(-ESRCH);
2261
2262 tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2263 if (!tp->sighand)
2264 return ERR_PTR(-ESRCH);
2265
2266 return seq_list_start(&tp->task->signal->posix_timers, *pos);
2267}
2268
2269static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2270{
2271 struct timers_private *tp = m->private;
2272 return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2273}
2274
2275static void timers_stop(struct seq_file *m, void *v)
2276{
2277 struct timers_private *tp = m->private;
2278
2279 if (tp->sighand) {
2280 unlock_task_sighand(tp->task, &tp->flags);
2281 tp->sighand = NULL;
2282 }
2283
2284 if (tp->task) {
2285 put_task_struct(tp->task);
2286 tp->task = NULL;
2287 }
2288}
2289
2290static int show_timer(struct seq_file *m, void *v)
2291{
2292 struct k_itimer *timer;
57b8015e
PE		 2293 struct timers_private *tp = m->private;
2294 int notify;
cedbccab 2295 static const char * const nstr[] = {
57b8015e
PE		 2296 [SIGEV_SIGNAL] = "signal",
2297 [SIGEV_NONE] = "none",
2298 [SIGEV_THREAD] = "thread",
2299 };
48f6a7a5
PE		 2300
2301 timer = list_entry((struct list_head *)v, struct k_itimer, list);
57b8015e
PE		 2302 notify = timer->it_sigev_notify;
2303
48f6a7a5 2304 seq_printf(m, "ID: %d\n", timer->it_id);
ba3edf1f 2305 seq_printf(m, "signal: %d/%px\n",
25ce3191
JP		 2306 timer->sigq->info.si_signo,
2307 timer->sigq->info.si_value.sival_ptr);
57b8015e 2308 seq_printf(m, "notify: %s/%s.%d\n",
25ce3191
JP		 2309 nstr[notify & ~SIGEV_THREAD_ID],
2310 (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2311 pid_nr_ns(timer->it_pid, tp->ns));
15ef0298 2312 seq_printf(m, "ClockID: %d\n", timer->it_clock);
48f6a7a5
PE		 2313
2314 return 0;
2315}
2316
2317static const struct seq_operations proc_timers_seq_ops = {
2318 .start = timers_start,
2319 .next = timers_next,
2320 .stop = timers_stop,
2321 .show = show_timer,
2322};
2323
2324static int proc_timers_open(struct inode *inode, struct file *file)
2325{
2326 struct timers_private *tp;
2327
2328 tp = __seq_open_private(file, &proc_timers_seq_ops,
2329 sizeof(struct timers_private));
2330 if (!tp)
2331 return -ENOMEM;
2332
2333 tp->pid = proc_pid(inode);
57b8015e 2334 tp->ns = inode->i_sb->s_fs_info;
48f6a7a5
PE		 2335 return 0;
2336}
2337
2338static const struct file_operations proc_timers_operations = {
2339 .open = proc_timers_open,
2340 .read = seq_read,
2341 .llseek = seq_lseek,
2342 .release = seq_release_private,
2343};
b5946bea 2344#endif
640708a2 2345
5de23d43
JS		 2346static ssize_t timerslack_ns_write(struct file *file, const char __user *buf,
2347 size_t count, loff_t *offset)
2348{
2349 struct inode *inode = file_inode(file);
2350 struct task_struct *p;
2351 u64 slack_ns;
2352 int err;
2353
2354 err = kstrtoull_from_user(buf, count, 10, &slack_ns);
2355 if (err < 0)
2356 return err;
2357
2358 p = get_proc_task(inode);
2359 if (!p)
2360 return -ESRCH;
2361
4b2bd5fe
JS		 2362 if (p != current) {
2363 if (!capable(CAP_SYS_NICE)) {
2364 count = -EPERM;
2365 goto out;
2366 }
5de23d43 2367
4b2bd5fe
JS		 2368 err = security_task_setscheduler(p);
2369 if (err) {
2370 count = err;
2371 goto out;
2372 }
904763e1
JS		 2373 }
2374
7abbaf94
JS		 2375 task_lock(p);
2376 if (slack_ns == 0)
2377 p->timer_slack_ns = p->default_timer_slack_ns;
2378 else
2379 p->timer_slack_ns = slack_ns;
2380 task_unlock(p);
2381
2382out:
5de23d43
JS		 2383 put_task_struct(p);
2384
2385 return count;
2386}
2387
2388static int timerslack_ns_show(struct seq_file *m, void *v)
2389{
2390 struct inode *inode = m->private;
2391 struct task_struct *p;
7abbaf94 2392 int err = 0;
5de23d43
JS		 2393
2394 p = get_proc_task(inode);
2395 if (!p)
2396 return -ESRCH;
2397
4b2bd5fe 2398 if (p != current) {
5de23d43 2399
4b2bd5fe
JS		 2400 if (!capable(CAP_SYS_NICE)) {
2401 err = -EPERM;
2402 goto out;
2403 }
2404 err = security_task_getscheduler(p);
2405 if (err)
2406 goto out;
2407 }
904763e1 2408
7abbaf94
JS		 2409 task_lock(p);
2410 seq_printf(m, "%llu\n", p->timer_slack_ns);
2411 task_unlock(p);
2412
2413out:
5de23d43
JS		 2414 put_task_struct(p);
2415
2416 return err;
2417}
2418
2419static int timerslack_ns_open(struct inode *inode, struct file *filp)
2420{
2421 return single_open(filp, timerslack_ns_show, inode);
2422}
2423
2424static const struct file_operations proc_pid_set_timerslack_ns_operations = {
2425 .open = timerslack_ns_open,
2426 .read = seq_read,
2427 .write = timerslack_ns_write,
2428 .llseek = seq_lseek,
2429 .release = single_release,
2430};
2431
c52a47ac 2432static int proc_pident_instantiate(struct inode *dir,
c5141e6d 2433 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8 2434{
c5141e6d 2435 const struct pid_entry *p = ptr;
444ceed8
EB		 2436 struct inode *inode;
2437 struct proc_inode *ei;
444ceed8 2438
db978da8 2439 inode = proc_pid_make_inode(dir->i_sb, task, p->mode);
444ceed8
EB		 2440 if (!inode)
2441 goto out;
2442
2443 ei = PROC_I(inode);
444ceed8 2444 if (S_ISDIR(inode->i_mode))
bfe86848 2445 set_nlink(inode, 2); /* Use getattr to fix if necessary */
444ceed8
EB		 2446 if (p->iop)
2447 inode->i_op = p->iop;
2448 if (p->fop)
2449 inode->i_fop = p->fop;
2450 ei->op = p->op;
fb045adb 2451 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 2452 d_add(dentry, inode);
2453 /* Close the race of the process dying before we return the dentry */
0b728e19 2454 if (pid_revalidate(dentry, 0))
c52a47ac 2455 return 0;
444ceed8 2456out:
c52a47ac 2457 return -ENOENT;
444ceed8
EB		 2458}
2459
1da177e4
LT		 2460static struct dentry *proc_pident_lookup(struct inode *dir,
2461 struct dentry *dentry,
c5141e6d 2462 const struct pid_entry *ents,
7bcd6b0e 2463 unsigned int nents)
1da177e4 2464{
c52a47ac 2465 int error;
99f89551 2466 struct task_struct *task = get_proc_task(dir);
c5141e6d 2467 const struct pid_entry *p, *last;
1da177e4 2468
c52a47ac 2469 error = -ENOENT;
1da177e4 2470
99f89551
EB		 2471 if (!task)
2472 goto out_no_task;
1da177e4 2473
20cdc894
EB		 2474 /*
2475 * Yes, it does not scale. And it should not. Don't add
2476 * new entries into /proc/<tgid>/ without very good reasons.
2477 */
bac5f5d5
AD		 2478 last = &ents[nents];
2479 for (p = ents; p < last; p++) {
1da177e4
LT		 2480 if (p->len != dentry->d_name.len)
2481 continue;
2482 if (!memcmp(dentry->d_name.name, p->name, p->len))
2483 break;
2484 }
bac5f5d5 2485 if (p >= last)
1da177e4
LT		 2486 goto out;
2487
444ceed8 2488 error = proc_pident_instantiate(dir, dentry, task, p);
1da177e4 2489out:
99f89551
EB		 2490 put_task_struct(task);
2491out_no_task:
c52a47ac 2492 return ERR_PTR(error);
1da177e4
LT		 2493}
2494
f0c3b509 2495static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
c5141e6d 2496 const struct pid_entry *ents, unsigned int nents)
28a6d671 2497{
f0c3b509
AV		 2498 struct task_struct *task = get_proc_task(file_inode(file));
2499 const struct pid_entry *p;
28a6d671 2500
28a6d671 2501 if (!task)
f0c3b509 2502 return -ENOENT;
28a6d671 2503
f0c3b509
AV		 2504 if (!dir_emit_dots(file, ctx))
2505 goto out;
2506
2507 if (ctx->pos >= nents + 2)
2508 goto out;
28a6d671 2509
bac5f5d5 2510 for (p = ents + (ctx->pos - 2); p < ents + nents; p++) {
f0c3b509
AV		 2511 if (!proc_fill_cache(file, ctx, p->name, p->len,
2512 proc_pident_instantiate, task, p))
2513 break;
2514 ctx->pos++;
2515 }
28a6d671 2516out:
61a28784 2517 put_task_struct(task);
f0c3b509 2518 return 0;
1da177e4
LT		 2519}
2520
28a6d671
EB		 2521#ifdef CONFIG_SECURITY
2522static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2523 size_t count, loff_t *ppos)
2524{
496ad9aa 2525 struct inode * inode = file_inode(file);
04ff9708 2526 char *p = NULL;
28a6d671
EB		 2527 ssize_t length;
2528 struct task_struct *task = get_proc_task(inode);
2529
28a6d671 2530 if (!task)
04ff9708 2531 return -ESRCH;
28a6d671
EB		 2532
2533 length = security_getprocattr(task,
2fddfeef 2534 (char*)file->f_path.dentry->d_name.name,
04ff9708 2535 &p);
28a6d671 2536 put_task_struct(task);
04ff9708
AV		 2537 if (length > 0)
2538 length = simple_read_from_buffer(buf, count, ppos, p, length);
2539 kfree(p);
28a6d671 2540 return length;
1da177e4
LT		 2541}
2542
28a6d671
EB		 2543static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2544 size_t count, loff_t *ppos)
2545{
496ad9aa 2546 struct inode * inode = file_inode(file);
bb646cdb 2547 void *page;
28a6d671
EB		 2548 ssize_t length;
2549 struct task_struct *task = get_proc_task(inode);
2550
2551 length = -ESRCH;
2552 if (!task)
2553 goto out_no_task;
b21507e2
SS		 2554
2555 /* A task may only write its own attributes. */
2556 length = -EACCES;
2557 if (current != task)
2558 goto out;
2559
28a6d671
EB		 2560 if (count > PAGE_SIZE)
2561 count = PAGE_SIZE;
2562
2563 /* No partial writes. */
2564 length = -EINVAL;
2565 if (*ppos != 0)
2566 goto out;
2567
bb646cdb
AV		 2568 page = memdup_user(buf, count);
2569 if (IS_ERR(page)) {
2570 length = PTR_ERR(page);
28a6d671 2571 goto out;
bb646cdb 2572 }
28a6d671 2573
107db7c7 2574 /* Guard against adverse ptrace interaction */
b21507e2 2575 length = mutex_lock_interruptible(&current->signal->cred_guard_mutex);
107db7c7
DH		 2576 if (length < 0)
2577 goto out_free;
2578
b21507e2 2579 length = security_setprocattr(file->f_path.dentry->d_name.name,
bb646cdb 2580 page, count);
b21507e2 2581 mutex_unlock(&current->signal->cred_guard_mutex);
28a6d671 2582out_free:
bb646cdb 2583 kfree(page);
28a6d671
EB		 2584out:
2585 put_task_struct(task);
2586out_no_task:
2587 return length;
2588}
2589
00977a59 2590static const struct file_operations proc_pid_attr_operations = {
28a6d671
EB		 2591 .read = proc_pid_attr_read,
2592 .write = proc_pid_attr_write,
87df8424 2593 .llseek = generic_file_llseek,
28a6d671
EB		 2594};
2595
c5141e6d 2596static const struct pid_entry attr_dir_stuff[] = {
631f9c18
AD		 2597 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2598 REG("prev", S_IRUGO, proc_pid_attr_operations),
2599 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2600 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2601 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2602 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
28a6d671
EB		 2603};
2604
f0c3b509 2605static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
28a6d671 2606{
f0c3b509
AV		 2607 return proc_pident_readdir(file, ctx,
2608 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2609}
2610
00977a59 2611static const struct file_operations proc_attr_dir_operations = {
1da177e4 2612 .read = generic_read_dir,
f50752ea
AV		 2613 .iterate_shared = proc_attr_dir_readdir,
2614 .llseek = generic_file_llseek,
1da177e4
LT		 2615};
2616
72d9dcfc 2617static struct dentry *proc_attr_dir_lookup(struct inode *dir,
00cd8dd3 2618 struct dentry *dentry, unsigned int flags)
28a6d671 2619{
7bcd6b0e
EB		 2620 return proc_pident_lookup(dir, dentry,
2621 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2622}
2623
c5ef1c42 2624static const struct inode_operations proc_attr_dir_inode_operations = {
72d9dcfc 2625 .lookup = proc_attr_dir_lookup,
99f89551 2626 .getattr = pid_getattr,
6d76fa58 2627 .setattr = proc_setattr,
1da177e4
LT		 2628};
2629
28a6d671
EB		 2630#endif
2631
698ba7b5 2632#ifdef CONFIG_ELF_CORE
3cb4a0bb
KH		 2633static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2634 size_t count, loff_t *ppos)
2635{
496ad9aa 2636 struct task_struct *task = get_proc_task(file_inode(file));
3cb4a0bb
KH		 2637 struct mm_struct *mm;
2638 char buffer[PROC_NUMBUF];
2639 size_t len;
2640 int ret;
2641
2642 if (!task)
2643 return -ESRCH;
2644
2645 ret = 0;
2646 mm = get_task_mm(task);
2647 if (mm) {
2648 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2649 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2650 MMF_DUMP_FILTER_SHIFT));
2651 mmput(mm);
2652 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2653 }
2654
2655 put_task_struct(task);
2656
2657 return ret;
2658}
2659
2660static ssize_t proc_coredump_filter_write(struct file *file,
2661 const char __user *buf,
2662 size_t count,
2663 loff_t *ppos)
2664{
2665 struct task_struct *task;
2666 struct mm_struct *mm;
3cb4a0bb
KH		 2667 unsigned int val;
2668 int ret;
2669 int i;
2670 unsigned long mask;
2671
774636e1
AD		 2672 ret = kstrtouint_from_user(buf, count, 0, &val);
2673 if (ret < 0)
2674 return ret;
3cb4a0bb
KH		 2675
2676 ret = -ESRCH;
496ad9aa 2677 task = get_proc_task(file_inode(file));
3cb4a0bb
KH		 2678 if (!task)
2679 goto out_no_task;
2680
3cb4a0bb
KH		 2681 mm = get_task_mm(task);
2682 if (!mm)
2683 goto out_no_mm;
41a0c249 2684 ret = 0;
3cb4a0bb
KH		 2685
2686 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2687 if (val & mask)
2688 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2689 else
2690 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2691 }
2692
2693 mmput(mm);
2694 out_no_mm:
2695 put_task_struct(task);
2696 out_no_task:
774636e1
AD		 2697 if (ret < 0)
2698 return ret;
2699 return count;
3cb4a0bb
KH		 2700}
2701
2702static const struct file_operations proc_coredump_filter_operations = {
2703 .read = proc_coredump_filter_read,
2704 .write = proc_coredump_filter_write,
87df8424 2705 .llseek = generic_file_llseek,
3cb4a0bb
KH		 2706};
2707#endif
2708
aba76fdb 2709#ifdef CONFIG_TASK_IO_ACCOUNTING
19aadc98 2710static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
297c5d92 2711{
940389b8 2712 struct task_io_accounting acct = task->ioac;
5995477a 2713 unsigned long flags;
293eb1e7 2714 int result;
5995477a 2715
293eb1e7
VK		 2716 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2717 if (result)
2718 return result;
2719
caaee623 2720 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) {
293eb1e7
VK		 2721 result = -EACCES;
2722 goto out_unlock;
2723 }
1d1221f3 2724
5995477a
AR		 2725 if (whole && lock_task_sighand(task, &flags)) {
2726 struct task_struct *t = task;
2727
2728 task_io_accounting_add(&acct, &task->signal->ioac);
2729 while_each_thread(task, t)
2730 task_io_accounting_add(&acct, &t->ioac);
2731
2732 unlock_task_sighand(task, &flags);
297c5d92 2733 }
25ce3191
JP		 2734 seq_printf(m,
2735 "rchar: %llu\n"
2736 "wchar: %llu\n"
2737 "syscr: %llu\n"
2738 "syscw: %llu\n"
2739 "read_bytes: %llu\n"
2740 "write_bytes: %llu\n"
2741 "cancelled_write_bytes: %llu\n",
2742 (unsigned long long)acct.rchar,
2743 (unsigned long long)acct.wchar,
2744 (unsigned long long)acct.syscr,
2745 (unsigned long long)acct.syscw,
2746 (unsigned long long)acct.read_bytes,
2747 (unsigned long long)acct.write_bytes,
2748 (unsigned long long)acct.cancelled_write_bytes);
2749 result = 0;
2750
293eb1e7
VK		 2751out_unlock:
2752 mutex_unlock(&task->signal->cred_guard_mutex);
2753 return result;
297c5d92
AR		 2754}
2755
19aadc98
AD		 2756static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2757 struct pid *pid, struct task_struct *task)
297c5d92 2758{
19aadc98 2759 return do_io_accounting(task, m, 0);
aba76fdb 2760}
297c5d92 2761
19aadc98
AD		 2762static int proc_tgid_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, 1);
297c5d92
AR		 2766}
2767#endif /* CONFIG_TASK_IO_ACCOUNTING */
aba76fdb 2768
22d917d8
EB		 2769#ifdef CONFIG_USER_NS
2770static int proc_id_map_open(struct inode *inode, struct file *file,
ccf94f1b 2771 const struct seq_operations *seq_ops)
22d917d8
EB		 2772{
2773 struct user_namespace *ns = NULL;
2774 struct task_struct *task;
2775 struct seq_file *seq;
2776 int ret = -EINVAL;
2777
2778 task = get_proc_task(inode);
2779 if (task) {
2780 rcu_read_lock();
2781 ns = get_user_ns(task_cred_xxx(task, user_ns));
2782 rcu_read_unlock();
2783 put_task_struct(task);
2784 }
2785 if (!ns)
2786 goto err;
2787
2788 ret = seq_open(file, seq_ops);
2789 if (ret)
2790 goto err_put_ns;
2791
2792 seq = file->private_data;
2793 seq->private = ns;
2794
2795 return 0;
2796err_put_ns:
2797 put_user_ns(ns);
2798err:
2799 return ret;
2800}
2801
2802static int proc_id_map_release(struct inode *inode, struct file *file)
2803{
2804 struct seq_file *seq = file->private_data;
2805 struct user_namespace *ns = seq->private;
2806 put_user_ns(ns);
2807 return seq_release(inode, file);
2808}
2809
2810static int proc_uid_map_open(struct inode *inode, struct file *file)
2811{
2812 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2813}
2814
2815static int proc_gid_map_open(struct inode *inode, struct file *file)
2816{
2817 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2818}
2819
f76d207a
EB		 2820static int proc_projid_map_open(struct inode *inode, struct file *file)
2821{
2822 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2823}
2824
22d917d8
EB		 2825static const struct file_operations proc_uid_map_operations = {
2826 .open = proc_uid_map_open,
2827 .write = proc_uid_map_write,
2828 .read = seq_read,
2829 .llseek = seq_lseek,
2830 .release = proc_id_map_release,
2831};
2832
2833static const struct file_operations proc_gid_map_operations = {
2834 .open = proc_gid_map_open,
2835 .write = proc_gid_map_write,
2836 .read = seq_read,
2837 .llseek = seq_lseek,
2838 .release = proc_id_map_release,
2839};
f76d207a
EB		 2840
2841static const struct file_operations proc_projid_map_operations = {
2842 .open = proc_projid_map_open,
2843 .write = proc_projid_map_write,
2844 .read = seq_read,
2845 .llseek = seq_lseek,
2846 .release = proc_id_map_release,
2847};
9cc46516
EB		 2848
2849static int proc_setgroups_open(struct inode *inode, struct file *file)
2850{
2851 struct user_namespace *ns = NULL;
2852 struct task_struct *task;
2853 int ret;
2854
2855 ret = -ESRCH;
2856 task = get_proc_task(inode);
2857 if (task) {
2858 rcu_read_lock();
2859 ns = get_user_ns(task_cred_xxx(task, user_ns));
2860 rcu_read_unlock();
2861 put_task_struct(task);
2862 }
2863 if (!ns)
2864 goto err;
2865
2866 if (file->f_mode & FMODE_WRITE) {
2867 ret = -EACCES;
2868 if (!ns_capable(ns, CAP_SYS_ADMIN))
2869 goto err_put_ns;
2870 }
2871
2872 ret = single_open(file, &proc_setgroups_show, ns);
2873 if (ret)
2874 goto err_put_ns;
2875
2876 return 0;
2877err_put_ns:
2878 put_user_ns(ns);
2879err:
2880 return ret;
2881}
2882
2883static int proc_setgroups_release(struct inode *inode, struct file *file)
2884{
2885 struct seq_file *seq = file->private_data;
2886 struct user_namespace *ns = seq->private;
2887 int ret = single_release(inode, file);
2888 put_user_ns(ns);
2889 return ret;
2890}
2891
2892static const struct file_operations proc_setgroups_operations = {
2893 .open = proc_setgroups_open,
2894 .write = proc_setgroups_write,
2895 .read = seq_read,
2896 .llseek = seq_lseek,
2897 .release = proc_setgroups_release,
2898};
22d917d8
EB		 2899#endif /* CONFIG_USER_NS */
2900
47830723
KC		 2901static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2902 struct pid *pid, struct task_struct *task)
2903{
a9712bc1
AV		 2904 int err = lock_trace(task);
2905 if (!err) {
2906 seq_printf(m, "%08x\n", task->personality);
2907 unlock_trace(task);
2908 }
2909 return err;
47830723
KC		 2910}
2911
7c23b330
JP		 2912#ifdef CONFIG_LIVEPATCH
2913static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns,
2914 struct pid *pid, struct task_struct *task)
2915{
2916 seq_printf(m, "%d\n", task->patch_state);
2917 return 0;
2918}
2919#endif /* CONFIG_LIVEPATCH */
2920
28a6d671
EB		 2921/*
2922 * Thread groups
2923 */
00977a59 2924static const struct file_operations proc_task_operations;
c5ef1c42 2925static const struct inode_operations proc_task_inode_operations;
20cdc894 2926
c5141e6d 2927static const struct pid_entry tgid_base_stuff[] = {
631f9c18
AD		 2928 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2929 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
640708a2 2930 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
631f9c18 2931 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 2932 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
b2211a36 2933#ifdef CONFIG_NET
631f9c18 2934 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
b2211a36 2935#endif
631f9c18 2936 REG("environ", S_IRUSR, proc_environ_operations),
c5317167 2937 REG("auxv", S_IRUSR, proc_auxv_operations),
631f9c18 2938 ONE("status", S_IRUGO, proc_pid_status),
35a35046 2939 ONE("personality", S_IRUSR, proc_pid_personality),
1c963eb1 2940 ONE("limits", S_IRUGO, proc_pid_limits),
43ae34cb 2941#ifdef CONFIG_SCHED_DEBUG
631f9c18 2942 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
5091faa4
MG		 2943#endif
2944#ifdef CONFIG_SCHED_AUTOGROUP
2945 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
ebcb6734 2946#endif
4614a696 2947 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 2948#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
09d93bd6 2949 ONE("syscall", S_IRUSR, proc_pid_syscall),
43ae34cb 2950#endif
c2c0bb44 2951 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
631f9c18
AD		 2952 ONE("stat", S_IRUGO, proc_tgid_stat),
2953 ONE("statm", S_IRUGO, proc_pid_statm),
b7643757 2954 REG("maps", S_IRUGO, proc_pid_maps_operations),
28a6d671 2955#ifdef CONFIG_NUMA
b7643757 2956 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
28a6d671 2957#endif
631f9c18
AD		 2958 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2959 LNK("cwd", proc_cwd_link),
2960 LNK("root", proc_root_link),
2961 LNK("exe", proc_exe_link),
2962 REG("mounts", S_IRUGO, proc_mounts_operations),
2963 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2964 REG("mountstats", S_IRUSR, proc_mountstats_operations),
1e883281 2965#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18 2966 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
b7643757 2967 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
493b0e9d 2968 REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
32ed74a4 2969 REG("pagemap", S_IRUSR, proc_pagemap_operations),
28a6d671
EB		 2970#endif
2971#ifdef CONFIG_SECURITY
631f9c18 2972 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 2973#endif
2974#ifdef CONFIG_KALLSYMS
edfcd606 2975 ONE("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 2976#endif
2ec220e2 2977#ifdef CONFIG_STACKTRACE
35a35046 2978 ONE("stack", S_IRUSR, proc_pid_stack),
28a6d671 2979#endif
5968cece 2980#ifdef CONFIG_SCHED_INFO
f6e826ca 2981 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 2982#endif
9745512c 2983#ifdef CONFIG_LATENCYTOP
631f9c18 2984 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 2985#endif
8793d854 2986#ifdef CONFIG_PROC_PID_CPUSET
52de4779 2987 ONE("cpuset", S_IRUGO, proc_cpuset_show),
a424316c
PM		 2988#endif
2989#ifdef CONFIG_CGROUPS
006f4ac4 2990 ONE("cgroup", S_IRUGO, proc_cgroup_show),
28a6d671 2991#endif
6ba51e37 2992 ONE("oom_score", S_IRUGO, proc_oom_score),
fa0cbbf1 2993 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
a63d83f4 2994 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 2995#ifdef CONFIG_AUDITSYSCALL
631f9c18
AD		 2996 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2997 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 2998#endif
f4f154fd 2999#ifdef CONFIG_FAULT_INJECTION
631f9c18 3000 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
168c42bc 3001 REG("fail-nth", 0644, proc_fail_nth_operations),
f4f154fd 3002#endif
698ba7b5 3003#ifdef CONFIG_ELF_CORE
631f9c18 3004 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3cb4a0bb 3005#endif
aba76fdb 3006#ifdef CONFIG_TASK_IO_ACCOUNTING
19aadc98 3007 ONE("io", S_IRUSR, proc_tgid_io_accounting),
aba76fdb 3008#endif
22d917d8
EB		 3009#ifdef CONFIG_USER_NS
3010 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3011 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
f76d207a 3012 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
9cc46516 3013 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
22d917d8 3014#endif
b18b6a9c 3015#if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
48f6a7a5
PE		 3016 REG("timers", S_IRUGO, proc_timers_operations),
3017#endif
5de23d43 3018 REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations),
7c23b330
JP		 3019#ifdef CONFIG_LIVEPATCH
3020 ONE("patch_state", S_IRUSR, proc_pid_patch_state),
3021#endif
28a6d671 3022};
1da177e4 3023
f0c3b509 3024static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
1da177e4 3025{
f0c3b509
AV		 3026 return proc_pident_readdir(file, ctx,
3027 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 3028}
3029
00977a59 3030static const struct file_operations proc_tgid_base_operations = {
1da177e4 3031 .read = generic_read_dir,
f50752ea
AV		 3032 .iterate_shared = proc_tgid_base_readdir,
3033 .llseek = generic_file_llseek,
1da177e4
LT		 3034};
3035
00cd8dd3
AV		 3036static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3037{
7bcd6b0e
EB		 3038 return proc_pident_lookup(dir, dentry,
3039 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 3040}
3041
c5ef1c42 3042static const struct inode_operations proc_tgid_base_inode_operations = {
28a6d671 3043 .lookup = proc_tgid_base_lookup,
99f89551 3044 .getattr = pid_getattr,
6d76fa58 3045 .setattr = proc_setattr,
0499680a 3046 .permission = proc_pid_permission,
1da177e4 3047};
1da177e4 3048
60347f67 3049static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
1da177e4 3050{
48e6484d 3051 struct dentry *dentry, *leader, *dir;
e3912ac3 3052 char buf[10 + 1];
48e6484d
EB		 3053 struct qstr name;
3054
3055 name.name = buf;
e3912ac3 3056 name.len = snprintf(buf, sizeof(buf), "%u", pid);
4f522a24 3057 /* no ->d_hash() rejects on procfs */
60347f67 3058 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d 3059 if (dentry) {
bbd51924 3060 d_invalidate(dentry);
48e6484d
EB		 3061 dput(dentry);
3062 }
1da177e4 3063
c35a7f18
ON		 3064 if (pid == tgid)
3065 return;
3066
48e6484d 3067 name.name = buf;
e3912ac3 3068 name.len = snprintf(buf, sizeof(buf), "%u", tgid);
60347f67 3069 leader = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d
EB		 3070 if (!leader)
3071 goto out;
1da177e4 3072
48e6484d
EB		 3073 name.name = "task";
3074 name.len = strlen(name.name);
3075 dir = d_hash_and_lookup(leader, &name);
3076 if (!dir)
3077 goto out_put_leader;
3078
3079 name.name = buf;
e3912ac3 3080 name.len = snprintf(buf, sizeof(buf), "%u", pid);
48e6484d
EB		 3081 dentry = d_hash_and_lookup(dir, &name);
3082 if (dentry) {
bbd51924 3083 d_invalidate(dentry);
48e6484d 3084 dput(dentry);
1da177e4 3085 }
48e6484d
EB		 3086
3087 dput(dir);
3088out_put_leader:
3089 dput(leader);
3090out:
3091 return;
1da177e4
LT		 3092}
3093
0895e91d
RD		 3094/**
3095 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3096 * @task: task that should be flushed.
3097 *
3098 * When flushing dentries from proc, one needs to flush them from global
60347f67 3099 * proc (proc_mnt) and from all the namespaces' procs this task was seen
0895e91d
RD		 3100 * in. This call is supposed to do all of this job.
3101 *
3102 * Looks in the dcache for
3103 * /proc/@pid
3104 * /proc/@tgid/task/@pid
3105 * if either directory is present flushes it and all of it'ts children
3106 * from the dcache.
3107 *
3108 * It is safe and reasonable to cache /proc entries for a task until
3109 * that task exits. After that they just clog up the dcache with
3110 * useless entries, possibly causing useful dcache entries to be
3111 * flushed instead. This routine is proved to flush those useless
3112 * dcache entries at process exit time.
3113 *
3114 * NOTE: This routine is just an optimization so it does not guarantee
3115 * that no dcache entries will exist at process exit time it
3116 * just makes it very unlikely that any will persist.
60347f67
PE		 3117 */
3118
3119void proc_flush_task(struct task_struct *task)
3120{
9fcc2d15 3121 int i;
9b4d1cbe 3122 struct pid *pid, *tgid;
130f77ec
PE		 3123 struct upid *upid;
3124
130f77ec 3125 pid = task_pid(task);
9b4d1cbe 3126 tgid = task_tgid(task);
130f77ec 3127
9fcc2d15 3128 for (i = 0; i <= pid->level; i++) {
130f77ec
PE		 3129 upid = &pid->numbers[i];
3130 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
9b4d1cbe 3131 tgid->numbers[i].nr);
130f77ec 3132 }
60347f67
PE		 3133}
3134
c52a47ac
AV		 3135static int proc_pid_instantiate(struct inode *dir,
3136 struct dentry * dentry,
3137 struct task_struct *task, const void *ptr)
444ceed8 3138{
444ceed8
EB		 3139 struct inode *inode;
3140
db978da8 3141 inode = proc_pid_make_inode(dir->i_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
444ceed8
EB		 3142 if (!inode)
3143 goto out;
3144
444ceed8
EB		 3145 inode->i_op = &proc_tgid_base_inode_operations;
3146 inode->i_fop = &proc_tgid_base_operations;
3147 inode->i_flags|=S_IMMUTABLE;
aed54175 3148
1270dd8d 3149 set_nlink(inode, nlink_tgid);
444ceed8 3150
fb045adb 3151 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 3152
3153 d_add(dentry, inode);
3154 /* Close the race of the process dying before we return the dentry */
0b728e19 3155 if (pid_revalidate(dentry, 0))
c52a47ac 3156 return 0;
444ceed8 3157out:
c52a47ac 3158 return -ENOENT;
444ceed8
EB		 3159}
3160
00cd8dd3 3161struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
1da177e4 3162{
335eb531 3163 int result = -ENOENT;
1da177e4 3164 struct task_struct *task;
1da177e4 3165 unsigned tgid;
b488893a 3166 struct pid_namespace *ns;
1da177e4 3167
dbcdb504 3168 tgid = name_to_int(&dentry->d_name);
1da177e4
LT		 3169 if (tgid == ~0U)
3170 goto out;
3171
b488893a 3172 ns = dentry->d_sb->s_fs_info;
de758734 3173 rcu_read_lock();
b488893a 3174 task = find_task_by_pid_ns(tgid, ns);
1da177e4
LT		 3175 if (task)
3176 get_task_struct(task);
de758734 3177 rcu_read_unlock();
1da177e4
LT		 3178 if (!task)
3179 goto out;
3180
444ceed8 3181 result = proc_pid_instantiate(dir, dentry, task, NULL);
1da177e4 3182 put_task_struct(task);
1da177e4 3183out:
c52a47ac 3184 return ERR_PTR(result);
1da177e4
LT		 3185}
3186
1da177e4 3187/*
0804ef4b 3188 * Find the first task with tgid >= tgid
0bc58a91 3189 *
1da177e4 3190 */
19fd4bb2
EB		 3191struct tgid_iter {
3192 unsigned int tgid;
0804ef4b 3193 struct task_struct *task;
19fd4bb2
EB		 3194};
3195static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3196{
0804ef4b 3197 struct pid *pid;
1da177e4 3198
19fd4bb2
EB		 3199 if (iter.task)
3200 put_task_struct(iter.task);
454cc105 3201 rcu_read_lock();
0804ef4b 3202retry:
19fd4bb2
EB		 3203 iter.task = NULL;
3204 pid = find_ge_pid(iter.tgid, ns);
0804ef4b 3205 if (pid) {
19fd4bb2
EB		 3206 iter.tgid = pid_nr_ns(pid, ns);
3207 iter.task = pid_task(pid, PIDTYPE_PID);
0804ef4b
EB		 3208 /* What we to know is if the pid we have find is the
3209 * pid of a thread_group_leader. Testing for task
3210 * being a thread_group_leader is the obvious thing
3211 * todo but there is a window when it fails, due to
3212 * the pid transfer logic in de_thread.
3213 *
3214 * So we perform the straight forward test of seeing
3215 * if the pid we have found is the pid of a thread
3216 * group leader, and don't worry if the task we have
3217 * found doesn't happen to be a thread group leader.
3218 * As we don't care in the case of readdir.
3219 */
19fd4bb2
EB		 3220 if (!iter.task || !has_group_leader_pid(iter.task)) {
3221 iter.tgid += 1;
0804ef4b 3222 goto retry;
19fd4bb2
EB		 3223 }
3224 get_task_struct(iter.task);
0bc58a91 3225 }
454cc105 3226 rcu_read_unlock();
19fd4bb2 3227 return iter;
1da177e4
LT		 3228}
3229
0097875b 3230#define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
0804ef4b 3231
1da177e4 3232/* for the /proc/ directory itself, after non-process stuff has been done */
f0c3b509 3233int proc_pid_readdir(struct file *file, struct dir_context *ctx)
1da177e4 3234{
19fd4bb2 3235 struct tgid_iter iter;
3aa3377f 3236 struct pid_namespace *ns = file_inode(file)->i_sb->s_fs_info;
f0c3b509 3237 loff_t pos = ctx->pos;
1da177e4 3238
021ada7d 3239 if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
f0c3b509 3240 return 0;
1da177e4 3241
0097875b 3242 if (pos == TGID_OFFSET - 2) {
2b0143b5 3243 struct inode *inode = d_inode(ns->proc_self);
db963164 3244 if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
f0c3b509 3245 return 0;
0097875b
EB		 3246 ctx->pos = pos = pos + 1;
3247 }
3248 if (pos == TGID_OFFSET - 1) {
2b0143b5 3249 struct inode *inode = d_inode(ns->proc_thread_self);
0097875b
EB		 3250 if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3251 return 0;
3252 ctx->pos = pos = pos + 1;
021ada7d 3253 }
0097875b 3254 iter.tgid = pos - TGID_OFFSET;
19fd4bb2 3255 iter.task = NULL;
19fd4bb2
EB		 3256 for (iter = next_tgid(ns, iter);
3257 iter.task;
3258 iter.tgid += 1, iter = next_tgid(ns, iter)) {
e3912ac3 3259 char name[10 + 1];
f0c3b509 3260 int len;
3ba4bcee
ED		 3261
3262 cond_resched();
796f571b 3263 if (!has_pid_permissions(ns, iter.task, HIDEPID_INVISIBLE))
f0c3b509 3264 continue;
0499680a 3265
e3912ac3 3266 len = snprintf(name, sizeof(name), "%u", iter.tgid);
f0c3b509
AV		 3267 ctx->pos = iter.tgid + TGID_OFFSET;
3268 if (!proc_fill_cache(file, ctx, name, len,
3269 proc_pid_instantiate, iter.task, NULL)) {
19fd4bb2 3270 put_task_struct(iter.task);
f0c3b509 3271 return 0;
1da177e4 3272 }
0bc58a91 3273 }
f0c3b509 3274 ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
0bc58a91
EB		 3275 return 0;
3276}
1da177e4 3277
1b3044e3
JD		 3278/*
3279 * proc_tid_comm_permission is a special permission function exclusively
3280 * used for the node /proc/<pid>/task/<tid>/comm.
3281 * It bypasses generic permission checks in the case where a task of the same
3282 * task group attempts to access the node.
3283 * The rationale behind this is that glibc and bionic access this node for
3284 * cross thread naming (pthread_set/getname_np(!self)). However, if
3285 * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3286 * which locks out the cross thread naming implementation.
3287 * This function makes sure that the node is always accessible for members of
3288 * same thread group.
3289 */
3290static int proc_tid_comm_permission(struct inode *inode, int mask)
3291{
3292 bool is_same_tgroup;
3293 struct task_struct *task;
3294
3295 task = get_proc_task(inode);
3296 if (!task)
3297 return -ESRCH;
3298 is_same_tgroup = same_thread_group(current, task);
3299 put_task_struct(task);
3300
3301 if (likely(is_same_tgroup && !(mask & MAY_EXEC))) {
3302 /* This file (/proc/<pid>/task/<tid>/comm) can always be
3303 * read or written by the members of the corresponding
3304 * thread group.
3305 */
3306 return 0;
3307 }
3308
3309 return generic_permission(inode, mask);
3310}
3311
3312static const struct inode_operations proc_tid_comm_inode_operations = {
3313 .permission = proc_tid_comm_permission,
3314};
3315
28a6d671
EB		 3316/*
3317 * Tasks
3318 */
c5141e6d 3319static const struct pid_entry tid_base_stuff[] = {
631f9c18 3320 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3835541d 3321 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 3322 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
6ba8ed79
EB		 3323#ifdef CONFIG_NET
3324 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3325#endif
631f9c18 3326 REG("environ", S_IRUSR, proc_environ_operations),
c5317167 3327 REG("auxv", S_IRUSR, proc_auxv_operations),
631f9c18 3328 ONE("status", S_IRUGO, proc_pid_status),
35a35046 3329 ONE("personality", S_IRUSR, proc_pid_personality),
1c963eb1 3330 ONE("limits", S_IRUGO, proc_pid_limits),
43ae34cb 3331#ifdef CONFIG_SCHED_DEBUG
631f9c18 3332 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
ebcb6734 3333#endif
1b3044e3
JD		 3334 NOD("comm", S_IFREG|S_IRUGO|S_IWUSR,
3335 &proc_tid_comm_inode_operations,
3336 &proc_pid_set_comm_operations, {}),
ebcb6734 3337#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
09d93bd6 3338 ONE("syscall", S_IRUSR, proc_pid_syscall),
43ae34cb 3339#endif
c2c0bb44 3340 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
631f9c18
AD		 3341 ONE("stat", S_IRUGO, proc_tid_stat),
3342 ONE("statm", S_IRUGO, proc_pid_statm),
b7643757 3343 REG("maps", S_IRUGO, proc_tid_maps_operations),
2e13ba54 3344#ifdef CONFIG_PROC_CHILDREN
81841161
CG		 3345 REG("children", S_IRUGO, proc_tid_children_operations),
3346#endif
28a6d671 3347#ifdef CONFIG_NUMA
b7643757 3348 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
28a6d671 3349#endif
631f9c18
AD		 3350 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3351 LNK("cwd", proc_cwd_link),
3352 LNK("root", proc_root_link),
3353 LNK("exe", proc_exe_link),
3354 REG("mounts", S_IRUGO, proc_mounts_operations),
3355 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
1e883281 3356#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18 3357 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
b7643757 3358 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
493b0e9d 3359 REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
32ed74a4 3360 REG("pagemap", S_IRUSR, proc_pagemap_operations),
28a6d671
EB		 3361#endif
3362#ifdef CONFIG_SECURITY
631f9c18 3363 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 3364#endif
3365#ifdef CONFIG_KALLSYMS
edfcd606 3366 ONE("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 3367#endif
2ec220e2 3368#ifdef CONFIG_STACKTRACE
35a35046 3369 ONE("stack", S_IRUSR, proc_pid_stack),
28a6d671 3370#endif
5968cece 3371#ifdef CONFIG_SCHED_INFO
f6e826ca 3372 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 3373#endif
9745512c 3374#ifdef CONFIG_LATENCYTOP
631f9c18 3375 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 3376#endif
8793d854 3377#ifdef CONFIG_PROC_PID_CPUSET
52de4779 3378 ONE("cpuset", S_IRUGO, proc_cpuset_show),
a424316c
PM		 3379#endif
3380#ifdef CONFIG_CGROUPS
006f4ac4 3381 ONE("cgroup", S_IRUGO, proc_cgroup_show),
28a6d671 3382#endif
6ba51e37 3383 ONE("oom_score", S_IRUGO, proc_oom_score),
fa0cbbf1 3384 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
a63d83f4 3385 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 3386#ifdef CONFIG_AUDITSYSCALL
631f9c18 3387 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
26ec3c64 3388 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 3389#endif
f4f154fd 3390#ifdef CONFIG_FAULT_INJECTION
631f9c18 3391 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
1203c8e6 3392 REG("fail-nth", 0644, proc_fail_nth_operations),
f4f154fd 3393#endif
297c5d92 3394#ifdef CONFIG_TASK_IO_ACCOUNTING
19aadc98 3395 ONE("io", S_IRUSR, proc_tid_io_accounting),
297c5d92 3396#endif
22d917d8
EB		 3397#ifdef CONFIG_USER_NS
3398 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3399 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
f76d207a 3400 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
9cc46516 3401 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
22d917d8 3402#endif
7c23b330
JP		 3403#ifdef CONFIG_LIVEPATCH
3404 ONE("patch_state", S_IRUSR, proc_pid_patch_state),
3405#endif
28a6d671
EB		 3406};
3407
f0c3b509 3408static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
28a6d671 3409{
f0c3b509
AV		 3410 return proc_pident_readdir(file, ctx,
3411 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
28a6d671
EB		 3412}
3413
00cd8dd3
AV		 3414static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3415{
7bcd6b0e
EB		 3416 return proc_pident_lookup(dir, dentry,
3417 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
28a6d671
EB		 3418}
3419
00977a59 3420static const struct file_operations proc_tid_base_operations = {
28a6d671 3421 .read = generic_read_dir,
f50752ea
AV		 3422 .iterate_shared = proc_tid_base_readdir,
3423 .llseek = generic_file_llseek,
28a6d671
EB		 3424};
3425
c5ef1c42 3426static const struct inode_operations proc_tid_base_inode_operations = {
28a6d671
EB		 3427 .lookup = proc_tid_base_lookup,
3428 .getattr = pid_getattr,
3429 .setattr = proc_setattr,
3430};
3431
c52a47ac 3432static int proc_task_instantiate(struct inode *dir,
c5141e6d 3433 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8 3434{
444ceed8 3435 struct inode *inode;
db978da8 3436 inode = proc_pid_make_inode(dir->i_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
444ceed8
EB		 3437
3438 if (!inode)
3439 goto out;
444ceed8
EB		 3440 inode->i_op = &proc_tid_base_inode_operations;
3441 inode->i_fop = &proc_tid_base_operations;
3442 inode->i_flags|=S_IMMUTABLE;
aed54175 3443
1270dd8d 3444 set_nlink(inode, nlink_tid);
444ceed8 3445
fb045adb 3446 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 3447
3448 d_add(dentry, inode);
3449 /* Close the race of the process dying before we return the dentry */
0b728e19 3450 if (pid_revalidate(dentry, 0))
c52a47ac 3451 return 0;
444ceed8 3452out:
c52a47ac 3453 return -ENOENT;
444ceed8
EB		 3454}
3455
00cd8dd3 3456static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
28a6d671 3457{
c52a47ac 3458 int result = -ENOENT;
28a6d671
EB		 3459 struct task_struct *task;
3460 struct task_struct *leader = get_proc_task(dir);
28a6d671 3461 unsigned tid;
b488893a 3462 struct pid_namespace *ns;
28a6d671
EB		 3463
3464 if (!leader)
3465 goto out_no_task;
3466
dbcdb504 3467 tid = name_to_int(&dentry->d_name);
28a6d671
EB		 3468 if (tid == ~0U)
3469 goto out;
3470
b488893a 3471 ns = dentry->d_sb->s_fs_info;
28a6d671 3472 rcu_read_lock();
b488893a 3473 task = find_task_by_pid_ns(tid, ns);
28a6d671
EB		 3474 if (task)
3475 get_task_struct(task);
3476 rcu_read_unlock();
3477 if (!task)
3478 goto out;
bac0abd6 3479 if (!same_thread_group(leader, task))
28a6d671
EB		 3480 goto out_drop_task;
3481
444ceed8 3482 result = proc_task_instantiate(dir, dentry, task, NULL);
28a6d671
EB		 3483out_drop_task:
3484 put_task_struct(task);
3485out:
3486 put_task_struct(leader);
3487out_no_task:
c52a47ac 3488 return ERR_PTR(result);
28a6d671
EB		 3489}
3490
0bc58a91
EB		 3491/*
3492 * Find the first tid of a thread group to return to user space.
3493 *
3494 * Usually this is just the thread group leader, but if the users
3495 * buffer was too small or there was a seek into the middle of the
3496 * directory we have more work todo.
3497 *
3498 * In the case of a short read we start with find_task_by_pid.
3499 *
3500 * In the case of a seek we start with the leader and walk nr
3501 * threads past it.
3502 */
9f6e963f
ON		 3503static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3504 struct pid_namespace *ns)
0bc58a91 3505{
d855a4b7 3506 struct task_struct *pos, *task;
9f6e963f
ON		 3507 unsigned long nr = f_pos;
3508
3509 if (nr != f_pos) /* 32bit overflow? */
3510 return NULL;
1da177e4 3511
cc288738 3512 rcu_read_lock();
d855a4b7
ON		 3513 task = pid_task(pid, PIDTYPE_PID);
3514 if (!task)
3515 goto fail;
3516
3517 /* Attempt to start with the tid of a thread */
9f6e963f 3518 if (tid && nr) {
b488893a 3519 pos = find_task_by_pid_ns(tid, ns);
d855a4b7 3520 if (pos && same_thread_group(pos, task))
a872ff0c 3521 goto found;
0bc58a91 3522 }
1da177e4 3523
0bc58a91 3524 /* If nr exceeds the number of threads there is nothing todo */
9f6e963f 3525 if (nr >= get_nr_threads(task))
c986c14a 3526 goto fail;
1da177e4 3527
a872ff0c
ON		 3528 /* If we haven't found our starting place yet start
3529 * with the leader and walk nr threads forward.
0bc58a91 3530 */
d855a4b7 3531 pos = task = task->group_leader;
c986c14a 3532 do {
9f6e963f 3533 if (!nr--)
c986c14a 3534 goto found;
d855a4b7 3535 } while_each_thread(task, pos);
c986c14a
ON		 3536fail:
3537 pos = NULL;
3538 goto out;
a872ff0c
ON		 3539found:
3540 get_task_struct(pos);
3541out:
cc288738 3542 rcu_read_unlock();
0bc58a91
EB		 3543 return pos;
3544}
3545
3546/*
3547 * Find the next thread in the thread list.
3548 * Return NULL if there is an error or no next thread.
3549 *
3550 * The reference to the input task_struct is released.
3551 */
3552static struct task_struct *next_tid(struct task_struct *start)
3553{
c1df7fb8 3554 struct task_struct *pos = NULL;
cc288738 3555 rcu_read_lock();
c1df7fb8 3556 if (pid_alive(start)) {
0bc58a91 3557 pos = next_thread(start);
c1df7fb8
ON		 3558 if (thread_group_leader(pos))
3559 pos = NULL;
3560 else
3561 get_task_struct(pos);
3562 }
cc288738 3563 rcu_read_unlock();
0bc58a91
EB		 3564 put_task_struct(start);
3565 return pos;
1da177e4
LT		 3566}
3567
3568/* for the /proc/TGID/task/ directories */
f0c3b509 3569static int proc_task_readdir(struct file *file, struct dir_context *ctx)
1da177e4 3570{
d855a4b7
ON		 3571 struct inode *inode = file_inode(file);
3572 struct task_struct *task;
b488893a 3573 struct pid_namespace *ns;
f0c3b509 3574 int tid;
1da177e4 3575
d855a4b7 3576 if (proc_inode_is_dead(inode))
f0c3b509 3577 return -ENOENT;
1da177e4 3578
f0c3b509 3579 if (!dir_emit_dots(file, ctx))
d855a4b7 3580 return 0;
1da177e4 3581
0bc58a91
EB		 3582 /* f_version caches the tgid value that the last readdir call couldn't
3583 * return. lseek aka telldir automagically resets f_version to 0.
3584 */
3aa3377f 3585 ns = inode->i_sb->s_fs_info;
f0c3b509
AV		 3586 tid = (int)file->f_version;
3587 file->f_version = 0;
d855a4b7 3588 for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
0bc58a91 3589 task;
f0c3b509 3590 task = next_tid(task), ctx->pos++) {
e3912ac3 3591 char name[10 + 1];
f0c3b509 3592 int len;
b488893a 3593 tid = task_pid_nr_ns(task, ns);
e3912ac3 3594 len = snprintf(name, sizeof(name), "%u", tid);
f0c3b509
AV		 3595 if (!proc_fill_cache(file, ctx, name, len,
3596 proc_task_instantiate, task, NULL)) {
0bc58a91
EB		 3597 /* returning this tgid failed, save it as the first
3598 * pid for the next readir call */
f0c3b509 3599 file->f_version = (u64)tid;
0bc58a91 3600 put_task_struct(task);
1da177e4 3601 break;
0bc58a91 3602 }
1da177e4 3603 }
d855a4b7 3604
f0c3b509 3605 return 0;
1da177e4 3606}
6e66b52b 3607
a528d35e
DH		 3608static int proc_task_getattr(const struct path *path, struct kstat *stat,
3609 u32 request_mask, unsigned int query_flags)
6e66b52b 3610{
a528d35e 3611 struct inode *inode = d_inode(path->dentry);
99f89551 3612 struct task_struct *p = get_proc_task(inode);
6e66b52b
EB		 3613 generic_fillattr(inode, stat);
3614
99f89551 3615 if (p) {
99f89551 3616 stat->nlink += get_nr_threads(p);
99f89551 3617 put_task_struct(p);
6e66b52b
EB		 3618 }
3619
3620 return 0;
3621}
28a6d671 3622
c5ef1c42 3623static const struct inode_operations proc_task_inode_operations = {
28a6d671
EB		 3624 .lookup = proc_task_lookup,
3625 .getattr = proc_task_getattr,
3626 .setattr = proc_setattr,
0499680a 3627 .permission = proc_pid_permission,
28a6d671
EB		 3628};
3629
00977a59 3630static const struct file_operations proc_task_operations = {
28a6d671 3631 .read = generic_read_dir,
f50752ea
AV		 3632 .iterate_shared = proc_task_readdir,
3633 .llseek = generic_file_llseek,
28a6d671 3634};
1270dd8d
AD		 3635
3636void __init set_proc_pid_nlink(void)
3637{
3638 nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3639 nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3640}
Linux 6.x block layer and io_uring tree(s)