projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
proc, oom: drop bogus sighand lock
[linux-block.git] / fs / proc / base.c
CommitLineData
1da177e4
LT		 1/*
2 * linux/fs/proc/base.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 *
6 * proc base directory handling functions
7 *
8 * 1999, Al Viro. Rewritten. Now it covers the whole per-process part.
9 * Instead of using magical inumbers to determine the kind of object
10 * we allocate and fill in-core inodes upon lookup. They don't even
11 * go into icache. We cache the reference to task_struct upon lookup too.
12 * Eventually it should become a filesystem in its own. We don't use the
13 * rest of procfs anymore.
e070ad49
ML		 14 *
15 *
16 * Changelog:
17 * 17-Jan-2005
18 * Allan Bezerra
19 * Bruna Moreira <bruna.moreira@indt.org.br>
20 * Edjard Mota <edjard.mota@indt.org.br>
21 * Ilias Biris <ilias.biris@indt.org.br>
22 * Mauricio Lin <mauricio.lin@indt.org.br>
23 *
24 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
25 *
26 * A new process specific entry (smaps) included in /proc. It shows the
27 * size of rss for each memory area. The maps entry lacks information
28 * about physical memory size (rss) for each mapped file, i.e.,
29 * rss information for executables and library files.
30 * This additional information is useful for any tools that need to know
31 * about physical memory consumption for a process specific library.
32 *
33 * Changelog:
34 * 21-Feb-2005
35 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
36 * Pud inclusion in the page table walking.
37 *
38 * ChangeLog:
39 * 10-Mar-2005
40 * 10LE Instituto Nokia de Tecnologia - INdT:
41 * A better way to walks through the page table as suggested by Hugh Dickins.
42 *
43 * Simo Piiroinen <simo.piiroinen@nokia.com>:
44 * Smaps information related to shared, private, clean and dirty pages.
45 *
46 * Paul Mundt <paul.mundt@nokia.com>:
47 * Overall revision about smaps.
1da177e4
LT		 48 */
49
50#include <asm/uaccess.h>
51
1da177e4
LT		 52#include <linux/errno.h>
53#include <linux/time.h>
54#include <linux/proc_fs.h>
55#include <linux/stat.h>
5995477a 56#include <linux/task_io_accounting_ops.h>
1da177e4 57#include <linux/init.h>
16f7e0fe 58#include <linux/capability.h>
1da177e4 59#include <linux/file.h>
9f3acc31 60#include <linux/fdtable.h>
1da177e4
LT		 61#include <linux/string.h>
62#include <linux/seq_file.h>
63#include <linux/namei.h>
6b3286ed 64#include <linux/mnt_namespace.h>
1da177e4 65#include <linux/mm.h>
a63d83f4 66#include <linux/swap.h>
b835996f 67#include <linux/rcupdate.h>
1da177e4 68#include <linux/kallsyms.h>
2ec220e2 69#include <linux/stacktrace.h>
d85f50d5 70#include <linux/resource.h>
5096add8 71#include <linux/module.h>
1da177e4
LT		 72#include <linux/mount.h>
73#include <linux/security.h>
74#include <linux/ptrace.h>
0d094efe 75#include <linux/tracehook.h>
87ebdc00 76#include <linux/printk.h>
a424316c 77#include <linux/cgroup.h>
1da177e4
LT		 78#include <linux/cpuset.h>
79#include <linux/audit.h>
5addc5dd 80#include <linux/poll.h>
1651e14e 81#include <linux/nsproxy.h>
8ac773b4 82#include <linux/oom.h>
3cb4a0bb 83#include <linux/elf.h>
60347f67 84#include <linux/pid_namespace.h>
22d917d8 85#include <linux/user_namespace.h>
5ad4e53b 86#include <linux/fs_struct.h>
5a0e3ad6 87#include <linux/slab.h>
640708a2 88#include <linux/flex_array.h>
48f6a7a5 89#include <linux/posix-timers.h>
f133ecca
CM		 90#ifdef CONFIG_HARDWALL
91#include <asm/hardwall.h>
92#endif
43d2b113 93#include <trace/events/oom.h>
1da177e4 94#include "internal.h"
faf60af1 95#include "fd.h"
1da177e4 96
0f2fe20f
EB		 97/* NOTE:
98 * Implementing inode permission operations in /proc is almost
99 * certainly an error. Permission checks need to happen during
100 * each system call not at open time. The reason is that most of
101 * what we wish to check for permissions in /proc varies at runtime.
102 *
103 * The classic example of a problem is opening file descriptors
104 * in /proc for a task before it execs a suid executable.
105 */
106
1da177e4 107struct pid_entry {
cedbccab 108 const char *name;
c5141e6d 109 int len;
d161a13f 110 umode_t mode;
c5ef1c42 111 const struct inode_operations *iop;
00977a59 112 const struct file_operations *fop;
20cdc894 113 union proc_op op;
1da177e4
LT		 114};
115
61a28784 116#define NOD(NAME, MODE, IOP, FOP, OP) { \
20cdc894 117 .name = (NAME), \
c5141e6d 118 .len = sizeof(NAME) - 1, \
20cdc894
EB		 119 .mode = MODE, \
120 .iop = IOP, \
121 .fop = FOP, \
122 .op = OP, \
123}
124
631f9c18
AD		 125#define DIR(NAME, MODE, iops, fops) \
126 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
127#define LNK(NAME, get_link) \
61a28784 128 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
20cdc894 129 &proc_pid_link_inode_operations, NULL, \
631f9c18
AD		 130 { .proc_get_link = get_link } )
131#define REG(NAME, MODE, fops) \
132 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
631f9c18 133#define ONE(NAME, MODE, show) \
be614086
EB		 134 NOD(NAME, (S_IFREG|(MODE)), \
135 NULL, &proc_single_file_operations, \
631f9c18 136 { .proc_show = show } )
1da177e4 137
aed54175
VN		 138/*
139 * Count the number of hardlinks for the pid_entry table, excluding the .
140 * and .. links.
141 */
142static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
143 unsigned int n)
144{
145 unsigned int i;
146 unsigned int count;
147
148 count = 0;
149 for (i = 0; i < n; ++i) {
150 if (S_ISDIR(entries[i].mode))
151 ++count;
152 }
153
154 return count;
155}
156
f7ad3c6b 157static int get_task_root(struct task_struct *task, struct path *root)
1da177e4 158{
7c2c7d99
HD		 159 int result = -ENOENT;
160
0494f6ec 161 task_lock(task);
f7ad3c6b
MS		 162 if (task->fs) {
163 get_fs_root(task->fs, root);
7c2c7d99
HD		 164 result = 0;
165 }
0494f6ec 166 task_unlock(task);
7c2c7d99 167 return result;
0494f6ec
MS		 168}
169
7773fbc5 170static int proc_cwd_link(struct dentry *dentry, struct path *path)
0494f6ec 171{
2b0143b5 172 struct task_struct *task = get_proc_task(d_inode(dentry));
0494f6ec 173 int result = -ENOENT;
99f89551
EB		 174
175 if (task) {
f7ad3c6b
MS		 176 task_lock(task);
177 if (task->fs) {
178 get_fs_pwd(task->fs, path);
179 result = 0;
180 }
181 task_unlock(task);
99f89551
EB		 182 put_task_struct(task);
183 }
1da177e4
LT		 184 return result;
185}
186
7773fbc5 187static int proc_root_link(struct dentry *dentry, struct path *path)
1da177e4 188{
2b0143b5 189 struct task_struct *task = get_proc_task(d_inode(dentry));
1da177e4 190 int result = -ENOENT;
99f89551
EB		 191
192 if (task) {
f7ad3c6b 193 result = get_task_root(task, path);
99f89551
EB		 194 put_task_struct(task);
195 }
1da177e4
LT		 196 return result;
197}
198
c2c0bb44
AD		 199static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,
200 size_t _count, loff_t *pos)
1da177e4 201{
c2c0bb44
AD		 202 struct task_struct *tsk;
203 struct mm_struct *mm;
204 char *page;
205 unsigned long count = _count;
206 unsigned long arg_start, arg_end, env_start, env_end;
207 unsigned long len1, len2, len;
208 unsigned long p;
209 char c;
210 ssize_t rv;
211
212 BUG_ON(*pos < 0);
213
214 tsk = get_proc_task(file_inode(file));
215 if (!tsk)
216 return -ESRCH;
217 mm = get_task_mm(tsk);
218 put_task_struct(tsk);
219 if (!mm)
220 return 0;
221 /* Check if process spawned far enough to have cmdline. */
222 if (!mm->env_end) {
223 rv = 0;
224 goto out_mmput;
225 }
226
227 page = (char *)__get_free_page(GFP_TEMPORARY);
228 if (!page) {
229 rv = -ENOMEM;
230 goto out_mmput;
231 }
232
233 down_read(&mm->mmap_sem);
234 arg_start = mm->arg_start;
235 arg_end = mm->arg_end;
236 env_start = mm->env_start;
237 env_end = mm->env_end;
238 up_read(&mm->mmap_sem);
239
240 BUG_ON(arg_start > arg_end);
241 BUG_ON(env_start > env_end);
242
243 len1 = arg_end - arg_start;
244 len2 = env_end - env_start;
245
3581d458
AD		 246 /* Empty ARGV. */
247 if (len1 == 0) {
248 rv = 0;
249 goto out_free_page;
250 }
2ca66ff7 251 /*
c2c0bb44
AD		 252 * Inherently racy -- command line shares address space
253 * with code and data.
2ca66ff7 254 */
c2c0bb44
AD		 255 rv = access_remote_vm(mm, arg_end - 1, &c, 1, 0);
256 if (rv <= 0)
257 goto out_free_page;
258
259 rv = 0;
260
261 if (c == '\0') {
262 /* Command line (set of strings) occupies whole ARGV. */
263 if (len1 <= *pos)
264 goto out_free_page;
265
266 p = arg_start + *pos;
267 len = len1 - *pos;
268 while (count > 0 && len > 0) {
269 unsigned int _count;
270 int nr_read;
271
272 _count = min3(count, len, PAGE_SIZE);
273 nr_read = access_remote_vm(mm, p, page, _count, 0);
274 if (nr_read < 0)
275 rv = nr_read;
276 if (nr_read <= 0)
277 goto out_free_page;
278
279 if (copy_to_user(buf, page, nr_read)) {
280 rv = -EFAULT;
281 goto out_free_page;
282 }
283
284 p += nr_read;
285 len -= nr_read;
286 buf += nr_read;
287 count -= nr_read;
288 rv += nr_read;
289 }
290 } else {
291 /*
292 * Command line (1 string) occupies ARGV and maybe
293 * extends into ENVP.
294 */
295 if (len1 + len2 <= *pos)
296 goto skip_argv_envp;
297 if (len1 <= *pos)
298 goto skip_argv;
299
300 p = arg_start + *pos;
301 len = len1 - *pos;
302 while (count > 0 && len > 0) {
303 unsigned int _count, l;
304 int nr_read;
305 bool final;
306
307 _count = min3(count, len, PAGE_SIZE);
308 nr_read = access_remote_vm(mm, p, page, _count, 0);
309 if (nr_read < 0)
310 rv = nr_read;
311 if (nr_read <= 0)
312 goto out_free_page;
313
314 /*
315 * Command line can be shorter than whole ARGV
316 * even if last "marker" byte says it is not.
317 */
318 final = false;
319 l = strnlen(page, nr_read);
320 if (l < nr_read) {
321 nr_read = l;
322 final = true;
323 }
324
325 if (copy_to_user(buf, page, nr_read)) {
326 rv = -EFAULT;
327 goto out_free_page;
328 }
329
330 p += nr_read;
331 len -= nr_read;
332 buf += nr_read;
333 count -= nr_read;
334 rv += nr_read;
335
336 if (final)
337 goto out_free_page;
338 }
339skip_argv:
340 /*
341 * Command line (1 string) occupies ARGV and
342 * extends into ENVP.
343 */
344 if (len1 <= *pos) {
345 p = env_start + *pos - len1;
346 len = len1 + len2 - *pos;
347 } else {
348 p = env_start;
349 len = len2;
350 }
351 while (count > 0 && len > 0) {
352 unsigned int _count, l;
353 int nr_read;
354 bool final;
355
356 _count = min3(count, len, PAGE_SIZE);
357 nr_read = access_remote_vm(mm, p, page, _count, 0);
358 if (nr_read < 0)
359 rv = nr_read;
360 if (nr_read <= 0)
361 goto out_free_page;
362
363 /* Find EOS. */
364 final = false;
365 l = strnlen(page, nr_read);
366 if (l < nr_read) {
367 nr_read = l;
368 final = true;
369 }
370
371 if (copy_to_user(buf, page, nr_read)) {
372 rv = -EFAULT;
373 goto out_free_page;
374 }
375
376 p += nr_read;
377 len -= nr_read;
378 buf += nr_read;
379 count -= nr_read;
380 rv += nr_read;
381
382 if (final)
383 goto out_free_page;
384 }
385skip_argv_envp:
386 ;
387 }
388
389out_free_page:
390 free_page((unsigned long)page);
391out_mmput:
392 mmput(mm);
393 if (rv > 0)
394 *pos += rv;
395 return rv;
1da177e4
LT		 396}
397
c2c0bb44
AD		 398static const struct file_operations proc_pid_cmdline_ops = {
399 .read = proc_pid_cmdline_read,
400 .llseek = generic_file_llseek,
401};
402
f9ea536e
AD		 403static int proc_pid_auxv(struct seq_file *m, struct pid_namespace *ns,
404 struct pid *pid, struct task_struct *task)
1da177e4 405{
caaee623 406 struct mm_struct *mm = mm_access(task, PTRACE_MODE_READ_FSCREDS);
2fadaef4 407 if (mm && !IS_ERR(mm)) {
1da177e4 408 unsigned int nwords = 0;
dfe6b7d9 409 do {
1da177e4 410 nwords += 2;
dfe6b7d9 411 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
f9ea536e 412 seq_write(m, mm->saved_auxv, nwords * sizeof(mm->saved_auxv[0]));
1da177e4 413 mmput(mm);
f9ea536e
AD		 414 return 0;
415 } else
416 return PTR_ERR(mm);
1da177e4
LT		 417}
418
419
420#ifdef CONFIG_KALLSYMS
421/*
422 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
423 * Returns the resolved symbol. If that fails, simply return the address.
424 */
edfcd606
AD		 425static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns,
426 struct pid *pid, struct task_struct *task)
1da177e4 427{
ffb45122 428 unsigned long wchan;
9281acea 429 char symname[KSYM_NAME_LEN];
1da177e4
LT		 430
431 wchan = get_wchan(task);
432
caaee623
JH		 433 if (wchan && ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)
434 && !lookup_symbol_name(wchan, symname))
25ce3191 435 seq_printf(m, "%s", symname);
b2f73922 436 else
1e92a61c 437 seq_putc(m, '0');
25ce3191
JP		 438
439 return 0;
1da177e4
LT		 440}
441#endif /* CONFIG_KALLSYMS */
442
a9712bc1
AV		 443static int lock_trace(struct task_struct *task)
444{
445 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
446 if (err)
447 return err;
caaee623 448 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) {
a9712bc1
AV		 449 mutex_unlock(&task->signal->cred_guard_mutex);
450 return -EPERM;
451 }
452 return 0;
453}
454
455static void unlock_trace(struct task_struct *task)
456{
457 mutex_unlock(&task->signal->cred_guard_mutex);
458}
459
2ec220e2
KC		 460#ifdef CONFIG_STACKTRACE
461
462#define MAX_STACK_TRACE_DEPTH 64
463
464static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
465 struct pid *pid, struct task_struct *task)
466{
467 struct stack_trace trace;
468 unsigned long *entries;
a9712bc1 469 int err;
2ec220e2
KC		 470 int i;
471
472 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
473 if (!entries)
474 return -ENOMEM;
475
476 trace.nr_entries = 0;
477 trace.max_entries = MAX_STACK_TRACE_DEPTH;
478 trace.entries = entries;
479 trace.skip = 0;
2ec220e2 480
a9712bc1
AV		 481 err = lock_trace(task);
482 if (!err) {
483 save_stack_trace_tsk(task, &trace);
484
485 for (i = 0; i < trace.nr_entries; i++) {
b81a618d 486 seq_printf(m, "[<%pK>] %pS\n",
a9712bc1
AV		 487 (void *)entries[i], (void *)entries[i]);
488 }
489 unlock_trace(task);
2ec220e2
KC		 490 }
491 kfree(entries);
492
a9712bc1 493 return err;
2ec220e2
KC		 494}
495#endif
496
5968cece 497#ifdef CONFIG_SCHED_INFO
1da177e4
LT		 498/*
499 * Provides /proc/PID/schedstat
500 */
f6e826ca
AD		 501static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
502 struct pid *pid, struct task_struct *task)
1da177e4 503{
5968cece
NR		 504 if (unlikely(!sched_info_on()))
505 seq_printf(m, "0 0 0\n");
506 else
507 seq_printf(m, "%llu %llu %lu\n",
25ce3191
JP		 508 (unsigned long long)task->se.sum_exec_runtime,
509 (unsigned long long)task->sched_info.run_delay,
510 task->sched_info.pcount);
511
512 return 0;
1da177e4
LT		 513}
514#endif
515
9745512c
AV		 516#ifdef CONFIG_LATENCYTOP
517static int lstats_show_proc(struct seq_file *m, void *v)
518{
519 int i;
13d77c37
HS		 520 struct inode *inode = m->private;
521 struct task_struct *task = get_proc_task(inode);
9745512c 522
13d77c37
HS		 523 if (!task)
524 return -ESRCH;
525 seq_puts(m, "Latency Top version : v0.1\n");
9745512c 526 for (i = 0; i < 32; i++) {
34e49d4f
JP		 527 struct latency_record *lr = &task->latency_record[i];
528 if (lr->backtrace[0]) {
9745512c 529 int q;
34e49d4f
JP		 530 seq_printf(m, "%i %li %li",
531 lr->count, lr->time, lr->max);
9745512c 532 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
34e49d4f
JP		 533 unsigned long bt = lr->backtrace[q];
534 if (!bt)
9745512c 535 break;
34e49d4f 536 if (bt == ULONG_MAX)
9745512c 537 break;
34e49d4f 538 seq_printf(m, " %ps", (void *)bt);
9745512c 539 }
9d6de12f 540 seq_putc(m, '\n');
9745512c
AV		 541 }
542
543 }
13d77c37 544 put_task_struct(task);
9745512c
AV		 545 return 0;
546}
547
548static int lstats_open(struct inode *inode, struct file *file)
549{
13d77c37 550 return single_open(file, lstats_show_proc, inode);
d6643d12
HS		 551}
552
9745512c
AV		 553static ssize_t lstats_write(struct file *file, const char __user *buf,
554 size_t count, loff_t *offs)
555{
496ad9aa 556 struct task_struct *task = get_proc_task(file_inode(file));
9745512c 557
13d77c37
HS		 558 if (!task)
559 return -ESRCH;
9745512c 560 clear_all_latency_tracing(task);
13d77c37 561 put_task_struct(task);
9745512c
AV		 562
563 return count;
564}
565
566static const struct file_operations proc_lstats_operations = {
567 .open = lstats_open,
568 .read = seq_read,
569 .write = lstats_write,
570 .llseek = seq_lseek,
13d77c37 571 .release = single_release,
9745512c
AV		 572};
573
574#endif
575
6ba51e37
AD		 576static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
577 struct pid *pid, struct task_struct *task)
1da177e4 578{
a7f638f9 579 unsigned long totalpages = totalram_pages + total_swap_pages;
b95c35e7 580 unsigned long points = 0;
1da177e4 581
19c5d45a 582 read_lock(&tasklist_lock);
b95c35e7 583 if (pid_alive(task))
a7f638f9
DR		 584 points = oom_badness(task, NULL, NULL, totalpages) *
585 1000 / totalpages;
19c5d45a 586 read_unlock(&tasklist_lock);
25ce3191
JP		 587 seq_printf(m, "%lu\n", points);
588
589 return 0;
1da177e4
LT		 590}
591
d85f50d5 592struct limit_names {
cedbccab
AD		 593 const char *name;
594 const char *unit;
d85f50d5
NH		 595};
596
597static const struct limit_names lnames[RLIM_NLIMITS] = {
cff4edb5 598 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
d85f50d5
NH		 599 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
600 [RLIMIT_DATA] = {"Max data size", "bytes"},
601 [RLIMIT_STACK] = {"Max stack size", "bytes"},
602 [RLIMIT_CORE] = {"Max core file size", "bytes"},
603 [RLIMIT_RSS] = {"Max resident set", "bytes"},
604 [RLIMIT_NPROC] = {"Max processes", "processes"},
605 [RLIMIT_NOFILE] = {"Max open files", "files"},
606 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
607 [RLIMIT_AS] = {"Max address space", "bytes"},
608 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
609 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
610 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
611 [RLIMIT_NICE] = {"Max nice priority", NULL},
612 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
8808117c 613 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
d85f50d5
NH		 614};
615
616/* Display limits for a process */
1c963eb1
AD		 617static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
618 struct pid *pid, struct task_struct *task)
d85f50d5
NH		 619{
620 unsigned int i;
d85f50d5 621 unsigned long flags;
d85f50d5
NH		 622
623 struct rlimit rlim[RLIM_NLIMITS];
624
a6bebbc8 625 if (!lock_task_sighand(task, &flags))
d85f50d5 626 return 0;
d85f50d5
NH		 627 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
628 unlock_task_sighand(task, &flags);
d85f50d5
NH		 629
630 /*
631 * print the file header
632 */
1c963eb1 633 seq_printf(m, "%-25s %-20s %-20s %-10s\n",
25ce3191 634 "Limit", "Soft Limit", "Hard Limit", "Units");
d85f50d5
NH		 635
636 for (i = 0; i < RLIM_NLIMITS; i++) {
637 if (rlim[i].rlim_cur == RLIM_INFINITY)
1c963eb1 638 seq_printf(m, "%-25s %-20s ",
25ce3191 639 lnames[i].name, "unlimited");
d85f50d5 640 else
1c963eb1 641 seq_printf(m, "%-25s %-20lu ",
25ce3191 642 lnames[i].name, rlim[i].rlim_cur);
d85f50d5
NH		 643
644 if (rlim[i].rlim_max == RLIM_INFINITY)
1c963eb1 645 seq_printf(m, "%-20s ", "unlimited");
d85f50d5 646 else
1c963eb1 647 seq_printf(m, "%-20lu ", rlim[i].rlim_max);
d85f50d5
NH		 648
649 if (lnames[i].unit)
1c963eb1 650 seq_printf(m, "%-10s\n", lnames[i].unit);
d85f50d5 651 else
1c963eb1 652 seq_putc(m, '\n');
d85f50d5
NH		 653 }
654
1c963eb1 655 return 0;
d85f50d5
NH		 656}
657
ebcb6734 658#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
09d93bd6
AD		 659static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
660 struct pid *pid, struct task_struct *task)
ebcb6734
RM		 661{
662 long nr;
663 unsigned long args[6], sp, pc;
25ce3191
JP		 664 int res;
665
666 res = lock_trace(task);
a9712bc1
AV		 667 if (res)
668 return res;
ebcb6734
RM		 669
670 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
09d93bd6 671 seq_puts(m, "running\n");
a9712bc1 672 else if (nr < 0)
09d93bd6 673 seq_printf(m, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
a9712bc1 674 else
09d93bd6 675 seq_printf(m,
ebcb6734
RM		 676 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
677 nr,
678 args[0], args[1], args[2], args[3], args[4], args[5],
679 sp, pc);
a9712bc1 680 unlock_trace(task);
25ce3191
JP		 681
682 return 0;
ebcb6734
RM		 683}
684#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
685
1da177e4
LT		 686/************************************************************************/
687/* Here the fs part begins */
688/************************************************************************/
689
690/* permission checks */
778c1144 691static int proc_fd_access_allowed(struct inode *inode)
1da177e4 692{
778c1144
EB		 693 struct task_struct *task;
694 int allowed = 0;
df26c40e
EB		 695 /* Allow access to a task's file descriptors if it is us or we
696 * may use ptrace attach to the process and find out that
697 * information.
778c1144
EB		 698 */
699 task = get_proc_task(inode);
df26c40e 700 if (task) {
caaee623 701 allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
778c1144 702 put_task_struct(task);
df26c40e 703 }
778c1144 704 return allowed;
1da177e4
LT		 705}
706
6b4e306a 707int proc_setattr(struct dentry *dentry, struct iattr *attr)
6d76fa58
LT		 708{
709 int error;
2b0143b5 710 struct inode *inode = d_inode(dentry);
6d76fa58
LT		 711
712 if (attr->ia_valid & ATTR_MODE)
713 return -EPERM;
714
715 error = inode_change_ok(inode, attr);
1025774c
CH		 716 if (error)
717 return error;
718
1025774c
CH		 719 setattr_copy(inode, attr);
720 mark_inode_dirty(inode);
721 return 0;
6d76fa58
LT		 722}
723
0499680a
VK		 724/*
725 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
726 * or euid/egid (for hide_pid_min=2)?
727 */
728static bool has_pid_permissions(struct pid_namespace *pid,
729 struct task_struct *task,
730 int hide_pid_min)
731{
732 if (pid->hide_pid < hide_pid_min)
733 return true;
734 if (in_group_p(pid->pid_gid))
735 return true;
caaee623 736 return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
0499680a
VK		 737}
738
739
740static int proc_pid_permission(struct inode *inode, int mask)
741{
742 struct pid_namespace *pid = inode->i_sb->s_fs_info;
743 struct task_struct *task;
744 bool has_perms;
745
746 task = get_proc_task(inode);
a2ef990a
XF		 747 if (!task)
748 return -ESRCH;
0499680a
VK		 749 has_perms = has_pid_permissions(pid, task, 1);
750 put_task_struct(task);
751
752 if (!has_perms) {
753 if (pid->hide_pid == 2) {
754 /*
755 * Let's make getdents(), stat(), and open()
756 * consistent with each other. If a process
757 * may not stat() a file, it shouldn't be seen
758 * in procfs at all.
759 */
760 return -ENOENT;
761 }
762
763 return -EPERM;
764 }
765 return generic_permission(inode, mask);
766}
767
768
769
c5ef1c42 770static const struct inode_operations proc_def_inode_operations = {
6d76fa58
LT		 771 .setattr = proc_setattr,
772};
773
be614086
EB		 774static int proc_single_show(struct seq_file *m, void *v)
775{
776 struct inode *inode = m->private;
777 struct pid_namespace *ns;
778 struct pid *pid;
779 struct task_struct *task;
780 int ret;
781
782 ns = inode->i_sb->s_fs_info;
783 pid = proc_pid(inode);
784 task = get_pid_task(pid, PIDTYPE_PID);
785 if (!task)
786 return -ESRCH;
787
788 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
789
790 put_task_struct(task);
791 return ret;
792}
793
794static int proc_single_open(struct inode *inode, struct file *filp)
795{
c6a34058 796 return single_open(filp, proc_single_show, inode);
be614086
EB		 797}
798
799static const struct file_operations proc_single_file_operations = {
800 .open = proc_single_open,
801 .read = seq_read,
802 .llseek = seq_lseek,
803 .release = single_release,
804};
805
5381e169
ON		 806
807struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
1da177e4 808{
5381e169
ON		 809 struct task_struct *task = get_proc_task(inode);
810 struct mm_struct *mm = ERR_PTR(-ESRCH);
e268337d 811
5381e169 812 if (task) {
caaee623 813 mm = mm_access(task, mode | PTRACE_MODE_FSCREDS);
5381e169 814 put_task_struct(task);
e268337d 815
5381e169
ON		 816 if (!IS_ERR_OR_NULL(mm)) {
817 /* ensure this mm_struct can't be freed */
818 atomic_inc(&mm->mm_count);
819 /* but do not pin its memory */
820 mmput(mm);
821 }
822 }
823
824 return mm;
825}
826
827static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
828{
829 struct mm_struct *mm = proc_mem_open(inode, mode);
e268337d
LT		 830
831 if (IS_ERR(mm))
832 return PTR_ERR(mm);
833
e268337d 834 file->private_data = mm;
1da177e4
LT		 835 return 0;
836}
837
b409e578
CW		 838static int mem_open(struct inode *inode, struct file *file)
839{
bc452b4b
DH		 840 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
841
842 /* OK to pass negative loff_t, we can catch out-of-range */
843 file->f_mode |= FMODE_UNSIGNED_OFFSET;
844
845 return ret;
b409e578
CW		 846}
847
572d34b9
ON		 848static ssize_t mem_rw(struct file *file, char __user *buf,
849 size_t count, loff_t *ppos, int write)
1da177e4 850{
e268337d 851 struct mm_struct *mm = file->private_data;
572d34b9
ON		 852 unsigned long addr = *ppos;
853 ssize_t copied;
1da177e4 854 char *page;
1da177e4 855
e268337d
LT		 856 if (!mm)
857 return 0;
99f89551 858
30cd8903
KM		 859 page = (char *)__get_free_page(GFP_TEMPORARY);
860 if (!page)
e268337d 861 return -ENOMEM;
1da177e4 862
f7ca54f4 863 copied = 0;
6d08f2c7
ON		 864 if (!atomic_inc_not_zero(&mm->mm_users))
865 goto free;
866
1da177e4 867 while (count > 0) {
572d34b9 868 int this_len = min_t(int, count, PAGE_SIZE);
1da177e4 869
572d34b9 870 if (write && copy_from_user(page, buf, this_len)) {
1da177e4
LT		 871 copied = -EFAULT;
872 break;
873 }
572d34b9
ON		 874
875 this_len = access_remote_vm(mm, addr, page, this_len, write);
876 if (!this_len) {
1da177e4
LT		 877 if (!copied)
878 copied = -EIO;
879 break;
880 }
572d34b9
ON		 881
882 if (!write && copy_to_user(buf, page, this_len)) {
883 copied = -EFAULT;
884 break;
885 }
886
887 buf += this_len;
888 addr += this_len;
889 copied += this_len;
890 count -= this_len;
1da177e4 891 }
572d34b9 892 *ppos = addr;
30cd8903 893
6d08f2c7
ON		 894 mmput(mm);
895free:
30cd8903 896 free_page((unsigned long) page);
1da177e4
LT		 897 return copied;
898}
1da177e4 899
572d34b9
ON		 900static ssize_t mem_read(struct file *file, char __user *buf,
901 size_t count, loff_t *ppos)
902{
903 return mem_rw(file, buf, count, ppos, 0);
904}
905
906static ssize_t mem_write(struct file *file, const char __user *buf,
907 size_t count, loff_t *ppos)
908{
909 return mem_rw(file, (char __user*)buf, count, ppos, 1);
910}
911
85863e47 912loff_t mem_lseek(struct file *file, loff_t offset, int orig)
1da177e4
LT		 913{
914 switch (orig) {
915 case 0:
916 file->f_pos = offset;
917 break;
918 case 1:
919 file->f_pos += offset;
920 break;
921 default:
922 return -EINVAL;
923 }
924 force_successful_syscall_return();
925 return file->f_pos;
926}
927
e268337d
LT		 928static int mem_release(struct inode *inode, struct file *file)
929{
930 struct mm_struct *mm = file->private_data;
71879d3c 931 if (mm)
6d08f2c7 932 mmdrop(mm);
e268337d
LT		 933 return 0;
934}
935
00977a59 936static const struct file_operations proc_mem_operations = {
1da177e4
LT		 937 .llseek = mem_lseek,
938 .read = mem_read,
939 .write = mem_write,
940 .open = mem_open,
e268337d 941 .release = mem_release,
1da177e4
LT		 942};
943
b409e578
CW		 944static int environ_open(struct inode *inode, struct file *file)
945{
946 return __mem_open(inode, file, PTRACE_MODE_READ);
947}
948
315e28c8
JP		 949static ssize_t environ_read(struct file *file, char __user *buf,
950 size_t count, loff_t *ppos)
951{
315e28c8
JP		 952 char *page;
953 unsigned long src = *ppos;
b409e578
CW		 954 int ret = 0;
955 struct mm_struct *mm = file->private_data;
a3b609ef 956 unsigned long env_start, env_end;
315e28c8 957
8148a73c
MK		 958 /* Ensure the process spawned far enough to have an environment. */
959 if (!mm || !mm->env_end)
b409e578 960 return 0;
315e28c8 961
315e28c8
JP		 962 page = (char *)__get_free_page(GFP_TEMPORARY);
963 if (!page)
b409e578 964 return -ENOMEM;
315e28c8 965
d6f64b89 966 ret = 0;
b409e578
CW		 967 if (!atomic_inc_not_zero(&mm->mm_users))
968 goto free;
a3b609ef
MG		 969
970 down_read(&mm->mmap_sem);
971 env_start = mm->env_start;
972 env_end = mm->env_end;
973 up_read(&mm->mmap_sem);
974
315e28c8 975 while (count > 0) {
e8905ec2
DH		 976 size_t this_len, max_len;
977 int retval;
315e28c8 978
a3b609ef 979 if (src >= (env_end - env_start))
315e28c8
JP		 980 break;
981
a3b609ef 982 this_len = env_end - (env_start + src);
e8905ec2
DH		 983
984 max_len = min_t(size_t, PAGE_SIZE, count);
985 this_len = min(max_len, this_len);
315e28c8 986
a3b609ef 987 retval = access_remote_vm(mm, (env_start + src),
315e28c8
JP		 988 page, this_len, 0);
989
990 if (retval <= 0) {
991 ret = retval;
992 break;
993 }
994
995 if (copy_to_user(buf, page, retval)) {
996 ret = -EFAULT;
997 break;
998 }
999
1000 ret += retval;
1001 src += retval;
1002 buf += retval;
1003 count -= retval;
1004 }
1005 *ppos = src;
315e28c8 1006 mmput(mm);
b409e578
CW		 1007
1008free:
315e28c8 1009 free_page((unsigned long) page);
315e28c8
JP		 1010 return ret;
1011}
1012
1013static const struct file_operations proc_environ_operations = {
b409e578 1014 .open = environ_open,
315e28c8 1015 .read = environ_read,
87df8424 1016 .llseek = generic_file_llseek,
b409e578 1017 .release = mem_release,
315e28c8
JP		 1018};
1019
fa0cbbf1
DR		 1020static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
1021 loff_t *ppos)
1022{
496ad9aa 1023 struct task_struct *task = get_proc_task(file_inode(file));
fa0cbbf1
DR		 1024 char buffer[PROC_NUMBUF];
1025 int oom_adj = OOM_ADJUST_MIN;
1026 size_t len;
fa0cbbf1
DR		 1027
1028 if (!task)
1029 return -ESRCH;
f913da59
MH		 1030 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
1031 oom_adj = OOM_ADJUST_MAX;
1032 else
1033 oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
1034 OOM_SCORE_ADJ_MAX;
fa0cbbf1
DR		 1035 put_task_struct(task);
1036 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
1037 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1038}
1039
f913da59
MH		 1040static DEFINE_MUTEX(oom_adj_mutex);
1041
b72bdfa7
DR		 1042/*
1043 * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1044 * kernels. The effective policy is defined by oom_score_adj, which has a
1045 * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1046 * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1047 * Processes that become oom disabled via oom_adj will still be oom disabled
1048 * with this implementation.
1049 *
1050 * oom_adj cannot be removed since existing userspace binaries use it.
1051 */
fa0cbbf1
DR		 1052static ssize_t oom_adj_write(struct file *file, const char __user *buf,
1053 size_t count, loff_t *ppos)
1054{
1055 struct task_struct *task;
1056 char buffer[PROC_NUMBUF];
1057 int oom_adj;
fa0cbbf1
DR		 1058 int err;
1059
1060 memset(buffer, 0, sizeof(buffer));
1061 if (count > sizeof(buffer) - 1)
1062 count = sizeof(buffer) - 1;
1063 if (copy_from_user(buffer, buf, count)) {
1064 err = -EFAULT;
1065 goto out;
1066 }
1067
1068 err = kstrtoint(strstrip(buffer), 0, &oom_adj);
1069 if (err)
1070 goto out;
1071 if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
1072 oom_adj != OOM_DISABLE) {
1073 err = -EINVAL;
1074 goto out;
1075 }
1076
496ad9aa 1077 task = get_proc_task(file_inode(file));
fa0cbbf1
DR		 1078 if (!task) {
1079 err = -ESRCH;
1080 goto out;
1081 }
1082
fa0cbbf1
DR		 1083 /*
1084 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1085 * value is always attainable.
1086 */
1087 if (oom_adj == OOM_ADJUST_MAX)
1088 oom_adj = OOM_SCORE_ADJ_MAX;
1089 else
1090 oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
1091
f913da59 1092 mutex_lock(&oom_adj_mutex);
fa0cbbf1
DR		 1093 if (oom_adj < task->signal->oom_score_adj &&
1094 !capable(CAP_SYS_RESOURCE)) {
1095 err = -EACCES;
f913da59 1096 goto err_unlock;
fa0cbbf1
DR		 1097 }
1098
1099 /*
1100 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1101 * /proc/pid/oom_score_adj instead.
1102 */
87ebdc00 1103 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
fa0cbbf1
DR		 1104 current->comm, task_pid_nr(current), task_pid_nr(task),
1105 task_pid_nr(task));
1106
1107 task->signal->oom_score_adj = oom_adj;
1108 trace_oom_score_adj_update(task);
f913da59
MH		 1109err_unlock:
1110 mutex_unlock(&oom_adj_mutex);
fa0cbbf1
DR		 1111 put_task_struct(task);
1112out:
1113 return err < 0 ? err : count;
1114}
1115
1116static const struct file_operations proc_oom_adj_operations = {
1117 .read = oom_adj_read,
1118 .write = oom_adj_write,
1119 .llseek = generic_file_llseek,
1120};
1121
a63d83f4
DR		 1122static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1123 size_t count, loff_t *ppos)
1124{
496ad9aa 1125 struct task_struct *task = get_proc_task(file_inode(file));
a63d83f4 1126 char buffer[PROC_NUMBUF];
a9c58b90 1127 short oom_score_adj = OOM_SCORE_ADJ_MIN;
a63d83f4
DR		 1128 size_t len;
1129
1130 if (!task)
1131 return -ESRCH;
f913da59 1132 oom_score_adj = task->signal->oom_score_adj;
a63d83f4 1133 put_task_struct(task);
a9c58b90 1134 len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
a63d83f4
DR		 1135 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1136}
1137
1138static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1139 size_t count, loff_t *ppos)
1140{
1141 struct task_struct *task;
1142 char buffer[PROC_NUMBUF];
0a8cb8e3 1143 int oom_score_adj;
a63d83f4
DR		 1144 int err;
1145
1146 memset(buffer, 0, sizeof(buffer));
1147 if (count > sizeof(buffer) - 1)
1148 count = sizeof(buffer) - 1;
723548bf
DR		 1149 if (copy_from_user(buffer, buf, count)) {
1150 err = -EFAULT;
1151 goto out;
1152 }
a63d83f4 1153
0a8cb8e3 1154 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
a63d83f4 1155 if (err)
723548bf 1156 goto out;
a63d83f4 1157 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
723548bf
DR		 1158 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1159 err = -EINVAL;
1160 goto out;
1161 }
a63d83f4 1162
496ad9aa 1163 task = get_proc_task(file_inode(file));
723548bf
DR		 1164 if (!task) {
1165 err = -ESRCH;
1166 goto out;
1167 }
d19d5476 1168
f913da59 1169 mutex_lock(&oom_adj_mutex);
a9c58b90 1170 if ((short)oom_score_adj < task->signal->oom_score_adj_min &&
a63d83f4 1171 !capable(CAP_SYS_RESOURCE)) {
723548bf 1172 err = -EACCES;
f913da59 1173 goto err_unlock;
a63d83f4
DR		 1174 }
1175
a9c58b90 1176 task->signal->oom_score_adj = (short)oom_score_adj;
dabb16f6 1177 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
a9c58b90 1178 task->signal->oom_score_adj_min = (short)oom_score_adj;
f913da59 1179
43d2b113 1180 trace_oom_score_adj_update(task);
01dc52eb 1181
f913da59
MH		 1182err_unlock:
1183 mutex_unlock(&oom_adj_mutex);
a63d83f4 1184 put_task_struct(task);
723548bf
DR		 1185out:
1186 return err < 0 ? err : count;
a63d83f4
DR		 1187}
1188
1189static const struct file_operations proc_oom_score_adj_operations = {
1190 .read = oom_score_adj_read,
1191 .write = oom_score_adj_write,
6038f373 1192 .llseek = default_llseek,
a63d83f4
DR		 1193};
1194
1da177e4
LT		 1195#ifdef CONFIG_AUDITSYSCALL
1196#define TMPBUFLEN 21
1197static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1198 size_t count, loff_t *ppos)
1199{
496ad9aa 1200 struct inode * inode = file_inode(file);
99f89551 1201 struct task_struct *task = get_proc_task(inode);
1da177e4
LT		 1202 ssize_t length;
1203 char tmpbuf[TMPBUFLEN];
1204
99f89551
EB		 1205 if (!task)
1206 return -ESRCH;
1da177e4 1207 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
e1760bd5
EB		 1208 from_kuid(file->f_cred->user_ns,
1209 audit_get_loginuid(task)));
99f89551 1210 put_task_struct(task);
1da177e4
LT		 1211 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1212}
1213
1214static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1215 size_t count, loff_t *ppos)
1216{
496ad9aa 1217 struct inode * inode = file_inode(file);
1da177e4 1218 uid_t loginuid;
e1760bd5 1219 kuid_t kloginuid;
774636e1 1220 int rv;
1da177e4 1221
7dc52157
PM		 1222 rcu_read_lock();
1223 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1224 rcu_read_unlock();
1da177e4 1225 return -EPERM;
7dc52157
PM		 1226 }
1227 rcu_read_unlock();
1da177e4 1228
1da177e4
LT		 1229 if (*ppos != 0) {
1230 /* No partial writes. */
1231 return -EINVAL;
1232 }
1da177e4 1233
774636e1
AD		 1234 rv = kstrtou32_from_user(buf, count, 10, &loginuid);
1235 if (rv < 0)
1236 return rv;
81407c84
EP		 1237
1238 /* is userspace tring to explicitly UNSET the loginuid? */
1239 if (loginuid == AUDIT_UID_UNSET) {
1240 kloginuid = INVALID_UID;
1241 } else {
1242 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
774636e1
AD		 1243 if (!uid_valid(kloginuid))
1244 return -EINVAL;
e1760bd5
EB		 1245 }
1246
774636e1
AD		 1247 rv = audit_set_loginuid(kloginuid);
1248 if (rv < 0)
1249 return rv;
1250 return count;
1da177e4
LT		 1251}
1252
00977a59 1253static const struct file_operations proc_loginuid_operations = {
1da177e4
LT		 1254 .read = proc_loginuid_read,
1255 .write = proc_loginuid_write,
87df8424 1256 .llseek = generic_file_llseek,
1da177e4 1257};
1e0bd755
EP		 1258
1259static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1260 size_t count, loff_t *ppos)
1261{
496ad9aa 1262 struct inode * inode = file_inode(file);
1e0bd755
EP		 1263 struct task_struct *task = get_proc_task(inode);
1264 ssize_t length;
1265 char tmpbuf[TMPBUFLEN];
1266
1267 if (!task)
1268 return -ESRCH;
1269 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1270 audit_get_sessionid(task));
1271 put_task_struct(task);
1272 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1273}
1274
1275static const struct file_operations proc_sessionid_operations = {
1276 .read = proc_sessionid_read,
87df8424 1277 .llseek = generic_file_llseek,
1e0bd755 1278};
1da177e4
LT		 1279#endif
1280
f4f154fd
AM		 1281#ifdef CONFIG_FAULT_INJECTION
1282static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1283 size_t count, loff_t *ppos)
1284{
496ad9aa 1285 struct task_struct *task = get_proc_task(file_inode(file));
f4f154fd
AM		 1286 char buffer[PROC_NUMBUF];
1287 size_t len;
1288 int make_it_fail;
f4f154fd
AM		 1289
1290 if (!task)
1291 return -ESRCH;
1292 make_it_fail = task->make_it_fail;
1293 put_task_struct(task);
1294
1295 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
0c28f287
AM		 1296
1297 return simple_read_from_buffer(buf, count, ppos, buffer, len);
f4f154fd
AM		 1298}
1299
1300static ssize_t proc_fault_inject_write(struct file * file,
1301 const char __user * buf, size_t count, loff_t *ppos)
1302{
1303 struct task_struct *task;
774636e1 1304 char buffer[PROC_NUMBUF];
f4f154fd 1305 int make_it_fail;
774636e1 1306 int rv;
f4f154fd
AM		 1307
1308 if (!capable(CAP_SYS_RESOURCE))
1309 return -EPERM;
1310 memset(buffer, 0, sizeof(buffer));
1311 if (count > sizeof(buffer) - 1)
1312 count = sizeof(buffer) - 1;
1313 if (copy_from_user(buffer, buf, count))
1314 return -EFAULT;
774636e1
AD		 1315 rv = kstrtoint(strstrip(buffer), 0, &make_it_fail);
1316 if (rv < 0)
1317 return rv;
16caed31
DJ		 1318 if (make_it_fail < 0 || make_it_fail > 1)
1319 return -EINVAL;
1320
496ad9aa 1321 task = get_proc_task(file_inode(file));
f4f154fd
AM		 1322 if (!task)
1323 return -ESRCH;
1324 task->make_it_fail = make_it_fail;
1325 put_task_struct(task);
cba8aafe
VL		 1326
1327 return count;
f4f154fd
AM		 1328}
1329
00977a59 1330static const struct file_operations proc_fault_inject_operations = {
f4f154fd
AM		 1331 .read = proc_fault_inject_read,
1332 .write = proc_fault_inject_write,
87df8424 1333 .llseek = generic_file_llseek,
f4f154fd
AM		 1334};
1335#endif
1336
9745512c 1337
43ae34cb
IM		 1338#ifdef CONFIG_SCHED_DEBUG
1339/*
1340 * Print out various scheduling related per-task fields:
1341 */
1342static int sched_show(struct seq_file *m, void *v)
1343{
1344 struct inode *inode = m->private;
1345 struct task_struct *p;
1346
43ae34cb
IM		 1347 p = get_proc_task(inode);
1348 if (!p)
1349 return -ESRCH;
1350 proc_sched_show_task(p, m);
1351
1352 put_task_struct(p);
1353
1354 return 0;
1355}
1356
1357static ssize_t
1358sched_write(struct file *file, const char __user *buf,
1359 size_t count, loff_t *offset)
1360{
496ad9aa 1361 struct inode *inode = file_inode(file);
43ae34cb
IM		 1362 struct task_struct *p;
1363
43ae34cb
IM		 1364 p = get_proc_task(inode);
1365 if (!p)
1366 return -ESRCH;
1367 proc_sched_set_task(p);
1368
1369 put_task_struct(p);
1370
1371 return count;
1372}
1373
1374static int sched_open(struct inode *inode, struct file *filp)
1375{
c6a34058 1376 return single_open(filp, sched_show, inode);
43ae34cb
IM		 1377}
1378
1379static const struct file_operations proc_pid_sched_operations = {
1380 .open = sched_open,
1381 .read = seq_read,
1382 .write = sched_write,
1383 .llseek = seq_lseek,
5ea473a1 1384 .release = single_release,
43ae34cb
IM		 1385};
1386
1387#endif
1388
5091faa4
MG		 1389#ifdef CONFIG_SCHED_AUTOGROUP
1390/*
1391 * Print out autogroup related information:
1392 */
1393static int sched_autogroup_show(struct seq_file *m, void *v)
1394{
1395 struct inode *inode = m->private;
1396 struct task_struct *p;
1397
1398 p = get_proc_task(inode);
1399 if (!p)
1400 return -ESRCH;
1401 proc_sched_autogroup_show_task(p, m);
1402
1403 put_task_struct(p);
1404
1405 return 0;
1406}
1407
1408static ssize_t
1409sched_autogroup_write(struct file *file, const char __user *buf,
1410 size_t count, loff_t *offset)
1411{
496ad9aa 1412 struct inode *inode = file_inode(file);
5091faa4
MG		 1413 struct task_struct *p;
1414 char buffer[PROC_NUMBUF];
0a8cb8e3 1415 int nice;
5091faa4
MG		 1416 int err;
1417
1418 memset(buffer, 0, sizeof(buffer));
1419 if (count > sizeof(buffer) - 1)
1420 count = sizeof(buffer) - 1;
1421 if (copy_from_user(buffer, buf, count))
1422 return -EFAULT;
1423
0a8cb8e3
AD		 1424 err = kstrtoint(strstrip(buffer), 0, &nice);
1425 if (err < 0)
1426 return err;
5091faa4
MG		 1427
1428 p = get_proc_task(inode);
1429 if (!p)
1430 return -ESRCH;
1431
2e5b5b3a 1432 err = proc_sched_autogroup_set_nice(p, nice);
5091faa4
MG		 1433 if (err)
1434 count = err;
1435
1436 put_task_struct(p);
1437
1438 return count;
1439}
1440
1441static int sched_autogroup_open(struct inode *inode, struct file *filp)
1442{
1443 int ret;
1444
1445 ret = single_open(filp, sched_autogroup_show, NULL);
1446 if (!ret) {
1447 struct seq_file *m = filp->private_data;
1448
1449 m->private = inode;
1450 }
1451 return ret;
1452}
1453
1454static const struct file_operations proc_pid_sched_autogroup_operations = {
1455 .open = sched_autogroup_open,
1456 .read = seq_read,
1457 .write = sched_autogroup_write,
1458 .llseek = seq_lseek,
1459 .release = single_release,
1460};
1461
1462#endif /* CONFIG_SCHED_AUTOGROUP */
1463
4614a696 1464static ssize_t comm_write(struct file *file, const char __user *buf,
1465 size_t count, loff_t *offset)
1466{
496ad9aa 1467 struct inode *inode = file_inode(file);
4614a696 1468 struct task_struct *p;
1469 char buffer[TASK_COMM_LEN];
830e0fc9 1470 const size_t maxlen = sizeof(buffer) - 1;
4614a696 1471
1472 memset(buffer, 0, sizeof(buffer));
830e0fc9 1473 if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
4614a696 1474 return -EFAULT;
1475
1476 p = get_proc_task(inode);
1477 if (!p)
1478 return -ESRCH;
1479
1480 if (same_thread_group(current, p))
1481 set_task_comm(p, buffer);
1482 else
1483 count = -EINVAL;
1484
1485 put_task_struct(p);
1486
1487 return count;
1488}
1489
1490static int comm_show(struct seq_file *m, void *v)
1491{
1492 struct inode *inode = m->private;
1493 struct task_struct *p;
1494
1495 p = get_proc_task(inode);
1496 if (!p)
1497 return -ESRCH;
1498
1499 task_lock(p);
1500 seq_printf(m, "%s\n", p->comm);
1501 task_unlock(p);
1502
1503 put_task_struct(p);
1504
1505 return 0;
1506}
1507
1508static int comm_open(struct inode *inode, struct file *filp)
1509{
c6a34058 1510 return single_open(filp, comm_show, inode);
4614a696 1511}
1512
1513static const struct file_operations proc_pid_set_comm_operations = {
1514 .open = comm_open,
1515 .read = seq_read,
1516 .write = comm_write,
1517 .llseek = seq_lseek,
1518 .release = single_release,
1519};
1520
7773fbc5 1521static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
925d1c40
MH		 1522{
1523 struct task_struct *task;
1524 struct mm_struct *mm;
1525 struct file *exe_file;
1526
2b0143b5 1527 task = get_proc_task(d_inode(dentry));
925d1c40
MH		 1528 if (!task)
1529 return -ENOENT;
1530 mm = get_task_mm(task);
1531 put_task_struct(task);
1532 if (!mm)
1533 return -ENOENT;
1534 exe_file = get_mm_exe_file(mm);
1535 mmput(mm);
1536 if (exe_file) {
1537 *exe_path = exe_file->f_path;
1538 path_get(&exe_file->f_path);
1539 fput(exe_file);
1540 return 0;
1541 } else
1542 return -ENOENT;
1543}
1544
6b255391 1545static const char *proc_pid_get_link(struct dentry *dentry,
fceef393
AV		 1546 struct inode *inode,
1547 struct delayed_call *done)
1da177e4 1548{
408ef013 1549 struct path path;
1da177e4
LT		 1550 int error = -EACCES;
1551
6b255391
AV		 1552 if (!dentry)
1553 return ERR_PTR(-ECHILD);
1554
778c1144
EB		 1555 /* Are we allowed to snoop on the tasks file descriptors? */
1556 if (!proc_fd_access_allowed(inode))
1da177e4 1557 goto out;
1da177e4 1558
408ef013
CH		 1559 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1560 if (error)
1561 goto out;
1562
6e77137b 1563 nd_jump_link(&path);
408ef013 1564 return NULL;
1da177e4 1565out:
008b150a 1566 return ERR_PTR(error);
1da177e4
LT		 1567}
1568
3dcd25f3 1569static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1da177e4 1570{
e12ba74d 1571 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
3dcd25f3 1572 char *pathname;
1da177e4
LT		 1573 int len;
1574
1575 if (!tmp)
1576 return -ENOMEM;
0c28f287 1577
7b2a69ba 1578 pathname = d_path(path, tmp, PAGE_SIZE);
3dcd25f3
JB		 1579 len = PTR_ERR(pathname);
1580 if (IS_ERR(pathname))
1da177e4 1581 goto out;
3dcd25f3 1582 len = tmp + PAGE_SIZE - 1 - pathname;
1da177e4
LT		 1583
1584 if (len > buflen)
1585 len = buflen;
3dcd25f3 1586 if (copy_to_user(buffer, pathname, len))
1da177e4
LT		 1587 len = -EFAULT;
1588 out:
1589 free_page((unsigned long)tmp);
1590 return len;
1591}
1592
1593static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1594{
1595 int error = -EACCES;
2b0143b5 1596 struct inode *inode = d_inode(dentry);
3dcd25f3 1597 struct path path;
1da177e4 1598
778c1144
EB		 1599 /* Are we allowed to snoop on the tasks file descriptors? */
1600 if (!proc_fd_access_allowed(inode))
1da177e4 1601 goto out;
1da177e4 1602
7773fbc5 1603 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1da177e4
LT		 1604 if (error)
1605 goto out;
1606
3dcd25f3
JB		 1607 error = do_proc_readlink(&path, buffer, buflen);
1608 path_put(&path);
1da177e4 1609out:
1da177e4
LT		 1610 return error;
1611}
1612
faf60af1 1613const struct inode_operations proc_pid_link_inode_operations = {
1da177e4 1614 .readlink = proc_pid_readlink,
6b255391 1615 .get_link = proc_pid_get_link,
6d76fa58 1616 .setattr = proc_setattr,
1da177e4
LT		 1617};
1618
28a6d671
EB		 1619
1620/* building an inode */
1621
6b4e306a 1622struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
28a6d671
EB		 1623{
1624 struct inode * inode;
1625 struct proc_inode *ei;
c69e8d9c 1626 const struct cred *cred;
1da177e4 1627
28a6d671 1628 /* We need a new inode */
1da177e4 1629
28a6d671
EB		 1630 inode = new_inode(sb);
1631 if (!inode)
1632 goto out;
1633
1634 /* Common stuff */
1635 ei = PROC_I(inode);
85fe4025 1636 inode->i_ino = get_next_ino();
28a6d671 1637 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
28a6d671
EB		 1638 inode->i_op = &proc_def_inode_operations;
1639
1640 /*
1641 * grab the reference to task.
1642 */
1a657f78 1643 ei->pid = get_task_pid(task, PIDTYPE_PID);
28a6d671
EB		 1644 if (!ei->pid)
1645 goto out_unlock;
1646
28a6d671 1647 if (task_dumpable(task)) {
c69e8d9c
DH		 1648 rcu_read_lock();
1649 cred = __task_cred(task);
1650 inode->i_uid = cred->euid;
1651 inode->i_gid = cred->egid;
1652 rcu_read_unlock();
1da177e4 1653 }
28a6d671
EB		 1654 security_task_to_inode(task, inode);
1655
1da177e4 1656out:
28a6d671
EB		 1657 return inode;
1658
1659out_unlock:
1660 iput(inode);
1661 return NULL;
1da177e4
LT		 1662}
1663
6b4e306a 1664int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1da177e4 1665{
2b0143b5 1666 struct inode *inode = d_inode(dentry);
28a6d671 1667 struct task_struct *task;
c69e8d9c 1668 const struct cred *cred;
0499680a 1669 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
c69e8d9c 1670
28a6d671 1671 generic_fillattr(inode, stat);
1da177e4 1672
28a6d671 1673 rcu_read_lock();
dcb0f222
EB		 1674 stat->uid = GLOBAL_ROOT_UID;
1675 stat->gid = GLOBAL_ROOT_GID;
28a6d671
EB		 1676 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1677 if (task) {
0499680a
VK		 1678 if (!has_pid_permissions(pid, task, 2)) {
1679 rcu_read_unlock();
1680 /*
1681 * This doesn't prevent learning whether PID exists,
1682 * it only makes getattr() consistent with readdir().
1683 */
1684 return -ENOENT;
1685 }
28a6d671
EB		 1686 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1687 task_dumpable(task)) {
c69e8d9c
DH		 1688 cred = __task_cred(task);
1689 stat->uid = cred->euid;
1690 stat->gid = cred->egid;
1da177e4
LT		 1691 }
1692 }
28a6d671 1693 rcu_read_unlock();
d6e71144 1694 return 0;
1da177e4
LT		 1695}
1696
1da177e4
LT		 1697/* dentry stuff */
1698
1699/*
1700 * Exceptional case: normally we are not allowed to unhash a busy
1701 * directory. In this case, however, we can do it - no aliasing problems
1702 * due to the way we treat inodes.
1703 *
1704 * Rewrite the inode's ownerships here because the owning task may have
1705 * performed a setuid(), etc.
99f89551
EB		 1706 *
1707 * Before the /proc/pid/status file was created the only way to read
1708 * the effective uid of a /process was to stat /proc/pid. Reading
1709 * /proc/pid/status is slow enough that procps and other packages
1710 * kept stating /proc/pid. To keep the rules in /proc simple I have
1711 * made this apply to all per process world readable and executable
1712 * directories.
1da177e4 1713 */
0b728e19 1714int pid_revalidate(struct dentry *dentry, unsigned int flags)
1da177e4 1715{
34286d66
NP		 1716 struct inode *inode;
1717 struct task_struct *task;
c69e8d9c
DH		 1718 const struct cred *cred;
1719
0b728e19 1720 if (flags & LOOKUP_RCU)
34286d66
NP		 1721 return -ECHILD;
1722
2b0143b5 1723 inode = d_inode(dentry);
34286d66
NP		 1724 task = get_proc_task(inode);
1725
99f89551
EB		 1726 if (task) {
1727 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1728 task_dumpable(task)) {
c69e8d9c
DH		 1729 rcu_read_lock();
1730 cred = __task_cred(task);
1731 inode->i_uid = cred->euid;
1732 inode->i_gid = cred->egid;
1733 rcu_read_unlock();
1da177e4 1734 } else {
dcb0f222
EB		 1735 inode->i_uid = GLOBAL_ROOT_UID;
1736 inode->i_gid = GLOBAL_ROOT_GID;
1da177e4 1737 }
9ee8ab9f 1738 inode->i_mode &= ~(S_ISUID | S_ISGID);
1da177e4 1739 security_task_to_inode(task, inode);
99f89551 1740 put_task_struct(task);
1da177e4
LT		 1741 return 1;
1742 }
1da177e4
LT		 1743 return 0;
1744}
1745
d855a4b7
ON		 1746static inline bool proc_inode_is_dead(struct inode *inode)
1747{
1748 return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
1749}
1750
1dd704b6
DH		 1751int pid_delete_dentry(const struct dentry *dentry)
1752{
1753 /* Is the task we represent dead?
1754 * If so, then don't put the dentry on the lru list,
1755 * kill it immediately.
1756 */
2b0143b5 1757 return proc_inode_is_dead(d_inode(dentry));
1dd704b6
DH		 1758}
1759
6b4e306a 1760const struct dentry_operations pid_dentry_operations =
28a6d671
EB		 1761{
1762 .d_revalidate = pid_revalidate,
1763 .d_delete = pid_delete_dentry,
1764};
1765
1766/* Lookups */
1767
1c0d04c9
EB		 1768/*
1769 * Fill a directory entry.
1770 *
1771 * If possible create the dcache entry and derive our inode number and
1772 * file type from dcache entry.
1773 *
1774 * Since all of the proc inode numbers are dynamically generated, the inode
1775 * numbers do not exist until the inode is cache. This means creating the
1776 * the dcache entry in readdir is necessary to keep the inode numbers
1777 * reported by readdir in sync with the inode numbers reported
1778 * by stat.
1779 */
f0c3b509 1780bool proc_fill_cache(struct file *file, struct dir_context *ctx,
6b4e306a 1781 const char *name, int len,
c5141e6d 1782 instantiate_t instantiate, struct task_struct *task, const void *ptr)
61a28784 1783{
f0c3b509 1784 struct dentry *child, *dir = file->f_path.dentry;
1df98b8b 1785 struct qstr qname = QSTR_INIT(name, len);
61a28784 1786 struct inode *inode;
1df98b8b
AV		 1787 unsigned type;
1788 ino_t ino;
61a28784 1789
1df98b8b 1790 child = d_hash_and_lookup(dir, &qname);
61a28784 1791 if (!child) {
3781764b
AV		 1792 DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq);
1793 child = d_alloc_parallel(dir, &qname, &wq);
1794 if (IS_ERR(child))
1df98b8b 1795 goto end_instantiate;
3781764b
AV		 1796 if (d_in_lookup(child)) {
1797 int err = instantiate(d_inode(dir), child, task, ptr);
1798 d_lookup_done(child);
1799 if (err < 0) {
1800 dput(child);
1801 goto end_instantiate;
1802 }
61a28784
EB		 1803 }
1804 }
2b0143b5 1805 inode = d_inode(child);
147ce699
AV		 1806 ino = inode->i_ino;
1807 type = inode->i_mode >> 12;
61a28784 1808 dput(child);
f0c3b509 1809 return dir_emit(ctx, name, len, ino, type);
1df98b8b
AV		 1810
1811end_instantiate:
1812 return dir_emit(ctx, name, len, 1, DT_UNKNOWN);
61a28784
EB		 1813}
1814
640708a2
PE		 1815/*
1816 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1817 * which represent vma start and end addresses.
1818 */
1819static int dname_to_vma_addr(struct dentry *dentry,
1820 unsigned long *start, unsigned long *end)
1821{
1822 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
1823 return -EINVAL;
1824
1825 return 0;
1826}
1827
0b728e19 1828static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
640708a2
PE		 1829{
1830 unsigned long vm_start, vm_end;
1831 bool exact_vma_exists = false;
1832 struct mm_struct *mm = NULL;
1833 struct task_struct *task;
1834 const struct cred *cred;
1835 struct inode *inode;
1836 int status = 0;
1837
0b728e19 1838 if (flags & LOOKUP_RCU)
640708a2
PE		 1839 return -ECHILD;
1840
2b0143b5 1841 inode = d_inode(dentry);
640708a2
PE		 1842 task = get_proc_task(inode);
1843 if (!task)
1844 goto out_notask;
1845
caaee623 1846 mm = mm_access(task, PTRACE_MODE_READ_FSCREDS);
2344bec7 1847 if (IS_ERR_OR_NULL(mm))
640708a2
PE		 1848 goto out;
1849
1850 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1851 down_read(&mm->mmap_sem);
1852 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1853 up_read(&mm->mmap_sem);
1854 }
1855
1856 mmput(mm);
1857
1858 if (exact_vma_exists) {
1859 if (task_dumpable(task)) {
1860 rcu_read_lock();
1861 cred = __task_cred(task);
1862 inode->i_uid = cred->euid;
1863 inode->i_gid = cred->egid;
1864 rcu_read_unlock();
1865 } else {
dcb0f222
EB		 1866 inode->i_uid = GLOBAL_ROOT_UID;
1867 inode->i_gid = GLOBAL_ROOT_GID;
640708a2
PE		 1868 }
1869 security_task_to_inode(task, inode);
1870 status = 1;
1871 }
1872
1873out:
1874 put_task_struct(task);
1875
1876out_notask:
640708a2
PE		 1877 return status;
1878}
1879
1880static const struct dentry_operations tid_map_files_dentry_operations = {
1881 .d_revalidate = map_files_d_revalidate,
1882 .d_delete = pid_delete_dentry,
1883};
1884
6b255391 1885static int map_files_get_link(struct dentry *dentry, struct path *path)
640708a2
PE		 1886{
1887 unsigned long vm_start, vm_end;
1888 struct vm_area_struct *vma;
1889 struct task_struct *task;
1890 struct mm_struct *mm;
1891 int rc;
1892
1893 rc = -ENOENT;
2b0143b5 1894 task = get_proc_task(d_inode(dentry));
640708a2
PE		 1895 if (!task)
1896 goto out;
1897
1898 mm = get_task_mm(task);
1899 put_task_struct(task);
1900 if (!mm)
1901 goto out;
1902
1903 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
1904 if (rc)
1905 goto out_mmput;
1906
70335abb 1907 rc = -ENOENT;
640708a2
PE		 1908 down_read(&mm->mmap_sem);
1909 vma = find_exact_vma(mm, vm_start, vm_end);
1910 if (vma && vma->vm_file) {
1911 *path = vma->vm_file->f_path;
1912 path_get(path);
1913 rc = 0;
1914 }
1915 up_read(&mm->mmap_sem);
1916
1917out_mmput:
1918 mmput(mm);
1919out:
1920 return rc;
1921}
1922
1923struct map_files_info {
7b540d06 1924 fmode_t mode;
640708a2
PE		 1925 unsigned long len;
1926 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
1927};
1928
bdb4d100
CO		 1929/*
1930 * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the
1931 * symlinks may be used to bypass permissions on ancestor directories in the
1932 * path to the file in question.
1933 */
1934static const char *
6b255391 1935proc_map_files_get_link(struct dentry *dentry,
fceef393
AV		 1936 struct inode *inode,
1937 struct delayed_call *done)
bdb4d100
CO		 1938{
1939 if (!capable(CAP_SYS_ADMIN))
1940 return ERR_PTR(-EPERM);
1941
fceef393 1942 return proc_pid_get_link(dentry, inode, done);
bdb4d100
CO		 1943}
1944
1945/*
6b255391 1946 * Identical to proc_pid_link_inode_operations except for get_link()
bdb4d100
CO		 1947 */
1948static const struct inode_operations proc_map_files_link_inode_operations = {
1949 .readlink = proc_pid_readlink,
6b255391 1950 .get_link = proc_map_files_get_link,
bdb4d100
CO		 1951 .setattr = proc_setattr,
1952};
1953
c52a47ac 1954static int
640708a2
PE		 1955proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
1956 struct task_struct *task, const void *ptr)
1957{
7b540d06 1958 fmode_t mode = (fmode_t)(unsigned long)ptr;
640708a2
PE		 1959 struct proc_inode *ei;
1960 struct inode *inode;
1961
640708a2
PE		 1962 inode = proc_pid_make_inode(dir->i_sb, task);
1963 if (!inode)
c52a47ac 1964 return -ENOENT;
640708a2
PE		 1965
1966 ei = PROC_I(inode);
6b255391 1967 ei->op.proc_get_link = map_files_get_link;
640708a2 1968
bdb4d100 1969 inode->i_op = &proc_map_files_link_inode_operations;
640708a2
PE		 1970 inode->i_size = 64;
1971 inode->i_mode = S_IFLNK;
1972
7b540d06 1973 if (mode & FMODE_READ)
640708a2 1974 inode->i_mode |= S_IRUSR;
7b540d06 1975 if (mode & FMODE_WRITE)
640708a2
PE		 1976 inode->i_mode |= S_IWUSR;
1977
1978 d_set_d_op(dentry, &tid_map_files_dentry_operations);
1979 d_add(dentry, inode);
1980
c52a47ac 1981 return 0;
640708a2
PE		 1982}
1983
1984static struct dentry *proc_map_files_lookup(struct inode *dir,
00cd8dd3 1985 struct dentry *dentry, unsigned int flags)
640708a2
PE		 1986{
1987 unsigned long vm_start, vm_end;
1988 struct vm_area_struct *vma;
1989 struct task_struct *task;
c52a47ac 1990 int result;
640708a2
PE		 1991 struct mm_struct *mm;
1992
c52a47ac 1993 result = -ENOENT;
640708a2
PE		 1994 task = get_proc_task(dir);
1995 if (!task)
1996 goto out;
1997
c52a47ac 1998 result = -EACCES;
caaee623 1999 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
640708a2
PE		 2000 goto out_put_task;
2001
c52a47ac 2002 result = -ENOENT;
640708a2 2003 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
eb94cd96 2004 goto out_put_task;
640708a2
PE		 2005
2006 mm = get_task_mm(task);
2007 if (!mm)
eb94cd96 2008 goto out_put_task;
640708a2
PE		 2009
2010 down_read(&mm->mmap_sem);
2011 vma = find_exact_vma(mm, vm_start, vm_end);
2012 if (!vma)
2013 goto out_no_vma;
2014
05f56484
SK		 2015 if (vma->vm_file)
2016 result = proc_map_files_instantiate(dir, dentry, task,
2017 (void *)(unsigned long)vma->vm_file->f_mode);
640708a2
PE		 2018
2019out_no_vma:
2020 up_read(&mm->mmap_sem);
2021 mmput(mm);
640708a2
PE		 2022out_put_task:
2023 put_task_struct(task);
2024out:
c52a47ac 2025 return ERR_PTR(result);
640708a2
PE		 2026}
2027
2028static const struct inode_operations proc_map_files_inode_operations = {
2029 .lookup = proc_map_files_lookup,
2030 .permission = proc_fd_permission,
2031 .setattr = proc_setattr,
2032};
2033
2034static int
f0c3b509 2035proc_map_files_readdir(struct file *file, struct dir_context *ctx)
640708a2 2036{
640708a2
PE		 2037 struct vm_area_struct *vma;
2038 struct task_struct *task;
2039 struct mm_struct *mm;
f0c3b509
AV		 2040 unsigned long nr_files, pos, i;
2041 struct flex_array *fa = NULL;
2042 struct map_files_info info;
2043 struct map_files_info *p;
640708a2
PE		 2044 int ret;
2045
640708a2 2046 ret = -ENOENT;
f0c3b509 2047 task = get_proc_task(file_inode(file));
640708a2
PE		 2048 if (!task)
2049 goto out;
2050
2051 ret = -EACCES;
caaee623 2052 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
640708a2
PE		 2053 goto out_put_task;
2054
2055 ret = 0;
f0c3b509
AV		 2056 if (!dir_emit_dots(file, ctx))
2057 goto out_put_task;
640708a2 2058
f0c3b509
AV		 2059 mm = get_task_mm(task);
2060 if (!mm)
2061 goto out_put_task;
2062 down_read(&mm->mmap_sem);
640708a2 2063
f0c3b509 2064 nr_files = 0;
640708a2 2065
f0c3b509
AV		 2066 /*
2067 * We need two passes here:
2068 *
2069 * 1) Collect vmas of mapped files with mmap_sem taken
2070 * 2) Release mmap_sem and instantiate entries
2071 *
2072 * otherwise we get lockdep complained, since filldir()
2073 * routine might require mmap_sem taken in might_fault().
2074 */
640708a2 2075
f0c3b509
AV		 2076 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2077 if (vma->vm_file && ++pos > ctx->pos)
2078 nr_files++;
2079 }
2080
2081 if (nr_files) {
2082 fa = flex_array_alloc(sizeof(info), nr_files,
2083 GFP_KERNEL);
2084 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2085 GFP_KERNEL)) {
2086 ret = -ENOMEM;
2087 if (fa)
2088 flex_array_free(fa);
2089 up_read(&mm->mmap_sem);
2090 mmput(mm);
2091 goto out_put_task;
640708a2 2092 }
f0c3b509
AV		 2093 for (i = 0, vma = mm->mmap, pos = 2; vma;
2094 vma = vma->vm_next) {
2095 if (!vma->vm_file)
2096 continue;
2097 if (++pos <= ctx->pos)
2098 continue;
2099
2100 info.mode = vma->vm_file->f_mode;
2101 info.len = snprintf(info.name,
2102 sizeof(info.name), "%lx-%lx",
2103 vma->vm_start, vma->vm_end);
2104 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2105 BUG();
640708a2 2106 }
640708a2 2107 }
f0c3b509
AV		 2108 up_read(&mm->mmap_sem);
2109
2110 for (i = 0; i < nr_files; i++) {
2111 p = flex_array_get(fa, i);
2112 if (!proc_fill_cache(file, ctx,
2113 p->name, p->len,
2114 proc_map_files_instantiate,
2115 task,
2116 (void *)(unsigned long)p->mode))
2117 break;
2118 ctx->pos++;
640708a2 2119 }
f0c3b509
AV		 2120 if (fa)
2121 flex_array_free(fa);
2122 mmput(mm);
640708a2 2123
640708a2
PE		 2124out_put_task:
2125 put_task_struct(task);
2126out:
2127 return ret;
2128}
2129
2130static const struct file_operations proc_map_files_operations = {
2131 .read = generic_read_dir,
f50752ea
AV		 2132 .iterate_shared = proc_map_files_readdir,
2133 .llseek = generic_file_llseek,
640708a2
PE		 2134};
2135
b5946bea 2136#ifdef CONFIG_CHECKPOINT_RESTORE
48f6a7a5
PE		 2137struct timers_private {
2138 struct pid *pid;
2139 struct task_struct *task;
2140 struct sighand_struct *sighand;
57b8015e 2141 struct pid_namespace *ns;
48f6a7a5
PE		 2142 unsigned long flags;
2143};
2144
2145static void *timers_start(struct seq_file *m, loff_t *pos)
2146{
2147 struct timers_private *tp = m->private;
2148
2149 tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2150 if (!tp->task)
2151 return ERR_PTR(-ESRCH);
2152
2153 tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2154 if (!tp->sighand)
2155 return ERR_PTR(-ESRCH);
2156
2157 return seq_list_start(&tp->task->signal->posix_timers, *pos);
2158}
2159
2160static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2161{
2162 struct timers_private *tp = m->private;
2163 return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2164}
2165
2166static void timers_stop(struct seq_file *m, void *v)
2167{
2168 struct timers_private *tp = m->private;
2169
2170 if (tp->sighand) {
2171 unlock_task_sighand(tp->task, &tp->flags);
2172 tp->sighand = NULL;
2173 }
2174
2175 if (tp->task) {
2176 put_task_struct(tp->task);
2177 tp->task = NULL;
2178 }
2179}
2180
2181static int show_timer(struct seq_file *m, void *v)
2182{
2183 struct k_itimer *timer;
57b8015e
PE		 2184 struct timers_private *tp = m->private;
2185 int notify;
cedbccab 2186 static const char * const nstr[] = {
57b8015e
PE		 2187 [SIGEV_SIGNAL] = "signal",
2188 [SIGEV_NONE] = "none",
2189 [SIGEV_THREAD] = "thread",
2190 };
48f6a7a5
PE		 2191
2192 timer = list_entry((struct list_head *)v, struct k_itimer, list);
57b8015e
PE		 2193 notify = timer->it_sigev_notify;
2194
48f6a7a5 2195 seq_printf(m, "ID: %d\n", timer->it_id);
25ce3191
JP		 2196 seq_printf(m, "signal: %d/%p\n",
2197 timer->sigq->info.si_signo,
2198 timer->sigq->info.si_value.sival_ptr);
57b8015e 2199 seq_printf(m, "notify: %s/%s.%d\n",
25ce3191
JP		 2200 nstr[notify & ~SIGEV_THREAD_ID],
2201 (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2202 pid_nr_ns(timer->it_pid, tp->ns));
15ef0298 2203 seq_printf(m, "ClockID: %d\n", timer->it_clock);
48f6a7a5
PE		 2204
2205 return 0;
2206}
2207
2208static const struct seq_operations proc_timers_seq_ops = {
2209 .start = timers_start,
2210 .next = timers_next,
2211 .stop = timers_stop,
2212 .show = show_timer,
2213};
2214
2215static int proc_timers_open(struct inode *inode, struct file *file)
2216{
2217 struct timers_private *tp;
2218
2219 tp = __seq_open_private(file, &proc_timers_seq_ops,
2220 sizeof(struct timers_private));
2221 if (!tp)
2222 return -ENOMEM;
2223
2224 tp->pid = proc_pid(inode);
57b8015e 2225 tp->ns = inode->i_sb->s_fs_info;
48f6a7a5
PE		 2226 return 0;
2227}
2228
2229static const struct file_operations proc_timers_operations = {
2230 .open = proc_timers_open,
2231 .read = seq_read,
2232 .llseek = seq_lseek,
2233 .release = seq_release_private,
2234};
b5946bea 2235#endif
640708a2 2236
5de23d43
JS		 2237static ssize_t timerslack_ns_write(struct file *file, const char __user *buf,
2238 size_t count, loff_t *offset)
2239{
2240 struct inode *inode = file_inode(file);
2241 struct task_struct *p;
2242 u64 slack_ns;
2243 int err;
2244
2245 err = kstrtoull_from_user(buf, count, 10, &slack_ns);
2246 if (err < 0)
2247 return err;
2248
2249 p = get_proc_task(inode);
2250 if (!p)
2251 return -ESRCH;
2252
2253 if (ptrace_may_access(p, PTRACE_MODE_ATTACH_FSCREDS)) {
2254 task_lock(p);
2255 if (slack_ns == 0)
2256 p->timer_slack_ns = p->default_timer_slack_ns;
2257 else
2258 p->timer_slack_ns = slack_ns;
2259 task_unlock(p);
2260 } else
2261 count = -EPERM;
2262
2263 put_task_struct(p);
2264
2265 return count;
2266}
2267
2268static int timerslack_ns_show(struct seq_file *m, void *v)
2269{
2270 struct inode *inode = m->private;
2271 struct task_struct *p;
2272 int err = 0;
2273
2274 p = get_proc_task(inode);
2275 if (!p)
2276 return -ESRCH;
2277
2278 if (ptrace_may_access(p, PTRACE_MODE_ATTACH_FSCREDS)) {
2279 task_lock(p);
2280 seq_printf(m, "%llu\n", p->timer_slack_ns);
2281 task_unlock(p);
2282 } else
2283 err = -EPERM;
2284
2285 put_task_struct(p);
2286
2287 return err;
2288}
2289
2290static int timerslack_ns_open(struct inode *inode, struct file *filp)
2291{
2292 return single_open(filp, timerslack_ns_show, inode);
2293}
2294
2295static const struct file_operations proc_pid_set_timerslack_ns_operations = {
2296 .open = timerslack_ns_open,
2297 .read = seq_read,
2298 .write = timerslack_ns_write,
2299 .llseek = seq_lseek,
2300 .release = single_release,
2301};
2302
c52a47ac 2303static int proc_pident_instantiate(struct inode *dir,
c5141e6d 2304 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8 2305{
c5141e6d 2306 const struct pid_entry *p = ptr;
444ceed8
EB		 2307 struct inode *inode;
2308 struct proc_inode *ei;
444ceed8 2309
61a28784 2310 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 2311 if (!inode)
2312 goto out;
2313
2314 ei = PROC_I(inode);
2315 inode->i_mode = p->mode;
2316 if (S_ISDIR(inode->i_mode))
bfe86848 2317 set_nlink(inode, 2); /* Use getattr to fix if necessary */
444ceed8
EB		 2318 if (p->iop)
2319 inode->i_op = p->iop;
2320 if (p->fop)
2321 inode->i_fop = p->fop;
2322 ei->op = p->op;
fb045adb 2323 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 2324 d_add(dentry, inode);
2325 /* Close the race of the process dying before we return the dentry */
0b728e19 2326 if (pid_revalidate(dentry, 0))
c52a47ac 2327 return 0;
444ceed8 2328out:
c52a47ac 2329 return -ENOENT;
444ceed8
EB		 2330}
2331
1da177e4
LT		 2332static struct dentry *proc_pident_lookup(struct inode *dir,
2333 struct dentry *dentry,
c5141e6d 2334 const struct pid_entry *ents,
7bcd6b0e 2335 unsigned int nents)
1da177e4 2336{
c52a47ac 2337 int error;
99f89551 2338 struct task_struct *task = get_proc_task(dir);
c5141e6d 2339 const struct pid_entry *p, *last;
1da177e4 2340
c52a47ac 2341 error = -ENOENT;
1da177e4 2342
99f89551
EB		 2343 if (!task)
2344 goto out_no_task;
1da177e4 2345
20cdc894
EB		 2346 /*
2347 * Yes, it does not scale. And it should not. Don't add
2348 * new entries into /proc/<tgid>/ without very good reasons.
2349 */
7bcd6b0e
EB		 2350 last = &ents[nents - 1];
2351 for (p = ents; p <= last; p++) {
1da177e4
LT		 2352 if (p->len != dentry->d_name.len)
2353 continue;
2354 if (!memcmp(dentry->d_name.name, p->name, p->len))
2355 break;
2356 }
7bcd6b0e 2357 if (p > last)
1da177e4
LT		 2358 goto out;
2359
444ceed8 2360 error = proc_pident_instantiate(dir, dentry, task, p);
1da177e4 2361out:
99f89551
EB		 2362 put_task_struct(task);
2363out_no_task:
c52a47ac 2364 return ERR_PTR(error);
1da177e4
LT		 2365}
2366
f0c3b509 2367static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
c5141e6d 2368 const struct pid_entry *ents, unsigned int nents)
28a6d671 2369{
f0c3b509
AV		 2370 struct task_struct *task = get_proc_task(file_inode(file));
2371 const struct pid_entry *p;
28a6d671 2372
28a6d671 2373 if (!task)
f0c3b509 2374 return -ENOENT;
28a6d671 2375
f0c3b509
AV		 2376 if (!dir_emit_dots(file, ctx))
2377 goto out;
2378
2379 if (ctx->pos >= nents + 2)
2380 goto out;
28a6d671 2381
f0c3b509
AV		 2382 for (p = ents + (ctx->pos - 2); p <= ents + nents - 1; p++) {
2383 if (!proc_fill_cache(file, ctx, p->name, p->len,
2384 proc_pident_instantiate, task, p))
2385 break;
2386 ctx->pos++;
2387 }
28a6d671 2388out:
61a28784 2389 put_task_struct(task);
f0c3b509 2390 return 0;
1da177e4
LT		 2391}
2392
28a6d671
EB		 2393#ifdef CONFIG_SECURITY
2394static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2395 size_t count, loff_t *ppos)
2396{
496ad9aa 2397 struct inode * inode = file_inode(file);
04ff9708 2398 char *p = NULL;
28a6d671
EB		 2399 ssize_t length;
2400 struct task_struct *task = get_proc_task(inode);
2401
28a6d671 2402 if (!task)
04ff9708 2403 return -ESRCH;
28a6d671
EB		 2404
2405 length = security_getprocattr(task,
2fddfeef 2406 (char*)file->f_path.dentry->d_name.name,
04ff9708 2407 &p);
28a6d671 2408 put_task_struct(task);
04ff9708
AV		 2409 if (length > 0)
2410 length = simple_read_from_buffer(buf, count, ppos, p, length);
2411 kfree(p);
28a6d671 2412 return length;
1da177e4
LT		 2413}
2414
28a6d671
EB		 2415static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2416 size_t count, loff_t *ppos)
2417{
496ad9aa 2418 struct inode * inode = file_inode(file);
bb646cdb 2419 void *page;
28a6d671
EB		 2420 ssize_t length;
2421 struct task_struct *task = get_proc_task(inode);
2422
2423 length = -ESRCH;
2424 if (!task)
2425 goto out_no_task;
2426 if (count > PAGE_SIZE)
2427 count = PAGE_SIZE;
2428
2429 /* No partial writes. */
2430 length = -EINVAL;
2431 if (*ppos != 0)
2432 goto out;
2433
bb646cdb
AV		 2434 page = memdup_user(buf, count);
2435 if (IS_ERR(page)) {
2436 length = PTR_ERR(page);
28a6d671 2437 goto out;
bb646cdb 2438 }
28a6d671 2439
107db7c7 2440 /* Guard against adverse ptrace interaction */
9b1bf12d 2441 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
107db7c7
DH		 2442 if (length < 0)
2443 goto out_free;
2444
28a6d671 2445 length = security_setprocattr(task,
2fddfeef 2446 (char*)file->f_path.dentry->d_name.name,
bb646cdb 2447 page, count);
9b1bf12d 2448 mutex_unlock(&task->signal->cred_guard_mutex);
28a6d671 2449out_free:
bb646cdb 2450 kfree(page);
28a6d671
EB		 2451out:
2452 put_task_struct(task);
2453out_no_task:
2454 return length;
2455}
2456
00977a59 2457static const struct file_operations proc_pid_attr_operations = {
28a6d671
EB		 2458 .read = proc_pid_attr_read,
2459 .write = proc_pid_attr_write,
87df8424 2460 .llseek = generic_file_llseek,
28a6d671
EB		 2461};
2462
c5141e6d 2463static const struct pid_entry attr_dir_stuff[] = {
631f9c18
AD		 2464 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2465 REG("prev", S_IRUGO, proc_pid_attr_operations),
2466 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2467 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2468 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2469 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
28a6d671
EB		 2470};
2471
f0c3b509 2472static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
28a6d671 2473{
f0c3b509
AV		 2474 return proc_pident_readdir(file, ctx,
2475 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2476}
2477
00977a59 2478static const struct file_operations proc_attr_dir_operations = {
1da177e4 2479 .read = generic_read_dir,
f50752ea
AV		 2480 .iterate_shared = proc_attr_dir_readdir,
2481 .llseek = generic_file_llseek,
1da177e4
LT		 2482};
2483
72d9dcfc 2484static struct dentry *proc_attr_dir_lookup(struct inode *dir,
00cd8dd3 2485 struct dentry *dentry, unsigned int flags)
28a6d671 2486{
7bcd6b0e
EB		 2487 return proc_pident_lookup(dir, dentry,
2488 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2489}
2490
c5ef1c42 2491static const struct inode_operations proc_attr_dir_inode_operations = {
72d9dcfc 2492 .lookup = proc_attr_dir_lookup,
99f89551 2493 .getattr = pid_getattr,
6d76fa58 2494 .setattr = proc_setattr,
1da177e4
LT		 2495};
2496
28a6d671
EB		 2497#endif
2498
698ba7b5 2499#ifdef CONFIG_ELF_CORE
3cb4a0bb
KH		 2500static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2501 size_t count, loff_t *ppos)
2502{
496ad9aa 2503 struct task_struct *task = get_proc_task(file_inode(file));
3cb4a0bb
KH		 2504 struct mm_struct *mm;
2505 char buffer[PROC_NUMBUF];
2506 size_t len;
2507 int ret;
2508
2509 if (!task)
2510 return -ESRCH;
2511
2512 ret = 0;
2513 mm = get_task_mm(task);
2514 if (mm) {
2515 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2516 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2517 MMF_DUMP_FILTER_SHIFT));
2518 mmput(mm);
2519 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2520 }
2521
2522 put_task_struct(task);
2523
2524 return ret;
2525}
2526
2527static ssize_t proc_coredump_filter_write(struct file *file,
2528 const char __user *buf,
2529 size_t count,
2530 loff_t *ppos)
2531{
2532 struct task_struct *task;
2533 struct mm_struct *mm;
3cb4a0bb
KH		 2534 unsigned int val;
2535 int ret;
2536 int i;
2537 unsigned long mask;
2538
774636e1
AD		 2539 ret = kstrtouint_from_user(buf, count, 0, &val);
2540 if (ret < 0)
2541 return ret;
3cb4a0bb
KH		 2542
2543 ret = -ESRCH;
496ad9aa 2544 task = get_proc_task(file_inode(file));
3cb4a0bb
KH		 2545 if (!task)
2546 goto out_no_task;
2547
3cb4a0bb
KH		 2548 mm = get_task_mm(task);
2549 if (!mm)
2550 goto out_no_mm;
41a0c249 2551 ret = 0;
3cb4a0bb
KH		 2552
2553 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2554 if (val & mask)
2555 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2556 else
2557 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2558 }
2559
2560 mmput(mm);
2561 out_no_mm:
2562 put_task_struct(task);
2563 out_no_task:
774636e1
AD		 2564 if (ret < 0)
2565 return ret;
2566 return count;
3cb4a0bb
KH		 2567}
2568
2569static const struct file_operations proc_coredump_filter_operations = {
2570 .read = proc_coredump_filter_read,
2571 .write = proc_coredump_filter_write,
87df8424 2572 .llseek = generic_file_llseek,
3cb4a0bb
KH		 2573};
2574#endif
2575
aba76fdb 2576#ifdef CONFIG_TASK_IO_ACCOUNTING
19aadc98 2577static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
297c5d92 2578{
940389b8 2579 struct task_io_accounting acct = task->ioac;
5995477a 2580 unsigned long flags;
293eb1e7 2581 int result;
5995477a 2582
293eb1e7
VK		 2583 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2584 if (result)
2585 return result;
2586
caaee623 2587 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) {
293eb1e7
VK		 2588 result = -EACCES;
2589 goto out_unlock;
2590 }
1d1221f3 2591
5995477a
AR		 2592 if (whole && lock_task_sighand(task, &flags)) {
2593 struct task_struct *t = task;
2594
2595 task_io_accounting_add(&acct, &task->signal->ioac);
2596 while_each_thread(task, t)
2597 task_io_accounting_add(&acct, &t->ioac);
2598
2599 unlock_task_sighand(task, &flags);
297c5d92 2600 }
25ce3191
JP		 2601 seq_printf(m,
2602 "rchar: %llu\n"
2603 "wchar: %llu\n"
2604 "syscr: %llu\n"
2605 "syscw: %llu\n"
2606 "read_bytes: %llu\n"
2607 "write_bytes: %llu\n"
2608 "cancelled_write_bytes: %llu\n",
2609 (unsigned long long)acct.rchar,
2610 (unsigned long long)acct.wchar,
2611 (unsigned long long)acct.syscr,
2612 (unsigned long long)acct.syscw,
2613 (unsigned long long)acct.read_bytes,
2614 (unsigned long long)acct.write_bytes,
2615 (unsigned long long)acct.cancelled_write_bytes);
2616 result = 0;
2617
293eb1e7
VK		 2618out_unlock:
2619 mutex_unlock(&task->signal->cred_guard_mutex);
2620 return result;
297c5d92
AR		 2621}
2622
19aadc98
AD		 2623static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2624 struct pid *pid, struct task_struct *task)
297c5d92 2625{
19aadc98 2626 return do_io_accounting(task, m, 0);
aba76fdb 2627}
297c5d92 2628
19aadc98
AD		 2629static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2630 struct pid *pid, struct task_struct *task)
297c5d92 2631{
19aadc98 2632 return do_io_accounting(task, m, 1);
297c5d92
AR		 2633}
2634#endif /* CONFIG_TASK_IO_ACCOUNTING */
aba76fdb 2635
22d917d8
EB		 2636#ifdef CONFIG_USER_NS
2637static int proc_id_map_open(struct inode *inode, struct file *file,
ccf94f1b 2638 const struct seq_operations *seq_ops)
22d917d8
EB		 2639{
2640 struct user_namespace *ns = NULL;
2641 struct task_struct *task;
2642 struct seq_file *seq;
2643 int ret = -EINVAL;
2644
2645 task = get_proc_task(inode);
2646 if (task) {
2647 rcu_read_lock();
2648 ns = get_user_ns(task_cred_xxx(task, user_ns));
2649 rcu_read_unlock();
2650 put_task_struct(task);
2651 }
2652 if (!ns)
2653 goto err;
2654
2655 ret = seq_open(file, seq_ops);
2656 if (ret)
2657 goto err_put_ns;
2658
2659 seq = file->private_data;
2660 seq->private = ns;
2661
2662 return 0;
2663err_put_ns:
2664 put_user_ns(ns);
2665err:
2666 return ret;
2667}
2668
2669static int proc_id_map_release(struct inode *inode, struct file *file)
2670{
2671 struct seq_file *seq = file->private_data;
2672 struct user_namespace *ns = seq->private;
2673 put_user_ns(ns);
2674 return seq_release(inode, file);
2675}
2676
2677static int proc_uid_map_open(struct inode *inode, struct file *file)
2678{
2679 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2680}
2681
2682static int proc_gid_map_open(struct inode *inode, struct file *file)
2683{
2684 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2685}
2686
f76d207a
EB		 2687static int proc_projid_map_open(struct inode *inode, struct file *file)
2688{
2689 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2690}
2691
22d917d8
EB		 2692static const struct file_operations proc_uid_map_operations = {
2693 .open = proc_uid_map_open,
2694 .write = proc_uid_map_write,
2695 .read = seq_read,
2696 .llseek = seq_lseek,
2697 .release = proc_id_map_release,
2698};
2699
2700static const struct file_operations proc_gid_map_operations = {
2701 .open = proc_gid_map_open,
2702 .write = proc_gid_map_write,
2703 .read = seq_read,
2704 .llseek = seq_lseek,
2705 .release = proc_id_map_release,
2706};
f76d207a
EB		 2707
2708static const struct file_operations proc_projid_map_operations = {
2709 .open = proc_projid_map_open,
2710 .write = proc_projid_map_write,
2711 .read = seq_read,
2712 .llseek = seq_lseek,
2713 .release = proc_id_map_release,
2714};
9cc46516
EB		 2715
2716static int proc_setgroups_open(struct inode *inode, struct file *file)
2717{
2718 struct user_namespace *ns = NULL;
2719 struct task_struct *task;
2720 int ret;
2721
2722 ret = -ESRCH;
2723 task = get_proc_task(inode);
2724 if (task) {
2725 rcu_read_lock();
2726 ns = get_user_ns(task_cred_xxx(task, user_ns));
2727 rcu_read_unlock();
2728 put_task_struct(task);
2729 }
2730 if (!ns)
2731 goto err;
2732
2733 if (file->f_mode & FMODE_WRITE) {
2734 ret = -EACCES;
2735 if (!ns_capable(ns, CAP_SYS_ADMIN))
2736 goto err_put_ns;
2737 }
2738
2739 ret = single_open(file, &proc_setgroups_show, ns);
2740 if (ret)
2741 goto err_put_ns;
2742
2743 return 0;
2744err_put_ns:
2745 put_user_ns(ns);
2746err:
2747 return ret;
2748}
2749
2750static int proc_setgroups_release(struct inode *inode, struct file *file)
2751{
2752 struct seq_file *seq = file->private_data;
2753 struct user_namespace *ns = seq->private;
2754 int ret = single_release(inode, file);
2755 put_user_ns(ns);
2756 return ret;
2757}
2758
2759static const struct file_operations proc_setgroups_operations = {
2760 .open = proc_setgroups_open,
2761 .write = proc_setgroups_write,
2762 .read = seq_read,
2763 .llseek = seq_lseek,
2764 .release = proc_setgroups_release,
2765};
22d917d8
EB		 2766#endif /* CONFIG_USER_NS */
2767
47830723
KC		 2768static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2769 struct pid *pid, struct task_struct *task)
2770{
a9712bc1
AV		 2771 int err = lock_trace(task);
2772 if (!err) {
2773 seq_printf(m, "%08x\n", task->personality);
2774 unlock_trace(task);
2775 }
2776 return err;
47830723
KC		 2777}
2778
28a6d671
EB		 2779/*
2780 * Thread groups
2781 */
00977a59 2782static const struct file_operations proc_task_operations;
c5ef1c42 2783static const struct inode_operations proc_task_inode_operations;
20cdc894 2784
c5141e6d 2785static const struct pid_entry tgid_base_stuff[] = {
631f9c18
AD		 2786 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2787 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
640708a2 2788 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
631f9c18 2789 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 2790 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
b2211a36 2791#ifdef CONFIG_NET
631f9c18 2792 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
b2211a36 2793#endif
631f9c18 2794 REG("environ", S_IRUSR, proc_environ_operations),
f9ea536e 2795 ONE("auxv", S_IRUSR, proc_pid_auxv),
631f9c18 2796 ONE("status", S_IRUGO, proc_pid_status),
35a35046 2797 ONE("personality", S_IRUSR, proc_pid_personality),
1c963eb1 2798 ONE("limits", S_IRUGO, proc_pid_limits),
43ae34cb 2799#ifdef CONFIG_SCHED_DEBUG
631f9c18 2800 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
5091faa4
MG		 2801#endif
2802#ifdef CONFIG_SCHED_AUTOGROUP
2803 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
ebcb6734 2804#endif
4614a696 2805 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 2806#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
09d93bd6 2807 ONE("syscall", S_IRUSR, proc_pid_syscall),
43ae34cb 2808#endif
c2c0bb44 2809 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
631f9c18
AD		 2810 ONE("stat", S_IRUGO, proc_tgid_stat),
2811 ONE("statm", S_IRUGO, proc_pid_statm),
b7643757 2812 REG("maps", S_IRUGO, proc_pid_maps_operations),
28a6d671 2813#ifdef CONFIG_NUMA
b7643757 2814 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
28a6d671 2815#endif
631f9c18
AD		 2816 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2817 LNK("cwd", proc_cwd_link),
2818 LNK("root", proc_root_link),
2819 LNK("exe", proc_exe_link),
2820 REG("mounts", S_IRUGO, proc_mounts_operations),
2821 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2822 REG("mountstats", S_IRUSR, proc_mountstats_operations),
1e883281 2823#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18 2824 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
b7643757 2825 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
32ed74a4 2826 REG("pagemap", S_IRUSR, proc_pagemap_operations),
28a6d671
EB		 2827#endif
2828#ifdef CONFIG_SECURITY
631f9c18 2829 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 2830#endif
2831#ifdef CONFIG_KALLSYMS
edfcd606 2832 ONE("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 2833#endif
2ec220e2 2834#ifdef CONFIG_STACKTRACE
35a35046 2835 ONE("stack", S_IRUSR, proc_pid_stack),
28a6d671 2836#endif
5968cece 2837#ifdef CONFIG_SCHED_INFO
f6e826ca 2838 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 2839#endif
9745512c 2840#ifdef CONFIG_LATENCYTOP
631f9c18 2841 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 2842#endif
8793d854 2843#ifdef CONFIG_PROC_PID_CPUSET
52de4779 2844 ONE("cpuset", S_IRUGO, proc_cpuset_show),
a424316c
PM		 2845#endif
2846#ifdef CONFIG_CGROUPS
006f4ac4 2847 ONE("cgroup", S_IRUGO, proc_cgroup_show),
28a6d671 2848#endif
6ba51e37 2849 ONE("oom_score", S_IRUGO, proc_oom_score),
fa0cbbf1 2850 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
a63d83f4 2851 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 2852#ifdef CONFIG_AUDITSYSCALL
631f9c18
AD		 2853 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2854 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 2855#endif
f4f154fd 2856#ifdef CONFIG_FAULT_INJECTION
631f9c18 2857 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 2858#endif
698ba7b5 2859#ifdef CONFIG_ELF_CORE
631f9c18 2860 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3cb4a0bb 2861#endif
aba76fdb 2862#ifdef CONFIG_TASK_IO_ACCOUNTING
19aadc98 2863 ONE("io", S_IRUSR, proc_tgid_io_accounting),
aba76fdb 2864#endif
f133ecca 2865#ifdef CONFIG_HARDWALL
d962c144 2866 ONE("hardwall", S_IRUGO, proc_pid_hardwall),
f133ecca 2867#endif
22d917d8
EB		 2868#ifdef CONFIG_USER_NS
2869 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
2870 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
f76d207a 2871 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
9cc46516 2872 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
22d917d8 2873#endif
48f6a7a5
PE		 2874#ifdef CONFIG_CHECKPOINT_RESTORE
2875 REG("timers", S_IRUGO, proc_timers_operations),
2876#endif
5de23d43 2877 REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations),
28a6d671 2878};
1da177e4 2879
f0c3b509 2880static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
1da177e4 2881{
f0c3b509
AV		 2882 return proc_pident_readdir(file, ctx,
2883 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 2884}
2885
00977a59 2886static const struct file_operations proc_tgid_base_operations = {
1da177e4 2887 .read = generic_read_dir,
f50752ea
AV		 2888 .iterate_shared = proc_tgid_base_readdir,
2889 .llseek = generic_file_llseek,
1da177e4
LT		 2890};
2891
00cd8dd3
AV		 2892static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2893{
7bcd6b0e
EB		 2894 return proc_pident_lookup(dir, dentry,
2895 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 2896}
2897
c5ef1c42 2898static const struct inode_operations proc_tgid_base_inode_operations = {
28a6d671 2899 .lookup = proc_tgid_base_lookup,
99f89551 2900 .getattr = pid_getattr,
6d76fa58 2901 .setattr = proc_setattr,
0499680a 2902 .permission = proc_pid_permission,
1da177e4 2903};
1da177e4 2904
60347f67 2905static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
1da177e4 2906{
48e6484d 2907 struct dentry *dentry, *leader, *dir;
8578cea7 2908 char buf[PROC_NUMBUF];
48e6484d
EB		 2909 struct qstr name;
2910
2911 name.name = buf;
60347f67 2912 name.len = snprintf(buf, sizeof(buf), "%d", pid);
4f522a24 2913 /* no ->d_hash() rejects on procfs */
60347f67 2914 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d 2915 if (dentry) {
bbd51924 2916 d_invalidate(dentry);
48e6484d
EB		 2917 dput(dentry);
2918 }
1da177e4 2919
c35a7f18
ON		 2920 if (pid == tgid)
2921 return;
2922
48e6484d 2923 name.name = buf;
60347f67
PE		 2924 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2925 leader = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d
EB		 2926 if (!leader)
2927 goto out;
1da177e4 2928
48e6484d
EB		 2929 name.name = "task";
2930 name.len = strlen(name.name);
2931 dir = d_hash_and_lookup(leader, &name);
2932 if (!dir)
2933 goto out_put_leader;
2934
2935 name.name = buf;
60347f67 2936 name.len = snprintf(buf, sizeof(buf), "%d", pid);
48e6484d
EB		 2937 dentry = d_hash_and_lookup(dir, &name);
2938 if (dentry) {
bbd51924 2939 d_invalidate(dentry);
48e6484d 2940 dput(dentry);
1da177e4 2941 }
48e6484d
EB		 2942
2943 dput(dir);
2944out_put_leader:
2945 dput(leader);
2946out:
2947 return;
1da177e4
LT		 2948}
2949
0895e91d
RD		 2950/**
2951 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2952 * @task: task that should be flushed.
2953 *
2954 * When flushing dentries from proc, one needs to flush them from global
60347f67 2955 * proc (proc_mnt) and from all the namespaces' procs this task was seen
0895e91d
RD		 2956 * in. This call is supposed to do all of this job.
2957 *
2958 * Looks in the dcache for
2959 * /proc/@pid
2960 * /proc/@tgid/task/@pid
2961 * if either directory is present flushes it and all of it'ts children
2962 * from the dcache.
2963 *
2964 * It is safe and reasonable to cache /proc entries for a task until
2965 * that task exits. After that they just clog up the dcache with
2966 * useless entries, possibly causing useful dcache entries to be
2967 * flushed instead. This routine is proved to flush those useless
2968 * dcache entries at process exit time.
2969 *
2970 * NOTE: This routine is just an optimization so it does not guarantee
2971 * that no dcache entries will exist at process exit time it
2972 * just makes it very unlikely that any will persist.
60347f67
PE		 2973 */
2974
2975void proc_flush_task(struct task_struct *task)
2976{
9fcc2d15 2977 int i;
9b4d1cbe 2978 struct pid *pid, *tgid;
130f77ec
PE		 2979 struct upid *upid;
2980
130f77ec 2981 pid = task_pid(task);
9b4d1cbe 2982 tgid = task_tgid(task);
130f77ec 2983
9fcc2d15 2984 for (i = 0; i <= pid->level; i++) {
130f77ec
PE		 2985 upid = &pid->numbers[i];
2986 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
9b4d1cbe 2987 tgid->numbers[i].nr);
130f77ec 2988 }
60347f67
PE		 2989}
2990
c52a47ac
AV		 2991static int proc_pid_instantiate(struct inode *dir,
2992 struct dentry * dentry,
2993 struct task_struct *task, const void *ptr)
444ceed8 2994{
444ceed8
EB		 2995 struct inode *inode;
2996
61a28784 2997 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 2998 if (!inode)
2999 goto out;
3000
3001 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3002 inode->i_op = &proc_tgid_base_inode_operations;
3003 inode->i_fop = &proc_tgid_base_operations;
3004 inode->i_flags|=S_IMMUTABLE;
aed54175 3005
bfe86848
MS		 3006 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
3007 ARRAY_SIZE(tgid_base_stuff)));
444ceed8 3008
fb045adb 3009 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 3010
3011 d_add(dentry, inode);
3012 /* Close the race of the process dying before we return the dentry */
0b728e19 3013 if (pid_revalidate(dentry, 0))
c52a47ac 3014 return 0;
444ceed8 3015out:
c52a47ac 3016 return -ENOENT;
444ceed8
EB		 3017}
3018
00cd8dd3 3019struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
1da177e4 3020{
335eb531 3021 int result = -ENOENT;
1da177e4 3022 struct task_struct *task;
1da177e4 3023 unsigned tgid;
b488893a 3024 struct pid_namespace *ns;
1da177e4 3025
dbcdb504 3026 tgid = name_to_int(&dentry->d_name);
1da177e4
LT		 3027 if (tgid == ~0U)
3028 goto out;
3029
b488893a 3030 ns = dentry->d_sb->s_fs_info;
de758734 3031 rcu_read_lock();
b488893a 3032 task = find_task_by_pid_ns(tgid, ns);
1da177e4
LT		 3033 if (task)
3034 get_task_struct(task);
de758734 3035 rcu_read_unlock();
1da177e4
LT		 3036 if (!task)
3037 goto out;
3038
444ceed8 3039 result = proc_pid_instantiate(dir, dentry, task, NULL);
1da177e4 3040 put_task_struct(task);
1da177e4 3041out:
c52a47ac 3042 return ERR_PTR(result);
1da177e4
LT		 3043}
3044
1da177e4 3045/*
0804ef4b 3046 * Find the first task with tgid >= tgid
0bc58a91 3047 *
1da177e4 3048 */
19fd4bb2
EB		 3049struct tgid_iter {
3050 unsigned int tgid;
0804ef4b 3051 struct task_struct *task;
19fd4bb2
EB		 3052};
3053static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3054{
0804ef4b 3055 struct pid *pid;
1da177e4 3056
19fd4bb2
EB		 3057 if (iter.task)
3058 put_task_struct(iter.task);
454cc105 3059 rcu_read_lock();
0804ef4b 3060retry:
19fd4bb2
EB		 3061 iter.task = NULL;
3062 pid = find_ge_pid(iter.tgid, ns);
0804ef4b 3063 if (pid) {
19fd4bb2
EB		 3064 iter.tgid = pid_nr_ns(pid, ns);
3065 iter.task = pid_task(pid, PIDTYPE_PID);
0804ef4b
EB		 3066 /* What we to know is if the pid we have find is the
3067 * pid of a thread_group_leader. Testing for task
3068 * being a thread_group_leader is the obvious thing
3069 * todo but there is a window when it fails, due to
3070 * the pid transfer logic in de_thread.
3071 *
3072 * So we perform the straight forward test of seeing
3073 * if the pid we have found is the pid of a thread
3074 * group leader, and don't worry if the task we have
3075 * found doesn't happen to be a thread group leader.
3076 * As we don't care in the case of readdir.
3077 */
19fd4bb2
EB		 3078 if (!iter.task || !has_group_leader_pid(iter.task)) {
3079 iter.tgid += 1;
0804ef4b 3080 goto retry;
19fd4bb2
EB		 3081 }
3082 get_task_struct(iter.task);
0bc58a91 3083 }
454cc105 3084 rcu_read_unlock();
19fd4bb2 3085 return iter;
1da177e4
LT		 3086}
3087
0097875b 3088#define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
0804ef4b 3089
1da177e4 3090/* for the /proc/ directory itself, after non-process stuff has been done */
f0c3b509 3091int proc_pid_readdir(struct file *file, struct dir_context *ctx)
1da177e4 3092{
19fd4bb2 3093 struct tgid_iter iter;
3aa3377f 3094 struct pid_namespace *ns = file_inode(file)->i_sb->s_fs_info;
f0c3b509 3095 loff_t pos = ctx->pos;
1da177e4 3096
021ada7d 3097 if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
f0c3b509 3098 return 0;
1da177e4 3099
0097875b 3100 if (pos == TGID_OFFSET - 2) {
2b0143b5 3101 struct inode *inode = d_inode(ns->proc_self);
db963164 3102 if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
f0c3b509 3103 return 0;
0097875b
EB		 3104 ctx->pos = pos = pos + 1;
3105 }
3106 if (pos == TGID_OFFSET - 1) {
2b0143b5 3107 struct inode *inode = d_inode(ns->proc_thread_self);
0097875b
EB		 3108 if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3109 return 0;
3110 ctx->pos = pos = pos + 1;
021ada7d 3111 }
0097875b 3112 iter.tgid = pos - TGID_OFFSET;
19fd4bb2 3113 iter.task = NULL;
19fd4bb2
EB		 3114 for (iter = next_tgid(ns, iter);
3115 iter.task;
3116 iter.tgid += 1, iter = next_tgid(ns, iter)) {
f0c3b509
AV		 3117 char name[PROC_NUMBUF];
3118 int len;
3119 if (!has_pid_permissions(ns, iter.task, 2))
3120 continue;
0499680a 3121
f0c3b509
AV		 3122 len = snprintf(name, sizeof(name), "%d", iter.tgid);
3123 ctx->pos = iter.tgid + TGID_OFFSET;
3124 if (!proc_fill_cache(file, ctx, name, len,
3125 proc_pid_instantiate, iter.task, NULL)) {
19fd4bb2 3126 put_task_struct(iter.task);
f0c3b509 3127 return 0;
1da177e4 3128 }
0bc58a91 3129 }
f0c3b509 3130 ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
0bc58a91
EB		 3131 return 0;
3132}
1da177e4 3133
1b3044e3
JD		 3134/*
3135 * proc_tid_comm_permission is a special permission function exclusively
3136 * used for the node /proc/<pid>/task/<tid>/comm.
3137 * It bypasses generic permission checks in the case where a task of the same
3138 * task group attempts to access the node.
3139 * The rationale behind this is that glibc and bionic access this node for
3140 * cross thread naming (pthread_set/getname_np(!self)). However, if
3141 * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3142 * which locks out the cross thread naming implementation.
3143 * This function makes sure that the node is always accessible for members of
3144 * same thread group.
3145 */
3146static int proc_tid_comm_permission(struct inode *inode, int mask)
3147{
3148 bool is_same_tgroup;
3149 struct task_struct *task;
3150
3151 task = get_proc_task(inode);
3152 if (!task)
3153 return -ESRCH;
3154 is_same_tgroup = same_thread_group(current, task);
3155 put_task_struct(task);
3156
3157 if (likely(is_same_tgroup && !(mask & MAY_EXEC))) {
3158 /* This file (/proc/<pid>/task/<tid>/comm) can always be
3159 * read or written by the members of the corresponding
3160 * thread group.
3161 */
3162 return 0;
3163 }
3164
3165 return generic_permission(inode, mask);
3166}
3167
3168static const struct inode_operations proc_tid_comm_inode_operations = {
3169 .permission = proc_tid_comm_permission,
3170};
3171
28a6d671
EB		 3172/*
3173 * Tasks
3174 */
c5141e6d 3175static const struct pid_entry tid_base_stuff[] = {
631f9c18 3176 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3835541d 3177 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 3178 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
6ba8ed79
EB		 3179#ifdef CONFIG_NET
3180 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3181#endif
631f9c18 3182 REG("environ", S_IRUSR, proc_environ_operations),
f9ea536e 3183 ONE("auxv", S_IRUSR, proc_pid_auxv),
631f9c18 3184 ONE("status", S_IRUGO, proc_pid_status),
35a35046 3185 ONE("personality", S_IRUSR, proc_pid_personality),
1c963eb1 3186 ONE("limits", S_IRUGO, proc_pid_limits),
43ae34cb 3187#ifdef CONFIG_SCHED_DEBUG
631f9c18 3188 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
ebcb6734 3189#endif
1b3044e3
JD		 3190 NOD("comm", S_IFREG|S_IRUGO|S_IWUSR,
3191 &proc_tid_comm_inode_operations,
3192 &proc_pid_set_comm_operations, {}),
ebcb6734 3193#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
09d93bd6 3194 ONE("syscall", S_IRUSR, proc_pid_syscall),
43ae34cb 3195#endif
c2c0bb44 3196 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
631f9c18
AD		 3197 ONE("stat", S_IRUGO, proc_tid_stat),
3198 ONE("statm", S_IRUGO, proc_pid_statm),
b7643757 3199 REG("maps", S_IRUGO, proc_tid_maps_operations),
2e13ba54 3200#ifdef CONFIG_PROC_CHILDREN
81841161
CG		 3201 REG("children", S_IRUGO, proc_tid_children_operations),
3202#endif
28a6d671 3203#ifdef CONFIG_NUMA
b7643757 3204 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
28a6d671 3205#endif
631f9c18
AD		 3206 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3207 LNK("cwd", proc_cwd_link),
3208 LNK("root", proc_root_link),
3209 LNK("exe", proc_exe_link),
3210 REG("mounts", S_IRUGO, proc_mounts_operations),
3211 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
1e883281 3212#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18 3213 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
b7643757 3214 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
32ed74a4 3215 REG("pagemap", S_IRUSR, proc_pagemap_operations),
28a6d671
EB		 3216#endif
3217#ifdef CONFIG_SECURITY
631f9c18 3218 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 3219#endif
3220#ifdef CONFIG_KALLSYMS
edfcd606 3221 ONE("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 3222#endif
2ec220e2 3223#ifdef CONFIG_STACKTRACE
35a35046 3224 ONE("stack", S_IRUSR, proc_pid_stack),
28a6d671 3225#endif
5968cece 3226#ifdef CONFIG_SCHED_INFO
f6e826ca 3227 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 3228#endif
9745512c 3229#ifdef CONFIG_LATENCYTOP
631f9c18 3230 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 3231#endif
8793d854 3232#ifdef CONFIG_PROC_PID_CPUSET
52de4779 3233 ONE("cpuset", S_IRUGO, proc_cpuset_show),
a424316c
PM		 3234#endif
3235#ifdef CONFIG_CGROUPS
006f4ac4 3236 ONE("cgroup", S_IRUGO, proc_cgroup_show),
28a6d671 3237#endif
6ba51e37 3238 ONE("oom_score", S_IRUGO, proc_oom_score),
fa0cbbf1 3239 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
a63d83f4 3240 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 3241#ifdef CONFIG_AUDITSYSCALL
631f9c18 3242 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
26ec3c64 3243 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 3244#endif
f4f154fd 3245#ifdef CONFIG_FAULT_INJECTION
631f9c18 3246 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 3247#endif
297c5d92 3248#ifdef CONFIG_TASK_IO_ACCOUNTING
19aadc98 3249 ONE("io", S_IRUSR, proc_tid_io_accounting),
297c5d92 3250#endif
f133ecca 3251#ifdef CONFIG_HARDWALL
d962c144 3252 ONE("hardwall", S_IRUGO, proc_pid_hardwall),
f133ecca 3253#endif
22d917d8
EB		 3254#ifdef CONFIG_USER_NS
3255 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3256 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
f76d207a 3257 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
9cc46516 3258 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
22d917d8 3259#endif
28a6d671
EB		 3260};
3261
f0c3b509 3262static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
28a6d671 3263{
f0c3b509
AV		 3264 return proc_pident_readdir(file, ctx,
3265 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
28a6d671
EB		 3266}
3267
00cd8dd3
AV		 3268static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3269{
7bcd6b0e
EB		 3270 return proc_pident_lookup(dir, dentry,
3271 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
28a6d671
EB		 3272}
3273
00977a59 3274static const struct file_operations proc_tid_base_operations = {
28a6d671 3275 .read = generic_read_dir,
f50752ea
AV		 3276 .iterate_shared = proc_tid_base_readdir,
3277 .llseek = generic_file_llseek,
28a6d671
EB		 3278};
3279
c5ef1c42 3280static const struct inode_operations proc_tid_base_inode_operations = {
28a6d671
EB		 3281 .lookup = proc_tid_base_lookup,
3282 .getattr = pid_getattr,
3283 .setattr = proc_setattr,
3284};
3285
c52a47ac 3286static int proc_task_instantiate(struct inode *dir,
c5141e6d 3287 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8 3288{
444ceed8 3289 struct inode *inode;
61a28784 3290 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 3291
3292 if (!inode)
3293 goto out;
3294 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3295 inode->i_op = &proc_tid_base_inode_operations;
3296 inode->i_fop = &proc_tid_base_operations;
3297 inode->i_flags|=S_IMMUTABLE;
aed54175 3298
bfe86848
MS		 3299 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3300 ARRAY_SIZE(tid_base_stuff)));
444ceed8 3301
fb045adb 3302 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 3303
3304 d_add(dentry, inode);
3305 /* Close the race of the process dying before we return the dentry */
0b728e19 3306 if (pid_revalidate(dentry, 0))
c52a47ac 3307 return 0;
444ceed8 3308out:
c52a47ac 3309 return -ENOENT;
444ceed8
EB		 3310}
3311
00cd8dd3 3312static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
28a6d671 3313{
c52a47ac 3314 int result = -ENOENT;
28a6d671
EB		 3315 struct task_struct *task;
3316 struct task_struct *leader = get_proc_task(dir);
28a6d671 3317 unsigned tid;
b488893a 3318 struct pid_namespace *ns;
28a6d671
EB		 3319
3320 if (!leader)
3321 goto out_no_task;
3322
dbcdb504 3323 tid = name_to_int(&dentry->d_name);
28a6d671
EB		 3324 if (tid == ~0U)
3325 goto out;
3326
b488893a 3327 ns = dentry->d_sb->s_fs_info;
28a6d671 3328 rcu_read_lock();
b488893a 3329 task = find_task_by_pid_ns(tid, ns);
28a6d671
EB		 3330 if (task)
3331 get_task_struct(task);
3332 rcu_read_unlock();
3333 if (!task)
3334 goto out;
bac0abd6 3335 if (!same_thread_group(leader, task))
28a6d671
EB		 3336 goto out_drop_task;
3337
444ceed8 3338 result = proc_task_instantiate(dir, dentry, task, NULL);
28a6d671
EB		 3339out_drop_task:
3340 put_task_struct(task);
3341out:
3342 put_task_struct(leader);
3343out_no_task:
c52a47ac 3344 return ERR_PTR(result);
28a6d671
EB		 3345}
3346
0bc58a91
EB		 3347/*
3348 * Find the first tid of a thread group to return to user space.
3349 *
3350 * Usually this is just the thread group leader, but if the users
3351 * buffer was too small or there was a seek into the middle of the
3352 * directory we have more work todo.
3353 *
3354 * In the case of a short read we start with find_task_by_pid.
3355 *
3356 * In the case of a seek we start with the leader and walk nr
3357 * threads past it.
3358 */
9f6e963f
ON		 3359static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3360 struct pid_namespace *ns)
0bc58a91 3361{
d855a4b7 3362 struct task_struct *pos, *task;
9f6e963f
ON		 3363 unsigned long nr = f_pos;
3364
3365 if (nr != f_pos) /* 32bit overflow? */
3366 return NULL;
1da177e4 3367
cc288738 3368 rcu_read_lock();
d855a4b7
ON		 3369 task = pid_task(pid, PIDTYPE_PID);
3370 if (!task)
3371 goto fail;
3372
3373 /* Attempt to start with the tid of a thread */
9f6e963f 3374 if (tid && nr) {
b488893a 3375 pos = find_task_by_pid_ns(tid, ns);
d855a4b7 3376 if (pos && same_thread_group(pos, task))
a872ff0c 3377 goto found;
0bc58a91 3378 }
1da177e4 3379
0bc58a91 3380 /* If nr exceeds the number of threads there is nothing todo */
9f6e963f 3381 if (nr >= get_nr_threads(task))
c986c14a 3382 goto fail;
1da177e4 3383
a872ff0c
ON		 3384 /* If we haven't found our starting place yet start
3385 * with the leader and walk nr threads forward.
0bc58a91 3386 */
d855a4b7 3387 pos = task = task->group_leader;
c986c14a 3388 do {
9f6e963f 3389 if (!nr--)
c986c14a 3390 goto found;
d855a4b7 3391 } while_each_thread(task, pos);
c986c14a
ON		 3392fail:
3393 pos = NULL;
3394 goto out;
a872ff0c
ON		 3395found:
3396 get_task_struct(pos);
3397out:
cc288738 3398 rcu_read_unlock();
0bc58a91
EB		 3399 return pos;
3400}
3401
3402/*
3403 * Find the next thread in the thread list.
3404 * Return NULL if there is an error or no next thread.
3405 *
3406 * The reference to the input task_struct is released.
3407 */
3408static struct task_struct *next_tid(struct task_struct *start)
3409{
c1df7fb8 3410 struct task_struct *pos = NULL;
cc288738 3411 rcu_read_lock();
c1df7fb8 3412 if (pid_alive(start)) {
0bc58a91 3413 pos = next_thread(start);
c1df7fb8
ON		 3414 if (thread_group_leader(pos))
3415 pos = NULL;
3416 else
3417 get_task_struct(pos);
3418 }
cc288738 3419 rcu_read_unlock();
0bc58a91
EB		 3420 put_task_struct(start);
3421 return pos;
1da177e4
LT		 3422}
3423
3424/* for the /proc/TGID/task/ directories */
f0c3b509 3425static int proc_task_readdir(struct file *file, struct dir_context *ctx)
1da177e4 3426{
d855a4b7
ON		 3427 struct inode *inode = file_inode(file);
3428 struct task_struct *task;
b488893a 3429 struct pid_namespace *ns;
f0c3b509 3430 int tid;
1da177e4 3431
d855a4b7 3432 if (proc_inode_is_dead(inode))
f0c3b509 3433 return -ENOENT;
1da177e4 3434
f0c3b509 3435 if (!dir_emit_dots(file, ctx))
d855a4b7 3436 return 0;
1da177e4 3437
0bc58a91
EB		 3438 /* f_version caches the tgid value that the last readdir call couldn't
3439 * return. lseek aka telldir automagically resets f_version to 0.
3440 */
3aa3377f 3441 ns = inode->i_sb->s_fs_info;
f0c3b509
AV		 3442 tid = (int)file->f_version;
3443 file->f_version = 0;
d855a4b7 3444 for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
0bc58a91 3445 task;
f0c3b509
AV		 3446 task = next_tid(task), ctx->pos++) {
3447 char name[PROC_NUMBUF];
3448 int len;
b488893a 3449 tid = task_pid_nr_ns(task, ns);
f0c3b509
AV		 3450 len = snprintf(name, sizeof(name), "%d", tid);
3451 if (!proc_fill_cache(file, ctx, name, len,
3452 proc_task_instantiate, task, NULL)) {
0bc58a91
EB		 3453 /* returning this tgid failed, save it as the first
3454 * pid for the next readir call */
f0c3b509 3455 file->f_version = (u64)tid;
0bc58a91 3456 put_task_struct(task);
1da177e4 3457 break;
0bc58a91 3458 }
1da177e4 3459 }
d855a4b7 3460
f0c3b509 3461 return 0;
1da177e4 3462}
6e66b52b
EB		 3463
3464static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3465{
2b0143b5 3466 struct inode *inode = d_inode(dentry);
99f89551 3467 struct task_struct *p = get_proc_task(inode);
6e66b52b
EB		 3468 generic_fillattr(inode, stat);
3469
99f89551 3470 if (p) {
99f89551 3471 stat->nlink += get_nr_threads(p);
99f89551 3472 put_task_struct(p);
6e66b52b
EB		 3473 }
3474
3475 return 0;
3476}
28a6d671 3477
c5ef1c42 3478static const struct inode_operations proc_task_inode_operations = {
28a6d671
EB		 3479 .lookup = proc_task_lookup,
3480 .getattr = proc_task_getattr,
3481 .setattr = proc_setattr,
0499680a 3482 .permission = proc_pid_permission,
28a6d671
EB		 3483};
3484
00977a59 3485static const struct file_operations proc_task_operations = {
28a6d671 3486 .read = generic_read_dir,
f50752ea
AV		 3487 .iterate_shared = proc_task_readdir,
3488 .llseek = generic_file_llseek,
28a6d671 3489};
Linux 6.x block layer and io_uring tree(s)