projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
proc: add cond_resched to /proc/kpage* read/write loop
[linux-2.6-block.git] / fs / proc / base.c
CommitLineData
1da177e4
LT		 1/*
2 * linux/fs/proc/base.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 *
6 * proc base directory handling functions
7 *
8 * 1999, Al Viro. Rewritten. Now it covers the whole per-process part.
9 * Instead of using magical inumbers to determine the kind of object
10 * we allocate and fill in-core inodes upon lookup. They don't even
11 * go into icache. We cache the reference to task_struct upon lookup too.
12 * Eventually it should become a filesystem in its own. We don't use the
13 * rest of procfs anymore.
e070ad49
ML		 14 *
15 *
16 * Changelog:
17 * 17-Jan-2005
18 * Allan Bezerra
19 * Bruna Moreira <bruna.moreira@indt.org.br>
20 * Edjard Mota <edjard.mota@indt.org.br>
21 * Ilias Biris <ilias.biris@indt.org.br>
22 * Mauricio Lin <mauricio.lin@indt.org.br>
23 *
24 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
25 *
26 * A new process specific entry (smaps) included in /proc. It shows the
27 * size of rss for each memory area. The maps entry lacks information
28 * about physical memory size (rss) for each mapped file, i.e.,
29 * rss information for executables and library files.
30 * This additional information is useful for any tools that need to know
31 * about physical memory consumption for a process specific library.
32 *
33 * Changelog:
34 * 21-Feb-2005
35 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
36 * Pud inclusion in the page table walking.
37 *
38 * ChangeLog:
39 * 10-Mar-2005
40 * 10LE Instituto Nokia de Tecnologia - INdT:
41 * A better way to walks through the page table as suggested by Hugh Dickins.
42 *
43 * Simo Piiroinen <simo.piiroinen@nokia.com>:
44 * Smaps information related to shared, private, clean and dirty pages.
45 *
46 * Paul Mundt <paul.mundt@nokia.com>:
47 * Overall revision about smaps.
1da177e4
LT		 48 */
49
50#include <asm/uaccess.h>
51
1da177e4
LT		 52#include <linux/errno.h>
53#include <linux/time.h>
54#include <linux/proc_fs.h>
55#include <linux/stat.h>
5995477a 56#include <linux/task_io_accounting_ops.h>
1da177e4 57#include <linux/init.h>
16f7e0fe 58#include <linux/capability.h>
1da177e4 59#include <linux/file.h>
9f3acc31 60#include <linux/fdtable.h>
1da177e4
LT		 61#include <linux/string.h>
62#include <linux/seq_file.h>
63#include <linux/namei.h>
6b3286ed 64#include <linux/mnt_namespace.h>
1da177e4 65#include <linux/mm.h>
a63d83f4 66#include <linux/swap.h>
b835996f 67#include <linux/rcupdate.h>
1da177e4 68#include <linux/kallsyms.h>
2ec220e2 69#include <linux/stacktrace.h>
d85f50d5 70#include <linux/resource.h>
5096add8 71#include <linux/module.h>
1da177e4
LT		 72#include <linux/mount.h>
73#include <linux/security.h>
74#include <linux/ptrace.h>
0d094efe 75#include <linux/tracehook.h>
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{
e7dcd999 406 struct mm_struct *mm = mm_access(task, PTRACE_MODE_READ);
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
25ce3191 433 if (lookup_symbol_name(wchan, symname) < 0) {
f83ce3e6
JE		 434 if (!ptrace_may_access(task, PTRACE_MODE_READ))
435 return 0;
25ce3191
JP		 436 seq_printf(m, "%lu", wchan);
437 } else {
438 seq_printf(m, "%s", symname);
439 }
440
441 return 0;
1da177e4
LT		 442}
443#endif /* CONFIG_KALLSYMS */
444
a9712bc1
AV		 445static int lock_trace(struct task_struct *task)
446{
447 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
448 if (err)
449 return err;
450 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
451 mutex_unlock(&task->signal->cred_guard_mutex);
452 return -EPERM;
453 }
454 return 0;
455}
456
457static void unlock_trace(struct task_struct *task)
458{
459 mutex_unlock(&task->signal->cred_guard_mutex);
460}
461
2ec220e2
KC		 462#ifdef CONFIG_STACKTRACE
463
464#define MAX_STACK_TRACE_DEPTH 64
465
466static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
467 struct pid *pid, struct task_struct *task)
468{
469 struct stack_trace trace;
470 unsigned long *entries;
a9712bc1 471 int err;
2ec220e2
KC		 472 int i;
473
474 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
475 if (!entries)
476 return -ENOMEM;
477
478 trace.nr_entries = 0;
479 trace.max_entries = MAX_STACK_TRACE_DEPTH;
480 trace.entries = entries;
481 trace.skip = 0;
2ec220e2 482
a9712bc1
AV		 483 err = lock_trace(task);
484 if (!err) {
485 save_stack_trace_tsk(task, &trace);
486
487 for (i = 0; i < trace.nr_entries; i++) {
b81a618d 488 seq_printf(m, "[<%pK>] %pS\n",
a9712bc1
AV		 489 (void *)entries[i], (void *)entries[i]);
490 }
491 unlock_trace(task);
2ec220e2
KC		 492 }
493 kfree(entries);
494
a9712bc1 495 return err;
2ec220e2
KC		 496}
497#endif
498
5968cece 499#ifdef CONFIG_SCHED_INFO
1da177e4
LT		 500/*
501 * Provides /proc/PID/schedstat
502 */
f6e826ca
AD		 503static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
504 struct pid *pid, struct task_struct *task)
1da177e4 505{
5968cece
NR		 506 if (unlikely(!sched_info_on()))
507 seq_printf(m, "0 0 0\n");
508 else
509 seq_printf(m, "%llu %llu %lu\n",
25ce3191
JP		 510 (unsigned long long)task->se.sum_exec_runtime,
511 (unsigned long long)task->sched_info.run_delay,
512 task->sched_info.pcount);
513
514 return 0;
1da177e4
LT		 515}
516#endif
517
9745512c
AV		 518#ifdef CONFIG_LATENCYTOP
519static int lstats_show_proc(struct seq_file *m, void *v)
520{
521 int i;
13d77c37
HS		 522 struct inode *inode = m->private;
523 struct task_struct *task = get_proc_task(inode);
9745512c 524
13d77c37
HS		 525 if (!task)
526 return -ESRCH;
527 seq_puts(m, "Latency Top version : v0.1\n");
9745512c 528 for (i = 0; i < 32; i++) {
34e49d4f
JP		 529 struct latency_record *lr = &task->latency_record[i];
530 if (lr->backtrace[0]) {
9745512c 531 int q;
34e49d4f
JP		 532 seq_printf(m, "%i %li %li",
533 lr->count, lr->time, lr->max);
9745512c 534 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
34e49d4f
JP		 535 unsigned long bt = lr->backtrace[q];
536 if (!bt)
9745512c 537 break;
34e49d4f 538 if (bt == ULONG_MAX)
9745512c 539 break;
34e49d4f 540 seq_printf(m, " %ps", (void *)bt);
9745512c 541 }
9d6de12f 542 seq_putc(m, '\n');
9745512c
AV		 543 }
544
545 }
13d77c37 546 put_task_struct(task);
9745512c
AV		 547 return 0;
548}
549
550static int lstats_open(struct inode *inode, struct file *file)
551{
13d77c37 552 return single_open(file, lstats_show_proc, inode);
d6643d12
HS		 553}
554
9745512c
AV		 555static ssize_t lstats_write(struct file *file, const char __user *buf,
556 size_t count, loff_t *offs)
557{
496ad9aa 558 struct task_struct *task = get_proc_task(file_inode(file));
9745512c 559
13d77c37
HS		 560 if (!task)
561 return -ESRCH;
9745512c 562 clear_all_latency_tracing(task);
13d77c37 563 put_task_struct(task);
9745512c
AV		 564
565 return count;
566}
567
568static const struct file_operations proc_lstats_operations = {
569 .open = lstats_open,
570 .read = seq_read,
571 .write = lstats_write,
572 .llseek = seq_lseek,
13d77c37 573 .release = single_release,
9745512c
AV		 574};
575
576#endif
577
6ba51e37
AD		 578static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
579 struct pid *pid, struct task_struct *task)
1da177e4 580{
a7f638f9 581 unsigned long totalpages = totalram_pages + total_swap_pages;
b95c35e7 582 unsigned long points = 0;
1da177e4 583
19c5d45a 584 read_lock(&tasklist_lock);
b95c35e7 585 if (pid_alive(task))
a7f638f9
DR		 586 points = oom_badness(task, NULL, NULL, totalpages) *
587 1000 / totalpages;
19c5d45a 588 read_unlock(&tasklist_lock);
25ce3191
JP		 589 seq_printf(m, "%lu\n", points);
590
591 return 0;
1da177e4
LT		 592}
593
d85f50d5 594struct limit_names {
cedbccab
AD		 595 const char *name;
596 const char *unit;
d85f50d5
NH		 597};
598
599static const struct limit_names lnames[RLIM_NLIMITS] = {
cff4edb5 600 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
d85f50d5
NH		 601 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
602 [RLIMIT_DATA] = {"Max data size", "bytes"},
603 [RLIMIT_STACK] = {"Max stack size", "bytes"},
604 [RLIMIT_CORE] = {"Max core file size", "bytes"},
605 [RLIMIT_RSS] = {"Max resident set", "bytes"},
606 [RLIMIT_NPROC] = {"Max processes", "processes"},
607 [RLIMIT_NOFILE] = {"Max open files", "files"},
608 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
609 [RLIMIT_AS] = {"Max address space", "bytes"},
610 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
611 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
612 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
613 [RLIMIT_NICE] = {"Max nice priority", NULL},
614 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
8808117c 615 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
d85f50d5
NH		 616};
617
618/* Display limits for a process */
1c963eb1
AD		 619static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
620 struct pid *pid, struct task_struct *task)
d85f50d5
NH		 621{
622 unsigned int i;
d85f50d5 623 unsigned long flags;
d85f50d5
NH		 624
625 struct rlimit rlim[RLIM_NLIMITS];
626
a6bebbc8 627 if (!lock_task_sighand(task, &flags))
d85f50d5 628 return 0;
d85f50d5
NH		 629 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
630 unlock_task_sighand(task, &flags);
d85f50d5
NH		 631
632 /*
633 * print the file header
634 */
1c963eb1 635 seq_printf(m, "%-25s %-20s %-20s %-10s\n",
25ce3191 636 "Limit", "Soft Limit", "Hard Limit", "Units");
d85f50d5
NH		 637
638 for (i = 0; i < RLIM_NLIMITS; i++) {
639 if (rlim[i].rlim_cur == RLIM_INFINITY)
1c963eb1 640 seq_printf(m, "%-25s %-20s ",
25ce3191 641 lnames[i].name, "unlimited");
d85f50d5 642 else
1c963eb1 643 seq_printf(m, "%-25s %-20lu ",
25ce3191 644 lnames[i].name, rlim[i].rlim_cur);
d85f50d5
NH		 645
646 if (rlim[i].rlim_max == RLIM_INFINITY)
1c963eb1 647 seq_printf(m, "%-20s ", "unlimited");
d85f50d5 648 else
1c963eb1 649 seq_printf(m, "%-20lu ", rlim[i].rlim_max);
d85f50d5
NH		 650
651 if (lnames[i].unit)
1c963eb1 652 seq_printf(m, "%-10s\n", lnames[i].unit);
d85f50d5 653 else
1c963eb1 654 seq_putc(m, '\n');
d85f50d5
NH		 655 }
656
1c963eb1 657 return 0;
d85f50d5
NH		 658}
659
ebcb6734 660#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
09d93bd6
AD		 661static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
662 struct pid *pid, struct task_struct *task)
ebcb6734
RM		 663{
664 long nr;
665 unsigned long args[6], sp, pc;
25ce3191
JP		 666 int res;
667
668 res = lock_trace(task);
a9712bc1
AV		 669 if (res)
670 return res;
ebcb6734
RM		 671
672 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
09d93bd6 673 seq_puts(m, "running\n");
a9712bc1 674 else if (nr < 0)
09d93bd6 675 seq_printf(m, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
a9712bc1 676 else
09d93bd6 677 seq_printf(m,
ebcb6734
RM		 678 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
679 nr,
680 args[0], args[1], args[2], args[3], args[4], args[5],
681 sp, pc);
a9712bc1 682 unlock_trace(task);
25ce3191
JP		 683
684 return 0;
ebcb6734
RM		 685}
686#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
687
1da177e4
LT		 688/************************************************************************/
689/* Here the fs part begins */
690/************************************************************************/
691
692/* permission checks */
778c1144 693static int proc_fd_access_allowed(struct inode *inode)
1da177e4 694{
778c1144
EB		 695 struct task_struct *task;
696 int allowed = 0;
df26c40e
EB		 697 /* Allow access to a task's file descriptors if it is us or we
698 * may use ptrace attach to the process and find out that
699 * information.
778c1144
EB		 700 */
701 task = get_proc_task(inode);
df26c40e 702 if (task) {
006ebb40 703 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
778c1144 704 put_task_struct(task);
df26c40e 705 }
778c1144 706 return allowed;
1da177e4
LT		 707}
708
6b4e306a 709int proc_setattr(struct dentry *dentry, struct iattr *attr)
6d76fa58
LT		 710{
711 int error;
2b0143b5 712 struct inode *inode = d_inode(dentry);
6d76fa58
LT		 713
714 if (attr->ia_valid & ATTR_MODE)
715 return -EPERM;
716
717 error = inode_change_ok(inode, attr);
1025774c
CH		 718 if (error)
719 return error;
720
1025774c
CH		 721 setattr_copy(inode, attr);
722 mark_inode_dirty(inode);
723 return 0;
6d76fa58
LT		 724}
725
0499680a
VK		 726/*
727 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
728 * or euid/egid (for hide_pid_min=2)?
729 */
730static bool has_pid_permissions(struct pid_namespace *pid,
731 struct task_struct *task,
732 int hide_pid_min)
733{
734 if (pid->hide_pid < hide_pid_min)
735 return true;
736 if (in_group_p(pid->pid_gid))
737 return true;
738 return ptrace_may_access(task, PTRACE_MODE_READ);
739}
740
741
742static int proc_pid_permission(struct inode *inode, int mask)
743{
744 struct pid_namespace *pid = inode->i_sb->s_fs_info;
745 struct task_struct *task;
746 bool has_perms;
747
748 task = get_proc_task(inode);
a2ef990a
XF		 749 if (!task)
750 return -ESRCH;
0499680a
VK		 751 has_perms = has_pid_permissions(pid, task, 1);
752 put_task_struct(task);
753
754 if (!has_perms) {
755 if (pid->hide_pid == 2) {
756 /*
757 * Let's make getdents(), stat(), and open()
758 * consistent with each other. If a process
759 * may not stat() a file, it shouldn't be seen
760 * in procfs at all.
761 */
762 return -ENOENT;
763 }
764
765 return -EPERM;
766 }
767 return generic_permission(inode, mask);
768}
769
770
771
c5ef1c42 772static const struct inode_operations proc_def_inode_operations = {
6d76fa58
LT		 773 .setattr = proc_setattr,
774};
775
be614086
EB		 776static int proc_single_show(struct seq_file *m, void *v)
777{
778 struct inode *inode = m->private;
779 struct pid_namespace *ns;
780 struct pid *pid;
781 struct task_struct *task;
782 int ret;
783
784 ns = inode->i_sb->s_fs_info;
785 pid = proc_pid(inode);
786 task = get_pid_task(pid, PIDTYPE_PID);
787 if (!task)
788 return -ESRCH;
789
790 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
791
792 put_task_struct(task);
793 return ret;
794}
795
796static int proc_single_open(struct inode *inode, struct file *filp)
797{
c6a34058 798 return single_open(filp, proc_single_show, inode);
be614086
EB		 799}
800
801static const struct file_operations proc_single_file_operations = {
802 .open = proc_single_open,
803 .read = seq_read,
804 .llseek = seq_lseek,
805 .release = single_release,
806};
807
5381e169
ON		 808
809struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
1da177e4 810{
5381e169
ON		 811 struct task_struct *task = get_proc_task(inode);
812 struct mm_struct *mm = ERR_PTR(-ESRCH);
e268337d 813
5381e169
ON		 814 if (task) {
815 mm = mm_access(task, mode);
816 put_task_struct(task);
e268337d 817
5381e169
ON		 818 if (!IS_ERR_OR_NULL(mm)) {
819 /* ensure this mm_struct can't be freed */
820 atomic_inc(&mm->mm_count);
821 /* but do not pin its memory */
822 mmput(mm);
823 }
824 }
825
826 return mm;
827}
828
829static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
830{
831 struct mm_struct *mm = proc_mem_open(inode, mode);
e268337d
LT		 832
833 if (IS_ERR(mm))
834 return PTR_ERR(mm);
835
e268337d 836 file->private_data = mm;
1da177e4
LT		 837 return 0;
838}
839
b409e578
CW		 840static int mem_open(struct inode *inode, struct file *file)
841{
bc452b4b
DH		 842 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
843
844 /* OK to pass negative loff_t, we can catch out-of-range */
845 file->f_mode |= FMODE_UNSIGNED_OFFSET;
846
847 return ret;
b409e578
CW		 848}
849
572d34b9
ON		 850static ssize_t mem_rw(struct file *file, char __user *buf,
851 size_t count, loff_t *ppos, int write)
1da177e4 852{
e268337d 853 struct mm_struct *mm = file->private_data;
572d34b9
ON		 854 unsigned long addr = *ppos;
855 ssize_t copied;
1da177e4 856 char *page;
1da177e4 857
e268337d
LT		 858 if (!mm)
859 return 0;
99f89551 860
30cd8903
KM		 861 page = (char *)__get_free_page(GFP_TEMPORARY);
862 if (!page)
e268337d 863 return -ENOMEM;
1da177e4 864
f7ca54f4 865 copied = 0;
6d08f2c7
ON		 866 if (!atomic_inc_not_zero(&mm->mm_users))
867 goto free;
868
1da177e4 869 while (count > 0) {
572d34b9 870 int this_len = min_t(int, count, PAGE_SIZE);
1da177e4 871
572d34b9 872 if (write && copy_from_user(page, buf, this_len)) {
1da177e4
LT		 873 copied = -EFAULT;
874 break;
875 }
572d34b9
ON		 876
877 this_len = access_remote_vm(mm, addr, page, this_len, write);
878 if (!this_len) {
1da177e4
LT		 879 if (!copied)
880 copied = -EIO;
881 break;
882 }
572d34b9
ON		 883
884 if (!write && copy_to_user(buf, page, this_len)) {
885 copied = -EFAULT;
886 break;
887 }
888
889 buf += this_len;
890 addr += this_len;
891 copied += this_len;
892 count -= this_len;
1da177e4 893 }
572d34b9 894 *ppos = addr;
30cd8903 895
6d08f2c7
ON		 896 mmput(mm);
897free:
30cd8903 898 free_page((unsigned long) page);
1da177e4
LT		 899 return copied;
900}
1da177e4 901
572d34b9
ON		 902static ssize_t mem_read(struct file *file, char __user *buf,
903 size_t count, loff_t *ppos)
904{
905 return mem_rw(file, buf, count, ppos, 0);
906}
907
908static ssize_t mem_write(struct file *file, const char __user *buf,
909 size_t count, loff_t *ppos)
910{
911 return mem_rw(file, (char __user*)buf, count, ppos, 1);
912}
913
85863e47 914loff_t mem_lseek(struct file *file, loff_t offset, int orig)
1da177e4
LT		 915{
916 switch (orig) {
917 case 0:
918 file->f_pos = offset;
919 break;
920 case 1:
921 file->f_pos += offset;
922 break;
923 default:
924 return -EINVAL;
925 }
926 force_successful_syscall_return();
927 return file->f_pos;
928}
929
e268337d
LT		 930static int mem_release(struct inode *inode, struct file *file)
931{
932 struct mm_struct *mm = file->private_data;
71879d3c 933 if (mm)
6d08f2c7 934 mmdrop(mm);
e268337d
LT		 935 return 0;
936}
937
00977a59 938static const struct file_operations proc_mem_operations = {
1da177e4
LT		 939 .llseek = mem_lseek,
940 .read = mem_read,
941 .write = mem_write,
942 .open = mem_open,
e268337d 943 .release = mem_release,
1da177e4
LT		 944};
945
b409e578
CW		 946static int environ_open(struct inode *inode, struct file *file)
947{
948 return __mem_open(inode, file, PTRACE_MODE_READ);
949}
950
315e28c8
JP		 951static ssize_t environ_read(struct file *file, char __user *buf,
952 size_t count, loff_t *ppos)
953{
315e28c8
JP		 954 char *page;
955 unsigned long src = *ppos;
b409e578
CW		 956 int ret = 0;
957 struct mm_struct *mm = file->private_data;
315e28c8 958
b409e578
CW		 959 if (!mm)
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;
315e28c8 969 while (count > 0) {
e8905ec2
DH		 970 size_t this_len, max_len;
971 int retval;
315e28c8 972
e8905ec2 973 if (src >= (mm->env_end - mm->env_start))
315e28c8
JP		 974 break;
975
e8905ec2
DH		 976 this_len = mm->env_end - (mm->env_start + src);
977
978 max_len = min_t(size_t, PAGE_SIZE, count);
979 this_len = min(max_len, this_len);
315e28c8 980
b409e578 981 retval = access_remote_vm(mm, (mm->env_start + src),
315e28c8
JP		 982 page, this_len, 0);
983
984 if (retval <= 0) {
985 ret = retval;
986 break;
987 }
988
989 if (copy_to_user(buf, page, retval)) {
990 ret = -EFAULT;
991 break;
992 }
993
994 ret += retval;
995 src += retval;
996 buf += retval;
997 count -= retval;
998 }
999 *ppos = src;
315e28c8 1000 mmput(mm);
b409e578
CW		 1001
1002free:
315e28c8 1003 free_page((unsigned long) page);
315e28c8
JP		 1004 return ret;
1005}
1006
1007static const struct file_operations proc_environ_operations = {
b409e578 1008 .open = environ_open,
315e28c8 1009 .read = environ_read,
87df8424 1010 .llseek = generic_file_llseek,
b409e578 1011 .release = mem_release,
315e28c8
JP		 1012};
1013
fa0cbbf1
DR		 1014static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
1015 loff_t *ppos)
1016{
496ad9aa 1017 struct task_struct *task = get_proc_task(file_inode(file));
fa0cbbf1
DR		 1018 char buffer[PROC_NUMBUF];
1019 int oom_adj = OOM_ADJUST_MIN;
1020 size_t len;
1021 unsigned long flags;
1022
1023 if (!task)
1024 return -ESRCH;
1025 if (lock_task_sighand(task, &flags)) {
1026 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
1027 oom_adj = OOM_ADJUST_MAX;
1028 else
1029 oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
1030 OOM_SCORE_ADJ_MAX;
1031 unlock_task_sighand(task, &flags);
1032 }
1033 put_task_struct(task);
1034 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
1035 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1036}
1037
1038static ssize_t oom_adj_write(struct file *file, const char __user *buf,
1039 size_t count, loff_t *ppos)
1040{
1041 struct task_struct *task;
1042 char buffer[PROC_NUMBUF];
1043 int oom_adj;
1044 unsigned long flags;
1045 int err;
1046
1047 memset(buffer, 0, sizeof(buffer));
1048 if (count > sizeof(buffer) - 1)
1049 count = sizeof(buffer) - 1;
1050 if (copy_from_user(buffer, buf, count)) {
1051 err = -EFAULT;
1052 goto out;
1053 }
1054
1055 err = kstrtoint(strstrip(buffer), 0, &oom_adj);
1056 if (err)
1057 goto out;
1058 if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
1059 oom_adj != OOM_DISABLE) {
1060 err = -EINVAL;
1061 goto out;
1062 }
1063
496ad9aa 1064 task = get_proc_task(file_inode(file));
fa0cbbf1
DR		 1065 if (!task) {
1066 err = -ESRCH;
1067 goto out;
1068 }
1069
1070 task_lock(task);
1071 if (!task->mm) {
1072 err = -EINVAL;
1073 goto err_task_lock;
1074 }
1075
1076 if (!lock_task_sighand(task, &flags)) {
1077 err = -ESRCH;
1078 goto err_task_lock;
1079 }
1080
1081 /*
1082 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1083 * value is always attainable.
1084 */
1085 if (oom_adj == OOM_ADJUST_MAX)
1086 oom_adj = OOM_SCORE_ADJ_MAX;
1087 else
1088 oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
1089
1090 if (oom_adj < task->signal->oom_score_adj &&
1091 !capable(CAP_SYS_RESOURCE)) {
1092 err = -EACCES;
1093 goto err_sighand;
1094 }
1095
1096 /*
1097 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1098 * /proc/pid/oom_score_adj instead.
1099 */
87ebdc00 1100 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
fa0cbbf1
DR		 1101 current->comm, task_pid_nr(current), task_pid_nr(task),
1102 task_pid_nr(task));
1103
1104 task->signal->oom_score_adj = oom_adj;
1105 trace_oom_score_adj_update(task);
1106err_sighand:
1107 unlock_task_sighand(task, &flags);
1108err_task_lock:
1109 task_unlock(task);
1110 put_task_struct(task);
1111out:
1112 return err < 0 ? err : count;
1113}
1114
1115static const struct file_operations proc_oom_adj_operations = {
1116 .read = oom_adj_read,
1117 .write = oom_adj_write,
1118 .llseek = generic_file_llseek,
1119};
1120
a63d83f4
DR		 1121static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1122 size_t count, loff_t *ppos)
1123{
496ad9aa 1124 struct task_struct *task = get_proc_task(file_inode(file));
a63d83f4 1125 char buffer[PROC_NUMBUF];
a9c58b90 1126 short oom_score_adj = OOM_SCORE_ADJ_MIN;
a63d83f4
DR		 1127 unsigned long flags;
1128 size_t len;
1129
1130 if (!task)
1131 return -ESRCH;
1132 if (lock_task_sighand(task, &flags)) {
1133 oom_score_adj = task->signal->oom_score_adj;
1134 unlock_task_sighand(task, &flags);
1135 }
1136 put_task_struct(task);
a9c58b90 1137 len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
a63d83f4
DR		 1138 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1139}
1140
1141static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1142 size_t count, loff_t *ppos)
1143{
1144 struct task_struct *task;
1145 char buffer[PROC_NUMBUF];
1146 unsigned long flags;
0a8cb8e3 1147 int oom_score_adj;
a63d83f4
DR		 1148 int err;
1149
1150 memset(buffer, 0, sizeof(buffer));
1151 if (count > sizeof(buffer) - 1)
1152 count = sizeof(buffer) - 1;
723548bf
DR		 1153 if (copy_from_user(buffer, buf, count)) {
1154 err = -EFAULT;
1155 goto out;
1156 }
a63d83f4 1157
0a8cb8e3 1158 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
a63d83f4 1159 if (err)
723548bf 1160 goto out;
a63d83f4 1161 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
723548bf
DR		 1162 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1163 err = -EINVAL;
1164 goto out;
1165 }
a63d83f4 1166
496ad9aa 1167 task = get_proc_task(file_inode(file));
723548bf
DR		 1168 if (!task) {
1169 err = -ESRCH;
1170 goto out;
1171 }
d19d5476
DR		 1172
1173 task_lock(task);
1174 if (!task->mm) {
1175 err = -EINVAL;
1176 goto err_task_lock;
1177 }
1178
a63d83f4 1179 if (!lock_task_sighand(task, &flags)) {
723548bf 1180 err = -ESRCH;
d19d5476 1181 goto err_task_lock;
a63d83f4 1182 }
d19d5476 1183
a9c58b90 1184 if ((short)oom_score_adj < task->signal->oom_score_adj_min &&
a63d83f4 1185 !capable(CAP_SYS_RESOURCE)) {
723548bf
DR		 1186 err = -EACCES;
1187 goto err_sighand;
a63d83f4
DR		 1188 }
1189
a9c58b90 1190 task->signal->oom_score_adj = (short)oom_score_adj;
dabb16f6 1191 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
a9c58b90 1192 task->signal->oom_score_adj_min = (short)oom_score_adj;
43d2b113 1193 trace_oom_score_adj_update(task);
01dc52eb 1194
723548bf 1195err_sighand:
a63d83f4 1196 unlock_task_sighand(task, &flags);
d19d5476
DR		 1197err_task_lock:
1198 task_unlock(task);
a63d83f4 1199 put_task_struct(task);
723548bf
DR		 1200out:
1201 return err < 0 ? err : count;
a63d83f4
DR		 1202}
1203
1204static const struct file_operations proc_oom_score_adj_operations = {
1205 .read = oom_score_adj_read,
1206 .write = oom_score_adj_write,
6038f373 1207 .llseek = default_llseek,
a63d83f4
DR		 1208};
1209
1da177e4
LT		 1210#ifdef CONFIG_AUDITSYSCALL
1211#define TMPBUFLEN 21
1212static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1213 size_t count, loff_t *ppos)
1214{
496ad9aa 1215 struct inode * inode = file_inode(file);
99f89551 1216 struct task_struct *task = get_proc_task(inode);
1da177e4
LT		 1217 ssize_t length;
1218 char tmpbuf[TMPBUFLEN];
1219
99f89551
EB		 1220 if (!task)
1221 return -ESRCH;
1da177e4 1222 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
e1760bd5
EB		 1223 from_kuid(file->f_cred->user_ns,
1224 audit_get_loginuid(task)));
99f89551 1225 put_task_struct(task);
1da177e4
LT		 1226 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1227}
1228
1229static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1230 size_t count, loff_t *ppos)
1231{
496ad9aa 1232 struct inode * inode = file_inode(file);
1da177e4
LT		 1233 char *page, *tmp;
1234 ssize_t length;
1da177e4 1235 uid_t loginuid;
e1760bd5 1236 kuid_t kloginuid;
1da177e4 1237
7dc52157
PM		 1238 rcu_read_lock();
1239 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1240 rcu_read_unlock();
1da177e4 1241 return -EPERM;
7dc52157
PM		 1242 }
1243 rcu_read_unlock();
1da177e4 1244
e0182909
AV		 1245 if (count >= PAGE_SIZE)
1246 count = PAGE_SIZE - 1;
1da177e4
LT		 1247
1248 if (*ppos != 0) {
1249 /* No partial writes. */
1250 return -EINVAL;
1251 }
e12ba74d 1252 page = (char*)__get_free_page(GFP_TEMPORARY);
1da177e4
LT		 1253 if (!page)
1254 return -ENOMEM;
1255 length = -EFAULT;
1256 if (copy_from_user(page, buf, count))
1257 goto out_free_page;
1258
e0182909 1259 page[count] = '\0';
1da177e4
LT		 1260 loginuid = simple_strtoul(page, &tmp, 10);
1261 if (tmp == page) {
1262 length = -EINVAL;
1263 goto out_free_page;
1264
1265 }
81407c84
EP		 1266
1267 /* is userspace tring to explicitly UNSET the loginuid? */
1268 if (loginuid == AUDIT_UID_UNSET) {
1269 kloginuid = INVALID_UID;
1270 } else {
1271 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1272 if (!uid_valid(kloginuid)) {
1273 length = -EINVAL;
1274 goto out_free_page;
1275 }
e1760bd5
EB		 1276 }
1277
1278 length = audit_set_loginuid(kloginuid);
1da177e4
LT		 1279 if (likely(length == 0))
1280 length = count;
1281
1282out_free_page:
1283 free_page((unsigned long) page);
1284 return length;
1285}
1286
00977a59 1287static const struct file_operations proc_loginuid_operations = {
1da177e4
LT		 1288 .read = proc_loginuid_read,
1289 .write = proc_loginuid_write,
87df8424 1290 .llseek = generic_file_llseek,
1da177e4 1291};
1e0bd755
EP		 1292
1293static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1294 size_t count, loff_t *ppos)
1295{
496ad9aa 1296 struct inode * inode = file_inode(file);
1e0bd755
EP		 1297 struct task_struct *task = get_proc_task(inode);
1298 ssize_t length;
1299 char tmpbuf[TMPBUFLEN];
1300
1301 if (!task)
1302 return -ESRCH;
1303 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1304 audit_get_sessionid(task));
1305 put_task_struct(task);
1306 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1307}
1308
1309static const struct file_operations proc_sessionid_operations = {
1310 .read = proc_sessionid_read,
87df8424 1311 .llseek = generic_file_llseek,
1e0bd755 1312};
1da177e4
LT		 1313#endif
1314
f4f154fd
AM		 1315#ifdef CONFIG_FAULT_INJECTION
1316static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1317 size_t count, loff_t *ppos)
1318{
496ad9aa 1319 struct task_struct *task = get_proc_task(file_inode(file));
f4f154fd
AM		 1320 char buffer[PROC_NUMBUF];
1321 size_t len;
1322 int make_it_fail;
f4f154fd
AM		 1323
1324 if (!task)
1325 return -ESRCH;
1326 make_it_fail = task->make_it_fail;
1327 put_task_struct(task);
1328
1329 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
0c28f287
AM		 1330
1331 return simple_read_from_buffer(buf, count, ppos, buffer, len);
f4f154fd
AM		 1332}
1333
1334static ssize_t proc_fault_inject_write(struct file * file,
1335 const char __user * buf, size_t count, loff_t *ppos)
1336{
1337 struct task_struct *task;
1338 char buffer[PROC_NUMBUF], *end;
1339 int make_it_fail;
1340
1341 if (!capable(CAP_SYS_RESOURCE))
1342 return -EPERM;
1343 memset(buffer, 0, sizeof(buffer));
1344 if (count > sizeof(buffer) - 1)
1345 count = sizeof(buffer) - 1;
1346 if (copy_from_user(buffer, buf, count))
1347 return -EFAULT;
cba8aafe
VL		 1348 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1349 if (*end)
1350 return -EINVAL;
16caed31
DJ		 1351 if (make_it_fail < 0 || make_it_fail > 1)
1352 return -EINVAL;
1353
496ad9aa 1354 task = get_proc_task(file_inode(file));
f4f154fd
AM		 1355 if (!task)
1356 return -ESRCH;
1357 task->make_it_fail = make_it_fail;
1358 put_task_struct(task);
cba8aafe
VL		 1359
1360 return count;
f4f154fd
AM		 1361}
1362
00977a59 1363static const struct file_operations proc_fault_inject_operations = {
f4f154fd
AM		 1364 .read = proc_fault_inject_read,
1365 .write = proc_fault_inject_write,
87df8424 1366 .llseek = generic_file_llseek,
f4f154fd
AM		 1367};
1368#endif
1369
9745512c 1370
43ae34cb
IM		 1371#ifdef CONFIG_SCHED_DEBUG
1372/*
1373 * Print out various scheduling related per-task fields:
1374 */
1375static int sched_show(struct seq_file *m, void *v)
1376{
1377 struct inode *inode = m->private;
1378 struct task_struct *p;
1379
43ae34cb
IM		 1380 p = get_proc_task(inode);
1381 if (!p)
1382 return -ESRCH;
1383 proc_sched_show_task(p, m);
1384
1385 put_task_struct(p);
1386
1387 return 0;
1388}
1389
1390static ssize_t
1391sched_write(struct file *file, const char __user *buf,
1392 size_t count, loff_t *offset)
1393{
496ad9aa 1394 struct inode *inode = file_inode(file);
43ae34cb
IM		 1395 struct task_struct *p;
1396
43ae34cb
IM		 1397 p = get_proc_task(inode);
1398 if (!p)
1399 return -ESRCH;
1400 proc_sched_set_task(p);
1401
1402 put_task_struct(p);
1403
1404 return count;
1405}
1406
1407static int sched_open(struct inode *inode, struct file *filp)
1408{
c6a34058 1409 return single_open(filp, sched_show, inode);
43ae34cb
IM		 1410}
1411
1412static const struct file_operations proc_pid_sched_operations = {
1413 .open = sched_open,
1414 .read = seq_read,
1415 .write = sched_write,
1416 .llseek = seq_lseek,
5ea473a1 1417 .release = single_release,
43ae34cb
IM		 1418};
1419
1420#endif
1421
5091faa4
MG		 1422#ifdef CONFIG_SCHED_AUTOGROUP
1423/*
1424 * Print out autogroup related information:
1425 */
1426static int sched_autogroup_show(struct seq_file *m, void *v)
1427{
1428 struct inode *inode = m->private;
1429 struct task_struct *p;
1430
1431 p = get_proc_task(inode);
1432 if (!p)
1433 return -ESRCH;
1434 proc_sched_autogroup_show_task(p, m);
1435
1436 put_task_struct(p);
1437
1438 return 0;
1439}
1440
1441static ssize_t
1442sched_autogroup_write(struct file *file, const char __user *buf,
1443 size_t count, loff_t *offset)
1444{
496ad9aa 1445 struct inode *inode = file_inode(file);
5091faa4
MG		 1446 struct task_struct *p;
1447 char buffer[PROC_NUMBUF];
0a8cb8e3 1448 int nice;
5091faa4
MG		 1449 int err;
1450
1451 memset(buffer, 0, sizeof(buffer));
1452 if (count > sizeof(buffer) - 1)
1453 count = sizeof(buffer) - 1;
1454 if (copy_from_user(buffer, buf, count))
1455 return -EFAULT;
1456
0a8cb8e3
AD		 1457 err = kstrtoint(strstrip(buffer), 0, &nice);
1458 if (err < 0)
1459 return err;
5091faa4
MG		 1460
1461 p = get_proc_task(inode);
1462 if (!p)
1463 return -ESRCH;
1464
2e5b5b3a 1465 err = proc_sched_autogroup_set_nice(p, nice);
5091faa4
MG		 1466 if (err)
1467 count = err;
1468
1469 put_task_struct(p);
1470
1471 return count;
1472}
1473
1474static int sched_autogroup_open(struct inode *inode, struct file *filp)
1475{
1476 int ret;
1477
1478 ret = single_open(filp, sched_autogroup_show, NULL);
1479 if (!ret) {
1480 struct seq_file *m = filp->private_data;
1481
1482 m->private = inode;
1483 }
1484 return ret;
1485}
1486
1487static const struct file_operations proc_pid_sched_autogroup_operations = {
1488 .open = sched_autogroup_open,
1489 .read = seq_read,
1490 .write = sched_autogroup_write,
1491 .llseek = seq_lseek,
1492 .release = single_release,
1493};
1494
1495#endif /* CONFIG_SCHED_AUTOGROUP */
1496
4614a696 1497static ssize_t comm_write(struct file *file, const char __user *buf,
1498 size_t count, loff_t *offset)
1499{
496ad9aa 1500 struct inode *inode = file_inode(file);
4614a696 1501 struct task_struct *p;
1502 char buffer[TASK_COMM_LEN];
830e0fc9 1503 const size_t maxlen = sizeof(buffer) - 1;
4614a696 1504
1505 memset(buffer, 0, sizeof(buffer));
830e0fc9 1506 if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
4614a696 1507 return -EFAULT;
1508
1509 p = get_proc_task(inode);
1510 if (!p)
1511 return -ESRCH;
1512
1513 if (same_thread_group(current, p))
1514 set_task_comm(p, buffer);
1515 else
1516 count = -EINVAL;
1517
1518 put_task_struct(p);
1519
1520 return count;
1521}
1522
1523static int comm_show(struct seq_file *m, void *v)
1524{
1525 struct inode *inode = m->private;
1526 struct task_struct *p;
1527
1528 p = get_proc_task(inode);
1529 if (!p)
1530 return -ESRCH;
1531
1532 task_lock(p);
1533 seq_printf(m, "%s\n", p->comm);
1534 task_unlock(p);
1535
1536 put_task_struct(p);
1537
1538 return 0;
1539}
1540
1541static int comm_open(struct inode *inode, struct file *filp)
1542{
c6a34058 1543 return single_open(filp, comm_show, inode);
4614a696 1544}
1545
1546static const struct file_operations proc_pid_set_comm_operations = {
1547 .open = comm_open,
1548 .read = seq_read,
1549 .write = comm_write,
1550 .llseek = seq_lseek,
1551 .release = single_release,
1552};
1553
7773fbc5 1554static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
925d1c40
MH		 1555{
1556 struct task_struct *task;
1557 struct mm_struct *mm;
1558 struct file *exe_file;
1559
2b0143b5 1560 task = get_proc_task(d_inode(dentry));
925d1c40
MH		 1561 if (!task)
1562 return -ENOENT;
1563 mm = get_task_mm(task);
1564 put_task_struct(task);
1565 if (!mm)
1566 return -ENOENT;
1567 exe_file = get_mm_exe_file(mm);
1568 mmput(mm);
1569 if (exe_file) {
1570 *exe_path = exe_file->f_path;
1571 path_get(&exe_file->f_path);
1572 fput(exe_file);
1573 return 0;
1574 } else
1575 return -ENOENT;
1576}
1577
6e77137b 1578static const char *proc_pid_follow_link(struct dentry *dentry, void **cookie)
1da177e4 1579{
2b0143b5 1580 struct inode *inode = d_inode(dentry);
408ef013 1581 struct path path;
1da177e4
LT		 1582 int error = -EACCES;
1583
778c1144
EB		 1584 /* Are we allowed to snoop on the tasks file descriptors? */
1585 if (!proc_fd_access_allowed(inode))
1da177e4 1586 goto out;
1da177e4 1587
408ef013
CH		 1588 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1589 if (error)
1590 goto out;
1591
6e77137b 1592 nd_jump_link(&path);
408ef013 1593 return NULL;
1da177e4 1594out:
008b150a 1595 return ERR_PTR(error);
1da177e4
LT		 1596}
1597
3dcd25f3 1598static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1da177e4 1599{
e12ba74d 1600 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
3dcd25f3 1601 char *pathname;
1da177e4
LT		 1602 int len;
1603
1604 if (!tmp)
1605 return -ENOMEM;
0c28f287 1606
7b2a69ba 1607 pathname = d_path(path, tmp, PAGE_SIZE);
3dcd25f3
JB		 1608 len = PTR_ERR(pathname);
1609 if (IS_ERR(pathname))
1da177e4 1610 goto out;
3dcd25f3 1611 len = tmp + PAGE_SIZE - 1 - pathname;
1da177e4
LT		 1612
1613 if (len > buflen)
1614 len = buflen;
3dcd25f3 1615 if (copy_to_user(buffer, pathname, len))
1da177e4
LT		 1616 len = -EFAULT;
1617 out:
1618 free_page((unsigned long)tmp);
1619 return len;
1620}
1621
1622static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1623{
1624 int error = -EACCES;
2b0143b5 1625 struct inode *inode = d_inode(dentry);
3dcd25f3 1626 struct path path;
1da177e4 1627
778c1144
EB		 1628 /* Are we allowed to snoop on the tasks file descriptors? */
1629 if (!proc_fd_access_allowed(inode))
1da177e4 1630 goto out;
1da177e4 1631
7773fbc5 1632 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1da177e4
LT		 1633 if (error)
1634 goto out;
1635
3dcd25f3
JB		 1636 error = do_proc_readlink(&path, buffer, buflen);
1637 path_put(&path);
1da177e4 1638out:
1da177e4
LT		 1639 return error;
1640}
1641
faf60af1 1642const struct inode_operations proc_pid_link_inode_operations = {
1da177e4 1643 .readlink = proc_pid_readlink,
6d76fa58
LT		 1644 .follow_link = proc_pid_follow_link,
1645 .setattr = proc_setattr,
1da177e4
LT		 1646};
1647
28a6d671
EB		 1648
1649/* building an inode */
1650
6b4e306a 1651struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
28a6d671
EB		 1652{
1653 struct inode * inode;
1654 struct proc_inode *ei;
c69e8d9c 1655 const struct cred *cred;
1da177e4 1656
28a6d671 1657 /* We need a new inode */
1da177e4 1658
28a6d671
EB		 1659 inode = new_inode(sb);
1660 if (!inode)
1661 goto out;
1662
1663 /* Common stuff */
1664 ei = PROC_I(inode);
85fe4025 1665 inode->i_ino = get_next_ino();
28a6d671 1666 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
28a6d671
EB		 1667 inode->i_op = &proc_def_inode_operations;
1668
1669 /*
1670 * grab the reference to task.
1671 */
1a657f78 1672 ei->pid = get_task_pid(task, PIDTYPE_PID);
28a6d671
EB		 1673 if (!ei->pid)
1674 goto out_unlock;
1675
28a6d671 1676 if (task_dumpable(task)) {
c69e8d9c
DH		 1677 rcu_read_lock();
1678 cred = __task_cred(task);
1679 inode->i_uid = cred->euid;
1680 inode->i_gid = cred->egid;
1681 rcu_read_unlock();
1da177e4 1682 }
28a6d671
EB		 1683 security_task_to_inode(task, inode);
1684
1da177e4 1685out:
28a6d671
EB		 1686 return inode;
1687
1688out_unlock:
1689 iput(inode);
1690 return NULL;
1da177e4
LT		 1691}
1692
6b4e306a 1693int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1da177e4 1694{
2b0143b5 1695 struct inode *inode = d_inode(dentry);
28a6d671 1696 struct task_struct *task;
c69e8d9c 1697 const struct cred *cred;
0499680a 1698 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
c69e8d9c 1699
28a6d671 1700 generic_fillattr(inode, stat);
1da177e4 1701
28a6d671 1702 rcu_read_lock();
dcb0f222
EB		 1703 stat->uid = GLOBAL_ROOT_UID;
1704 stat->gid = GLOBAL_ROOT_GID;
28a6d671
EB		 1705 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1706 if (task) {
0499680a
VK		 1707 if (!has_pid_permissions(pid, task, 2)) {
1708 rcu_read_unlock();
1709 /*
1710 * This doesn't prevent learning whether PID exists,
1711 * it only makes getattr() consistent with readdir().
1712 */
1713 return -ENOENT;
1714 }
28a6d671
EB		 1715 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1716 task_dumpable(task)) {
c69e8d9c
DH		 1717 cred = __task_cred(task);
1718 stat->uid = cred->euid;
1719 stat->gid = cred->egid;
1da177e4
LT		 1720 }
1721 }
28a6d671 1722 rcu_read_unlock();
d6e71144 1723 return 0;
1da177e4
LT		 1724}
1725
1da177e4
LT		 1726/* dentry stuff */
1727
1728/*
1729 * Exceptional case: normally we are not allowed to unhash a busy
1730 * directory. In this case, however, we can do it - no aliasing problems
1731 * due to the way we treat inodes.
1732 *
1733 * Rewrite the inode's ownerships here because the owning task may have
1734 * performed a setuid(), etc.
99f89551
EB		 1735 *
1736 * Before the /proc/pid/status file was created the only way to read
1737 * the effective uid of a /process was to stat /proc/pid. Reading
1738 * /proc/pid/status is slow enough that procps and other packages
1739 * kept stating /proc/pid. To keep the rules in /proc simple I have
1740 * made this apply to all per process world readable and executable
1741 * directories.
1da177e4 1742 */
0b728e19 1743int pid_revalidate(struct dentry *dentry, unsigned int flags)
1da177e4 1744{
34286d66
NP		 1745 struct inode *inode;
1746 struct task_struct *task;
c69e8d9c
DH		 1747 const struct cred *cred;
1748
0b728e19 1749 if (flags & LOOKUP_RCU)
34286d66
NP		 1750 return -ECHILD;
1751
2b0143b5 1752 inode = d_inode(dentry);
34286d66
NP		 1753 task = get_proc_task(inode);
1754
99f89551
EB		 1755 if (task) {
1756 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1757 task_dumpable(task)) {
c69e8d9c
DH		 1758 rcu_read_lock();
1759 cred = __task_cred(task);
1760 inode->i_uid = cred->euid;
1761 inode->i_gid = cred->egid;
1762 rcu_read_unlock();
1da177e4 1763 } else {
dcb0f222
EB		 1764 inode->i_uid = GLOBAL_ROOT_UID;
1765 inode->i_gid = GLOBAL_ROOT_GID;
1da177e4 1766 }
9ee8ab9f 1767 inode->i_mode &= ~(S_ISUID | S_ISGID);
1da177e4 1768 security_task_to_inode(task, inode);
99f89551 1769 put_task_struct(task);
1da177e4
LT		 1770 return 1;
1771 }
1da177e4
LT		 1772 return 0;
1773}
1774
d855a4b7
ON		 1775static inline bool proc_inode_is_dead(struct inode *inode)
1776{
1777 return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
1778}
1779
1dd704b6
DH		 1780int pid_delete_dentry(const struct dentry *dentry)
1781{
1782 /* Is the task we represent dead?
1783 * If so, then don't put the dentry on the lru list,
1784 * kill it immediately.
1785 */
2b0143b5 1786 return proc_inode_is_dead(d_inode(dentry));
1dd704b6
DH		 1787}
1788
6b4e306a 1789const struct dentry_operations pid_dentry_operations =
28a6d671
EB		 1790{
1791 .d_revalidate = pid_revalidate,
1792 .d_delete = pid_delete_dentry,
1793};
1794
1795/* Lookups */
1796
1c0d04c9
EB		 1797/*
1798 * Fill a directory entry.
1799 *
1800 * If possible create the dcache entry and derive our inode number and
1801 * file type from dcache entry.
1802 *
1803 * Since all of the proc inode numbers are dynamically generated, the inode
1804 * numbers do not exist until the inode is cache. This means creating the
1805 * the dcache entry in readdir is necessary to keep the inode numbers
1806 * reported by readdir in sync with the inode numbers reported
1807 * by stat.
1808 */
f0c3b509 1809bool proc_fill_cache(struct file *file, struct dir_context *ctx,
6b4e306a 1810 const char *name, int len,
c5141e6d 1811 instantiate_t instantiate, struct task_struct *task, const void *ptr)
61a28784 1812{
f0c3b509 1813 struct dentry *child, *dir = file->f_path.dentry;
1df98b8b 1814 struct qstr qname = QSTR_INIT(name, len);
61a28784 1815 struct inode *inode;
1df98b8b
AV		 1816 unsigned type;
1817 ino_t ino;
61a28784 1818
1df98b8b 1819 child = d_hash_and_lookup(dir, &qname);
61a28784 1820 if (!child) {
1df98b8b
AV		 1821 child = d_alloc(dir, &qname);
1822 if (!child)
1823 goto end_instantiate;
2b0143b5 1824 if (instantiate(d_inode(dir), child, task, ptr) < 0) {
1df98b8b
AV		 1825 dput(child);
1826 goto end_instantiate;
61a28784
EB		 1827 }
1828 }
2b0143b5 1829 inode = d_inode(child);
147ce699
AV		 1830 ino = inode->i_ino;
1831 type = inode->i_mode >> 12;
61a28784 1832 dput(child);
f0c3b509 1833 return dir_emit(ctx, name, len, ino, type);
1df98b8b
AV		 1834
1835end_instantiate:
1836 return dir_emit(ctx, name, len, 1, DT_UNKNOWN);
61a28784
EB		 1837}
1838
640708a2
PE		 1839#ifdef CONFIG_CHECKPOINT_RESTORE
1840
1841/*
1842 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1843 * which represent vma start and end addresses.
1844 */
1845static int dname_to_vma_addr(struct dentry *dentry,
1846 unsigned long *start, unsigned long *end)
1847{
1848 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
1849 return -EINVAL;
1850
1851 return 0;
1852}
1853
0b728e19 1854static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
640708a2
PE		 1855{
1856 unsigned long vm_start, vm_end;
1857 bool exact_vma_exists = false;
1858 struct mm_struct *mm = NULL;
1859 struct task_struct *task;
1860 const struct cred *cred;
1861 struct inode *inode;
1862 int status = 0;
1863
0b728e19 1864 if (flags & LOOKUP_RCU)
640708a2
PE		 1865 return -ECHILD;
1866
1867 if (!capable(CAP_SYS_ADMIN)) {
41735818 1868 status = -EPERM;
640708a2
PE		 1869 goto out_notask;
1870 }
1871
2b0143b5 1872 inode = d_inode(dentry);
640708a2
PE		 1873 task = get_proc_task(inode);
1874 if (!task)
1875 goto out_notask;
1876
2344bec7
CW		 1877 mm = mm_access(task, PTRACE_MODE_READ);
1878 if (IS_ERR_OR_NULL(mm))
640708a2
PE		 1879 goto out;
1880
1881 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1882 down_read(&mm->mmap_sem);
1883 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1884 up_read(&mm->mmap_sem);
1885 }
1886
1887 mmput(mm);
1888
1889 if (exact_vma_exists) {
1890 if (task_dumpable(task)) {
1891 rcu_read_lock();
1892 cred = __task_cred(task);
1893 inode->i_uid = cred->euid;
1894 inode->i_gid = cred->egid;
1895 rcu_read_unlock();
1896 } else {
dcb0f222
EB		 1897 inode->i_uid = GLOBAL_ROOT_UID;
1898 inode->i_gid = GLOBAL_ROOT_GID;
640708a2
PE		 1899 }
1900 security_task_to_inode(task, inode);
1901 status = 1;
1902 }
1903
1904out:
1905 put_task_struct(task);
1906
1907out_notask:
640708a2
PE		 1908 return status;
1909}
1910
1911static const struct dentry_operations tid_map_files_dentry_operations = {
1912 .d_revalidate = map_files_d_revalidate,
1913 .d_delete = pid_delete_dentry,
1914};
1915
1916static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
1917{
1918 unsigned long vm_start, vm_end;
1919 struct vm_area_struct *vma;
1920 struct task_struct *task;
1921 struct mm_struct *mm;
1922 int rc;
1923
1924 rc = -ENOENT;
2b0143b5 1925 task = get_proc_task(d_inode(dentry));
640708a2
PE		 1926 if (!task)
1927 goto out;
1928
1929 mm = get_task_mm(task);
1930 put_task_struct(task);
1931 if (!mm)
1932 goto out;
1933
1934 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
1935 if (rc)
1936 goto out_mmput;
1937
70335abb 1938 rc = -ENOENT;
640708a2
PE		 1939 down_read(&mm->mmap_sem);
1940 vma = find_exact_vma(mm, vm_start, vm_end);
1941 if (vma && vma->vm_file) {
1942 *path = vma->vm_file->f_path;
1943 path_get(path);
1944 rc = 0;
1945 }
1946 up_read(&mm->mmap_sem);
1947
1948out_mmput:
1949 mmput(mm);
1950out:
1951 return rc;
1952}
1953
1954struct map_files_info {
7b540d06 1955 fmode_t mode;
640708a2
PE		 1956 unsigned long len;
1957 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
1958};
1959
c52a47ac 1960static int
640708a2
PE		 1961proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
1962 struct task_struct *task, const void *ptr)
1963{
7b540d06 1964 fmode_t mode = (fmode_t)(unsigned long)ptr;
640708a2
PE		 1965 struct proc_inode *ei;
1966 struct inode *inode;
1967
640708a2
PE		 1968 inode = proc_pid_make_inode(dir->i_sb, task);
1969 if (!inode)
c52a47ac 1970 return -ENOENT;
640708a2
PE		 1971
1972 ei = PROC_I(inode);
1973 ei->op.proc_get_link = proc_map_files_get_link;
1974
1975 inode->i_op = &proc_pid_link_inode_operations;
1976 inode->i_size = 64;
1977 inode->i_mode = S_IFLNK;
1978
7b540d06 1979 if (mode & FMODE_READ)
640708a2 1980 inode->i_mode |= S_IRUSR;
7b540d06 1981 if (mode & FMODE_WRITE)
640708a2
PE		 1982 inode->i_mode |= S_IWUSR;
1983
1984 d_set_d_op(dentry, &tid_map_files_dentry_operations);
1985 d_add(dentry, inode);
1986
c52a47ac 1987 return 0;
640708a2
PE		 1988}
1989
1990static struct dentry *proc_map_files_lookup(struct inode *dir,
00cd8dd3 1991 struct dentry *dentry, unsigned int flags)
640708a2
PE		 1992{
1993 unsigned long vm_start, vm_end;
1994 struct vm_area_struct *vma;
1995 struct task_struct *task;
c52a47ac 1996 int result;
640708a2
PE		 1997 struct mm_struct *mm;
1998
c52a47ac 1999 result = -EPERM;
640708a2
PE		 2000 if (!capable(CAP_SYS_ADMIN))
2001 goto out;
2002
c52a47ac 2003 result = -ENOENT;
640708a2
PE		 2004 task = get_proc_task(dir);
2005 if (!task)
2006 goto out;
2007
c52a47ac 2008 result = -EACCES;
eb94cd96 2009 if (!ptrace_may_access(task, PTRACE_MODE_READ))
640708a2
PE		 2010 goto out_put_task;
2011
c52a47ac 2012 result = -ENOENT;
640708a2 2013 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
eb94cd96 2014 goto out_put_task;
640708a2
PE		 2015
2016 mm = get_task_mm(task);
2017 if (!mm)
eb94cd96 2018 goto out_put_task;
640708a2
PE		 2019
2020 down_read(&mm->mmap_sem);
2021 vma = find_exact_vma(mm, vm_start, vm_end);
2022 if (!vma)
2023 goto out_no_vma;
2024
05f56484
SK		 2025 if (vma->vm_file)
2026 result = proc_map_files_instantiate(dir, dentry, task,
2027 (void *)(unsigned long)vma->vm_file->f_mode);
640708a2
PE		 2028
2029out_no_vma:
2030 up_read(&mm->mmap_sem);
2031 mmput(mm);
640708a2
PE		 2032out_put_task:
2033 put_task_struct(task);
2034out:
c52a47ac 2035 return ERR_PTR(result);
640708a2
PE		 2036}
2037
2038static const struct inode_operations proc_map_files_inode_operations = {
2039 .lookup = proc_map_files_lookup,
2040 .permission = proc_fd_permission,
2041 .setattr = proc_setattr,
2042};
2043
2044static int
f0c3b509 2045proc_map_files_readdir(struct file *file, struct dir_context *ctx)
640708a2 2046{
640708a2
PE		 2047 struct vm_area_struct *vma;
2048 struct task_struct *task;
2049 struct mm_struct *mm;
f0c3b509
AV		 2050 unsigned long nr_files, pos, i;
2051 struct flex_array *fa = NULL;
2052 struct map_files_info info;
2053 struct map_files_info *p;
640708a2
PE		 2054 int ret;
2055
41735818 2056 ret = -EPERM;
640708a2
PE		 2057 if (!capable(CAP_SYS_ADMIN))
2058 goto out;
2059
2060 ret = -ENOENT;
f0c3b509 2061 task = get_proc_task(file_inode(file));
640708a2
PE		 2062 if (!task)
2063 goto out;
2064
2065 ret = -EACCES;
eb94cd96 2066 if (!ptrace_may_access(task, PTRACE_MODE_READ))
640708a2
PE		 2067 goto out_put_task;
2068
2069 ret = 0;
f0c3b509
AV		 2070 if (!dir_emit_dots(file, ctx))
2071 goto out_put_task;
640708a2 2072
f0c3b509
AV		 2073 mm = get_task_mm(task);
2074 if (!mm)
2075 goto out_put_task;
2076 down_read(&mm->mmap_sem);
640708a2 2077
f0c3b509 2078 nr_files = 0;
640708a2 2079
f0c3b509
AV		 2080 /*
2081 * We need two passes here:
2082 *
2083 * 1) Collect vmas of mapped files with mmap_sem taken
2084 * 2) Release mmap_sem and instantiate entries
2085 *
2086 * otherwise we get lockdep complained, since filldir()
2087 * routine might require mmap_sem taken in might_fault().
2088 */
640708a2 2089
f0c3b509
AV		 2090 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2091 if (vma->vm_file && ++pos > ctx->pos)
2092 nr_files++;
2093 }
2094
2095 if (nr_files) {
2096 fa = flex_array_alloc(sizeof(info), nr_files,
2097 GFP_KERNEL);
2098 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2099 GFP_KERNEL)) {
2100 ret = -ENOMEM;
2101 if (fa)
2102 flex_array_free(fa);
2103 up_read(&mm->mmap_sem);
2104 mmput(mm);
2105 goto out_put_task;
640708a2 2106 }
f0c3b509
AV		 2107 for (i = 0, vma = mm->mmap, pos = 2; vma;
2108 vma = vma->vm_next) {
2109 if (!vma->vm_file)
2110 continue;
2111 if (++pos <= ctx->pos)
2112 continue;
2113
2114 info.mode = vma->vm_file->f_mode;
2115 info.len = snprintf(info.name,
2116 sizeof(info.name), "%lx-%lx",
2117 vma->vm_start, vma->vm_end);
2118 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2119 BUG();
640708a2 2120 }
640708a2 2121 }
f0c3b509
AV		 2122 up_read(&mm->mmap_sem);
2123
2124 for (i = 0; i < nr_files; i++) {
2125 p = flex_array_get(fa, i);
2126 if (!proc_fill_cache(file, ctx,
2127 p->name, p->len,
2128 proc_map_files_instantiate,
2129 task,
2130 (void *)(unsigned long)p->mode))
2131 break;
2132 ctx->pos++;
640708a2 2133 }
f0c3b509
AV		 2134 if (fa)
2135 flex_array_free(fa);
2136 mmput(mm);
640708a2 2137
640708a2
PE		 2138out_put_task:
2139 put_task_struct(task);
2140out:
2141 return ret;
2142}
2143
2144static const struct file_operations proc_map_files_operations = {
2145 .read = generic_read_dir,
f0c3b509 2146 .iterate = proc_map_files_readdir,
640708a2
PE		 2147 .llseek = default_llseek,
2148};
2149
48f6a7a5
PE		 2150struct timers_private {
2151 struct pid *pid;
2152 struct task_struct *task;
2153 struct sighand_struct *sighand;
57b8015e 2154 struct pid_namespace *ns;
48f6a7a5
PE		 2155 unsigned long flags;
2156};
2157
2158static void *timers_start(struct seq_file *m, loff_t *pos)
2159{
2160 struct timers_private *tp = m->private;
2161
2162 tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2163 if (!tp->task)
2164 return ERR_PTR(-ESRCH);
2165
2166 tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2167 if (!tp->sighand)
2168 return ERR_PTR(-ESRCH);
2169
2170 return seq_list_start(&tp->task->signal->posix_timers, *pos);
2171}
2172
2173static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2174{
2175 struct timers_private *tp = m->private;
2176 return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2177}
2178
2179static void timers_stop(struct seq_file *m, void *v)
2180{
2181 struct timers_private *tp = m->private;
2182
2183 if (tp->sighand) {
2184 unlock_task_sighand(tp->task, &tp->flags);
2185 tp->sighand = NULL;
2186 }
2187
2188 if (tp->task) {
2189 put_task_struct(tp->task);
2190 tp->task = NULL;
2191 }
2192}
2193
2194static int show_timer(struct seq_file *m, void *v)
2195{
2196 struct k_itimer *timer;
57b8015e
PE		 2197 struct timers_private *tp = m->private;
2198 int notify;
cedbccab 2199 static const char * const nstr[] = {
57b8015e
PE		 2200 [SIGEV_SIGNAL] = "signal",
2201 [SIGEV_NONE] = "none",
2202 [SIGEV_THREAD] = "thread",
2203 };
48f6a7a5
PE		 2204
2205 timer = list_entry((struct list_head *)v, struct k_itimer, list);
57b8015e
PE		 2206 notify = timer->it_sigev_notify;
2207
48f6a7a5 2208 seq_printf(m, "ID: %d\n", timer->it_id);
25ce3191
JP		 2209 seq_printf(m, "signal: %d/%p\n",
2210 timer->sigq->info.si_signo,
2211 timer->sigq->info.si_value.sival_ptr);
57b8015e 2212 seq_printf(m, "notify: %s/%s.%d\n",
25ce3191
JP		 2213 nstr[notify & ~SIGEV_THREAD_ID],
2214 (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2215 pid_nr_ns(timer->it_pid, tp->ns));
15ef0298 2216 seq_printf(m, "ClockID: %d\n", timer->it_clock);
48f6a7a5
PE		 2217
2218 return 0;
2219}
2220
2221static const struct seq_operations proc_timers_seq_ops = {
2222 .start = timers_start,
2223 .next = timers_next,
2224 .stop = timers_stop,
2225 .show = show_timer,
2226};
2227
2228static int proc_timers_open(struct inode *inode, struct file *file)
2229{
2230 struct timers_private *tp;
2231
2232 tp = __seq_open_private(file, &proc_timers_seq_ops,
2233 sizeof(struct timers_private));
2234 if (!tp)
2235 return -ENOMEM;
2236
2237 tp->pid = proc_pid(inode);
57b8015e 2238 tp->ns = inode->i_sb->s_fs_info;
48f6a7a5
PE		 2239 return 0;
2240}
2241
2242static const struct file_operations proc_timers_operations = {
2243 .open = proc_timers_open,
2244 .read = seq_read,
2245 .llseek = seq_lseek,
2246 .release = seq_release_private,
2247};
640708a2
PE		 2248#endif /* CONFIG_CHECKPOINT_RESTORE */
2249
c52a47ac 2250static int proc_pident_instantiate(struct inode *dir,
c5141e6d 2251 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8 2252{
c5141e6d 2253 const struct pid_entry *p = ptr;
444ceed8
EB		 2254 struct inode *inode;
2255 struct proc_inode *ei;
444ceed8 2256
61a28784 2257 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 2258 if (!inode)
2259 goto out;
2260
2261 ei = PROC_I(inode);
2262 inode->i_mode = p->mode;
2263 if (S_ISDIR(inode->i_mode))
bfe86848 2264 set_nlink(inode, 2); /* Use getattr to fix if necessary */
444ceed8
EB		 2265 if (p->iop)
2266 inode->i_op = p->iop;
2267 if (p->fop)
2268 inode->i_fop = p->fop;
2269 ei->op = p->op;
fb045adb 2270 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 2271 d_add(dentry, inode);
2272 /* Close the race of the process dying before we return the dentry */
0b728e19 2273 if (pid_revalidate(dentry, 0))
c52a47ac 2274 return 0;
444ceed8 2275out:
c52a47ac 2276 return -ENOENT;
444ceed8
EB		 2277}
2278
1da177e4
LT		 2279static struct dentry *proc_pident_lookup(struct inode *dir,
2280 struct dentry *dentry,
c5141e6d 2281 const struct pid_entry *ents,
7bcd6b0e 2282 unsigned int nents)
1da177e4 2283{
c52a47ac 2284 int error;
99f89551 2285 struct task_struct *task = get_proc_task(dir);
c5141e6d 2286 const struct pid_entry *p, *last;
1da177e4 2287
c52a47ac 2288 error = -ENOENT;
1da177e4 2289
99f89551
EB		 2290 if (!task)
2291 goto out_no_task;
1da177e4 2292
20cdc894
EB		 2293 /*
2294 * Yes, it does not scale. And it should not. Don't add
2295 * new entries into /proc/<tgid>/ without very good reasons.
2296 */
7bcd6b0e
EB		 2297 last = &ents[nents - 1];
2298 for (p = ents; p <= last; p++) {
1da177e4
LT		 2299 if (p->len != dentry->d_name.len)
2300 continue;
2301 if (!memcmp(dentry->d_name.name, p->name, p->len))
2302 break;
2303 }
7bcd6b0e 2304 if (p > last)
1da177e4
LT		 2305 goto out;
2306
444ceed8 2307 error = proc_pident_instantiate(dir, dentry, task, p);
1da177e4 2308out:
99f89551
EB		 2309 put_task_struct(task);
2310out_no_task:
c52a47ac 2311 return ERR_PTR(error);
1da177e4
LT		 2312}
2313
f0c3b509 2314static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
c5141e6d 2315 const struct pid_entry *ents, unsigned int nents)
28a6d671 2316{
f0c3b509
AV		 2317 struct task_struct *task = get_proc_task(file_inode(file));
2318 const struct pid_entry *p;
28a6d671 2319
28a6d671 2320 if (!task)
f0c3b509 2321 return -ENOENT;
28a6d671 2322
f0c3b509
AV		 2323 if (!dir_emit_dots(file, ctx))
2324 goto out;
2325
2326 if (ctx->pos >= nents + 2)
2327 goto out;
28a6d671 2328
f0c3b509
AV		 2329 for (p = ents + (ctx->pos - 2); p <= ents + nents - 1; p++) {
2330 if (!proc_fill_cache(file, ctx, p->name, p->len,
2331 proc_pident_instantiate, task, p))
2332 break;
2333 ctx->pos++;
2334 }
28a6d671 2335out:
61a28784 2336 put_task_struct(task);
f0c3b509 2337 return 0;
1da177e4
LT		 2338}
2339
28a6d671
EB		 2340#ifdef CONFIG_SECURITY
2341static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2342 size_t count, loff_t *ppos)
2343{
496ad9aa 2344 struct inode * inode = file_inode(file);
04ff9708 2345 char *p = NULL;
28a6d671
EB		 2346 ssize_t length;
2347 struct task_struct *task = get_proc_task(inode);
2348
28a6d671 2349 if (!task)
04ff9708 2350 return -ESRCH;
28a6d671
EB		 2351
2352 length = security_getprocattr(task,
2fddfeef 2353 (char*)file->f_path.dentry->d_name.name,
04ff9708 2354 &p);
28a6d671 2355 put_task_struct(task);
04ff9708
AV		 2356 if (length > 0)
2357 length = simple_read_from_buffer(buf, count, ppos, p, length);
2358 kfree(p);
28a6d671 2359 return length;
1da177e4
LT		 2360}
2361
28a6d671
EB		 2362static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2363 size_t count, loff_t *ppos)
2364{
496ad9aa 2365 struct inode * inode = file_inode(file);
28a6d671
EB		 2366 char *page;
2367 ssize_t length;
2368 struct task_struct *task = get_proc_task(inode);
2369
2370 length = -ESRCH;
2371 if (!task)
2372 goto out_no_task;
2373 if (count > PAGE_SIZE)
2374 count = PAGE_SIZE;
2375
2376 /* No partial writes. */
2377 length = -EINVAL;
2378 if (*ppos != 0)
2379 goto out;
2380
2381 length = -ENOMEM;
e12ba74d 2382 page = (char*)__get_free_page(GFP_TEMPORARY);
28a6d671
EB		 2383 if (!page)
2384 goto out;
2385
2386 length = -EFAULT;
2387 if (copy_from_user(page, buf, count))
2388 goto out_free;
2389
107db7c7 2390 /* Guard against adverse ptrace interaction */
9b1bf12d 2391 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
107db7c7
DH		 2392 if (length < 0)
2393 goto out_free;
2394
28a6d671 2395 length = security_setprocattr(task,
2fddfeef 2396 (char*)file->f_path.dentry->d_name.name,
28a6d671 2397 (void*)page, count);
9b1bf12d 2398 mutex_unlock(&task->signal->cred_guard_mutex);
28a6d671
EB		 2399out_free:
2400 free_page((unsigned long) page);
2401out:
2402 put_task_struct(task);
2403out_no_task:
2404 return length;
2405}
2406
00977a59 2407static const struct file_operations proc_pid_attr_operations = {
28a6d671
EB		 2408 .read = proc_pid_attr_read,
2409 .write = proc_pid_attr_write,
87df8424 2410 .llseek = generic_file_llseek,
28a6d671
EB		 2411};
2412
c5141e6d 2413static const struct pid_entry attr_dir_stuff[] = {
631f9c18
AD		 2414 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2415 REG("prev", S_IRUGO, proc_pid_attr_operations),
2416 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2417 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2418 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2419 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
28a6d671
EB		 2420};
2421
f0c3b509 2422static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
28a6d671 2423{
f0c3b509
AV		 2424 return proc_pident_readdir(file, ctx,
2425 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2426}
2427
00977a59 2428static const struct file_operations proc_attr_dir_operations = {
1da177e4 2429 .read = generic_read_dir,
f0c3b509 2430 .iterate = proc_attr_dir_readdir,
6038f373 2431 .llseek = default_llseek,
1da177e4
LT		 2432};
2433
72d9dcfc 2434static struct dentry *proc_attr_dir_lookup(struct inode *dir,
00cd8dd3 2435 struct dentry *dentry, unsigned int flags)
28a6d671 2436{
7bcd6b0e
EB		 2437 return proc_pident_lookup(dir, dentry,
2438 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2439}
2440
c5ef1c42 2441static const struct inode_operations proc_attr_dir_inode_operations = {
72d9dcfc 2442 .lookup = proc_attr_dir_lookup,
99f89551 2443 .getattr = pid_getattr,
6d76fa58 2444 .setattr = proc_setattr,
1da177e4
LT		 2445};
2446
28a6d671
EB		 2447#endif
2448
698ba7b5 2449#ifdef CONFIG_ELF_CORE
3cb4a0bb
KH		 2450static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2451 size_t count, loff_t *ppos)
2452{
496ad9aa 2453 struct task_struct *task = get_proc_task(file_inode(file));
3cb4a0bb
KH		 2454 struct mm_struct *mm;
2455 char buffer[PROC_NUMBUF];
2456 size_t len;
2457 int ret;
2458
2459 if (!task)
2460 return -ESRCH;
2461
2462 ret = 0;
2463 mm = get_task_mm(task);
2464 if (mm) {
2465 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2466 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2467 MMF_DUMP_FILTER_SHIFT));
2468 mmput(mm);
2469 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2470 }
2471
2472 put_task_struct(task);
2473
2474 return ret;
2475}
2476
2477static ssize_t proc_coredump_filter_write(struct file *file,
2478 const char __user *buf,
2479 size_t count,
2480 loff_t *ppos)
2481{
2482 struct task_struct *task;
2483 struct mm_struct *mm;
2484 char buffer[PROC_NUMBUF], *end;
2485 unsigned int val;
2486 int ret;
2487 int i;
2488 unsigned long mask;
2489
2490 ret = -EFAULT;
2491 memset(buffer, 0, sizeof(buffer));
2492 if (count > sizeof(buffer) - 1)
2493 count = sizeof(buffer) - 1;
2494 if (copy_from_user(buffer, buf, count))
2495 goto out_no_task;
2496
2497 ret = -EINVAL;
2498 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2499 if (*end == '\n')
2500 end++;
2501 if (end - buffer == 0)
2502 goto out_no_task;
2503
2504 ret = -ESRCH;
496ad9aa 2505 task = get_proc_task(file_inode(file));
3cb4a0bb
KH		 2506 if (!task)
2507 goto out_no_task;
2508
2509 ret = end - buffer;
2510 mm = get_task_mm(task);
2511 if (!mm)
2512 goto out_no_mm;
2513
2514 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2515 if (val & mask)
2516 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2517 else
2518 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2519 }
2520
2521 mmput(mm);
2522 out_no_mm:
2523 put_task_struct(task);
2524 out_no_task:
2525 return ret;
2526}
2527
2528static const struct file_operations proc_coredump_filter_operations = {
2529 .read = proc_coredump_filter_read,
2530 .write = proc_coredump_filter_write,
87df8424 2531 .llseek = generic_file_llseek,
3cb4a0bb
KH		 2532};
2533#endif
2534
aba76fdb 2535#ifdef CONFIG_TASK_IO_ACCOUNTING
19aadc98 2536static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
297c5d92 2537{
940389b8 2538 struct task_io_accounting acct = task->ioac;
5995477a 2539 unsigned long flags;
293eb1e7 2540 int result;
5995477a 2541
293eb1e7
VK		 2542 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2543 if (result)
2544 return result;
2545
2546 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2547 result = -EACCES;
2548 goto out_unlock;
2549 }
1d1221f3 2550
5995477a
AR		 2551 if (whole && lock_task_sighand(task, &flags)) {
2552 struct task_struct *t = task;
2553
2554 task_io_accounting_add(&acct, &task->signal->ioac);
2555 while_each_thread(task, t)
2556 task_io_accounting_add(&acct, &t->ioac);
2557
2558 unlock_task_sighand(task, &flags);
297c5d92 2559 }
25ce3191
JP		 2560 seq_printf(m,
2561 "rchar: %llu\n"
2562 "wchar: %llu\n"
2563 "syscr: %llu\n"
2564 "syscw: %llu\n"
2565 "read_bytes: %llu\n"
2566 "write_bytes: %llu\n"
2567 "cancelled_write_bytes: %llu\n",
2568 (unsigned long long)acct.rchar,
2569 (unsigned long long)acct.wchar,
2570 (unsigned long long)acct.syscr,
2571 (unsigned long long)acct.syscw,
2572 (unsigned long long)acct.read_bytes,
2573 (unsigned long long)acct.write_bytes,
2574 (unsigned long long)acct.cancelled_write_bytes);
2575 result = 0;
2576
293eb1e7
VK		 2577out_unlock:
2578 mutex_unlock(&task->signal->cred_guard_mutex);
2579 return result;
297c5d92
AR		 2580}
2581
19aadc98
AD		 2582static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2583 struct pid *pid, struct task_struct *task)
297c5d92 2584{
19aadc98 2585 return do_io_accounting(task, m, 0);
aba76fdb 2586}
297c5d92 2587
19aadc98
AD		 2588static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2589 struct pid *pid, struct task_struct *task)
297c5d92 2590{
19aadc98 2591 return do_io_accounting(task, m, 1);
297c5d92
AR		 2592}
2593#endif /* CONFIG_TASK_IO_ACCOUNTING */
aba76fdb 2594
22d917d8
EB		 2595#ifdef CONFIG_USER_NS
2596static int proc_id_map_open(struct inode *inode, struct file *file,
ccf94f1b 2597 const struct seq_operations *seq_ops)
22d917d8
EB		 2598{
2599 struct user_namespace *ns = NULL;
2600 struct task_struct *task;
2601 struct seq_file *seq;
2602 int ret = -EINVAL;
2603
2604 task = get_proc_task(inode);
2605 if (task) {
2606 rcu_read_lock();
2607 ns = get_user_ns(task_cred_xxx(task, user_ns));
2608 rcu_read_unlock();
2609 put_task_struct(task);
2610 }
2611 if (!ns)
2612 goto err;
2613
2614 ret = seq_open(file, seq_ops);
2615 if (ret)
2616 goto err_put_ns;
2617
2618 seq = file->private_data;
2619 seq->private = ns;
2620
2621 return 0;
2622err_put_ns:
2623 put_user_ns(ns);
2624err:
2625 return ret;
2626}
2627
2628static int proc_id_map_release(struct inode *inode, struct file *file)
2629{
2630 struct seq_file *seq = file->private_data;
2631 struct user_namespace *ns = seq->private;
2632 put_user_ns(ns);
2633 return seq_release(inode, file);
2634}
2635
2636static int proc_uid_map_open(struct inode *inode, struct file *file)
2637{
2638 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2639}
2640
2641static int proc_gid_map_open(struct inode *inode, struct file *file)
2642{
2643 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2644}
2645
f76d207a
EB		 2646static int proc_projid_map_open(struct inode *inode, struct file *file)
2647{
2648 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2649}
2650
22d917d8
EB		 2651static const struct file_operations proc_uid_map_operations = {
2652 .open = proc_uid_map_open,
2653 .write = proc_uid_map_write,
2654 .read = seq_read,
2655 .llseek = seq_lseek,
2656 .release = proc_id_map_release,
2657};
2658
2659static const struct file_operations proc_gid_map_operations = {
2660 .open = proc_gid_map_open,
2661 .write = proc_gid_map_write,
2662 .read = seq_read,
2663 .llseek = seq_lseek,
2664 .release = proc_id_map_release,
2665};
f76d207a
EB		 2666
2667static const struct file_operations proc_projid_map_operations = {
2668 .open = proc_projid_map_open,
2669 .write = proc_projid_map_write,
2670 .read = seq_read,
2671 .llseek = seq_lseek,
2672 .release = proc_id_map_release,
2673};
9cc46516
EB		 2674
2675static int proc_setgroups_open(struct inode *inode, struct file *file)
2676{
2677 struct user_namespace *ns = NULL;
2678 struct task_struct *task;
2679 int ret;
2680
2681 ret = -ESRCH;
2682 task = get_proc_task(inode);
2683 if (task) {
2684 rcu_read_lock();
2685 ns = get_user_ns(task_cred_xxx(task, user_ns));
2686 rcu_read_unlock();
2687 put_task_struct(task);
2688 }
2689 if (!ns)
2690 goto err;
2691
2692 if (file->f_mode & FMODE_WRITE) {
2693 ret = -EACCES;
2694 if (!ns_capable(ns, CAP_SYS_ADMIN))
2695 goto err_put_ns;
2696 }
2697
2698 ret = single_open(file, &proc_setgroups_show, ns);
2699 if (ret)
2700 goto err_put_ns;
2701
2702 return 0;
2703err_put_ns:
2704 put_user_ns(ns);
2705err:
2706 return ret;
2707}
2708
2709static int proc_setgroups_release(struct inode *inode, struct file *file)
2710{
2711 struct seq_file *seq = file->private_data;
2712 struct user_namespace *ns = seq->private;
2713 int ret = single_release(inode, file);
2714 put_user_ns(ns);
2715 return ret;
2716}
2717
2718static const struct file_operations proc_setgroups_operations = {
2719 .open = proc_setgroups_open,
2720 .write = proc_setgroups_write,
2721 .read = seq_read,
2722 .llseek = seq_lseek,
2723 .release = proc_setgroups_release,
2724};
22d917d8
EB		 2725#endif /* CONFIG_USER_NS */
2726
47830723
KC		 2727static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2728 struct pid *pid, struct task_struct *task)
2729{
a9712bc1
AV		 2730 int err = lock_trace(task);
2731 if (!err) {
2732 seq_printf(m, "%08x\n", task->personality);
2733 unlock_trace(task);
2734 }
2735 return err;
47830723
KC		 2736}
2737
28a6d671
EB		 2738/*
2739 * Thread groups
2740 */
00977a59 2741static const struct file_operations proc_task_operations;
c5ef1c42 2742static const struct inode_operations proc_task_inode_operations;
20cdc894 2743
c5141e6d 2744static const struct pid_entry tgid_base_stuff[] = {
631f9c18
AD		 2745 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2746 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
640708a2
PE		 2747#ifdef CONFIG_CHECKPOINT_RESTORE
2748 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
2749#endif
631f9c18 2750 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 2751 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
b2211a36 2752#ifdef CONFIG_NET
631f9c18 2753 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
b2211a36 2754#endif
631f9c18 2755 REG("environ", S_IRUSR, proc_environ_operations),
f9ea536e 2756 ONE("auxv", S_IRUSR, proc_pid_auxv),
631f9c18 2757 ONE("status", S_IRUGO, proc_pid_status),
35a35046 2758 ONE("personality", S_IRUSR, proc_pid_personality),
1c963eb1 2759 ONE("limits", S_IRUGO, proc_pid_limits),
43ae34cb 2760#ifdef CONFIG_SCHED_DEBUG
631f9c18 2761 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
5091faa4
MG		 2762#endif
2763#ifdef CONFIG_SCHED_AUTOGROUP
2764 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
ebcb6734 2765#endif
4614a696 2766 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 2767#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
09d93bd6 2768 ONE("syscall", S_IRUSR, proc_pid_syscall),
43ae34cb 2769#endif
c2c0bb44 2770 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
631f9c18
AD		 2771 ONE("stat", S_IRUGO, proc_tgid_stat),
2772 ONE("statm", S_IRUGO, proc_pid_statm),
b7643757 2773 REG("maps", S_IRUGO, proc_pid_maps_operations),
28a6d671 2774#ifdef CONFIG_NUMA
b7643757 2775 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
28a6d671 2776#endif
631f9c18
AD		 2777 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2778 LNK("cwd", proc_cwd_link),
2779 LNK("root", proc_root_link),
2780 LNK("exe", proc_exe_link),
2781 REG("mounts", S_IRUGO, proc_mounts_operations),
2782 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2783 REG("mountstats", S_IRUSR, proc_mountstats_operations),
1e883281 2784#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18 2785 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
b7643757 2786 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
32ed74a4 2787 REG("pagemap", S_IRUSR, proc_pagemap_operations),
28a6d671
EB		 2788#endif
2789#ifdef CONFIG_SECURITY
631f9c18 2790 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 2791#endif
2792#ifdef CONFIG_KALLSYMS
edfcd606 2793 ONE("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 2794#endif
2ec220e2 2795#ifdef CONFIG_STACKTRACE
35a35046 2796 ONE("stack", S_IRUSR, proc_pid_stack),
28a6d671 2797#endif
5968cece 2798#ifdef CONFIG_SCHED_INFO
f6e826ca 2799 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 2800#endif
9745512c 2801#ifdef CONFIG_LATENCYTOP
631f9c18 2802 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 2803#endif
8793d854 2804#ifdef CONFIG_PROC_PID_CPUSET
52de4779 2805 ONE("cpuset", S_IRUGO, proc_cpuset_show),
a424316c
PM		 2806#endif
2807#ifdef CONFIG_CGROUPS
006f4ac4 2808 ONE("cgroup", S_IRUGO, proc_cgroup_show),
28a6d671 2809#endif
6ba51e37 2810 ONE("oom_score", S_IRUGO, proc_oom_score),
fa0cbbf1 2811 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
a63d83f4 2812 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 2813#ifdef CONFIG_AUDITSYSCALL
631f9c18
AD		 2814 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2815 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 2816#endif
f4f154fd 2817#ifdef CONFIG_FAULT_INJECTION
631f9c18 2818 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 2819#endif
698ba7b5 2820#ifdef CONFIG_ELF_CORE
631f9c18 2821 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3cb4a0bb 2822#endif
aba76fdb 2823#ifdef CONFIG_TASK_IO_ACCOUNTING
19aadc98 2824 ONE("io", S_IRUSR, proc_tgid_io_accounting),
aba76fdb 2825#endif
f133ecca 2826#ifdef CONFIG_HARDWALL
d962c144 2827 ONE("hardwall", S_IRUGO, proc_pid_hardwall),
f133ecca 2828#endif
22d917d8
EB		 2829#ifdef CONFIG_USER_NS
2830 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
2831 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
f76d207a 2832 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
9cc46516 2833 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
22d917d8 2834#endif
48f6a7a5
PE		 2835#ifdef CONFIG_CHECKPOINT_RESTORE
2836 REG("timers", S_IRUGO, proc_timers_operations),
2837#endif
28a6d671 2838};
1da177e4 2839
f0c3b509 2840static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
1da177e4 2841{
f0c3b509
AV		 2842 return proc_pident_readdir(file, ctx,
2843 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 2844}
2845
00977a59 2846static const struct file_operations proc_tgid_base_operations = {
1da177e4 2847 .read = generic_read_dir,
f0c3b509 2848 .iterate = proc_tgid_base_readdir,
6038f373 2849 .llseek = default_llseek,
1da177e4
LT		 2850};
2851
00cd8dd3
AV		 2852static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2853{
7bcd6b0e
EB		 2854 return proc_pident_lookup(dir, dentry,
2855 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 2856}
2857
c5ef1c42 2858static const struct inode_operations proc_tgid_base_inode_operations = {
28a6d671 2859 .lookup = proc_tgid_base_lookup,
99f89551 2860 .getattr = pid_getattr,
6d76fa58 2861 .setattr = proc_setattr,
0499680a 2862 .permission = proc_pid_permission,
1da177e4 2863};
1da177e4 2864
60347f67 2865static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
1da177e4 2866{
48e6484d 2867 struct dentry *dentry, *leader, *dir;
8578cea7 2868 char buf[PROC_NUMBUF];
48e6484d
EB		 2869 struct qstr name;
2870
2871 name.name = buf;
60347f67 2872 name.len = snprintf(buf, sizeof(buf), "%d", pid);
4f522a24 2873 /* no ->d_hash() rejects on procfs */
60347f67 2874 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d 2875 if (dentry) {
bbd51924 2876 d_invalidate(dentry);
48e6484d
EB		 2877 dput(dentry);
2878 }
1da177e4 2879
c35a7f18
ON		 2880 if (pid == tgid)
2881 return;
2882
48e6484d 2883 name.name = buf;
60347f67
PE		 2884 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2885 leader = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d
EB		 2886 if (!leader)
2887 goto out;
1da177e4 2888
48e6484d
EB		 2889 name.name = "task";
2890 name.len = strlen(name.name);
2891 dir = d_hash_and_lookup(leader, &name);
2892 if (!dir)
2893 goto out_put_leader;
2894
2895 name.name = buf;
60347f67 2896 name.len = snprintf(buf, sizeof(buf), "%d", pid);
48e6484d
EB		 2897 dentry = d_hash_and_lookup(dir, &name);
2898 if (dentry) {
bbd51924 2899 d_invalidate(dentry);
48e6484d 2900 dput(dentry);
1da177e4 2901 }
48e6484d
EB		 2902
2903 dput(dir);
2904out_put_leader:
2905 dput(leader);
2906out:
2907 return;
1da177e4
LT		 2908}
2909
0895e91d
RD		 2910/**
2911 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2912 * @task: task that should be flushed.
2913 *
2914 * When flushing dentries from proc, one needs to flush them from global
60347f67 2915 * proc (proc_mnt) and from all the namespaces' procs this task was seen
0895e91d
RD		 2916 * in. This call is supposed to do all of this job.
2917 *
2918 * Looks in the dcache for
2919 * /proc/@pid
2920 * /proc/@tgid/task/@pid
2921 * if either directory is present flushes it and all of it'ts children
2922 * from the dcache.
2923 *
2924 * It is safe and reasonable to cache /proc entries for a task until
2925 * that task exits. After that they just clog up the dcache with
2926 * useless entries, possibly causing useful dcache entries to be
2927 * flushed instead. This routine is proved to flush those useless
2928 * dcache entries at process exit time.
2929 *
2930 * NOTE: This routine is just an optimization so it does not guarantee
2931 * that no dcache entries will exist at process exit time it
2932 * just makes it very unlikely that any will persist.
60347f67
PE		 2933 */
2934
2935void proc_flush_task(struct task_struct *task)
2936{
9fcc2d15 2937 int i;
9b4d1cbe 2938 struct pid *pid, *tgid;
130f77ec
PE		 2939 struct upid *upid;
2940
130f77ec 2941 pid = task_pid(task);
9b4d1cbe 2942 tgid = task_tgid(task);
130f77ec 2943
9fcc2d15 2944 for (i = 0; i <= pid->level; i++) {
130f77ec
PE		 2945 upid = &pid->numbers[i];
2946 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
9b4d1cbe 2947 tgid->numbers[i].nr);
130f77ec 2948 }
60347f67
PE		 2949}
2950
c52a47ac
AV		 2951static int proc_pid_instantiate(struct inode *dir,
2952 struct dentry * dentry,
2953 struct task_struct *task, const void *ptr)
444ceed8 2954{
444ceed8
EB		 2955 struct inode *inode;
2956
61a28784 2957 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 2958 if (!inode)
2959 goto out;
2960
2961 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2962 inode->i_op = &proc_tgid_base_inode_operations;
2963 inode->i_fop = &proc_tgid_base_operations;
2964 inode->i_flags|=S_IMMUTABLE;
aed54175 2965
bfe86848
MS		 2966 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
2967 ARRAY_SIZE(tgid_base_stuff)));
444ceed8 2968
fb045adb 2969 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 2970
2971 d_add(dentry, inode);
2972 /* Close the race of the process dying before we return the dentry */
0b728e19 2973 if (pid_revalidate(dentry, 0))
c52a47ac 2974 return 0;
444ceed8 2975out:
c52a47ac 2976 return -ENOENT;
444ceed8
EB		 2977}
2978
00cd8dd3 2979struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
1da177e4 2980{
335eb531 2981 int result = -ENOENT;
1da177e4 2982 struct task_struct *task;
1da177e4 2983 unsigned tgid;
b488893a 2984 struct pid_namespace *ns;
1da177e4 2985
dbcdb504 2986 tgid = name_to_int(&dentry->d_name);
1da177e4
LT		 2987 if (tgid == ~0U)
2988 goto out;
2989
b488893a 2990 ns = dentry->d_sb->s_fs_info;
de758734 2991 rcu_read_lock();
b488893a 2992 task = find_task_by_pid_ns(tgid, ns);
1da177e4
LT		 2993 if (task)
2994 get_task_struct(task);
de758734 2995 rcu_read_unlock();
1da177e4
LT		 2996 if (!task)
2997 goto out;
2998
444ceed8 2999 result = proc_pid_instantiate(dir, dentry, task, NULL);
1da177e4 3000 put_task_struct(task);
1da177e4 3001out:
c52a47ac 3002 return ERR_PTR(result);
1da177e4
LT		 3003}
3004
1da177e4 3005/*
0804ef4b 3006 * Find the first task with tgid >= tgid
0bc58a91 3007 *
1da177e4 3008 */
19fd4bb2
EB		 3009struct tgid_iter {
3010 unsigned int tgid;
0804ef4b 3011 struct task_struct *task;
19fd4bb2
EB		 3012};
3013static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3014{
0804ef4b 3015 struct pid *pid;
1da177e4 3016
19fd4bb2
EB		 3017 if (iter.task)
3018 put_task_struct(iter.task);
454cc105 3019 rcu_read_lock();
0804ef4b 3020retry:
19fd4bb2
EB		 3021 iter.task = NULL;
3022 pid = find_ge_pid(iter.tgid, ns);
0804ef4b 3023 if (pid) {
19fd4bb2
EB		 3024 iter.tgid = pid_nr_ns(pid, ns);
3025 iter.task = pid_task(pid, PIDTYPE_PID);
0804ef4b
EB		 3026 /* What we to know is if the pid we have find is the
3027 * pid of a thread_group_leader. Testing for task
3028 * being a thread_group_leader is the obvious thing
3029 * todo but there is a window when it fails, due to
3030 * the pid transfer logic in de_thread.
3031 *
3032 * So we perform the straight forward test of seeing
3033 * if the pid we have found is the pid of a thread
3034 * group leader, and don't worry if the task we have
3035 * found doesn't happen to be a thread group leader.
3036 * As we don't care in the case of readdir.
3037 */
19fd4bb2
EB		 3038 if (!iter.task || !has_group_leader_pid(iter.task)) {
3039 iter.tgid += 1;
0804ef4b 3040 goto retry;
19fd4bb2
EB		 3041 }
3042 get_task_struct(iter.task);
0bc58a91 3043 }
454cc105 3044 rcu_read_unlock();
19fd4bb2 3045 return iter;
1da177e4
LT		 3046}
3047
0097875b 3048#define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
0804ef4b 3049
1da177e4 3050/* for the /proc/ directory itself, after non-process stuff has been done */
f0c3b509 3051int proc_pid_readdir(struct file *file, struct dir_context *ctx)
1da177e4 3052{
19fd4bb2 3053 struct tgid_iter iter;
3aa3377f 3054 struct pid_namespace *ns = file_inode(file)->i_sb->s_fs_info;
f0c3b509 3055 loff_t pos = ctx->pos;
1da177e4 3056
021ada7d 3057 if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
f0c3b509 3058 return 0;
1da177e4 3059
0097875b 3060 if (pos == TGID_OFFSET - 2) {
2b0143b5 3061 struct inode *inode = d_inode(ns->proc_self);
db963164 3062 if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
f0c3b509 3063 return 0;
0097875b
EB		 3064 ctx->pos = pos = pos + 1;
3065 }
3066 if (pos == TGID_OFFSET - 1) {
2b0143b5 3067 struct inode *inode = d_inode(ns->proc_thread_self);
0097875b
EB		 3068 if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3069 return 0;
3070 ctx->pos = pos = pos + 1;
021ada7d 3071 }
0097875b 3072 iter.tgid = pos - TGID_OFFSET;
19fd4bb2 3073 iter.task = NULL;
19fd4bb2
EB		 3074 for (iter = next_tgid(ns, iter);
3075 iter.task;
3076 iter.tgid += 1, iter = next_tgid(ns, iter)) {
f0c3b509
AV		 3077 char name[PROC_NUMBUF];
3078 int len;
3079 if (!has_pid_permissions(ns, iter.task, 2))
3080 continue;
0499680a 3081
f0c3b509
AV		 3082 len = snprintf(name, sizeof(name), "%d", iter.tgid);
3083 ctx->pos = iter.tgid + TGID_OFFSET;
3084 if (!proc_fill_cache(file, ctx, name, len,
3085 proc_pid_instantiate, iter.task, NULL)) {
19fd4bb2 3086 put_task_struct(iter.task);
f0c3b509 3087 return 0;
1da177e4 3088 }
0bc58a91 3089 }
f0c3b509 3090 ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
0bc58a91
EB		 3091 return 0;
3092}
1da177e4 3093
28a6d671
EB		 3094/*
3095 * Tasks
3096 */
c5141e6d 3097static const struct pid_entry tid_base_stuff[] = {
631f9c18 3098 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3835541d 3099 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 3100 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
6ba8ed79
EB		 3101#ifdef CONFIG_NET
3102 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3103#endif
631f9c18 3104 REG("environ", S_IRUSR, proc_environ_operations),
f9ea536e 3105 ONE("auxv", S_IRUSR, proc_pid_auxv),
631f9c18 3106 ONE("status", S_IRUGO, proc_pid_status),
35a35046 3107 ONE("personality", S_IRUSR, proc_pid_personality),
1c963eb1 3108 ONE("limits", S_IRUGO, proc_pid_limits),
43ae34cb 3109#ifdef CONFIG_SCHED_DEBUG
631f9c18 3110 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
ebcb6734 3111#endif
4614a696 3112 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 3113#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
09d93bd6 3114 ONE("syscall", S_IRUSR, proc_pid_syscall),
43ae34cb 3115#endif
c2c0bb44 3116 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
631f9c18
AD		 3117 ONE("stat", S_IRUGO, proc_tid_stat),
3118 ONE("statm", S_IRUGO, proc_pid_statm),
b7643757 3119 REG("maps", S_IRUGO, proc_tid_maps_operations),
2e13ba54 3120#ifdef CONFIG_PROC_CHILDREN
81841161
CG		 3121 REG("children", S_IRUGO, proc_tid_children_operations),
3122#endif
28a6d671 3123#ifdef CONFIG_NUMA
b7643757 3124 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
28a6d671 3125#endif
631f9c18
AD		 3126 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3127 LNK("cwd", proc_cwd_link),
3128 LNK("root", proc_root_link),
3129 LNK("exe", proc_exe_link),
3130 REG("mounts", S_IRUGO, proc_mounts_operations),
3131 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
1e883281 3132#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18 3133 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
b7643757 3134 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
32ed74a4 3135 REG("pagemap", S_IRUSR, proc_pagemap_operations),
28a6d671
EB		 3136#endif
3137#ifdef CONFIG_SECURITY
631f9c18 3138 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 3139#endif
3140#ifdef CONFIG_KALLSYMS
edfcd606 3141 ONE("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 3142#endif
2ec220e2 3143#ifdef CONFIG_STACKTRACE
35a35046 3144 ONE("stack", S_IRUSR, proc_pid_stack),
28a6d671 3145#endif
5968cece 3146#ifdef CONFIG_SCHED_INFO
f6e826ca 3147 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 3148#endif
9745512c 3149#ifdef CONFIG_LATENCYTOP
631f9c18 3150 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 3151#endif
8793d854 3152#ifdef CONFIG_PROC_PID_CPUSET
52de4779 3153 ONE("cpuset", S_IRUGO, proc_cpuset_show),
a424316c
PM		 3154#endif
3155#ifdef CONFIG_CGROUPS
006f4ac4 3156 ONE("cgroup", S_IRUGO, proc_cgroup_show),
28a6d671 3157#endif
6ba51e37 3158 ONE("oom_score", S_IRUGO, proc_oom_score),
fa0cbbf1 3159 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
a63d83f4 3160 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 3161#ifdef CONFIG_AUDITSYSCALL
631f9c18 3162 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
26ec3c64 3163 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 3164#endif
f4f154fd 3165#ifdef CONFIG_FAULT_INJECTION
631f9c18 3166 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 3167#endif
297c5d92 3168#ifdef CONFIG_TASK_IO_ACCOUNTING
19aadc98 3169 ONE("io", S_IRUSR, proc_tid_io_accounting),
297c5d92 3170#endif
f133ecca 3171#ifdef CONFIG_HARDWALL
d962c144 3172 ONE("hardwall", S_IRUGO, proc_pid_hardwall),
f133ecca 3173#endif
22d917d8
EB		 3174#ifdef CONFIG_USER_NS
3175 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3176 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
f76d207a 3177 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
9cc46516 3178 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
22d917d8 3179#endif
28a6d671
EB		 3180};
3181
f0c3b509 3182static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
28a6d671 3183{
f0c3b509
AV		 3184 return proc_pident_readdir(file, ctx,
3185 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
28a6d671
EB		 3186}
3187
00cd8dd3
AV		 3188static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3189{
7bcd6b0e
EB		 3190 return proc_pident_lookup(dir, dentry,
3191 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
28a6d671
EB		 3192}
3193
00977a59 3194static const struct file_operations proc_tid_base_operations = {
28a6d671 3195 .read = generic_read_dir,
f0c3b509 3196 .iterate = proc_tid_base_readdir,
6038f373 3197 .llseek = default_llseek,
28a6d671
EB		 3198};
3199
c5ef1c42 3200static const struct inode_operations proc_tid_base_inode_operations = {
28a6d671
EB		 3201 .lookup = proc_tid_base_lookup,
3202 .getattr = pid_getattr,
3203 .setattr = proc_setattr,
3204};
3205
c52a47ac 3206static int proc_task_instantiate(struct inode *dir,
c5141e6d 3207 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8 3208{
444ceed8 3209 struct inode *inode;
61a28784 3210 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 3211
3212 if (!inode)
3213 goto out;
3214 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3215 inode->i_op = &proc_tid_base_inode_operations;
3216 inode->i_fop = &proc_tid_base_operations;
3217 inode->i_flags|=S_IMMUTABLE;
aed54175 3218
bfe86848
MS		 3219 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3220 ARRAY_SIZE(tid_base_stuff)));
444ceed8 3221
fb045adb 3222 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 3223
3224 d_add(dentry, inode);
3225 /* Close the race of the process dying before we return the dentry */
0b728e19 3226 if (pid_revalidate(dentry, 0))
c52a47ac 3227 return 0;
444ceed8 3228out:
c52a47ac 3229 return -ENOENT;
444ceed8
EB		 3230}
3231
00cd8dd3 3232static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
28a6d671 3233{
c52a47ac 3234 int result = -ENOENT;
28a6d671
EB		 3235 struct task_struct *task;
3236 struct task_struct *leader = get_proc_task(dir);
28a6d671 3237 unsigned tid;
b488893a 3238 struct pid_namespace *ns;
28a6d671
EB		 3239
3240 if (!leader)
3241 goto out_no_task;
3242
dbcdb504 3243 tid = name_to_int(&dentry->d_name);
28a6d671
EB		 3244 if (tid == ~0U)
3245 goto out;
3246
b488893a 3247 ns = dentry->d_sb->s_fs_info;
28a6d671 3248 rcu_read_lock();
b488893a 3249 task = find_task_by_pid_ns(tid, ns);
28a6d671
EB		 3250 if (task)
3251 get_task_struct(task);
3252 rcu_read_unlock();
3253 if (!task)
3254 goto out;
bac0abd6 3255 if (!same_thread_group(leader, task))
28a6d671
EB		 3256 goto out_drop_task;
3257
444ceed8 3258 result = proc_task_instantiate(dir, dentry, task, NULL);
28a6d671
EB		 3259out_drop_task:
3260 put_task_struct(task);
3261out:
3262 put_task_struct(leader);
3263out_no_task:
c52a47ac 3264 return ERR_PTR(result);
28a6d671
EB		 3265}
3266
0bc58a91
EB		 3267/*
3268 * Find the first tid of a thread group to return to user space.
3269 *
3270 * Usually this is just the thread group leader, but if the users
3271 * buffer was too small or there was a seek into the middle of the
3272 * directory we have more work todo.
3273 *
3274 * In the case of a short read we start with find_task_by_pid.
3275 *
3276 * In the case of a seek we start with the leader and walk nr
3277 * threads past it.
3278 */
9f6e963f
ON		 3279static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3280 struct pid_namespace *ns)
0bc58a91 3281{
d855a4b7 3282 struct task_struct *pos, *task;
9f6e963f
ON		 3283 unsigned long nr = f_pos;
3284
3285 if (nr != f_pos) /* 32bit overflow? */
3286 return NULL;
1da177e4 3287
cc288738 3288 rcu_read_lock();
d855a4b7
ON		 3289 task = pid_task(pid, PIDTYPE_PID);
3290 if (!task)
3291 goto fail;
3292
3293 /* Attempt to start with the tid of a thread */
9f6e963f 3294 if (tid && nr) {
b488893a 3295 pos = find_task_by_pid_ns(tid, ns);
d855a4b7 3296 if (pos && same_thread_group(pos, task))
a872ff0c 3297 goto found;
0bc58a91 3298 }
1da177e4 3299
0bc58a91 3300 /* If nr exceeds the number of threads there is nothing todo */
9f6e963f 3301 if (nr >= get_nr_threads(task))
c986c14a 3302 goto fail;
1da177e4 3303
a872ff0c
ON		 3304 /* If we haven't found our starting place yet start
3305 * with the leader and walk nr threads forward.
0bc58a91 3306 */
d855a4b7 3307 pos = task = task->group_leader;
c986c14a 3308 do {
9f6e963f 3309 if (!nr--)
c986c14a 3310 goto found;
d855a4b7 3311 } while_each_thread(task, pos);
c986c14a
ON		 3312fail:
3313 pos = NULL;
3314 goto out;
a872ff0c
ON		 3315found:
3316 get_task_struct(pos);
3317out:
cc288738 3318 rcu_read_unlock();
0bc58a91
EB		 3319 return pos;
3320}
3321
3322/*
3323 * Find the next thread in the thread list.
3324 * Return NULL if there is an error or no next thread.
3325 *
3326 * The reference to the input task_struct is released.
3327 */
3328static struct task_struct *next_tid(struct task_struct *start)
3329{
c1df7fb8 3330 struct task_struct *pos = NULL;
cc288738 3331 rcu_read_lock();
c1df7fb8 3332 if (pid_alive(start)) {
0bc58a91 3333 pos = next_thread(start);
c1df7fb8
ON		 3334 if (thread_group_leader(pos))
3335 pos = NULL;
3336 else
3337 get_task_struct(pos);
3338 }
cc288738 3339 rcu_read_unlock();
0bc58a91
EB		 3340 put_task_struct(start);
3341 return pos;
1da177e4
LT		 3342}
3343
3344/* for the /proc/TGID/task/ directories */
f0c3b509 3345static int proc_task_readdir(struct file *file, struct dir_context *ctx)
1da177e4 3346{
d855a4b7
ON		 3347 struct inode *inode = file_inode(file);
3348 struct task_struct *task;
b488893a 3349 struct pid_namespace *ns;
f0c3b509 3350 int tid;
1da177e4 3351
d855a4b7 3352 if (proc_inode_is_dead(inode))
f0c3b509 3353 return -ENOENT;
1da177e4 3354
f0c3b509 3355 if (!dir_emit_dots(file, ctx))
d855a4b7 3356 return 0;
1da177e4 3357
0bc58a91
EB		 3358 /* f_version caches the tgid value that the last readdir call couldn't
3359 * return. lseek aka telldir automagically resets f_version to 0.
3360 */
3aa3377f 3361 ns = inode->i_sb->s_fs_info;
f0c3b509
AV		 3362 tid = (int)file->f_version;
3363 file->f_version = 0;
d855a4b7 3364 for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
0bc58a91 3365 task;
f0c3b509
AV		 3366 task = next_tid(task), ctx->pos++) {
3367 char name[PROC_NUMBUF];
3368 int len;
b488893a 3369 tid = task_pid_nr_ns(task, ns);
f0c3b509
AV		 3370 len = snprintf(name, sizeof(name), "%d", tid);
3371 if (!proc_fill_cache(file, ctx, name, len,
3372 proc_task_instantiate, task, NULL)) {
0bc58a91
EB		 3373 /* returning this tgid failed, save it as the first
3374 * pid for the next readir call */
f0c3b509 3375 file->f_version = (u64)tid;
0bc58a91 3376 put_task_struct(task);
1da177e4 3377 break;
0bc58a91 3378 }
1da177e4 3379 }
d855a4b7 3380
f0c3b509 3381 return 0;
1da177e4 3382}
6e66b52b
EB		 3383
3384static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3385{
2b0143b5 3386 struct inode *inode = d_inode(dentry);
99f89551 3387 struct task_struct *p = get_proc_task(inode);
6e66b52b
EB		 3388 generic_fillattr(inode, stat);
3389
99f89551 3390 if (p) {
99f89551 3391 stat->nlink += get_nr_threads(p);
99f89551 3392 put_task_struct(p);
6e66b52b
EB		 3393 }
3394
3395 return 0;
3396}
28a6d671 3397
c5ef1c42 3398static const struct inode_operations proc_task_inode_operations = {
28a6d671
EB		 3399 .lookup = proc_task_lookup,
3400 .getattr = proc_task_getattr,
3401 .setattr = proc_setattr,
0499680a 3402 .permission = proc_pid_permission,
28a6d671
EB		 3403};
3404
00977a59 3405static const struct file_operations proc_task_operations = {
28a6d671 3406 .read = generic_read_dir,
f0c3b509 3407 .iterate = proc_task_readdir,
6038f373 3408 .llseek = default_llseek,
28a6d671 3409};
Linux 6.x block layer and io_uring tree(s)