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