projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge branch 'nowait-aio-btrfs-fixup' of git://git.kernel.org/pub/scm/linux/kernel...
[linux-2.6-block.git] / fs / proc / base.c
CommitLineData
1da177e4
LT		 1/*
2 * linux/fs/proc/base.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 *
6 * proc base directory handling functions
7 *
8 * 1999, Al Viro. Rewritten. Now it covers the whole per-process part.
9 * Instead of using magical inumbers to determine the kind of object
10 * we allocate and fill in-core inodes upon lookup. They don't even
11 * go into icache. We cache the reference to task_struct upon lookup too.
12 * Eventually it should become a filesystem in its own. We don't use the
13 * rest of procfs anymore.
e070ad49
ML		 14 *
15 *
16 * Changelog:
17 * 17-Jan-2005
18 * Allan Bezerra
19 * Bruna Moreira <bruna.moreira@indt.org.br>
20 * Edjard Mota <edjard.mota@indt.org.br>
21 * Ilias Biris <ilias.biris@indt.org.br>
22 * Mauricio Lin <mauricio.lin@indt.org.br>
23 *
24 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
25 *
26 * A new process specific entry (smaps) included in /proc. It shows the
27 * size of rss for each memory area. The maps entry lacks information
28 * about physical memory size (rss) for each mapped file, i.e.,
29 * rss information for executables and library files.
30 * This additional information is useful for any tools that need to know
31 * about physical memory consumption for a process specific library.
32 *
33 * Changelog:
34 * 21-Feb-2005
35 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
36 * Pud inclusion in the page table walking.
37 *
38 * ChangeLog:
39 * 10-Mar-2005
40 * 10LE Instituto Nokia de Tecnologia - INdT:
41 * A better way to walks through the page table as suggested by Hugh Dickins.
42 *
43 * Simo Piiroinen <simo.piiroinen@nokia.com>:
44 * Smaps information related to shared, private, clean and dirty pages.
45 *
46 * Paul Mundt <paul.mundt@nokia.com>:
47 * Overall revision about smaps.
1da177e4
LT		 48 */
49
7c0f6ba6 50#include <linux/uaccess.h>
1da177e4 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>
4eb5aaa3 88#include <linux/sched/autogroup.h>
6e84f315 89#include <linux/sched/mm.h>
f7ccbae4 90#include <linux/sched/coredump.h>
b17b0153 91#include <linux/sched/debug.h>
3905f9ad 92#include <linux/sched/stat.h>
640708a2 93#include <linux/flex_array.h>
48f6a7a5 94#include <linux/posix-timers.h>
f133ecca
CM		 95#ifdef CONFIG_HARDWALL
96#include <asm/hardwall.h>
97#endif
43d2b113 98#include <trace/events/oom.h>
1da177e4 99#include "internal.h"
faf60af1 100#include "fd.h"
1da177e4 101
0f2fe20f
EB		 102/* NOTE:
103 * Implementing inode permission operations in /proc is almost
104 * certainly an error. Permission checks need to happen during
105 * each system call not at open time. The reason is that most of
106 * what we wish to check for permissions in /proc varies at runtime.
107 *
108 * The classic example of a problem is opening file descriptors
109 * in /proc for a task before it execs a suid executable.
110 */
111
1270dd8d
AD		 112static u8 nlink_tid;
113static u8 nlink_tgid;
114
1da177e4 115struct pid_entry {
cedbccab 116 const char *name;
623f594e 117 unsigned int len;
d161a13f 118 umode_t mode;
c5ef1c42 119 const struct inode_operations *iop;
00977a59 120 const struct file_operations *fop;
20cdc894 121 union proc_op op;
1da177e4
LT		 122};
123
61a28784 124#define NOD(NAME, MODE, IOP, FOP, OP) { \
20cdc894 125 .name = (NAME), \
c5141e6d 126 .len = sizeof(NAME) - 1, \
20cdc894
EB		 127 .mode = MODE, \
128 .iop = IOP, \
129 .fop = FOP, \
130 .op = OP, \
131}
132
631f9c18
AD		 133#define DIR(NAME, MODE, iops, fops) \
134 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
135#define LNK(NAME, get_link) \
61a28784 136 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
20cdc894 137 &proc_pid_link_inode_operations, NULL, \
631f9c18
AD		 138 { .proc_get_link = get_link } )
139#define REG(NAME, MODE, fops) \
140 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
631f9c18 141#define ONE(NAME, MODE, show) \
be614086
EB		 142 NOD(NAME, (S_IFREG|(MODE)), \
143 NULL, &proc_single_file_operations, \
631f9c18 144 { .proc_show = show } )
1da177e4 145
aed54175
VN		 146/*
147 * Count the number of hardlinks for the pid_entry table, excluding the .
148 * and .. links.
149 */
1270dd8d 150static unsigned int __init pid_entry_nlink(const struct pid_entry *entries,
aed54175
VN		 151 unsigned int n)
152{
153 unsigned int i;
154 unsigned int count;
155
1270dd8d 156 count = 2;
aed54175
VN		 157 for (i = 0; i < n; ++i) {
158 if (S_ISDIR(entries[i].mode))
159 ++count;
160 }
161
162 return count;
163}
164
f7ad3c6b 165static int get_task_root(struct task_struct *task, struct path *root)
1da177e4 166{
7c2c7d99
HD		 167 int result = -ENOENT;
168
0494f6ec 169 task_lock(task);
f7ad3c6b
MS		 170 if (task->fs) {
171 get_fs_root(task->fs, root);
7c2c7d99
HD		 172 result = 0;
173 }
0494f6ec 174 task_unlock(task);
7c2c7d99 175 return result;
0494f6ec
MS		 176}
177
7773fbc5 178static int proc_cwd_link(struct dentry *dentry, struct path *path)
0494f6ec 179{
2b0143b5 180 struct task_struct *task = get_proc_task(d_inode(dentry));
0494f6ec 181 int result = -ENOENT;
99f89551
EB		 182
183 if (task) {
f7ad3c6b
MS		 184 task_lock(task);
185 if (task->fs) {
186 get_fs_pwd(task->fs, path);
187 result = 0;
188 }
189 task_unlock(task);
99f89551
EB		 190 put_task_struct(task);
191 }
1da177e4
LT		 192 return result;
193}
194
7773fbc5 195static int proc_root_link(struct dentry *dentry, struct path *path)
1da177e4 196{
2b0143b5 197 struct task_struct *task = get_proc_task(d_inode(dentry));
1da177e4 198 int result = -ENOENT;
99f89551
EB		 199
200 if (task) {
f7ad3c6b 201 result = get_task_root(task, path);
99f89551
EB		 202 put_task_struct(task);
203 }
1da177e4
LT		 204 return result;
205}
206
c2c0bb44
AD		 207static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,
208 size_t _count, loff_t *pos)
1da177e4 209{
c2c0bb44
AD		 210 struct task_struct *tsk;
211 struct mm_struct *mm;
212 char *page;
213 unsigned long count = _count;
214 unsigned long arg_start, arg_end, env_start, env_end;
215 unsigned long len1, len2, len;
216 unsigned long p;
217 char c;
218 ssize_t rv;
219
220 BUG_ON(*pos < 0);
221
222 tsk = get_proc_task(file_inode(file));
223 if (!tsk)
224 return -ESRCH;
225 mm = get_task_mm(tsk);
226 put_task_struct(tsk);
227 if (!mm)
228 return 0;
229 /* Check if process spawned far enough to have cmdline. */
230 if (!mm->env_end) {
231 rv = 0;
232 goto out_mmput;
233 }
234
235 page = (char *)__get_free_page(GFP_TEMPORARY);
236 if (!page) {
237 rv = -ENOMEM;
238 goto out_mmput;
239 }
240
241 down_read(&mm->mmap_sem);
242 arg_start = mm->arg_start;
243 arg_end = mm->arg_end;
244 env_start = mm->env_start;
245 env_end = mm->env_end;
246 up_read(&mm->mmap_sem);
247
248 BUG_ON(arg_start > arg_end);
249 BUG_ON(env_start > env_end);
250
251 len1 = arg_end - arg_start;
252 len2 = env_end - env_start;
253
3581d458
AD		 254 /* Empty ARGV. */
255 if (len1 == 0) {
256 rv = 0;
257 goto out_free_page;
258 }
2ca66ff7 259 /*
c2c0bb44
AD		 260 * Inherently racy -- command line shares address space
261 * with code and data.
2ca66ff7 262 */
272ddc8b 263 rv = access_remote_vm(mm, arg_end - 1, &c, 1, 0);
c2c0bb44
AD		 264 if (rv <= 0)
265 goto out_free_page;
266
267 rv = 0;
268
269 if (c == '\0') {
270 /* Command line (set of strings) occupies whole ARGV. */
271 if (len1 <= *pos)
272 goto out_free_page;
273
274 p = arg_start + *pos;
275 len = len1 - *pos;
276 while (count > 0 && len > 0) {
277 unsigned int _count;
278 int nr_read;
279
280 _count = min3(count, len, PAGE_SIZE);
272ddc8b 281 nr_read = access_remote_vm(mm, p, page, _count, 0);
c2c0bb44
AD		 282 if (nr_read < 0)
283 rv = nr_read;
284 if (nr_read <= 0)
285 goto out_free_page;
286
287 if (copy_to_user(buf, page, nr_read)) {
288 rv = -EFAULT;
289 goto out_free_page;
290 }
291
292 p += nr_read;
293 len -= nr_read;
294 buf += nr_read;
295 count -= nr_read;
296 rv += nr_read;
297 }
298 } else {
c2c0bb44
AD		 299 /*
300 * Command line (1 string) occupies ARGV and
301 * extends into ENVP.
302 */
a0a07b87
AD		 303 struct {
304 unsigned long p;
305 unsigned long len;
306 } cmdline[2] = {
307 { .p = arg_start, .len = len1 },
308 { .p = env_start, .len = len2 },
309 };
310 loff_t pos1 = *pos;
311 unsigned int i;
312
313 i = 0;
314 while (i < 2 && pos1 >= cmdline[i].len) {
315 pos1 -= cmdline[i].len;
316 i++;
c2c0bb44 317 }
a0a07b87
AD		 318 while (i < 2) {
319 p = cmdline[i].p + pos1;
320 len = cmdline[i].len - pos1;
321 while (count > 0 && len > 0) {
322 unsigned int _count, l;
323 int nr_read;
324 bool final;
325
326 _count = min3(count, len, PAGE_SIZE);
327 nr_read = access_remote_vm(mm, p, page, _count, 0);
328 if (nr_read < 0)
329 rv = nr_read;
330 if (nr_read <= 0)
331 goto out_free_page;
332
333 /*
334 * Command line can be shorter than whole ARGV
335 * even if last "marker" byte says it is not.
336 */
337 final = false;
338 l = strnlen(page, nr_read);
339 if (l < nr_read) {
340 nr_read = l;
341 final = true;
342 }
343
344 if (copy_to_user(buf, page, nr_read)) {
345 rv = -EFAULT;
346 goto out_free_page;
347 }
348
349 p += nr_read;
350 len -= nr_read;
351 buf += nr_read;
352 count -= nr_read;
353 rv += nr_read;
354
355 if (final)
356 goto out_free_page;
c2c0bb44
AD		 357 }
358
a0a07b87
AD		 359 /* Only first chunk can be read partially. */
360 pos1 = 0;
361 i++;
c2c0bb44 362 }
c2c0bb44
AD		 363 }
364
365out_free_page:
366 free_page((unsigned long)page);
367out_mmput:
368 mmput(mm);
369 if (rv > 0)
370 *pos += rv;
371 return rv;
1da177e4
LT		 372}
373
c2c0bb44
AD		 374static const struct file_operations proc_pid_cmdline_ops = {
375 .read = proc_pid_cmdline_read,
376 .llseek = generic_file_llseek,
377};
378
1da177e4
LT		 379#ifdef CONFIG_KALLSYMS
380/*
381 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
382 * Returns the resolved symbol. If that fails, simply return the address.
383 */
edfcd606
AD		 384static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns,
385 struct pid *pid, struct task_struct *task)
1da177e4 386{
ffb45122 387 unsigned long wchan;
9281acea 388 char symname[KSYM_NAME_LEN];
1da177e4
LT		 389
390 wchan = get_wchan(task);
391
caaee623
JH		 392 if (wchan && ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)
393 && !lookup_symbol_name(wchan, symname))
25ce3191 394 seq_printf(m, "%s", symname);
b2f73922 395 else
1e92a61c 396 seq_putc(m, '0');
25ce3191
JP		 397
398 return 0;
1da177e4
LT		 399}
400#endif /* CONFIG_KALLSYMS */
401
a9712bc1
AV		 402static int lock_trace(struct task_struct *task)
403{
404 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
405 if (err)
406 return err;
caaee623 407 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) {
a9712bc1
AV		 408 mutex_unlock(&task->signal->cred_guard_mutex);
409 return -EPERM;
410 }
411 return 0;
412}
413
414static void unlock_trace(struct task_struct *task)
415{
416 mutex_unlock(&task->signal->cred_guard_mutex);
417}
418
2ec220e2
KC		 419#ifdef CONFIG_STACKTRACE
420
421#define MAX_STACK_TRACE_DEPTH 64
422
423static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
424 struct pid *pid, struct task_struct *task)
425{
426 struct stack_trace trace;
427 unsigned long *entries;
a9712bc1 428 int err;
2ec220e2
KC		 429 int i;
430
431 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
432 if (!entries)
433 return -ENOMEM;
434
435 trace.nr_entries = 0;
436 trace.max_entries = MAX_STACK_TRACE_DEPTH;
437 trace.entries = entries;
438 trace.skip = 0;
2ec220e2 439
a9712bc1
AV		 440 err = lock_trace(task);
441 if (!err) {
442 save_stack_trace_tsk(task, &trace);
443
444 for (i = 0; i < trace.nr_entries; i++) {
8b927d73 445 seq_printf(m, "[<%pK>] %pB\n",
a9712bc1
AV		 446 (void *)entries[i], (void *)entries[i]);
447 }
448 unlock_trace(task);
2ec220e2
KC		 449 }
450 kfree(entries);
451
a9712bc1 452 return err;
2ec220e2
KC		 453}
454#endif
455
5968cece 456#ifdef CONFIG_SCHED_INFO
1da177e4
LT		 457/*
458 * Provides /proc/PID/schedstat
459 */
f6e826ca
AD		 460static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
461 struct pid *pid, struct task_struct *task)
1da177e4 462{
5968cece
NR		 463 if (unlikely(!sched_info_on()))
464 seq_printf(m, "0 0 0\n");
465 else
466 seq_printf(m, "%llu %llu %lu\n",
25ce3191
JP		 467 (unsigned long long)task->se.sum_exec_runtime,
468 (unsigned long long)task->sched_info.run_delay,
469 task->sched_info.pcount);
470
471 return 0;
1da177e4
LT		 472}
473#endif
474
9745512c
AV		 475#ifdef CONFIG_LATENCYTOP
476static int lstats_show_proc(struct seq_file *m, void *v)
477{
478 int i;
13d77c37
HS		 479 struct inode *inode = m->private;
480 struct task_struct *task = get_proc_task(inode);
9745512c 481
13d77c37
HS		 482 if (!task)
483 return -ESRCH;
484 seq_puts(m, "Latency Top version : v0.1\n");
9745512c 485 for (i = 0; i < 32; i++) {
34e49d4f
JP		 486 struct latency_record *lr = &task->latency_record[i];
487 if (lr->backtrace[0]) {
9745512c 488 int q;
34e49d4f
JP		 489 seq_printf(m, "%i %li %li",
490 lr->count, lr->time, lr->max);
9745512c 491 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
34e49d4f
JP		 492 unsigned long bt = lr->backtrace[q];
493 if (!bt)
9745512c 494 break;
34e49d4f 495 if (bt == ULONG_MAX)
9745512c 496 break;
34e49d4f 497 seq_printf(m, " %ps", (void *)bt);
9745512c 498 }
9d6de12f 499 seq_putc(m, '\n');
9745512c
AV		 500 }
501
502 }
13d77c37 503 put_task_struct(task);
9745512c
AV		 504 return 0;
505}
506
507static int lstats_open(struct inode *inode, struct file *file)
508{
13d77c37 509 return single_open(file, lstats_show_proc, inode);
d6643d12
HS		 510}
511
9745512c
AV		 512static ssize_t lstats_write(struct file *file, const char __user *buf,
513 size_t count, loff_t *offs)
514{
496ad9aa 515 struct task_struct *task = get_proc_task(file_inode(file));
9745512c 516
13d77c37
HS		 517 if (!task)
518 return -ESRCH;
9745512c 519 clear_all_latency_tracing(task);
13d77c37 520 put_task_struct(task);
9745512c
AV		 521
522 return count;
523}
524
525static const struct file_operations proc_lstats_operations = {
526 .open = lstats_open,
527 .read = seq_read,
528 .write = lstats_write,
529 .llseek = seq_lseek,
13d77c37 530 .release = single_release,
9745512c
AV		 531};
532
533#endif
534
6ba51e37
AD		 535static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
536 struct pid *pid, struct task_struct *task)
1da177e4 537{
a7f638f9 538 unsigned long totalpages = totalram_pages + total_swap_pages;
b95c35e7 539 unsigned long points = 0;
1da177e4 540
ef419398
ON		 541 points = oom_badness(task, NULL, NULL, totalpages) *
542 1000 / totalpages;
25ce3191
JP		 543 seq_printf(m, "%lu\n", points);
544
545 return 0;
1da177e4
LT		 546}
547
d85f50d5 548struct limit_names {
cedbccab
AD		 549 const char *name;
550 const char *unit;
d85f50d5
NH		 551};
552
553static const struct limit_names lnames[RLIM_NLIMITS] = {
cff4edb5 554 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
d85f50d5
NH		 555 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
556 [RLIMIT_DATA] = {"Max data size", "bytes"},
557 [RLIMIT_STACK] = {"Max stack size", "bytes"},
558 [RLIMIT_CORE] = {"Max core file size", "bytes"},
559 [RLIMIT_RSS] = {"Max resident set", "bytes"},
560 [RLIMIT_NPROC] = {"Max processes", "processes"},
561 [RLIMIT_NOFILE] = {"Max open files", "files"},
562 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
563 [RLIMIT_AS] = {"Max address space", "bytes"},
564 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
565 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
566 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
567 [RLIMIT_NICE] = {"Max nice priority", NULL},
568 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
8808117c 569 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
d85f50d5
NH		 570};
571
572/* Display limits for a process */
1c963eb1
AD		 573static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
574 struct pid *pid, struct task_struct *task)
d85f50d5
NH		 575{
576 unsigned int i;
d85f50d5 577 unsigned long flags;
d85f50d5
NH		 578
579 struct rlimit rlim[RLIM_NLIMITS];
580
a6bebbc8 581 if (!lock_task_sighand(task, &flags))
d85f50d5 582 return 0;
d85f50d5
NH		 583 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
584 unlock_task_sighand(task, &flags);
d85f50d5
NH		 585
586 /*
587 * print the file header
588 */
1c963eb1 589 seq_printf(m, "%-25s %-20s %-20s %-10s\n",
25ce3191 590 "Limit", "Soft Limit", "Hard Limit", "Units");
d85f50d5
NH		 591
592 for (i = 0; i < RLIM_NLIMITS; i++) {
593 if (rlim[i].rlim_cur == RLIM_INFINITY)
1c963eb1 594 seq_printf(m, "%-25s %-20s ",
25ce3191 595 lnames[i].name, "unlimited");
d85f50d5 596 else
1c963eb1 597 seq_printf(m, "%-25s %-20lu ",
25ce3191 598 lnames[i].name, rlim[i].rlim_cur);
d85f50d5
NH		 599
600 if (rlim[i].rlim_max == RLIM_INFINITY)
1c963eb1 601 seq_printf(m, "%-20s ", "unlimited");
d85f50d5 602 else
1c963eb1 603 seq_printf(m, "%-20lu ", rlim[i].rlim_max);
d85f50d5
NH		 604
605 if (lnames[i].unit)
1c963eb1 606 seq_printf(m, "%-10s\n", lnames[i].unit);
d85f50d5 607 else
1c963eb1 608 seq_putc(m, '\n');
d85f50d5
NH		 609 }
610
1c963eb1 611 return 0;
d85f50d5
NH		 612}
613
ebcb6734 614#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
09d93bd6
AD		 615static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
616 struct pid *pid, struct task_struct *task)
ebcb6734
RM		 617{
618 long nr;
619 unsigned long args[6], sp, pc;
25ce3191
JP		 620 int res;
621
622 res = lock_trace(task);
a9712bc1
AV		 623 if (res)
624 return res;
ebcb6734
RM		 625
626 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
09d93bd6 627 seq_puts(m, "running\n");
a9712bc1 628 else if (nr < 0)
09d93bd6 629 seq_printf(m, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
a9712bc1 630 else
09d93bd6 631 seq_printf(m,
ebcb6734
RM		 632 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
633 nr,
634 args[0], args[1], args[2], args[3], args[4], args[5],
635 sp, pc);
a9712bc1 636 unlock_trace(task);
25ce3191
JP		 637
638 return 0;
ebcb6734
RM		 639}
640#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
641
1da177e4
LT		 642/************************************************************************/
643/* Here the fs part begins */
644/************************************************************************/
645
646/* permission checks */
778c1144 647static int proc_fd_access_allowed(struct inode *inode)
1da177e4 648{
778c1144
EB		 649 struct task_struct *task;
650 int allowed = 0;
df26c40e
EB		 651 /* Allow access to a task's file descriptors if it is us or we
652 * may use ptrace attach to the process and find out that
653 * information.
778c1144
EB		 654 */
655 task = get_proc_task(inode);
df26c40e 656 if (task) {
caaee623 657 allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
778c1144 658 put_task_struct(task);
df26c40e 659 }
778c1144 660 return allowed;
1da177e4
LT		 661}
662
6b4e306a 663int proc_setattr(struct dentry *dentry, struct iattr *attr)
6d76fa58
LT		 664{
665 int error;
2b0143b5 666 struct inode *inode = d_inode(dentry);
6d76fa58
LT		 667
668 if (attr->ia_valid & ATTR_MODE)
669 return -EPERM;
670
31051c85 671 error = setattr_prepare(dentry, attr);
1025774c
CH		 672 if (error)
673 return error;
674
1025774c
CH		 675 setattr_copy(inode, attr);
676 mark_inode_dirty(inode);
677 return 0;
6d76fa58
LT		 678}
679
0499680a
VK		 680/*
681 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
682 * or euid/egid (for hide_pid_min=2)?
683 */
684static bool has_pid_permissions(struct pid_namespace *pid,
685 struct task_struct *task,
686 int hide_pid_min)
687{
688 if (pid->hide_pid < hide_pid_min)
689 return true;
690 if (in_group_p(pid->pid_gid))
691 return true;
caaee623 692 return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
0499680a
VK		 693}
694
695
696static int proc_pid_permission(struct inode *inode, int mask)
697{
698 struct pid_namespace *pid = inode->i_sb->s_fs_info;
699 struct task_struct *task;
700 bool has_perms;
701
702 task = get_proc_task(inode);
a2ef990a
XF		 703 if (!task)
704 return -ESRCH;
796f571b 705 has_perms = has_pid_permissions(pid, task, HIDEPID_NO_ACCESS);
0499680a
VK		 706 put_task_struct(task);
707
708 if (!has_perms) {
796f571b 709 if (pid->hide_pid == HIDEPID_INVISIBLE) {
0499680a
VK		 710 /*
711 * Let's make getdents(), stat(), and open()
712 * consistent with each other. If a process
713 * may not stat() a file, it shouldn't be seen
714 * in procfs at all.
715 */
716 return -ENOENT;
717 }
718
719 return -EPERM;
720 }
721 return generic_permission(inode, mask);
722}
723
724
725
c5ef1c42 726static const struct inode_operations proc_def_inode_operations = {
6d76fa58
LT		 727 .setattr = proc_setattr,
728};
729
be614086
EB		 730static int proc_single_show(struct seq_file *m, void *v)
731{
732 struct inode *inode = m->private;
733 struct pid_namespace *ns;
734 struct pid *pid;
735 struct task_struct *task;
736 int ret;
737
738 ns = inode->i_sb->s_fs_info;
739 pid = proc_pid(inode);
740 task = get_pid_task(pid, PIDTYPE_PID);
741 if (!task)
742 return -ESRCH;
743
744 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
745
746 put_task_struct(task);
747 return ret;
748}
749
750static int proc_single_open(struct inode *inode, struct file *filp)
751{
c6a34058 752 return single_open(filp, proc_single_show, inode);
be614086
EB		 753}
754
755static const struct file_operations proc_single_file_operations = {
756 .open = proc_single_open,
757 .read = seq_read,
758 .llseek = seq_lseek,
759 .release = single_release,
760};
761
5381e169
ON		 762
763struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
1da177e4 764{
5381e169
ON		 765 struct task_struct *task = get_proc_task(inode);
766 struct mm_struct *mm = ERR_PTR(-ESRCH);
e268337d 767
5381e169 768 if (task) {
caaee623 769 mm = mm_access(task, mode | PTRACE_MODE_FSCREDS);
5381e169 770 put_task_struct(task);
e268337d 771
5381e169
ON		 772 if (!IS_ERR_OR_NULL(mm)) {
773 /* ensure this mm_struct can't be freed */
f1f10076 774 mmgrab(mm);
5381e169
ON		 775 /* but do not pin its memory */
776 mmput(mm);
777 }
778 }
779
780 return mm;
781}
782
783static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
784{
785 struct mm_struct *mm = proc_mem_open(inode, mode);
e268337d
LT		 786
787 if (IS_ERR(mm))
788 return PTR_ERR(mm);
789
e268337d 790 file->private_data = mm;
1da177e4
LT		 791 return 0;
792}
793
b409e578
CW		 794static int mem_open(struct inode *inode, struct file *file)
795{
bc452b4b
DH		 796 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
797
798 /* OK to pass negative loff_t, we can catch out-of-range */
799 file->f_mode |= FMODE_UNSIGNED_OFFSET;
800
801 return ret;
b409e578
CW		 802}
803
572d34b9
ON		 804static ssize_t mem_rw(struct file *file, char __user *buf,
805 size_t count, loff_t *ppos, int write)
1da177e4 806{
e268337d 807 struct mm_struct *mm = file->private_data;
572d34b9
ON		 808 unsigned long addr = *ppos;
809 ssize_t copied;
1da177e4 810 char *page;
272ddc8b 811 unsigned int flags;
1da177e4 812
e268337d
LT		 813 if (!mm)
814 return 0;
99f89551 815
30cd8903
KM		 816 page = (char *)__get_free_page(GFP_TEMPORARY);
817 if (!page)
e268337d 818 return -ENOMEM;
1da177e4 819
f7ca54f4 820 copied = 0;
388f7934 821 if (!mmget_not_zero(mm))
6d08f2c7
ON		 822 goto free;
823
f511c0b1 824 flags = FOLL_FORCE | (write ? FOLL_WRITE : 0);
6347e8d5 825
1da177e4 826 while (count > 0) {
572d34b9 827 int this_len = min_t(int, count, PAGE_SIZE);
1da177e4 828
572d34b9 829 if (write && copy_from_user(page, buf, this_len)) {
1da177e4
LT		 830 copied = -EFAULT;
831 break;
832 }
572d34b9 833
6347e8d5 834 this_len = access_remote_vm(mm, addr, page, this_len, flags);
572d34b9 835 if (!this_len) {
1da177e4
LT		 836 if (!copied)
837 copied = -EIO;
838 break;
839 }
572d34b9
ON		 840
841 if (!write && copy_to_user(buf, page, this_len)) {
842 copied = -EFAULT;
843 break;
844 }
845
846 buf += this_len;
847 addr += this_len;
848 copied += this_len;
849 count -= this_len;
1da177e4 850 }
572d34b9 851 *ppos = addr;
30cd8903 852
6d08f2c7
ON		 853 mmput(mm);
854free:
30cd8903 855 free_page((unsigned long) page);
1da177e4
LT		 856 return copied;
857}
1da177e4 858
572d34b9
ON		 859static ssize_t mem_read(struct file *file, char __user *buf,
860 size_t count, loff_t *ppos)
861{
862 return mem_rw(file, buf, count, ppos, 0);
863}
864
865static ssize_t mem_write(struct file *file, const char __user *buf,
866 size_t count, loff_t *ppos)
867{
868 return mem_rw(file, (char __user*)buf, count, ppos, 1);
869}
870
85863e47 871loff_t mem_lseek(struct file *file, loff_t offset, int orig)
1da177e4
LT		 872{
873 switch (orig) {
874 case 0:
875 file->f_pos = offset;
876 break;
877 case 1:
878 file->f_pos += offset;
879 break;
880 default:
881 return -EINVAL;
882 }
883 force_successful_syscall_return();
884 return file->f_pos;
885}
886
e268337d
LT		 887static int mem_release(struct inode *inode, struct file *file)
888{
889 struct mm_struct *mm = file->private_data;
71879d3c 890 if (mm)
6d08f2c7 891 mmdrop(mm);
e268337d
LT		 892 return 0;
893}
894
00977a59 895static const struct file_operations proc_mem_operations = {
1da177e4
LT		 896 .llseek = mem_lseek,
897 .read = mem_read,
898 .write = mem_write,
899 .open = mem_open,
e268337d 900 .release = mem_release,
1da177e4
LT		 901};
902
b409e578
CW		 903static int environ_open(struct inode *inode, struct file *file)
904{
905 return __mem_open(inode, file, PTRACE_MODE_READ);
906}
907
315e28c8
JP		 908static ssize_t environ_read(struct file *file, char __user *buf,
909 size_t count, loff_t *ppos)
910{
315e28c8
JP		 911 char *page;
912 unsigned long src = *ppos;
b409e578
CW		 913 int ret = 0;
914 struct mm_struct *mm = file->private_data;
a3b609ef 915 unsigned long env_start, env_end;
315e28c8 916
8148a73c
MK		 917 /* Ensure the process spawned far enough to have an environment. */
918 if (!mm || !mm->env_end)
b409e578 919 return 0;
315e28c8 920
315e28c8
JP		 921 page = (char *)__get_free_page(GFP_TEMPORARY);
922 if (!page)
b409e578 923 return -ENOMEM;
315e28c8 924
d6f64b89 925 ret = 0;
388f7934 926 if (!mmget_not_zero(mm))
b409e578 927 goto free;
a3b609ef
MG		 928
929 down_read(&mm->mmap_sem);
930 env_start = mm->env_start;
931 env_end = mm->env_end;
932 up_read(&mm->mmap_sem);
933
315e28c8 934 while (count > 0) {
e8905ec2
DH		 935 size_t this_len, max_len;
936 int retval;
315e28c8 937
a3b609ef 938 if (src >= (env_end - env_start))
315e28c8
JP		 939 break;
940
a3b609ef 941 this_len = env_end - (env_start + src);
e8905ec2
DH		 942
943 max_len = min_t(size_t, PAGE_SIZE, count);
944 this_len = min(max_len, this_len);
315e28c8 945
272ddc8b 946 retval = access_remote_vm(mm, (env_start + src), page, this_len, 0);
315e28c8
JP		 947
948 if (retval <= 0) {
949 ret = retval;
950 break;
951 }
952
953 if (copy_to_user(buf, page, retval)) {
954 ret = -EFAULT;
955 break;
956 }
957
958 ret += retval;
959 src += retval;
960 buf += retval;
961 count -= retval;
962 }
963 *ppos = src;
315e28c8 964 mmput(mm);
b409e578
CW		 965
966free:
315e28c8 967 free_page((unsigned long) page);
315e28c8
JP		 968 return ret;
969}
970
971static const struct file_operations proc_environ_operations = {
b409e578 972 .open = environ_open,
315e28c8 973 .read = environ_read,
87df8424 974 .llseek = generic_file_llseek,
b409e578 975 .release = mem_release,
315e28c8
JP		 976};
977
c5317167
AV		 978static int auxv_open(struct inode *inode, struct file *file)
979{
980 return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS);
981}
982
983static ssize_t auxv_read(struct file *file, char __user *buf,
984 size_t count, loff_t *ppos)
985{
986 struct mm_struct *mm = file->private_data;
987 unsigned int nwords = 0;
06b2849d
LY		 988
989 if (!mm)
990 return 0;
c5317167
AV		 991 do {
992 nwords += 2;
993 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
994 return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv,
995 nwords * sizeof(mm->saved_auxv[0]));
996}
997
998static const struct file_operations proc_auxv_operations = {
999 .open = auxv_open,
1000 .read = auxv_read,
1001 .llseek = generic_file_llseek,
1002 .release = mem_release,
1003};
1004
fa0cbbf1
DR		 1005static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
1006 loff_t *ppos)
1007{
496ad9aa 1008 struct task_struct *task = get_proc_task(file_inode(file));
fa0cbbf1
DR		 1009 char buffer[PROC_NUMBUF];
1010 int oom_adj = OOM_ADJUST_MIN;
1011 size_t len;
fa0cbbf1
DR		 1012
1013 if (!task)
1014 return -ESRCH;
f913da59
MH		 1015 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
1016 oom_adj = OOM_ADJUST_MAX;
1017 else
1018 oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
1019 OOM_SCORE_ADJ_MAX;
fa0cbbf1
DR		 1020 put_task_struct(task);
1021 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
1022 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1023}
1024
1d5f0acb
MH		 1025static int __set_oom_adj(struct file *file, int oom_adj, bool legacy)
1026{
1027 static DEFINE_MUTEX(oom_adj_mutex);
44a70ade 1028 struct mm_struct *mm = NULL;
1d5f0acb
MH		 1029 struct task_struct *task;
1030 int err = 0;
1031
1032 task = get_proc_task(file_inode(file));
1033 if (!task)
1034 return -ESRCH;
1035
1036 mutex_lock(&oom_adj_mutex);
1037 if (legacy) {
1038 if (oom_adj < task->signal->oom_score_adj &&
1039 !capable(CAP_SYS_RESOURCE)) {
1040 err = -EACCES;
1041 goto err_unlock;
1042 }
1043 /*
1044 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1045 * /proc/pid/oom_score_adj instead.
1046 */
1047 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1048 current->comm, task_pid_nr(current), task_pid_nr(task),
1049 task_pid_nr(task));
1050 } else {
1051 if ((short)oom_adj < task->signal->oom_score_adj_min &&
1052 !capable(CAP_SYS_RESOURCE)) {
1053 err = -EACCES;
1054 goto err_unlock;
1055 }
1056 }
1057
44a70ade
MH		 1058 /*
1059 * Make sure we will check other processes sharing the mm if this is
1060 * not vfrok which wants its own oom_score_adj.
1061 * pin the mm so it doesn't go away and get reused after task_unlock
1062 */
1063 if (!task->vfork_done) {
1064 struct task_struct *p = find_lock_task_mm(task);
1065
1066 if (p) {
1067 if (atomic_read(&p->mm->mm_users) > 1) {
1068 mm = p->mm;
f1f10076 1069 mmgrab(mm);
44a70ade
MH		 1070 }
1071 task_unlock(p);
1072 }
1073 }
1074
1d5f0acb
MH		 1075 task->signal->oom_score_adj = oom_adj;
1076 if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1077 task->signal->oom_score_adj_min = (short)oom_adj;
1078 trace_oom_score_adj_update(task);
44a70ade
MH		 1079
1080 if (mm) {
1081 struct task_struct *p;
1082
1083 rcu_read_lock();
1084 for_each_process(p) {
1085 if (same_thread_group(task, p))
1086 continue;
1087
1088 /* do not touch kernel threads or the global init */
1089 if (p->flags & PF_KTHREAD || is_global_init(p))
1090 continue;
1091
1092 task_lock(p);
1093 if (!p->vfork_done && process_shares_mm(p, mm)) {
1094 pr_info("updating oom_score_adj for %d (%s) from %d to %d because it shares mm with %d (%s). Report if this is unexpected.\n",
1095 task_pid_nr(p), p->comm,
1096 p->signal->oom_score_adj, oom_adj,
1097 task_pid_nr(task), task->comm);
1098 p->signal->oom_score_adj = oom_adj;
1099 if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1100 p->signal->oom_score_adj_min = (short)oom_adj;
1101 }
1102 task_unlock(p);
1103 }
1104 rcu_read_unlock();
1105 mmdrop(mm);
1106 }
1d5f0acb
MH		 1107err_unlock:
1108 mutex_unlock(&oom_adj_mutex);
1109 put_task_struct(task);
1110 return err;
1111}
f913da59 1112
b72bdfa7
DR		 1113/*
1114 * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1115 * kernels. The effective policy is defined by oom_score_adj, which has a
1116 * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1117 * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1118 * Processes that become oom disabled via oom_adj will still be oom disabled
1119 * with this implementation.
1120 *
1121 * oom_adj cannot be removed since existing userspace binaries use it.
1122 */
fa0cbbf1
DR		 1123static ssize_t oom_adj_write(struct file *file, const char __user *buf,
1124 size_t count, loff_t *ppos)
1125{
fa0cbbf1
DR		 1126 char buffer[PROC_NUMBUF];
1127 int oom_adj;
fa0cbbf1
DR		 1128 int err;
1129
1130 memset(buffer, 0, sizeof(buffer));
1131 if (count > sizeof(buffer) - 1)
1132 count = sizeof(buffer) - 1;
1133 if (copy_from_user(buffer, buf, count)) {
1134 err = -EFAULT;
1135 goto out;
1136 }
1137
1138 err = kstrtoint(strstrip(buffer), 0, &oom_adj);
1139 if (err)
1140 goto out;
1141 if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
1142 oom_adj != OOM_DISABLE) {
1143 err = -EINVAL;
1144 goto out;
1145 }
1146
fa0cbbf1
DR		 1147 /*
1148 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1149 * value is always attainable.
1150 */
1151 if (oom_adj == OOM_ADJUST_MAX)
1152 oom_adj = OOM_SCORE_ADJ_MAX;
1153 else
1154 oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
1155
1d5f0acb 1156 err = __set_oom_adj(file, oom_adj, true);
fa0cbbf1
DR		 1157out:
1158 return err < 0 ? err : count;
1159}
1160
1161static const struct file_operations proc_oom_adj_operations = {
1162 .read = oom_adj_read,
1163 .write = oom_adj_write,
1164 .llseek = generic_file_llseek,
1165};
1166
a63d83f4
DR		 1167static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1168 size_t count, loff_t *ppos)
1169{
496ad9aa 1170 struct task_struct *task = get_proc_task(file_inode(file));
a63d83f4 1171 char buffer[PROC_NUMBUF];
a9c58b90 1172 short oom_score_adj = OOM_SCORE_ADJ_MIN;
a63d83f4
DR		 1173 size_t len;
1174
1175 if (!task)
1176 return -ESRCH;
f913da59 1177 oom_score_adj = task->signal->oom_score_adj;
a63d83f4 1178 put_task_struct(task);
a9c58b90 1179 len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
a63d83f4
DR		 1180 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1181}
1182
1183static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1184 size_t count, loff_t *ppos)
1185{
a63d83f4 1186 char buffer[PROC_NUMBUF];
0a8cb8e3 1187 int oom_score_adj;
a63d83f4
DR		 1188 int err;
1189
1190 memset(buffer, 0, sizeof(buffer));
1191 if (count > sizeof(buffer) - 1)
1192 count = sizeof(buffer) - 1;
723548bf
DR		 1193 if (copy_from_user(buffer, buf, count)) {
1194 err = -EFAULT;
1195 goto out;
1196 }
a63d83f4 1197
0a8cb8e3 1198 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
a63d83f4 1199 if (err)
723548bf 1200 goto out;
a63d83f4 1201 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
723548bf
DR		 1202 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1203 err = -EINVAL;
1204 goto out;
1205 }
a63d83f4 1206
1d5f0acb 1207 err = __set_oom_adj(file, oom_score_adj, false);
723548bf
DR		 1208out:
1209 return err < 0 ? err : count;
a63d83f4
DR		 1210}
1211
1212static const struct file_operations proc_oom_score_adj_operations = {
1213 .read = oom_score_adj_read,
1214 .write = oom_score_adj_write,
6038f373 1215 .llseek = default_llseek,
a63d83f4
DR		 1216};
1217
1da177e4 1218#ifdef CONFIG_AUDITSYSCALL
b4eb4f7f 1219#define TMPBUFLEN 11
1da177e4
LT		 1220static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1221 size_t count, loff_t *ppos)
1222{
496ad9aa 1223 struct inode * inode = file_inode(file);
99f89551 1224 struct task_struct *task = get_proc_task(inode);
1da177e4
LT		 1225 ssize_t length;
1226 char tmpbuf[TMPBUFLEN];
1227
99f89551
EB		 1228 if (!task)
1229 return -ESRCH;
1da177e4 1230 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
e1760bd5
EB		 1231 from_kuid(file->f_cred->user_ns,
1232 audit_get_loginuid(task)));
99f89551 1233 put_task_struct(task);
1da177e4
LT		 1234 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1235}
1236
1237static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1238 size_t count, loff_t *ppos)
1239{
496ad9aa 1240 struct inode * inode = file_inode(file);
1da177e4 1241 uid_t loginuid;
e1760bd5 1242 kuid_t kloginuid;
774636e1 1243 int rv;
1da177e4 1244
7dc52157
PM		 1245 rcu_read_lock();
1246 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1247 rcu_read_unlock();
1da177e4 1248 return -EPERM;
7dc52157
PM		 1249 }
1250 rcu_read_unlock();
1da177e4 1251
1da177e4
LT		 1252 if (*ppos != 0) {
1253 /* No partial writes. */
1254 return -EINVAL;
1255 }
1da177e4 1256
774636e1
AD		 1257 rv = kstrtou32_from_user(buf, count, 10, &loginuid);
1258 if (rv < 0)
1259 return rv;
81407c84
EP		 1260
1261 /* is userspace tring to explicitly UNSET the loginuid? */
1262 if (loginuid == AUDIT_UID_UNSET) {
1263 kloginuid = INVALID_UID;
1264 } else {
1265 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
774636e1
AD		 1266 if (!uid_valid(kloginuid))
1267 return -EINVAL;
e1760bd5
EB		 1268 }
1269
774636e1
AD		 1270 rv = audit_set_loginuid(kloginuid);
1271 if (rv < 0)
1272 return rv;
1273 return count;
1da177e4
LT		 1274}
1275
00977a59 1276static const struct file_operations proc_loginuid_operations = {
1da177e4
LT		 1277 .read = proc_loginuid_read,
1278 .write = proc_loginuid_write,
87df8424 1279 .llseek = generic_file_llseek,
1da177e4 1280};
1e0bd755
EP		 1281
1282static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1283 size_t count, loff_t *ppos)
1284{
496ad9aa 1285 struct inode * inode = file_inode(file);
1e0bd755
EP		 1286 struct task_struct *task = get_proc_task(inode);
1287 ssize_t length;
1288 char tmpbuf[TMPBUFLEN];
1289
1290 if (!task)
1291 return -ESRCH;
1292 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1293 audit_get_sessionid(task));
1294 put_task_struct(task);
1295 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1296}
1297
1298static const struct file_operations proc_sessionid_operations = {
1299 .read = proc_sessionid_read,
87df8424 1300 .llseek = generic_file_llseek,
1e0bd755 1301};
1da177e4
LT		 1302#endif
1303
f4f154fd
AM		 1304#ifdef CONFIG_FAULT_INJECTION
1305static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1306 size_t count, loff_t *ppos)
1307{
496ad9aa 1308 struct task_struct *task = get_proc_task(file_inode(file));
f4f154fd
AM		 1309 char buffer[PROC_NUMBUF];
1310 size_t len;
1311 int make_it_fail;
f4f154fd
AM		 1312
1313 if (!task)
1314 return -ESRCH;
1315 make_it_fail = task->make_it_fail;
1316 put_task_struct(task);
1317
1318 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
0c28f287
AM		 1319
1320 return simple_read_from_buffer(buf, count, ppos, buffer, len);
f4f154fd
AM		 1321}
1322
1323static ssize_t proc_fault_inject_write(struct file * file,
1324 const char __user * buf, size_t count, loff_t *ppos)
1325{
1326 struct task_struct *task;
774636e1 1327 char buffer[PROC_NUMBUF];
f4f154fd 1328 int make_it_fail;
774636e1 1329 int rv;
f4f154fd
AM		 1330
1331 if (!capable(CAP_SYS_RESOURCE))
1332 return -EPERM;
1333 memset(buffer, 0, sizeof(buffer));
1334 if (count > sizeof(buffer) - 1)
1335 count = sizeof(buffer) - 1;
1336 if (copy_from_user(buffer, buf, count))
1337 return -EFAULT;
774636e1
AD		 1338 rv = kstrtoint(strstrip(buffer), 0, &make_it_fail);
1339 if (rv < 0)
1340 return rv;
16caed31
DJ		 1341 if (make_it_fail < 0 || make_it_fail > 1)
1342 return -EINVAL;
1343
496ad9aa 1344 task = get_proc_task(file_inode(file));
f4f154fd
AM		 1345 if (!task)
1346 return -ESRCH;
1347 task->make_it_fail = make_it_fail;
1348 put_task_struct(task);
cba8aafe
VL		 1349
1350 return count;
f4f154fd
AM		 1351}
1352
00977a59 1353static const struct file_operations proc_fault_inject_operations = {
f4f154fd
AM		 1354 .read = proc_fault_inject_read,
1355 .write = proc_fault_inject_write,
87df8424 1356 .llseek = generic_file_llseek,
f4f154fd
AM		 1357};
1358#endif
1359
9745512c 1360
43ae34cb
IM		 1361#ifdef CONFIG_SCHED_DEBUG
1362/*
1363 * Print out various scheduling related per-task fields:
1364 */
1365static int sched_show(struct seq_file *m, void *v)
1366{
1367 struct inode *inode = m->private;
1368 struct task_struct *p;
1369
43ae34cb
IM		 1370 p = get_proc_task(inode);
1371 if (!p)
1372 return -ESRCH;
1373 proc_sched_show_task(p, m);
1374
1375 put_task_struct(p);
1376
1377 return 0;
1378}
1379
1380static ssize_t
1381sched_write(struct file *file, const char __user *buf,
1382 size_t count, loff_t *offset)
1383{
496ad9aa 1384 struct inode *inode = file_inode(file);
43ae34cb
IM		 1385 struct task_struct *p;
1386
43ae34cb
IM		 1387 p = get_proc_task(inode);
1388 if (!p)
1389 return -ESRCH;
1390 proc_sched_set_task(p);
1391
1392 put_task_struct(p);
1393
1394 return count;
1395}
1396
1397static int sched_open(struct inode *inode, struct file *filp)
1398{
c6a34058 1399 return single_open(filp, sched_show, inode);
43ae34cb
IM		 1400}
1401
1402static const struct file_operations proc_pid_sched_operations = {
1403 .open = sched_open,
1404 .read = seq_read,
1405 .write = sched_write,
1406 .llseek = seq_lseek,
5ea473a1 1407 .release = single_release,
43ae34cb
IM		 1408};
1409
1410#endif
1411
5091faa4
MG		 1412#ifdef CONFIG_SCHED_AUTOGROUP
1413/*
1414 * Print out autogroup related information:
1415 */
1416static int sched_autogroup_show(struct seq_file *m, void *v)
1417{
1418 struct inode *inode = m->private;
1419 struct task_struct *p;
1420
1421 p = get_proc_task(inode);
1422 if (!p)
1423 return -ESRCH;
1424 proc_sched_autogroup_show_task(p, m);
1425
1426 put_task_struct(p);
1427
1428 return 0;
1429}
1430
1431static ssize_t
1432sched_autogroup_write(struct file *file, const char __user *buf,
1433 size_t count, loff_t *offset)
1434{
496ad9aa 1435 struct inode *inode = file_inode(file);
5091faa4
MG		 1436 struct task_struct *p;
1437 char buffer[PROC_NUMBUF];
0a8cb8e3 1438 int nice;
5091faa4
MG		 1439 int err;
1440
1441 memset(buffer, 0, sizeof(buffer));
1442 if (count > sizeof(buffer) - 1)
1443 count = sizeof(buffer) - 1;
1444 if (copy_from_user(buffer, buf, count))
1445 return -EFAULT;
1446
0a8cb8e3
AD		 1447 err = kstrtoint(strstrip(buffer), 0, &nice);
1448 if (err < 0)
1449 return err;
5091faa4
MG		 1450
1451 p = get_proc_task(inode);
1452 if (!p)
1453 return -ESRCH;
1454
2e5b5b3a 1455 err = proc_sched_autogroup_set_nice(p, nice);
5091faa4
MG		 1456 if (err)
1457 count = err;
1458
1459 put_task_struct(p);
1460
1461 return count;
1462}
1463
1464static int sched_autogroup_open(struct inode *inode, struct file *filp)
1465{
1466 int ret;
1467
1468 ret = single_open(filp, sched_autogroup_show, NULL);
1469 if (!ret) {
1470 struct seq_file *m = filp->private_data;
1471
1472 m->private = inode;
1473 }
1474 return ret;
1475}
1476
1477static const struct file_operations proc_pid_sched_autogroup_operations = {
1478 .open = sched_autogroup_open,
1479 .read = seq_read,
1480 .write = sched_autogroup_write,
1481 .llseek = seq_lseek,
1482 .release = single_release,
1483};
1484
1485#endif /* CONFIG_SCHED_AUTOGROUP */
1486
4614a696 1487static ssize_t comm_write(struct file *file, const char __user *buf,
1488 size_t count, loff_t *offset)
1489{
496ad9aa 1490 struct inode *inode = file_inode(file);
4614a696 1491 struct task_struct *p;
1492 char buffer[TASK_COMM_LEN];
830e0fc9 1493 const size_t maxlen = sizeof(buffer) - 1;
4614a696 1494
1495 memset(buffer, 0, sizeof(buffer));
830e0fc9 1496 if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
4614a696 1497 return -EFAULT;
1498
1499 p = get_proc_task(inode);
1500 if (!p)
1501 return -ESRCH;
1502
1503 if (same_thread_group(current, p))
1504 set_task_comm(p, buffer);
1505 else
1506 count = -EINVAL;
1507
1508 put_task_struct(p);
1509
1510 return count;
1511}
1512
1513static int comm_show(struct seq_file *m, void *v)
1514{
1515 struct inode *inode = m->private;
1516 struct task_struct *p;
1517
1518 p = get_proc_task(inode);
1519 if (!p)
1520 return -ESRCH;
1521
1522 task_lock(p);
1523 seq_printf(m, "%s\n", p->comm);
1524 task_unlock(p);
1525
1526 put_task_struct(p);
1527
1528 return 0;
1529}
1530
1531static int comm_open(struct inode *inode, struct file *filp)
1532{
c6a34058 1533 return single_open(filp, comm_show, inode);
4614a696 1534}
1535
1536static const struct file_operations proc_pid_set_comm_operations = {
1537 .open = comm_open,
1538 .read = seq_read,
1539 .write = comm_write,
1540 .llseek = seq_lseek,
1541 .release = single_release,
1542};
1543
7773fbc5 1544static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
925d1c40
MH		 1545{
1546 struct task_struct *task;
925d1c40
MH		 1547 struct file *exe_file;
1548
2b0143b5 1549 task = get_proc_task(d_inode(dentry));
925d1c40
MH		 1550 if (!task)
1551 return -ENOENT;
cd81a917 1552 exe_file = get_task_exe_file(task);
925d1c40 1553 put_task_struct(task);
925d1c40
MH		 1554 if (exe_file) {
1555 *exe_path = exe_file->f_path;
1556 path_get(&exe_file->f_path);
1557 fput(exe_file);
1558 return 0;
1559 } else
1560 return -ENOENT;
1561}
1562
6b255391 1563static const char *proc_pid_get_link(struct dentry *dentry,
fceef393
AV		 1564 struct inode *inode,
1565 struct delayed_call *done)
1da177e4 1566{
408ef013 1567 struct path path;
1da177e4
LT		 1568 int error = -EACCES;
1569
6b255391
AV		 1570 if (!dentry)
1571 return ERR_PTR(-ECHILD);
1572
778c1144
EB		 1573 /* Are we allowed to snoop on the tasks file descriptors? */
1574 if (!proc_fd_access_allowed(inode))
1da177e4 1575 goto out;
1da177e4 1576
408ef013
CH		 1577 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1578 if (error)
1579 goto out;
1580
6e77137b 1581 nd_jump_link(&path);
408ef013 1582 return NULL;
1da177e4 1583out:
008b150a 1584 return ERR_PTR(error);
1da177e4
LT		 1585}
1586
3dcd25f3 1587static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1da177e4 1588{
e12ba74d 1589 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
3dcd25f3 1590 char *pathname;
1da177e4
LT		 1591 int len;
1592
1593 if (!tmp)
1594 return -ENOMEM;
0c28f287 1595
7b2a69ba 1596 pathname = d_path(path, tmp, PAGE_SIZE);
3dcd25f3
JB		 1597 len = PTR_ERR(pathname);
1598 if (IS_ERR(pathname))
1da177e4 1599 goto out;
3dcd25f3 1600 len = tmp + PAGE_SIZE - 1 - pathname;
1da177e4
LT		 1601
1602 if (len > buflen)
1603 len = buflen;
3dcd25f3 1604 if (copy_to_user(buffer, pathname, len))
1da177e4
LT		 1605 len = -EFAULT;
1606 out:
1607 free_page((unsigned long)tmp);
1608 return len;
1609}
1610
1611static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1612{
1613 int error = -EACCES;
2b0143b5 1614 struct inode *inode = d_inode(dentry);
3dcd25f3 1615 struct path path;
1da177e4 1616
778c1144
EB		 1617 /* Are we allowed to snoop on the tasks file descriptors? */
1618 if (!proc_fd_access_allowed(inode))
1da177e4 1619 goto out;
1da177e4 1620
7773fbc5 1621 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1da177e4
LT		 1622 if (error)
1623 goto out;
1624
3dcd25f3
JB		 1625 error = do_proc_readlink(&path, buffer, buflen);
1626 path_put(&path);
1da177e4 1627out:
1da177e4
LT		 1628 return error;
1629}
1630
faf60af1 1631const struct inode_operations proc_pid_link_inode_operations = {
1da177e4 1632 .readlink = proc_pid_readlink,
6b255391 1633 .get_link = proc_pid_get_link,
6d76fa58 1634 .setattr = proc_setattr,
1da177e4
LT		 1635};
1636
28a6d671
EB		 1637
1638/* building an inode */
1639
68eb94f1
EB		 1640void task_dump_owner(struct task_struct *task, mode_t mode,
1641 kuid_t *ruid, kgid_t *rgid)
1642{
1643 /* Depending on the state of dumpable compute who should own a
1644 * proc file for a task.
1645 */
1646 const struct cred *cred;
1647 kuid_t uid;
1648 kgid_t gid;
1649
1650 /* Default to the tasks effective ownership */
1651 rcu_read_lock();
1652 cred = __task_cred(task);
1653 uid = cred->euid;
1654 gid = cred->egid;
1655 rcu_read_unlock();
1656
1657 /*
1658 * Before the /proc/pid/status file was created the only way to read
1659 * the effective uid of a /process was to stat /proc/pid. Reading
1660 * /proc/pid/status is slow enough that procps and other packages
1661 * kept stating /proc/pid. To keep the rules in /proc simple I have
1662 * made this apply to all per process world readable and executable
1663 * directories.
1664 */
1665 if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) {
1666 struct mm_struct *mm;
1667 task_lock(task);
1668 mm = task->mm;
1669 /* Make non-dumpable tasks owned by some root */
1670 if (mm) {
1671 if (get_dumpable(mm) != SUID_DUMP_USER) {
1672 struct user_namespace *user_ns = mm->user_ns;
1673
1674 uid = make_kuid(user_ns, 0);
1675 if (!uid_valid(uid))
1676 uid = GLOBAL_ROOT_UID;
1677
1678 gid = make_kgid(user_ns, 0);
1679 if (!gid_valid(gid))
1680 gid = GLOBAL_ROOT_GID;
1681 }
1682 } else {
1683 uid = GLOBAL_ROOT_UID;
1684 gid = GLOBAL_ROOT_GID;
1685 }
1686 task_unlock(task);
1687 }
1688 *ruid = uid;
1689 *rgid = gid;
1690}
1691
db978da8
AG		 1692struct inode *proc_pid_make_inode(struct super_block * sb,
1693 struct task_struct *task, umode_t mode)
28a6d671
EB		 1694{
1695 struct inode * inode;
1696 struct proc_inode *ei;
1da177e4 1697
28a6d671 1698 /* We need a new inode */
1da177e4 1699
28a6d671
EB		 1700 inode = new_inode(sb);
1701 if (!inode)
1702 goto out;
1703
1704 /* Common stuff */
1705 ei = PROC_I(inode);
db978da8 1706 inode->i_mode = mode;
85fe4025 1707 inode->i_ino = get_next_ino();
078cd827 1708 inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode);
28a6d671
EB		 1709 inode->i_op = &proc_def_inode_operations;
1710
1711 /*
1712 * grab the reference to task.
1713 */
1a657f78 1714 ei->pid = get_task_pid(task, PIDTYPE_PID);
28a6d671
EB		 1715 if (!ei->pid)
1716 goto out_unlock;
1717
68eb94f1 1718 task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
28a6d671
EB		 1719 security_task_to_inode(task, inode);
1720
1da177e4 1721out:
28a6d671
EB		 1722 return inode;
1723
1724out_unlock:
1725 iput(inode);
1726 return NULL;
1da177e4
LT		 1727}
1728
a528d35e
DH		 1729int pid_getattr(const struct path *path, struct kstat *stat,
1730 u32 request_mask, unsigned int query_flags)
1da177e4 1731{
a528d35e 1732 struct inode *inode = d_inode(path->dentry);
28a6d671 1733 struct task_struct *task;
a528d35e 1734 struct pid_namespace *pid = path->dentry->d_sb->s_fs_info;
c69e8d9c 1735
28a6d671 1736 generic_fillattr(inode, stat);
1da177e4 1737
28a6d671 1738 rcu_read_lock();
dcb0f222
EB		 1739 stat->uid = GLOBAL_ROOT_UID;
1740 stat->gid = GLOBAL_ROOT_GID;
28a6d671
EB		 1741 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1742 if (task) {
796f571b 1743 if (!has_pid_permissions(pid, task, HIDEPID_INVISIBLE)) {
0499680a
VK		 1744 rcu_read_unlock();
1745 /*
1746 * This doesn't prevent learning whether PID exists,
1747 * it only makes getattr() consistent with readdir().
1748 */
1749 return -ENOENT;
1750 }
68eb94f1 1751 task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid);
1da177e4 1752 }
28a6d671 1753 rcu_read_unlock();
d6e71144 1754 return 0;
1da177e4
LT		 1755}
1756
1da177e4
LT		 1757/* dentry stuff */
1758
1759/*
1760 * Exceptional case: normally we are not allowed to unhash a busy
1761 * directory. In this case, however, we can do it - no aliasing problems
1762 * due to the way we treat inodes.
1763 *
1764 * Rewrite the inode's ownerships here because the owning task may have
1765 * performed a setuid(), etc.
99f89551 1766 *
1da177e4 1767 */
0b728e19 1768int pid_revalidate(struct dentry *dentry, unsigned int flags)
1da177e4 1769{
34286d66
NP		 1770 struct inode *inode;
1771 struct task_struct *task;
c69e8d9c 1772
0b728e19 1773 if (flags & LOOKUP_RCU)
34286d66
NP		 1774 return -ECHILD;
1775
2b0143b5 1776 inode = d_inode(dentry);
34286d66
NP		 1777 task = get_proc_task(inode);
1778
99f89551 1779 if (task) {
68eb94f1
EB		 1780 task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid);
1781
9ee8ab9f 1782 inode->i_mode &= ~(S_ISUID | S_ISGID);
1da177e4 1783 security_task_to_inode(task, inode);
99f89551 1784 put_task_struct(task);
1da177e4
LT		 1785 return 1;
1786 }
1da177e4
LT		 1787 return 0;
1788}
1789
d855a4b7
ON		 1790static inline bool proc_inode_is_dead(struct inode *inode)
1791{
1792 return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
1793}
1794
1dd704b6
DH		 1795int pid_delete_dentry(const struct dentry *dentry)
1796{
1797 /* Is the task we represent dead?
1798 * If so, then don't put the dentry on the lru list,
1799 * kill it immediately.
1800 */
2b0143b5 1801 return proc_inode_is_dead(d_inode(dentry));
1dd704b6
DH		 1802}
1803
6b4e306a 1804const struct dentry_operations pid_dentry_operations =
28a6d671
EB		 1805{
1806 .d_revalidate = pid_revalidate,
1807 .d_delete = pid_delete_dentry,
1808};
1809
1810/* Lookups */
1811
1c0d04c9
EB		 1812/*
1813 * Fill a directory entry.
1814 *
1815 * If possible create the dcache entry and derive our inode number and
1816 * file type from dcache entry.
1817 *
1818 * Since all of the proc inode numbers are dynamically generated, the inode
1819 * numbers do not exist until the inode is cache. This means creating the
1820 * the dcache entry in readdir is necessary to keep the inode numbers
1821 * reported by readdir in sync with the inode numbers reported
1822 * by stat.
1823 */
f0c3b509 1824bool proc_fill_cache(struct file *file, struct dir_context *ctx,
6b4e306a 1825 const char *name, int len,
c5141e6d 1826 instantiate_t instantiate, struct task_struct *task, const void *ptr)
61a28784 1827{
f0c3b509 1828 struct dentry *child, *dir = file->f_path.dentry;
1df98b8b 1829 struct qstr qname = QSTR_INIT(name, len);
61a28784 1830 struct inode *inode;
1df98b8b
AV		 1831 unsigned type;
1832 ino_t ino;
61a28784 1833
1df98b8b 1834 child = d_hash_and_lookup(dir, &qname);
61a28784 1835 if (!child) {
3781764b
AV		 1836 DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq);
1837 child = d_alloc_parallel(dir, &qname, &wq);
1838 if (IS_ERR(child))
1df98b8b 1839 goto end_instantiate;
3781764b
AV		 1840 if (d_in_lookup(child)) {
1841 int err = instantiate(d_inode(dir), child, task, ptr);
1842 d_lookup_done(child);
1843 if (err < 0) {
1844 dput(child);
1845 goto end_instantiate;
1846 }
61a28784
EB		 1847 }
1848 }
2b0143b5 1849 inode = d_inode(child);
147ce699
AV		 1850 ino = inode->i_ino;
1851 type = inode->i_mode >> 12;
61a28784 1852 dput(child);
f0c3b509 1853 return dir_emit(ctx, name, len, ino, type);
1df98b8b
AV		 1854
1855end_instantiate:
1856 return dir_emit(ctx, name, len, 1, DT_UNKNOWN);
61a28784
EB		 1857}
1858
640708a2
PE		 1859/*
1860 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1861 * which represent vma start and end addresses.
1862 */
1863static int dname_to_vma_addr(struct dentry *dentry,
1864 unsigned long *start, unsigned long *end)
1865{
1866 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
1867 return -EINVAL;
1868
1869 return 0;
1870}
1871
0b728e19 1872static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
640708a2
PE		 1873{
1874 unsigned long vm_start, vm_end;
1875 bool exact_vma_exists = false;
1876 struct mm_struct *mm = NULL;
1877 struct task_struct *task;
640708a2
PE		 1878 struct inode *inode;
1879 int status = 0;
1880
0b728e19 1881 if (flags & LOOKUP_RCU)
640708a2
PE		 1882 return -ECHILD;
1883
2b0143b5 1884 inode = d_inode(dentry);
640708a2
PE		 1885 task = get_proc_task(inode);
1886 if (!task)
1887 goto out_notask;
1888
caaee623 1889 mm = mm_access(task, PTRACE_MODE_READ_FSCREDS);
2344bec7 1890 if (IS_ERR_OR_NULL(mm))
640708a2
PE		 1891 goto out;
1892
1893 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1894 down_read(&mm->mmap_sem);
1895 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1896 up_read(&mm->mmap_sem);
1897 }
1898
1899 mmput(mm);
1900
1901 if (exact_vma_exists) {
68eb94f1
EB		 1902 task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
1903
640708a2
PE		 1904 security_task_to_inode(task, inode);
1905 status = 1;
1906 }
1907
1908out:
1909 put_task_struct(task);
1910
1911out_notask:
640708a2
PE		 1912 return status;
1913}
1914
1915static const struct dentry_operations tid_map_files_dentry_operations = {
1916 .d_revalidate = map_files_d_revalidate,
1917 .d_delete = pid_delete_dentry,
1918};
1919
6b255391 1920static int map_files_get_link(struct dentry *dentry, struct path *path)
640708a2
PE		 1921{
1922 unsigned long vm_start, vm_end;
1923 struct vm_area_struct *vma;
1924 struct task_struct *task;
1925 struct mm_struct *mm;
1926 int rc;
1927
1928 rc = -ENOENT;
2b0143b5 1929 task = get_proc_task(d_inode(dentry));
640708a2
PE		 1930 if (!task)
1931 goto out;
1932
1933 mm = get_task_mm(task);
1934 put_task_struct(task);
1935 if (!mm)
1936 goto out;
1937
1938 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
1939 if (rc)
1940 goto out_mmput;
1941
70335abb 1942 rc = -ENOENT;
640708a2
PE		 1943 down_read(&mm->mmap_sem);
1944 vma = find_exact_vma(mm, vm_start, vm_end);
1945 if (vma && vma->vm_file) {
1946 *path = vma->vm_file->f_path;
1947 path_get(path);
1948 rc = 0;
1949 }
1950 up_read(&mm->mmap_sem);
1951
1952out_mmput:
1953 mmput(mm);
1954out:
1955 return rc;
1956}
1957
1958struct map_files_info {
7b540d06 1959 fmode_t mode;
9a87fe0d 1960 unsigned int len;
640708a2
PE		 1961 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
1962};
1963
bdb4d100
CO		 1964/*
1965 * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the
1966 * symlinks may be used to bypass permissions on ancestor directories in the
1967 * path to the file in question.
1968 */
1969static const char *
6b255391 1970proc_map_files_get_link(struct dentry *dentry,
fceef393
AV		 1971 struct inode *inode,
1972 struct delayed_call *done)
bdb4d100
CO		 1973{
1974 if (!capable(CAP_SYS_ADMIN))
1975 return ERR_PTR(-EPERM);
1976
fceef393 1977 return proc_pid_get_link(dentry, inode, done);
bdb4d100
CO		 1978}
1979
1980/*
6b255391 1981 * Identical to proc_pid_link_inode_operations except for get_link()
bdb4d100
CO		 1982 */
1983static const struct inode_operations proc_map_files_link_inode_operations = {
1984 .readlink = proc_pid_readlink,
6b255391 1985 .get_link = proc_map_files_get_link,
bdb4d100
CO		 1986 .setattr = proc_setattr,
1987};
1988
c52a47ac 1989static int
640708a2
PE		 1990proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
1991 struct task_struct *task, const void *ptr)
1992{
7b540d06 1993 fmode_t mode = (fmode_t)(unsigned long)ptr;
640708a2
PE		 1994 struct proc_inode *ei;
1995 struct inode *inode;
1996
db978da8
AG		 1997 inode = proc_pid_make_inode(dir->i_sb, task, S_IFLNK |
1998 ((mode & FMODE_READ ) ? S_IRUSR : 0) |
1999 ((mode & FMODE_WRITE) ? S_IWUSR : 0));
640708a2 2000 if (!inode)
c52a47ac 2001 return -ENOENT;
640708a2
PE		 2002
2003 ei = PROC_I(inode);
6b255391 2004 ei->op.proc_get_link = map_files_get_link;
640708a2 2005
bdb4d100 2006 inode->i_op = &proc_map_files_link_inode_operations;
640708a2 2007 inode->i_size = 64;
640708a2
PE		 2008
2009 d_set_d_op(dentry, &tid_map_files_dentry_operations);
2010 d_add(dentry, inode);
2011
c52a47ac 2012 return 0;
640708a2
PE		 2013}
2014
2015static struct dentry *proc_map_files_lookup(struct inode *dir,
00cd8dd3 2016 struct dentry *dentry, unsigned int flags)
640708a2
PE		 2017{
2018 unsigned long vm_start, vm_end;
2019 struct vm_area_struct *vma;
2020 struct task_struct *task;
c52a47ac 2021 int result;
640708a2
PE		 2022 struct mm_struct *mm;
2023
c52a47ac 2024 result = -ENOENT;
640708a2
PE		 2025 task = get_proc_task(dir);
2026 if (!task)
2027 goto out;
2028
c52a47ac 2029 result = -EACCES;
caaee623 2030 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
640708a2
PE		 2031 goto out_put_task;
2032
c52a47ac 2033 result = -ENOENT;
640708a2 2034 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
eb94cd96 2035 goto out_put_task;
640708a2
PE		 2036
2037 mm = get_task_mm(task);
2038 if (!mm)
eb94cd96 2039 goto out_put_task;
640708a2
PE		 2040
2041 down_read(&mm->mmap_sem);
2042 vma = find_exact_vma(mm, vm_start, vm_end);
2043 if (!vma)
2044 goto out_no_vma;
2045
05f56484
SK		 2046 if (vma->vm_file)
2047 result = proc_map_files_instantiate(dir, dentry, task,
2048 (void *)(unsigned long)vma->vm_file->f_mode);
640708a2
PE		 2049
2050out_no_vma:
2051 up_read(&mm->mmap_sem);
2052 mmput(mm);
640708a2
PE		 2053out_put_task:
2054 put_task_struct(task);
2055out:
c52a47ac 2056 return ERR_PTR(result);
640708a2
PE		 2057}
2058
2059static const struct inode_operations proc_map_files_inode_operations = {
2060 .lookup = proc_map_files_lookup,
2061 .permission = proc_fd_permission,
2062 .setattr = proc_setattr,
2063};
2064
2065static int
f0c3b509 2066proc_map_files_readdir(struct file *file, struct dir_context *ctx)
640708a2 2067{
640708a2
PE		 2068 struct vm_area_struct *vma;
2069 struct task_struct *task;
2070 struct mm_struct *mm;
f0c3b509
AV		 2071 unsigned long nr_files, pos, i;
2072 struct flex_array *fa = NULL;
2073 struct map_files_info info;
2074 struct map_files_info *p;
640708a2
PE		 2075 int ret;
2076
640708a2 2077 ret = -ENOENT;
f0c3b509 2078 task = get_proc_task(file_inode(file));
640708a2
PE		 2079 if (!task)
2080 goto out;
2081
2082 ret = -EACCES;
caaee623 2083 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
640708a2
PE		 2084 goto out_put_task;
2085
2086 ret = 0;
f0c3b509
AV		 2087 if (!dir_emit_dots(file, ctx))
2088 goto out_put_task;
640708a2 2089
f0c3b509
AV		 2090 mm = get_task_mm(task);
2091 if (!mm)
2092 goto out_put_task;
2093 down_read(&mm->mmap_sem);
640708a2 2094
f0c3b509 2095 nr_files = 0;
640708a2 2096
f0c3b509
AV		 2097 /*
2098 * We need two passes here:
2099 *
2100 * 1) Collect vmas of mapped files with mmap_sem taken
2101 * 2) Release mmap_sem and instantiate entries
2102 *
2103 * otherwise we get lockdep complained, since filldir()
2104 * routine might require mmap_sem taken in might_fault().
2105 */
640708a2 2106
f0c3b509
AV		 2107 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2108 if (vma->vm_file && ++pos > ctx->pos)
2109 nr_files++;
2110 }
2111
2112 if (nr_files) {
2113 fa = flex_array_alloc(sizeof(info), nr_files,
2114 GFP_KERNEL);
2115 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2116 GFP_KERNEL)) {
2117 ret = -ENOMEM;
2118 if (fa)
2119 flex_array_free(fa);
2120 up_read(&mm->mmap_sem);
2121 mmput(mm);
2122 goto out_put_task;
640708a2 2123 }
f0c3b509
AV		 2124 for (i = 0, vma = mm->mmap, pos = 2; vma;
2125 vma = vma->vm_next) {
2126 if (!vma->vm_file)
2127 continue;
2128 if (++pos <= ctx->pos)
2129 continue;
2130
2131 info.mode = vma->vm_file->f_mode;
2132 info.len = snprintf(info.name,
2133 sizeof(info.name), "%lx-%lx",
2134 vma->vm_start, vma->vm_end);
2135 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2136 BUG();
640708a2 2137 }
640708a2 2138 }
f0c3b509
AV		 2139 up_read(&mm->mmap_sem);
2140
2141 for (i = 0; i < nr_files; i++) {
2142 p = flex_array_get(fa, i);
2143 if (!proc_fill_cache(file, ctx,
2144 p->name, p->len,
2145 proc_map_files_instantiate,
2146 task,
2147 (void *)(unsigned long)p->mode))
2148 break;
2149 ctx->pos++;
640708a2 2150 }
f0c3b509
AV		 2151 if (fa)
2152 flex_array_free(fa);
2153 mmput(mm);
640708a2 2154
640708a2
PE		 2155out_put_task:
2156 put_task_struct(task);
2157out:
2158 return ret;
2159}
2160
2161static const struct file_operations proc_map_files_operations = {
2162 .read = generic_read_dir,
f50752ea
AV		 2163 .iterate_shared = proc_map_files_readdir,
2164 .llseek = generic_file_llseek,
640708a2
PE		 2165};
2166
b18b6a9c 2167#if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
48f6a7a5
PE		 2168struct timers_private {
2169 struct pid *pid;
2170 struct task_struct *task;
2171 struct sighand_struct *sighand;
57b8015e 2172 struct pid_namespace *ns;
48f6a7a5
PE		 2173 unsigned long flags;
2174};
2175
2176static void *timers_start(struct seq_file *m, loff_t *pos)
2177{
2178 struct timers_private *tp = m->private;
2179
2180 tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2181 if (!tp->task)
2182 return ERR_PTR(-ESRCH);
2183
2184 tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2185 if (!tp->sighand)
2186 return ERR_PTR(-ESRCH);
2187
2188 return seq_list_start(&tp->task->signal->posix_timers, *pos);
2189}
2190
2191static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2192{
2193 struct timers_private *tp = m->private;
2194 return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2195}
2196
2197static void timers_stop(struct seq_file *m, void *v)
2198{
2199 struct timers_private *tp = m->private;
2200
2201 if (tp->sighand) {
2202 unlock_task_sighand(tp->task, &tp->flags);
2203 tp->sighand = NULL;
2204 }
2205
2206 if (tp->task) {
2207 put_task_struct(tp->task);
2208 tp->task = NULL;
2209 }
2210}
2211
2212static int show_timer(struct seq_file *m, void *v)
2213{
2214 struct k_itimer *timer;
57b8015e
PE		 2215 struct timers_private *tp = m->private;
2216 int notify;
cedbccab 2217 static const char * const nstr[] = {
57b8015e
PE		 2218 [SIGEV_SIGNAL] = "signal",
2219 [SIGEV_NONE] = "none",
2220 [SIGEV_THREAD] = "thread",
2221 };
48f6a7a5
PE		 2222
2223 timer = list_entry((struct list_head *)v, struct k_itimer, list);
57b8015e
PE		 2224 notify = timer->it_sigev_notify;
2225
48f6a7a5 2226 seq_printf(m, "ID: %d\n", timer->it_id);
25ce3191
JP		 2227 seq_printf(m, "signal: %d/%p\n",
2228 timer->sigq->info.si_signo,
2229 timer->sigq->info.si_value.sival_ptr);
57b8015e 2230 seq_printf(m, "notify: %s/%s.%d\n",
25ce3191
JP		 2231 nstr[notify & ~SIGEV_THREAD_ID],
2232 (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2233 pid_nr_ns(timer->it_pid, tp->ns));
15ef0298 2234 seq_printf(m, "ClockID: %d\n", timer->it_clock);
48f6a7a5
PE		 2235
2236 return 0;
2237}
2238
2239static const struct seq_operations proc_timers_seq_ops = {
2240 .start = timers_start,
2241 .next = timers_next,
2242 .stop = timers_stop,
2243 .show = show_timer,
2244};
2245
2246static int proc_timers_open(struct inode *inode, struct file *file)
2247{
2248 struct timers_private *tp;
2249
2250 tp = __seq_open_private(file, &proc_timers_seq_ops,
2251 sizeof(struct timers_private));
2252 if (!tp)
2253 return -ENOMEM;
2254
2255 tp->pid = proc_pid(inode);
57b8015e 2256 tp->ns = inode->i_sb->s_fs_info;
48f6a7a5
PE		 2257 return 0;
2258}
2259
2260static const struct file_operations proc_timers_operations = {
2261 .open = proc_timers_open,
2262 .read = seq_read,
2263 .llseek = seq_lseek,
2264 .release = seq_release_private,
2265};
b5946bea 2266#endif
640708a2 2267
5de23d43
JS		 2268static ssize_t timerslack_ns_write(struct file *file, const char __user *buf,
2269 size_t count, loff_t *offset)
2270{
2271 struct inode *inode = file_inode(file);
2272 struct task_struct *p;
2273 u64 slack_ns;
2274 int err;
2275
2276 err = kstrtoull_from_user(buf, count, 10, &slack_ns);
2277 if (err < 0)
2278 return err;
2279
2280 p = get_proc_task(inode);
2281 if (!p)
2282 return -ESRCH;
2283
4b2bd5fe
JS		 2284 if (p != current) {
2285 if (!capable(CAP_SYS_NICE)) {
2286 count = -EPERM;
2287 goto out;
2288 }
5de23d43 2289
4b2bd5fe
JS		 2290 err = security_task_setscheduler(p);
2291 if (err) {
2292 count = err;
2293 goto out;
2294 }
904763e1
JS		 2295 }
2296
7abbaf94
JS		 2297 task_lock(p);
2298 if (slack_ns == 0)
2299 p->timer_slack_ns = p->default_timer_slack_ns;
2300 else
2301 p->timer_slack_ns = slack_ns;
2302 task_unlock(p);
2303
2304out:
5de23d43
JS		 2305 put_task_struct(p);
2306
2307 return count;
2308}
2309
2310static int timerslack_ns_show(struct seq_file *m, void *v)
2311{
2312 struct inode *inode = m->private;
2313 struct task_struct *p;
7abbaf94 2314 int err = 0;
5de23d43
JS		 2315
2316 p = get_proc_task(inode);
2317 if (!p)
2318 return -ESRCH;
2319
4b2bd5fe 2320 if (p != current) {
5de23d43 2321
4b2bd5fe
JS		 2322 if (!capable(CAP_SYS_NICE)) {
2323 err = -EPERM;
2324 goto out;
2325 }
2326 err = security_task_getscheduler(p);
2327 if (err)
2328 goto out;
2329 }
904763e1 2330
7abbaf94
JS		 2331 task_lock(p);
2332 seq_printf(m, "%llu\n", p->timer_slack_ns);
2333 task_unlock(p);
2334
2335out:
5de23d43
JS		 2336 put_task_struct(p);
2337
2338 return err;
2339}
2340
2341static int timerslack_ns_open(struct inode *inode, struct file *filp)
2342{
2343 return single_open(filp, timerslack_ns_show, inode);
2344}
2345
2346static const struct file_operations proc_pid_set_timerslack_ns_operations = {
2347 .open = timerslack_ns_open,
2348 .read = seq_read,
2349 .write = timerslack_ns_write,
2350 .llseek = seq_lseek,
2351 .release = single_release,
2352};
2353
c52a47ac 2354static int proc_pident_instantiate(struct inode *dir,
c5141e6d 2355 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8 2356{
c5141e6d 2357 const struct pid_entry *p = ptr;
444ceed8
EB		 2358 struct inode *inode;
2359 struct proc_inode *ei;
444ceed8 2360
db978da8 2361 inode = proc_pid_make_inode(dir->i_sb, task, p->mode);
444ceed8
EB		 2362 if (!inode)
2363 goto out;
2364
2365 ei = PROC_I(inode);
444ceed8 2366 if (S_ISDIR(inode->i_mode))
bfe86848 2367 set_nlink(inode, 2); /* Use getattr to fix if necessary */
444ceed8
EB		 2368 if (p->iop)
2369 inode->i_op = p->iop;
2370 if (p->fop)
2371 inode->i_fop = p->fop;
2372 ei->op = p->op;
fb045adb 2373 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 2374 d_add(dentry, inode);
2375 /* Close the race of the process dying before we return the dentry */
0b728e19 2376 if (pid_revalidate(dentry, 0))
c52a47ac 2377 return 0;
444ceed8 2378out:
c52a47ac 2379 return -ENOENT;
444ceed8
EB		 2380}
2381
1da177e4
LT		 2382static struct dentry *proc_pident_lookup(struct inode *dir,
2383 struct dentry *dentry,
c5141e6d 2384 const struct pid_entry *ents,
7bcd6b0e 2385 unsigned int nents)
1da177e4 2386{
c52a47ac 2387 int error;
99f89551 2388 struct task_struct *task = get_proc_task(dir);
c5141e6d 2389 const struct pid_entry *p, *last;
1da177e4 2390
c52a47ac 2391 error = -ENOENT;
1da177e4 2392
99f89551
EB		 2393 if (!task)
2394 goto out_no_task;
1da177e4 2395
20cdc894
EB		 2396 /*
2397 * Yes, it does not scale. And it should not. Don't add
2398 * new entries into /proc/<tgid>/ without very good reasons.
2399 */
bac5f5d5
AD		 2400 last = &ents[nents];
2401 for (p = ents; p < last; p++) {
1da177e4
LT		 2402 if (p->len != dentry->d_name.len)
2403 continue;
2404 if (!memcmp(dentry->d_name.name, p->name, p->len))
2405 break;
2406 }
bac5f5d5 2407 if (p >= last)
1da177e4
LT		 2408 goto out;
2409
444ceed8 2410 error = proc_pident_instantiate(dir, dentry, task, p);
1da177e4 2411out:
99f89551
EB		 2412 put_task_struct(task);
2413out_no_task:
c52a47ac 2414 return ERR_PTR(error);
1da177e4
LT		 2415}
2416
f0c3b509 2417static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
c5141e6d 2418 const struct pid_entry *ents, unsigned int nents)
28a6d671 2419{
f0c3b509
AV		 2420 struct task_struct *task = get_proc_task(file_inode(file));
2421 const struct pid_entry *p;
28a6d671 2422
28a6d671 2423 if (!task)
f0c3b509 2424 return -ENOENT;
28a6d671 2425
f0c3b509
AV		 2426 if (!dir_emit_dots(file, ctx))
2427 goto out;
2428
2429 if (ctx->pos >= nents + 2)
2430 goto out;
28a6d671 2431
bac5f5d5 2432 for (p = ents + (ctx->pos - 2); p < ents + nents; p++) {
f0c3b509
AV		 2433 if (!proc_fill_cache(file, ctx, p->name, p->len,
2434 proc_pident_instantiate, task, p))
2435 break;
2436 ctx->pos++;
2437 }
28a6d671 2438out:
61a28784 2439 put_task_struct(task);
f0c3b509 2440 return 0;
1da177e4
LT		 2441}
2442
28a6d671
EB		 2443#ifdef CONFIG_SECURITY
2444static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2445 size_t count, loff_t *ppos)
2446{
496ad9aa 2447 struct inode * inode = file_inode(file);
04ff9708 2448 char *p = NULL;
28a6d671
EB		 2449 ssize_t length;
2450 struct task_struct *task = get_proc_task(inode);
2451
28a6d671 2452 if (!task)
04ff9708 2453 return -ESRCH;
28a6d671
EB		 2454
2455 length = security_getprocattr(task,
2fddfeef 2456 (char*)file->f_path.dentry->d_name.name,
04ff9708 2457 &p);
28a6d671 2458 put_task_struct(task);
04ff9708
AV		 2459 if (length > 0)
2460 length = simple_read_from_buffer(buf, count, ppos, p, length);
2461 kfree(p);
28a6d671 2462 return length;
1da177e4
LT		 2463}
2464
28a6d671
EB		 2465static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2466 size_t count, loff_t *ppos)
2467{
496ad9aa 2468 struct inode * inode = file_inode(file);
bb646cdb 2469 void *page;
28a6d671
EB		 2470 ssize_t length;
2471 struct task_struct *task = get_proc_task(inode);
2472
2473 length = -ESRCH;
2474 if (!task)
2475 goto out_no_task;
b21507e2
SS		 2476
2477 /* A task may only write its own attributes. */
2478 length = -EACCES;
2479 if (current != task)
2480 goto out;
2481
28a6d671
EB		 2482 if (count > PAGE_SIZE)
2483 count = PAGE_SIZE;
2484
2485 /* No partial writes. */
2486 length = -EINVAL;
2487 if (*ppos != 0)
2488 goto out;
2489
bb646cdb
AV		 2490 page = memdup_user(buf, count);
2491 if (IS_ERR(page)) {
2492 length = PTR_ERR(page);
28a6d671 2493 goto out;
bb646cdb 2494 }
28a6d671 2495
107db7c7 2496 /* Guard against adverse ptrace interaction */
b21507e2 2497 length = mutex_lock_interruptible(&current->signal->cred_guard_mutex);
107db7c7
DH		 2498 if (length < 0)
2499 goto out_free;
2500
b21507e2 2501 length = security_setprocattr(file->f_path.dentry->d_name.name,
bb646cdb 2502 page, count);
b21507e2 2503 mutex_unlock(&current->signal->cred_guard_mutex);
28a6d671 2504out_free:
bb646cdb 2505 kfree(page);
28a6d671
EB		 2506out:
2507 put_task_struct(task);
2508out_no_task:
2509 return length;
2510}
2511
00977a59 2512static const struct file_operations proc_pid_attr_operations = {
28a6d671
EB		 2513 .read = proc_pid_attr_read,
2514 .write = proc_pid_attr_write,
87df8424 2515 .llseek = generic_file_llseek,
28a6d671
EB		 2516};
2517
c5141e6d 2518static const struct pid_entry attr_dir_stuff[] = {
631f9c18
AD		 2519 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2520 REG("prev", S_IRUGO, proc_pid_attr_operations),
2521 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2522 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2523 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2524 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
28a6d671
EB		 2525};
2526
f0c3b509 2527static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
28a6d671 2528{
f0c3b509
AV		 2529 return proc_pident_readdir(file, ctx,
2530 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2531}
2532
00977a59 2533static const struct file_operations proc_attr_dir_operations = {
1da177e4 2534 .read = generic_read_dir,
f50752ea
AV		 2535 .iterate_shared = proc_attr_dir_readdir,
2536 .llseek = generic_file_llseek,
1da177e4
LT		 2537};
2538
72d9dcfc 2539static struct dentry *proc_attr_dir_lookup(struct inode *dir,
00cd8dd3 2540 struct dentry *dentry, unsigned int flags)
28a6d671 2541{
7bcd6b0e
EB		 2542 return proc_pident_lookup(dir, dentry,
2543 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2544}
2545
c5ef1c42 2546static const struct inode_operations proc_attr_dir_inode_operations = {
72d9dcfc 2547 .lookup = proc_attr_dir_lookup,
99f89551 2548 .getattr = pid_getattr,
6d76fa58 2549 .setattr = proc_setattr,
1da177e4
LT		 2550};
2551
28a6d671
EB		 2552#endif
2553
698ba7b5 2554#ifdef CONFIG_ELF_CORE
3cb4a0bb
KH		 2555static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2556 size_t count, loff_t *ppos)
2557{
496ad9aa 2558 struct task_struct *task = get_proc_task(file_inode(file));
3cb4a0bb
KH		 2559 struct mm_struct *mm;
2560 char buffer[PROC_NUMBUF];
2561 size_t len;
2562 int ret;
2563
2564 if (!task)
2565 return -ESRCH;
2566
2567 ret = 0;
2568 mm = get_task_mm(task);
2569 if (mm) {
2570 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2571 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2572 MMF_DUMP_FILTER_SHIFT));
2573 mmput(mm);
2574 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2575 }
2576
2577 put_task_struct(task);
2578
2579 return ret;
2580}
2581
2582static ssize_t proc_coredump_filter_write(struct file *file,
2583 const char __user *buf,
2584 size_t count,
2585 loff_t *ppos)
2586{
2587 struct task_struct *task;
2588 struct mm_struct *mm;
3cb4a0bb
KH		 2589 unsigned int val;
2590 int ret;
2591 int i;
2592 unsigned long mask;
2593
774636e1
AD		 2594 ret = kstrtouint_from_user(buf, count, 0, &val);
2595 if (ret < 0)
2596 return ret;
3cb4a0bb
KH		 2597
2598 ret = -ESRCH;
496ad9aa 2599 task = get_proc_task(file_inode(file));
3cb4a0bb
KH		 2600 if (!task)
2601 goto out_no_task;
2602
3cb4a0bb
KH		 2603 mm = get_task_mm(task);
2604 if (!mm)
2605 goto out_no_mm;
41a0c249 2606 ret = 0;
3cb4a0bb
KH		 2607
2608 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2609 if (val & mask)
2610 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2611 else
2612 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2613 }
2614
2615 mmput(mm);
2616 out_no_mm:
2617 put_task_struct(task);
2618 out_no_task:
774636e1
AD		 2619 if (ret < 0)
2620 return ret;
2621 return count;
3cb4a0bb
KH		 2622}
2623
2624static const struct file_operations proc_coredump_filter_operations = {
2625 .read = proc_coredump_filter_read,
2626 .write = proc_coredump_filter_write,
87df8424 2627 .llseek = generic_file_llseek,
3cb4a0bb
KH		 2628};
2629#endif
2630
aba76fdb 2631#ifdef CONFIG_TASK_IO_ACCOUNTING
19aadc98 2632static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
297c5d92 2633{
940389b8 2634 struct task_io_accounting acct = task->ioac;
5995477a 2635 unsigned long flags;
293eb1e7 2636 int result;
5995477a 2637
293eb1e7
VK		 2638 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2639 if (result)
2640 return result;
2641
caaee623 2642 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) {
293eb1e7
VK		 2643 result = -EACCES;
2644 goto out_unlock;
2645 }
1d1221f3 2646
5995477a
AR		 2647 if (whole && lock_task_sighand(task, &flags)) {
2648 struct task_struct *t = task;
2649
2650 task_io_accounting_add(&acct, &task->signal->ioac);
2651 while_each_thread(task, t)
2652 task_io_accounting_add(&acct, &t->ioac);
2653
2654 unlock_task_sighand(task, &flags);
297c5d92 2655 }
25ce3191
JP		 2656 seq_printf(m,
2657 "rchar: %llu\n"
2658 "wchar: %llu\n"
2659 "syscr: %llu\n"
2660 "syscw: %llu\n"
2661 "read_bytes: %llu\n"
2662 "write_bytes: %llu\n"
2663 "cancelled_write_bytes: %llu\n",
2664 (unsigned long long)acct.rchar,
2665 (unsigned long long)acct.wchar,
2666 (unsigned long long)acct.syscr,
2667 (unsigned long long)acct.syscw,
2668 (unsigned long long)acct.read_bytes,
2669 (unsigned long long)acct.write_bytes,
2670 (unsigned long long)acct.cancelled_write_bytes);
2671 result = 0;
2672
293eb1e7
VK		 2673out_unlock:
2674 mutex_unlock(&task->signal->cred_guard_mutex);
2675 return result;
297c5d92
AR		 2676}
2677
19aadc98
AD		 2678static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2679 struct pid *pid, struct task_struct *task)
297c5d92 2680{
19aadc98 2681 return do_io_accounting(task, m, 0);
aba76fdb 2682}
297c5d92 2683
19aadc98
AD		 2684static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2685 struct pid *pid, struct task_struct *task)
297c5d92 2686{
19aadc98 2687 return do_io_accounting(task, m, 1);
297c5d92
AR		 2688}
2689#endif /* CONFIG_TASK_IO_ACCOUNTING */
aba76fdb 2690
22d917d8
EB		 2691#ifdef CONFIG_USER_NS
2692static int proc_id_map_open(struct inode *inode, struct file *file,
ccf94f1b 2693 const struct seq_operations *seq_ops)
22d917d8
EB		 2694{
2695 struct user_namespace *ns = NULL;
2696 struct task_struct *task;
2697 struct seq_file *seq;
2698 int ret = -EINVAL;
2699
2700 task = get_proc_task(inode);
2701 if (task) {
2702 rcu_read_lock();
2703 ns = get_user_ns(task_cred_xxx(task, user_ns));
2704 rcu_read_unlock();
2705 put_task_struct(task);
2706 }
2707 if (!ns)
2708 goto err;
2709
2710 ret = seq_open(file, seq_ops);
2711 if (ret)
2712 goto err_put_ns;
2713
2714 seq = file->private_data;
2715 seq->private = ns;
2716
2717 return 0;
2718err_put_ns:
2719 put_user_ns(ns);
2720err:
2721 return ret;
2722}
2723
2724static int proc_id_map_release(struct inode *inode, struct file *file)
2725{
2726 struct seq_file *seq = file->private_data;
2727 struct user_namespace *ns = seq->private;
2728 put_user_ns(ns);
2729 return seq_release(inode, file);
2730}
2731
2732static int proc_uid_map_open(struct inode *inode, struct file *file)
2733{
2734 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2735}
2736
2737static int proc_gid_map_open(struct inode *inode, struct file *file)
2738{
2739 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2740}
2741
f76d207a
EB		 2742static int proc_projid_map_open(struct inode *inode, struct file *file)
2743{
2744 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2745}
2746
22d917d8
EB		 2747static const struct file_operations proc_uid_map_operations = {
2748 .open = proc_uid_map_open,
2749 .write = proc_uid_map_write,
2750 .read = seq_read,
2751 .llseek = seq_lseek,
2752 .release = proc_id_map_release,
2753};
2754
2755static const struct file_operations proc_gid_map_operations = {
2756 .open = proc_gid_map_open,
2757 .write = proc_gid_map_write,
2758 .read = seq_read,
2759 .llseek = seq_lseek,
2760 .release = proc_id_map_release,
2761};
f76d207a
EB		 2762
2763static const struct file_operations proc_projid_map_operations = {
2764 .open = proc_projid_map_open,
2765 .write = proc_projid_map_write,
2766 .read = seq_read,
2767 .llseek = seq_lseek,
2768 .release = proc_id_map_release,
2769};
9cc46516
EB		 2770
2771static int proc_setgroups_open(struct inode *inode, struct file *file)
2772{
2773 struct user_namespace *ns = NULL;
2774 struct task_struct *task;
2775 int ret;
2776
2777 ret = -ESRCH;
2778 task = get_proc_task(inode);
2779 if (task) {
2780 rcu_read_lock();
2781 ns = get_user_ns(task_cred_xxx(task, user_ns));
2782 rcu_read_unlock();
2783 put_task_struct(task);
2784 }
2785 if (!ns)
2786 goto err;
2787
2788 if (file->f_mode & FMODE_WRITE) {
2789 ret = -EACCES;
2790 if (!ns_capable(ns, CAP_SYS_ADMIN))
2791 goto err_put_ns;
2792 }
2793
2794 ret = single_open(file, &proc_setgroups_show, ns);
2795 if (ret)
2796 goto err_put_ns;
2797
2798 return 0;
2799err_put_ns:
2800 put_user_ns(ns);
2801err:
2802 return ret;
2803}
2804
2805static int proc_setgroups_release(struct inode *inode, struct file *file)
2806{
2807 struct seq_file *seq = file->private_data;
2808 struct user_namespace *ns = seq->private;
2809 int ret = single_release(inode, file);
2810 put_user_ns(ns);
2811 return ret;
2812}
2813
2814static const struct file_operations proc_setgroups_operations = {
2815 .open = proc_setgroups_open,
2816 .write = proc_setgroups_write,
2817 .read = seq_read,
2818 .llseek = seq_lseek,
2819 .release = proc_setgroups_release,
2820};
22d917d8
EB		 2821#endif /* CONFIG_USER_NS */
2822
47830723
KC		 2823static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2824 struct pid *pid, struct task_struct *task)
2825{
a9712bc1
AV		 2826 int err = lock_trace(task);
2827 if (!err) {
2828 seq_printf(m, "%08x\n", task->personality);
2829 unlock_trace(task);
2830 }
2831 return err;
47830723
KC		 2832}
2833
7c23b330
JP		 2834#ifdef CONFIG_LIVEPATCH
2835static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns,
2836 struct pid *pid, struct task_struct *task)
2837{
2838 seq_printf(m, "%d\n", task->patch_state);
2839 return 0;
2840}
2841#endif /* CONFIG_LIVEPATCH */
2842
28a6d671
EB		 2843/*
2844 * Thread groups
2845 */
00977a59 2846static const struct file_operations proc_task_operations;
c5ef1c42 2847static const struct inode_operations proc_task_inode_operations;
20cdc894 2848
c5141e6d 2849static const struct pid_entry tgid_base_stuff[] = {
631f9c18
AD		 2850 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2851 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
640708a2 2852 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
631f9c18 2853 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 2854 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
b2211a36 2855#ifdef CONFIG_NET
631f9c18 2856 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
b2211a36 2857#endif
631f9c18 2858 REG("environ", S_IRUSR, proc_environ_operations),
c5317167 2859 REG("auxv", S_IRUSR, proc_auxv_operations),
631f9c18 2860 ONE("status", S_IRUGO, proc_pid_status),
35a35046 2861 ONE("personality", S_IRUSR, proc_pid_personality),
1c963eb1 2862 ONE("limits", S_IRUGO, proc_pid_limits),
43ae34cb 2863#ifdef CONFIG_SCHED_DEBUG
631f9c18 2864 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
5091faa4
MG		 2865#endif
2866#ifdef CONFIG_SCHED_AUTOGROUP
2867 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
ebcb6734 2868#endif
4614a696 2869 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 2870#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
09d93bd6 2871 ONE("syscall", S_IRUSR, proc_pid_syscall),
43ae34cb 2872#endif
c2c0bb44 2873 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
631f9c18
AD		 2874 ONE("stat", S_IRUGO, proc_tgid_stat),
2875 ONE("statm", S_IRUGO, proc_pid_statm),
b7643757 2876 REG("maps", S_IRUGO, proc_pid_maps_operations),
28a6d671 2877#ifdef CONFIG_NUMA
b7643757 2878 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
28a6d671 2879#endif
631f9c18
AD		 2880 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2881 LNK("cwd", proc_cwd_link),
2882 LNK("root", proc_root_link),
2883 LNK("exe", proc_exe_link),
2884 REG("mounts", S_IRUGO, proc_mounts_operations),
2885 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2886 REG("mountstats", S_IRUSR, proc_mountstats_operations),
1e883281 2887#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18 2888 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
b7643757 2889 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
32ed74a4 2890 REG("pagemap", S_IRUSR, proc_pagemap_operations),
28a6d671
EB		 2891#endif
2892#ifdef CONFIG_SECURITY
631f9c18 2893 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 2894#endif
2895#ifdef CONFIG_KALLSYMS
edfcd606 2896 ONE("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 2897#endif
2ec220e2 2898#ifdef CONFIG_STACKTRACE
35a35046 2899 ONE("stack", S_IRUSR, proc_pid_stack),
28a6d671 2900#endif
5968cece 2901#ifdef CONFIG_SCHED_INFO
f6e826ca 2902 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 2903#endif
9745512c 2904#ifdef CONFIG_LATENCYTOP
631f9c18 2905 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 2906#endif
8793d854 2907#ifdef CONFIG_PROC_PID_CPUSET
52de4779 2908 ONE("cpuset", S_IRUGO, proc_cpuset_show),
a424316c
PM		 2909#endif
2910#ifdef CONFIG_CGROUPS
006f4ac4 2911 ONE("cgroup", S_IRUGO, proc_cgroup_show),
28a6d671 2912#endif
6ba51e37 2913 ONE("oom_score", S_IRUGO, proc_oom_score),
fa0cbbf1 2914 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
a63d83f4 2915 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 2916#ifdef CONFIG_AUDITSYSCALL
631f9c18
AD		 2917 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2918 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 2919#endif
f4f154fd 2920#ifdef CONFIG_FAULT_INJECTION
631f9c18 2921 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 2922#endif
698ba7b5 2923#ifdef CONFIG_ELF_CORE
631f9c18 2924 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3cb4a0bb 2925#endif
aba76fdb 2926#ifdef CONFIG_TASK_IO_ACCOUNTING
19aadc98 2927 ONE("io", S_IRUSR, proc_tgid_io_accounting),
aba76fdb 2928#endif
f133ecca 2929#ifdef CONFIG_HARDWALL
d962c144 2930 ONE("hardwall", S_IRUGO, proc_pid_hardwall),
f133ecca 2931#endif
22d917d8
EB		 2932#ifdef CONFIG_USER_NS
2933 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
2934 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
f76d207a 2935 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
9cc46516 2936 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
22d917d8 2937#endif
b18b6a9c 2938#if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
48f6a7a5
PE		 2939 REG("timers", S_IRUGO, proc_timers_operations),
2940#endif
5de23d43 2941 REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations),
7c23b330
JP		 2942#ifdef CONFIG_LIVEPATCH
2943 ONE("patch_state", S_IRUSR, proc_pid_patch_state),
2944#endif
28a6d671 2945};
1da177e4 2946
f0c3b509 2947static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
1da177e4 2948{
f0c3b509
AV		 2949 return proc_pident_readdir(file, ctx,
2950 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 2951}
2952
00977a59 2953static const struct file_operations proc_tgid_base_operations = {
1da177e4 2954 .read = generic_read_dir,
f50752ea
AV		 2955 .iterate_shared = proc_tgid_base_readdir,
2956 .llseek = generic_file_llseek,
1da177e4
LT		 2957};
2958
00cd8dd3
AV		 2959static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2960{
7bcd6b0e
EB		 2961 return proc_pident_lookup(dir, dentry,
2962 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 2963}
2964
c5ef1c42 2965static const struct inode_operations proc_tgid_base_inode_operations = {
28a6d671 2966 .lookup = proc_tgid_base_lookup,
99f89551 2967 .getattr = pid_getattr,
6d76fa58 2968 .setattr = proc_setattr,
0499680a 2969 .permission = proc_pid_permission,
1da177e4 2970};
1da177e4 2971
60347f67 2972static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
1da177e4 2973{
48e6484d 2974 struct dentry *dentry, *leader, *dir;
8578cea7 2975 char buf[PROC_NUMBUF];
48e6484d
EB		 2976 struct qstr name;
2977
2978 name.name = buf;
60347f67 2979 name.len = snprintf(buf, sizeof(buf), "%d", pid);
4f522a24 2980 /* no ->d_hash() rejects on procfs */
60347f67 2981 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d 2982 if (dentry) {
bbd51924 2983 d_invalidate(dentry);
48e6484d
EB		 2984 dput(dentry);
2985 }
1da177e4 2986
c35a7f18
ON		 2987 if (pid == tgid)
2988 return;
2989
48e6484d 2990 name.name = buf;
60347f67
PE		 2991 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2992 leader = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d
EB		 2993 if (!leader)
2994 goto out;
1da177e4 2995
48e6484d
EB		 2996 name.name = "task";
2997 name.len = strlen(name.name);
2998 dir = d_hash_and_lookup(leader, &name);
2999 if (!dir)
3000 goto out_put_leader;
3001
3002 name.name = buf;
60347f67 3003 name.len = snprintf(buf, sizeof(buf), "%d", pid);
48e6484d
EB		 3004 dentry = d_hash_and_lookup(dir, &name);
3005 if (dentry) {
bbd51924 3006 d_invalidate(dentry);
48e6484d 3007 dput(dentry);
1da177e4 3008 }
48e6484d
EB		 3009
3010 dput(dir);
3011out_put_leader:
3012 dput(leader);
3013out:
3014 return;
1da177e4
LT		 3015}
3016
0895e91d
RD		 3017/**
3018 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3019 * @task: task that should be flushed.
3020 *
3021 * When flushing dentries from proc, one needs to flush them from global
60347f67 3022 * proc (proc_mnt) and from all the namespaces' procs this task was seen
0895e91d
RD		 3023 * in. This call is supposed to do all of this job.
3024 *
3025 * Looks in the dcache for
3026 * /proc/@pid
3027 * /proc/@tgid/task/@pid
3028 * if either directory is present flushes it and all of it'ts children
3029 * from the dcache.
3030 *
3031 * It is safe and reasonable to cache /proc entries for a task until
3032 * that task exits. After that they just clog up the dcache with
3033 * useless entries, possibly causing useful dcache entries to be
3034 * flushed instead. This routine is proved to flush those useless
3035 * dcache entries at process exit time.
3036 *
3037 * NOTE: This routine is just an optimization so it does not guarantee
3038 * that no dcache entries will exist at process exit time it
3039 * just makes it very unlikely that any will persist.
60347f67
PE		 3040 */
3041
3042void proc_flush_task(struct task_struct *task)
3043{
9fcc2d15 3044 int i;
9b4d1cbe 3045 struct pid *pid, *tgid;
130f77ec
PE		 3046 struct upid *upid;
3047
130f77ec 3048 pid = task_pid(task);
9b4d1cbe 3049 tgid = task_tgid(task);
130f77ec 3050
9fcc2d15 3051 for (i = 0; i <= pid->level; i++) {
130f77ec
PE		 3052 upid = &pid->numbers[i];
3053 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
9b4d1cbe 3054 tgid->numbers[i].nr);
130f77ec 3055 }
60347f67
PE		 3056}
3057
c52a47ac
AV		 3058static int proc_pid_instantiate(struct inode *dir,
3059 struct dentry * dentry,
3060 struct task_struct *task, const void *ptr)
444ceed8 3061{
444ceed8
EB		 3062 struct inode *inode;
3063
db978da8 3064 inode = proc_pid_make_inode(dir->i_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
444ceed8
EB		 3065 if (!inode)
3066 goto out;
3067
444ceed8
EB		 3068 inode->i_op = &proc_tgid_base_inode_operations;
3069 inode->i_fop = &proc_tgid_base_operations;
3070 inode->i_flags|=S_IMMUTABLE;
aed54175 3071
1270dd8d 3072 set_nlink(inode, nlink_tgid);
444ceed8 3073
fb045adb 3074 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 3075
3076 d_add(dentry, inode);
3077 /* Close the race of the process dying before we return the dentry */
0b728e19 3078 if (pid_revalidate(dentry, 0))
c52a47ac 3079 return 0;
444ceed8 3080out:
c52a47ac 3081 return -ENOENT;
444ceed8
EB		 3082}
3083
00cd8dd3 3084struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
1da177e4 3085{
335eb531 3086 int result = -ENOENT;
1da177e4 3087 struct task_struct *task;
1da177e4 3088 unsigned tgid;
b488893a 3089 struct pid_namespace *ns;
1da177e4 3090
dbcdb504 3091 tgid = name_to_int(&dentry->d_name);
1da177e4
LT		 3092 if (tgid == ~0U)
3093 goto out;
3094
b488893a 3095 ns = dentry->d_sb->s_fs_info;
de758734 3096 rcu_read_lock();
b488893a 3097 task = find_task_by_pid_ns(tgid, ns);
1da177e4
LT		 3098 if (task)
3099 get_task_struct(task);
de758734 3100 rcu_read_unlock();
1da177e4
LT		 3101 if (!task)
3102 goto out;
3103
444ceed8 3104 result = proc_pid_instantiate(dir, dentry, task, NULL);
1da177e4 3105 put_task_struct(task);
1da177e4 3106out:
c52a47ac 3107 return ERR_PTR(result);
1da177e4
LT		 3108}
3109
1da177e4 3110/*
0804ef4b 3111 * Find the first task with tgid >= tgid
0bc58a91 3112 *
1da177e4 3113 */
19fd4bb2
EB		 3114struct tgid_iter {
3115 unsigned int tgid;
0804ef4b 3116 struct task_struct *task;
19fd4bb2
EB		 3117};
3118static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3119{
0804ef4b 3120 struct pid *pid;
1da177e4 3121
19fd4bb2
EB		 3122 if (iter.task)
3123 put_task_struct(iter.task);
454cc105 3124 rcu_read_lock();
0804ef4b 3125retry:
19fd4bb2
EB		 3126 iter.task = NULL;
3127 pid = find_ge_pid(iter.tgid, ns);
0804ef4b 3128 if (pid) {
19fd4bb2
EB		 3129 iter.tgid = pid_nr_ns(pid, ns);
3130 iter.task = pid_task(pid, PIDTYPE_PID);
0804ef4b
EB		 3131 /* What we to know is if the pid we have find is the
3132 * pid of a thread_group_leader. Testing for task
3133 * being a thread_group_leader is the obvious thing
3134 * todo but there is a window when it fails, due to
3135 * the pid transfer logic in de_thread.
3136 *
3137 * So we perform the straight forward test of seeing
3138 * if the pid we have found is the pid of a thread
3139 * group leader, and don't worry if the task we have
3140 * found doesn't happen to be a thread group leader.
3141 * As we don't care in the case of readdir.
3142 */
19fd4bb2
EB		 3143 if (!iter.task || !has_group_leader_pid(iter.task)) {
3144 iter.tgid += 1;
0804ef4b 3145 goto retry;
19fd4bb2
EB		 3146 }
3147 get_task_struct(iter.task);
0bc58a91 3148 }
454cc105 3149 rcu_read_unlock();
19fd4bb2 3150 return iter;
1da177e4
LT		 3151}
3152
0097875b 3153#define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
0804ef4b 3154
1da177e4 3155/* for the /proc/ directory itself, after non-process stuff has been done */
f0c3b509 3156int proc_pid_readdir(struct file *file, struct dir_context *ctx)
1da177e4 3157{
19fd4bb2 3158 struct tgid_iter iter;
3aa3377f 3159 struct pid_namespace *ns = file_inode(file)->i_sb->s_fs_info;
f0c3b509 3160 loff_t pos = ctx->pos;
1da177e4 3161
021ada7d 3162 if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
f0c3b509 3163 return 0;
1da177e4 3164
0097875b 3165 if (pos == TGID_OFFSET - 2) {
2b0143b5 3166 struct inode *inode = d_inode(ns->proc_self);
db963164 3167 if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
f0c3b509 3168 return 0;
0097875b
EB		 3169 ctx->pos = pos = pos + 1;
3170 }
3171 if (pos == TGID_OFFSET - 1) {
2b0143b5 3172 struct inode *inode = d_inode(ns->proc_thread_self);
0097875b
EB		 3173 if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3174 return 0;
3175 ctx->pos = pos = pos + 1;
021ada7d 3176 }
0097875b 3177 iter.tgid = pos - TGID_OFFSET;
19fd4bb2 3178 iter.task = NULL;
19fd4bb2
EB		 3179 for (iter = next_tgid(ns, iter);
3180 iter.task;
3181 iter.tgid += 1, iter = next_tgid(ns, iter)) {
f0c3b509
AV		 3182 char name[PROC_NUMBUF];
3183 int len;
3ba4bcee
ED		 3184
3185 cond_resched();
796f571b 3186 if (!has_pid_permissions(ns, iter.task, HIDEPID_INVISIBLE))
f0c3b509 3187 continue;
0499680a 3188
f0c3b509
AV		 3189 len = snprintf(name, sizeof(name), "%d", iter.tgid);
3190 ctx->pos = iter.tgid + TGID_OFFSET;
3191 if (!proc_fill_cache(file, ctx, name, len,
3192 proc_pid_instantiate, iter.task, NULL)) {
19fd4bb2 3193 put_task_struct(iter.task);
f0c3b509 3194 return 0;
1da177e4 3195 }
0bc58a91 3196 }
f0c3b509 3197 ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
0bc58a91
EB		 3198 return 0;
3199}
1da177e4 3200
1b3044e3
JD		 3201/*
3202 * proc_tid_comm_permission is a special permission function exclusively
3203 * used for the node /proc/<pid>/task/<tid>/comm.
3204 * It bypasses generic permission checks in the case where a task of the same
3205 * task group attempts to access the node.
3206 * The rationale behind this is that glibc and bionic access this node for
3207 * cross thread naming (pthread_set/getname_np(!self)). However, if
3208 * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3209 * which locks out the cross thread naming implementation.
3210 * This function makes sure that the node is always accessible for members of
3211 * same thread group.
3212 */
3213static int proc_tid_comm_permission(struct inode *inode, int mask)
3214{
3215 bool is_same_tgroup;
3216 struct task_struct *task;
3217
3218 task = get_proc_task(inode);
3219 if (!task)
3220 return -ESRCH;
3221 is_same_tgroup = same_thread_group(current, task);
3222 put_task_struct(task);
3223
3224 if (likely(is_same_tgroup && !(mask & MAY_EXEC))) {
3225 /* This file (/proc/<pid>/task/<tid>/comm) can always be
3226 * read or written by the members of the corresponding
3227 * thread group.
3228 */
3229 return 0;
3230 }
3231
3232 return generic_permission(inode, mask);
3233}
3234
3235static const struct inode_operations proc_tid_comm_inode_operations = {
3236 .permission = proc_tid_comm_permission,
3237};
3238
28a6d671
EB		 3239/*
3240 * Tasks
3241 */
c5141e6d 3242static const struct pid_entry tid_base_stuff[] = {
631f9c18 3243 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3835541d 3244 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 3245 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
6ba8ed79
EB		 3246#ifdef CONFIG_NET
3247 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3248#endif
631f9c18 3249 REG("environ", S_IRUSR, proc_environ_operations),
c5317167 3250 REG("auxv", S_IRUSR, proc_auxv_operations),
631f9c18 3251 ONE("status", S_IRUGO, proc_pid_status),
35a35046 3252 ONE("personality", S_IRUSR, proc_pid_personality),
1c963eb1 3253 ONE("limits", S_IRUGO, proc_pid_limits),
43ae34cb 3254#ifdef CONFIG_SCHED_DEBUG
631f9c18 3255 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
ebcb6734 3256#endif
1b3044e3
JD		 3257 NOD("comm", S_IFREG|S_IRUGO|S_IWUSR,
3258 &proc_tid_comm_inode_operations,
3259 &proc_pid_set_comm_operations, {}),
ebcb6734 3260#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
09d93bd6 3261 ONE("syscall", S_IRUSR, proc_pid_syscall),
43ae34cb 3262#endif
c2c0bb44 3263 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
631f9c18
AD		 3264 ONE("stat", S_IRUGO, proc_tid_stat),
3265 ONE("statm", S_IRUGO, proc_pid_statm),
b7643757 3266 REG("maps", S_IRUGO, proc_tid_maps_operations),
2e13ba54 3267#ifdef CONFIG_PROC_CHILDREN
81841161
CG		 3268 REG("children", S_IRUGO, proc_tid_children_operations),
3269#endif
28a6d671 3270#ifdef CONFIG_NUMA
b7643757 3271 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
28a6d671 3272#endif
631f9c18
AD		 3273 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3274 LNK("cwd", proc_cwd_link),
3275 LNK("root", proc_root_link),
3276 LNK("exe", proc_exe_link),
3277 REG("mounts", S_IRUGO, proc_mounts_operations),
3278 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
1e883281 3279#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18 3280 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
b7643757 3281 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
32ed74a4 3282 REG("pagemap", S_IRUSR, proc_pagemap_operations),
28a6d671
EB		 3283#endif
3284#ifdef CONFIG_SECURITY
631f9c18 3285 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 3286#endif
3287#ifdef CONFIG_KALLSYMS
edfcd606 3288 ONE("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 3289#endif
2ec220e2 3290#ifdef CONFIG_STACKTRACE
35a35046 3291 ONE("stack", S_IRUSR, proc_pid_stack),
28a6d671 3292#endif
5968cece 3293#ifdef CONFIG_SCHED_INFO
f6e826ca 3294 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 3295#endif
9745512c 3296#ifdef CONFIG_LATENCYTOP
631f9c18 3297 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 3298#endif
8793d854 3299#ifdef CONFIG_PROC_PID_CPUSET
52de4779 3300 ONE("cpuset", S_IRUGO, proc_cpuset_show),
a424316c
PM		 3301#endif
3302#ifdef CONFIG_CGROUPS
006f4ac4 3303 ONE("cgroup", S_IRUGO, proc_cgroup_show),
28a6d671 3304#endif
6ba51e37 3305 ONE("oom_score", S_IRUGO, proc_oom_score),
fa0cbbf1 3306 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
a63d83f4 3307 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 3308#ifdef CONFIG_AUDITSYSCALL
631f9c18 3309 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
26ec3c64 3310 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 3311#endif
f4f154fd 3312#ifdef CONFIG_FAULT_INJECTION
631f9c18 3313 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 3314#endif
297c5d92 3315#ifdef CONFIG_TASK_IO_ACCOUNTING
19aadc98 3316 ONE("io", S_IRUSR, proc_tid_io_accounting),
297c5d92 3317#endif
f133ecca 3318#ifdef CONFIG_HARDWALL
d962c144 3319 ONE("hardwall", S_IRUGO, proc_pid_hardwall),
f133ecca 3320#endif
22d917d8
EB		 3321#ifdef CONFIG_USER_NS
3322 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3323 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
f76d207a 3324 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
9cc46516 3325 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
22d917d8 3326#endif
7c23b330
JP		 3327#ifdef CONFIG_LIVEPATCH
3328 ONE("patch_state", S_IRUSR, proc_pid_patch_state),
3329#endif
28a6d671
EB		 3330};
3331
f0c3b509 3332static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
28a6d671 3333{
f0c3b509
AV		 3334 return proc_pident_readdir(file, ctx,
3335 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
28a6d671
EB		 3336}
3337
00cd8dd3
AV		 3338static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3339{
7bcd6b0e
EB		 3340 return proc_pident_lookup(dir, dentry,
3341 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
28a6d671
EB		 3342}
3343
00977a59 3344static const struct file_operations proc_tid_base_operations = {
28a6d671 3345 .read = generic_read_dir,
f50752ea
AV		 3346 .iterate_shared = proc_tid_base_readdir,
3347 .llseek = generic_file_llseek,
28a6d671
EB		 3348};
3349
c5ef1c42 3350static const struct inode_operations proc_tid_base_inode_operations = {
28a6d671
EB		 3351 .lookup = proc_tid_base_lookup,
3352 .getattr = pid_getattr,
3353 .setattr = proc_setattr,
3354};
3355
c52a47ac 3356static int proc_task_instantiate(struct inode *dir,
c5141e6d 3357 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8 3358{
444ceed8 3359 struct inode *inode;
db978da8 3360 inode = proc_pid_make_inode(dir->i_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
444ceed8
EB		 3361
3362 if (!inode)
3363 goto out;
444ceed8
EB		 3364 inode->i_op = &proc_tid_base_inode_operations;
3365 inode->i_fop = &proc_tid_base_operations;
3366 inode->i_flags|=S_IMMUTABLE;
aed54175 3367
1270dd8d 3368 set_nlink(inode, nlink_tid);
444ceed8 3369
fb045adb 3370 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 3371
3372 d_add(dentry, inode);
3373 /* Close the race of the process dying before we return the dentry */
0b728e19 3374 if (pid_revalidate(dentry, 0))
c52a47ac 3375 return 0;
444ceed8 3376out:
c52a47ac 3377 return -ENOENT;
444ceed8
EB		 3378}
3379
00cd8dd3 3380static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
28a6d671 3381{
c52a47ac 3382 int result = -ENOENT;
28a6d671
EB		 3383 struct task_struct *task;
3384 struct task_struct *leader = get_proc_task(dir);
28a6d671 3385 unsigned tid;
b488893a 3386 struct pid_namespace *ns;
28a6d671
EB		 3387
3388 if (!leader)
3389 goto out_no_task;
3390
dbcdb504 3391 tid = name_to_int(&dentry->d_name);
28a6d671
EB		 3392 if (tid == ~0U)
3393 goto out;
3394
b488893a 3395 ns = dentry->d_sb->s_fs_info;
28a6d671 3396 rcu_read_lock();
b488893a 3397 task = find_task_by_pid_ns(tid, ns);
28a6d671
EB		 3398 if (task)
3399 get_task_struct(task);
3400 rcu_read_unlock();
3401 if (!task)
3402 goto out;
bac0abd6 3403 if (!same_thread_group(leader, task))
28a6d671
EB		 3404 goto out_drop_task;
3405
444ceed8 3406 result = proc_task_instantiate(dir, dentry, task, NULL);
28a6d671
EB		 3407out_drop_task:
3408 put_task_struct(task);
3409out:
3410 put_task_struct(leader);
3411out_no_task:
c52a47ac 3412 return ERR_PTR(result);
28a6d671
EB		 3413}
3414
0bc58a91
EB		 3415/*
3416 * Find the first tid of a thread group to return to user space.
3417 *
3418 * Usually this is just the thread group leader, but if the users
3419 * buffer was too small or there was a seek into the middle of the
3420 * directory we have more work todo.
3421 *
3422 * In the case of a short read we start with find_task_by_pid.
3423 *
3424 * In the case of a seek we start with the leader and walk nr
3425 * threads past it.
3426 */
9f6e963f
ON		 3427static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3428 struct pid_namespace *ns)
0bc58a91 3429{
d855a4b7 3430 struct task_struct *pos, *task;
9f6e963f
ON		 3431 unsigned long nr = f_pos;
3432
3433 if (nr != f_pos) /* 32bit overflow? */
3434 return NULL;
1da177e4 3435
cc288738 3436 rcu_read_lock();
d855a4b7
ON		 3437 task = pid_task(pid, PIDTYPE_PID);
3438 if (!task)
3439 goto fail;
3440
3441 /* Attempt to start with the tid of a thread */
9f6e963f 3442 if (tid && nr) {
b488893a 3443 pos = find_task_by_pid_ns(tid, ns);
d855a4b7 3444 if (pos && same_thread_group(pos, task))
a872ff0c 3445 goto found;
0bc58a91 3446 }
1da177e4 3447
0bc58a91 3448 /* If nr exceeds the number of threads there is nothing todo */
9f6e963f 3449 if (nr >= get_nr_threads(task))
c986c14a 3450 goto fail;
1da177e4 3451
a872ff0c
ON		 3452 /* If we haven't found our starting place yet start
3453 * with the leader and walk nr threads forward.
0bc58a91 3454 */
d855a4b7 3455 pos = task = task->group_leader;
c986c14a 3456 do {
9f6e963f 3457 if (!nr--)
c986c14a 3458 goto found;
d855a4b7 3459 } while_each_thread(task, pos);
c986c14a
ON		 3460fail:
3461 pos = NULL;
3462 goto out;
a872ff0c
ON		 3463found:
3464 get_task_struct(pos);
3465out:
cc288738 3466 rcu_read_unlock();
0bc58a91
EB		 3467 return pos;
3468}
3469
3470/*
3471 * Find the next thread in the thread list.
3472 * Return NULL if there is an error or no next thread.
3473 *
3474 * The reference to the input task_struct is released.
3475 */
3476static struct task_struct *next_tid(struct task_struct *start)
3477{
c1df7fb8 3478 struct task_struct *pos = NULL;
cc288738 3479 rcu_read_lock();
c1df7fb8 3480 if (pid_alive(start)) {
0bc58a91 3481 pos = next_thread(start);
c1df7fb8
ON		 3482 if (thread_group_leader(pos))
3483 pos = NULL;
3484 else
3485 get_task_struct(pos);
3486 }
cc288738 3487 rcu_read_unlock();
0bc58a91
EB		 3488 put_task_struct(start);
3489 return pos;
1da177e4
LT		 3490}
3491
3492/* for the /proc/TGID/task/ directories */
f0c3b509 3493static int proc_task_readdir(struct file *file, struct dir_context *ctx)
1da177e4 3494{
d855a4b7
ON		 3495 struct inode *inode = file_inode(file);
3496 struct task_struct *task;
b488893a 3497 struct pid_namespace *ns;
f0c3b509 3498 int tid;
1da177e4 3499
d855a4b7 3500 if (proc_inode_is_dead(inode))
f0c3b509 3501 return -ENOENT;
1da177e4 3502
f0c3b509 3503 if (!dir_emit_dots(file, ctx))
d855a4b7 3504 return 0;
1da177e4 3505
0bc58a91
EB		 3506 /* f_version caches the tgid value that the last readdir call couldn't
3507 * return. lseek aka telldir automagically resets f_version to 0.
3508 */
3aa3377f 3509 ns = inode->i_sb->s_fs_info;
f0c3b509
AV		 3510 tid = (int)file->f_version;
3511 file->f_version = 0;
d855a4b7 3512 for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
0bc58a91 3513 task;
f0c3b509
AV		 3514 task = next_tid(task), ctx->pos++) {
3515 char name[PROC_NUMBUF];
3516 int len;
b488893a 3517 tid = task_pid_nr_ns(task, ns);
f0c3b509
AV		 3518 len = snprintf(name, sizeof(name), "%d", tid);
3519 if (!proc_fill_cache(file, ctx, name, len,
3520 proc_task_instantiate, task, NULL)) {
0bc58a91
EB		 3521 /* returning this tgid failed, save it as the first
3522 * pid for the next readir call */
f0c3b509 3523 file->f_version = (u64)tid;
0bc58a91 3524 put_task_struct(task);
1da177e4 3525 break;
0bc58a91 3526 }
1da177e4 3527 }
d855a4b7 3528
f0c3b509 3529 return 0;
1da177e4 3530}
6e66b52b 3531
a528d35e
DH		 3532static int proc_task_getattr(const struct path *path, struct kstat *stat,
3533 u32 request_mask, unsigned int query_flags)
6e66b52b 3534{
a528d35e 3535 struct inode *inode = d_inode(path->dentry);
99f89551 3536 struct task_struct *p = get_proc_task(inode);
6e66b52b
EB		 3537 generic_fillattr(inode, stat);
3538
99f89551 3539 if (p) {
99f89551 3540 stat->nlink += get_nr_threads(p);
99f89551 3541 put_task_struct(p);
6e66b52b
EB		 3542 }
3543
3544 return 0;
3545}
28a6d671 3546
c5ef1c42 3547static const struct inode_operations proc_task_inode_operations = {
28a6d671
EB		 3548 .lookup = proc_task_lookup,
3549 .getattr = proc_task_getattr,
3550 .setattr = proc_setattr,
0499680a 3551 .permission = proc_pid_permission,
28a6d671
EB		 3552};
3553
00977a59 3554static const struct file_operations proc_task_operations = {
28a6d671 3555 .read = generic_read_dir,
f50752ea
AV		 3556 .iterate_shared = proc_task_readdir,
3557 .llseek = generic_file_llseek,
28a6d671 3558};
1270dd8d
AD		 3559
3560void __init set_proc_pid_nlink(void)
3561{
3562 nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3563 nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3564}
Linux 6.x block layer and io_uring tree(s)