]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - fs/proc/base.c
projects / mirror_ubuntu-bionic-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
mm: simplify find_vma_prev()
[mirror_ubuntu-bionic-kernel.git] / fs / proc / base.c
CommitLineData
1da177e4
LT		 1/*
2 * linux/fs/proc/base.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 *
6 * proc base directory handling functions
7 *
8 * 1999, Al Viro. Rewritten. Now it covers the whole per-process part.
9 * Instead of using magical inumbers to determine the kind of object
10 * we allocate and fill in-core inodes upon lookup. They don't even
11 * go into icache. We cache the reference to task_struct upon lookup too.
12 * Eventually it should become a filesystem in its own. We don't use the
13 * rest of procfs anymore.
e070ad49
ML		 14 *
15 *
16 * Changelog:
17 * 17-Jan-2005
18 * Allan Bezerra
19 * Bruna Moreira <bruna.moreira@indt.org.br>
20 * Edjard Mota <edjard.mota@indt.org.br>
21 * Ilias Biris <ilias.biris@indt.org.br>
22 * Mauricio Lin <mauricio.lin@indt.org.br>
23 *
24 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
25 *
26 * A new process specific entry (smaps) included in /proc. It shows the
27 * size of rss for each memory area. The maps entry lacks information
28 * about physical memory size (rss) for each mapped file, i.e.,
29 * rss information for executables and library files.
30 * This additional information is useful for any tools that need to know
31 * about physical memory consumption for a process specific library.
32 *
33 * Changelog:
34 * 21-Feb-2005
35 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
36 * Pud inclusion in the page table walking.
37 *
38 * ChangeLog:
39 * 10-Mar-2005
40 * 10LE Instituto Nokia de Tecnologia - INdT:
41 * A better way to walks through the page table as suggested by Hugh Dickins.
42 *
43 * Simo Piiroinen <simo.piiroinen@nokia.com>:
44 * Smaps information related to shared, private, clean and dirty pages.
45 *
46 * Paul Mundt <paul.mundt@nokia.com>:
47 * Overall revision about smaps.
1da177e4
LT		 48 */
49
50#include <asm/uaccess.h>
51
1da177e4
LT		 52#include <linux/errno.h>
53#include <linux/time.h>
54#include <linux/proc_fs.h>
55#include <linux/stat.h>
5995477a 56#include <linux/task_io_accounting_ops.h>
1da177e4 57#include <linux/init.h>
16f7e0fe 58#include <linux/capability.h>
1da177e4 59#include <linux/file.h>
9f3acc31 60#include <linux/fdtable.h>
1da177e4
LT		 61#include <linux/string.h>
62#include <linux/seq_file.h>
63#include <linux/namei.h>
6b3286ed 64#include <linux/mnt_namespace.h>
1da177e4 65#include <linux/mm.h>
a63d83f4 66#include <linux/swap.h>
b835996f 67#include <linux/rcupdate.h>
1da177e4 68#include <linux/kallsyms.h>
2ec220e2 69#include <linux/stacktrace.h>
d85f50d5 70#include <linux/resource.h>
5096add8 71#include <linux/module.h>
1da177e4
LT		 72#include <linux/mount.h>
73#include <linux/security.h>
74#include <linux/ptrace.h>
0d094efe 75#include <linux/tracehook.h>
a424316c 76#include <linux/cgroup.h>
1da177e4
LT		 77#include <linux/cpuset.h>
78#include <linux/audit.h>
5addc5dd 79#include <linux/poll.h>
1651e14e 80#include <linux/nsproxy.h>
8ac773b4 81#include <linux/oom.h>
3cb4a0bb 82#include <linux/elf.h>
60347f67 83#include <linux/pid_namespace.h>
5ad4e53b 84#include <linux/fs_struct.h>
5a0e3ad6 85#include <linux/slab.h>
f133ecca
CM		 86#ifdef CONFIG_HARDWALL
87#include <asm/hardwall.h>
88#endif
1da177e4
LT		 89#include "internal.h"
90
0f2fe20f
EB		 91/* NOTE:
92 * Implementing inode permission operations in /proc is almost
93 * certainly an error. Permission checks need to happen during
94 * each system call not at open time. The reason is that most of
95 * what we wish to check for permissions in /proc varies at runtime.
96 *
97 * The classic example of a problem is opening file descriptors
98 * in /proc for a task before it execs a suid executable.
99 */
100
1da177e4 101struct pid_entry {
1da177e4 102 char *name;
c5141e6d 103 int len;
d161a13f 104 umode_t mode;
c5ef1c42 105 const struct inode_operations *iop;
00977a59 106 const struct file_operations *fop;
20cdc894 107 union proc_op op;
1da177e4
LT		 108};
109
61a28784 110#define NOD(NAME, MODE, IOP, FOP, OP) { \
20cdc894 111 .name = (NAME), \
c5141e6d 112 .len = sizeof(NAME) - 1, \
20cdc894
EB		 113 .mode = MODE, \
114 .iop = IOP, \
115 .fop = FOP, \
116 .op = OP, \
117}
118
631f9c18
AD		 119#define DIR(NAME, MODE, iops, fops) \
120 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
121#define LNK(NAME, get_link) \
61a28784 122 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
20cdc894 123 &proc_pid_link_inode_operations, NULL, \
631f9c18
AD		 124 { .proc_get_link = get_link } )
125#define REG(NAME, MODE, fops) \
126 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
127#define INF(NAME, MODE, read) \
61a28784 128 NOD(NAME, (S_IFREG|(MODE)), \
20cdc894 129 NULL, &proc_info_file_operations, \
631f9c18
AD		 130 { .proc_read = read } )
131#define ONE(NAME, MODE, show) \
be614086
EB		 132 NOD(NAME, (S_IFREG|(MODE)), \
133 NULL, &proc_single_file_operations, \
631f9c18 134 { .proc_show = show } )
1da177e4 135
aed54175
VN		 136/*
137 * Count the number of hardlinks for the pid_entry table, excluding the .
138 * and .. links.
139 */
140static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
141 unsigned int n)
142{
143 unsigned int i;
144 unsigned int count;
145
146 count = 0;
147 for (i = 0; i < n; ++i) {
148 if (S_ISDIR(entries[i].mode))
149 ++count;
150 }
151
152 return count;
153}
154
f7ad3c6b 155static int get_task_root(struct task_struct *task, struct path *root)
1da177e4 156{
7c2c7d99
HD		 157 int result = -ENOENT;
158
0494f6ec 159 task_lock(task);
f7ad3c6b
MS		 160 if (task->fs) {
161 get_fs_root(task->fs, root);
7c2c7d99
HD		 162 result = 0;
163 }
0494f6ec 164 task_unlock(task);
7c2c7d99 165 return result;
0494f6ec
MS		 166}
167
3dcd25f3 168static int proc_cwd_link(struct inode *inode, struct path *path)
0494f6ec 169{
99f89551 170 struct task_struct *task = get_proc_task(inode);
0494f6ec 171 int result = -ENOENT;
99f89551
EB		 172
173 if (task) {
f7ad3c6b
MS		 174 task_lock(task);
175 if (task->fs) {
176 get_fs_pwd(task->fs, path);
177 result = 0;
178 }
179 task_unlock(task);
99f89551
EB		 180 put_task_struct(task);
181 }
1da177e4
LT		 182 return result;
183}
184
3dcd25f3 185static int proc_root_link(struct inode *inode, struct path *path)
1da177e4 186{
99f89551 187 struct task_struct *task = get_proc_task(inode);
1da177e4 188 int result = -ENOENT;
99f89551
EB		 189
190 if (task) {
f7ad3c6b 191 result = get_task_root(task, path);
99f89551
EB		 192 put_task_struct(task);
193 }
1da177e4
LT		 194 return result;
195}
196
8b0db9db 197static struct mm_struct *__check_mem_permission(struct task_struct *task)
638fa202 198{
8b0db9db
SW		 199 struct mm_struct *mm;
200
201 mm = get_task_mm(task);
202 if (!mm)
203 return ERR_PTR(-EINVAL);
204
638fa202
RM		 205 /*
206 * A task can always look at itself, in case it chooses
207 * to use system calls instead of load instructions.
208 */
209 if (task == current)
8b0db9db 210 return mm;
638fa202
RM		 211
212 /*
213 * If current is actively ptrace'ing, and would also be
214 * permitted to freshly attach with ptrace now, permit it.
215 */
0d094efe
RM		 216 if (task_is_stopped_or_traced(task)) {
217 int match;
218 rcu_read_lock();
06d98473 219 match = (ptrace_parent(task) == current);
0d094efe
RM		 220 rcu_read_unlock();
221 if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH))
8b0db9db 222 return mm;
0d094efe 223 }
638fa202
RM		 224
225 /*
25985edc 226 * No one else is allowed.
638fa202 227 */
8b0db9db
SW		 228 mmput(mm);
229 return ERR_PTR(-EPERM);
638fa202 230}
1da177e4 231
18f661bc 232/*
8b0db9db
SW		 233 * If current may access user memory in @task return a reference to the
234 * corresponding mm, otherwise ERR_PTR.
18f661bc 235 */
8b0db9db 236static struct mm_struct *check_mem_permission(struct task_struct *task)
18f661bc 237{
8b0db9db 238 struct mm_struct *mm;
18f661bc
SW		 239 int err;
240
241 /*
242 * Avoid racing if task exec's as we might get a new mm but validate
243 * against old credentials.
244 */
245 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
246 if (err)
8b0db9db 247 return ERR_PTR(err);
18f661bc 248
8b0db9db 249 mm = __check_mem_permission(task);
18f661bc
SW		 250 mutex_unlock(&task->signal->cred_guard_mutex);
251
8b0db9db 252 return mm;
638fa202 253}
1da177e4 254
831830b5
AV		 255struct mm_struct *mm_for_maps(struct task_struct *task)
256{
704b836c 257 struct mm_struct *mm;
ec6fd8a4 258 int err;
704b836c 259
ec6fd8a4
AV		 260 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
261 if (err)
262 return ERR_PTR(err);
00f89d21 263
704b836c
ON		 264 mm = get_task_mm(task);
265 if (mm && mm != current->mm &&
266 !ptrace_may_access(task, PTRACE_MODE_READ)) {
267 mmput(mm);
ec6fd8a4 268 mm = ERR_PTR(-EACCES);
13f0feaf 269 }
9b1bf12d 270 mutex_unlock(&task->signal->cred_guard_mutex);
704b836c 271
831830b5 272 return mm;
831830b5
AV		 273}
274
1da177e4
LT		 275static int proc_pid_cmdline(struct task_struct *task, char * buffer)
276{
277 int res = 0;
278 unsigned int len;
279 struct mm_struct *mm = get_task_mm(task);
280 if (!mm)
281 goto out;
282 if (!mm->arg_end)
283 goto out_mm; /* Shh! No looking before we're done */
284
285 len = mm->arg_end - mm->arg_start;
286
287 if (len > PAGE_SIZE)
288 len = PAGE_SIZE;
289
290 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
291
292 // If the nul at the end of args has been overwritten, then
293 // assume application is using setproctitle(3).
294 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
295 len = strnlen(buffer, res);
296 if (len < res) {
297 res = len;
298 } else {
299 len = mm->env_end - mm->env_start;
300 if (len > PAGE_SIZE - res)
301 len = PAGE_SIZE - res;
302 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
303 res = strnlen(buffer, res);
304 }
305 }
306out_mm:
307 mmput(mm);
308out:
309 return res;
310}
311
312static int proc_pid_auxv(struct task_struct *task, char *buffer)
313{
2fadaef4
AV		 314 struct mm_struct *mm = mm_for_maps(task);
315 int res = PTR_ERR(mm);
316 if (mm && !IS_ERR(mm)) {
1da177e4 317 unsigned int nwords = 0;
dfe6b7d9 318 do {
1da177e4 319 nwords += 2;
dfe6b7d9 320 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
1da177e4
LT		 321 res = nwords * sizeof(mm->saved_auxv[0]);
322 if (res > PAGE_SIZE)
323 res = PAGE_SIZE;
324 memcpy(buffer, mm->saved_auxv, res);
325 mmput(mm);
326 }
327 return res;
328}
329
330
331#ifdef CONFIG_KALLSYMS
332/*
333 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
334 * Returns the resolved symbol. If that fails, simply return the address.
335 */
336static int proc_pid_wchan(struct task_struct *task, char *buffer)
337{
ffb45122 338 unsigned long wchan;
9281acea 339 char symname[KSYM_NAME_LEN];
1da177e4
LT		 340
341 wchan = get_wchan(task);
342
9d65cb4a 343 if (lookup_symbol_name(wchan, symname) < 0)
f83ce3e6
JE		 344 if (!ptrace_may_access(task, PTRACE_MODE_READ))
345 return 0;
346 else
347 return sprintf(buffer, "%lu", wchan);
9d65cb4a
AD		 348 else
349 return sprintf(buffer, "%s", symname);
1da177e4
LT		 350}
351#endif /* CONFIG_KALLSYMS */
352
a9712bc1
AV		 353static int lock_trace(struct task_struct *task)
354{
355 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
356 if (err)
357 return err;
358 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
359 mutex_unlock(&task->signal->cred_guard_mutex);
360 return -EPERM;
361 }
362 return 0;
363}
364
365static void unlock_trace(struct task_struct *task)
366{
367 mutex_unlock(&task->signal->cred_guard_mutex);
368}
369
2ec220e2
KC		 370#ifdef CONFIG_STACKTRACE
371
372#define MAX_STACK_TRACE_DEPTH 64
373
374static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
375 struct pid *pid, struct task_struct *task)
376{
377 struct stack_trace trace;
378 unsigned long *entries;
a9712bc1 379 int err;
2ec220e2
KC		 380 int i;
381
382 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
383 if (!entries)
384 return -ENOMEM;
385
386 trace.nr_entries = 0;
387 trace.max_entries = MAX_STACK_TRACE_DEPTH;
388 trace.entries = entries;
389 trace.skip = 0;
2ec220e2 390
a9712bc1
AV		 391 err = lock_trace(task);
392 if (!err) {
393 save_stack_trace_tsk(task, &trace);
394
395 for (i = 0; i < trace.nr_entries; i++) {
b81a618d 396 seq_printf(m, "[<%pK>] %pS\n",
a9712bc1
AV		 397 (void *)entries[i], (void *)entries[i]);
398 }
399 unlock_trace(task);
2ec220e2
KC		 400 }
401 kfree(entries);
402
a9712bc1 403 return err;
2ec220e2
KC		 404}
405#endif
406
1da177e4
LT		 407#ifdef CONFIG_SCHEDSTATS
408/*
409 * Provides /proc/PID/schedstat
410 */
411static int proc_pid_schedstat(struct task_struct *task, char *buffer)
412{
172ba844 413 return sprintf(buffer, "%llu %llu %lu\n",
826e08b0
IM		 414 (unsigned long long)task->se.sum_exec_runtime,
415 (unsigned long long)task->sched_info.run_delay,
2d72376b 416 task->sched_info.pcount);
1da177e4
LT		 417}
418#endif
419
9745512c
AV		 420#ifdef CONFIG_LATENCYTOP
421static int lstats_show_proc(struct seq_file *m, void *v)
422{
423 int i;
13d77c37
HS		 424 struct inode *inode = m->private;
425 struct task_struct *task = get_proc_task(inode);
9745512c 426
13d77c37
HS		 427 if (!task)
428 return -ESRCH;
429 seq_puts(m, "Latency Top version : v0.1\n");
9745512c 430 for (i = 0; i < 32; i++) {
34e49d4f
JP		 431 struct latency_record *lr = &task->latency_record[i];
432 if (lr->backtrace[0]) {
9745512c 433 int q;
34e49d4f
JP		 434 seq_printf(m, "%i %li %li",
435 lr->count, lr->time, lr->max);
9745512c 436 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
34e49d4f
JP		 437 unsigned long bt = lr->backtrace[q];
438 if (!bt)
9745512c 439 break;
34e49d4f 440 if (bt == ULONG_MAX)
9745512c 441 break;
34e49d4f 442 seq_printf(m, " %ps", (void *)bt);
9745512c 443 }
9d6de12f 444 seq_putc(m, '\n');
9745512c
AV		 445 }
446
447 }
13d77c37 448 put_task_struct(task);
9745512c
AV		 449 return 0;
450}
451
452static int lstats_open(struct inode *inode, struct file *file)
453{
13d77c37 454 return single_open(file, lstats_show_proc, inode);
d6643d12
HS		 455}
456
9745512c
AV		 457static ssize_t lstats_write(struct file *file, const char __user *buf,
458 size_t count, loff_t *offs)
459{
13d77c37 460 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
9745512c 461
13d77c37
HS		 462 if (!task)
463 return -ESRCH;
9745512c 464 clear_all_latency_tracing(task);
13d77c37 465 put_task_struct(task);
9745512c
AV		 466
467 return count;
468}
469
470static const struct file_operations proc_lstats_operations = {
471 .open = lstats_open,
472 .read = seq_read,
473 .write = lstats_write,
474 .llseek = seq_lseek,
13d77c37 475 .release = single_release,
9745512c
AV		 476};
477
478#endif
479
1da177e4
LT		 480static int proc_oom_score(struct task_struct *task, char *buffer)
481{
b95c35e7 482 unsigned long points = 0;
1da177e4 483
19c5d45a 484 read_lock(&tasklist_lock);
b95c35e7 485 if (pid_alive(task))
a63d83f4
DR		 486 points = oom_badness(task, NULL, NULL,
487 totalram_pages + total_swap_pages);
19c5d45a 488 read_unlock(&tasklist_lock);
1da177e4
LT		 489 return sprintf(buffer, "%lu\n", points);
490}
491
d85f50d5
NH		 492struct limit_names {
493 char *name;
494 char *unit;
495};
496
497static const struct limit_names lnames[RLIM_NLIMITS] = {
cff4edb5 498 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
d85f50d5
NH		 499 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
500 [RLIMIT_DATA] = {"Max data size", "bytes"},
501 [RLIMIT_STACK] = {"Max stack size", "bytes"},
502 [RLIMIT_CORE] = {"Max core file size", "bytes"},
503 [RLIMIT_RSS] = {"Max resident set", "bytes"},
504 [RLIMIT_NPROC] = {"Max processes", "processes"},
505 [RLIMIT_NOFILE] = {"Max open files", "files"},
506 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
507 [RLIMIT_AS] = {"Max address space", "bytes"},
508 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
509 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
510 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
511 [RLIMIT_NICE] = {"Max nice priority", NULL},
512 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
8808117c 513 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
d85f50d5
NH		 514};
515
516/* Display limits for a process */
517static int proc_pid_limits(struct task_struct *task, char *buffer)
518{
519 unsigned int i;
520 int count = 0;
521 unsigned long flags;
522 char *bufptr = buffer;
523
524 struct rlimit rlim[RLIM_NLIMITS];
525
a6bebbc8 526 if (!lock_task_sighand(task, &flags))
d85f50d5 527 return 0;
d85f50d5
NH		 528 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
529 unlock_task_sighand(task, &flags);
d85f50d5
NH		 530
531 /*
532 * print the file header
533 */
534 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
535 "Limit", "Soft Limit", "Hard Limit", "Units");
536
537 for (i = 0; i < RLIM_NLIMITS; i++) {
538 if (rlim[i].rlim_cur == RLIM_INFINITY)
539 count += sprintf(&bufptr[count], "%-25s %-20s ",
540 lnames[i].name, "unlimited");
541 else
542 count += sprintf(&bufptr[count], "%-25s %-20lu ",
543 lnames[i].name, rlim[i].rlim_cur);
544
545 if (rlim[i].rlim_max == RLIM_INFINITY)
546 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
547 else
548 count += sprintf(&bufptr[count], "%-20lu ",
549 rlim[i].rlim_max);
550
551 if (lnames[i].unit)
552 count += sprintf(&bufptr[count], "%-10s\n",
553 lnames[i].unit);
554 else
555 count += sprintf(&bufptr[count], "\n");
556 }
557
558 return count;
559}
560
ebcb6734
RM		 561#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
562static int proc_pid_syscall(struct task_struct *task, char *buffer)
563{
564 long nr;
565 unsigned long args[6], sp, pc;
a9712bc1
AV		 566 int res = lock_trace(task);
567 if (res)
568 return res;
ebcb6734
RM		 569
570 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
a9712bc1
AV		 571 res = sprintf(buffer, "running\n");
572 else if (nr < 0)
573 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
574 else
575 res = sprintf(buffer,
ebcb6734
RM		 576 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
577 nr,
578 args[0], args[1], args[2], args[3], args[4], args[5],
579 sp, pc);
a9712bc1
AV		 580 unlock_trace(task);
581 return res;
ebcb6734
RM		 582}
583#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
584
1da177e4
LT		 585/************************************************************************/
586/* Here the fs part begins */
587/************************************************************************/
588
589/* permission checks */
778c1144 590static int proc_fd_access_allowed(struct inode *inode)
1da177e4 591{
778c1144
EB		 592 struct task_struct *task;
593 int allowed = 0;
df26c40e
EB		 594 /* Allow access to a task's file descriptors if it is us or we
595 * may use ptrace attach to the process and find out that
596 * information.
778c1144
EB		 597 */
598 task = get_proc_task(inode);
df26c40e 599 if (task) {
006ebb40 600 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
778c1144 601 put_task_struct(task);
df26c40e 602 }
778c1144 603 return allowed;
1da177e4
LT		 604}
605
6b4e306a 606int proc_setattr(struct dentry *dentry, struct iattr *attr)
6d76fa58
LT		 607{
608 int error;
609 struct inode *inode = dentry->d_inode;
610
611 if (attr->ia_valid & ATTR_MODE)
612 return -EPERM;
613
614 error = inode_change_ok(inode, attr);
1025774c
CH		 615 if (error)
616 return error;
617
618 if ((attr->ia_valid & ATTR_SIZE) &&
619 attr->ia_size != i_size_read(inode)) {
620 error = vmtruncate(inode, attr->ia_size);
621 if (error)
622 return error;
623 }
624
625 setattr_copy(inode, attr);
626 mark_inode_dirty(inode);
627 return 0;
6d76fa58
LT		 628}
629
c5ef1c42 630static const struct inode_operations proc_def_inode_operations = {
6d76fa58
LT		 631 .setattr = proc_setattr,
632};
633
1da177e4
LT		 634#define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
635
636static ssize_t proc_info_read(struct file * file, char __user * buf,
637 size_t count, loff_t *ppos)
638{
2fddfeef 639 struct inode * inode = file->f_path.dentry->d_inode;
1da177e4
LT		 640 unsigned long page;
641 ssize_t length;
99f89551
EB		 642 struct task_struct *task = get_proc_task(inode);
643
644 length = -ESRCH;
645 if (!task)
646 goto out_no_task;
1da177e4
LT		 647
648 if (count > PROC_BLOCK_SIZE)
649 count = PROC_BLOCK_SIZE;
99f89551
EB		 650
651 length = -ENOMEM;
e12ba74d 652 if (!(page = __get_free_page(GFP_TEMPORARY)))
99f89551 653 goto out;
1da177e4
LT		 654
655 length = PROC_I(inode)->op.proc_read(task, (char*)page);
656
657 if (length >= 0)
658 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
659 free_page(page);
99f89551
EB		 660out:
661 put_task_struct(task);
662out_no_task:
1da177e4
LT		 663 return length;
664}
665
00977a59 666static const struct file_operations proc_info_file_operations = {
1da177e4 667 .read = proc_info_read,
87df8424 668 .llseek = generic_file_llseek,
1da177e4
LT		 669};
670
be614086
EB		 671static int proc_single_show(struct seq_file *m, void *v)
672{
673 struct inode *inode = m->private;
674 struct pid_namespace *ns;
675 struct pid *pid;
676 struct task_struct *task;
677 int ret;
678
679 ns = inode->i_sb->s_fs_info;
680 pid = proc_pid(inode);
681 task = get_pid_task(pid, PIDTYPE_PID);
682 if (!task)
683 return -ESRCH;
684
685 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
686
687 put_task_struct(task);
688 return ret;
689}
690
691static int proc_single_open(struct inode *inode, struct file *filp)
692{
c6a34058 693 return single_open(filp, proc_single_show, inode);
be614086
EB		 694}
695
696static const struct file_operations proc_single_file_operations = {
697 .open = proc_single_open,
698 .read = seq_read,
699 .llseek = seq_lseek,
700 .release = single_release,
701};
702
1da177e4
LT		 703static int mem_open(struct inode* inode, struct file* file)
704{
705 file->private_data = (void*)((long)current->self_exec_id);
4a3956c7
KH		 706 /* OK to pass negative loff_t, we can catch out-of-range */
707 file->f_mode |= FMODE_UNSIGNED_OFFSET;
1da177e4
LT		 708 return 0;
709}
710
711static ssize_t mem_read(struct file * file, char __user * buf,
712 size_t count, loff_t *ppos)
713{
2fddfeef 714 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1da177e4
LT		 715 char *page;
716 unsigned long src = *ppos;
717 int ret = -ESRCH;
718 struct mm_struct *mm;
719
99f89551
EB		 720 if (!task)
721 goto out_no_task;
722
1da177e4 723 ret = -ENOMEM;
e12ba74d 724 page = (char *)__get_free_page(GFP_TEMPORARY);
1da177e4
LT		 725 if (!page)
726 goto out;
727
8b0db9db
SW		 728 mm = check_mem_permission(task);
729 ret = PTR_ERR(mm);
730 if (IS_ERR(mm))
1da177e4
LT		 731 goto out_free;
732
733 ret = -EIO;
734
735 if (file->private_data != (void*)((long)current->self_exec_id))
736 goto out_put;
737
738 ret = 0;
739
740 while (count > 0) {
741 int this_len, retval;
742
743 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
8b0db9db
SW		 744 retval = access_remote_vm(mm, src, page, this_len, 0);
745 if (!retval) {
1da177e4
LT		 746 if (!ret)
747 ret = -EIO;
748 break;
749 }
750
751 if (copy_to_user(buf, page, retval)) {
752 ret = -EFAULT;
753 break;
754 }
755
756 ret += retval;
757 src += retval;
758 buf += retval;
759 count -= retval;
760 }
761 *ppos = src;
762
763out_put:
764 mmput(mm);
765out_free:
766 free_page((unsigned long) page);
767out:
99f89551
EB		 768 put_task_struct(task);
769out_no_task:
1da177e4
LT		 770 return ret;
771}
772
63967fa9 773static ssize_t mem_write(struct file * file, const char __user *buf,
1da177e4
LT		 774 size_t count, loff_t *ppos)
775{
f7ca54f4 776 int copied;
1da177e4 777 char *page;
2fddfeef 778 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1da177e4 779 unsigned long dst = *ppos;
8b0db9db 780 struct mm_struct *mm;
1da177e4 781
99f89551
EB		 782 copied = -ESRCH;
783 if (!task)
784 goto out_no_task;
785
30cd8903
KM		 786 copied = -ENOMEM;
787 page = (char *)__get_free_page(GFP_TEMPORARY);
788 if (!page)
789 goto out_task;
790
8b0db9db
SW		 791 mm = check_mem_permission(task);
792 copied = PTR_ERR(mm);
793 if (IS_ERR(mm))
30cd8903 794 goto out_free;
1da177e4 795
26947f8c
SW		 796 copied = -EIO;
797 if (file->private_data != (void *)((long)current->self_exec_id))
8b0db9db 798 goto out_mm;
1da177e4 799
f7ca54f4 800 copied = 0;
1da177e4
LT		 801 while (count > 0) {
802 int this_len, retval;
803
804 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
805 if (copy_from_user(page, buf, this_len)) {
806 copied = -EFAULT;
807 break;
808 }
8b0db9db 809 retval = access_remote_vm(mm, dst, page, this_len, 1);
1da177e4
LT		 810 if (!retval) {
811 if (!copied)
812 copied = -EIO;
813 break;
814 }
815 copied += retval;
816 buf += retval;
817 dst += retval;
818 count -= retval;
819 }
820 *ppos = dst;
30cd8903 821
8b0db9db
SW		 822out_mm:
823 mmput(mm);
30cd8903
KM		 824out_free:
825 free_page((unsigned long) page);
8b0db9db 826out_task:
99f89551
EB		 827 put_task_struct(task);
828out_no_task:
1da177e4
LT		 829 return copied;
830}
1da177e4 831
85863e47 832loff_t mem_lseek(struct file *file, loff_t offset, int orig)
1da177e4
LT		 833{
834 switch (orig) {
835 case 0:
836 file->f_pos = offset;
837 break;
838 case 1:
839 file->f_pos += offset;
840 break;
841 default:
842 return -EINVAL;
843 }
844 force_successful_syscall_return();
845 return file->f_pos;
846}
847
00977a59 848static const struct file_operations proc_mem_operations = {
1da177e4
LT		 849 .llseek = mem_lseek,
850 .read = mem_read,
851 .write = mem_write,
852 .open = mem_open,
853};
854
315e28c8
JP		 855static ssize_t environ_read(struct file *file, char __user *buf,
856 size_t count, loff_t *ppos)
857{
858 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
859 char *page;
860 unsigned long src = *ppos;
861 int ret = -ESRCH;
862 struct mm_struct *mm;
863
864 if (!task)
865 goto out_no_task;
866
315e28c8
JP		 867 ret = -ENOMEM;
868 page = (char *)__get_free_page(GFP_TEMPORARY);
869 if (!page)
870 goto out;
871
315e28c8 872
d6f64b89
AV		 873 mm = mm_for_maps(task);
874 ret = PTR_ERR(mm);
875 if (!mm || IS_ERR(mm))
315e28c8
JP		 876 goto out_free;
877
d6f64b89 878 ret = 0;
315e28c8
JP		 879 while (count > 0) {
880 int this_len, retval, max_len;
881
882 this_len = mm->env_end - (mm->env_start + src);
883
884 if (this_len <= 0)
885 break;
886
887 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
888 this_len = (this_len > max_len) ? max_len : this_len;
889
890 retval = access_process_vm(task, (mm->env_start + src),
891 page, this_len, 0);
892
893 if (retval <= 0) {
894 ret = retval;
895 break;
896 }
897
898 if (copy_to_user(buf, page, retval)) {
899 ret = -EFAULT;
900 break;
901 }
902
903 ret += retval;
904 src += retval;
905 buf += retval;
906 count -= retval;
907 }
908 *ppos = src;
909
910 mmput(mm);
911out_free:
912 free_page((unsigned long) page);
913out:
914 put_task_struct(task);
915out_no_task:
916 return ret;
917}
918
919static const struct file_operations proc_environ_operations = {
920 .read = environ_read,
87df8424 921 .llseek = generic_file_llseek,
315e28c8
JP		 922};
923
1da177e4
LT		 924static ssize_t oom_adjust_read(struct file *file, char __user *buf,
925 size_t count, loff_t *ppos)
926{
2fddfeef 927 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
8578cea7 928 char buffer[PROC_NUMBUF];
1da177e4 929 size_t len;
28b83c51
KM		 930 int oom_adjust = OOM_DISABLE;
931 unsigned long flags;
1da177e4 932
99f89551
EB		 933 if (!task)
934 return -ESRCH;
28b83c51
KM		 935
936 if (lock_task_sighand(task, &flags)) {
937 oom_adjust = task->signal->oom_adj;
938 unlock_task_sighand(task, &flags);
939 }
940
99f89551
EB		 941 put_task_struct(task);
942
8578cea7 943 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
0c28f287
AM		 944
945 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1da177e4
LT		 946}
947
948static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
949 size_t count, loff_t *ppos)
950{
99f89551 951 struct task_struct *task;
5d863b89 952 char buffer[PROC_NUMBUF];
0a8cb8e3 953 int oom_adjust;
28b83c51 954 unsigned long flags;
5d863b89 955 int err;
1da177e4 956
8578cea7
EB		 957 memset(buffer, 0, sizeof(buffer));
958 if (count > sizeof(buffer) - 1)
959 count = sizeof(buffer) - 1;
723548bf
DR		 960 if (copy_from_user(buffer, buf, count)) {
961 err = -EFAULT;
962 goto out;
963 }
5d863b89 964
0a8cb8e3 965 err = kstrtoint(strstrip(buffer), 0, &oom_adjust);
5d863b89 966 if (err)
723548bf 967 goto out;
8ac773b4 968 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
723548bf
DR		 969 oom_adjust != OOM_DISABLE) {
970 err = -EINVAL;
971 goto out;
972 }
5d863b89 973
2fddfeef 974 task = get_proc_task(file->f_path.dentry->d_inode);
723548bf
DR		 975 if (!task) {
976 err = -ESRCH;
977 goto out;
978 }
d19d5476
DR		 979
980 task_lock(task);
981 if (!task->mm) {
982 err = -EINVAL;
983 goto err_task_lock;
984 }
985
28b83c51 986 if (!lock_task_sighand(task, &flags)) {
723548bf 987 err = -ESRCH;
d19d5476 988 goto err_task_lock;
28b83c51
KM		 989 }
990
991 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
723548bf
DR		 992 err = -EACCES;
993 goto err_sighand;
8fb4fc68 994 }
28b83c51 995
51b1bd2a
DR		 996 /*
997 * Warn that /proc/pid/oom_adj is deprecated, see
998 * Documentation/feature-removal-schedule.txt.
999 */
c2142704 1000 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
be8f684d
DR		 1001 current->comm, task_pid_nr(current), task_pid_nr(task),
1002 task_pid_nr(task));
28b83c51 1003 task->signal->oom_adj = oom_adjust;
a63d83f4
DR		 1004 /*
1005 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1006 * value is always attainable.
1007 */
1008 if (task->signal->oom_adj == OOM_ADJUST_MAX)
1009 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
1010 else
1011 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
1012 -OOM_DISABLE;
723548bf 1013err_sighand:
28b83c51 1014 unlock_task_sighand(task, &flags);
d19d5476
DR		 1015err_task_lock:
1016 task_unlock(task);
99f89551 1017 put_task_struct(task);
723548bf
DR		 1018out:
1019 return err < 0 ? err : count;
1da177e4
LT		 1020}
1021
00977a59 1022static const struct file_operations proc_oom_adjust_operations = {
1da177e4
LT		 1023 .read = oom_adjust_read,
1024 .write = oom_adjust_write,
87df8424 1025 .llseek = generic_file_llseek,
1da177e4
LT		 1026};
1027
a63d83f4
DR		 1028static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1029 size_t count, loff_t *ppos)
1030{
1031 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1032 char buffer[PROC_NUMBUF];
1033 int oom_score_adj = OOM_SCORE_ADJ_MIN;
1034 unsigned long flags;
1035 size_t len;
1036
1037 if (!task)
1038 return -ESRCH;
1039 if (lock_task_sighand(task, &flags)) {
1040 oom_score_adj = task->signal->oom_score_adj;
1041 unlock_task_sighand(task, &flags);
1042 }
1043 put_task_struct(task);
1044 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
1045 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1046}
1047
1048static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1049 size_t count, loff_t *ppos)
1050{
1051 struct task_struct *task;
1052 char buffer[PROC_NUMBUF];
1053 unsigned long flags;
0a8cb8e3 1054 int oom_score_adj;
a63d83f4
DR		 1055 int err;
1056
1057 memset(buffer, 0, sizeof(buffer));
1058 if (count > sizeof(buffer) - 1)
1059 count = sizeof(buffer) - 1;
723548bf
DR		 1060 if (copy_from_user(buffer, buf, count)) {
1061 err = -EFAULT;
1062 goto out;
1063 }
a63d83f4 1064
0a8cb8e3 1065 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
a63d83f4 1066 if (err)
723548bf 1067 goto out;
a63d83f4 1068 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
723548bf
DR		 1069 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1070 err = -EINVAL;
1071 goto out;
1072 }
a63d83f4
DR		 1073
1074 task = get_proc_task(file->f_path.dentry->d_inode);
723548bf
DR		 1075 if (!task) {
1076 err = -ESRCH;
1077 goto out;
1078 }
d19d5476
DR		 1079
1080 task_lock(task);
1081 if (!task->mm) {
1082 err = -EINVAL;
1083 goto err_task_lock;
1084 }
1085
a63d83f4 1086 if (!lock_task_sighand(task, &flags)) {
723548bf 1087 err = -ESRCH;
d19d5476 1088 goto err_task_lock;
a63d83f4 1089 }
d19d5476 1090
dabb16f6 1091 if (oom_score_adj < task->signal->oom_score_adj_min &&
a63d83f4 1092 !capable(CAP_SYS_RESOURCE)) {
723548bf
DR		 1093 err = -EACCES;
1094 goto err_sighand;
a63d83f4
DR		 1095 }
1096
1097 task->signal->oom_score_adj = oom_score_adj;
dabb16f6
MSB		 1098 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1099 task->signal->oom_score_adj_min = oom_score_adj;
a63d83f4
DR		 1100 /*
1101 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1102 * always attainable.
1103 */
1104 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1105 task->signal->oom_adj = OOM_DISABLE;
1106 else
1107 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1108 OOM_SCORE_ADJ_MAX;
723548bf 1109err_sighand:
a63d83f4 1110 unlock_task_sighand(task, &flags);
d19d5476
DR		 1111err_task_lock:
1112 task_unlock(task);
a63d83f4 1113 put_task_struct(task);
723548bf
DR		 1114out:
1115 return err < 0 ? err : count;
a63d83f4
DR		 1116}
1117
1118static const struct file_operations proc_oom_score_adj_operations = {
1119 .read = oom_score_adj_read,
1120 .write = oom_score_adj_write,
6038f373 1121 .llseek = default_llseek,
a63d83f4
DR		 1122};
1123
1da177e4
LT		 1124#ifdef CONFIG_AUDITSYSCALL
1125#define TMPBUFLEN 21
1126static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1127 size_t count, loff_t *ppos)
1128{
2fddfeef 1129 struct inode * inode = file->f_path.dentry->d_inode;
99f89551 1130 struct task_struct *task = get_proc_task(inode);
1da177e4
LT		 1131 ssize_t length;
1132 char tmpbuf[TMPBUFLEN];
1133
99f89551
EB		 1134 if (!task)
1135 return -ESRCH;
1da177e4 1136 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
0c11b942 1137 audit_get_loginuid(task));
99f89551 1138 put_task_struct(task);
1da177e4
LT		 1139 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1140}
1141
1142static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1143 size_t count, loff_t *ppos)
1144{
2fddfeef 1145 struct inode * inode = file->f_path.dentry->d_inode;
1da177e4
LT		 1146 char *page, *tmp;
1147 ssize_t length;
1da177e4
LT		 1148 uid_t loginuid;
1149
1150 if (!capable(CAP_AUDIT_CONTROL))
1151 return -EPERM;
1152
7dc52157
PM		 1153 rcu_read_lock();
1154 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1155 rcu_read_unlock();
1da177e4 1156 return -EPERM;
7dc52157
PM		 1157 }
1158 rcu_read_unlock();
1da177e4 1159
e0182909
AV		 1160 if (count >= PAGE_SIZE)
1161 count = PAGE_SIZE - 1;
1da177e4
LT		 1162
1163 if (*ppos != 0) {
1164 /* No partial writes. */
1165 return -EINVAL;
1166 }
e12ba74d 1167 page = (char*)__get_free_page(GFP_TEMPORARY);
1da177e4
LT		 1168 if (!page)
1169 return -ENOMEM;
1170 length = -EFAULT;
1171 if (copy_from_user(page, buf, count))
1172 goto out_free_page;
1173
e0182909 1174 page[count] = '\0';
1da177e4
LT		 1175 loginuid = simple_strtoul(page, &tmp, 10);
1176 if (tmp == page) {
1177 length = -EINVAL;
1178 goto out_free_page;
1179
1180 }
99f89551 1181 length = audit_set_loginuid(current, loginuid);
1da177e4
LT		 1182 if (likely(length == 0))
1183 length = count;
1184
1185out_free_page:
1186 free_page((unsigned long) page);
1187 return length;
1188}
1189
00977a59 1190static const struct file_operations proc_loginuid_operations = {
1da177e4
LT		 1191 .read = proc_loginuid_read,
1192 .write = proc_loginuid_write,
87df8424 1193 .llseek = generic_file_llseek,
1da177e4 1194};
1e0bd755
EP		 1195
1196static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1197 size_t count, loff_t *ppos)
1198{
1199 struct inode * inode = file->f_path.dentry->d_inode;
1200 struct task_struct *task = get_proc_task(inode);
1201 ssize_t length;
1202 char tmpbuf[TMPBUFLEN];
1203
1204 if (!task)
1205 return -ESRCH;
1206 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1207 audit_get_sessionid(task));
1208 put_task_struct(task);
1209 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1210}
1211
1212static const struct file_operations proc_sessionid_operations = {
1213 .read = proc_sessionid_read,
87df8424 1214 .llseek = generic_file_llseek,
1e0bd755 1215};
1da177e4
LT		 1216#endif
1217
f4f154fd
AM		 1218#ifdef CONFIG_FAULT_INJECTION
1219static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1220 size_t count, loff_t *ppos)
1221{
1222 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1223 char buffer[PROC_NUMBUF];
1224 size_t len;
1225 int make_it_fail;
f4f154fd
AM		 1226
1227 if (!task)
1228 return -ESRCH;
1229 make_it_fail = task->make_it_fail;
1230 put_task_struct(task);
1231
1232 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
0c28f287
AM		 1233
1234 return simple_read_from_buffer(buf, count, ppos, buffer, len);
f4f154fd
AM		 1235}
1236
1237static ssize_t proc_fault_inject_write(struct file * file,
1238 const char __user * buf, size_t count, loff_t *ppos)
1239{
1240 struct task_struct *task;
1241 char buffer[PROC_NUMBUF], *end;
1242 int make_it_fail;
1243
1244 if (!capable(CAP_SYS_RESOURCE))
1245 return -EPERM;
1246 memset(buffer, 0, sizeof(buffer));
1247 if (count > sizeof(buffer) - 1)
1248 count = sizeof(buffer) - 1;
1249 if (copy_from_user(buffer, buf, count))
1250 return -EFAULT;
cba8aafe
VL		 1251 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1252 if (*end)
1253 return -EINVAL;
f4f154fd
AM		 1254 task = get_proc_task(file->f_dentry->d_inode);
1255 if (!task)
1256 return -ESRCH;
1257 task->make_it_fail = make_it_fail;
1258 put_task_struct(task);
cba8aafe
VL		 1259
1260 return count;
f4f154fd
AM		 1261}
1262
00977a59 1263static const struct file_operations proc_fault_inject_operations = {
f4f154fd
AM		 1264 .read = proc_fault_inject_read,
1265 .write = proc_fault_inject_write,
87df8424 1266 .llseek = generic_file_llseek,
f4f154fd
AM		 1267};
1268#endif
1269
9745512c 1270
43ae34cb
IM		 1271#ifdef CONFIG_SCHED_DEBUG
1272/*
1273 * Print out various scheduling related per-task fields:
1274 */
1275static int sched_show(struct seq_file *m, void *v)
1276{
1277 struct inode *inode = m->private;
1278 struct task_struct *p;
1279
43ae34cb
IM		 1280 p = get_proc_task(inode);
1281 if (!p)
1282 return -ESRCH;
1283 proc_sched_show_task(p, m);
1284
1285 put_task_struct(p);
1286
1287 return 0;
1288}
1289
1290static ssize_t
1291sched_write(struct file *file, const char __user *buf,
1292 size_t count, loff_t *offset)
1293{
1294 struct inode *inode = file->f_path.dentry->d_inode;
1295 struct task_struct *p;
1296
43ae34cb
IM		 1297 p = get_proc_task(inode);
1298 if (!p)
1299 return -ESRCH;
1300 proc_sched_set_task(p);
1301
1302 put_task_struct(p);
1303
1304 return count;
1305}
1306
1307static int sched_open(struct inode *inode, struct file *filp)
1308{
c6a34058 1309 return single_open(filp, sched_show, inode);
43ae34cb
IM		 1310}
1311
1312static const struct file_operations proc_pid_sched_operations = {
1313 .open = sched_open,
1314 .read = seq_read,
1315 .write = sched_write,
1316 .llseek = seq_lseek,
5ea473a1 1317 .release = single_release,
43ae34cb
IM		 1318};
1319
1320#endif
1321
5091faa4
MG		 1322#ifdef CONFIG_SCHED_AUTOGROUP
1323/*
1324 * Print out autogroup related information:
1325 */
1326static int sched_autogroup_show(struct seq_file *m, void *v)
1327{
1328 struct inode *inode = m->private;
1329 struct task_struct *p;
1330
1331 p = get_proc_task(inode);
1332 if (!p)
1333 return -ESRCH;
1334 proc_sched_autogroup_show_task(p, m);
1335
1336 put_task_struct(p);
1337
1338 return 0;
1339}
1340
1341static ssize_t
1342sched_autogroup_write(struct file *file, const char __user *buf,
1343 size_t count, loff_t *offset)
1344{
1345 struct inode *inode = file->f_path.dentry->d_inode;
1346 struct task_struct *p;
1347 char buffer[PROC_NUMBUF];
0a8cb8e3 1348 int nice;
5091faa4
MG		 1349 int err;
1350
1351 memset(buffer, 0, sizeof(buffer));
1352 if (count > sizeof(buffer) - 1)
1353 count = sizeof(buffer) - 1;
1354 if (copy_from_user(buffer, buf, count))
1355 return -EFAULT;
1356
0a8cb8e3
AD		 1357 err = kstrtoint(strstrip(buffer), 0, &nice);
1358 if (err < 0)
1359 return err;
5091faa4
MG		 1360
1361 p = get_proc_task(inode);
1362 if (!p)
1363 return -ESRCH;
1364
1365 err = nice;
1366 err = proc_sched_autogroup_set_nice(p, &err);
1367 if (err)
1368 count = err;
1369
1370 put_task_struct(p);
1371
1372 return count;
1373}
1374
1375static int sched_autogroup_open(struct inode *inode, struct file *filp)
1376{
1377 int ret;
1378
1379 ret = single_open(filp, sched_autogroup_show, NULL);
1380 if (!ret) {
1381 struct seq_file *m = filp->private_data;
1382
1383 m->private = inode;
1384 }
1385 return ret;
1386}
1387
1388static const struct file_operations proc_pid_sched_autogroup_operations = {
1389 .open = sched_autogroup_open,
1390 .read = seq_read,
1391 .write = sched_autogroup_write,
1392 .llseek = seq_lseek,
1393 .release = single_release,
1394};
1395
1396#endif /* CONFIG_SCHED_AUTOGROUP */
1397
4614a696
JS		 1398static ssize_t comm_write(struct file *file, const char __user *buf,
1399 size_t count, loff_t *offset)
1400{
1401 struct inode *inode = file->f_path.dentry->d_inode;
1402 struct task_struct *p;
1403 char buffer[TASK_COMM_LEN];
1404
1405 memset(buffer, 0, sizeof(buffer));
1406 if (count > sizeof(buffer) - 1)
1407 count = sizeof(buffer) - 1;
1408 if (copy_from_user(buffer, buf, count))
1409 return -EFAULT;
1410
1411 p = get_proc_task(inode);
1412 if (!p)
1413 return -ESRCH;
1414
1415 if (same_thread_group(current, p))
1416 set_task_comm(p, buffer);
1417 else
1418 count = -EINVAL;
1419
1420 put_task_struct(p);
1421
1422 return count;
1423}
1424
1425static int comm_show(struct seq_file *m, void *v)
1426{
1427 struct inode *inode = m->private;
1428 struct task_struct *p;
1429
1430 p = get_proc_task(inode);
1431 if (!p)
1432 return -ESRCH;
1433
1434 task_lock(p);
1435 seq_printf(m, "%s\n", p->comm);
1436 task_unlock(p);
1437
1438 put_task_struct(p);
1439
1440 return 0;
1441}
1442
1443static int comm_open(struct inode *inode, struct file *filp)
1444{
c6a34058 1445 return single_open(filp, comm_show, inode);
4614a696
JS		 1446}
1447
1448static const struct file_operations proc_pid_set_comm_operations = {
1449 .open = comm_open,
1450 .read = seq_read,
1451 .write = comm_write,
1452 .llseek = seq_lseek,
1453 .release = single_release,
1454};
1455
925d1c40
MH		 1456static int proc_exe_link(struct inode *inode, struct path *exe_path)
1457{
1458 struct task_struct *task;
1459 struct mm_struct *mm;
1460 struct file *exe_file;
1461
1462 task = get_proc_task(inode);
1463 if (!task)
1464 return -ENOENT;
1465 mm = get_task_mm(task);
1466 put_task_struct(task);
1467 if (!mm)
1468 return -ENOENT;
1469 exe_file = get_mm_exe_file(mm);
1470 mmput(mm);
1471 if (exe_file) {
1472 *exe_path = exe_file->f_path;
1473 path_get(&exe_file->f_path);
1474 fput(exe_file);
1475 return 0;
1476 } else
1477 return -ENOENT;
1478}
1479
008b150a 1480static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1da177e4
LT		 1481{
1482 struct inode *inode = dentry->d_inode;
1483 int error = -EACCES;
1484
1485 /* We don't need a base pointer in the /proc filesystem */
1d957f9b 1486 path_put(&nd->path);
1da177e4 1487
778c1144
EB		 1488 /* Are we allowed to snoop on the tasks file descriptors? */
1489 if (!proc_fd_access_allowed(inode))
1da177e4 1490 goto out;
1da177e4 1491
3dcd25f3 1492 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
1da177e4 1493out:
008b150a 1494 return ERR_PTR(error);
1da177e4
LT		 1495}
1496
3dcd25f3 1497static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1da177e4 1498{
e12ba74d 1499 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
3dcd25f3 1500 char *pathname;
1da177e4
LT		 1501 int len;
1502
1503 if (!tmp)
1504 return -ENOMEM;
0c28f287 1505
7b2a69ba 1506 pathname = d_path(path, tmp, PAGE_SIZE);
3dcd25f3
JB		 1507 len = PTR_ERR(pathname);
1508 if (IS_ERR(pathname))
1da177e4 1509 goto out;
3dcd25f3 1510 len = tmp + PAGE_SIZE - 1 - pathname;
1da177e4
LT		 1511
1512 if (len > buflen)
1513 len = buflen;
3dcd25f3 1514 if (copy_to_user(buffer, pathname, len))
1da177e4
LT		 1515 len = -EFAULT;
1516 out:
1517 free_page((unsigned long)tmp);
1518 return len;
1519}
1520
1521static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1522{
1523 int error = -EACCES;
1524 struct inode *inode = dentry->d_inode;
3dcd25f3 1525 struct path path;
1da177e4 1526
778c1144
EB		 1527 /* Are we allowed to snoop on the tasks file descriptors? */
1528 if (!proc_fd_access_allowed(inode))
1da177e4 1529 goto out;
1da177e4 1530
3dcd25f3 1531 error = PROC_I(inode)->op.proc_get_link(inode, &path);
1da177e4
LT		 1532 if (error)
1533 goto out;
1534
3dcd25f3
JB		 1535 error = do_proc_readlink(&path, buffer, buflen);
1536 path_put(&path);
1da177e4 1537out:
1da177e4
LT		 1538 return error;
1539}
1540
c5ef1c42 1541static const struct inode_operations proc_pid_link_inode_operations = {
1da177e4 1542 .readlink = proc_pid_readlink,
6d76fa58
LT		 1543 .follow_link = proc_pid_follow_link,
1544 .setattr = proc_setattr,
1da177e4
LT		 1545};
1546
28a6d671
EB		 1547
1548/* building an inode */
1549
1550static int task_dumpable(struct task_struct *task)
1da177e4 1551{
28a6d671
EB		 1552 int dumpable = 0;
1553 struct mm_struct *mm;
1da177e4 1554
28a6d671
EB		 1555 task_lock(task);
1556 mm = task->mm;
1557 if (mm)
6c5d5238 1558 dumpable = get_dumpable(mm);
28a6d671
EB		 1559 task_unlock(task);
1560 if(dumpable == 1)
1561 return 1;
1562 return 0;
1563}
1da177e4 1564
6b4e306a 1565struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
28a6d671
EB		 1566{
1567 struct inode * inode;
1568 struct proc_inode *ei;
c69e8d9c 1569 const struct cred *cred;
1da177e4 1570
28a6d671 1571 /* We need a new inode */
1da177e4 1572
28a6d671
EB		 1573 inode = new_inode(sb);
1574 if (!inode)
1575 goto out;
1576
1577 /* Common stuff */
1578 ei = PROC_I(inode);
85fe4025 1579 inode->i_ino = get_next_ino();
28a6d671 1580 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
28a6d671
EB		 1581 inode->i_op = &proc_def_inode_operations;
1582
1583 /*
1584 * grab the reference to task.
1585 */
1a657f78 1586 ei->pid = get_task_pid(task, PIDTYPE_PID);
28a6d671
EB		 1587 if (!ei->pid)
1588 goto out_unlock;
1589
28a6d671 1590 if (task_dumpable(task)) {
c69e8d9c
DH		 1591 rcu_read_lock();
1592 cred = __task_cred(task);
1593 inode->i_uid = cred->euid;
1594 inode->i_gid = cred->egid;
1595 rcu_read_unlock();
1da177e4 1596 }
28a6d671
EB		 1597 security_task_to_inode(task, inode);
1598
1da177e4 1599out:
28a6d671
EB		 1600 return inode;
1601
1602out_unlock:
1603 iput(inode);
1604 return NULL;
1da177e4
LT		 1605}
1606
6b4e306a 1607int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1da177e4 1608{
1da177e4 1609 struct inode *inode = dentry->d_inode;
28a6d671 1610 struct task_struct *task;
c69e8d9c
DH		 1611 const struct cred *cred;
1612
28a6d671 1613 generic_fillattr(inode, stat);
1da177e4 1614
28a6d671
EB		 1615 rcu_read_lock();
1616 stat->uid = 0;
1617 stat->gid = 0;
1618 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1619 if (task) {
1620 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1621 task_dumpable(task)) {
c69e8d9c
DH		 1622 cred = __task_cred(task);
1623 stat->uid = cred->euid;
1624 stat->gid = cred->egid;
1da177e4
LT		 1625 }
1626 }
28a6d671 1627 rcu_read_unlock();
d6e71144 1628 return 0;
1da177e4
LT		 1629}
1630
1da177e4
LT		 1631/* dentry stuff */
1632
1633/*
1634 * Exceptional case: normally we are not allowed to unhash a busy
1635 * directory. In this case, however, we can do it - no aliasing problems
1636 * due to the way we treat inodes.
1637 *
1638 * Rewrite the inode's ownerships here because the owning task may have
1639 * performed a setuid(), etc.
99f89551
EB		 1640 *
1641 * Before the /proc/pid/status file was created the only way to read
1642 * the effective uid of a /process was to stat /proc/pid. Reading
1643 * /proc/pid/status is slow enough that procps and other packages
1644 * kept stating /proc/pid. To keep the rules in /proc simple I have
1645 * made this apply to all per process world readable and executable
1646 * directories.
1da177e4 1647 */
6b4e306a 1648int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1da177e4 1649{
34286d66
NP		 1650 struct inode *inode;
1651 struct task_struct *task;
c69e8d9c
DH		 1652 const struct cred *cred;
1653
34286d66
NP		 1654 if (nd && nd->flags & LOOKUP_RCU)
1655 return -ECHILD;
1656
1657 inode = dentry->d_inode;
1658 task = get_proc_task(inode);
1659
99f89551
EB		 1660 if (task) {
1661 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1662 task_dumpable(task)) {
c69e8d9c
DH		 1663 rcu_read_lock();
1664 cred = __task_cred(task);
1665 inode->i_uid = cred->euid;
1666 inode->i_gid = cred->egid;
1667 rcu_read_unlock();
1da177e4
LT		 1668 } else {
1669 inode->i_uid = 0;
1670 inode->i_gid = 0;
1671 }
9ee8ab9f 1672 inode->i_mode &= ~(S_ISUID | S_ISGID);
1da177e4 1673 security_task_to_inode(task, inode);
99f89551 1674 put_task_struct(task);
1da177e4
LT		 1675 return 1;
1676 }
1677 d_drop(dentry);
1678 return 0;
1679}
1680
fe15ce44 1681static int pid_delete_dentry(const struct dentry * dentry)
99f89551 1682{
28a6d671
EB		 1683 /* Is the task we represent dead?
1684 * If so, then don't put the dentry on the lru list,
1685 * kill it immediately.
1686 */
1687 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1688}
1689
6b4e306a 1690const struct dentry_operations pid_dentry_operations =
28a6d671
EB		 1691{
1692 .d_revalidate = pid_revalidate,
1693 .d_delete = pid_delete_dentry,
1694};
1695
1696/* Lookups */
1697
1c0d04c9
EB		 1698/*
1699 * Fill a directory entry.
1700 *
1701 * If possible create the dcache entry and derive our inode number and
1702 * file type from dcache entry.
1703 *
1704 * Since all of the proc inode numbers are dynamically generated, the inode
1705 * numbers do not exist until the inode is cache. This means creating the
1706 * the dcache entry in readdir is necessary to keep the inode numbers
1707 * reported by readdir in sync with the inode numbers reported
1708 * by stat.
1709 */
6b4e306a
EB		 1710int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1711 const char *name, int len,
c5141e6d 1712 instantiate_t instantiate, struct task_struct *task, const void *ptr)
61a28784 1713{
2fddfeef 1714 struct dentry *child, *dir = filp->f_path.dentry;
61a28784
EB		 1715 struct inode *inode;
1716 struct qstr qname;
1717 ino_t ino = 0;
1718 unsigned type = DT_UNKNOWN;
1719
1720 qname.name = name;
1721 qname.len = len;
1722 qname.hash = full_name_hash(name, len);
1723
1724 child = d_lookup(dir, &qname);
1725 if (!child) {
1726 struct dentry *new;
1727 new = d_alloc(dir, &qname);
1728 if (new) {
1729 child = instantiate(dir->d_inode, new, task, ptr);
1730 if (child)
1731 dput(new);
1732 else
1733 child = new;
1734 }
1735 }
1736 if (!child || IS_ERR(child) || !child->d_inode)
1737 goto end_instantiate;
1738 inode = child->d_inode;
1739 if (inode) {
1740 ino = inode->i_ino;
1741 type = inode->i_mode >> 12;
1742 }
1743 dput(child);
1744end_instantiate:
1745 if (!ino)
1746 ino = find_inode_number(dir, &qname);
1747 if (!ino)
1748 ino = 1;
1749 return filldir(dirent, name, len, filp->f_pos, ino, type);
1750}
1751
28a6d671
EB		 1752static unsigned name_to_int(struct dentry *dentry)
1753{
1754 const char *name = dentry->d_name.name;
1755 int len = dentry->d_name.len;
1756 unsigned n = 0;
1757
1758 if (len > 1 && *name == '0')
1759 goto out;
1760 while (len-- > 0) {
1761 unsigned c = *name++ - '0';
1762 if (c > 9)
1763 goto out;
1764 if (n >= (~0U-9)/10)
1765 goto out;
1766 n *= 10;
1767 n += c;
1768 }
1769 return n;
1770out:
1771 return ~0U;
1772}
1773
27932742
MS		 1774#define PROC_FDINFO_MAX 64
1775
3dcd25f3 1776static int proc_fd_info(struct inode *inode, struct path *path, char *info)
28a6d671 1777{
5e442a49
LT		 1778 struct task_struct *task = get_proc_task(inode);
1779 struct files_struct *files = NULL;
28a6d671
EB		 1780 struct file *file;
1781 int fd = proc_fd(inode);
aa6afca5 1782
5e442a49
LT		 1783 if (task) {
1784 files = get_files_struct(task);
1785 put_task_struct(task);
1786 }
1787 if (files) {
1788 /*
1789 * We are not taking a ref to the file structure, so we must
1790 * hold ->file_lock.
1791 */
1792 spin_lock(&files->file_lock);
1793 file = fcheck_files(files, fd);
1794 if (file) {
1795 unsigned int f_flags;
1796 struct fdtable *fdt;
1797
1798 fdt = files_fdtable(files);
1799 f_flags = file->f_flags & ~O_CLOEXEC;
1800 if (FD_ISSET(fd, fdt->close_on_exec))
1801 f_flags |= O_CLOEXEC;
1802
1803 if (path) {
1804 *path = file->f_path;
1805 path_get(&file->f_path);
1806 }
1807 if (info)
1808 snprintf(info, PROC_FDINFO_MAX,
1809 "pos:\t%lli\n"
1810 "flags:\t0%o\n",
1811 (long long) file->f_pos,
1812 f_flags);
1813 spin_unlock(&files->file_lock);
1814 put_files_struct(files);
1815 return 0;
99f89551 1816 }
5e442a49
LT		 1817 spin_unlock(&files->file_lock);
1818 put_files_struct(files);
1819 }
1820 return -ENOENT;
99f89551
EB		 1821}
1822
3dcd25f3 1823static int proc_fd_link(struct inode *inode, struct path *path)
27932742 1824{
3dcd25f3 1825 return proc_fd_info(inode, path, NULL);
27932742
MS		 1826}
1827
1da177e4
LT		 1828static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1829{
34286d66
NP		 1830 struct inode *inode;
1831 struct task_struct *task;
1832 int fd;
1da177e4 1833 struct files_struct *files;
c69e8d9c 1834 const struct cred *cred;
1da177e4 1835
34286d66
NP		 1836 if (nd && nd->flags & LOOKUP_RCU)
1837 return -ECHILD;
1838
1839 inode = dentry->d_inode;
1840 task = get_proc_task(inode);
1841 fd = proc_fd(inode);
1842
99f89551
EB		 1843 if (task) {
1844 files = get_files_struct(task);
1845 if (files) {
1846 rcu_read_lock();
1847 if (fcheck_files(files, fd)) {
1848 rcu_read_unlock();
1849 put_files_struct(files);
1850 if (task_dumpable(task)) {
c69e8d9c
DH		 1851 rcu_read_lock();
1852 cred = __task_cred(task);
1853 inode->i_uid = cred->euid;
1854 inode->i_gid = cred->egid;
1855 rcu_read_unlock();
99f89551
EB		 1856 } else {
1857 inode->i_uid = 0;
1858 inode->i_gid = 0;
1859 }
9ee8ab9f 1860 inode->i_mode &= ~(S_ISUID | S_ISGID);
99f89551
EB		 1861 security_task_to_inode(task, inode);
1862 put_task_struct(task);
1863 return 1;
1864 }
b835996f 1865 rcu_read_unlock();
1da177e4 1866 put_files_struct(files);
1da177e4 1867 }
99f89551 1868 put_task_struct(task);
1da177e4
LT		 1869 }
1870 d_drop(dentry);
1871 return 0;
1872}
1873
d72f71eb 1874static const struct dentry_operations tid_fd_dentry_operations =
1da177e4
LT		 1875{
1876 .d_revalidate = tid_fd_revalidate,
1877 .d_delete = pid_delete_dentry,
1878};
1879
444ceed8 1880static struct dentry *proc_fd_instantiate(struct inode *dir,
c5141e6d 1881 struct dentry *dentry, struct task_struct *task, const void *ptr)
1da177e4 1882{
c5141e6d 1883 unsigned fd = *(const unsigned *)ptr;
444ceed8
EB		 1884 struct file *file;
1885 struct files_struct *files;
1886 struct inode *inode;
1887 struct proc_inode *ei;
1888 struct dentry *error = ERR_PTR(-ENOENT);
1da177e4 1889
61a28784 1890 inode = proc_pid_make_inode(dir->i_sb, task);
1da177e4
LT		 1891 if (!inode)
1892 goto out;
1893 ei = PROC_I(inode);
aed7a6c4 1894 ei->fd = fd;
1da177e4
LT		 1895 files = get_files_struct(task);
1896 if (!files)
444ceed8 1897 goto out_iput;
1da177e4 1898 inode->i_mode = S_IFLNK;
ca99c1da
DS		 1899
1900 /*
1901 * We are not taking a ref to the file structure, so we must
1902 * hold ->file_lock.
1903 */
1904 spin_lock(&files->file_lock);
1da177e4
LT		 1905 file = fcheck_files(files, fd);
1906 if (!file)
444ceed8 1907 goto out_unlock;
aeb5d727 1908 if (file->f_mode & FMODE_READ)
1da177e4 1909 inode->i_mode |= S_IRUSR | S_IXUSR;
aeb5d727 1910 if (file->f_mode & FMODE_WRITE)
1da177e4 1911 inode->i_mode |= S_IWUSR | S_IXUSR;
ca99c1da 1912 spin_unlock(&files->file_lock);
1da177e4 1913 put_files_struct(files);
444ceed8 1914
5e442a49 1915 inode->i_op = &proc_pid_link_inode_operations;
1da177e4
LT		 1916 inode->i_size = 64;
1917 ei->op.proc_get_link = proc_fd_link;
fb045adb 1918 d_set_d_op(dentry, &tid_fd_dentry_operations);
1da177e4 1919 d_add(dentry, inode);
cd6a3ce9
EB		 1920 /* Close the race of the process dying before we return the dentry */
1921 if (tid_fd_revalidate(dentry, NULL))
444ceed8 1922 error = NULL;
1da177e4 1923
444ceed8
EB		 1924 out:
1925 return error;
1926out_unlock:
ca99c1da 1927 spin_unlock(&files->file_lock);
1da177e4 1928 put_files_struct(files);
444ceed8 1929out_iput:
1da177e4 1930 iput(inode);
cd6a3ce9 1931 goto out;
1da177e4
LT		 1932}
1933
27932742
MS		 1934static struct dentry *proc_lookupfd_common(struct inode *dir,
1935 struct dentry *dentry,
1936 instantiate_t instantiate)
444ceed8
EB		 1937{
1938 struct task_struct *task = get_proc_task(dir);
1939 unsigned fd = name_to_int(dentry);
1940 struct dentry *result = ERR_PTR(-ENOENT);
1941
1942 if (!task)
1943 goto out_no_task;
1944 if (fd == ~0U)
1945 goto out;
1946
27932742 1947 result = instantiate(dir, dentry, task, &fd);
444ceed8
EB		 1948out:
1949 put_task_struct(task);
1950out_no_task:
1951 return result;
1952}
1953
27932742
MS		 1954static int proc_readfd_common(struct file * filp, void * dirent,
1955 filldir_t filldir, instantiate_t instantiate)
28a6d671 1956{
2fddfeef 1957 struct dentry *dentry = filp->f_path.dentry;
28a6d671
EB		 1958 struct inode *inode = dentry->d_inode;
1959 struct task_struct *p = get_proc_task(inode);
457c2510 1960 unsigned int fd, ino;
28a6d671 1961 int retval;
28a6d671 1962 struct files_struct * files;
1da177e4 1963
28a6d671
EB		 1964 retval = -ENOENT;
1965 if (!p)
1966 goto out_no_task;
1967 retval = 0;
28a6d671
EB		 1968
1969 fd = filp->f_pos;
1970 switch (fd) {
1971 case 0:
1972 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
5e442a49 1973 goto out;
28a6d671
EB		 1974 filp->f_pos++;
1975 case 1:
1976 ino = parent_ino(dentry);
1977 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
5e442a49 1978 goto out;
28a6d671
EB		 1979 filp->f_pos++;
1980 default:
1981 files = get_files_struct(p);
1982 if (!files)
5e442a49 1983 goto out;
28a6d671 1984 rcu_read_lock();
28a6d671 1985 for (fd = filp->f_pos-2;
9b4f526c 1986 fd < files_fdtable(files)->max_fds;
28a6d671 1987 fd++, filp->f_pos++) {
27932742
MS		 1988 char name[PROC_NUMBUF];
1989 int len;
28a6d671
EB		 1990
1991 if (!fcheck_files(files, fd))
1992 continue;
1993 rcu_read_unlock();
1994
27932742
MS		 1995 len = snprintf(name, sizeof(name), "%d", fd);
1996 if (proc_fill_cache(filp, dirent, filldir,
1997 name, len, instantiate,
1998 p, &fd) < 0) {
28a6d671
EB		 1999 rcu_read_lock();
2000 break;
2001 }
2002 rcu_read_lock();
2003 }
2004 rcu_read_unlock();
2005 put_files_struct(files);
2006 }
2007out:
2008 put_task_struct(p);
2009out_no_task:
2010 return retval;
2011}
2012
27932742
MS		 2013static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
2014 struct nameidata *nd)
2015{
2016 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
2017}
2018
2019static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
2020{
2021 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
2022}
2023
2024static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
2025 size_t len, loff_t *ppos)
2026{
2027 char tmp[PROC_FDINFO_MAX];
3dcd25f3 2028 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
27932742
MS		 2029 if (!err)
2030 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
2031 return err;
2032}
2033
2034static const struct file_operations proc_fdinfo_file_operations = {
87df8424 2035 .open = nonseekable_open,
27932742 2036 .read = proc_fdinfo_read,
6038f373 2037 .llseek = no_llseek,
27932742
MS		 2038};
2039
00977a59 2040static const struct file_operations proc_fd_operations = {
28a6d671
EB		 2041 .read = generic_read_dir,
2042 .readdir = proc_readfd,
6038f373 2043 .llseek = default_llseek,
1da177e4
LT		 2044};
2045
8948e11f
AD		 2046/*
2047 * /proc/pid/fd needs a special permission handler so that a process can still
2048 * access /proc/self/fd after it has executed a setuid().
2049 */
10556cb2 2050static int proc_fd_permission(struct inode *inode, int mask)
8948e11f 2051{
2830ba7f 2052 int rv = generic_permission(inode, mask);
8948e11f
AD		 2053 if (rv == 0)
2054 return 0;
2055 if (task_pid(current) == proc_pid(inode))
2056 rv = 0;
2057 return rv;
2058}
2059
1da177e4
LT		 2060/*
2061 * proc directories can do almost nothing..
2062 */
c5ef1c42 2063static const struct inode_operations proc_fd_inode_operations = {
1da177e4 2064 .lookup = proc_lookupfd,
8948e11f 2065 .permission = proc_fd_permission,
6d76fa58 2066 .setattr = proc_setattr,
1da177e4
LT		 2067};
2068
27932742
MS		 2069static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2070 struct dentry *dentry, struct task_struct *task, const void *ptr)
2071{
2072 unsigned fd = *(unsigned *)ptr;
2073 struct inode *inode;
2074 struct proc_inode *ei;
2075 struct dentry *error = ERR_PTR(-ENOENT);
2076
2077 inode = proc_pid_make_inode(dir->i_sb, task);
2078 if (!inode)
2079 goto out;
2080 ei = PROC_I(inode);
2081 ei->fd = fd;
2082 inode->i_mode = S_IFREG | S_IRUSR;
2083 inode->i_fop = &proc_fdinfo_file_operations;
fb045adb 2084 d_set_d_op(dentry, &tid_fd_dentry_operations);
27932742
MS		 2085 d_add(dentry, inode);
2086 /* Close the race of the process dying before we return the dentry */
2087 if (tid_fd_revalidate(dentry, NULL))
2088 error = NULL;
2089
2090 out:
2091 return error;
2092}
2093
2094static struct dentry *proc_lookupfdinfo(struct inode *dir,
2095 struct dentry *dentry,
2096 struct nameidata *nd)
2097{
2098 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2099}
2100
2101static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2102{
2103 return proc_readfd_common(filp, dirent, filldir,
2104 proc_fdinfo_instantiate);
2105}
2106
2107static const struct file_operations proc_fdinfo_operations = {
2108 .read = generic_read_dir,
2109 .readdir = proc_readfdinfo,
6038f373 2110 .llseek = default_llseek,
27932742
MS		 2111};
2112
2113/*
2114 * proc directories can do almost nothing..
2115 */
2116static const struct inode_operations proc_fdinfo_inode_operations = {
2117 .lookup = proc_lookupfdinfo,
2118 .setattr = proc_setattr,
2119};
2120
2121
444ceed8 2122static struct dentry *proc_pident_instantiate(struct inode *dir,
c5141e6d 2123 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8 2124{
c5141e6d 2125 const struct pid_entry *p = ptr;
444ceed8
EB		 2126 struct inode *inode;
2127 struct proc_inode *ei;
bd6daba9 2128 struct dentry *error = ERR_PTR(-ENOENT);
444ceed8 2129
61a28784 2130 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 2131 if (!inode)
2132 goto out;
2133
2134 ei = PROC_I(inode);
2135 inode->i_mode = p->mode;
2136 if (S_ISDIR(inode->i_mode))
bfe86848 2137 set_nlink(inode, 2); /* Use getattr to fix if necessary */
444ceed8
EB		 2138 if (p->iop)
2139 inode->i_op = p->iop;
2140 if (p->fop)
2141 inode->i_fop = p->fop;
2142 ei->op = p->op;
fb045adb 2143 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 2144 d_add(dentry, inode);
2145 /* Close the race of the process dying before we return the dentry */
2146 if (pid_revalidate(dentry, NULL))
2147 error = NULL;
2148out:
2149 return error;
2150}
2151
1da177e4
LT		 2152static struct dentry *proc_pident_lookup(struct inode *dir,
2153 struct dentry *dentry,
c5141e6d 2154 const struct pid_entry *ents,
7bcd6b0e 2155 unsigned int nents)
1da177e4 2156{
cd6a3ce9 2157 struct dentry *error;
99f89551 2158 struct task_struct *task = get_proc_task(dir);
c5141e6d 2159 const struct pid_entry *p, *last;
1da177e4 2160
cd6a3ce9 2161 error = ERR_PTR(-ENOENT);
1da177e4 2162
99f89551
EB		 2163 if (!task)
2164 goto out_no_task;
1da177e4 2165
20cdc894
EB		 2166 /*
2167 * Yes, it does not scale. And it should not. Don't add
2168 * new entries into /proc/<tgid>/ without very good reasons.
2169 */
7bcd6b0e
EB		 2170 last = &ents[nents - 1];
2171 for (p = ents; p <= last; p++) {
1da177e4
LT		 2172 if (p->len != dentry->d_name.len)
2173 continue;
2174 if (!memcmp(dentry->d_name.name, p->name, p->len))
2175 break;
2176 }
7bcd6b0e 2177 if (p > last)
1da177e4
LT		 2178 goto out;
2179
444ceed8 2180 error = proc_pident_instantiate(dir, dentry, task, p);
1da177e4 2181out:
99f89551
EB		 2182 put_task_struct(task);
2183out_no_task:
cd6a3ce9 2184 return error;
1da177e4
LT		 2185}
2186
c5141e6d
ED		 2187static int proc_pident_fill_cache(struct file *filp, void *dirent,
2188 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
61a28784
EB		 2189{
2190 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2191 proc_pident_instantiate, task, p);
2192}
2193
28a6d671
EB		 2194static int proc_pident_readdir(struct file *filp,
2195 void *dirent, filldir_t filldir,
c5141e6d 2196 const struct pid_entry *ents, unsigned int nents)
28a6d671
EB		 2197{
2198 int i;
2fddfeef 2199 struct dentry *dentry = filp->f_path.dentry;
28a6d671
EB		 2200 struct inode *inode = dentry->d_inode;
2201 struct task_struct *task = get_proc_task(inode);
c5141e6d 2202 const struct pid_entry *p, *last;
28a6d671
EB		 2203 ino_t ino;
2204 int ret;
2205
2206 ret = -ENOENT;
2207 if (!task)
61a28784 2208 goto out_no_task;
28a6d671
EB		 2209
2210 ret = 0;
28a6d671
EB		 2211 i = filp->f_pos;
2212 switch (i) {
2213 case 0:
2214 ino = inode->i_ino;
2215 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2216 goto out;
2217 i++;
2218 filp->f_pos++;
2219 /* fall through */
2220 case 1:
2221 ino = parent_ino(dentry);
2222 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2223 goto out;
2224 i++;
2225 filp->f_pos++;
2226 /* fall through */
2227 default:
2228 i -= 2;
2229 if (i >= nents) {
2230 ret = 1;
2231 goto out;
2232 }
2233 p = ents + i;
7bcd6b0e
EB		 2234 last = &ents[nents - 1];
2235 while (p <= last) {
61a28784 2236 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
28a6d671
EB		 2237 goto out;
2238 filp->f_pos++;
2239 p++;
2240 }
2241 }
2242
2243 ret = 1;
2244out:
61a28784
EB		 2245 put_task_struct(task);
2246out_no_task:
28a6d671 2247 return ret;
1da177e4
LT		 2248}
2249
28a6d671
EB		 2250#ifdef CONFIG_SECURITY
2251static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2252 size_t count, loff_t *ppos)
2253{
2fddfeef 2254 struct inode * inode = file->f_path.dentry->d_inode;
04ff9708 2255 char *p = NULL;
28a6d671
EB		 2256 ssize_t length;
2257 struct task_struct *task = get_proc_task(inode);
2258
28a6d671 2259 if (!task)
04ff9708 2260 return -ESRCH;
28a6d671
EB		 2261
2262 length = security_getprocattr(task,
2fddfeef 2263 (char*)file->f_path.dentry->d_name.name,
04ff9708 2264 &p);
28a6d671 2265 put_task_struct(task);
04ff9708
AV		 2266 if (length > 0)
2267 length = simple_read_from_buffer(buf, count, ppos, p, length);
2268 kfree(p);
28a6d671 2269 return length;
1da177e4
LT		 2270}
2271
28a6d671
EB		 2272static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2273 size_t count, loff_t *ppos)
2274{
2fddfeef 2275 struct inode * inode = file->f_path.dentry->d_inode;
28a6d671
EB		 2276 char *page;
2277 ssize_t length;
2278 struct task_struct *task = get_proc_task(inode);
2279
2280 length = -ESRCH;
2281 if (!task)
2282 goto out_no_task;
2283 if (count > PAGE_SIZE)
2284 count = PAGE_SIZE;
2285
2286 /* No partial writes. */
2287 length = -EINVAL;
2288 if (*ppos != 0)
2289 goto out;
2290
2291 length = -ENOMEM;
e12ba74d 2292 page = (char*)__get_free_page(GFP_TEMPORARY);
28a6d671
EB		 2293 if (!page)
2294 goto out;
2295
2296 length = -EFAULT;
2297 if (copy_from_user(page, buf, count))
2298 goto out_free;
2299
107db7c7 2300 /* Guard against adverse ptrace interaction */
9b1bf12d 2301 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
107db7c7
DH		 2302 if (length < 0)
2303 goto out_free;
2304
28a6d671 2305 length = security_setprocattr(task,
2fddfeef 2306 (char*)file->f_path.dentry->d_name.name,
28a6d671 2307 (void*)page, count);
9b1bf12d 2308 mutex_unlock(&task->signal->cred_guard_mutex);
28a6d671
EB		 2309out_free:
2310 free_page((unsigned long) page);
2311out:
2312 put_task_struct(task);
2313out_no_task:
2314 return length;
2315}
2316
00977a59 2317static const struct file_operations proc_pid_attr_operations = {
28a6d671
EB		 2318 .read = proc_pid_attr_read,
2319 .write = proc_pid_attr_write,
87df8424 2320 .llseek = generic_file_llseek,
28a6d671
EB		 2321};
2322
c5141e6d 2323static const struct pid_entry attr_dir_stuff[] = {
631f9c18
AD		 2324 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2325 REG("prev", S_IRUGO, proc_pid_attr_operations),
2326 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2327 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2328 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2329 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
28a6d671
EB		 2330};
2331
72d9dcfc 2332static int proc_attr_dir_readdir(struct file * filp,
28a6d671
EB		 2333 void * dirent, filldir_t filldir)
2334{
2335 return proc_pident_readdir(filp,dirent,filldir,
72d9dcfc 2336 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2337}
2338
00977a59 2339static const struct file_operations proc_attr_dir_operations = {
1da177e4 2340 .read = generic_read_dir,
72d9dcfc 2341 .readdir = proc_attr_dir_readdir,
6038f373 2342 .llseek = default_llseek,
1da177e4
LT		 2343};
2344
72d9dcfc 2345static struct dentry *proc_attr_dir_lookup(struct inode *dir,
28a6d671
EB		 2346 struct dentry *dentry, struct nameidata *nd)
2347{
7bcd6b0e
EB		 2348 return proc_pident_lookup(dir, dentry,
2349 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2350}
2351
c5ef1c42 2352static const struct inode_operations proc_attr_dir_inode_operations = {
72d9dcfc 2353 .lookup = proc_attr_dir_lookup,
99f89551 2354 .getattr = pid_getattr,
6d76fa58 2355 .setattr = proc_setattr,
1da177e4
LT		 2356};
2357
28a6d671
EB		 2358#endif
2359
698ba7b5 2360#ifdef CONFIG_ELF_CORE
3cb4a0bb
KH		 2361static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2362 size_t count, loff_t *ppos)
2363{
2364 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2365 struct mm_struct *mm;
2366 char buffer[PROC_NUMBUF];
2367 size_t len;
2368 int ret;
2369
2370 if (!task)
2371 return -ESRCH;
2372
2373 ret = 0;
2374 mm = get_task_mm(task);
2375 if (mm) {
2376 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2377 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2378 MMF_DUMP_FILTER_SHIFT));
2379 mmput(mm);
2380 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2381 }
2382
2383 put_task_struct(task);
2384
2385 return ret;
2386}
2387
2388static ssize_t proc_coredump_filter_write(struct file *file,
2389 const char __user *buf,
2390 size_t count,
2391 loff_t *ppos)
2392{
2393 struct task_struct *task;
2394 struct mm_struct *mm;
2395 char buffer[PROC_NUMBUF], *end;
2396 unsigned int val;
2397 int ret;
2398 int i;
2399 unsigned long mask;
2400
2401 ret = -EFAULT;
2402 memset(buffer, 0, sizeof(buffer));
2403 if (count > sizeof(buffer) - 1)
2404 count = sizeof(buffer) - 1;
2405 if (copy_from_user(buffer, buf, count))
2406 goto out_no_task;
2407
2408 ret = -EINVAL;
2409 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2410 if (*end == '\n')
2411 end++;
2412 if (end - buffer == 0)
2413 goto out_no_task;
2414
2415 ret = -ESRCH;
2416 task = get_proc_task(file->f_dentry->d_inode);
2417 if (!task)
2418 goto out_no_task;
2419
2420 ret = end - buffer;
2421 mm = get_task_mm(task);
2422 if (!mm)
2423 goto out_no_mm;
2424
2425 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2426 if (val & mask)
2427 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2428 else
2429 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2430 }
2431
2432 mmput(mm);
2433 out_no_mm:
2434 put_task_struct(task);
2435 out_no_task:
2436 return ret;
2437}
2438
2439static const struct file_operations proc_coredump_filter_operations = {
2440 .read = proc_coredump_filter_read,
2441 .write = proc_coredump_filter_write,
87df8424 2442 .llseek = generic_file_llseek,
3cb4a0bb
KH		 2443};
2444#endif
2445
28a6d671
EB		 2446/*
2447 * /proc/self:
2448 */
2449static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2450 int buflen)
2451{
488e5bc4 2452 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
b55fcb22 2453 pid_t tgid = task_tgid_nr_ns(current, ns);
28a6d671 2454 char tmp[PROC_NUMBUF];
b55fcb22 2455 if (!tgid)
488e5bc4 2456 return -ENOENT;
b55fcb22 2457 sprintf(tmp, "%d", tgid);
28a6d671
EB		 2458 return vfs_readlink(dentry,buffer,buflen,tmp);
2459}
2460
2461static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2462{
488e5bc4 2463 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
b55fcb22 2464 pid_t tgid = task_tgid_nr_ns(current, ns);
7fee4868
AV		 2465 char *name = ERR_PTR(-ENOENT);
2466 if (tgid) {
2467 name = __getname();
2468 if (!name)
2469 name = ERR_PTR(-ENOMEM);
2470 else
2471 sprintf(name, "%d", tgid);
2472 }
2473 nd_set_link(nd, name);
2474 return NULL;
2475}
2476
2477static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2478 void *cookie)
2479{
2480 char *s = nd_get_link(nd);
2481 if (!IS_ERR(s))
2482 __putname(s);
28a6d671
EB		 2483}
2484
c5ef1c42 2485static const struct inode_operations proc_self_inode_operations = {
28a6d671
EB		 2486 .readlink = proc_self_readlink,
2487 .follow_link = proc_self_follow_link,
7fee4868 2488 .put_link = proc_self_put_link,
28a6d671
EB		 2489};
2490
801199ce
EB		 2491/*
2492 * proc base
2493 *
2494 * These are the directory entries in the root directory of /proc
2495 * that properly belong to the /proc filesystem, as they describe
2496 * describe something that is process related.
2497 */
c5141e6d 2498static const struct pid_entry proc_base_stuff[] = {
61a28784 2499 NOD("self", S_IFLNK|S_IRWXUGO,
801199ce 2500 &proc_self_inode_operations, NULL, {}),
801199ce
EB		 2501};
2502
444ceed8 2503static struct dentry *proc_base_instantiate(struct inode *dir,
c5141e6d 2504 struct dentry *dentry, struct task_struct *task, const void *ptr)
801199ce 2505{
c5141e6d 2506 const struct pid_entry *p = ptr;
801199ce 2507 struct inode *inode;
801199ce 2508 struct proc_inode *ei;
73d36460 2509 struct dentry *error;
801199ce
EB		 2510
2511 /* Allocate the inode */
2512 error = ERR_PTR(-ENOMEM);
2513 inode = new_inode(dir->i_sb);
2514 if (!inode)
2515 goto out;
2516
2517 /* Initialize the inode */
2518 ei = PROC_I(inode);
85fe4025 2519 inode->i_ino = get_next_ino();
801199ce 2520 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
801199ce
EB		 2521
2522 /*
2523 * grab the reference to the task.
2524 */
1a657f78 2525 ei->pid = get_task_pid(task, PIDTYPE_PID);
801199ce
EB		 2526 if (!ei->pid)
2527 goto out_iput;
2528
801199ce
EB		 2529 inode->i_mode = p->mode;
2530 if (S_ISDIR(inode->i_mode))
bfe86848 2531 set_nlink(inode, 2);
801199ce
EB		 2532 if (S_ISLNK(inode->i_mode))
2533 inode->i_size = 64;
2534 if (p->iop)
2535 inode->i_op = p->iop;
2536 if (p->fop)
2537 inode->i_fop = p->fop;
2538 ei->op = p->op;
801199ce
EB		 2539 d_add(dentry, inode);
2540 error = NULL;
2541out:
801199ce
EB		 2542 return error;
2543out_iput:
2544 iput(inode);
2545 goto out;
2546}
2547
444ceed8
EB		 2548static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2549{
2550 struct dentry *error;
2551 struct task_struct *task = get_proc_task(dir);
c5141e6d 2552 const struct pid_entry *p, *last;
444ceed8
EB		 2553
2554 error = ERR_PTR(-ENOENT);
2555
2556 if (!task)
2557 goto out_no_task;
2558
2559 /* Lookup the directory entry */
7bcd6b0e
EB		 2560 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2561 for (p = proc_base_stuff; p <= last; p++) {
444ceed8
EB		 2562 if (p->len != dentry->d_name.len)
2563 continue;
2564 if (!memcmp(dentry->d_name.name, p->name, p->len))
2565 break;
2566 }
7bcd6b0e 2567 if (p > last)
444ceed8
EB		 2568 goto out;
2569
2570 error = proc_base_instantiate(dir, dentry, task, p);
2571
2572out:
2573 put_task_struct(task);
2574out_no_task:
2575 return error;
2576}
2577
c5141e6d
ED		 2578static int proc_base_fill_cache(struct file *filp, void *dirent,
2579 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
61a28784
EB		 2580{
2581 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2582 proc_base_instantiate, task, p);
2583}
2584
aba76fdb 2585#ifdef CONFIG_TASK_IO_ACCOUNTING
297c5d92
AR		 2586static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2587{
940389b8 2588 struct task_io_accounting acct = task->ioac;
5995477a 2589 unsigned long flags;
293eb1e7 2590 int result;
5995477a 2591
293eb1e7
VK		 2592 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2593 if (result)
2594 return result;
2595
2596 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2597 result = -EACCES;
2598 goto out_unlock;
2599 }
1d1221f3 2600
5995477a
AR		 2601 if (whole && lock_task_sighand(task, &flags)) {
2602 struct task_struct *t = task;
2603
2604 task_io_accounting_add(&acct, &task->signal->ioac);
2605 while_each_thread(task, t)
2606 task_io_accounting_add(&acct, &t->ioac);
2607
2608 unlock_task_sighand(task, &flags);
297c5d92 2609 }
293eb1e7 2610 result = sprintf(buffer,
aba76fdb
AM		 2611 "rchar: %llu\n"
2612 "wchar: %llu\n"
2613 "syscr: %llu\n"
2614 "syscw: %llu\n"
2615 "read_bytes: %llu\n"
2616 "write_bytes: %llu\n"
2617 "cancelled_write_bytes: %llu\n",
7c44319d
AB		 2618 (unsigned long long)acct.rchar,
2619 (unsigned long long)acct.wchar,
2620 (unsigned long long)acct.syscr,
2621 (unsigned long long)acct.syscw,
2622 (unsigned long long)acct.read_bytes,
2623 (unsigned long long)acct.write_bytes,
2624 (unsigned long long)acct.cancelled_write_bytes);
293eb1e7
VK		 2625out_unlock:
2626 mutex_unlock(&task->signal->cred_guard_mutex);
2627 return result;
297c5d92
AR		 2628}
2629
2630static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2631{
2632 return do_io_accounting(task, buffer, 0);
aba76fdb 2633}
297c5d92
AR		 2634
2635static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2636{
2637 return do_io_accounting(task, buffer, 1);
2638}
2639#endif /* CONFIG_TASK_IO_ACCOUNTING */
aba76fdb 2640
47830723
KC		 2641static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2642 struct pid *pid, struct task_struct *task)
2643{
a9712bc1
AV		 2644 int err = lock_trace(task);
2645 if (!err) {
2646 seq_printf(m, "%08x\n", task->personality);
2647 unlock_trace(task);
2648 }
2649 return err;
47830723
KC		 2650}
2651
28a6d671
EB		 2652/*
2653 * Thread groups
2654 */
00977a59 2655static const struct file_operations proc_task_operations;
c5ef1c42 2656static const struct inode_operations proc_task_inode_operations;
20cdc894 2657
c5141e6d 2658static const struct pid_entry tgid_base_stuff[] = {
631f9c18
AD		 2659 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2660 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2661 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 2662 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
b2211a36 2663#ifdef CONFIG_NET
631f9c18 2664 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
b2211a36 2665#endif
631f9c18
AD		 2666 REG("environ", S_IRUSR, proc_environ_operations),
2667 INF("auxv", S_IRUSR, proc_pid_auxv),
2668 ONE("status", S_IRUGO, proc_pid_status),
a9712bc1 2669 ONE("personality", S_IRUGO, proc_pid_personality),
3036e7b4 2670 INF("limits", S_IRUGO, proc_pid_limits),
43ae34cb 2671#ifdef CONFIG_SCHED_DEBUG
631f9c18 2672 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
5091faa4
MG		 2673#endif
2674#ifdef CONFIG_SCHED_AUTOGROUP
2675 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
ebcb6734 2676#endif
4614a696 2677 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 2678#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
a9712bc1 2679 INF("syscall", S_IRUGO, proc_pid_syscall),
43ae34cb 2680#endif
631f9c18
AD		 2681 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2682 ONE("stat", S_IRUGO, proc_tgid_stat),
2683 ONE("statm", S_IRUGO, proc_pid_statm),
2684 REG("maps", S_IRUGO, proc_maps_operations),
28a6d671 2685#ifdef CONFIG_NUMA
631f9c18 2686 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
28a6d671 2687#endif
631f9c18
AD		 2688 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2689 LNK("cwd", proc_cwd_link),
2690 LNK("root", proc_root_link),
2691 LNK("exe", proc_exe_link),
2692 REG("mounts", S_IRUGO, proc_mounts_operations),
2693 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2694 REG("mountstats", S_IRUSR, proc_mountstats_operations),
1e883281 2695#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18
AD		 2696 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2697 REG("smaps", S_IRUGO, proc_smaps_operations),
ca6b0bf0 2698 REG("pagemap", S_IRUGO, proc_pagemap_operations),
28a6d671
EB		 2699#endif
2700#ifdef CONFIG_SECURITY
631f9c18 2701 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 2702#endif
2703#ifdef CONFIG_KALLSYMS
631f9c18 2704 INF("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 2705#endif
2ec220e2 2706#ifdef CONFIG_STACKTRACE
a9712bc1 2707 ONE("stack", S_IRUGO, proc_pid_stack),
28a6d671
EB		 2708#endif
2709#ifdef CONFIG_SCHEDSTATS
631f9c18 2710 INF("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 2711#endif
9745512c 2712#ifdef CONFIG_LATENCYTOP
631f9c18 2713 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 2714#endif
8793d854 2715#ifdef CONFIG_PROC_PID_CPUSET
631f9c18 2716 REG("cpuset", S_IRUGO, proc_cpuset_operations),
a424316c
PM		 2717#endif
2718#ifdef CONFIG_CGROUPS
631f9c18 2719 REG("cgroup", S_IRUGO, proc_cgroup_operations),
28a6d671 2720#endif
631f9c18
AD		 2721 INF("oom_score", S_IRUGO, proc_oom_score),
2722 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
a63d83f4 2723 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 2724#ifdef CONFIG_AUDITSYSCALL
631f9c18
AD		 2725 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2726 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 2727#endif
f4f154fd 2728#ifdef CONFIG_FAULT_INJECTION
631f9c18 2729 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 2730#endif
698ba7b5 2731#ifdef CONFIG_ELF_CORE
631f9c18 2732 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3cb4a0bb 2733#endif
aba76fdb 2734#ifdef CONFIG_TASK_IO_ACCOUNTING
1d1221f3 2735 INF("io", S_IRUSR, proc_tgid_io_accounting),
aba76fdb 2736#endif
f133ecca
CM		 2737#ifdef CONFIG_HARDWALL
2738 INF("hardwall", S_IRUGO, proc_pid_hardwall),
2739#endif
28a6d671 2740};
1da177e4 2741
28a6d671 2742static int proc_tgid_base_readdir(struct file * filp,
1da177e4
LT		 2743 void * dirent, filldir_t filldir)
2744{
2745 return proc_pident_readdir(filp,dirent,filldir,
28a6d671 2746 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 2747}
2748
00977a59 2749static const struct file_operations proc_tgid_base_operations = {
1da177e4 2750 .read = generic_read_dir,
28a6d671 2751 .readdir = proc_tgid_base_readdir,
6038f373 2752 .llseek = default_llseek,
1da177e4
LT		 2753};
2754
28a6d671 2755static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
7bcd6b0e
EB		 2756 return proc_pident_lookup(dir, dentry,
2757 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 2758}
2759
c5ef1c42 2760static const struct inode_operations proc_tgid_base_inode_operations = {
28a6d671 2761 .lookup = proc_tgid_base_lookup,
99f89551 2762 .getattr = pid_getattr,
6d76fa58 2763 .setattr = proc_setattr,
1da177e4 2764};
1da177e4 2765
60347f67 2766static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
1da177e4 2767{
48e6484d 2768 struct dentry *dentry, *leader, *dir;
8578cea7 2769 char buf[PROC_NUMBUF];
48e6484d
EB		 2770 struct qstr name;
2771
2772 name.name = buf;
60347f67
PE		 2773 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2774 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d 2775 if (dentry) {
29f12ca3 2776 shrink_dcache_parent(dentry);
48e6484d
EB		 2777 d_drop(dentry);
2778 dput(dentry);
2779 }
1da177e4 2780
48e6484d 2781 name.name = buf;
60347f67
PE		 2782 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2783 leader = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d
EB		 2784 if (!leader)
2785 goto out;
1da177e4 2786
48e6484d
EB		 2787 name.name = "task";
2788 name.len = strlen(name.name);
2789 dir = d_hash_and_lookup(leader, &name);
2790 if (!dir)
2791 goto out_put_leader;
2792
2793 name.name = buf;
60347f67 2794 name.len = snprintf(buf, sizeof(buf), "%d", pid);
48e6484d
EB		 2795 dentry = d_hash_and_lookup(dir, &name);
2796 if (dentry) {
2797 shrink_dcache_parent(dentry);
2798 d_drop(dentry);
2799 dput(dentry);
1da177e4 2800 }
48e6484d
EB		 2801
2802 dput(dir);
2803out_put_leader:
2804 dput(leader);
2805out:
2806 return;
1da177e4
LT		 2807}
2808
0895e91d
RD		 2809/**
2810 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2811 * @task: task that should be flushed.
2812 *
2813 * When flushing dentries from proc, one needs to flush them from global
60347f67 2814 * proc (proc_mnt) and from all the namespaces' procs this task was seen
0895e91d
RD		 2815 * in. This call is supposed to do all of this job.
2816 *
2817 * Looks in the dcache for
2818 * /proc/@pid
2819 * /proc/@tgid/task/@pid
2820 * if either directory is present flushes it and all of it'ts children
2821 * from the dcache.
2822 *
2823 * It is safe and reasonable to cache /proc entries for a task until
2824 * that task exits. After that they just clog up the dcache with
2825 * useless entries, possibly causing useful dcache entries to be
2826 * flushed instead. This routine is proved to flush those useless
2827 * dcache entries at process exit time.
2828 *
2829 * NOTE: This routine is just an optimization so it does not guarantee
2830 * that no dcache entries will exist at process exit time it
2831 * just makes it very unlikely that any will persist.
60347f67
PE		 2832 */
2833
2834void proc_flush_task(struct task_struct *task)
2835{
9fcc2d15 2836 int i;
9b4d1cbe 2837 struct pid *pid, *tgid;
130f77ec
PE		 2838 struct upid *upid;
2839
130f77ec 2840 pid = task_pid(task);
9b4d1cbe 2841 tgid = task_tgid(task);
130f77ec 2842
9fcc2d15 2843 for (i = 0; i <= pid->level; i++) {
130f77ec
PE		 2844 upid = &pid->numbers[i];
2845 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
9b4d1cbe 2846 tgid->numbers[i].nr);
130f77ec 2847 }
6f4e6433
PE		 2848
2849 upid = &pid->numbers[pid->level];
2850 if (upid->nr == 1)
2851 pid_ns_release_proc(upid->ns);
60347f67
PE		 2852}
2853
9711ef99
AB		 2854static struct dentry *proc_pid_instantiate(struct inode *dir,
2855 struct dentry * dentry,
c5141e6d 2856 struct task_struct *task, const void *ptr)
444ceed8
EB		 2857{
2858 struct dentry *error = ERR_PTR(-ENOENT);
2859 struct inode *inode;
2860
61a28784 2861 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 2862 if (!inode)
2863 goto out;
2864
2865 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2866 inode->i_op = &proc_tgid_base_inode_operations;
2867 inode->i_fop = &proc_tgid_base_operations;
2868 inode->i_flags|=S_IMMUTABLE;
aed54175 2869
bfe86848
MS		 2870 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
2871 ARRAY_SIZE(tgid_base_stuff)));
444ceed8 2872
fb045adb 2873 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 2874
2875 d_add(dentry, inode);
2876 /* Close the race of the process dying before we return the dentry */
2877 if (pid_revalidate(dentry, NULL))
2878 error = NULL;
2879out:
2880 return error;
2881}
2882
1da177e4
LT		 2883struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2884{
73d36460 2885 struct dentry *result;
1da177e4 2886 struct task_struct *task;
1da177e4 2887 unsigned tgid;
b488893a 2888 struct pid_namespace *ns;
1da177e4 2889
801199ce
EB		 2890 result = proc_base_lookup(dir, dentry);
2891 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2892 goto out;
2893
1da177e4
LT		 2894 tgid = name_to_int(dentry);
2895 if (tgid == ~0U)
2896 goto out;
2897
b488893a 2898 ns = dentry->d_sb->s_fs_info;
de758734 2899 rcu_read_lock();
b488893a 2900 task = find_task_by_pid_ns(tgid, ns);
1da177e4
LT		 2901 if (task)
2902 get_task_struct(task);
de758734 2903 rcu_read_unlock();
1da177e4
LT		 2904 if (!task)
2905 goto out;
2906
444ceed8 2907 result = proc_pid_instantiate(dir, dentry, task, NULL);
1da177e4 2908 put_task_struct(task);
1da177e4 2909out:
cd6a3ce9 2910 return result;
1da177e4
LT		 2911}
2912
1da177e4 2913/*
0804ef4b 2914 * Find the first task with tgid >= tgid
0bc58a91 2915 *
1da177e4 2916 */
19fd4bb2
EB		 2917struct tgid_iter {
2918 unsigned int tgid;
0804ef4b 2919 struct task_struct *task;
19fd4bb2
EB		 2920};
2921static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2922{
0804ef4b 2923 struct pid *pid;
1da177e4 2924
19fd4bb2
EB		 2925 if (iter.task)
2926 put_task_struct(iter.task);
454cc105 2927 rcu_read_lock();
0804ef4b 2928retry:
19fd4bb2
EB		 2929 iter.task = NULL;
2930 pid = find_ge_pid(iter.tgid, ns);
0804ef4b 2931 if (pid) {
19fd4bb2
EB		 2932 iter.tgid = pid_nr_ns(pid, ns);
2933 iter.task = pid_task(pid, PIDTYPE_PID);
0804ef4b
EB		 2934 /* What we to know is if the pid we have find is the
2935 * pid of a thread_group_leader. Testing for task
2936 * being a thread_group_leader is the obvious thing
2937 * todo but there is a window when it fails, due to
2938 * the pid transfer logic in de_thread.
2939 *
2940 * So we perform the straight forward test of seeing
2941 * if the pid we have found is the pid of a thread
2942 * group leader, and don't worry if the task we have
2943 * found doesn't happen to be a thread group leader.
2944 * As we don't care in the case of readdir.
2945 */
19fd4bb2
EB		 2946 if (!iter.task || !has_group_leader_pid(iter.task)) {
2947 iter.tgid += 1;
0804ef4b 2948 goto retry;
19fd4bb2
EB		 2949 }
2950 get_task_struct(iter.task);
0bc58a91 2951 }
454cc105 2952 rcu_read_unlock();
19fd4bb2 2953 return iter;
1da177e4
LT		 2954}
2955
7bcd6b0e 2956#define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
0804ef4b 2957
61a28784 2958static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
19fd4bb2 2959 struct tgid_iter iter)
61a28784
EB		 2960{
2961 char name[PROC_NUMBUF];
19fd4bb2 2962 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
61a28784 2963 return proc_fill_cache(filp, dirent, filldir, name, len,
19fd4bb2 2964 proc_pid_instantiate, iter.task, NULL);
61a28784
EB		 2965}
2966
1da177e4
LT		 2967/* for the /proc/ directory itself, after non-process stuff has been done */
2968int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2969{
d8bdc59f
LT		 2970 unsigned int nr;
2971 struct task_struct *reaper;
19fd4bb2 2972 struct tgid_iter iter;
b488893a 2973 struct pid_namespace *ns;
1da177e4 2974
d8bdc59f
LT		 2975 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
2976 goto out_no_task;
2977 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2978
2979 reaper = get_proc_task(filp->f_path.dentry->d_inode);
61a28784
EB		 2980 if (!reaper)
2981 goto out_no_task;
2982
7bcd6b0e 2983 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
c5141e6d 2984 const struct pid_entry *p = &proc_base_stuff[nr];
61a28784 2985 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
801199ce 2986 goto out;
1da177e4
LT		 2987 }
2988
b488893a 2989 ns = filp->f_dentry->d_sb->s_fs_info;
19fd4bb2
EB		 2990 iter.task = NULL;
2991 iter.tgid = filp->f_pos - TGID_OFFSET;
2992 for (iter = next_tgid(ns, iter);
2993 iter.task;
2994 iter.tgid += 1, iter = next_tgid(ns, iter)) {
2995 filp->f_pos = iter.tgid + TGID_OFFSET;
2996 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
2997 put_task_struct(iter.task);
0804ef4b 2998 goto out;
1da177e4 2999 }
0bc58a91 3000 }
0804ef4b
EB		 3001 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3002out:
61a28784
EB		 3003 put_task_struct(reaper);
3004out_no_task:
0bc58a91
EB		 3005 return 0;
3006}
1da177e4 3007
28a6d671
EB		 3008/*
3009 * Tasks
3010 */
c5141e6d 3011static const struct pid_entry tid_base_stuff[] = {
631f9c18 3012 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3835541d 3013 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 3014 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
631f9c18
AD		 3015 REG("environ", S_IRUSR, proc_environ_operations),
3016 INF("auxv", S_IRUSR, proc_pid_auxv),
3017 ONE("status", S_IRUGO, proc_pid_status),
a9712bc1 3018 ONE("personality", S_IRUGO, proc_pid_personality),
3036e7b4 3019 INF("limits", S_IRUGO, proc_pid_limits),
43ae34cb 3020#ifdef CONFIG_SCHED_DEBUG
631f9c18 3021 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
ebcb6734 3022#endif
4614a696 3023 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 3024#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
a9712bc1 3025 INF("syscall", S_IRUGO, proc_pid_syscall),
43ae34cb 3026#endif
631f9c18
AD		 3027 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3028 ONE("stat", S_IRUGO, proc_tid_stat),
3029 ONE("statm", S_IRUGO, proc_pid_statm),
3030 REG("maps", S_IRUGO, proc_maps_operations),
28a6d671 3031#ifdef CONFIG_NUMA
631f9c18 3032 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
28a6d671 3033#endif
631f9c18
AD		 3034 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3035 LNK("cwd", proc_cwd_link),
3036 LNK("root", proc_root_link),
3037 LNK("exe", proc_exe_link),
3038 REG("mounts", S_IRUGO, proc_mounts_operations),
3039 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
1e883281 3040#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18
AD		 3041 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3042 REG("smaps", S_IRUGO, proc_smaps_operations),
ca6b0bf0 3043 REG("pagemap", S_IRUGO, proc_pagemap_operations),
28a6d671
EB		 3044#endif
3045#ifdef CONFIG_SECURITY
631f9c18 3046 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 3047#endif
3048#ifdef CONFIG_KALLSYMS
631f9c18 3049 INF("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 3050#endif
2ec220e2 3051#ifdef CONFIG_STACKTRACE
a9712bc1 3052 ONE("stack", S_IRUGO, proc_pid_stack),
28a6d671
EB		 3053#endif
3054#ifdef CONFIG_SCHEDSTATS
631f9c18 3055 INF("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 3056#endif
9745512c 3057#ifdef CONFIG_LATENCYTOP
631f9c18 3058 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 3059#endif
8793d854 3060#ifdef CONFIG_PROC_PID_CPUSET
631f9c18 3061 REG("cpuset", S_IRUGO, proc_cpuset_operations),
a424316c
PM		 3062#endif
3063#ifdef CONFIG_CGROUPS
631f9c18 3064 REG("cgroup", S_IRUGO, proc_cgroup_operations),
28a6d671 3065#endif
631f9c18
AD		 3066 INF("oom_score", S_IRUGO, proc_oom_score),
3067 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
a63d83f4 3068 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 3069#ifdef CONFIG_AUDITSYSCALL
631f9c18 3070 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
26ec3c64 3071 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 3072#endif
f4f154fd 3073#ifdef CONFIG_FAULT_INJECTION
631f9c18 3074 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 3075#endif
297c5d92 3076#ifdef CONFIG_TASK_IO_ACCOUNTING
1d1221f3 3077 INF("io", S_IRUSR, proc_tid_io_accounting),
297c5d92 3078#endif
f133ecca
CM		 3079#ifdef CONFIG_HARDWALL
3080 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3081#endif
28a6d671
EB		 3082};
3083
3084static int proc_tid_base_readdir(struct file * filp,
3085 void * dirent, filldir_t filldir)
3086{
3087 return proc_pident_readdir(filp,dirent,filldir,
3088 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3089}
3090
3091static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
7bcd6b0e
EB		 3092 return proc_pident_lookup(dir, dentry,
3093 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
28a6d671
EB		 3094}
3095
00977a59 3096static const struct file_operations proc_tid_base_operations = {
28a6d671
EB		 3097 .read = generic_read_dir,
3098 .readdir = proc_tid_base_readdir,
6038f373 3099 .llseek = default_llseek,
28a6d671
EB		 3100};
3101
c5ef1c42 3102static const struct inode_operations proc_tid_base_inode_operations = {
28a6d671
EB		 3103 .lookup = proc_tid_base_lookup,
3104 .getattr = pid_getattr,
3105 .setattr = proc_setattr,
3106};
3107
444ceed8 3108static struct dentry *proc_task_instantiate(struct inode *dir,
c5141e6d 3109 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8
EB		 3110{
3111 struct dentry *error = ERR_PTR(-ENOENT);
3112 struct inode *inode;
61a28784 3113 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 3114
3115 if (!inode)
3116 goto out;
3117 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3118 inode->i_op = &proc_tid_base_inode_operations;
3119 inode->i_fop = &proc_tid_base_operations;
3120 inode->i_flags|=S_IMMUTABLE;
aed54175 3121
bfe86848
MS		 3122 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3123 ARRAY_SIZE(tid_base_stuff)));
444ceed8 3124
fb045adb 3125 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 3126
3127 d_add(dentry, inode);
3128 /* Close the race of the process dying before we return the dentry */
3129 if (pid_revalidate(dentry, NULL))
3130 error = NULL;
3131out:
3132 return error;
3133}
3134
28a6d671
EB		 3135static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3136{
3137 struct dentry *result = ERR_PTR(-ENOENT);
3138 struct task_struct *task;
3139 struct task_struct *leader = get_proc_task(dir);
28a6d671 3140 unsigned tid;
b488893a 3141 struct pid_namespace *ns;
28a6d671
EB		 3142
3143 if (!leader)
3144 goto out_no_task;
3145
3146 tid = name_to_int(dentry);
3147 if (tid == ~0U)
3148 goto out;
3149
b488893a 3150 ns = dentry->d_sb->s_fs_info;
28a6d671 3151 rcu_read_lock();
b488893a 3152 task = find_task_by_pid_ns(tid, ns);
28a6d671
EB		 3153 if (task)
3154 get_task_struct(task);
3155 rcu_read_unlock();
3156 if (!task)
3157 goto out;
bac0abd6 3158 if (!same_thread_group(leader, task))
28a6d671
EB		 3159 goto out_drop_task;
3160
444ceed8 3161 result = proc_task_instantiate(dir, dentry, task, NULL);
28a6d671
EB		 3162out_drop_task:
3163 put_task_struct(task);
3164out:
3165 put_task_struct(leader);
3166out_no_task:
3167 return result;
3168}
3169
0bc58a91
EB		 3170/*
3171 * Find the first tid of a thread group to return to user space.
3172 *
3173 * Usually this is just the thread group leader, but if the users
3174 * buffer was too small or there was a seek into the middle of the
3175 * directory we have more work todo.
3176 *
3177 * In the case of a short read we start with find_task_by_pid.
3178 *
3179 * In the case of a seek we start with the leader and walk nr
3180 * threads past it.
3181 */
cc288738 3182static struct task_struct *first_tid(struct task_struct *leader,
b488893a 3183 int tid, int nr, struct pid_namespace *ns)
0bc58a91 3184{
a872ff0c 3185 struct task_struct *pos;
1da177e4 3186
cc288738 3187 rcu_read_lock();
0bc58a91
EB		 3188 /* Attempt to start with the pid of a thread */
3189 if (tid && (nr > 0)) {
b488893a 3190 pos = find_task_by_pid_ns(tid, ns);
a872ff0c
ON		 3191 if (pos && (pos->group_leader == leader))
3192 goto found;
0bc58a91 3193 }
1da177e4 3194
0bc58a91 3195 /* If nr exceeds the number of threads there is nothing todo */
a872ff0c
ON		 3196 pos = NULL;
3197 if (nr && nr >= get_nr_threads(leader))
3198 goto out;
1da177e4 3199
a872ff0c
ON		 3200 /* If we haven't found our starting place yet start
3201 * with the leader and walk nr threads forward.
0bc58a91 3202 */
a872ff0c
ON		 3203 for (pos = leader; nr > 0; --nr) {
3204 pos = next_thread(pos);
3205 if (pos == leader) {
3206 pos = NULL;
3207 goto out;
3208 }
1da177e4 3209 }
a872ff0c
ON		 3210found:
3211 get_task_struct(pos);
3212out:
cc288738 3213 rcu_read_unlock();
0bc58a91
EB		 3214 return pos;
3215}
3216
3217/*
3218 * Find the next thread in the thread list.
3219 * Return NULL if there is an error or no next thread.
3220 *
3221 * The reference to the input task_struct is released.
3222 */
3223static struct task_struct *next_tid(struct task_struct *start)
3224{
c1df7fb8 3225 struct task_struct *pos = NULL;
cc288738 3226 rcu_read_lock();
c1df7fb8 3227 if (pid_alive(start)) {
0bc58a91 3228 pos = next_thread(start);
c1df7fb8
ON		 3229 if (thread_group_leader(pos))
3230 pos = NULL;
3231 else
3232 get_task_struct(pos);
3233 }
cc288738 3234 rcu_read_unlock();
0bc58a91
EB		 3235 put_task_struct(start);
3236 return pos;
1da177e4
LT		 3237}
3238
61a28784
EB		 3239static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3240 struct task_struct *task, int tid)
3241{
3242 char name[PROC_NUMBUF];
3243 int len = snprintf(name, sizeof(name), "%d", tid);
3244 return proc_fill_cache(filp, dirent, filldir, name, len,
3245 proc_task_instantiate, task, NULL);
3246}
3247
1da177e4
LT		 3248/* for the /proc/TGID/task/ directories */
3249static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3250{
2fddfeef 3251 struct dentry *dentry = filp->f_path.dentry;
1da177e4 3252 struct inode *inode = dentry->d_inode;
7d895244 3253 struct task_struct *leader = NULL;
0bc58a91 3254 struct task_struct *task;
1da177e4
LT		 3255 int retval = -ENOENT;
3256 ino_t ino;
0bc58a91 3257 int tid;
b488893a 3258 struct pid_namespace *ns;
1da177e4 3259
7d895244
GC		 3260 task = get_proc_task(inode);
3261 if (!task)
3262 goto out_no_task;
3263 rcu_read_lock();
3264 if (pid_alive(task)) {
3265 leader = task->group_leader;
3266 get_task_struct(leader);
3267 }
3268 rcu_read_unlock();
3269 put_task_struct(task);
99f89551
EB		 3270 if (!leader)
3271 goto out_no_task;
1da177e4
LT		 3272 retval = 0;
3273
ee568b25 3274 switch ((unsigned long)filp->f_pos) {
1da177e4
LT		 3275 case 0:
3276 ino = inode->i_ino;
ee6f779b 3277 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
1da177e4 3278 goto out;
ee6f779b 3279 filp->f_pos++;
1da177e4
LT		 3280 /* fall through */
3281 case 1:
3282 ino = parent_ino(dentry);
ee6f779b 3283 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
1da177e4 3284 goto out;
ee6f779b 3285 filp->f_pos++;
1da177e4
LT		 3286 /* fall through */
3287 }
3288
0bc58a91
EB		 3289 /* f_version caches the tgid value that the last readdir call couldn't
3290 * return. lseek aka telldir automagically resets f_version to 0.
3291 */
b488893a 3292 ns = filp->f_dentry->d_sb->s_fs_info;
2b47c361 3293 tid = (int)filp->f_version;
0bc58a91 3294 filp->f_version = 0;
ee6f779b 3295 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
0bc58a91 3296 task;
ee6f779b 3297 task = next_tid(task), filp->f_pos++) {
b488893a 3298 tid = task_pid_nr_ns(task, ns);
61a28784 3299 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
0bc58a91
EB		 3300 /* returning this tgid failed, save it as the first
3301 * pid for the next readir call */
2b47c361 3302 filp->f_version = (u64)tid;
0bc58a91 3303 put_task_struct(task);
1da177e4 3304 break;
0bc58a91 3305 }
1da177e4
LT		 3306 }
3307out:
99f89551
EB		 3308 put_task_struct(leader);
3309out_no_task:
1da177e4
LT		 3310 return retval;
3311}
6e66b52b
EB		 3312
3313static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3314{
3315 struct inode *inode = dentry->d_inode;
99f89551 3316 struct task_struct *p = get_proc_task(inode);
6e66b52b
EB		 3317 generic_fillattr(inode, stat);
3318
99f89551 3319 if (p) {
99f89551 3320 stat->nlink += get_nr_threads(p);
99f89551 3321 put_task_struct(p);
6e66b52b
EB		 3322 }
3323
3324 return 0;
3325}
28a6d671 3326
c5ef1c42 3327static const struct inode_operations proc_task_inode_operations = {
28a6d671
EB		 3328 .lookup = proc_task_lookup,
3329 .getattr = proc_task_getattr,
3330 .setattr = proc_setattr,
3331};
3332
00977a59 3333static const struct file_operations proc_task_operations = {
28a6d671
EB		 3334 .read = generic_read_dir,
3335 .readdir = proc_task_readdir,
6038f373 3336 .llseek = default_llseek,
28a6d671 3337};
ubuntu-bionic-kernel mirror