projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
fs: move path_put on failure out of ->follow_link
[deliverable/linux.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>
22d917d8 84#include <linux/user_namespace.h>
5ad4e53b 85#include <linux/fs_struct.h>
5a0e3ad6 86#include <linux/slab.h>
640708a2 87#include <linux/flex_array.h>
f133ecca
CM		 88#ifdef CONFIG_HARDWALL
89#include <asm/hardwall.h>
90#endif
43d2b113 91#include <trace/events/oom.h>
1da177e4
LT		 92#include "internal.h"
93
0f2fe20f
EB		 94/* NOTE:
95 * Implementing inode permission operations in /proc is almost
96 * certainly an error. Permission checks need to happen during
97 * each system call not at open time. The reason is that most of
98 * what we wish to check for permissions in /proc varies at runtime.
99 *
100 * The classic example of a problem is opening file descriptors
101 * in /proc for a task before it execs a suid executable.
102 */
103
1da177e4 104struct pid_entry {
1da177e4 105 char *name;
c5141e6d 106 int len;
d161a13f 107 umode_t mode;
c5ef1c42 108 const struct inode_operations *iop;
00977a59 109 const struct file_operations *fop;
20cdc894 110 union proc_op op;
1da177e4
LT		 111};
112
61a28784 113#define NOD(NAME, MODE, IOP, FOP, OP) { \
20cdc894 114 .name = (NAME), \
c5141e6d 115 .len = sizeof(NAME) - 1, \
20cdc894
EB		 116 .mode = MODE, \
117 .iop = IOP, \
118 .fop = FOP, \
119 .op = OP, \
120}
121
631f9c18
AD		 122#define DIR(NAME, MODE, iops, fops) \
123 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
124#define LNK(NAME, get_link) \
61a28784 125 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
20cdc894 126 &proc_pid_link_inode_operations, NULL, \
631f9c18
AD		 127 { .proc_get_link = get_link } )
128#define REG(NAME, MODE, fops) \
129 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
130#define INF(NAME, MODE, read) \
61a28784 131 NOD(NAME, (S_IFREG|(MODE)), \
20cdc894 132 NULL, &proc_info_file_operations, \
631f9c18
AD		 133 { .proc_read = read } )
134#define ONE(NAME, MODE, show) \
be614086
EB		 135 NOD(NAME, (S_IFREG|(MODE)), \
136 NULL, &proc_single_file_operations, \
631f9c18 137 { .proc_show = show } )
1da177e4 138
640708a2
PE		 139static int proc_fd_permission(struct inode *inode, int mask);
140
aed54175
VN		 141/*
142 * Count the number of hardlinks for the pid_entry table, excluding the .
143 * and .. links.
144 */
145static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
146 unsigned int n)
147{
148 unsigned int i;
149 unsigned int count;
150
151 count = 0;
152 for (i = 0; i < n; ++i) {
153 if (S_ISDIR(entries[i].mode))
154 ++count;
155 }
156
157 return count;
158}
159
f7ad3c6b 160static int get_task_root(struct task_struct *task, struct path *root)
1da177e4 161{
7c2c7d99
HD		 162 int result = -ENOENT;
163
0494f6ec 164 task_lock(task);
f7ad3c6b
MS		 165 if (task->fs) {
166 get_fs_root(task->fs, root);
7c2c7d99
HD		 167 result = 0;
168 }
0494f6ec 169 task_unlock(task);
7c2c7d99 170 return result;
0494f6ec
MS		 171}
172
7773fbc5 173static int proc_cwd_link(struct dentry *dentry, struct path *path)
0494f6ec 174{
7773fbc5 175 struct task_struct *task = get_proc_task(dentry->d_inode);
0494f6ec 176 int result = -ENOENT;
99f89551
EB		 177
178 if (task) {
f7ad3c6b
MS		 179 task_lock(task);
180 if (task->fs) {
181 get_fs_pwd(task->fs, path);
182 result = 0;
183 }
184 task_unlock(task);
99f89551
EB		 185 put_task_struct(task);
186 }
1da177e4
LT		 187 return result;
188}
189
7773fbc5 190static int proc_root_link(struct dentry *dentry, struct path *path)
1da177e4 191{
7773fbc5 192 struct task_struct *task = get_proc_task(dentry->d_inode);
1da177e4 193 int result = -ENOENT;
99f89551
EB		 194
195 if (task) {
f7ad3c6b 196 result = get_task_root(task, path);
99f89551
EB		 197 put_task_struct(task);
198 }
1da177e4
LT		 199 return result;
200}
201
1da177e4
LT		 202static int proc_pid_cmdline(struct task_struct *task, char * buffer)
203{
204 int res = 0;
205 unsigned int len;
206 struct mm_struct *mm = get_task_mm(task);
207 if (!mm)
208 goto out;
209 if (!mm->arg_end)
210 goto out_mm; /* Shh! No looking before we're done */
211
212 len = mm->arg_end - mm->arg_start;
213
214 if (len > PAGE_SIZE)
215 len = PAGE_SIZE;
216
217 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
218
219 // If the nul at the end of args has been overwritten, then
220 // assume application is using setproctitle(3).
221 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
222 len = strnlen(buffer, res);
223 if (len < res) {
224 res = len;
225 } else {
226 len = mm->env_end - mm->env_start;
227 if (len > PAGE_SIZE - res)
228 len = PAGE_SIZE - res;
229 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
230 res = strnlen(buffer, res);
231 }
232 }
233out_mm:
234 mmput(mm);
235out:
236 return res;
237}
238
239static int proc_pid_auxv(struct task_struct *task, char *buffer)
240{
e7dcd999 241 struct mm_struct *mm = mm_access(task, PTRACE_MODE_READ);
2fadaef4
AV		 242 int res = PTR_ERR(mm);
243 if (mm && !IS_ERR(mm)) {
1da177e4 244 unsigned int nwords = 0;
dfe6b7d9 245 do {
1da177e4 246 nwords += 2;
dfe6b7d9 247 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
1da177e4
LT		 248 res = nwords * sizeof(mm->saved_auxv[0]);
249 if (res > PAGE_SIZE)
250 res = PAGE_SIZE;
251 memcpy(buffer, mm->saved_auxv, res);
252 mmput(mm);
253 }
254 return res;
255}
256
257
258#ifdef CONFIG_KALLSYMS
259/*
260 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
261 * Returns the resolved symbol. If that fails, simply return the address.
262 */
263static int proc_pid_wchan(struct task_struct *task, char *buffer)
264{
ffb45122 265 unsigned long wchan;
9281acea 266 char symname[KSYM_NAME_LEN];
1da177e4
LT		 267
268 wchan = get_wchan(task);
269
9d65cb4a 270 if (lookup_symbol_name(wchan, symname) < 0)
f83ce3e6
JE		 271 if (!ptrace_may_access(task, PTRACE_MODE_READ))
272 return 0;
273 else
274 return sprintf(buffer, "%lu", wchan);
9d65cb4a
AD		 275 else
276 return sprintf(buffer, "%s", symname);
1da177e4
LT		 277}
278#endif /* CONFIG_KALLSYMS */
279
a9712bc1
AV		 280static int lock_trace(struct task_struct *task)
281{
282 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
283 if (err)
284 return err;
285 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
286 mutex_unlock(&task->signal->cred_guard_mutex);
287 return -EPERM;
288 }
289 return 0;
290}
291
292static void unlock_trace(struct task_struct *task)
293{
294 mutex_unlock(&task->signal->cred_guard_mutex);
295}
296
2ec220e2
KC		 297#ifdef CONFIG_STACKTRACE
298
299#define MAX_STACK_TRACE_DEPTH 64
300
301static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
302 struct pid *pid, struct task_struct *task)
303{
304 struct stack_trace trace;
305 unsigned long *entries;
a9712bc1 306 int err;
2ec220e2
KC		 307 int i;
308
309 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
310 if (!entries)
311 return -ENOMEM;
312
313 trace.nr_entries = 0;
314 trace.max_entries = MAX_STACK_TRACE_DEPTH;
315 trace.entries = entries;
316 trace.skip = 0;
2ec220e2 317
a9712bc1
AV		 318 err = lock_trace(task);
319 if (!err) {
320 save_stack_trace_tsk(task, &trace);
321
322 for (i = 0; i < trace.nr_entries; i++) {
b81a618d 323 seq_printf(m, "[<%pK>] %pS\n",
a9712bc1
AV		 324 (void *)entries[i], (void *)entries[i]);
325 }
326 unlock_trace(task);
2ec220e2
KC		 327 }
328 kfree(entries);
329
a9712bc1 330 return err;
2ec220e2
KC		 331}
332#endif
333
1da177e4
LT		 334#ifdef CONFIG_SCHEDSTATS
335/*
336 * Provides /proc/PID/schedstat
337 */
338static int proc_pid_schedstat(struct task_struct *task, char *buffer)
339{
172ba844 340 return sprintf(buffer, "%llu %llu %lu\n",
826e08b0
IM		 341 (unsigned long long)task->se.sum_exec_runtime,
342 (unsigned long long)task->sched_info.run_delay,
2d72376b 343 task->sched_info.pcount);
1da177e4
LT		 344}
345#endif
346
9745512c
AV		 347#ifdef CONFIG_LATENCYTOP
348static int lstats_show_proc(struct seq_file *m, void *v)
349{
350 int i;
13d77c37
HS		 351 struct inode *inode = m->private;
352 struct task_struct *task = get_proc_task(inode);
9745512c 353
13d77c37
HS		 354 if (!task)
355 return -ESRCH;
356 seq_puts(m, "Latency Top version : v0.1\n");
9745512c 357 for (i = 0; i < 32; i++) {
34e49d4f
JP		 358 struct latency_record *lr = &task->latency_record[i];
359 if (lr->backtrace[0]) {
9745512c 360 int q;
34e49d4f
JP		 361 seq_printf(m, "%i %li %li",
362 lr->count, lr->time, lr->max);
9745512c 363 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
34e49d4f
JP		 364 unsigned long bt = lr->backtrace[q];
365 if (!bt)
9745512c 366 break;
34e49d4f 367 if (bt == ULONG_MAX)
9745512c 368 break;
34e49d4f 369 seq_printf(m, " %ps", (void *)bt);
9745512c 370 }
9d6de12f 371 seq_putc(m, '\n');
9745512c
AV		 372 }
373
374 }
13d77c37 375 put_task_struct(task);
9745512c
AV		 376 return 0;
377}
378
379static int lstats_open(struct inode *inode, struct file *file)
380{
13d77c37 381 return single_open(file, lstats_show_proc, inode);
d6643d12
HS		 382}
383
9745512c
AV		 384static ssize_t lstats_write(struct file *file, const char __user *buf,
385 size_t count, loff_t *offs)
386{
13d77c37 387 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
9745512c 388
13d77c37
HS		 389 if (!task)
390 return -ESRCH;
9745512c 391 clear_all_latency_tracing(task);
13d77c37 392 put_task_struct(task);
9745512c
AV		 393
394 return count;
395}
396
397static const struct file_operations proc_lstats_operations = {
398 .open = lstats_open,
399 .read = seq_read,
400 .write = lstats_write,
401 .llseek = seq_lseek,
13d77c37 402 .release = single_release,
9745512c
AV		 403};
404
405#endif
406
1da177e4
LT		 407static int proc_oom_score(struct task_struct *task, char *buffer)
408{
a7f638f9 409 unsigned long totalpages = totalram_pages + total_swap_pages;
b95c35e7 410 unsigned long points = 0;
1da177e4 411
19c5d45a 412 read_lock(&tasklist_lock);
b95c35e7 413 if (pid_alive(task))
a7f638f9
DR		 414 points = oom_badness(task, NULL, NULL, totalpages) *
415 1000 / totalpages;
19c5d45a 416 read_unlock(&tasklist_lock);
1da177e4
LT		 417 return sprintf(buffer, "%lu\n", points);
418}
419
d85f50d5
NH		 420struct limit_names {
421 char *name;
422 char *unit;
423};
424
425static const struct limit_names lnames[RLIM_NLIMITS] = {
cff4edb5 426 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
d85f50d5
NH		 427 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
428 [RLIMIT_DATA] = {"Max data size", "bytes"},
429 [RLIMIT_STACK] = {"Max stack size", "bytes"},
430 [RLIMIT_CORE] = {"Max core file size", "bytes"},
431 [RLIMIT_RSS] = {"Max resident set", "bytes"},
432 [RLIMIT_NPROC] = {"Max processes", "processes"},
433 [RLIMIT_NOFILE] = {"Max open files", "files"},
434 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
435 [RLIMIT_AS] = {"Max address space", "bytes"},
436 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
437 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
438 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
439 [RLIMIT_NICE] = {"Max nice priority", NULL},
440 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
8808117c 441 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
d85f50d5
NH		 442};
443
444/* Display limits for a process */
445static int proc_pid_limits(struct task_struct *task, char *buffer)
446{
447 unsigned int i;
448 int count = 0;
449 unsigned long flags;
450 char *bufptr = buffer;
451
452 struct rlimit rlim[RLIM_NLIMITS];
453
a6bebbc8 454 if (!lock_task_sighand(task, &flags))
d85f50d5 455 return 0;
d85f50d5
NH		 456 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
457 unlock_task_sighand(task, &flags);
d85f50d5
NH		 458
459 /*
460 * print the file header
461 */
462 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
463 "Limit", "Soft Limit", "Hard Limit", "Units");
464
465 for (i = 0; i < RLIM_NLIMITS; i++) {
466 if (rlim[i].rlim_cur == RLIM_INFINITY)
467 count += sprintf(&bufptr[count], "%-25s %-20s ",
468 lnames[i].name, "unlimited");
469 else
470 count += sprintf(&bufptr[count], "%-25s %-20lu ",
471 lnames[i].name, rlim[i].rlim_cur);
472
473 if (rlim[i].rlim_max == RLIM_INFINITY)
474 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
475 else
476 count += sprintf(&bufptr[count], "%-20lu ",
477 rlim[i].rlim_max);
478
479 if (lnames[i].unit)
480 count += sprintf(&bufptr[count], "%-10s\n",
481 lnames[i].unit);
482 else
483 count += sprintf(&bufptr[count], "\n");
484 }
485
486 return count;
487}
488
ebcb6734
RM		 489#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
490static int proc_pid_syscall(struct task_struct *task, char *buffer)
491{
492 long nr;
493 unsigned long args[6], sp, pc;
a9712bc1
AV		 494 int res = lock_trace(task);
495 if (res)
496 return res;
ebcb6734
RM		 497
498 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
a9712bc1
AV		 499 res = sprintf(buffer, "running\n");
500 else if (nr < 0)
501 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
502 else
503 res = sprintf(buffer,
ebcb6734
RM		 504 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
505 nr,
506 args[0], args[1], args[2], args[3], args[4], args[5],
507 sp, pc);
a9712bc1
AV		 508 unlock_trace(task);
509 return res;
ebcb6734
RM		 510}
511#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
512
1da177e4
LT		 513/************************************************************************/
514/* Here the fs part begins */
515/************************************************************************/
516
517/* permission checks */
778c1144 518static int proc_fd_access_allowed(struct inode *inode)
1da177e4 519{
778c1144
EB		 520 struct task_struct *task;
521 int allowed = 0;
df26c40e
EB		 522 /* Allow access to a task's file descriptors if it is us or we
523 * may use ptrace attach to the process and find out that
524 * information.
778c1144
EB		 525 */
526 task = get_proc_task(inode);
df26c40e 527 if (task) {
006ebb40 528 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
778c1144 529 put_task_struct(task);
df26c40e 530 }
778c1144 531 return allowed;
1da177e4
LT		 532}
533
6b4e306a 534int proc_setattr(struct dentry *dentry, struct iattr *attr)
6d76fa58
LT		 535{
536 int error;
537 struct inode *inode = dentry->d_inode;
538
539 if (attr->ia_valid & ATTR_MODE)
540 return -EPERM;
541
542 error = inode_change_ok(inode, attr);
1025774c
CH		 543 if (error)
544 return error;
545
546 if ((attr->ia_valid & ATTR_SIZE) &&
547 attr->ia_size != i_size_read(inode)) {
548 error = vmtruncate(inode, attr->ia_size);
549 if (error)
550 return error;
551 }
552
553 setattr_copy(inode, attr);
554 mark_inode_dirty(inode);
555 return 0;
6d76fa58
LT		 556}
557
0499680a
VK		 558/*
559 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
560 * or euid/egid (for hide_pid_min=2)?
561 */
562static bool has_pid_permissions(struct pid_namespace *pid,
563 struct task_struct *task,
564 int hide_pid_min)
565{
566 if (pid->hide_pid < hide_pid_min)
567 return true;
568 if (in_group_p(pid->pid_gid))
569 return true;
570 return ptrace_may_access(task, PTRACE_MODE_READ);
571}
572
573
574static int proc_pid_permission(struct inode *inode, int mask)
575{
576 struct pid_namespace *pid = inode->i_sb->s_fs_info;
577 struct task_struct *task;
578 bool has_perms;
579
580 task = get_proc_task(inode);
a2ef990a
XF		 581 if (!task)
582 return -ESRCH;
0499680a
VK		 583 has_perms = has_pid_permissions(pid, task, 1);
584 put_task_struct(task);
585
586 if (!has_perms) {
587 if (pid->hide_pid == 2) {
588 /*
589 * Let's make getdents(), stat(), and open()
590 * consistent with each other. If a process
591 * may not stat() a file, it shouldn't be seen
592 * in procfs at all.
593 */
594 return -ENOENT;
595 }
596
597 return -EPERM;
598 }
599 return generic_permission(inode, mask);
600}
601
602
603
c5ef1c42 604static const struct inode_operations proc_def_inode_operations = {
6d76fa58
LT		 605 .setattr = proc_setattr,
606};
607
1da177e4
LT		 608#define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
609
610static ssize_t proc_info_read(struct file * file, char __user * buf,
611 size_t count, loff_t *ppos)
612{
2fddfeef 613 struct inode * inode = file->f_path.dentry->d_inode;
1da177e4
LT		 614 unsigned long page;
615 ssize_t length;
99f89551
EB		 616 struct task_struct *task = get_proc_task(inode);
617
618 length = -ESRCH;
619 if (!task)
620 goto out_no_task;
1da177e4
LT		 621
622 if (count > PROC_BLOCK_SIZE)
623 count = PROC_BLOCK_SIZE;
99f89551
EB		 624
625 length = -ENOMEM;
e12ba74d 626 if (!(page = __get_free_page(GFP_TEMPORARY)))
99f89551 627 goto out;
1da177e4
LT		 628
629 length = PROC_I(inode)->op.proc_read(task, (char*)page);
630
631 if (length >= 0)
632 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
633 free_page(page);
99f89551
EB		 634out:
635 put_task_struct(task);
636out_no_task:
1da177e4
LT		 637 return length;
638}
639
00977a59 640static const struct file_operations proc_info_file_operations = {
1da177e4 641 .read = proc_info_read,
87df8424 642 .llseek = generic_file_llseek,
1da177e4
LT		 643};
644
be614086
EB		 645static int proc_single_show(struct seq_file *m, void *v)
646{
647 struct inode *inode = m->private;
648 struct pid_namespace *ns;
649 struct pid *pid;
650 struct task_struct *task;
651 int ret;
652
653 ns = inode->i_sb->s_fs_info;
654 pid = proc_pid(inode);
655 task = get_pid_task(pid, PIDTYPE_PID);
656 if (!task)
657 return -ESRCH;
658
659 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
660
661 put_task_struct(task);
662 return ret;
663}
664
665static int proc_single_open(struct inode *inode, struct file *filp)
666{
c6a34058 667 return single_open(filp, proc_single_show, inode);
be614086
EB		 668}
669
670static const struct file_operations proc_single_file_operations = {
671 .open = proc_single_open,
672 .read = seq_read,
673 .llseek = seq_lseek,
674 .release = single_release,
675};
676
b409e578 677static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
1da177e4 678{
e268337d
LT		 679 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
680 struct mm_struct *mm;
681
682 if (!task)
683 return -ESRCH;
684
b409e578 685 mm = mm_access(task, mode);
e268337d
LT		 686 put_task_struct(task);
687
688 if (IS_ERR(mm))
689 return PTR_ERR(mm);
690
6d08f2c7
ON		 691 if (mm) {
692 /* ensure this mm_struct can't be freed */
693 atomic_inc(&mm->mm_count);
694 /* but do not pin its memory */
695 mmput(mm);
696 }
697
4a3956c7
KH		 698 /* OK to pass negative loff_t, we can catch out-of-range */
699 file->f_mode |= FMODE_UNSIGNED_OFFSET;
e268337d
LT		 700 file->private_data = mm;
701
1da177e4
LT		 702 return 0;
703}
704
b409e578
CW		 705static int mem_open(struct inode *inode, struct file *file)
706{
707 return __mem_open(inode, file, PTRACE_MODE_ATTACH);
708}
709
572d34b9
ON		 710static ssize_t mem_rw(struct file *file, char __user *buf,
711 size_t count, loff_t *ppos, int write)
1da177e4 712{
e268337d 713 struct mm_struct *mm = file->private_data;
572d34b9
ON		 714 unsigned long addr = *ppos;
715 ssize_t copied;
1da177e4 716 char *page;
1da177e4 717
e268337d
LT		 718 if (!mm)
719 return 0;
99f89551 720
30cd8903
KM		 721 page = (char *)__get_free_page(GFP_TEMPORARY);
722 if (!page)
e268337d 723 return -ENOMEM;
1da177e4 724
f7ca54f4 725 copied = 0;
6d08f2c7
ON		 726 if (!atomic_inc_not_zero(&mm->mm_users))
727 goto free;
728
1da177e4 729 while (count > 0) {
572d34b9 730 int this_len = min_t(int, count, PAGE_SIZE);
1da177e4 731
572d34b9 732 if (write && copy_from_user(page, buf, this_len)) {
1da177e4
LT		 733 copied = -EFAULT;
734 break;
735 }
572d34b9
ON		 736
737 this_len = access_remote_vm(mm, addr, page, this_len, write);
738 if (!this_len) {
1da177e4
LT		 739 if (!copied)
740 copied = -EIO;
741 break;
742 }
572d34b9
ON		 743
744 if (!write && copy_to_user(buf, page, this_len)) {
745 copied = -EFAULT;
746 break;
747 }
748
749 buf += this_len;
750 addr += this_len;
751 copied += this_len;
752 count -= this_len;
1da177e4 753 }
572d34b9 754 *ppos = addr;
30cd8903 755
6d08f2c7
ON		 756 mmput(mm);
757free:
30cd8903 758 free_page((unsigned long) page);
1da177e4
LT		 759 return copied;
760}
1da177e4 761
572d34b9
ON		 762static ssize_t mem_read(struct file *file, char __user *buf,
763 size_t count, loff_t *ppos)
764{
765 return mem_rw(file, buf, count, ppos, 0);
766}
767
768static ssize_t mem_write(struct file *file, const char __user *buf,
769 size_t count, loff_t *ppos)
770{
771 return mem_rw(file, (char __user*)buf, count, ppos, 1);
772}
773
85863e47 774loff_t mem_lseek(struct file *file, loff_t offset, int orig)
1da177e4
LT		 775{
776 switch (orig) {
777 case 0:
778 file->f_pos = offset;
779 break;
780 case 1:
781 file->f_pos += offset;
782 break;
783 default:
784 return -EINVAL;
785 }
786 force_successful_syscall_return();
787 return file->f_pos;
788}
789
e268337d
LT		 790static int mem_release(struct inode *inode, struct file *file)
791{
792 struct mm_struct *mm = file->private_data;
71879d3c 793 if (mm)
6d08f2c7 794 mmdrop(mm);
e268337d
LT		 795 return 0;
796}
797
00977a59 798static const struct file_operations proc_mem_operations = {
1da177e4
LT		 799 .llseek = mem_lseek,
800 .read = mem_read,
801 .write = mem_write,
802 .open = mem_open,
e268337d 803 .release = mem_release,
1da177e4
LT		 804};
805
b409e578
CW		 806static int environ_open(struct inode *inode, struct file *file)
807{
808 return __mem_open(inode, file, PTRACE_MODE_READ);
809}
810
315e28c8
JP		 811static ssize_t environ_read(struct file *file, char __user *buf,
812 size_t count, loff_t *ppos)
813{
315e28c8
JP		 814 char *page;
815 unsigned long src = *ppos;
b409e578
CW		 816 int ret = 0;
817 struct mm_struct *mm = file->private_data;
315e28c8 818
b409e578
CW		 819 if (!mm)
820 return 0;
315e28c8 821
315e28c8
JP		 822 page = (char *)__get_free_page(GFP_TEMPORARY);
823 if (!page)
b409e578 824 return -ENOMEM;
315e28c8 825
d6f64b89 826 ret = 0;
b409e578
CW		 827 if (!atomic_inc_not_zero(&mm->mm_users))
828 goto free;
315e28c8
JP		 829 while (count > 0) {
830 int this_len, retval, max_len;
831
832 this_len = mm->env_end - (mm->env_start + src);
833
834 if (this_len <= 0)
835 break;
836
837 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
838 this_len = (this_len > max_len) ? max_len : this_len;
839
b409e578 840 retval = access_remote_vm(mm, (mm->env_start + src),
315e28c8
JP		 841 page, this_len, 0);
842
843 if (retval <= 0) {
844 ret = retval;
845 break;
846 }
847
848 if (copy_to_user(buf, page, retval)) {
849 ret = -EFAULT;
850 break;
851 }
852
853 ret += retval;
854 src += retval;
855 buf += retval;
856 count -= retval;
857 }
858 *ppos = src;
315e28c8 859 mmput(mm);
b409e578
CW		 860
861free:
315e28c8 862 free_page((unsigned long) page);
315e28c8
JP		 863 return ret;
864}
865
866static const struct file_operations proc_environ_operations = {
b409e578 867 .open = environ_open,
315e28c8 868 .read = environ_read,
87df8424 869 .llseek = generic_file_llseek,
b409e578 870 .release = mem_release,
315e28c8
JP		 871};
872
1da177e4
LT		 873static ssize_t oom_adjust_read(struct file *file, char __user *buf,
874 size_t count, loff_t *ppos)
875{
2fddfeef 876 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
8578cea7 877 char buffer[PROC_NUMBUF];
1da177e4 878 size_t len;
28b83c51
KM		 879 int oom_adjust = OOM_DISABLE;
880 unsigned long flags;
1da177e4 881
99f89551
EB		 882 if (!task)
883 return -ESRCH;
28b83c51
KM		 884
885 if (lock_task_sighand(task, &flags)) {
886 oom_adjust = task->signal->oom_adj;
887 unlock_task_sighand(task, &flags);
888 }
889
99f89551
EB		 890 put_task_struct(task);
891
8578cea7 892 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
0c28f287
AM		 893
894 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1da177e4
LT		 895}
896
897static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
898 size_t count, loff_t *ppos)
899{
99f89551 900 struct task_struct *task;
5d863b89 901 char buffer[PROC_NUMBUF];
0a8cb8e3 902 int oom_adjust;
28b83c51 903 unsigned long flags;
5d863b89 904 int err;
1da177e4 905
8578cea7
EB		 906 memset(buffer, 0, sizeof(buffer));
907 if (count > sizeof(buffer) - 1)
908 count = sizeof(buffer) - 1;
723548bf
DR		 909 if (copy_from_user(buffer, buf, count)) {
910 err = -EFAULT;
911 goto out;
912 }
5d863b89 913
0a8cb8e3 914 err = kstrtoint(strstrip(buffer), 0, &oom_adjust);
5d863b89 915 if (err)
723548bf 916 goto out;
8ac773b4 917 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
723548bf
DR		 918 oom_adjust != OOM_DISABLE) {
919 err = -EINVAL;
920 goto out;
921 }
5d863b89 922
2fddfeef 923 task = get_proc_task(file->f_path.dentry->d_inode);
723548bf
DR		 924 if (!task) {
925 err = -ESRCH;
926 goto out;
927 }
d19d5476
DR		 928
929 task_lock(task);
930 if (!task->mm) {
931 err = -EINVAL;
932 goto err_task_lock;
933 }
934
28b83c51 935 if (!lock_task_sighand(task, &flags)) {
723548bf 936 err = -ESRCH;
d19d5476 937 goto err_task_lock;
28b83c51
KM		 938 }
939
940 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
723548bf
DR		 941 err = -EACCES;
942 goto err_sighand;
8fb4fc68 943 }
28b83c51 944
51b1bd2a
DR		 945 /*
946 * Warn that /proc/pid/oom_adj is deprecated, see
947 * Documentation/feature-removal-schedule.txt.
948 */
c2142704 949 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
be8f684d
DR		 950 current->comm, task_pid_nr(current), task_pid_nr(task),
951 task_pid_nr(task));
28b83c51 952 task->signal->oom_adj = oom_adjust;
a63d83f4
DR		 953 /*
954 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
955 * value is always attainable.
956 */
957 if (task->signal->oom_adj == OOM_ADJUST_MAX)
958 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
959 else
960 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
961 -OOM_DISABLE;
43d2b113 962 trace_oom_score_adj_update(task);
723548bf 963err_sighand:
28b83c51 964 unlock_task_sighand(task, &flags);
d19d5476
DR		 965err_task_lock:
966 task_unlock(task);
99f89551 967 put_task_struct(task);
723548bf
DR		 968out:
969 return err < 0 ? err : count;
1da177e4
LT		 970}
971
00977a59 972static const struct file_operations proc_oom_adjust_operations = {
1da177e4
LT		 973 .read = oom_adjust_read,
974 .write = oom_adjust_write,
87df8424 975 .llseek = generic_file_llseek,
1da177e4
LT		 976};
977
a63d83f4
DR		 978static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
979 size_t count, loff_t *ppos)
980{
981 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
982 char buffer[PROC_NUMBUF];
983 int oom_score_adj = OOM_SCORE_ADJ_MIN;
984 unsigned long flags;
985 size_t len;
986
987 if (!task)
988 return -ESRCH;
989 if (lock_task_sighand(task, &flags)) {
990 oom_score_adj = task->signal->oom_score_adj;
991 unlock_task_sighand(task, &flags);
992 }
993 put_task_struct(task);
994 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
995 return simple_read_from_buffer(buf, count, ppos, buffer, len);
996}
997
998static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
999 size_t count, loff_t *ppos)
1000{
1001 struct task_struct *task;
1002 char buffer[PROC_NUMBUF];
1003 unsigned long flags;
0a8cb8e3 1004 int oom_score_adj;
a63d83f4
DR		 1005 int err;
1006
1007 memset(buffer, 0, sizeof(buffer));
1008 if (count > sizeof(buffer) - 1)
1009 count = sizeof(buffer) - 1;
723548bf
DR		 1010 if (copy_from_user(buffer, buf, count)) {
1011 err = -EFAULT;
1012 goto out;
1013 }
a63d83f4 1014
0a8cb8e3 1015 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
a63d83f4 1016 if (err)
723548bf 1017 goto out;
a63d83f4 1018 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
723548bf
DR		 1019 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1020 err = -EINVAL;
1021 goto out;
1022 }
a63d83f4
DR		 1023
1024 task = get_proc_task(file->f_path.dentry->d_inode);
723548bf
DR		 1025 if (!task) {
1026 err = -ESRCH;
1027 goto out;
1028 }
d19d5476
DR		 1029
1030 task_lock(task);
1031 if (!task->mm) {
1032 err = -EINVAL;
1033 goto err_task_lock;
1034 }
1035
a63d83f4 1036 if (!lock_task_sighand(task, &flags)) {
723548bf 1037 err = -ESRCH;
d19d5476 1038 goto err_task_lock;
a63d83f4 1039 }
d19d5476 1040
dabb16f6 1041 if (oom_score_adj < task->signal->oom_score_adj_min &&
a63d83f4 1042 !capable(CAP_SYS_RESOURCE)) {
723548bf
DR		 1043 err = -EACCES;
1044 goto err_sighand;
a63d83f4
DR		 1045 }
1046
1047 task->signal->oom_score_adj = oom_score_adj;
dabb16f6
MSB		 1048 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1049 task->signal->oom_score_adj_min = oom_score_adj;
43d2b113 1050 trace_oom_score_adj_update(task);
a63d83f4
DR		 1051 /*
1052 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1053 * always attainable.
1054 */
1055 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1056 task->signal->oom_adj = OOM_DISABLE;
1057 else
1058 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1059 OOM_SCORE_ADJ_MAX;
723548bf 1060err_sighand:
a63d83f4 1061 unlock_task_sighand(task, &flags);
d19d5476
DR		 1062err_task_lock:
1063 task_unlock(task);
a63d83f4 1064 put_task_struct(task);
723548bf
DR		 1065out:
1066 return err < 0 ? err : count;
a63d83f4
DR		 1067}
1068
1069static const struct file_operations proc_oom_score_adj_operations = {
1070 .read = oom_score_adj_read,
1071 .write = oom_score_adj_write,
6038f373 1072 .llseek = default_llseek,
a63d83f4
DR		 1073};
1074
1da177e4
LT		 1075#ifdef CONFIG_AUDITSYSCALL
1076#define TMPBUFLEN 21
1077static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1078 size_t count, loff_t *ppos)
1079{
2fddfeef 1080 struct inode * inode = file->f_path.dentry->d_inode;
99f89551 1081 struct task_struct *task = get_proc_task(inode);
1da177e4
LT		 1082 ssize_t length;
1083 char tmpbuf[TMPBUFLEN];
1084
99f89551
EB		 1085 if (!task)
1086 return -ESRCH;
1da177e4 1087 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
0c11b942 1088 audit_get_loginuid(task));
99f89551 1089 put_task_struct(task);
1da177e4
LT		 1090 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1091}
1092
1093static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1094 size_t count, loff_t *ppos)
1095{
2fddfeef 1096 struct inode * inode = file->f_path.dentry->d_inode;
1da177e4
LT		 1097 char *page, *tmp;
1098 ssize_t length;
1da177e4
LT		 1099 uid_t loginuid;
1100
7dc52157
PM		 1101 rcu_read_lock();
1102 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1103 rcu_read_unlock();
1da177e4 1104 return -EPERM;
7dc52157
PM		 1105 }
1106 rcu_read_unlock();
1da177e4 1107
e0182909
AV		 1108 if (count >= PAGE_SIZE)
1109 count = PAGE_SIZE - 1;
1da177e4
LT		 1110
1111 if (*ppos != 0) {
1112 /* No partial writes. */
1113 return -EINVAL;
1114 }
e12ba74d 1115 page = (char*)__get_free_page(GFP_TEMPORARY);
1da177e4
LT		 1116 if (!page)
1117 return -ENOMEM;
1118 length = -EFAULT;
1119 if (copy_from_user(page, buf, count))
1120 goto out_free_page;
1121
e0182909 1122 page[count] = '\0';
1da177e4
LT		 1123 loginuid = simple_strtoul(page, &tmp, 10);
1124 if (tmp == page) {
1125 length = -EINVAL;
1126 goto out_free_page;
1127
1128 }
0a300be6 1129 length = audit_set_loginuid(loginuid);
1da177e4
LT		 1130 if (likely(length == 0))
1131 length = count;
1132
1133out_free_page:
1134 free_page((unsigned long) page);
1135 return length;
1136}
1137
00977a59 1138static const struct file_operations proc_loginuid_operations = {
1da177e4
LT		 1139 .read = proc_loginuid_read,
1140 .write = proc_loginuid_write,
87df8424 1141 .llseek = generic_file_llseek,
1da177e4 1142};
1e0bd755
EP		 1143
1144static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1145 size_t count, loff_t *ppos)
1146{
1147 struct inode * inode = file->f_path.dentry->d_inode;
1148 struct task_struct *task = get_proc_task(inode);
1149 ssize_t length;
1150 char tmpbuf[TMPBUFLEN];
1151
1152 if (!task)
1153 return -ESRCH;
1154 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1155 audit_get_sessionid(task));
1156 put_task_struct(task);
1157 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1158}
1159
1160static const struct file_operations proc_sessionid_operations = {
1161 .read = proc_sessionid_read,
87df8424 1162 .llseek = generic_file_llseek,
1e0bd755 1163};
1da177e4
LT		 1164#endif
1165
f4f154fd
AM		 1166#ifdef CONFIG_FAULT_INJECTION
1167static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1168 size_t count, loff_t *ppos)
1169{
1170 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1171 char buffer[PROC_NUMBUF];
1172 size_t len;
1173 int make_it_fail;
f4f154fd
AM		 1174
1175 if (!task)
1176 return -ESRCH;
1177 make_it_fail = task->make_it_fail;
1178 put_task_struct(task);
1179
1180 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
0c28f287
AM		 1181
1182 return simple_read_from_buffer(buf, count, ppos, buffer, len);
f4f154fd
AM		 1183}
1184
1185static ssize_t proc_fault_inject_write(struct file * file,
1186 const char __user * buf, size_t count, loff_t *ppos)
1187{
1188 struct task_struct *task;
1189 char buffer[PROC_NUMBUF], *end;
1190 int make_it_fail;
1191
1192 if (!capable(CAP_SYS_RESOURCE))
1193 return -EPERM;
1194 memset(buffer, 0, sizeof(buffer));
1195 if (count > sizeof(buffer) - 1)
1196 count = sizeof(buffer) - 1;
1197 if (copy_from_user(buffer, buf, count))
1198 return -EFAULT;
cba8aafe
VL		 1199 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1200 if (*end)
1201 return -EINVAL;
f4f154fd
AM		 1202 task = get_proc_task(file->f_dentry->d_inode);
1203 if (!task)
1204 return -ESRCH;
1205 task->make_it_fail = make_it_fail;
1206 put_task_struct(task);
cba8aafe
VL		 1207
1208 return count;
f4f154fd
AM		 1209}
1210
00977a59 1211static const struct file_operations proc_fault_inject_operations = {
f4f154fd
AM		 1212 .read = proc_fault_inject_read,
1213 .write = proc_fault_inject_write,
87df8424 1214 .llseek = generic_file_llseek,
f4f154fd
AM		 1215};
1216#endif
1217
9745512c 1218
43ae34cb
IM		 1219#ifdef CONFIG_SCHED_DEBUG
1220/*
1221 * Print out various scheduling related per-task fields:
1222 */
1223static int sched_show(struct seq_file *m, void *v)
1224{
1225 struct inode *inode = m->private;
1226 struct task_struct *p;
1227
43ae34cb
IM		 1228 p = get_proc_task(inode);
1229 if (!p)
1230 return -ESRCH;
1231 proc_sched_show_task(p, m);
1232
1233 put_task_struct(p);
1234
1235 return 0;
1236}
1237
1238static ssize_t
1239sched_write(struct file *file, const char __user *buf,
1240 size_t count, loff_t *offset)
1241{
1242 struct inode *inode = file->f_path.dentry->d_inode;
1243 struct task_struct *p;
1244
43ae34cb
IM		 1245 p = get_proc_task(inode);
1246 if (!p)
1247 return -ESRCH;
1248 proc_sched_set_task(p);
1249
1250 put_task_struct(p);
1251
1252 return count;
1253}
1254
1255static int sched_open(struct inode *inode, struct file *filp)
1256{
c6a34058 1257 return single_open(filp, sched_show, inode);
43ae34cb
IM		 1258}
1259
1260static const struct file_operations proc_pid_sched_operations = {
1261 .open = sched_open,
1262 .read = seq_read,
1263 .write = sched_write,
1264 .llseek = seq_lseek,
5ea473a1 1265 .release = single_release,
43ae34cb
IM		 1266};
1267
1268#endif
1269
5091faa4
MG		 1270#ifdef CONFIG_SCHED_AUTOGROUP
1271/*
1272 * Print out autogroup related information:
1273 */
1274static int sched_autogroup_show(struct seq_file *m, void *v)
1275{
1276 struct inode *inode = m->private;
1277 struct task_struct *p;
1278
1279 p = get_proc_task(inode);
1280 if (!p)
1281 return -ESRCH;
1282 proc_sched_autogroup_show_task(p, m);
1283
1284 put_task_struct(p);
1285
1286 return 0;
1287}
1288
1289static ssize_t
1290sched_autogroup_write(struct file *file, const char __user *buf,
1291 size_t count, loff_t *offset)
1292{
1293 struct inode *inode = file->f_path.dentry->d_inode;
1294 struct task_struct *p;
1295 char buffer[PROC_NUMBUF];
0a8cb8e3 1296 int nice;
5091faa4
MG		 1297 int err;
1298
1299 memset(buffer, 0, sizeof(buffer));
1300 if (count > sizeof(buffer) - 1)
1301 count = sizeof(buffer) - 1;
1302 if (copy_from_user(buffer, buf, count))
1303 return -EFAULT;
1304
0a8cb8e3
AD		 1305 err = kstrtoint(strstrip(buffer), 0, &nice);
1306 if (err < 0)
1307 return err;
5091faa4
MG		 1308
1309 p = get_proc_task(inode);
1310 if (!p)
1311 return -ESRCH;
1312
2e5b5b3a 1313 err = proc_sched_autogroup_set_nice(p, nice);
5091faa4
MG		 1314 if (err)
1315 count = err;
1316
1317 put_task_struct(p);
1318
1319 return count;
1320}
1321
1322static int sched_autogroup_open(struct inode *inode, struct file *filp)
1323{
1324 int ret;
1325
1326 ret = single_open(filp, sched_autogroup_show, NULL);
1327 if (!ret) {
1328 struct seq_file *m = filp->private_data;
1329
1330 m->private = inode;
1331 }
1332 return ret;
1333}
1334
1335static const struct file_operations proc_pid_sched_autogroup_operations = {
1336 .open = sched_autogroup_open,
1337 .read = seq_read,
1338 .write = sched_autogroup_write,
1339 .llseek = seq_lseek,
1340 .release = single_release,
1341};
1342
1343#endif /* CONFIG_SCHED_AUTOGROUP */
1344
4614a696 1345static ssize_t comm_write(struct file *file, const char __user *buf,
1346 size_t count, loff_t *offset)
1347{
1348 struct inode *inode = file->f_path.dentry->d_inode;
1349 struct task_struct *p;
1350 char buffer[TASK_COMM_LEN];
1351
1352 memset(buffer, 0, sizeof(buffer));
1353 if (count > sizeof(buffer) - 1)
1354 count = sizeof(buffer) - 1;
1355 if (copy_from_user(buffer, buf, count))
1356 return -EFAULT;
1357
1358 p = get_proc_task(inode);
1359 if (!p)
1360 return -ESRCH;
1361
1362 if (same_thread_group(current, p))
1363 set_task_comm(p, buffer);
1364 else
1365 count = -EINVAL;
1366
1367 put_task_struct(p);
1368
1369 return count;
1370}
1371
1372static int comm_show(struct seq_file *m, void *v)
1373{
1374 struct inode *inode = m->private;
1375 struct task_struct *p;
1376
1377 p = get_proc_task(inode);
1378 if (!p)
1379 return -ESRCH;
1380
1381 task_lock(p);
1382 seq_printf(m, "%s\n", p->comm);
1383 task_unlock(p);
1384
1385 put_task_struct(p);
1386
1387 return 0;
1388}
1389
1390static int comm_open(struct inode *inode, struct file *filp)
1391{
c6a34058 1392 return single_open(filp, comm_show, inode);
4614a696 1393}
1394
1395static const struct file_operations proc_pid_set_comm_operations = {
1396 .open = comm_open,
1397 .read = seq_read,
1398 .write = comm_write,
1399 .llseek = seq_lseek,
1400 .release = single_release,
1401};
1402
7773fbc5 1403static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
925d1c40
MH		 1404{
1405 struct task_struct *task;
1406 struct mm_struct *mm;
1407 struct file *exe_file;
1408
7773fbc5 1409 task = get_proc_task(dentry->d_inode);
925d1c40
MH		 1410 if (!task)
1411 return -ENOENT;
1412 mm = get_task_mm(task);
1413 put_task_struct(task);
1414 if (!mm)
1415 return -ENOENT;
1416 exe_file = get_mm_exe_file(mm);
1417 mmput(mm);
1418 if (exe_file) {
1419 *exe_path = exe_file->f_path;
1420 path_get(&exe_file->f_path);
1421 fput(exe_file);
1422 return 0;
1423 } else
1424 return -ENOENT;
1425}
1426
008b150a 1427static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1da177e4
LT		 1428{
1429 struct inode *inode = dentry->d_inode;
408ef013 1430 struct path path;
1da177e4
LT		 1431 int error = -EACCES;
1432
778c1144
EB		 1433 /* Are we allowed to snoop on the tasks file descriptors? */
1434 if (!proc_fd_access_allowed(inode))
1da177e4 1435 goto out;
1da177e4 1436
408ef013
CH		 1437 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1438 if (error)
1439 goto out;
1440
1441 path_put(&nd->path);
1442 nd->path = path;
1443 return NULL;
1da177e4 1444out:
008b150a 1445 return ERR_PTR(error);
1da177e4
LT		 1446}
1447
3dcd25f3 1448static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1da177e4 1449{
e12ba74d 1450 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
3dcd25f3 1451 char *pathname;
1da177e4
LT		 1452 int len;
1453
1454 if (!tmp)
1455 return -ENOMEM;
0c28f287 1456
7b2a69ba 1457 pathname = d_path(path, tmp, PAGE_SIZE);
3dcd25f3
JB		 1458 len = PTR_ERR(pathname);
1459 if (IS_ERR(pathname))
1da177e4 1460 goto out;
3dcd25f3 1461 len = tmp + PAGE_SIZE - 1 - pathname;
1da177e4
LT		 1462
1463 if (len > buflen)
1464 len = buflen;
3dcd25f3 1465 if (copy_to_user(buffer, pathname, len))
1da177e4
LT		 1466 len = -EFAULT;
1467 out:
1468 free_page((unsigned long)tmp);
1469 return len;
1470}
1471
1472static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1473{
1474 int error = -EACCES;
1475 struct inode *inode = dentry->d_inode;
3dcd25f3 1476 struct path path;
1da177e4 1477
778c1144
EB		 1478 /* Are we allowed to snoop on the tasks file descriptors? */
1479 if (!proc_fd_access_allowed(inode))
1da177e4 1480 goto out;
1da177e4 1481
7773fbc5 1482 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1da177e4
LT		 1483 if (error)
1484 goto out;
1485
3dcd25f3
JB		 1486 error = do_proc_readlink(&path, buffer, buflen);
1487 path_put(&path);
1da177e4 1488out:
1da177e4
LT		 1489 return error;
1490}
1491
c5ef1c42 1492static const struct inode_operations proc_pid_link_inode_operations = {
1da177e4 1493 .readlink = proc_pid_readlink,
6d76fa58
LT		 1494 .follow_link = proc_pid_follow_link,
1495 .setattr = proc_setattr,
1da177e4
LT		 1496};
1497
28a6d671
EB		 1498
1499/* building an inode */
1500
1501static int task_dumpable(struct task_struct *task)
1da177e4 1502{
28a6d671
EB		 1503 int dumpable = 0;
1504 struct mm_struct *mm;
1da177e4 1505
28a6d671
EB		 1506 task_lock(task);
1507 mm = task->mm;
1508 if (mm)
6c5d5238 1509 dumpable = get_dumpable(mm);
28a6d671
EB		 1510 task_unlock(task);
1511 if(dumpable == 1)
1512 return 1;
1513 return 0;
1514}
1da177e4 1515
6b4e306a 1516struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
28a6d671
EB		 1517{
1518 struct inode * inode;
1519 struct proc_inode *ei;
c69e8d9c 1520 const struct cred *cred;
1da177e4 1521
28a6d671 1522 /* We need a new inode */
1da177e4 1523
28a6d671
EB		 1524 inode = new_inode(sb);
1525 if (!inode)
1526 goto out;
1527
1528 /* Common stuff */
1529 ei = PROC_I(inode);
85fe4025 1530 inode->i_ino = get_next_ino();
28a6d671 1531 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
28a6d671
EB		 1532 inode->i_op = &proc_def_inode_operations;
1533
1534 /*
1535 * grab the reference to task.
1536 */
1a657f78 1537 ei->pid = get_task_pid(task, PIDTYPE_PID);
28a6d671
EB		 1538 if (!ei->pid)
1539 goto out_unlock;
1540
28a6d671 1541 if (task_dumpable(task)) {
c69e8d9c
DH		 1542 rcu_read_lock();
1543 cred = __task_cred(task);
1544 inode->i_uid = cred->euid;
1545 inode->i_gid = cred->egid;
1546 rcu_read_unlock();
1da177e4 1547 }
28a6d671
EB		 1548 security_task_to_inode(task, inode);
1549
1da177e4 1550out:
28a6d671
EB		 1551 return inode;
1552
1553out_unlock:
1554 iput(inode);
1555 return NULL;
1da177e4
LT		 1556}
1557
6b4e306a 1558int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1da177e4 1559{
1da177e4 1560 struct inode *inode = dentry->d_inode;
28a6d671 1561 struct task_struct *task;
c69e8d9c 1562 const struct cred *cred;
0499680a 1563 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
c69e8d9c 1564
28a6d671 1565 generic_fillattr(inode, stat);
1da177e4 1566
28a6d671 1567 rcu_read_lock();
dcb0f222
EB		 1568 stat->uid = GLOBAL_ROOT_UID;
1569 stat->gid = GLOBAL_ROOT_GID;
28a6d671
EB		 1570 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1571 if (task) {
0499680a
VK		 1572 if (!has_pid_permissions(pid, task, 2)) {
1573 rcu_read_unlock();
1574 /*
1575 * This doesn't prevent learning whether PID exists,
1576 * it only makes getattr() consistent with readdir().
1577 */
1578 return -ENOENT;
1579 }
28a6d671
EB		 1580 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1581 task_dumpable(task)) {
c69e8d9c
DH		 1582 cred = __task_cred(task);
1583 stat->uid = cred->euid;
1584 stat->gid = cred->egid;
1da177e4
LT		 1585 }
1586 }
28a6d671 1587 rcu_read_unlock();
d6e71144 1588 return 0;
1da177e4
LT		 1589}
1590
1da177e4
LT		 1591/* dentry stuff */
1592
1593/*
1594 * Exceptional case: normally we are not allowed to unhash a busy
1595 * directory. In this case, however, we can do it - no aliasing problems
1596 * due to the way we treat inodes.
1597 *
1598 * Rewrite the inode's ownerships here because the owning task may have
1599 * performed a setuid(), etc.
99f89551
EB		 1600 *
1601 * Before the /proc/pid/status file was created the only way to read
1602 * the effective uid of a /process was to stat /proc/pid. Reading
1603 * /proc/pid/status is slow enough that procps and other packages
1604 * kept stating /proc/pid. To keep the rules in /proc simple I have
1605 * made this apply to all per process world readable and executable
1606 * directories.
1da177e4 1607 */
0b728e19 1608int pid_revalidate(struct dentry *dentry, unsigned int flags)
1da177e4 1609{
34286d66
NP		 1610 struct inode *inode;
1611 struct task_struct *task;
c69e8d9c
DH		 1612 const struct cred *cred;
1613
0b728e19 1614 if (flags & LOOKUP_RCU)
34286d66
NP		 1615 return -ECHILD;
1616
1617 inode = dentry->d_inode;
1618 task = get_proc_task(inode);
1619
99f89551
EB		 1620 if (task) {
1621 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1622 task_dumpable(task)) {
c69e8d9c
DH		 1623 rcu_read_lock();
1624 cred = __task_cred(task);
1625 inode->i_uid = cred->euid;
1626 inode->i_gid = cred->egid;
1627 rcu_read_unlock();
1da177e4 1628 } else {
dcb0f222
EB		 1629 inode->i_uid = GLOBAL_ROOT_UID;
1630 inode->i_gid = GLOBAL_ROOT_GID;
1da177e4 1631 }
9ee8ab9f 1632 inode->i_mode &= ~(S_ISUID | S_ISGID);
1da177e4 1633 security_task_to_inode(task, inode);
99f89551 1634 put_task_struct(task);
1da177e4
LT		 1635 return 1;
1636 }
1637 d_drop(dentry);
1638 return 0;
1639}
1640
fe15ce44 1641static int pid_delete_dentry(const struct dentry * dentry)
99f89551 1642{
28a6d671
EB		 1643 /* Is the task we represent dead?
1644 * If so, then don't put the dentry on the lru list,
1645 * kill it immediately.
1646 */
1647 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1648}
1649
6b4e306a 1650const struct dentry_operations pid_dentry_operations =
28a6d671
EB		 1651{
1652 .d_revalidate = pid_revalidate,
1653 .d_delete = pid_delete_dentry,
1654};
1655
1656/* Lookups */
1657
1c0d04c9
EB		 1658/*
1659 * Fill a directory entry.
1660 *
1661 * If possible create the dcache entry and derive our inode number and
1662 * file type from dcache entry.
1663 *
1664 * Since all of the proc inode numbers are dynamically generated, the inode
1665 * numbers do not exist until the inode is cache. This means creating the
1666 * the dcache entry in readdir is necessary to keep the inode numbers
1667 * reported by readdir in sync with the inode numbers reported
1668 * by stat.
1669 */
6b4e306a
EB		 1670int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1671 const char *name, int len,
c5141e6d 1672 instantiate_t instantiate, struct task_struct *task, const void *ptr)
61a28784 1673{
2fddfeef 1674 struct dentry *child, *dir = filp->f_path.dentry;
61a28784
EB		 1675 struct inode *inode;
1676 struct qstr qname;
1677 ino_t ino = 0;
1678 unsigned type = DT_UNKNOWN;
1679
1680 qname.name = name;
1681 qname.len = len;
1682 qname.hash = full_name_hash(name, len);
1683
1684 child = d_lookup(dir, &qname);
1685 if (!child) {
1686 struct dentry *new;
1687 new = d_alloc(dir, &qname);
1688 if (new) {
1689 child = instantiate(dir->d_inode, new, task, ptr);
1690 if (child)
1691 dput(new);
1692 else
1693 child = new;
1694 }
1695 }
1696 if (!child || IS_ERR(child) || !child->d_inode)
1697 goto end_instantiate;
1698 inode = child->d_inode;
1699 if (inode) {
1700 ino = inode->i_ino;
1701 type = inode->i_mode >> 12;
1702 }
1703 dput(child);
1704end_instantiate:
1705 if (!ino)
1706 ino = find_inode_number(dir, &qname);
1707 if (!ino)
1708 ino = 1;
1709 return filldir(dirent, name, len, filp->f_pos, ino, type);
1710}
1711
28a6d671
EB		 1712static unsigned name_to_int(struct dentry *dentry)
1713{
1714 const char *name = dentry->d_name.name;
1715 int len = dentry->d_name.len;
1716 unsigned n = 0;
1717
1718 if (len > 1 && *name == '0')
1719 goto out;
1720 while (len-- > 0) {
1721 unsigned c = *name++ - '0';
1722 if (c > 9)
1723 goto out;
1724 if (n >= (~0U-9)/10)
1725 goto out;
1726 n *= 10;
1727 n += c;
1728 }
1729 return n;
1730out:
1731 return ~0U;
1732}
1733
27932742
MS		 1734#define PROC_FDINFO_MAX 64
1735
3dcd25f3 1736static int proc_fd_info(struct inode *inode, struct path *path, char *info)
28a6d671 1737{
5e442a49
LT		 1738 struct task_struct *task = get_proc_task(inode);
1739 struct files_struct *files = NULL;
28a6d671
EB		 1740 struct file *file;
1741 int fd = proc_fd(inode);
aa6afca5 1742
5e442a49
LT		 1743 if (task) {
1744 files = get_files_struct(task);
1745 put_task_struct(task);
1746 }
1747 if (files) {
1748 /*
1749 * We are not taking a ref to the file structure, so we must
1750 * hold ->file_lock.
1751 */
1752 spin_lock(&files->file_lock);
1753 file = fcheck_files(files, fd);
1754 if (file) {
1755 unsigned int f_flags;
1756 struct fdtable *fdt;
1757
1758 fdt = files_fdtable(files);
1759 f_flags = file->f_flags & ~O_CLOEXEC;
1dce27c5 1760 if (close_on_exec(fd, fdt))
5e442a49
LT		 1761 f_flags |= O_CLOEXEC;
1762
1763 if (path) {
1764 *path = file->f_path;
1765 path_get(&file->f_path);
1766 }
1767 if (info)
1768 snprintf(info, PROC_FDINFO_MAX,
1769 "pos:\t%lli\n"
1770 "flags:\t0%o\n",
1771 (long long) file->f_pos,
1772 f_flags);
1773 spin_unlock(&files->file_lock);
1774 put_files_struct(files);
1775 return 0;
99f89551 1776 }
5e442a49
LT		 1777 spin_unlock(&files->file_lock);
1778 put_files_struct(files);
1779 }
1780 return -ENOENT;
99f89551
EB		 1781}
1782
7773fbc5 1783static int proc_fd_link(struct dentry *dentry, struct path *path)
27932742 1784{
7773fbc5 1785 return proc_fd_info(dentry->d_inode, path, NULL);
27932742
MS		 1786}
1787
0b728e19 1788static int tid_fd_revalidate(struct dentry *dentry, unsigned int flags)
1da177e4 1789{
34286d66
NP		 1790 struct inode *inode;
1791 struct task_struct *task;
1792 int fd;
1da177e4 1793 struct files_struct *files;
c69e8d9c 1794 const struct cred *cred;
1da177e4 1795
0b728e19 1796 if (flags & LOOKUP_RCU)
34286d66
NP		 1797 return -ECHILD;
1798
1799 inode = dentry->d_inode;
1800 task = get_proc_task(inode);
1801 fd = proc_fd(inode);
1802
99f89551
EB		 1803 if (task) {
1804 files = get_files_struct(task);
1805 if (files) {
30a08bf2 1806 struct file *file;
99f89551 1807 rcu_read_lock();
30a08bf2
LT		 1808 file = fcheck_files(files, fd);
1809 if (file) {
0640113b 1810 unsigned f_mode = file->f_mode;
30a08bf2 1811
99f89551
EB		 1812 rcu_read_unlock();
1813 put_files_struct(files);
30a08bf2 1814
99f89551 1815 if (task_dumpable(task)) {
c69e8d9c
DH		 1816 rcu_read_lock();
1817 cred = __task_cred(task);
1818 inode->i_uid = cred->euid;
1819 inode->i_gid = cred->egid;
1820 rcu_read_unlock();
99f89551 1821 } else {
dcb0f222
EB		 1822 inode->i_uid = GLOBAL_ROOT_UID;
1823 inode->i_gid = GLOBAL_ROOT_GID;
99f89551 1824 }
30a08bf2 1825
0640113b
LT		 1826 if (S_ISLNK(inode->i_mode)) {
1827 unsigned i_mode = S_IFLNK;
1828 if (f_mode & FMODE_READ)
1829 i_mode |= S_IRUSR | S_IXUSR;
1830 if (f_mode & FMODE_WRITE)
1831 i_mode |= S_IWUSR | S_IXUSR;
1832 inode->i_mode = i_mode;
1833 }
30a08bf2 1834
99f89551
EB		 1835 security_task_to_inode(task, inode);
1836 put_task_struct(task);
1837 return 1;
1838 }
b835996f 1839 rcu_read_unlock();
1da177e4 1840 put_files_struct(files);
1da177e4 1841 }
99f89551 1842 put_task_struct(task);
1da177e4
LT		 1843 }
1844 d_drop(dentry);
1845 return 0;
1846}
1847
d72f71eb 1848static const struct dentry_operations tid_fd_dentry_operations =
1da177e4
LT		 1849{
1850 .d_revalidate = tid_fd_revalidate,
1851 .d_delete = pid_delete_dentry,
1852};
1853
444ceed8 1854static struct dentry *proc_fd_instantiate(struct inode *dir,
c5141e6d 1855 struct dentry *dentry, struct task_struct *task, const void *ptr)
1da177e4 1856{
af5e6171 1857 unsigned fd = (unsigned long)ptr;
444ceed8
EB		 1858 struct inode *inode;
1859 struct proc_inode *ei;
1860 struct dentry *error = ERR_PTR(-ENOENT);
1da177e4 1861
61a28784 1862 inode = proc_pid_make_inode(dir->i_sb, task);
1da177e4
LT		 1863 if (!inode)
1864 goto out;
1865 ei = PROC_I(inode);
aed7a6c4 1866 ei->fd = fd;
444ceed8 1867
0640113b 1868 inode->i_mode = S_IFLNK;
5e442a49 1869 inode->i_op = &proc_pid_link_inode_operations;
1da177e4
LT		 1870 inode->i_size = 64;
1871 ei->op.proc_get_link = proc_fd_link;
fb045adb 1872 d_set_d_op(dentry, &tid_fd_dentry_operations);
1da177e4 1873 d_add(dentry, inode);
cd6a3ce9 1874 /* Close the race of the process dying before we return the dentry */
0b728e19 1875 if (tid_fd_revalidate(dentry, 0))
444ceed8 1876 error = NULL;
1da177e4 1877
444ceed8
EB		 1878 out:
1879 return error;
1da177e4
LT		 1880}
1881
27932742
MS		 1882static struct dentry *proc_lookupfd_common(struct inode *dir,
1883 struct dentry *dentry,
1884 instantiate_t instantiate)
444ceed8
EB		 1885{
1886 struct task_struct *task = get_proc_task(dir);
1887 unsigned fd = name_to_int(dentry);
1888 struct dentry *result = ERR_PTR(-ENOENT);
1889
1890 if (!task)
1891 goto out_no_task;
1892 if (fd == ~0U)
1893 goto out;
1894
af5e6171 1895 result = instantiate(dir, dentry, task, (void *)(unsigned long)fd);
444ceed8
EB		 1896out:
1897 put_task_struct(task);
1898out_no_task:
1899 return result;
1900}
1901
27932742
MS		 1902static int proc_readfd_common(struct file * filp, void * dirent,
1903 filldir_t filldir, instantiate_t instantiate)
28a6d671 1904{
2fddfeef 1905 struct dentry *dentry = filp->f_path.dentry;
28a6d671
EB		 1906 struct inode *inode = dentry->d_inode;
1907 struct task_struct *p = get_proc_task(inode);
457c2510 1908 unsigned int fd, ino;
28a6d671 1909 int retval;
28a6d671 1910 struct files_struct * files;
1da177e4 1911
28a6d671
EB		 1912 retval = -ENOENT;
1913 if (!p)
1914 goto out_no_task;
1915 retval = 0;
28a6d671
EB		 1916
1917 fd = filp->f_pos;
1918 switch (fd) {
1919 case 0:
1920 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
5e442a49 1921 goto out;
28a6d671
EB		 1922 filp->f_pos++;
1923 case 1:
1924 ino = parent_ino(dentry);
1925 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
5e442a49 1926 goto out;
28a6d671
EB		 1927 filp->f_pos++;
1928 default:
1929 files = get_files_struct(p);
1930 if (!files)
5e442a49 1931 goto out;
28a6d671 1932 rcu_read_lock();
28a6d671 1933 for (fd = filp->f_pos-2;
9b4f526c 1934 fd < files_fdtable(files)->max_fds;
28a6d671 1935 fd++, filp->f_pos++) {
27932742
MS		 1936 char name[PROC_NUMBUF];
1937 int len;
f05ed3f1 1938 int rv;
28a6d671
EB		 1939
1940 if (!fcheck_files(files, fd))
1941 continue;
1942 rcu_read_unlock();
1943
27932742 1944 len = snprintf(name, sizeof(name), "%d", fd);
f05ed3f1
AD		 1945 rv = proc_fill_cache(filp, dirent, filldir,
1946 name, len, instantiate, p,
af5e6171 1947 (void *)(unsigned long)fd);
f05ed3f1
AD		 1948 if (rv < 0)
1949 goto out_fd_loop;
28a6d671
EB		 1950 rcu_read_lock();
1951 }
1952 rcu_read_unlock();
f05ed3f1 1953out_fd_loop:
28a6d671
EB		 1954 put_files_struct(files);
1955 }
1956out:
1957 put_task_struct(p);
1958out_no_task:
1959 return retval;
1960}
1961
27932742 1962static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
00cd8dd3 1963 unsigned int flags)
27932742
MS		 1964{
1965 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1966}
1967
1968static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1969{
1970 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1971}
1972
1973static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1974 size_t len, loff_t *ppos)
1975{
1976 char tmp[PROC_FDINFO_MAX];
3dcd25f3 1977 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
27932742
MS		 1978 if (!err)
1979 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1980 return err;
1981}
1982
1983static const struct file_operations proc_fdinfo_file_operations = {
87df8424 1984 .open = nonseekable_open,
27932742 1985 .read = proc_fdinfo_read,
6038f373 1986 .llseek = no_llseek,
27932742
MS		 1987};
1988
00977a59 1989static const struct file_operations proc_fd_operations = {
28a6d671
EB		 1990 .read = generic_read_dir,
1991 .readdir = proc_readfd,
6038f373 1992 .llseek = default_llseek,
1da177e4
LT		 1993};
1994
640708a2
PE		 1995#ifdef CONFIG_CHECKPOINT_RESTORE
1996
1997/*
1998 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1999 * which represent vma start and end addresses.
2000 */
2001static int dname_to_vma_addr(struct dentry *dentry,
2002 unsigned long *start, unsigned long *end)
2003{
2004 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
2005 return -EINVAL;
2006
2007 return 0;
2008}
2009
0b728e19 2010static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
640708a2
PE		 2011{
2012 unsigned long vm_start, vm_end;
2013 bool exact_vma_exists = false;
2014 struct mm_struct *mm = NULL;
2015 struct task_struct *task;
2016 const struct cred *cred;
2017 struct inode *inode;
2018 int status = 0;
2019
0b728e19 2020 if (flags & LOOKUP_RCU)
640708a2
PE		 2021 return -ECHILD;
2022
2023 if (!capable(CAP_SYS_ADMIN)) {
2024 status = -EACCES;
2025 goto out_notask;
2026 }
2027
2028 inode = dentry->d_inode;
2029 task = get_proc_task(inode);
2030 if (!task)
2031 goto out_notask;
2032
2344bec7
CW		 2033 mm = mm_access(task, PTRACE_MODE_READ);
2034 if (IS_ERR_OR_NULL(mm))
640708a2
PE		 2035 goto out;
2036
2037 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
2038 down_read(&mm->mmap_sem);
2039 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
2040 up_read(&mm->mmap_sem);
2041 }
2042
2043 mmput(mm);
2044
2045 if (exact_vma_exists) {
2046 if (task_dumpable(task)) {
2047 rcu_read_lock();
2048 cred = __task_cred(task);
2049 inode->i_uid = cred->euid;
2050 inode->i_gid = cred->egid;
2051 rcu_read_unlock();
2052 } else {
dcb0f222
EB		 2053 inode->i_uid = GLOBAL_ROOT_UID;
2054 inode->i_gid = GLOBAL_ROOT_GID;
640708a2
PE		 2055 }
2056 security_task_to_inode(task, inode);
2057 status = 1;
2058 }
2059
2060out:
2061 put_task_struct(task);
2062
2063out_notask:
2064 if (status <= 0)
2065 d_drop(dentry);
2066
2067 return status;
2068}
2069
2070static const struct dentry_operations tid_map_files_dentry_operations = {
2071 .d_revalidate = map_files_d_revalidate,
2072 .d_delete = pid_delete_dentry,
2073};
2074
2075static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
2076{
2077 unsigned long vm_start, vm_end;
2078 struct vm_area_struct *vma;
2079 struct task_struct *task;
2080 struct mm_struct *mm;
2081 int rc;
2082
2083 rc = -ENOENT;
2084 task = get_proc_task(dentry->d_inode);
2085 if (!task)
2086 goto out;
2087
2088 mm = get_task_mm(task);
2089 put_task_struct(task);
2090 if (!mm)
2091 goto out;
2092
2093 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2094 if (rc)
2095 goto out_mmput;
2096
2097 down_read(&mm->mmap_sem);
2098 vma = find_exact_vma(mm, vm_start, vm_end);
2099 if (vma && vma->vm_file) {
2100 *path = vma->vm_file->f_path;
2101 path_get(path);
2102 rc = 0;
2103 }
2104 up_read(&mm->mmap_sem);
2105
2106out_mmput:
2107 mmput(mm);
2108out:
2109 return rc;
2110}
2111
2112struct map_files_info {
2113 struct file *file;
2114 unsigned long len;
2115 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
2116};
2117
2118static struct dentry *
2119proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
2120 struct task_struct *task, const void *ptr)
2121{
2122 const struct file *file = ptr;
2123 struct proc_inode *ei;
2124 struct inode *inode;
2125
2126 if (!file)
2127 return ERR_PTR(-ENOENT);
2128
2129 inode = proc_pid_make_inode(dir->i_sb, task);
2130 if (!inode)
2131 return ERR_PTR(-ENOENT);
2132
2133 ei = PROC_I(inode);
2134 ei->op.proc_get_link = proc_map_files_get_link;
2135
2136 inode->i_op = &proc_pid_link_inode_operations;
2137 inode->i_size = 64;
2138 inode->i_mode = S_IFLNK;
2139
2140 if (file->f_mode & FMODE_READ)
2141 inode->i_mode |= S_IRUSR;
2142 if (file->f_mode & FMODE_WRITE)
2143 inode->i_mode |= S_IWUSR;
2144
2145 d_set_d_op(dentry, &tid_map_files_dentry_operations);
2146 d_add(dentry, inode);
2147
2148 return NULL;
2149}
2150
2151static struct dentry *proc_map_files_lookup(struct inode *dir,
00cd8dd3 2152 struct dentry *dentry, unsigned int flags)
640708a2
PE		 2153{
2154 unsigned long vm_start, vm_end;
2155 struct vm_area_struct *vma;
2156 struct task_struct *task;
2157 struct dentry *result;
2158 struct mm_struct *mm;
2159
2160 result = ERR_PTR(-EACCES);
2161 if (!capable(CAP_SYS_ADMIN))
2162 goto out;
2163
2164 result = ERR_PTR(-ENOENT);
2165 task = get_proc_task(dir);
2166 if (!task)
2167 goto out;
2168
2169 result = ERR_PTR(-EACCES);
eb94cd96 2170 if (!ptrace_may_access(task, PTRACE_MODE_READ))
640708a2
PE		 2171 goto out_put_task;
2172
2173 result = ERR_PTR(-ENOENT);
2174 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
eb94cd96 2175 goto out_put_task;
640708a2
PE		 2176
2177 mm = get_task_mm(task);
2178 if (!mm)
eb94cd96 2179 goto out_put_task;
640708a2
PE		 2180
2181 down_read(&mm->mmap_sem);
2182 vma = find_exact_vma(mm, vm_start, vm_end);
2183 if (!vma)
2184 goto out_no_vma;
2185
2186 result = proc_map_files_instantiate(dir, dentry, task, vma->vm_file);
2187
2188out_no_vma:
2189 up_read(&mm->mmap_sem);
2190 mmput(mm);
640708a2
PE		 2191out_put_task:
2192 put_task_struct(task);
2193out:
2194 return result;
2195}
2196
2197static const struct inode_operations proc_map_files_inode_operations = {
2198 .lookup = proc_map_files_lookup,
2199 .permission = proc_fd_permission,
2200 .setattr = proc_setattr,
2201};
2202
2203static int
2204proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
2205{
2206 struct dentry *dentry = filp->f_path.dentry;
2207 struct inode *inode = dentry->d_inode;
2208 struct vm_area_struct *vma;
2209 struct task_struct *task;
2210 struct mm_struct *mm;
2211 ino_t ino;
2212 int ret;
2213
2214 ret = -EACCES;
2215 if (!capable(CAP_SYS_ADMIN))
2216 goto out;
2217
2218 ret = -ENOENT;
2219 task = get_proc_task(inode);
2220 if (!task)
2221 goto out;
2222
2223 ret = -EACCES;
eb94cd96 2224 if (!ptrace_may_access(task, PTRACE_MODE_READ))
640708a2
PE		 2225 goto out_put_task;
2226
2227 ret = 0;
2228 switch (filp->f_pos) {
2229 case 0:
2230 ino = inode->i_ino;
2231 if (filldir(dirent, ".", 1, 0, ino, DT_DIR) < 0)
eb94cd96 2232 goto out_put_task;
640708a2
PE		 2233 filp->f_pos++;
2234 case 1:
2235 ino = parent_ino(dentry);
2236 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
eb94cd96 2237 goto out_put_task;
640708a2
PE		 2238 filp->f_pos++;
2239 default:
2240 {
2241 unsigned long nr_files, pos, i;
2242 struct flex_array *fa = NULL;
2243 struct map_files_info info;
2244 struct map_files_info *p;
2245
2246 mm = get_task_mm(task);
2247 if (!mm)
eb94cd96 2248 goto out_put_task;
640708a2
PE		 2249 down_read(&mm->mmap_sem);
2250
2251 nr_files = 0;
2252
2253 /*
2254 * We need two passes here:
2255 *
2256 * 1) Collect vmas of mapped files with mmap_sem taken
2257 * 2) Release mmap_sem and instantiate entries
2258 *
2259 * otherwise we get lockdep complained, since filldir()
2260 * routine might require mmap_sem taken in might_fault().
2261 */
2262
2263 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2264 if (vma->vm_file && ++pos > filp->f_pos)
2265 nr_files++;
2266 }
2267
2268 if (nr_files) {
2269 fa = flex_array_alloc(sizeof(info), nr_files,
2270 GFP_KERNEL);
2271 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2272 GFP_KERNEL)) {
2273 ret = -ENOMEM;
2274 if (fa)
2275 flex_array_free(fa);
2276 up_read(&mm->mmap_sem);
2277 mmput(mm);
eb94cd96 2278 goto out_put_task;
640708a2
PE		 2279 }
2280 for (i = 0, vma = mm->mmap, pos = 2; vma;
2281 vma = vma->vm_next) {
2282 if (!vma->vm_file)
2283 continue;
2284 if (++pos <= filp->f_pos)
2285 continue;
2286
2287 get_file(vma->vm_file);
2288 info.file = vma->vm_file;
2289 info.len = snprintf(info.name,
2290 sizeof(info.name), "%lx-%lx",
2291 vma->vm_start, vma->vm_end);
2292 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2293 BUG();
2294 }
2295 }
2296 up_read(&mm->mmap_sem);
2297
2298 for (i = 0; i < nr_files; i++) {
2299 p = flex_array_get(fa, i);
2300 ret = proc_fill_cache(filp, dirent, filldir,
2301 p->name, p->len,
2302 proc_map_files_instantiate,
2303 task, p->file);
2304 if (ret)
2305 break;
2306 filp->f_pos++;
2307 fput(p->file);
2308 }
2309 for (; i < nr_files; i++) {
2310 /*
2311 * In case of error don't forget
2312 * to put rest of file refs.
2313 */
2314 p = flex_array_get(fa, i);
2315 fput(p->file);
2316 }
2317 if (fa)
2318 flex_array_free(fa);
2319 mmput(mm);
2320 }
2321 }
2322
640708a2
PE		 2323out_put_task:
2324 put_task_struct(task);
2325out:
2326 return ret;
2327}
2328
2329static const struct file_operations proc_map_files_operations = {
2330 .read = generic_read_dir,
2331 .readdir = proc_map_files_readdir,
2332 .llseek = default_llseek,
2333};
2334
2335#endif /* CONFIG_CHECKPOINT_RESTORE */
2336
8948e11f
AD		 2337/*
2338 * /proc/pid/fd needs a special permission handler so that a process can still
2339 * access /proc/self/fd after it has executed a setuid().
2340 */
10556cb2 2341static int proc_fd_permission(struct inode *inode, int mask)
8948e11f 2342{
2830ba7f 2343 int rv = generic_permission(inode, mask);
8948e11f
AD		 2344 if (rv == 0)
2345 return 0;
2346 if (task_pid(current) == proc_pid(inode))
2347 rv = 0;
2348 return rv;
2349}
2350
1da177e4
LT		 2351/*
2352 * proc directories can do almost nothing..
2353 */
c5ef1c42 2354static const struct inode_operations proc_fd_inode_operations = {
1da177e4 2355 .lookup = proc_lookupfd,
8948e11f 2356 .permission = proc_fd_permission,
6d76fa58 2357 .setattr = proc_setattr,
1da177e4
LT		 2358};
2359
27932742
MS		 2360static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2361 struct dentry *dentry, struct task_struct *task, const void *ptr)
2362{
af5e6171 2363 unsigned fd = (unsigned long)ptr;
27932742
MS		 2364 struct inode *inode;
2365 struct proc_inode *ei;
2366 struct dentry *error = ERR_PTR(-ENOENT);
2367
2368 inode = proc_pid_make_inode(dir->i_sb, task);
2369 if (!inode)
2370 goto out;
2371 ei = PROC_I(inode);
2372 ei->fd = fd;
2373 inode->i_mode = S_IFREG | S_IRUSR;
2374 inode->i_fop = &proc_fdinfo_file_operations;
fb045adb 2375 d_set_d_op(dentry, &tid_fd_dentry_operations);
27932742
MS		 2376 d_add(dentry, inode);
2377 /* Close the race of the process dying before we return the dentry */
0b728e19 2378 if (tid_fd_revalidate(dentry, 0))
27932742
MS		 2379 error = NULL;
2380
2381 out:
2382 return error;
2383}
2384
2385static struct dentry *proc_lookupfdinfo(struct inode *dir,
2386 struct dentry *dentry,
00cd8dd3 2387 unsigned int flags)
27932742
MS		 2388{
2389 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2390}
2391
2392static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2393{
2394 return proc_readfd_common(filp, dirent, filldir,
2395 proc_fdinfo_instantiate);
2396}
2397
2398static const struct file_operations proc_fdinfo_operations = {
2399 .read = generic_read_dir,
2400 .readdir = proc_readfdinfo,
6038f373 2401 .llseek = default_llseek,
27932742
MS		 2402};
2403
2404/*
2405 * proc directories can do almost nothing..
2406 */
2407static const struct inode_operations proc_fdinfo_inode_operations = {
2408 .lookup = proc_lookupfdinfo,
2409 .setattr = proc_setattr,
2410};
2411
2412
444ceed8 2413static struct dentry *proc_pident_instantiate(struct inode *dir,
c5141e6d 2414 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8 2415{
c5141e6d 2416 const struct pid_entry *p = ptr;
444ceed8
EB		 2417 struct inode *inode;
2418 struct proc_inode *ei;
bd6daba9 2419 struct dentry *error = ERR_PTR(-ENOENT);
444ceed8 2420
61a28784 2421 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 2422 if (!inode)
2423 goto out;
2424
2425 ei = PROC_I(inode);
2426 inode->i_mode = p->mode;
2427 if (S_ISDIR(inode->i_mode))
bfe86848 2428 set_nlink(inode, 2); /* Use getattr to fix if necessary */
444ceed8
EB		 2429 if (p->iop)
2430 inode->i_op = p->iop;
2431 if (p->fop)
2432 inode->i_fop = p->fop;
2433 ei->op = p->op;
fb045adb 2434 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 2435 d_add(dentry, inode);
2436 /* Close the race of the process dying before we return the dentry */
0b728e19 2437 if (pid_revalidate(dentry, 0))
444ceed8
EB		 2438 error = NULL;
2439out:
2440 return error;
2441}
2442
1da177e4
LT		 2443static struct dentry *proc_pident_lookup(struct inode *dir,
2444 struct dentry *dentry,
c5141e6d 2445 const struct pid_entry *ents,
7bcd6b0e 2446 unsigned int nents)
1da177e4 2447{
cd6a3ce9 2448 struct dentry *error;
99f89551 2449 struct task_struct *task = get_proc_task(dir);
c5141e6d 2450 const struct pid_entry *p, *last;
1da177e4 2451
cd6a3ce9 2452 error = ERR_PTR(-ENOENT);
1da177e4 2453
99f89551
EB		 2454 if (!task)
2455 goto out_no_task;
1da177e4 2456
20cdc894
EB		 2457 /*
2458 * Yes, it does not scale. And it should not. Don't add
2459 * new entries into /proc/<tgid>/ without very good reasons.
2460 */
7bcd6b0e
EB		 2461 last = &ents[nents - 1];
2462 for (p = ents; p <= last; p++) {
1da177e4
LT		 2463 if (p->len != dentry->d_name.len)
2464 continue;
2465 if (!memcmp(dentry->d_name.name, p->name, p->len))
2466 break;
2467 }
7bcd6b0e 2468 if (p > last)
1da177e4
LT		 2469 goto out;
2470
444ceed8 2471 error = proc_pident_instantiate(dir, dentry, task, p);
1da177e4 2472out:
99f89551
EB		 2473 put_task_struct(task);
2474out_no_task:
cd6a3ce9 2475 return error;
1da177e4
LT		 2476}
2477
c5141e6d
ED		 2478static int proc_pident_fill_cache(struct file *filp, void *dirent,
2479 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
61a28784
EB		 2480{
2481 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2482 proc_pident_instantiate, task, p);
2483}
2484
28a6d671
EB		 2485static int proc_pident_readdir(struct file *filp,
2486 void *dirent, filldir_t filldir,
c5141e6d 2487 const struct pid_entry *ents, unsigned int nents)
28a6d671
EB		 2488{
2489 int i;
2fddfeef 2490 struct dentry *dentry = filp->f_path.dentry;
28a6d671
EB		 2491 struct inode *inode = dentry->d_inode;
2492 struct task_struct *task = get_proc_task(inode);
c5141e6d 2493 const struct pid_entry *p, *last;
28a6d671
EB		 2494 ino_t ino;
2495 int ret;
2496
2497 ret = -ENOENT;
2498 if (!task)
61a28784 2499 goto out_no_task;
28a6d671
EB		 2500
2501 ret = 0;
28a6d671
EB		 2502 i = filp->f_pos;
2503 switch (i) {
2504 case 0:
2505 ino = inode->i_ino;
2506 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2507 goto out;
2508 i++;
2509 filp->f_pos++;
2510 /* fall through */
2511 case 1:
2512 ino = parent_ino(dentry);
2513 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2514 goto out;
2515 i++;
2516 filp->f_pos++;
2517 /* fall through */
2518 default:
2519 i -= 2;
2520 if (i >= nents) {
2521 ret = 1;
2522 goto out;
2523 }
2524 p = ents + i;
7bcd6b0e
EB		 2525 last = &ents[nents - 1];
2526 while (p <= last) {
61a28784 2527 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
28a6d671
EB		 2528 goto out;
2529 filp->f_pos++;
2530 p++;
2531 }
2532 }
2533
2534 ret = 1;
2535out:
61a28784
EB		 2536 put_task_struct(task);
2537out_no_task:
28a6d671 2538 return ret;
1da177e4
LT		 2539}
2540
28a6d671
EB		 2541#ifdef CONFIG_SECURITY
2542static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2543 size_t count, loff_t *ppos)
2544{
2fddfeef 2545 struct inode * inode = file->f_path.dentry->d_inode;
04ff9708 2546 char *p = NULL;
28a6d671
EB		 2547 ssize_t length;
2548 struct task_struct *task = get_proc_task(inode);
2549
28a6d671 2550 if (!task)
04ff9708 2551 return -ESRCH;
28a6d671
EB		 2552
2553 length = security_getprocattr(task,
2fddfeef 2554 (char*)file->f_path.dentry->d_name.name,
04ff9708 2555 &p);
28a6d671 2556 put_task_struct(task);
04ff9708
AV		 2557 if (length > 0)
2558 length = simple_read_from_buffer(buf, count, ppos, p, length);
2559 kfree(p);
28a6d671 2560 return length;
1da177e4
LT		 2561}
2562
28a6d671
EB		 2563static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2564 size_t count, loff_t *ppos)
2565{
2fddfeef 2566 struct inode * inode = file->f_path.dentry->d_inode;
28a6d671
EB		 2567 char *page;
2568 ssize_t length;
2569 struct task_struct *task = get_proc_task(inode);
2570
2571 length = -ESRCH;
2572 if (!task)
2573 goto out_no_task;
2574 if (count > PAGE_SIZE)
2575 count = PAGE_SIZE;
2576
2577 /* No partial writes. */
2578 length = -EINVAL;
2579 if (*ppos != 0)
2580 goto out;
2581
2582 length = -ENOMEM;
e12ba74d 2583 page = (char*)__get_free_page(GFP_TEMPORARY);
28a6d671
EB		 2584 if (!page)
2585 goto out;
2586
2587 length = -EFAULT;
2588 if (copy_from_user(page, buf, count))
2589 goto out_free;
2590
107db7c7 2591 /* Guard against adverse ptrace interaction */
9b1bf12d 2592 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
107db7c7
DH		 2593 if (length < 0)
2594 goto out_free;
2595
28a6d671 2596 length = security_setprocattr(task,
2fddfeef 2597 (char*)file->f_path.dentry->d_name.name,
28a6d671 2598 (void*)page, count);
9b1bf12d 2599 mutex_unlock(&task->signal->cred_guard_mutex);
28a6d671
EB		 2600out_free:
2601 free_page((unsigned long) page);
2602out:
2603 put_task_struct(task);
2604out_no_task:
2605 return length;
2606}
2607
00977a59 2608static const struct file_operations proc_pid_attr_operations = {
28a6d671
EB		 2609 .read = proc_pid_attr_read,
2610 .write = proc_pid_attr_write,
87df8424 2611 .llseek = generic_file_llseek,
28a6d671
EB		 2612};
2613
c5141e6d 2614static const struct pid_entry attr_dir_stuff[] = {
631f9c18
AD		 2615 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2616 REG("prev", S_IRUGO, proc_pid_attr_operations),
2617 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2618 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2619 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2620 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
28a6d671
EB		 2621};
2622
72d9dcfc 2623static int proc_attr_dir_readdir(struct file * filp,
28a6d671
EB		 2624 void * dirent, filldir_t filldir)
2625{
2626 return proc_pident_readdir(filp,dirent,filldir,
72d9dcfc 2627 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2628}
2629
00977a59 2630static const struct file_operations proc_attr_dir_operations = {
1da177e4 2631 .read = generic_read_dir,
72d9dcfc 2632 .readdir = proc_attr_dir_readdir,
6038f373 2633 .llseek = default_llseek,
1da177e4
LT		 2634};
2635
72d9dcfc 2636static struct dentry *proc_attr_dir_lookup(struct inode *dir,
00cd8dd3 2637 struct dentry *dentry, unsigned int flags)
28a6d671 2638{
7bcd6b0e
EB		 2639 return proc_pident_lookup(dir, dentry,
2640 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2641}
2642
c5ef1c42 2643static const struct inode_operations proc_attr_dir_inode_operations = {
72d9dcfc 2644 .lookup = proc_attr_dir_lookup,
99f89551 2645 .getattr = pid_getattr,
6d76fa58 2646 .setattr = proc_setattr,
1da177e4
LT		 2647};
2648
28a6d671
EB		 2649#endif
2650
698ba7b5 2651#ifdef CONFIG_ELF_CORE
3cb4a0bb
KH		 2652static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2653 size_t count, loff_t *ppos)
2654{
2655 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2656 struct mm_struct *mm;
2657 char buffer[PROC_NUMBUF];
2658 size_t len;
2659 int ret;
2660
2661 if (!task)
2662 return -ESRCH;
2663
2664 ret = 0;
2665 mm = get_task_mm(task);
2666 if (mm) {
2667 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2668 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2669 MMF_DUMP_FILTER_SHIFT));
2670 mmput(mm);
2671 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2672 }
2673
2674 put_task_struct(task);
2675
2676 return ret;
2677}
2678
2679static ssize_t proc_coredump_filter_write(struct file *file,
2680 const char __user *buf,
2681 size_t count,
2682 loff_t *ppos)
2683{
2684 struct task_struct *task;
2685 struct mm_struct *mm;
2686 char buffer[PROC_NUMBUF], *end;
2687 unsigned int val;
2688 int ret;
2689 int i;
2690 unsigned long mask;
2691
2692 ret = -EFAULT;
2693 memset(buffer, 0, sizeof(buffer));
2694 if (count > sizeof(buffer) - 1)
2695 count = sizeof(buffer) - 1;
2696 if (copy_from_user(buffer, buf, count))
2697 goto out_no_task;
2698
2699 ret = -EINVAL;
2700 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2701 if (*end == '\n')
2702 end++;
2703 if (end - buffer == 0)
2704 goto out_no_task;
2705
2706 ret = -ESRCH;
2707 task = get_proc_task(file->f_dentry->d_inode);
2708 if (!task)
2709 goto out_no_task;
2710
2711 ret = end - buffer;
2712 mm = get_task_mm(task);
2713 if (!mm)
2714 goto out_no_mm;
2715
2716 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2717 if (val & mask)
2718 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2719 else
2720 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2721 }
2722
2723 mmput(mm);
2724 out_no_mm:
2725 put_task_struct(task);
2726 out_no_task:
2727 return ret;
2728}
2729
2730static const struct file_operations proc_coredump_filter_operations = {
2731 .read = proc_coredump_filter_read,
2732 .write = proc_coredump_filter_write,
87df8424 2733 .llseek = generic_file_llseek,
3cb4a0bb
KH		 2734};
2735#endif
2736
28a6d671
EB		 2737/*
2738 * /proc/self:
2739 */
2740static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2741 int buflen)
2742{
488e5bc4 2743 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
b55fcb22 2744 pid_t tgid = task_tgid_nr_ns(current, ns);
28a6d671 2745 char tmp[PROC_NUMBUF];
b55fcb22 2746 if (!tgid)
488e5bc4 2747 return -ENOENT;
b55fcb22 2748 sprintf(tmp, "%d", tgid);
28a6d671
EB		 2749 return vfs_readlink(dentry,buffer,buflen,tmp);
2750}
2751
2752static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2753{
488e5bc4 2754 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
b55fcb22 2755 pid_t tgid = task_tgid_nr_ns(current, ns);
7fee4868
AV		 2756 char *name = ERR_PTR(-ENOENT);
2757 if (tgid) {
2758 name = __getname();
2759 if (!name)
2760 name = ERR_PTR(-ENOMEM);
2761 else
2762 sprintf(name, "%d", tgid);
2763 }
2764 nd_set_link(nd, name);
2765 return NULL;
2766}
2767
2768static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2769 void *cookie)
2770{
2771 char *s = nd_get_link(nd);
2772 if (!IS_ERR(s))
2773 __putname(s);
28a6d671
EB		 2774}
2775
c5ef1c42 2776static const struct inode_operations proc_self_inode_operations = {
28a6d671
EB		 2777 .readlink = proc_self_readlink,
2778 .follow_link = proc_self_follow_link,
7fee4868 2779 .put_link = proc_self_put_link,
28a6d671
EB		 2780};
2781
801199ce
EB		 2782/*
2783 * proc base
2784 *
2785 * These are the directory entries in the root directory of /proc
2786 * that properly belong to the /proc filesystem, as they describe
2787 * describe something that is process related.
2788 */
c5141e6d 2789static const struct pid_entry proc_base_stuff[] = {
61a28784 2790 NOD("self", S_IFLNK|S_IRWXUGO,
801199ce 2791 &proc_self_inode_operations, NULL, {}),
801199ce
EB		 2792};
2793
444ceed8 2794static struct dentry *proc_base_instantiate(struct inode *dir,
c5141e6d 2795 struct dentry *dentry, struct task_struct *task, const void *ptr)
801199ce 2796{
c5141e6d 2797 const struct pid_entry *p = ptr;
801199ce 2798 struct inode *inode;
801199ce 2799 struct proc_inode *ei;
73d36460 2800 struct dentry *error;
801199ce
EB		 2801
2802 /* Allocate the inode */
2803 error = ERR_PTR(-ENOMEM);
2804 inode = new_inode(dir->i_sb);
2805 if (!inode)
2806 goto out;
2807
2808 /* Initialize the inode */
2809 ei = PROC_I(inode);
85fe4025 2810 inode->i_ino = get_next_ino();
801199ce 2811 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
801199ce
EB		 2812
2813 /*
2814 * grab the reference to the task.
2815 */
1a657f78 2816 ei->pid = get_task_pid(task, PIDTYPE_PID);
801199ce
EB		 2817 if (!ei->pid)
2818 goto out_iput;
2819
801199ce
EB		 2820 inode->i_mode = p->mode;
2821 if (S_ISDIR(inode->i_mode))
bfe86848 2822 set_nlink(inode, 2);
801199ce
EB		 2823 if (S_ISLNK(inode->i_mode))
2824 inode->i_size = 64;
2825 if (p->iop)
2826 inode->i_op = p->iop;
2827 if (p->fop)
2828 inode->i_fop = p->fop;
2829 ei->op = p->op;
801199ce
EB		 2830 d_add(dentry, inode);
2831 error = NULL;
2832out:
801199ce
EB		 2833 return error;
2834out_iput:
2835 iput(inode);
2836 goto out;
2837}
2838
444ceed8
EB		 2839static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2840{
2841 struct dentry *error;
2842 struct task_struct *task = get_proc_task(dir);
c5141e6d 2843 const struct pid_entry *p, *last;
444ceed8
EB		 2844
2845 error = ERR_PTR(-ENOENT);
2846
2847 if (!task)
2848 goto out_no_task;
2849
2850 /* Lookup the directory entry */
7bcd6b0e
EB		 2851 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2852 for (p = proc_base_stuff; p <= last; p++) {
444ceed8
EB		 2853 if (p->len != dentry->d_name.len)
2854 continue;
2855 if (!memcmp(dentry->d_name.name, p->name, p->len))
2856 break;
2857 }
7bcd6b0e 2858 if (p > last)
444ceed8
EB		 2859 goto out;
2860
2861 error = proc_base_instantiate(dir, dentry, task, p);
2862
2863out:
2864 put_task_struct(task);
2865out_no_task:
2866 return error;
2867}
2868
c5141e6d
ED		 2869static int proc_base_fill_cache(struct file *filp, void *dirent,
2870 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
61a28784
EB		 2871{
2872 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2873 proc_base_instantiate, task, p);
2874}
2875
aba76fdb 2876#ifdef CONFIG_TASK_IO_ACCOUNTING
297c5d92
AR		 2877static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2878{
940389b8 2879 struct task_io_accounting acct = task->ioac;
5995477a 2880 unsigned long flags;
293eb1e7 2881 int result;
5995477a 2882
293eb1e7
VK		 2883 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2884 if (result)
2885 return result;
2886
2887 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2888 result = -EACCES;
2889 goto out_unlock;
2890 }
1d1221f3 2891
5995477a
AR		 2892 if (whole && lock_task_sighand(task, &flags)) {
2893 struct task_struct *t = task;
2894
2895 task_io_accounting_add(&acct, &task->signal->ioac);
2896 while_each_thread(task, t)
2897 task_io_accounting_add(&acct, &t->ioac);
2898
2899 unlock_task_sighand(task, &flags);
297c5d92 2900 }
293eb1e7 2901 result = sprintf(buffer,
aba76fdb
AM		 2902 "rchar: %llu\n"
2903 "wchar: %llu\n"
2904 "syscr: %llu\n"
2905 "syscw: %llu\n"
2906 "read_bytes: %llu\n"
2907 "write_bytes: %llu\n"
2908 "cancelled_write_bytes: %llu\n",
7c44319d
AB		 2909 (unsigned long long)acct.rchar,
2910 (unsigned long long)acct.wchar,
2911 (unsigned long long)acct.syscr,
2912 (unsigned long long)acct.syscw,
2913 (unsigned long long)acct.read_bytes,
2914 (unsigned long long)acct.write_bytes,
2915 (unsigned long long)acct.cancelled_write_bytes);
293eb1e7
VK		 2916out_unlock:
2917 mutex_unlock(&task->signal->cred_guard_mutex);
2918 return result;
297c5d92
AR		 2919}
2920
2921static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2922{
2923 return do_io_accounting(task, buffer, 0);
aba76fdb 2924}
297c5d92
AR		 2925
2926static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2927{
2928 return do_io_accounting(task, buffer, 1);
2929}
2930#endif /* CONFIG_TASK_IO_ACCOUNTING */
aba76fdb 2931
22d917d8
EB		 2932#ifdef CONFIG_USER_NS
2933static int proc_id_map_open(struct inode *inode, struct file *file,
2934 struct seq_operations *seq_ops)
2935{
2936 struct user_namespace *ns = NULL;
2937 struct task_struct *task;
2938 struct seq_file *seq;
2939 int ret = -EINVAL;
2940
2941 task = get_proc_task(inode);
2942 if (task) {
2943 rcu_read_lock();
2944 ns = get_user_ns(task_cred_xxx(task, user_ns));
2945 rcu_read_unlock();
2946 put_task_struct(task);
2947 }
2948 if (!ns)
2949 goto err;
2950
2951 ret = seq_open(file, seq_ops);
2952 if (ret)
2953 goto err_put_ns;
2954
2955 seq = file->private_data;
2956 seq->private = ns;
2957
2958 return 0;
2959err_put_ns:
2960 put_user_ns(ns);
2961err:
2962 return ret;
2963}
2964
2965static int proc_id_map_release(struct inode *inode, struct file *file)
2966{
2967 struct seq_file *seq = file->private_data;
2968 struct user_namespace *ns = seq->private;
2969 put_user_ns(ns);
2970 return seq_release(inode, file);
2971}
2972
2973static int proc_uid_map_open(struct inode *inode, struct file *file)
2974{
2975 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2976}
2977
2978static int proc_gid_map_open(struct inode *inode, struct file *file)
2979{
2980 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2981}
2982
2983static const struct file_operations proc_uid_map_operations = {
2984 .open = proc_uid_map_open,
2985 .write = proc_uid_map_write,
2986 .read = seq_read,
2987 .llseek = seq_lseek,
2988 .release = proc_id_map_release,
2989};
2990
2991static const struct file_operations proc_gid_map_operations = {
2992 .open = proc_gid_map_open,
2993 .write = proc_gid_map_write,
2994 .read = seq_read,
2995 .llseek = seq_lseek,
2996 .release = proc_id_map_release,
2997};
2998#endif /* CONFIG_USER_NS */
2999
47830723
KC		 3000static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
3001 struct pid *pid, struct task_struct *task)
3002{
a9712bc1
AV		 3003 int err = lock_trace(task);
3004 if (!err) {
3005 seq_printf(m, "%08x\n", task->personality);
3006 unlock_trace(task);
3007 }
3008 return err;
47830723
KC		 3009}
3010
28a6d671
EB		 3011/*
3012 * Thread groups
3013 */
00977a59 3014static const struct file_operations proc_task_operations;
c5ef1c42 3015static const struct inode_operations proc_task_inode_operations;
20cdc894 3016
c5141e6d 3017static const struct pid_entry tgid_base_stuff[] = {
631f9c18
AD		 3018 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
3019 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
640708a2
PE		 3020#ifdef CONFIG_CHECKPOINT_RESTORE
3021 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
3022#endif
631f9c18 3023 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 3024 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
b2211a36 3025#ifdef CONFIG_NET
631f9c18 3026 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
b2211a36 3027#endif
631f9c18
AD		 3028 REG("environ", S_IRUSR, proc_environ_operations),
3029 INF("auxv", S_IRUSR, proc_pid_auxv),
3030 ONE("status", S_IRUGO, proc_pid_status),
a9712bc1 3031 ONE("personality", S_IRUGO, proc_pid_personality),
3036e7b4 3032 INF("limits", S_IRUGO, proc_pid_limits),
43ae34cb 3033#ifdef CONFIG_SCHED_DEBUG
631f9c18 3034 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
5091faa4
MG		 3035#endif
3036#ifdef CONFIG_SCHED_AUTOGROUP
3037 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
ebcb6734 3038#endif
4614a696 3039 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 3040#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
a9712bc1 3041 INF("syscall", S_IRUGO, proc_pid_syscall),
43ae34cb 3042#endif
631f9c18
AD		 3043 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3044 ONE("stat", S_IRUGO, proc_tgid_stat),
3045 ONE("statm", S_IRUGO, proc_pid_statm),
b7643757 3046 REG("maps", S_IRUGO, proc_pid_maps_operations),
28a6d671 3047#ifdef CONFIG_NUMA
b7643757 3048 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
28a6d671 3049#endif
631f9c18
AD		 3050 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3051 LNK("cwd", proc_cwd_link),
3052 LNK("root", proc_root_link),
3053 LNK("exe", proc_exe_link),
3054 REG("mounts", S_IRUGO, proc_mounts_operations),
3055 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3056 REG("mountstats", S_IRUSR, proc_mountstats_operations),
1e883281 3057#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18 3058 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
b7643757 3059 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
ca6b0bf0 3060 REG("pagemap", S_IRUGO, proc_pagemap_operations),
28a6d671
EB		 3061#endif
3062#ifdef CONFIG_SECURITY
631f9c18 3063 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 3064#endif
3065#ifdef CONFIG_KALLSYMS
631f9c18 3066 INF("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 3067#endif
2ec220e2 3068#ifdef CONFIG_STACKTRACE
a9712bc1 3069 ONE("stack", S_IRUGO, proc_pid_stack),
28a6d671
EB		 3070#endif
3071#ifdef CONFIG_SCHEDSTATS
631f9c18 3072 INF("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 3073#endif
9745512c 3074#ifdef CONFIG_LATENCYTOP
631f9c18 3075 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 3076#endif
8793d854 3077#ifdef CONFIG_PROC_PID_CPUSET
631f9c18 3078 REG("cpuset", S_IRUGO, proc_cpuset_operations),
a424316c
PM		 3079#endif
3080#ifdef CONFIG_CGROUPS
631f9c18 3081 REG("cgroup", S_IRUGO, proc_cgroup_operations),
28a6d671 3082#endif
631f9c18
AD		 3083 INF("oom_score", S_IRUGO, proc_oom_score),
3084 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
a63d83f4 3085 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 3086#ifdef CONFIG_AUDITSYSCALL
631f9c18
AD		 3087 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3088 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 3089#endif
f4f154fd 3090#ifdef CONFIG_FAULT_INJECTION
631f9c18 3091 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 3092#endif
698ba7b5 3093#ifdef CONFIG_ELF_CORE
631f9c18 3094 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3cb4a0bb 3095#endif
aba76fdb 3096#ifdef CONFIG_TASK_IO_ACCOUNTING
1d1221f3 3097 INF("io", S_IRUSR, proc_tgid_io_accounting),
aba76fdb 3098#endif
f133ecca
CM		 3099#ifdef CONFIG_HARDWALL
3100 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3101#endif
22d917d8
EB		 3102#ifdef CONFIG_USER_NS
3103 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3104 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3105#endif
28a6d671 3106};
1da177e4 3107
28a6d671 3108static int proc_tgid_base_readdir(struct file * filp,
1da177e4
LT		 3109 void * dirent, filldir_t filldir)
3110{
3111 return proc_pident_readdir(filp,dirent,filldir,
28a6d671 3112 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 3113}
3114
00977a59 3115static const struct file_operations proc_tgid_base_operations = {
1da177e4 3116 .read = generic_read_dir,
28a6d671 3117 .readdir = proc_tgid_base_readdir,
6038f373 3118 .llseek = default_llseek,
1da177e4
LT		 3119};
3120
00cd8dd3
AV		 3121static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3122{
7bcd6b0e
EB		 3123 return proc_pident_lookup(dir, dentry,
3124 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 3125}
3126
c5ef1c42 3127static const struct inode_operations proc_tgid_base_inode_operations = {
28a6d671 3128 .lookup = proc_tgid_base_lookup,
99f89551 3129 .getattr = pid_getattr,
6d76fa58 3130 .setattr = proc_setattr,
0499680a 3131 .permission = proc_pid_permission,
1da177e4 3132};
1da177e4 3133
60347f67 3134static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
1da177e4 3135{
48e6484d 3136 struct dentry *dentry, *leader, *dir;
8578cea7 3137 char buf[PROC_NUMBUF];
48e6484d
EB		 3138 struct qstr name;
3139
3140 name.name = buf;
60347f67
PE		 3141 name.len = snprintf(buf, sizeof(buf), "%d", pid);
3142 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d 3143 if (dentry) {
29f12ca3 3144 shrink_dcache_parent(dentry);
48e6484d
EB		 3145 d_drop(dentry);
3146 dput(dentry);
3147 }
1da177e4 3148
48e6484d 3149 name.name = buf;
60347f67
PE		 3150 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
3151 leader = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d
EB		 3152 if (!leader)
3153 goto out;
1da177e4 3154
48e6484d
EB		 3155 name.name = "task";
3156 name.len = strlen(name.name);
3157 dir = d_hash_and_lookup(leader, &name);
3158 if (!dir)
3159 goto out_put_leader;
3160
3161 name.name = buf;
60347f67 3162 name.len = snprintf(buf, sizeof(buf), "%d", pid);
48e6484d
EB		 3163 dentry = d_hash_and_lookup(dir, &name);
3164 if (dentry) {
3165 shrink_dcache_parent(dentry);
3166 d_drop(dentry);
3167 dput(dentry);
1da177e4 3168 }
48e6484d
EB		 3169
3170 dput(dir);
3171out_put_leader:
3172 dput(leader);
3173out:
3174 return;
1da177e4
LT		 3175}
3176
0895e91d
RD		 3177/**
3178 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3179 * @task: task that should be flushed.
3180 *
3181 * When flushing dentries from proc, one needs to flush them from global
60347f67 3182 * proc (proc_mnt) and from all the namespaces' procs this task was seen
0895e91d
RD		 3183 * in. This call is supposed to do all of this job.
3184 *
3185 * Looks in the dcache for
3186 * /proc/@pid
3187 * /proc/@tgid/task/@pid
3188 * if either directory is present flushes it and all of it'ts children
3189 * from the dcache.
3190 *
3191 * It is safe and reasonable to cache /proc entries for a task until
3192 * that task exits. After that they just clog up the dcache with
3193 * useless entries, possibly causing useful dcache entries to be
3194 * flushed instead. This routine is proved to flush those useless
3195 * dcache entries at process exit time.
3196 *
3197 * NOTE: This routine is just an optimization so it does not guarantee
3198 * that no dcache entries will exist at process exit time it
3199 * just makes it very unlikely that any will persist.
60347f67
PE		 3200 */
3201
3202void proc_flush_task(struct task_struct *task)
3203{
9fcc2d15 3204 int i;
9b4d1cbe 3205 struct pid *pid, *tgid;
130f77ec
PE		 3206 struct upid *upid;
3207
130f77ec 3208 pid = task_pid(task);
9b4d1cbe 3209 tgid = task_tgid(task);
130f77ec 3210
9fcc2d15 3211 for (i = 0; i <= pid->level; i++) {
130f77ec
PE		 3212 upid = &pid->numbers[i];
3213 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
9b4d1cbe 3214 tgid->numbers[i].nr);
130f77ec 3215 }
6f4e6433
PE		 3216
3217 upid = &pid->numbers[pid->level];
3218 if (upid->nr == 1)
3219 pid_ns_release_proc(upid->ns);
60347f67
PE		 3220}
3221
9711ef99
AB		 3222static struct dentry *proc_pid_instantiate(struct inode *dir,
3223 struct dentry * dentry,
c5141e6d 3224 struct task_struct *task, const void *ptr)
444ceed8
EB		 3225{
3226 struct dentry *error = ERR_PTR(-ENOENT);
3227 struct inode *inode;
3228
61a28784 3229 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 3230 if (!inode)
3231 goto out;
3232
3233 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3234 inode->i_op = &proc_tgid_base_inode_operations;
3235 inode->i_fop = &proc_tgid_base_operations;
3236 inode->i_flags|=S_IMMUTABLE;
aed54175 3237
bfe86848
MS		 3238 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
3239 ARRAY_SIZE(tgid_base_stuff)));
444ceed8 3240
fb045adb 3241 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 3242
3243 d_add(dentry, inode);
3244 /* Close the race of the process dying before we return the dentry */
0b728e19 3245 if (pid_revalidate(dentry, 0))
444ceed8
EB		 3246 error = NULL;
3247out:
3248 return error;
3249}
3250
00cd8dd3 3251struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
1da177e4 3252{
73d36460 3253 struct dentry *result;
1da177e4 3254 struct task_struct *task;
1da177e4 3255 unsigned tgid;
b488893a 3256 struct pid_namespace *ns;
1da177e4 3257
801199ce
EB		 3258 result = proc_base_lookup(dir, dentry);
3259 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3260 goto out;
3261
1da177e4
LT		 3262 tgid = name_to_int(dentry);
3263 if (tgid == ~0U)
3264 goto out;
3265
b488893a 3266 ns = dentry->d_sb->s_fs_info;
de758734 3267 rcu_read_lock();
b488893a 3268 task = find_task_by_pid_ns(tgid, ns);
1da177e4
LT		 3269 if (task)
3270 get_task_struct(task);
de758734 3271 rcu_read_unlock();
1da177e4
LT		 3272 if (!task)
3273 goto out;
3274
444ceed8 3275 result = proc_pid_instantiate(dir, dentry, task, NULL);
1da177e4 3276 put_task_struct(task);
1da177e4 3277out:
cd6a3ce9 3278 return result;
1da177e4
LT		 3279}
3280
1da177e4 3281/*
0804ef4b 3282 * Find the first task with tgid >= tgid
0bc58a91 3283 *
1da177e4 3284 */
19fd4bb2
EB		 3285struct tgid_iter {
3286 unsigned int tgid;
0804ef4b 3287 struct task_struct *task;
19fd4bb2
EB		 3288};
3289static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3290{
0804ef4b 3291 struct pid *pid;
1da177e4 3292
19fd4bb2
EB		 3293 if (iter.task)
3294 put_task_struct(iter.task);
454cc105 3295 rcu_read_lock();
0804ef4b 3296retry:
19fd4bb2
EB		 3297 iter.task = NULL;
3298 pid = find_ge_pid(iter.tgid, ns);
0804ef4b 3299 if (pid) {
19fd4bb2
EB		 3300 iter.tgid = pid_nr_ns(pid, ns);
3301 iter.task = pid_task(pid, PIDTYPE_PID);
0804ef4b
EB		 3302 /* What we to know is if the pid we have find is the
3303 * pid of a thread_group_leader. Testing for task
3304 * being a thread_group_leader is the obvious thing
3305 * todo but there is a window when it fails, due to
3306 * the pid transfer logic in de_thread.
3307 *
3308 * So we perform the straight forward test of seeing
3309 * if the pid we have found is the pid of a thread
3310 * group leader, and don't worry if the task we have
3311 * found doesn't happen to be a thread group leader.
3312 * As we don't care in the case of readdir.
3313 */
19fd4bb2
EB		 3314 if (!iter.task || !has_group_leader_pid(iter.task)) {
3315 iter.tgid += 1;
0804ef4b 3316 goto retry;
19fd4bb2
EB		 3317 }
3318 get_task_struct(iter.task);
0bc58a91 3319 }
454cc105 3320 rcu_read_unlock();
19fd4bb2 3321 return iter;
1da177e4
LT		 3322}
3323
7bcd6b0e 3324#define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
0804ef4b 3325
61a28784 3326static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
19fd4bb2 3327 struct tgid_iter iter)
61a28784
EB		 3328{
3329 char name[PROC_NUMBUF];
19fd4bb2 3330 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
61a28784 3331 return proc_fill_cache(filp, dirent, filldir, name, len,
19fd4bb2 3332 proc_pid_instantiate, iter.task, NULL);
61a28784
EB		 3333}
3334
0499680a
VK		 3335static int fake_filldir(void *buf, const char *name, int namelen,
3336 loff_t offset, u64 ino, unsigned d_type)
3337{
3338 return 0;
3339}
3340
1da177e4
LT		 3341/* for the /proc/ directory itself, after non-process stuff has been done */
3342int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3343{
d8bdc59f
LT		 3344 unsigned int nr;
3345 struct task_struct *reaper;
19fd4bb2 3346 struct tgid_iter iter;
b488893a 3347 struct pid_namespace *ns;
0499680a 3348 filldir_t __filldir;
1da177e4 3349
d8bdc59f
LT		 3350 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3351 goto out_no_task;
3352 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3353
3354 reaper = get_proc_task(filp->f_path.dentry->d_inode);
61a28784
EB		 3355 if (!reaper)
3356 goto out_no_task;
3357
7bcd6b0e 3358 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
c5141e6d 3359 const struct pid_entry *p = &proc_base_stuff[nr];
61a28784 3360 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
801199ce 3361 goto out;
1da177e4
LT		 3362 }
3363
b488893a 3364 ns = filp->f_dentry->d_sb->s_fs_info;
19fd4bb2
EB		 3365 iter.task = NULL;
3366 iter.tgid = filp->f_pos - TGID_OFFSET;
3367 for (iter = next_tgid(ns, iter);
3368 iter.task;
3369 iter.tgid += 1, iter = next_tgid(ns, iter)) {
0499680a
VK		 3370 if (has_pid_permissions(ns, iter.task, 2))
3371 __filldir = filldir;
3372 else
3373 __filldir = fake_filldir;
3374
19fd4bb2 3375 filp->f_pos = iter.tgid + TGID_OFFSET;
0499680a 3376 if (proc_pid_fill_cache(filp, dirent, __filldir, iter) < 0) {
19fd4bb2 3377 put_task_struct(iter.task);
0804ef4b 3378 goto out;
1da177e4 3379 }
0bc58a91 3380 }
0804ef4b
EB		 3381 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3382out:
61a28784
EB		 3383 put_task_struct(reaper);
3384out_no_task:
0bc58a91
EB		 3385 return 0;
3386}
1da177e4 3387
28a6d671
EB		 3388/*
3389 * Tasks
3390 */
c5141e6d 3391static const struct pid_entry tid_base_stuff[] = {
631f9c18 3392 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3835541d 3393 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 3394 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
631f9c18
AD		 3395 REG("environ", S_IRUSR, proc_environ_operations),
3396 INF("auxv", S_IRUSR, proc_pid_auxv),
3397 ONE("status", S_IRUGO, proc_pid_status),
a9712bc1 3398 ONE("personality", S_IRUGO, proc_pid_personality),
3036e7b4 3399 INF("limits", S_IRUGO, proc_pid_limits),
43ae34cb 3400#ifdef CONFIG_SCHED_DEBUG
631f9c18 3401 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
ebcb6734 3402#endif
4614a696 3403 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 3404#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
a9712bc1 3405 INF("syscall", S_IRUGO, proc_pid_syscall),
43ae34cb 3406#endif
631f9c18
AD		 3407 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3408 ONE("stat", S_IRUGO, proc_tid_stat),
3409 ONE("statm", S_IRUGO, proc_pid_statm),
b7643757 3410 REG("maps", S_IRUGO, proc_tid_maps_operations),
81841161
CG		 3411#ifdef CONFIG_CHECKPOINT_RESTORE
3412 REG("children", S_IRUGO, proc_tid_children_operations),
3413#endif
28a6d671 3414#ifdef CONFIG_NUMA
b7643757 3415 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
28a6d671 3416#endif
631f9c18
AD		 3417 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3418 LNK("cwd", proc_cwd_link),
3419 LNK("root", proc_root_link),
3420 LNK("exe", proc_exe_link),
3421 REG("mounts", S_IRUGO, proc_mounts_operations),
3422 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
1e883281 3423#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18 3424 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
b7643757 3425 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
ca6b0bf0 3426 REG("pagemap", S_IRUGO, proc_pagemap_operations),
28a6d671
EB		 3427#endif
3428#ifdef CONFIG_SECURITY
631f9c18 3429 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 3430#endif
3431#ifdef CONFIG_KALLSYMS
631f9c18 3432 INF("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 3433#endif
2ec220e2 3434#ifdef CONFIG_STACKTRACE
a9712bc1 3435 ONE("stack", S_IRUGO, proc_pid_stack),
28a6d671
EB		 3436#endif
3437#ifdef CONFIG_SCHEDSTATS
631f9c18 3438 INF("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 3439#endif
9745512c 3440#ifdef CONFIG_LATENCYTOP
631f9c18 3441 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 3442#endif
8793d854 3443#ifdef CONFIG_PROC_PID_CPUSET
631f9c18 3444 REG("cpuset", S_IRUGO, proc_cpuset_operations),
a424316c
PM		 3445#endif
3446#ifdef CONFIG_CGROUPS
631f9c18 3447 REG("cgroup", S_IRUGO, proc_cgroup_operations),
28a6d671 3448#endif
631f9c18
AD		 3449 INF("oom_score", S_IRUGO, proc_oom_score),
3450 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
a63d83f4 3451 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 3452#ifdef CONFIG_AUDITSYSCALL
631f9c18 3453 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
26ec3c64 3454 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 3455#endif
f4f154fd 3456#ifdef CONFIG_FAULT_INJECTION
631f9c18 3457 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 3458#endif
297c5d92 3459#ifdef CONFIG_TASK_IO_ACCOUNTING
1d1221f3 3460 INF("io", S_IRUSR, proc_tid_io_accounting),
297c5d92 3461#endif
f133ecca
CM		 3462#ifdef CONFIG_HARDWALL
3463 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3464#endif
22d917d8
EB		 3465#ifdef CONFIG_USER_NS
3466 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3467 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3468#endif
28a6d671
EB		 3469};
3470
3471static int proc_tid_base_readdir(struct file * filp,
3472 void * dirent, filldir_t filldir)
3473{
3474 return proc_pident_readdir(filp,dirent,filldir,
3475 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3476}
3477
00cd8dd3
AV		 3478static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3479{
7bcd6b0e
EB		 3480 return proc_pident_lookup(dir, dentry,
3481 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
28a6d671
EB		 3482}
3483
00977a59 3484static const struct file_operations proc_tid_base_operations = {
28a6d671
EB		 3485 .read = generic_read_dir,
3486 .readdir = proc_tid_base_readdir,
6038f373 3487 .llseek = default_llseek,
28a6d671
EB		 3488};
3489
c5ef1c42 3490static const struct inode_operations proc_tid_base_inode_operations = {
28a6d671
EB		 3491 .lookup = proc_tid_base_lookup,
3492 .getattr = pid_getattr,
3493 .setattr = proc_setattr,
3494};
3495
444ceed8 3496static struct dentry *proc_task_instantiate(struct inode *dir,
c5141e6d 3497 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8
EB		 3498{
3499 struct dentry *error = ERR_PTR(-ENOENT);
3500 struct inode *inode;
61a28784 3501 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 3502
3503 if (!inode)
3504 goto out;
3505 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3506 inode->i_op = &proc_tid_base_inode_operations;
3507 inode->i_fop = &proc_tid_base_operations;
3508 inode->i_flags|=S_IMMUTABLE;
aed54175 3509
bfe86848
MS		 3510 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3511 ARRAY_SIZE(tid_base_stuff)));
444ceed8 3512
fb045adb 3513 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 3514
3515 d_add(dentry, inode);
3516 /* Close the race of the process dying before we return the dentry */
0b728e19 3517 if (pid_revalidate(dentry, 0))
444ceed8
EB		 3518 error = NULL;
3519out:
3520 return error;
3521}
3522
00cd8dd3 3523static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
28a6d671
EB		 3524{
3525 struct dentry *result = ERR_PTR(-ENOENT);
3526 struct task_struct *task;
3527 struct task_struct *leader = get_proc_task(dir);
28a6d671 3528 unsigned tid;
b488893a 3529 struct pid_namespace *ns;
28a6d671
EB		 3530
3531 if (!leader)
3532 goto out_no_task;
3533
3534 tid = name_to_int(dentry);
3535 if (tid == ~0U)
3536 goto out;
3537
b488893a 3538 ns = dentry->d_sb->s_fs_info;
28a6d671 3539 rcu_read_lock();
b488893a 3540 task = find_task_by_pid_ns(tid, ns);
28a6d671
EB		 3541 if (task)
3542 get_task_struct(task);
3543 rcu_read_unlock();
3544 if (!task)
3545 goto out;
bac0abd6 3546 if (!same_thread_group(leader, task))
28a6d671
EB		 3547 goto out_drop_task;
3548
444ceed8 3549 result = proc_task_instantiate(dir, dentry, task, NULL);
28a6d671
EB		 3550out_drop_task:
3551 put_task_struct(task);
3552out:
3553 put_task_struct(leader);
3554out_no_task:
3555 return result;
3556}
3557
0bc58a91
EB		 3558/*
3559 * Find the first tid of a thread group to return to user space.
3560 *
3561 * Usually this is just the thread group leader, but if the users
3562 * buffer was too small or there was a seek into the middle of the
3563 * directory we have more work todo.
3564 *
3565 * In the case of a short read we start with find_task_by_pid.
3566 *
3567 * In the case of a seek we start with the leader and walk nr
3568 * threads past it.
3569 */
cc288738 3570static struct task_struct *first_tid(struct task_struct *leader,
b488893a 3571 int tid, int nr, struct pid_namespace *ns)
0bc58a91 3572{
a872ff0c 3573 struct task_struct *pos;
1da177e4 3574
cc288738 3575 rcu_read_lock();
0bc58a91
EB		 3576 /* Attempt to start with the pid of a thread */
3577 if (tid && (nr > 0)) {
b488893a 3578 pos = find_task_by_pid_ns(tid, ns);
a872ff0c
ON		 3579 if (pos && (pos->group_leader == leader))
3580 goto found;
0bc58a91 3581 }
1da177e4 3582
0bc58a91 3583 /* If nr exceeds the number of threads there is nothing todo */
a872ff0c
ON		 3584 pos = NULL;
3585 if (nr && nr >= get_nr_threads(leader))
3586 goto out;
1da177e4 3587
a872ff0c
ON		 3588 /* If we haven't found our starting place yet start
3589 * with the leader and walk nr threads forward.
0bc58a91 3590 */
a872ff0c
ON		 3591 for (pos = leader; nr > 0; --nr) {
3592 pos = next_thread(pos);
3593 if (pos == leader) {
3594 pos = NULL;
3595 goto out;
3596 }
1da177e4 3597 }
a872ff0c
ON		 3598found:
3599 get_task_struct(pos);
3600out:
cc288738 3601 rcu_read_unlock();
0bc58a91
EB		 3602 return pos;
3603}
3604
3605/*
3606 * Find the next thread in the thread list.
3607 * Return NULL if there is an error or no next thread.
3608 *
3609 * The reference to the input task_struct is released.
3610 */
3611static struct task_struct *next_tid(struct task_struct *start)
3612{
c1df7fb8 3613 struct task_struct *pos = NULL;
cc288738 3614 rcu_read_lock();
c1df7fb8 3615 if (pid_alive(start)) {
0bc58a91 3616 pos = next_thread(start);
c1df7fb8
ON		 3617 if (thread_group_leader(pos))
3618 pos = NULL;
3619 else
3620 get_task_struct(pos);
3621 }
cc288738 3622 rcu_read_unlock();
0bc58a91
EB		 3623 put_task_struct(start);
3624 return pos;
1da177e4
LT		 3625}
3626
61a28784
EB		 3627static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3628 struct task_struct *task, int tid)
3629{
3630 char name[PROC_NUMBUF];
3631 int len = snprintf(name, sizeof(name), "%d", tid);
3632 return proc_fill_cache(filp, dirent, filldir, name, len,
3633 proc_task_instantiate, task, NULL);
3634}
3635
1da177e4
LT		 3636/* for the /proc/TGID/task/ directories */
3637static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3638{
2fddfeef 3639 struct dentry *dentry = filp->f_path.dentry;
1da177e4 3640 struct inode *inode = dentry->d_inode;
7d895244 3641 struct task_struct *leader = NULL;
0bc58a91 3642 struct task_struct *task;
1da177e4
LT		 3643 int retval = -ENOENT;
3644 ino_t ino;
0bc58a91 3645 int tid;
b488893a 3646 struct pid_namespace *ns;
1da177e4 3647
7d895244
GC		 3648 task = get_proc_task(inode);
3649 if (!task)
3650 goto out_no_task;
3651 rcu_read_lock();
3652 if (pid_alive(task)) {
3653 leader = task->group_leader;
3654 get_task_struct(leader);
3655 }
3656 rcu_read_unlock();
3657 put_task_struct(task);
99f89551
EB		 3658 if (!leader)
3659 goto out_no_task;
1da177e4
LT		 3660 retval = 0;
3661
ee568b25 3662 switch ((unsigned long)filp->f_pos) {
1da177e4
LT		 3663 case 0:
3664 ino = inode->i_ino;
ee6f779b 3665 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
1da177e4 3666 goto out;
ee6f779b 3667 filp->f_pos++;
1da177e4
LT		 3668 /* fall through */
3669 case 1:
3670 ino = parent_ino(dentry);
ee6f779b 3671 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
1da177e4 3672 goto out;
ee6f779b 3673 filp->f_pos++;
1da177e4
LT		 3674 /* fall through */
3675 }
3676
0bc58a91
EB		 3677 /* f_version caches the tgid value that the last readdir call couldn't
3678 * return. lseek aka telldir automagically resets f_version to 0.
3679 */
b488893a 3680 ns = filp->f_dentry->d_sb->s_fs_info;
2b47c361 3681 tid = (int)filp->f_version;
0bc58a91 3682 filp->f_version = 0;
ee6f779b 3683 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
0bc58a91 3684 task;
ee6f779b 3685 task = next_tid(task), filp->f_pos++) {
b488893a 3686 tid = task_pid_nr_ns(task, ns);
61a28784 3687 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
0bc58a91
EB		 3688 /* returning this tgid failed, save it as the first
3689 * pid for the next readir call */
2b47c361 3690 filp->f_version = (u64)tid;
0bc58a91 3691 put_task_struct(task);
1da177e4 3692 break;
0bc58a91 3693 }
1da177e4
LT		 3694 }
3695out:
99f89551
EB		 3696 put_task_struct(leader);
3697out_no_task:
1da177e4
LT		 3698 return retval;
3699}
6e66b52b
EB		 3700
3701static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3702{
3703 struct inode *inode = dentry->d_inode;
99f89551 3704 struct task_struct *p = get_proc_task(inode);
6e66b52b
EB		 3705 generic_fillattr(inode, stat);
3706
99f89551 3707 if (p) {
99f89551 3708 stat->nlink += get_nr_threads(p);
99f89551 3709 put_task_struct(p);
6e66b52b
EB		 3710 }
3711
3712 return 0;
3713}
28a6d671 3714
c5ef1c42 3715static const struct inode_operations proc_task_inode_operations = {
28a6d671
EB		 3716 .lookup = proc_task_lookup,
3717 .getattr = proc_task_getattr,
3718 .setattr = proc_setattr,
0499680a 3719 .permission = proc_pid_permission,
28a6d671
EB		 3720};
3721
00977a59 3722static const struct file_operations proc_task_operations = {
28a6d671
EB		 3723 .read = generic_read_dir,
3724 .readdir = proc_task_readdir,
6038f373 3725 .llseek = default_llseek,
28a6d671 3726};
Linux kernel - Deliverables
This page took 0.939077 seconds and 5 git commands to generate.