projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
new helper: daemonize_descriptors()
[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
e268337d
LT		 698 file->private_data = mm;
699
1da177e4
LT		 700 return 0;
701}
702
b409e578
CW		 703static int mem_open(struct inode *inode, struct file *file)
704{
bc452b4b
DH		 705 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
706
707 /* OK to pass negative loff_t, we can catch out-of-range */
708 file->f_mode |= FMODE_UNSIGNED_OFFSET;
709
710 return ret;
b409e578
CW		 711}
712
572d34b9
ON		 713static ssize_t mem_rw(struct file *file, char __user *buf,
714 size_t count, loff_t *ppos, int write)
1da177e4 715{
e268337d 716 struct mm_struct *mm = file->private_data;
572d34b9
ON		 717 unsigned long addr = *ppos;
718 ssize_t copied;
1da177e4 719 char *page;
1da177e4 720
e268337d
LT		 721 if (!mm)
722 return 0;
99f89551 723
30cd8903
KM		 724 page = (char *)__get_free_page(GFP_TEMPORARY);
725 if (!page)
e268337d 726 return -ENOMEM;
1da177e4 727
f7ca54f4 728 copied = 0;
6d08f2c7
ON		 729 if (!atomic_inc_not_zero(&mm->mm_users))
730 goto free;
731
1da177e4 732 while (count > 0) {
572d34b9 733 int this_len = min_t(int, count, PAGE_SIZE);
1da177e4 734
572d34b9 735 if (write && copy_from_user(page, buf, this_len)) {
1da177e4
LT		 736 copied = -EFAULT;
737 break;
738 }
572d34b9
ON		 739
740 this_len = access_remote_vm(mm, addr, page, this_len, write);
741 if (!this_len) {
1da177e4
LT		 742 if (!copied)
743 copied = -EIO;
744 break;
745 }
572d34b9
ON		 746
747 if (!write && copy_to_user(buf, page, this_len)) {
748 copied = -EFAULT;
749 break;
750 }
751
752 buf += this_len;
753 addr += this_len;
754 copied += this_len;
755 count -= this_len;
1da177e4 756 }
572d34b9 757 *ppos = addr;
30cd8903 758
6d08f2c7
ON		 759 mmput(mm);
760free:
30cd8903 761 free_page((unsigned long) page);
1da177e4
LT		 762 return copied;
763}
1da177e4 764
572d34b9
ON		 765static ssize_t mem_read(struct file *file, char __user *buf,
766 size_t count, loff_t *ppos)
767{
768 return mem_rw(file, buf, count, ppos, 0);
769}
770
771static ssize_t mem_write(struct file *file, const char __user *buf,
772 size_t count, loff_t *ppos)
773{
774 return mem_rw(file, (char __user*)buf, count, ppos, 1);
775}
776
85863e47 777loff_t mem_lseek(struct file *file, loff_t offset, int orig)
1da177e4
LT		 778{
779 switch (orig) {
780 case 0:
781 file->f_pos = offset;
782 break;
783 case 1:
784 file->f_pos += offset;
785 break;
786 default:
787 return -EINVAL;
788 }
789 force_successful_syscall_return();
790 return file->f_pos;
791}
792
e268337d
LT		 793static int mem_release(struct inode *inode, struct file *file)
794{
795 struct mm_struct *mm = file->private_data;
71879d3c 796 if (mm)
6d08f2c7 797 mmdrop(mm);
e268337d
LT		 798 return 0;
799}
800
00977a59 801static const struct file_operations proc_mem_operations = {
1da177e4
LT		 802 .llseek = mem_lseek,
803 .read = mem_read,
804 .write = mem_write,
805 .open = mem_open,
e268337d 806 .release = mem_release,
1da177e4
LT		 807};
808
b409e578
CW		 809static int environ_open(struct inode *inode, struct file *file)
810{
811 return __mem_open(inode, file, PTRACE_MODE_READ);
812}
813
315e28c8
JP		 814static ssize_t environ_read(struct file *file, char __user *buf,
815 size_t count, loff_t *ppos)
816{
315e28c8
JP		 817 char *page;
818 unsigned long src = *ppos;
b409e578
CW		 819 int ret = 0;
820 struct mm_struct *mm = file->private_data;
315e28c8 821
b409e578
CW		 822 if (!mm)
823 return 0;
315e28c8 824
315e28c8
JP		 825 page = (char *)__get_free_page(GFP_TEMPORARY);
826 if (!page)
b409e578 827 return -ENOMEM;
315e28c8 828
d6f64b89 829 ret = 0;
b409e578
CW		 830 if (!atomic_inc_not_zero(&mm->mm_users))
831 goto free;
315e28c8 832 while (count > 0) {
e8905ec2
DH		 833 size_t this_len, max_len;
834 int retval;
315e28c8 835
e8905ec2 836 if (src >= (mm->env_end - mm->env_start))
315e28c8
JP		 837 break;
838
e8905ec2
DH		 839 this_len = mm->env_end - (mm->env_start + src);
840
841 max_len = min_t(size_t, PAGE_SIZE, count);
842 this_len = min(max_len, this_len);
315e28c8 843
b409e578 844 retval = access_remote_vm(mm, (mm->env_start + src),
315e28c8
JP		 845 page, this_len, 0);
846
847 if (retval <= 0) {
848 ret = retval;
849 break;
850 }
851
852 if (copy_to_user(buf, page, retval)) {
853 ret = -EFAULT;
854 break;
855 }
856
857 ret += retval;
858 src += retval;
859 buf += retval;
860 count -= retval;
861 }
862 *ppos = src;
315e28c8 863 mmput(mm);
b409e578
CW		 864
865free:
315e28c8 866 free_page((unsigned long) page);
315e28c8
JP		 867 return ret;
868}
869
870static const struct file_operations proc_environ_operations = {
b409e578 871 .open = environ_open,
315e28c8 872 .read = environ_read,
87df8424 873 .llseek = generic_file_llseek,
b409e578 874 .release = mem_release,
315e28c8
JP		 875};
876
1da177e4
LT		 877static ssize_t oom_adjust_read(struct file *file, char __user *buf,
878 size_t count, loff_t *ppos)
879{
2fddfeef 880 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
8578cea7 881 char buffer[PROC_NUMBUF];
1da177e4 882 size_t len;
28b83c51
KM		 883 int oom_adjust = OOM_DISABLE;
884 unsigned long flags;
1da177e4 885
99f89551
EB		 886 if (!task)
887 return -ESRCH;
28b83c51
KM		 888
889 if (lock_task_sighand(task, &flags)) {
890 oom_adjust = task->signal->oom_adj;
891 unlock_task_sighand(task, &flags);
892 }
893
99f89551
EB		 894 put_task_struct(task);
895
8578cea7 896 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
0c28f287
AM		 897
898 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1da177e4
LT		 899}
900
901static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
902 size_t count, loff_t *ppos)
903{
99f89551 904 struct task_struct *task;
5d863b89 905 char buffer[PROC_NUMBUF];
0a8cb8e3 906 int oom_adjust;
28b83c51 907 unsigned long flags;
5d863b89 908 int err;
1da177e4 909
8578cea7
EB		 910 memset(buffer, 0, sizeof(buffer));
911 if (count > sizeof(buffer) - 1)
912 count = sizeof(buffer) - 1;
723548bf
DR		 913 if (copy_from_user(buffer, buf, count)) {
914 err = -EFAULT;
915 goto out;
916 }
5d863b89 917
0a8cb8e3 918 err = kstrtoint(strstrip(buffer), 0, &oom_adjust);
5d863b89 919 if (err)
723548bf 920 goto out;
8ac773b4 921 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
723548bf
DR		 922 oom_adjust != OOM_DISABLE) {
923 err = -EINVAL;
924 goto out;
925 }
5d863b89 926
2fddfeef 927 task = get_proc_task(file->f_path.dentry->d_inode);
723548bf
DR		 928 if (!task) {
929 err = -ESRCH;
930 goto out;
931 }
d19d5476
DR		 932
933 task_lock(task);
934 if (!task->mm) {
935 err = -EINVAL;
936 goto err_task_lock;
937 }
938
28b83c51 939 if (!lock_task_sighand(task, &flags)) {
723548bf 940 err = -ESRCH;
d19d5476 941 goto err_task_lock;
28b83c51
KM		 942 }
943
944 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
723548bf
DR		 945 err = -EACCES;
946 goto err_sighand;
8fb4fc68 947 }
28b83c51 948
51b1bd2a
DR		 949 /*
950 * Warn that /proc/pid/oom_adj is deprecated, see
951 * Documentation/feature-removal-schedule.txt.
952 */
c2142704 953 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
be8f684d
DR		 954 current->comm, task_pid_nr(current), task_pid_nr(task),
955 task_pid_nr(task));
28b83c51 956 task->signal->oom_adj = oom_adjust;
a63d83f4
DR		 957 /*
958 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
959 * value is always attainable.
960 */
961 if (task->signal->oom_adj == OOM_ADJUST_MAX)
962 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
963 else
964 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
965 -OOM_DISABLE;
43d2b113 966 trace_oom_score_adj_update(task);
723548bf 967err_sighand:
28b83c51 968 unlock_task_sighand(task, &flags);
d19d5476
DR		 969err_task_lock:
970 task_unlock(task);
99f89551 971 put_task_struct(task);
723548bf
DR		 972out:
973 return err < 0 ? err : count;
1da177e4
LT		 974}
975
00977a59 976static const struct file_operations proc_oom_adjust_operations = {
1da177e4
LT		 977 .read = oom_adjust_read,
978 .write = oom_adjust_write,
87df8424 979 .llseek = generic_file_llseek,
1da177e4
LT		 980};
981
a63d83f4
DR		 982static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
983 size_t count, loff_t *ppos)
984{
985 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
986 char buffer[PROC_NUMBUF];
987 int oom_score_adj = OOM_SCORE_ADJ_MIN;
988 unsigned long flags;
989 size_t len;
990
991 if (!task)
992 return -ESRCH;
993 if (lock_task_sighand(task, &flags)) {
994 oom_score_adj = task->signal->oom_score_adj;
995 unlock_task_sighand(task, &flags);
996 }
997 put_task_struct(task);
998 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
999 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1000}
1001
1002static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1003 size_t count, loff_t *ppos)
1004{
1005 struct task_struct *task;
1006 char buffer[PROC_NUMBUF];
1007 unsigned long flags;
0a8cb8e3 1008 int oom_score_adj;
a63d83f4
DR		 1009 int err;
1010
1011 memset(buffer, 0, sizeof(buffer));
1012 if (count > sizeof(buffer) - 1)
1013 count = sizeof(buffer) - 1;
723548bf
DR		 1014 if (copy_from_user(buffer, buf, count)) {
1015 err = -EFAULT;
1016 goto out;
1017 }
a63d83f4 1018
0a8cb8e3 1019 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
a63d83f4 1020 if (err)
723548bf 1021 goto out;
a63d83f4 1022 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
723548bf
DR		 1023 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1024 err = -EINVAL;
1025 goto out;
1026 }
a63d83f4
DR		 1027
1028 task = get_proc_task(file->f_path.dentry->d_inode);
723548bf
DR		 1029 if (!task) {
1030 err = -ESRCH;
1031 goto out;
1032 }
d19d5476
DR		 1033
1034 task_lock(task);
1035 if (!task->mm) {
1036 err = -EINVAL;
1037 goto err_task_lock;
1038 }
1039
a63d83f4 1040 if (!lock_task_sighand(task, &flags)) {
723548bf 1041 err = -ESRCH;
d19d5476 1042 goto err_task_lock;
a63d83f4 1043 }
d19d5476 1044
dabb16f6 1045 if (oom_score_adj < task->signal->oom_score_adj_min &&
a63d83f4 1046 !capable(CAP_SYS_RESOURCE)) {
723548bf
DR		 1047 err = -EACCES;
1048 goto err_sighand;
a63d83f4
DR		 1049 }
1050
1051 task->signal->oom_score_adj = oom_score_adj;
dabb16f6
MSB		 1052 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1053 task->signal->oom_score_adj_min = oom_score_adj;
43d2b113 1054 trace_oom_score_adj_update(task);
a63d83f4
DR		 1055 /*
1056 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1057 * always attainable.
1058 */
1059 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1060 task->signal->oom_adj = OOM_DISABLE;
1061 else
1062 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1063 OOM_SCORE_ADJ_MAX;
723548bf 1064err_sighand:
a63d83f4 1065 unlock_task_sighand(task, &flags);
d19d5476
DR		 1066err_task_lock:
1067 task_unlock(task);
a63d83f4 1068 put_task_struct(task);
723548bf
DR		 1069out:
1070 return err < 0 ? err : count;
a63d83f4
DR		 1071}
1072
1073static const struct file_operations proc_oom_score_adj_operations = {
1074 .read = oom_score_adj_read,
1075 .write = oom_score_adj_write,
6038f373 1076 .llseek = default_llseek,
a63d83f4
DR		 1077};
1078
1da177e4
LT		 1079#ifdef CONFIG_AUDITSYSCALL
1080#define TMPBUFLEN 21
1081static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1082 size_t count, loff_t *ppos)
1083{
2fddfeef 1084 struct inode * inode = file->f_path.dentry->d_inode;
99f89551 1085 struct task_struct *task = get_proc_task(inode);
1da177e4
LT		 1086 ssize_t length;
1087 char tmpbuf[TMPBUFLEN];
1088
99f89551
EB		 1089 if (!task)
1090 return -ESRCH;
1da177e4 1091 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
0c11b942 1092 audit_get_loginuid(task));
99f89551 1093 put_task_struct(task);
1da177e4
LT		 1094 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1095}
1096
1097static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1098 size_t count, loff_t *ppos)
1099{
2fddfeef 1100 struct inode * inode = file->f_path.dentry->d_inode;
1da177e4
LT		 1101 char *page, *tmp;
1102 ssize_t length;
1da177e4
LT		 1103 uid_t loginuid;
1104
7dc52157
PM		 1105 rcu_read_lock();
1106 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1107 rcu_read_unlock();
1da177e4 1108 return -EPERM;
7dc52157
PM		 1109 }
1110 rcu_read_unlock();
1da177e4 1111
e0182909
AV		 1112 if (count >= PAGE_SIZE)
1113 count = PAGE_SIZE - 1;
1da177e4
LT		 1114
1115 if (*ppos != 0) {
1116 /* No partial writes. */
1117 return -EINVAL;
1118 }
e12ba74d 1119 page = (char*)__get_free_page(GFP_TEMPORARY);
1da177e4
LT		 1120 if (!page)
1121 return -ENOMEM;
1122 length = -EFAULT;
1123 if (copy_from_user(page, buf, count))
1124 goto out_free_page;
1125
e0182909 1126 page[count] = '\0';
1da177e4
LT		 1127 loginuid = simple_strtoul(page, &tmp, 10);
1128 if (tmp == page) {
1129 length = -EINVAL;
1130 goto out_free_page;
1131
1132 }
0a300be6 1133 length = audit_set_loginuid(loginuid);
1da177e4
LT		 1134 if (likely(length == 0))
1135 length = count;
1136
1137out_free_page:
1138 free_page((unsigned long) page);
1139 return length;
1140}
1141
00977a59 1142static const struct file_operations proc_loginuid_operations = {
1da177e4
LT		 1143 .read = proc_loginuid_read,
1144 .write = proc_loginuid_write,
87df8424 1145 .llseek = generic_file_llseek,
1da177e4 1146};
1e0bd755
EP		 1147
1148static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1149 size_t count, loff_t *ppos)
1150{
1151 struct inode * inode = file->f_path.dentry->d_inode;
1152 struct task_struct *task = get_proc_task(inode);
1153 ssize_t length;
1154 char tmpbuf[TMPBUFLEN];
1155
1156 if (!task)
1157 return -ESRCH;
1158 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1159 audit_get_sessionid(task));
1160 put_task_struct(task);
1161 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1162}
1163
1164static const struct file_operations proc_sessionid_operations = {
1165 .read = proc_sessionid_read,
87df8424 1166 .llseek = generic_file_llseek,
1e0bd755 1167};
1da177e4
LT		 1168#endif
1169
f4f154fd
AM		 1170#ifdef CONFIG_FAULT_INJECTION
1171static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1172 size_t count, loff_t *ppos)
1173{
1174 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1175 char buffer[PROC_NUMBUF];
1176 size_t len;
1177 int make_it_fail;
f4f154fd
AM		 1178
1179 if (!task)
1180 return -ESRCH;
1181 make_it_fail = task->make_it_fail;
1182 put_task_struct(task);
1183
1184 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
0c28f287
AM		 1185
1186 return simple_read_from_buffer(buf, count, ppos, buffer, len);
f4f154fd
AM		 1187}
1188
1189static ssize_t proc_fault_inject_write(struct file * file,
1190 const char __user * buf, size_t count, loff_t *ppos)
1191{
1192 struct task_struct *task;
1193 char buffer[PROC_NUMBUF], *end;
1194 int make_it_fail;
1195
1196 if (!capable(CAP_SYS_RESOURCE))
1197 return -EPERM;
1198 memset(buffer, 0, sizeof(buffer));
1199 if (count > sizeof(buffer) - 1)
1200 count = sizeof(buffer) - 1;
1201 if (copy_from_user(buffer, buf, count))
1202 return -EFAULT;
cba8aafe
VL		 1203 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1204 if (*end)
1205 return -EINVAL;
f4f154fd
AM		 1206 task = get_proc_task(file->f_dentry->d_inode);
1207 if (!task)
1208 return -ESRCH;
1209 task->make_it_fail = make_it_fail;
1210 put_task_struct(task);
cba8aafe
VL		 1211
1212 return count;
f4f154fd
AM		 1213}
1214
00977a59 1215static const struct file_operations proc_fault_inject_operations = {
f4f154fd
AM		 1216 .read = proc_fault_inject_read,
1217 .write = proc_fault_inject_write,
87df8424 1218 .llseek = generic_file_llseek,
f4f154fd
AM		 1219};
1220#endif
1221
9745512c 1222
43ae34cb
IM		 1223#ifdef CONFIG_SCHED_DEBUG
1224/*
1225 * Print out various scheduling related per-task fields:
1226 */
1227static int sched_show(struct seq_file *m, void *v)
1228{
1229 struct inode *inode = m->private;
1230 struct task_struct *p;
1231
43ae34cb
IM		 1232 p = get_proc_task(inode);
1233 if (!p)
1234 return -ESRCH;
1235 proc_sched_show_task(p, m);
1236
1237 put_task_struct(p);
1238
1239 return 0;
1240}
1241
1242static ssize_t
1243sched_write(struct file *file, const char __user *buf,
1244 size_t count, loff_t *offset)
1245{
1246 struct inode *inode = file->f_path.dentry->d_inode;
1247 struct task_struct *p;
1248
43ae34cb
IM		 1249 p = get_proc_task(inode);
1250 if (!p)
1251 return -ESRCH;
1252 proc_sched_set_task(p);
1253
1254 put_task_struct(p);
1255
1256 return count;
1257}
1258
1259static int sched_open(struct inode *inode, struct file *filp)
1260{
c6a34058 1261 return single_open(filp, sched_show, inode);
43ae34cb
IM		 1262}
1263
1264static const struct file_operations proc_pid_sched_operations = {
1265 .open = sched_open,
1266 .read = seq_read,
1267 .write = sched_write,
1268 .llseek = seq_lseek,
5ea473a1 1269 .release = single_release,
43ae34cb
IM		 1270};
1271
1272#endif
1273
5091faa4
MG		 1274#ifdef CONFIG_SCHED_AUTOGROUP
1275/*
1276 * Print out autogroup related information:
1277 */
1278static int sched_autogroup_show(struct seq_file *m, void *v)
1279{
1280 struct inode *inode = m->private;
1281 struct task_struct *p;
1282
1283 p = get_proc_task(inode);
1284 if (!p)
1285 return -ESRCH;
1286 proc_sched_autogroup_show_task(p, m);
1287
1288 put_task_struct(p);
1289
1290 return 0;
1291}
1292
1293static ssize_t
1294sched_autogroup_write(struct file *file, const char __user *buf,
1295 size_t count, loff_t *offset)
1296{
1297 struct inode *inode = file->f_path.dentry->d_inode;
1298 struct task_struct *p;
1299 char buffer[PROC_NUMBUF];
0a8cb8e3 1300 int nice;
5091faa4
MG		 1301 int err;
1302
1303 memset(buffer, 0, sizeof(buffer));
1304 if (count > sizeof(buffer) - 1)
1305 count = sizeof(buffer) - 1;
1306 if (copy_from_user(buffer, buf, count))
1307 return -EFAULT;
1308
0a8cb8e3
AD		 1309 err = kstrtoint(strstrip(buffer), 0, &nice);
1310 if (err < 0)
1311 return err;
5091faa4
MG		 1312
1313 p = get_proc_task(inode);
1314 if (!p)
1315 return -ESRCH;
1316
2e5b5b3a 1317 err = proc_sched_autogroup_set_nice(p, nice);
5091faa4
MG		 1318 if (err)
1319 count = err;
1320
1321 put_task_struct(p);
1322
1323 return count;
1324}
1325
1326static int sched_autogroup_open(struct inode *inode, struct file *filp)
1327{
1328 int ret;
1329
1330 ret = single_open(filp, sched_autogroup_show, NULL);
1331 if (!ret) {
1332 struct seq_file *m = filp->private_data;
1333
1334 m->private = inode;
1335 }
1336 return ret;
1337}
1338
1339static const struct file_operations proc_pid_sched_autogroup_operations = {
1340 .open = sched_autogroup_open,
1341 .read = seq_read,
1342 .write = sched_autogroup_write,
1343 .llseek = seq_lseek,
1344 .release = single_release,
1345};
1346
1347#endif /* CONFIG_SCHED_AUTOGROUP */
1348
4614a696 1349static ssize_t comm_write(struct file *file, const char __user *buf,
1350 size_t count, loff_t *offset)
1351{
1352 struct inode *inode = file->f_path.dentry->d_inode;
1353 struct task_struct *p;
1354 char buffer[TASK_COMM_LEN];
1355
1356 memset(buffer, 0, sizeof(buffer));
1357 if (count > sizeof(buffer) - 1)
1358 count = sizeof(buffer) - 1;
1359 if (copy_from_user(buffer, buf, count))
1360 return -EFAULT;
1361
1362 p = get_proc_task(inode);
1363 if (!p)
1364 return -ESRCH;
1365
1366 if (same_thread_group(current, p))
1367 set_task_comm(p, buffer);
1368 else
1369 count = -EINVAL;
1370
1371 put_task_struct(p);
1372
1373 return count;
1374}
1375
1376static int comm_show(struct seq_file *m, void *v)
1377{
1378 struct inode *inode = m->private;
1379 struct task_struct *p;
1380
1381 p = get_proc_task(inode);
1382 if (!p)
1383 return -ESRCH;
1384
1385 task_lock(p);
1386 seq_printf(m, "%s\n", p->comm);
1387 task_unlock(p);
1388
1389 put_task_struct(p);
1390
1391 return 0;
1392}
1393
1394static int comm_open(struct inode *inode, struct file *filp)
1395{
c6a34058 1396 return single_open(filp, comm_show, inode);
4614a696 1397}
1398
1399static const struct file_operations proc_pid_set_comm_operations = {
1400 .open = comm_open,
1401 .read = seq_read,
1402 .write = comm_write,
1403 .llseek = seq_lseek,
1404 .release = single_release,
1405};
1406
7773fbc5 1407static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
925d1c40
MH		 1408{
1409 struct task_struct *task;
1410 struct mm_struct *mm;
1411 struct file *exe_file;
1412
7773fbc5 1413 task = get_proc_task(dentry->d_inode);
925d1c40
MH		 1414 if (!task)
1415 return -ENOENT;
1416 mm = get_task_mm(task);
1417 put_task_struct(task);
1418 if (!mm)
1419 return -ENOENT;
1420 exe_file = get_mm_exe_file(mm);
1421 mmput(mm);
1422 if (exe_file) {
1423 *exe_path = exe_file->f_path;
1424 path_get(&exe_file->f_path);
1425 fput(exe_file);
1426 return 0;
1427 } else
1428 return -ENOENT;
1429}
1430
008b150a 1431static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1da177e4
LT		 1432{
1433 struct inode *inode = dentry->d_inode;
408ef013 1434 struct path path;
1da177e4
LT		 1435 int error = -EACCES;
1436
778c1144
EB		 1437 /* Are we allowed to snoop on the tasks file descriptors? */
1438 if (!proc_fd_access_allowed(inode))
1da177e4 1439 goto out;
1da177e4 1440
408ef013
CH		 1441 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1442 if (error)
1443 goto out;
1444
b5fb63c1 1445 nd_jump_link(nd, &path);
408ef013 1446 return NULL;
1da177e4 1447out:
008b150a 1448 return ERR_PTR(error);
1da177e4
LT		 1449}
1450
3dcd25f3 1451static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1da177e4 1452{
e12ba74d 1453 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
3dcd25f3 1454 char *pathname;
1da177e4
LT		 1455 int len;
1456
1457 if (!tmp)
1458 return -ENOMEM;
0c28f287 1459
7b2a69ba 1460 pathname = d_path(path, tmp, PAGE_SIZE);
3dcd25f3
JB		 1461 len = PTR_ERR(pathname);
1462 if (IS_ERR(pathname))
1da177e4 1463 goto out;
3dcd25f3 1464 len = tmp + PAGE_SIZE - 1 - pathname;
1da177e4
LT		 1465
1466 if (len > buflen)
1467 len = buflen;
3dcd25f3 1468 if (copy_to_user(buffer, pathname, len))
1da177e4
LT		 1469 len = -EFAULT;
1470 out:
1471 free_page((unsigned long)tmp);
1472 return len;
1473}
1474
1475static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1476{
1477 int error = -EACCES;
1478 struct inode *inode = dentry->d_inode;
3dcd25f3 1479 struct path path;
1da177e4 1480
778c1144
EB		 1481 /* Are we allowed to snoop on the tasks file descriptors? */
1482 if (!proc_fd_access_allowed(inode))
1da177e4 1483 goto out;
1da177e4 1484
7773fbc5 1485 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1da177e4
LT		 1486 if (error)
1487 goto out;
1488
3dcd25f3
JB		 1489 error = do_proc_readlink(&path, buffer, buflen);
1490 path_put(&path);
1da177e4 1491out:
1da177e4
LT		 1492 return error;
1493}
1494
c5ef1c42 1495static const struct inode_operations proc_pid_link_inode_operations = {
1da177e4 1496 .readlink = proc_pid_readlink,
6d76fa58
LT		 1497 .follow_link = proc_pid_follow_link,
1498 .setattr = proc_setattr,
1da177e4
LT		 1499};
1500
28a6d671
EB		 1501
1502/* building an inode */
1503
1504static int task_dumpable(struct task_struct *task)
1da177e4 1505{
28a6d671
EB		 1506 int dumpable = 0;
1507 struct mm_struct *mm;
1da177e4 1508
28a6d671
EB		 1509 task_lock(task);
1510 mm = task->mm;
1511 if (mm)
6c5d5238 1512 dumpable = get_dumpable(mm);
28a6d671
EB		 1513 task_unlock(task);
1514 if(dumpable == 1)
1515 return 1;
1516 return 0;
1517}
1da177e4 1518
6b4e306a 1519struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
28a6d671
EB		 1520{
1521 struct inode * inode;
1522 struct proc_inode *ei;
c69e8d9c 1523 const struct cred *cred;
1da177e4 1524
28a6d671 1525 /* We need a new inode */
1da177e4 1526
28a6d671
EB		 1527 inode = new_inode(sb);
1528 if (!inode)
1529 goto out;
1530
1531 /* Common stuff */
1532 ei = PROC_I(inode);
85fe4025 1533 inode->i_ino = get_next_ino();
28a6d671 1534 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
28a6d671
EB		 1535 inode->i_op = &proc_def_inode_operations;
1536
1537 /*
1538 * grab the reference to task.
1539 */
1a657f78 1540 ei->pid = get_task_pid(task, PIDTYPE_PID);
28a6d671
EB		 1541 if (!ei->pid)
1542 goto out_unlock;
1543
28a6d671 1544 if (task_dumpable(task)) {
c69e8d9c
DH		 1545 rcu_read_lock();
1546 cred = __task_cred(task);
1547 inode->i_uid = cred->euid;
1548 inode->i_gid = cred->egid;
1549 rcu_read_unlock();
1da177e4 1550 }
28a6d671
EB		 1551 security_task_to_inode(task, inode);
1552
1da177e4 1553out:
28a6d671
EB		 1554 return inode;
1555
1556out_unlock:
1557 iput(inode);
1558 return NULL;
1da177e4
LT		 1559}
1560
6b4e306a 1561int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1da177e4 1562{
1da177e4 1563 struct inode *inode = dentry->d_inode;
28a6d671 1564 struct task_struct *task;
c69e8d9c 1565 const struct cred *cred;
0499680a 1566 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
c69e8d9c 1567
28a6d671 1568 generic_fillattr(inode, stat);
1da177e4 1569
28a6d671 1570 rcu_read_lock();
dcb0f222
EB		 1571 stat->uid = GLOBAL_ROOT_UID;
1572 stat->gid = GLOBAL_ROOT_GID;
28a6d671
EB		 1573 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1574 if (task) {
0499680a
VK		 1575 if (!has_pid_permissions(pid, task, 2)) {
1576 rcu_read_unlock();
1577 /*
1578 * This doesn't prevent learning whether PID exists,
1579 * it only makes getattr() consistent with readdir().
1580 */
1581 return -ENOENT;
1582 }
28a6d671
EB		 1583 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1584 task_dumpable(task)) {
c69e8d9c
DH		 1585 cred = __task_cred(task);
1586 stat->uid = cred->euid;
1587 stat->gid = cred->egid;
1da177e4
LT		 1588 }
1589 }
28a6d671 1590 rcu_read_unlock();
d6e71144 1591 return 0;
1da177e4
LT		 1592}
1593
1da177e4
LT		 1594/* dentry stuff */
1595
1596/*
1597 * Exceptional case: normally we are not allowed to unhash a busy
1598 * directory. In this case, however, we can do it - no aliasing problems
1599 * due to the way we treat inodes.
1600 *
1601 * Rewrite the inode's ownerships here because the owning task may have
1602 * performed a setuid(), etc.
99f89551
EB		 1603 *
1604 * Before the /proc/pid/status file was created the only way to read
1605 * the effective uid of a /process was to stat /proc/pid. Reading
1606 * /proc/pid/status is slow enough that procps and other packages
1607 * kept stating /proc/pid. To keep the rules in /proc simple I have
1608 * made this apply to all per process world readable and executable
1609 * directories.
1da177e4 1610 */
0b728e19 1611int pid_revalidate(struct dentry *dentry, unsigned int flags)
1da177e4 1612{
34286d66
NP		 1613 struct inode *inode;
1614 struct task_struct *task;
c69e8d9c
DH		 1615 const struct cred *cred;
1616
0b728e19 1617 if (flags & LOOKUP_RCU)
34286d66
NP		 1618 return -ECHILD;
1619
1620 inode = dentry->d_inode;
1621 task = get_proc_task(inode);
1622
99f89551
EB		 1623 if (task) {
1624 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1625 task_dumpable(task)) {
c69e8d9c
DH		 1626 rcu_read_lock();
1627 cred = __task_cred(task);
1628 inode->i_uid = cred->euid;
1629 inode->i_gid = cred->egid;
1630 rcu_read_unlock();
1da177e4 1631 } else {
dcb0f222
EB		 1632 inode->i_uid = GLOBAL_ROOT_UID;
1633 inode->i_gid = GLOBAL_ROOT_GID;
1da177e4 1634 }
9ee8ab9f 1635 inode->i_mode &= ~(S_ISUID | S_ISGID);
1da177e4 1636 security_task_to_inode(task, inode);
99f89551 1637 put_task_struct(task);
1da177e4
LT		 1638 return 1;
1639 }
1640 d_drop(dentry);
1641 return 0;
1642}
1643
fe15ce44 1644static int pid_delete_dentry(const struct dentry * dentry)
99f89551 1645{
28a6d671
EB		 1646 /* Is the task we represent dead?
1647 * If so, then don't put the dentry on the lru list,
1648 * kill it immediately.
1649 */
1650 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1651}
1652
6b4e306a 1653const struct dentry_operations pid_dentry_operations =
28a6d671
EB		 1654{
1655 .d_revalidate = pid_revalidate,
1656 .d_delete = pid_delete_dentry,
1657};
1658
1659/* Lookups */
1660
1c0d04c9
EB		 1661/*
1662 * Fill a directory entry.
1663 *
1664 * If possible create the dcache entry and derive our inode number and
1665 * file type from dcache entry.
1666 *
1667 * Since all of the proc inode numbers are dynamically generated, the inode
1668 * numbers do not exist until the inode is cache. This means creating the
1669 * the dcache entry in readdir is necessary to keep the inode numbers
1670 * reported by readdir in sync with the inode numbers reported
1671 * by stat.
1672 */
6b4e306a
EB		 1673int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1674 const char *name, int len,
c5141e6d 1675 instantiate_t instantiate, struct task_struct *task, const void *ptr)
61a28784 1676{
2fddfeef 1677 struct dentry *child, *dir = filp->f_path.dentry;
61a28784
EB		 1678 struct inode *inode;
1679 struct qstr qname;
1680 ino_t ino = 0;
1681 unsigned type = DT_UNKNOWN;
1682
1683 qname.name = name;
1684 qname.len = len;
1685 qname.hash = full_name_hash(name, len);
1686
1687 child = d_lookup(dir, &qname);
1688 if (!child) {
1689 struct dentry *new;
1690 new = d_alloc(dir, &qname);
1691 if (new) {
1692 child = instantiate(dir->d_inode, new, task, ptr);
1693 if (child)
1694 dput(new);
1695 else
1696 child = new;
1697 }
1698 }
1699 if (!child || IS_ERR(child) || !child->d_inode)
1700 goto end_instantiate;
1701 inode = child->d_inode;
1702 if (inode) {
1703 ino = inode->i_ino;
1704 type = inode->i_mode >> 12;
1705 }
1706 dput(child);
1707end_instantiate:
1708 if (!ino)
1709 ino = find_inode_number(dir, &qname);
1710 if (!ino)
1711 ino = 1;
1712 return filldir(dirent, name, len, filp->f_pos, ino, type);
1713}
1714
28a6d671
EB		 1715static unsigned name_to_int(struct dentry *dentry)
1716{
1717 const char *name = dentry->d_name.name;
1718 int len = dentry->d_name.len;
1719 unsigned n = 0;
1720
1721 if (len > 1 && *name == '0')
1722 goto out;
1723 while (len-- > 0) {
1724 unsigned c = *name++ - '0';
1725 if (c > 9)
1726 goto out;
1727 if (n >= (~0U-9)/10)
1728 goto out;
1729 n *= 10;
1730 n += c;
1731 }
1732 return n;
1733out:
1734 return ~0U;
1735}
1736
27932742
MS		 1737#define PROC_FDINFO_MAX 64
1738
3dcd25f3 1739static int proc_fd_info(struct inode *inode, struct path *path, char *info)
28a6d671 1740{
5e442a49
LT		 1741 struct task_struct *task = get_proc_task(inode);
1742 struct files_struct *files = NULL;
28a6d671
EB		 1743 struct file *file;
1744 int fd = proc_fd(inode);
aa6afca5 1745
5e442a49
LT		 1746 if (task) {
1747 files = get_files_struct(task);
1748 put_task_struct(task);
1749 }
1750 if (files) {
1751 /*
1752 * We are not taking a ref to the file structure, so we must
1753 * hold ->file_lock.
1754 */
1755 spin_lock(&files->file_lock);
1756 file = fcheck_files(files, fd);
1757 if (file) {
1758 unsigned int f_flags;
1759 struct fdtable *fdt;
1760
1761 fdt = files_fdtable(files);
1762 f_flags = file->f_flags & ~O_CLOEXEC;
1dce27c5 1763 if (close_on_exec(fd, fdt))
5e442a49
LT		 1764 f_flags |= O_CLOEXEC;
1765
1766 if (path) {
1767 *path = file->f_path;
1768 path_get(&file->f_path);
1769 }
1770 if (info)
1771 snprintf(info, PROC_FDINFO_MAX,
1772 "pos:\t%lli\n"
1773 "flags:\t0%o\n",
1774 (long long) file->f_pos,
1775 f_flags);
1776 spin_unlock(&files->file_lock);
1777 put_files_struct(files);
1778 return 0;
99f89551 1779 }
5e442a49
LT		 1780 spin_unlock(&files->file_lock);
1781 put_files_struct(files);
1782 }
1783 return -ENOENT;
99f89551
EB		 1784}
1785
7773fbc5 1786static int proc_fd_link(struct dentry *dentry, struct path *path)
27932742 1787{
7773fbc5 1788 return proc_fd_info(dentry->d_inode, path, NULL);
27932742
MS		 1789}
1790
0b728e19 1791static int tid_fd_revalidate(struct dentry *dentry, unsigned int flags)
1da177e4 1792{
34286d66
NP		 1793 struct inode *inode;
1794 struct task_struct *task;
1795 int fd;
1da177e4 1796 struct files_struct *files;
c69e8d9c 1797 const struct cred *cred;
1da177e4 1798
0b728e19 1799 if (flags & LOOKUP_RCU)
34286d66
NP		 1800 return -ECHILD;
1801
1802 inode = dentry->d_inode;
1803 task = get_proc_task(inode);
1804 fd = proc_fd(inode);
1805
99f89551
EB		 1806 if (task) {
1807 files = get_files_struct(task);
1808 if (files) {
30a08bf2 1809 struct file *file;
99f89551 1810 rcu_read_lock();
30a08bf2
LT		 1811 file = fcheck_files(files, fd);
1812 if (file) {
0640113b 1813 unsigned f_mode = file->f_mode;
30a08bf2 1814
99f89551
EB		 1815 rcu_read_unlock();
1816 put_files_struct(files);
30a08bf2 1817
99f89551 1818 if (task_dumpable(task)) {
c69e8d9c
DH		 1819 rcu_read_lock();
1820 cred = __task_cred(task);
1821 inode->i_uid = cred->euid;
1822 inode->i_gid = cred->egid;
1823 rcu_read_unlock();
99f89551 1824 } else {
dcb0f222
EB		 1825 inode->i_uid = GLOBAL_ROOT_UID;
1826 inode->i_gid = GLOBAL_ROOT_GID;
99f89551 1827 }
30a08bf2 1828
0640113b
LT		 1829 if (S_ISLNK(inode->i_mode)) {
1830 unsigned i_mode = S_IFLNK;
1831 if (f_mode & FMODE_READ)
1832 i_mode |= S_IRUSR | S_IXUSR;
1833 if (f_mode & FMODE_WRITE)
1834 i_mode |= S_IWUSR | S_IXUSR;
1835 inode->i_mode = i_mode;
1836 }
30a08bf2 1837
99f89551
EB		 1838 security_task_to_inode(task, inode);
1839 put_task_struct(task);
1840 return 1;
1841 }
b835996f 1842 rcu_read_unlock();
1da177e4 1843 put_files_struct(files);
1da177e4 1844 }
99f89551 1845 put_task_struct(task);
1da177e4
LT		 1846 }
1847 d_drop(dentry);
1848 return 0;
1849}
1850
d72f71eb 1851static const struct dentry_operations tid_fd_dentry_operations =
1da177e4
LT		 1852{
1853 .d_revalidate = tid_fd_revalidate,
1854 .d_delete = pid_delete_dentry,
1855};
1856
444ceed8 1857static struct dentry *proc_fd_instantiate(struct inode *dir,
c5141e6d 1858 struct dentry *dentry, struct task_struct *task, const void *ptr)
1da177e4 1859{
af5e6171 1860 unsigned fd = (unsigned long)ptr;
444ceed8
EB		 1861 struct inode *inode;
1862 struct proc_inode *ei;
1863 struct dentry *error = ERR_PTR(-ENOENT);
1da177e4 1864
61a28784 1865 inode = proc_pid_make_inode(dir->i_sb, task);
1da177e4
LT		 1866 if (!inode)
1867 goto out;
1868 ei = PROC_I(inode);
aed7a6c4 1869 ei->fd = fd;
444ceed8 1870
0640113b 1871 inode->i_mode = S_IFLNK;
5e442a49 1872 inode->i_op = &proc_pid_link_inode_operations;
1da177e4
LT		 1873 inode->i_size = 64;
1874 ei->op.proc_get_link = proc_fd_link;
fb045adb 1875 d_set_d_op(dentry, &tid_fd_dentry_operations);
1da177e4 1876 d_add(dentry, inode);
cd6a3ce9 1877 /* Close the race of the process dying before we return the dentry */
0b728e19 1878 if (tid_fd_revalidate(dentry, 0))
444ceed8 1879 error = NULL;
1da177e4 1880
444ceed8
EB		 1881 out:
1882 return error;
1da177e4
LT		 1883}
1884
27932742
MS		 1885static struct dentry *proc_lookupfd_common(struct inode *dir,
1886 struct dentry *dentry,
1887 instantiate_t instantiate)
444ceed8
EB		 1888{
1889 struct task_struct *task = get_proc_task(dir);
1890 unsigned fd = name_to_int(dentry);
1891 struct dentry *result = ERR_PTR(-ENOENT);
1892
1893 if (!task)
1894 goto out_no_task;
1895 if (fd == ~0U)
1896 goto out;
1897
af5e6171 1898 result = instantiate(dir, dentry, task, (void *)(unsigned long)fd);
444ceed8
EB		 1899out:
1900 put_task_struct(task);
1901out_no_task:
1902 return result;
1903}
1904
27932742
MS		 1905static int proc_readfd_common(struct file * filp, void * dirent,
1906 filldir_t filldir, instantiate_t instantiate)
28a6d671 1907{
2fddfeef 1908 struct dentry *dentry = filp->f_path.dentry;
28a6d671
EB		 1909 struct inode *inode = dentry->d_inode;
1910 struct task_struct *p = get_proc_task(inode);
457c2510 1911 unsigned int fd, ino;
28a6d671 1912 int retval;
28a6d671 1913 struct files_struct * files;
1da177e4 1914
28a6d671
EB		 1915 retval = -ENOENT;
1916 if (!p)
1917 goto out_no_task;
1918 retval = 0;
28a6d671
EB		 1919
1920 fd = filp->f_pos;
1921 switch (fd) {
1922 case 0:
1923 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
5e442a49 1924 goto out;
28a6d671
EB		 1925 filp->f_pos++;
1926 case 1:
1927 ino = parent_ino(dentry);
1928 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
5e442a49 1929 goto out;
28a6d671
EB		 1930 filp->f_pos++;
1931 default:
1932 files = get_files_struct(p);
1933 if (!files)
5e442a49 1934 goto out;
28a6d671 1935 rcu_read_lock();
28a6d671 1936 for (fd = filp->f_pos-2;
9b4f526c 1937 fd < files_fdtable(files)->max_fds;
28a6d671 1938 fd++, filp->f_pos++) {
27932742
MS		 1939 char name[PROC_NUMBUF];
1940 int len;
f05ed3f1 1941 int rv;
28a6d671
EB		 1942
1943 if (!fcheck_files(files, fd))
1944 continue;
1945 rcu_read_unlock();
1946
27932742 1947 len = snprintf(name, sizeof(name), "%d", fd);
f05ed3f1
AD		 1948 rv = proc_fill_cache(filp, dirent, filldir,
1949 name, len, instantiate, p,
af5e6171 1950 (void *)(unsigned long)fd);
f05ed3f1
AD		 1951 if (rv < 0)
1952 goto out_fd_loop;
28a6d671
EB		 1953 rcu_read_lock();
1954 }
1955 rcu_read_unlock();
f05ed3f1 1956out_fd_loop:
28a6d671
EB		 1957 put_files_struct(files);
1958 }
1959out:
1960 put_task_struct(p);
1961out_no_task:
1962 return retval;
1963}
1964
27932742 1965static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
00cd8dd3 1966 unsigned int flags)
27932742
MS		 1967{
1968 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1969}
1970
1971static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1972{
1973 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1974}
1975
1976static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1977 size_t len, loff_t *ppos)
1978{
1979 char tmp[PROC_FDINFO_MAX];
3dcd25f3 1980 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
27932742
MS		 1981 if (!err)
1982 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1983 return err;
1984}
1985
1986static const struct file_operations proc_fdinfo_file_operations = {
87df8424 1987 .open = nonseekable_open,
27932742 1988 .read = proc_fdinfo_read,
6038f373 1989 .llseek = no_llseek,
27932742
MS		 1990};
1991
00977a59 1992static const struct file_operations proc_fd_operations = {
28a6d671
EB		 1993 .read = generic_read_dir,
1994 .readdir = proc_readfd,
6038f373 1995 .llseek = default_llseek,
1da177e4
LT		 1996};
1997
640708a2
PE		 1998#ifdef CONFIG_CHECKPOINT_RESTORE
1999
2000/*
2001 * dname_to_vma_addr - maps a dentry name into two unsigned longs
2002 * which represent vma start and end addresses.
2003 */
2004static int dname_to_vma_addr(struct dentry *dentry,
2005 unsigned long *start, unsigned long *end)
2006{
2007 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
2008 return -EINVAL;
2009
2010 return 0;
2011}
2012
0b728e19 2013static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
640708a2
PE		 2014{
2015 unsigned long vm_start, vm_end;
2016 bool exact_vma_exists = false;
2017 struct mm_struct *mm = NULL;
2018 struct task_struct *task;
2019 const struct cred *cred;
2020 struct inode *inode;
2021 int status = 0;
2022
0b728e19 2023 if (flags & LOOKUP_RCU)
640708a2
PE		 2024 return -ECHILD;
2025
2026 if (!capable(CAP_SYS_ADMIN)) {
2027 status = -EACCES;
2028 goto out_notask;
2029 }
2030
2031 inode = dentry->d_inode;
2032 task = get_proc_task(inode);
2033 if (!task)
2034 goto out_notask;
2035
2344bec7
CW		 2036 mm = mm_access(task, PTRACE_MODE_READ);
2037 if (IS_ERR_OR_NULL(mm))
640708a2
PE		 2038 goto out;
2039
2040 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
2041 down_read(&mm->mmap_sem);
2042 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
2043 up_read(&mm->mmap_sem);
2044 }
2045
2046 mmput(mm);
2047
2048 if (exact_vma_exists) {
2049 if (task_dumpable(task)) {
2050 rcu_read_lock();
2051 cred = __task_cred(task);
2052 inode->i_uid = cred->euid;
2053 inode->i_gid = cred->egid;
2054 rcu_read_unlock();
2055 } else {
dcb0f222
EB		 2056 inode->i_uid = GLOBAL_ROOT_UID;
2057 inode->i_gid = GLOBAL_ROOT_GID;
640708a2
PE		 2058 }
2059 security_task_to_inode(task, inode);
2060 status = 1;
2061 }
2062
2063out:
2064 put_task_struct(task);
2065
2066out_notask:
2067 if (status <= 0)
2068 d_drop(dentry);
2069
2070 return status;
2071}
2072
2073static const struct dentry_operations tid_map_files_dentry_operations = {
2074 .d_revalidate = map_files_d_revalidate,
2075 .d_delete = pid_delete_dentry,
2076};
2077
2078static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
2079{
2080 unsigned long vm_start, vm_end;
2081 struct vm_area_struct *vma;
2082 struct task_struct *task;
2083 struct mm_struct *mm;
2084 int rc;
2085
2086 rc = -ENOENT;
2087 task = get_proc_task(dentry->d_inode);
2088 if (!task)
2089 goto out;
2090
2091 mm = get_task_mm(task);
2092 put_task_struct(task);
2093 if (!mm)
2094 goto out;
2095
2096 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2097 if (rc)
2098 goto out_mmput;
2099
2100 down_read(&mm->mmap_sem);
2101 vma = find_exact_vma(mm, vm_start, vm_end);
2102 if (vma && vma->vm_file) {
2103 *path = vma->vm_file->f_path;
2104 path_get(path);
2105 rc = 0;
2106 }
2107 up_read(&mm->mmap_sem);
2108
2109out_mmput:
2110 mmput(mm);
2111out:
2112 return rc;
2113}
2114
2115struct map_files_info {
2116 struct file *file;
2117 unsigned long len;
2118 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
2119};
2120
2121static struct dentry *
2122proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
2123 struct task_struct *task, const void *ptr)
2124{
2125 const struct file *file = ptr;
2126 struct proc_inode *ei;
2127 struct inode *inode;
2128
2129 if (!file)
2130 return ERR_PTR(-ENOENT);
2131
2132 inode = proc_pid_make_inode(dir->i_sb, task);
2133 if (!inode)
2134 return ERR_PTR(-ENOENT);
2135
2136 ei = PROC_I(inode);
2137 ei->op.proc_get_link = proc_map_files_get_link;
2138
2139 inode->i_op = &proc_pid_link_inode_operations;
2140 inode->i_size = 64;
2141 inode->i_mode = S_IFLNK;
2142
2143 if (file->f_mode & FMODE_READ)
2144 inode->i_mode |= S_IRUSR;
2145 if (file->f_mode & FMODE_WRITE)
2146 inode->i_mode |= S_IWUSR;
2147
2148 d_set_d_op(dentry, &tid_map_files_dentry_operations);
2149 d_add(dentry, inode);
2150
2151 return NULL;
2152}
2153
2154static struct dentry *proc_map_files_lookup(struct inode *dir,
00cd8dd3 2155 struct dentry *dentry, unsigned int flags)
640708a2
PE		 2156{
2157 unsigned long vm_start, vm_end;
2158 struct vm_area_struct *vma;
2159 struct task_struct *task;
2160 struct dentry *result;
2161 struct mm_struct *mm;
2162
2163 result = ERR_PTR(-EACCES);
2164 if (!capable(CAP_SYS_ADMIN))
2165 goto out;
2166
2167 result = ERR_PTR(-ENOENT);
2168 task = get_proc_task(dir);
2169 if (!task)
2170 goto out;
2171
2172 result = ERR_PTR(-EACCES);
eb94cd96 2173 if (!ptrace_may_access(task, PTRACE_MODE_READ))
640708a2
PE		 2174 goto out_put_task;
2175
2176 result = ERR_PTR(-ENOENT);
2177 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
eb94cd96 2178 goto out_put_task;
640708a2
PE		 2179
2180 mm = get_task_mm(task);
2181 if (!mm)
eb94cd96 2182 goto out_put_task;
640708a2
PE		 2183
2184 down_read(&mm->mmap_sem);
2185 vma = find_exact_vma(mm, vm_start, vm_end);
2186 if (!vma)
2187 goto out_no_vma;
2188
2189 result = proc_map_files_instantiate(dir, dentry, task, vma->vm_file);
2190
2191out_no_vma:
2192 up_read(&mm->mmap_sem);
2193 mmput(mm);
640708a2
PE		 2194out_put_task:
2195 put_task_struct(task);
2196out:
2197 return result;
2198}
2199
2200static const struct inode_operations proc_map_files_inode_operations = {
2201 .lookup = proc_map_files_lookup,
2202 .permission = proc_fd_permission,
2203 .setattr = proc_setattr,
2204};
2205
2206static int
2207proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
2208{
2209 struct dentry *dentry = filp->f_path.dentry;
2210 struct inode *inode = dentry->d_inode;
2211 struct vm_area_struct *vma;
2212 struct task_struct *task;
2213 struct mm_struct *mm;
2214 ino_t ino;
2215 int ret;
2216
2217 ret = -EACCES;
2218 if (!capable(CAP_SYS_ADMIN))
2219 goto out;
2220
2221 ret = -ENOENT;
2222 task = get_proc_task(inode);
2223 if (!task)
2224 goto out;
2225
2226 ret = -EACCES;
eb94cd96 2227 if (!ptrace_may_access(task, PTRACE_MODE_READ))
640708a2
PE		 2228 goto out_put_task;
2229
2230 ret = 0;
2231 switch (filp->f_pos) {
2232 case 0:
2233 ino = inode->i_ino;
2234 if (filldir(dirent, ".", 1, 0, ino, DT_DIR) < 0)
eb94cd96 2235 goto out_put_task;
640708a2
PE		 2236 filp->f_pos++;
2237 case 1:
2238 ino = parent_ino(dentry);
2239 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
eb94cd96 2240 goto out_put_task;
640708a2
PE		 2241 filp->f_pos++;
2242 default:
2243 {
2244 unsigned long nr_files, pos, i;
2245 struct flex_array *fa = NULL;
2246 struct map_files_info info;
2247 struct map_files_info *p;
2248
2249 mm = get_task_mm(task);
2250 if (!mm)
eb94cd96 2251 goto out_put_task;
640708a2
PE		 2252 down_read(&mm->mmap_sem);
2253
2254 nr_files = 0;
2255
2256 /*
2257 * We need two passes here:
2258 *
2259 * 1) Collect vmas of mapped files with mmap_sem taken
2260 * 2) Release mmap_sem and instantiate entries
2261 *
2262 * otherwise we get lockdep complained, since filldir()
2263 * routine might require mmap_sem taken in might_fault().
2264 */
2265
2266 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2267 if (vma->vm_file && ++pos > filp->f_pos)
2268 nr_files++;
2269 }
2270
2271 if (nr_files) {
2272 fa = flex_array_alloc(sizeof(info), nr_files,
2273 GFP_KERNEL);
2274 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2275 GFP_KERNEL)) {
2276 ret = -ENOMEM;
2277 if (fa)
2278 flex_array_free(fa);
2279 up_read(&mm->mmap_sem);
2280 mmput(mm);
eb94cd96 2281 goto out_put_task;
640708a2
PE		 2282 }
2283 for (i = 0, vma = mm->mmap, pos = 2; vma;
2284 vma = vma->vm_next) {
2285 if (!vma->vm_file)
2286 continue;
2287 if (++pos <= filp->f_pos)
2288 continue;
2289
2290 get_file(vma->vm_file);
2291 info.file = vma->vm_file;
2292 info.len = snprintf(info.name,
2293 sizeof(info.name), "%lx-%lx",
2294 vma->vm_start, vma->vm_end);
2295 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2296 BUG();
2297 }
2298 }
2299 up_read(&mm->mmap_sem);
2300
2301 for (i = 0; i < nr_files; i++) {
2302 p = flex_array_get(fa, i);
2303 ret = proc_fill_cache(filp, dirent, filldir,
2304 p->name, p->len,
2305 proc_map_files_instantiate,
2306 task, p->file);
2307 if (ret)
2308 break;
2309 filp->f_pos++;
2310 fput(p->file);
2311 }
2312 for (; i < nr_files; i++) {
2313 /*
2314 * In case of error don't forget
2315 * to put rest of file refs.
2316 */
2317 p = flex_array_get(fa, i);
2318 fput(p->file);
2319 }
2320 if (fa)
2321 flex_array_free(fa);
2322 mmput(mm);
2323 }
2324 }
2325
640708a2
PE		 2326out_put_task:
2327 put_task_struct(task);
2328out:
2329 return ret;
2330}
2331
2332static const struct file_operations proc_map_files_operations = {
2333 .read = generic_read_dir,
2334 .readdir = proc_map_files_readdir,
2335 .llseek = default_llseek,
2336};
2337
2338#endif /* CONFIG_CHECKPOINT_RESTORE */
2339
8948e11f
AD		 2340/*
2341 * /proc/pid/fd needs a special permission handler so that a process can still
2342 * access /proc/self/fd after it has executed a setuid().
2343 */
10556cb2 2344static int proc_fd_permission(struct inode *inode, int mask)
8948e11f 2345{
2830ba7f 2346 int rv = generic_permission(inode, mask);
8948e11f
AD		 2347 if (rv == 0)
2348 return 0;
2349 if (task_pid(current) == proc_pid(inode))
2350 rv = 0;
2351 return rv;
2352}
2353
1da177e4
LT		 2354/*
2355 * proc directories can do almost nothing..
2356 */
c5ef1c42 2357static const struct inode_operations proc_fd_inode_operations = {
1da177e4 2358 .lookup = proc_lookupfd,
8948e11f 2359 .permission = proc_fd_permission,
6d76fa58 2360 .setattr = proc_setattr,
1da177e4
LT		 2361};
2362
27932742
MS		 2363static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2364 struct dentry *dentry, struct task_struct *task, const void *ptr)
2365{
af5e6171 2366 unsigned fd = (unsigned long)ptr;
27932742
MS		 2367 struct inode *inode;
2368 struct proc_inode *ei;
2369 struct dentry *error = ERR_PTR(-ENOENT);
2370
2371 inode = proc_pid_make_inode(dir->i_sb, task);
2372 if (!inode)
2373 goto out;
2374 ei = PROC_I(inode);
2375 ei->fd = fd;
2376 inode->i_mode = S_IFREG | S_IRUSR;
2377 inode->i_fop = &proc_fdinfo_file_operations;
fb045adb 2378 d_set_d_op(dentry, &tid_fd_dentry_operations);
27932742
MS		 2379 d_add(dentry, inode);
2380 /* Close the race of the process dying before we return the dentry */
0b728e19 2381 if (tid_fd_revalidate(dentry, 0))
27932742
MS		 2382 error = NULL;
2383
2384 out:
2385 return error;
2386}
2387
2388static struct dentry *proc_lookupfdinfo(struct inode *dir,
2389 struct dentry *dentry,
00cd8dd3 2390 unsigned int flags)
27932742
MS		 2391{
2392 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2393}
2394
2395static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2396{
2397 return proc_readfd_common(filp, dirent, filldir,
2398 proc_fdinfo_instantiate);
2399}
2400
2401static const struct file_operations proc_fdinfo_operations = {
2402 .read = generic_read_dir,
2403 .readdir = proc_readfdinfo,
6038f373 2404 .llseek = default_llseek,
27932742
MS		 2405};
2406
2407/*
2408 * proc directories can do almost nothing..
2409 */
2410static const struct inode_operations proc_fdinfo_inode_operations = {
2411 .lookup = proc_lookupfdinfo,
2412 .setattr = proc_setattr,
2413};
2414
2415
444ceed8 2416static struct dentry *proc_pident_instantiate(struct inode *dir,
c5141e6d 2417 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8 2418{
c5141e6d 2419 const struct pid_entry *p = ptr;
444ceed8
EB		 2420 struct inode *inode;
2421 struct proc_inode *ei;
bd6daba9 2422 struct dentry *error = ERR_PTR(-ENOENT);
444ceed8 2423
61a28784 2424 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 2425 if (!inode)
2426 goto out;
2427
2428 ei = PROC_I(inode);
2429 inode->i_mode = p->mode;
2430 if (S_ISDIR(inode->i_mode))
bfe86848 2431 set_nlink(inode, 2); /* Use getattr to fix if necessary */
444ceed8
EB		 2432 if (p->iop)
2433 inode->i_op = p->iop;
2434 if (p->fop)
2435 inode->i_fop = p->fop;
2436 ei->op = p->op;
fb045adb 2437 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 2438 d_add(dentry, inode);
2439 /* Close the race of the process dying before we return the dentry */
0b728e19 2440 if (pid_revalidate(dentry, 0))
444ceed8
EB		 2441 error = NULL;
2442out:
2443 return error;
2444}
2445
1da177e4
LT		 2446static struct dentry *proc_pident_lookup(struct inode *dir,
2447 struct dentry *dentry,
c5141e6d 2448 const struct pid_entry *ents,
7bcd6b0e 2449 unsigned int nents)
1da177e4 2450{
cd6a3ce9 2451 struct dentry *error;
99f89551 2452 struct task_struct *task = get_proc_task(dir);
c5141e6d 2453 const struct pid_entry *p, *last;
1da177e4 2454
cd6a3ce9 2455 error = ERR_PTR(-ENOENT);
1da177e4 2456
99f89551
EB		 2457 if (!task)
2458 goto out_no_task;
1da177e4 2459
20cdc894
EB		 2460 /*
2461 * Yes, it does not scale. And it should not. Don't add
2462 * new entries into /proc/<tgid>/ without very good reasons.
2463 */
7bcd6b0e
EB		 2464 last = &ents[nents - 1];
2465 for (p = ents; p <= last; p++) {
1da177e4
LT		 2466 if (p->len != dentry->d_name.len)
2467 continue;
2468 if (!memcmp(dentry->d_name.name, p->name, p->len))
2469 break;
2470 }
7bcd6b0e 2471 if (p > last)
1da177e4
LT		 2472 goto out;
2473
444ceed8 2474 error = proc_pident_instantiate(dir, dentry, task, p);
1da177e4 2475out:
99f89551
EB		 2476 put_task_struct(task);
2477out_no_task:
cd6a3ce9 2478 return error;
1da177e4
LT		 2479}
2480
c5141e6d
ED		 2481static int proc_pident_fill_cache(struct file *filp, void *dirent,
2482 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
61a28784
EB		 2483{
2484 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2485 proc_pident_instantiate, task, p);
2486}
2487
28a6d671
EB		 2488static int proc_pident_readdir(struct file *filp,
2489 void *dirent, filldir_t filldir,
c5141e6d 2490 const struct pid_entry *ents, unsigned int nents)
28a6d671
EB		 2491{
2492 int i;
2fddfeef 2493 struct dentry *dentry = filp->f_path.dentry;
28a6d671
EB		 2494 struct inode *inode = dentry->d_inode;
2495 struct task_struct *task = get_proc_task(inode);
c5141e6d 2496 const struct pid_entry *p, *last;
28a6d671
EB		 2497 ino_t ino;
2498 int ret;
2499
2500 ret = -ENOENT;
2501 if (!task)
61a28784 2502 goto out_no_task;
28a6d671
EB		 2503
2504 ret = 0;
28a6d671
EB		 2505 i = filp->f_pos;
2506 switch (i) {
2507 case 0:
2508 ino = inode->i_ino;
2509 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2510 goto out;
2511 i++;
2512 filp->f_pos++;
2513 /* fall through */
2514 case 1:
2515 ino = parent_ino(dentry);
2516 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2517 goto out;
2518 i++;
2519 filp->f_pos++;
2520 /* fall through */
2521 default:
2522 i -= 2;
2523 if (i >= nents) {
2524 ret = 1;
2525 goto out;
2526 }
2527 p = ents + i;
7bcd6b0e
EB		 2528 last = &ents[nents - 1];
2529 while (p <= last) {
61a28784 2530 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
28a6d671
EB		 2531 goto out;
2532 filp->f_pos++;
2533 p++;
2534 }
2535 }
2536
2537 ret = 1;
2538out:
61a28784
EB		 2539 put_task_struct(task);
2540out_no_task:
28a6d671 2541 return ret;
1da177e4
LT		 2542}
2543
28a6d671
EB		 2544#ifdef CONFIG_SECURITY
2545static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2546 size_t count, loff_t *ppos)
2547{
2fddfeef 2548 struct inode * inode = file->f_path.dentry->d_inode;
04ff9708 2549 char *p = NULL;
28a6d671
EB		 2550 ssize_t length;
2551 struct task_struct *task = get_proc_task(inode);
2552
28a6d671 2553 if (!task)
04ff9708 2554 return -ESRCH;
28a6d671
EB		 2555
2556 length = security_getprocattr(task,
2fddfeef 2557 (char*)file->f_path.dentry->d_name.name,
04ff9708 2558 &p);
28a6d671 2559 put_task_struct(task);
04ff9708
AV		 2560 if (length > 0)
2561 length = simple_read_from_buffer(buf, count, ppos, p, length);
2562 kfree(p);
28a6d671 2563 return length;
1da177e4
LT		 2564}
2565
28a6d671
EB		 2566static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2567 size_t count, loff_t *ppos)
2568{
2fddfeef 2569 struct inode * inode = file->f_path.dentry->d_inode;
28a6d671
EB		 2570 char *page;
2571 ssize_t length;
2572 struct task_struct *task = get_proc_task(inode);
2573
2574 length = -ESRCH;
2575 if (!task)
2576 goto out_no_task;
2577 if (count > PAGE_SIZE)
2578 count = PAGE_SIZE;
2579
2580 /* No partial writes. */
2581 length = -EINVAL;
2582 if (*ppos != 0)
2583 goto out;
2584
2585 length = -ENOMEM;
e12ba74d 2586 page = (char*)__get_free_page(GFP_TEMPORARY);
28a6d671
EB		 2587 if (!page)
2588 goto out;
2589
2590 length = -EFAULT;
2591 if (copy_from_user(page, buf, count))
2592 goto out_free;
2593
107db7c7 2594 /* Guard against adverse ptrace interaction */
9b1bf12d 2595 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
107db7c7
DH		 2596 if (length < 0)
2597 goto out_free;
2598
28a6d671 2599 length = security_setprocattr(task,
2fddfeef 2600 (char*)file->f_path.dentry->d_name.name,
28a6d671 2601 (void*)page, count);
9b1bf12d 2602 mutex_unlock(&task->signal->cred_guard_mutex);
28a6d671
EB		 2603out_free:
2604 free_page((unsigned long) page);
2605out:
2606 put_task_struct(task);
2607out_no_task:
2608 return length;
2609}
2610
00977a59 2611static const struct file_operations proc_pid_attr_operations = {
28a6d671
EB		 2612 .read = proc_pid_attr_read,
2613 .write = proc_pid_attr_write,
87df8424 2614 .llseek = generic_file_llseek,
28a6d671
EB		 2615};
2616
c5141e6d 2617static const struct pid_entry attr_dir_stuff[] = {
631f9c18
AD		 2618 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2619 REG("prev", S_IRUGO, proc_pid_attr_operations),
2620 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2621 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2622 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2623 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
28a6d671
EB		 2624};
2625
72d9dcfc 2626static int proc_attr_dir_readdir(struct file * filp,
28a6d671
EB		 2627 void * dirent, filldir_t filldir)
2628{
2629 return proc_pident_readdir(filp,dirent,filldir,
72d9dcfc 2630 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2631}
2632
00977a59 2633static const struct file_operations proc_attr_dir_operations = {
1da177e4 2634 .read = generic_read_dir,
72d9dcfc 2635 .readdir = proc_attr_dir_readdir,
6038f373 2636 .llseek = default_llseek,
1da177e4
LT		 2637};
2638
72d9dcfc 2639static struct dentry *proc_attr_dir_lookup(struct inode *dir,
00cd8dd3 2640 struct dentry *dentry, unsigned int flags)
28a6d671 2641{
7bcd6b0e
EB		 2642 return proc_pident_lookup(dir, dentry,
2643 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2644}
2645
c5ef1c42 2646static const struct inode_operations proc_attr_dir_inode_operations = {
72d9dcfc 2647 .lookup = proc_attr_dir_lookup,
99f89551 2648 .getattr = pid_getattr,
6d76fa58 2649 .setattr = proc_setattr,
1da177e4
LT		 2650};
2651
28a6d671
EB		 2652#endif
2653
698ba7b5 2654#ifdef CONFIG_ELF_CORE
3cb4a0bb
KH		 2655static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2656 size_t count, loff_t *ppos)
2657{
2658 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2659 struct mm_struct *mm;
2660 char buffer[PROC_NUMBUF];
2661 size_t len;
2662 int ret;
2663
2664 if (!task)
2665 return -ESRCH;
2666
2667 ret = 0;
2668 mm = get_task_mm(task);
2669 if (mm) {
2670 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2671 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2672 MMF_DUMP_FILTER_SHIFT));
2673 mmput(mm);
2674 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2675 }
2676
2677 put_task_struct(task);
2678
2679 return ret;
2680}
2681
2682static ssize_t proc_coredump_filter_write(struct file *file,
2683 const char __user *buf,
2684 size_t count,
2685 loff_t *ppos)
2686{
2687 struct task_struct *task;
2688 struct mm_struct *mm;
2689 char buffer[PROC_NUMBUF], *end;
2690 unsigned int val;
2691 int ret;
2692 int i;
2693 unsigned long mask;
2694
2695 ret = -EFAULT;
2696 memset(buffer, 0, sizeof(buffer));
2697 if (count > sizeof(buffer) - 1)
2698 count = sizeof(buffer) - 1;
2699 if (copy_from_user(buffer, buf, count))
2700 goto out_no_task;
2701
2702 ret = -EINVAL;
2703 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2704 if (*end == '\n')
2705 end++;
2706 if (end - buffer == 0)
2707 goto out_no_task;
2708
2709 ret = -ESRCH;
2710 task = get_proc_task(file->f_dentry->d_inode);
2711 if (!task)
2712 goto out_no_task;
2713
2714 ret = end - buffer;
2715 mm = get_task_mm(task);
2716 if (!mm)
2717 goto out_no_mm;
2718
2719 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2720 if (val & mask)
2721 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2722 else
2723 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2724 }
2725
2726 mmput(mm);
2727 out_no_mm:
2728 put_task_struct(task);
2729 out_no_task:
2730 return ret;
2731}
2732
2733static const struct file_operations proc_coredump_filter_operations = {
2734 .read = proc_coredump_filter_read,
2735 .write = proc_coredump_filter_write,
87df8424 2736 .llseek = generic_file_llseek,
3cb4a0bb
KH		 2737};
2738#endif
2739
28a6d671
EB		 2740/*
2741 * /proc/self:
2742 */
2743static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2744 int buflen)
2745{
488e5bc4 2746 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
b55fcb22 2747 pid_t tgid = task_tgid_nr_ns(current, ns);
28a6d671 2748 char tmp[PROC_NUMBUF];
b55fcb22 2749 if (!tgid)
488e5bc4 2750 return -ENOENT;
b55fcb22 2751 sprintf(tmp, "%d", tgid);
28a6d671
EB		 2752 return vfs_readlink(dentry,buffer,buflen,tmp);
2753}
2754
2755static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2756{
488e5bc4 2757 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
b55fcb22 2758 pid_t tgid = task_tgid_nr_ns(current, ns);
7fee4868
AV		 2759 char *name = ERR_PTR(-ENOENT);
2760 if (tgid) {
2761 name = __getname();
2762 if (!name)
2763 name = ERR_PTR(-ENOMEM);
2764 else
2765 sprintf(name, "%d", tgid);
2766 }
2767 nd_set_link(nd, name);
2768 return NULL;
2769}
2770
2771static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2772 void *cookie)
2773{
2774 char *s = nd_get_link(nd);
2775 if (!IS_ERR(s))
2776 __putname(s);
28a6d671
EB		 2777}
2778
c5ef1c42 2779static const struct inode_operations proc_self_inode_operations = {
28a6d671
EB		 2780 .readlink = proc_self_readlink,
2781 .follow_link = proc_self_follow_link,
7fee4868 2782 .put_link = proc_self_put_link,
28a6d671
EB		 2783};
2784
801199ce
EB		 2785/*
2786 * proc base
2787 *
2788 * These are the directory entries in the root directory of /proc
2789 * that properly belong to the /proc filesystem, as they describe
2790 * describe something that is process related.
2791 */
c5141e6d 2792static const struct pid_entry proc_base_stuff[] = {
61a28784 2793 NOD("self", S_IFLNK|S_IRWXUGO,
801199ce 2794 &proc_self_inode_operations, NULL, {}),
801199ce
EB		 2795};
2796
444ceed8 2797static struct dentry *proc_base_instantiate(struct inode *dir,
c5141e6d 2798 struct dentry *dentry, struct task_struct *task, const void *ptr)
801199ce 2799{
c5141e6d 2800 const struct pid_entry *p = ptr;
801199ce 2801 struct inode *inode;
801199ce 2802 struct proc_inode *ei;
73d36460 2803 struct dentry *error;
801199ce
EB		 2804
2805 /* Allocate the inode */
2806 error = ERR_PTR(-ENOMEM);
2807 inode = new_inode(dir->i_sb);
2808 if (!inode)
2809 goto out;
2810
2811 /* Initialize the inode */
2812 ei = PROC_I(inode);
85fe4025 2813 inode->i_ino = get_next_ino();
801199ce 2814 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
801199ce
EB		 2815
2816 /*
2817 * grab the reference to the task.
2818 */
1a657f78 2819 ei->pid = get_task_pid(task, PIDTYPE_PID);
801199ce
EB		 2820 if (!ei->pid)
2821 goto out_iput;
2822
801199ce
EB		 2823 inode->i_mode = p->mode;
2824 if (S_ISDIR(inode->i_mode))
bfe86848 2825 set_nlink(inode, 2);
801199ce
EB		 2826 if (S_ISLNK(inode->i_mode))
2827 inode->i_size = 64;
2828 if (p->iop)
2829 inode->i_op = p->iop;
2830 if (p->fop)
2831 inode->i_fop = p->fop;
2832 ei->op = p->op;
801199ce
EB		 2833 d_add(dentry, inode);
2834 error = NULL;
2835out:
801199ce
EB		 2836 return error;
2837out_iput:
2838 iput(inode);
2839 goto out;
2840}
2841
444ceed8
EB		 2842static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2843{
2844 struct dentry *error;
2845 struct task_struct *task = get_proc_task(dir);
c5141e6d 2846 const struct pid_entry *p, *last;
444ceed8
EB		 2847
2848 error = ERR_PTR(-ENOENT);
2849
2850 if (!task)
2851 goto out_no_task;
2852
2853 /* Lookup the directory entry */
7bcd6b0e
EB		 2854 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2855 for (p = proc_base_stuff; p <= last; p++) {
444ceed8
EB		 2856 if (p->len != dentry->d_name.len)
2857 continue;
2858 if (!memcmp(dentry->d_name.name, p->name, p->len))
2859 break;
2860 }
7bcd6b0e 2861 if (p > last)
444ceed8
EB		 2862 goto out;
2863
2864 error = proc_base_instantiate(dir, dentry, task, p);
2865
2866out:
2867 put_task_struct(task);
2868out_no_task:
2869 return error;
2870}
2871
c5141e6d
ED		 2872static int proc_base_fill_cache(struct file *filp, void *dirent,
2873 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
61a28784
EB		 2874{
2875 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2876 proc_base_instantiate, task, p);
2877}
2878
aba76fdb 2879#ifdef CONFIG_TASK_IO_ACCOUNTING
297c5d92
AR		 2880static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2881{
940389b8 2882 struct task_io_accounting acct = task->ioac;
5995477a 2883 unsigned long flags;
293eb1e7 2884 int result;
5995477a 2885
293eb1e7
VK		 2886 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2887 if (result)
2888 return result;
2889
2890 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2891 result = -EACCES;
2892 goto out_unlock;
2893 }
1d1221f3 2894
5995477a
AR		 2895 if (whole && lock_task_sighand(task, &flags)) {
2896 struct task_struct *t = task;
2897
2898 task_io_accounting_add(&acct, &task->signal->ioac);
2899 while_each_thread(task, t)
2900 task_io_accounting_add(&acct, &t->ioac);
2901
2902 unlock_task_sighand(task, &flags);
297c5d92 2903 }
293eb1e7 2904 result = sprintf(buffer,
aba76fdb
AM		 2905 "rchar: %llu\n"
2906 "wchar: %llu\n"
2907 "syscr: %llu\n"
2908 "syscw: %llu\n"
2909 "read_bytes: %llu\n"
2910 "write_bytes: %llu\n"
2911 "cancelled_write_bytes: %llu\n",
7c44319d
AB		 2912 (unsigned long long)acct.rchar,
2913 (unsigned long long)acct.wchar,
2914 (unsigned long long)acct.syscr,
2915 (unsigned long long)acct.syscw,
2916 (unsigned long long)acct.read_bytes,
2917 (unsigned long long)acct.write_bytes,
2918 (unsigned long long)acct.cancelled_write_bytes);
293eb1e7
VK		 2919out_unlock:
2920 mutex_unlock(&task->signal->cred_guard_mutex);
2921 return result;
297c5d92
AR		 2922}
2923
2924static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2925{
2926 return do_io_accounting(task, buffer, 0);
aba76fdb 2927}
297c5d92
AR		 2928
2929static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2930{
2931 return do_io_accounting(task, buffer, 1);
2932}
2933#endif /* CONFIG_TASK_IO_ACCOUNTING */
aba76fdb 2934
22d917d8
EB		 2935#ifdef CONFIG_USER_NS
2936static int proc_id_map_open(struct inode *inode, struct file *file,
2937 struct seq_operations *seq_ops)
2938{
2939 struct user_namespace *ns = NULL;
2940 struct task_struct *task;
2941 struct seq_file *seq;
2942 int ret = -EINVAL;
2943
2944 task = get_proc_task(inode);
2945 if (task) {
2946 rcu_read_lock();
2947 ns = get_user_ns(task_cred_xxx(task, user_ns));
2948 rcu_read_unlock();
2949 put_task_struct(task);
2950 }
2951 if (!ns)
2952 goto err;
2953
2954 ret = seq_open(file, seq_ops);
2955 if (ret)
2956 goto err_put_ns;
2957
2958 seq = file->private_data;
2959 seq->private = ns;
2960
2961 return 0;
2962err_put_ns:
2963 put_user_ns(ns);
2964err:
2965 return ret;
2966}
2967
2968static int proc_id_map_release(struct inode *inode, struct file *file)
2969{
2970 struct seq_file *seq = file->private_data;
2971 struct user_namespace *ns = seq->private;
2972 put_user_ns(ns);
2973 return seq_release(inode, file);
2974}
2975
2976static int proc_uid_map_open(struct inode *inode, struct file *file)
2977{
2978 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2979}
2980
2981static int proc_gid_map_open(struct inode *inode, struct file *file)
2982{
2983 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2984}
2985
2986static const struct file_operations proc_uid_map_operations = {
2987 .open = proc_uid_map_open,
2988 .write = proc_uid_map_write,
2989 .read = seq_read,
2990 .llseek = seq_lseek,
2991 .release = proc_id_map_release,
2992};
2993
2994static const struct file_operations proc_gid_map_operations = {
2995 .open = proc_gid_map_open,
2996 .write = proc_gid_map_write,
2997 .read = seq_read,
2998 .llseek = seq_lseek,
2999 .release = proc_id_map_release,
3000};
3001#endif /* CONFIG_USER_NS */
3002
47830723
KC		 3003static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
3004 struct pid *pid, struct task_struct *task)
3005{
a9712bc1
AV		 3006 int err = lock_trace(task);
3007 if (!err) {
3008 seq_printf(m, "%08x\n", task->personality);
3009 unlock_trace(task);
3010 }
3011 return err;
47830723
KC		 3012}
3013
28a6d671
EB		 3014/*
3015 * Thread groups
3016 */
00977a59 3017static const struct file_operations proc_task_operations;
c5ef1c42 3018static const struct inode_operations proc_task_inode_operations;
20cdc894 3019
c5141e6d 3020static const struct pid_entry tgid_base_stuff[] = {
631f9c18
AD		 3021 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
3022 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
640708a2
PE		 3023#ifdef CONFIG_CHECKPOINT_RESTORE
3024 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
3025#endif
631f9c18 3026 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 3027 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
b2211a36 3028#ifdef CONFIG_NET
631f9c18 3029 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
b2211a36 3030#endif
631f9c18
AD		 3031 REG("environ", S_IRUSR, proc_environ_operations),
3032 INF("auxv", S_IRUSR, proc_pid_auxv),
3033 ONE("status", S_IRUGO, proc_pid_status),
a9712bc1 3034 ONE("personality", S_IRUGO, proc_pid_personality),
3036e7b4 3035 INF("limits", S_IRUGO, proc_pid_limits),
43ae34cb 3036#ifdef CONFIG_SCHED_DEBUG
631f9c18 3037 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
5091faa4
MG		 3038#endif
3039#ifdef CONFIG_SCHED_AUTOGROUP
3040 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
ebcb6734 3041#endif
4614a696 3042 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 3043#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
a9712bc1 3044 INF("syscall", S_IRUGO, proc_pid_syscall),
43ae34cb 3045#endif
631f9c18
AD		 3046 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3047 ONE("stat", S_IRUGO, proc_tgid_stat),
3048 ONE("statm", S_IRUGO, proc_pid_statm),
b7643757 3049 REG("maps", S_IRUGO, proc_pid_maps_operations),
28a6d671 3050#ifdef CONFIG_NUMA
b7643757 3051 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
28a6d671 3052#endif
631f9c18
AD		 3053 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3054 LNK("cwd", proc_cwd_link),
3055 LNK("root", proc_root_link),
3056 LNK("exe", proc_exe_link),
3057 REG("mounts", S_IRUGO, proc_mounts_operations),
3058 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3059 REG("mountstats", S_IRUSR, proc_mountstats_operations),
1e883281 3060#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18 3061 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
b7643757 3062 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
ca6b0bf0 3063 REG("pagemap", S_IRUGO, proc_pagemap_operations),
28a6d671
EB		 3064#endif
3065#ifdef CONFIG_SECURITY
631f9c18 3066 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 3067#endif
3068#ifdef CONFIG_KALLSYMS
631f9c18 3069 INF("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 3070#endif
2ec220e2 3071#ifdef CONFIG_STACKTRACE
a9712bc1 3072 ONE("stack", S_IRUGO, proc_pid_stack),
28a6d671
EB		 3073#endif
3074#ifdef CONFIG_SCHEDSTATS
631f9c18 3075 INF("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 3076#endif
9745512c 3077#ifdef CONFIG_LATENCYTOP
631f9c18 3078 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 3079#endif
8793d854 3080#ifdef CONFIG_PROC_PID_CPUSET
631f9c18 3081 REG("cpuset", S_IRUGO, proc_cpuset_operations),
a424316c
PM		 3082#endif
3083#ifdef CONFIG_CGROUPS
631f9c18 3084 REG("cgroup", S_IRUGO, proc_cgroup_operations),
28a6d671 3085#endif
631f9c18
AD		 3086 INF("oom_score", S_IRUGO, proc_oom_score),
3087 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
a63d83f4 3088 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 3089#ifdef CONFIG_AUDITSYSCALL
631f9c18
AD		 3090 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3091 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 3092#endif
f4f154fd 3093#ifdef CONFIG_FAULT_INJECTION
631f9c18 3094 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 3095#endif
698ba7b5 3096#ifdef CONFIG_ELF_CORE
631f9c18 3097 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3cb4a0bb 3098#endif
aba76fdb 3099#ifdef CONFIG_TASK_IO_ACCOUNTING
1d1221f3 3100 INF("io", S_IRUSR, proc_tgid_io_accounting),
aba76fdb 3101#endif
f133ecca
CM		 3102#ifdef CONFIG_HARDWALL
3103 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3104#endif
22d917d8
EB		 3105#ifdef CONFIG_USER_NS
3106 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3107 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3108#endif
28a6d671 3109};
1da177e4 3110
28a6d671 3111static int proc_tgid_base_readdir(struct file * filp,
1da177e4
LT		 3112 void * dirent, filldir_t filldir)
3113{
3114 return proc_pident_readdir(filp,dirent,filldir,
28a6d671 3115 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 3116}
3117
00977a59 3118static const struct file_operations proc_tgid_base_operations = {
1da177e4 3119 .read = generic_read_dir,
28a6d671 3120 .readdir = proc_tgid_base_readdir,
6038f373 3121 .llseek = default_llseek,
1da177e4
LT		 3122};
3123
00cd8dd3
AV		 3124static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3125{
7bcd6b0e
EB		 3126 return proc_pident_lookup(dir, dentry,
3127 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 3128}
3129
c5ef1c42 3130static const struct inode_operations proc_tgid_base_inode_operations = {
28a6d671 3131 .lookup = proc_tgid_base_lookup,
99f89551 3132 .getattr = pid_getattr,
6d76fa58 3133 .setattr = proc_setattr,
0499680a 3134 .permission = proc_pid_permission,
1da177e4 3135};
1da177e4 3136
60347f67 3137static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
1da177e4 3138{
48e6484d 3139 struct dentry *dentry, *leader, *dir;
8578cea7 3140 char buf[PROC_NUMBUF];
48e6484d
EB		 3141 struct qstr name;
3142
3143 name.name = buf;
60347f67
PE		 3144 name.len = snprintf(buf, sizeof(buf), "%d", pid);
3145 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d 3146 if (dentry) {
29f12ca3 3147 shrink_dcache_parent(dentry);
48e6484d
EB		 3148 d_drop(dentry);
3149 dput(dentry);
3150 }
1da177e4 3151
48e6484d 3152 name.name = buf;
60347f67
PE		 3153 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
3154 leader = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d
EB		 3155 if (!leader)
3156 goto out;
1da177e4 3157
48e6484d
EB		 3158 name.name = "task";
3159 name.len = strlen(name.name);
3160 dir = d_hash_and_lookup(leader, &name);
3161 if (!dir)
3162 goto out_put_leader;
3163
3164 name.name = buf;
60347f67 3165 name.len = snprintf(buf, sizeof(buf), "%d", pid);
48e6484d
EB		 3166 dentry = d_hash_and_lookup(dir, &name);
3167 if (dentry) {
3168 shrink_dcache_parent(dentry);
3169 d_drop(dentry);
3170 dput(dentry);
1da177e4 3171 }
48e6484d
EB		 3172
3173 dput(dir);
3174out_put_leader:
3175 dput(leader);
3176out:
3177 return;
1da177e4
LT		 3178}
3179
0895e91d
RD		 3180/**
3181 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3182 * @task: task that should be flushed.
3183 *
3184 * When flushing dentries from proc, one needs to flush them from global
60347f67 3185 * proc (proc_mnt) and from all the namespaces' procs this task was seen
0895e91d
RD		 3186 * in. This call is supposed to do all of this job.
3187 *
3188 * Looks in the dcache for
3189 * /proc/@pid
3190 * /proc/@tgid/task/@pid
3191 * if either directory is present flushes it and all of it'ts children
3192 * from the dcache.
3193 *
3194 * It is safe and reasonable to cache /proc entries for a task until
3195 * that task exits. After that they just clog up the dcache with
3196 * useless entries, possibly causing useful dcache entries to be
3197 * flushed instead. This routine is proved to flush those useless
3198 * dcache entries at process exit time.
3199 *
3200 * NOTE: This routine is just an optimization so it does not guarantee
3201 * that no dcache entries will exist at process exit time it
3202 * just makes it very unlikely that any will persist.
60347f67
PE		 3203 */
3204
3205void proc_flush_task(struct task_struct *task)
3206{
9fcc2d15 3207 int i;
9b4d1cbe 3208 struct pid *pid, *tgid;
130f77ec
PE		 3209 struct upid *upid;
3210
130f77ec 3211 pid = task_pid(task);
9b4d1cbe 3212 tgid = task_tgid(task);
130f77ec 3213
9fcc2d15 3214 for (i = 0; i <= pid->level; i++) {
130f77ec
PE		 3215 upid = &pid->numbers[i];
3216 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
9b4d1cbe 3217 tgid->numbers[i].nr);
130f77ec 3218 }
6f4e6433
PE		 3219
3220 upid = &pid->numbers[pid->level];
3221 if (upid->nr == 1)
3222 pid_ns_release_proc(upid->ns);
60347f67
PE		 3223}
3224
9711ef99
AB		 3225static struct dentry *proc_pid_instantiate(struct inode *dir,
3226 struct dentry * dentry,
c5141e6d 3227 struct task_struct *task, const void *ptr)
444ceed8
EB		 3228{
3229 struct dentry *error = ERR_PTR(-ENOENT);
3230 struct inode *inode;
3231
61a28784 3232 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 3233 if (!inode)
3234 goto out;
3235
3236 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3237 inode->i_op = &proc_tgid_base_inode_operations;
3238 inode->i_fop = &proc_tgid_base_operations;
3239 inode->i_flags|=S_IMMUTABLE;
aed54175 3240
bfe86848
MS		 3241 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
3242 ARRAY_SIZE(tgid_base_stuff)));
444ceed8 3243
fb045adb 3244 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 3245
3246 d_add(dentry, inode);
3247 /* Close the race of the process dying before we return the dentry */
0b728e19 3248 if (pid_revalidate(dentry, 0))
444ceed8
EB		 3249 error = NULL;
3250out:
3251 return error;
3252}
3253
00cd8dd3 3254struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
1da177e4 3255{
73d36460 3256 struct dentry *result;
1da177e4 3257 struct task_struct *task;
1da177e4 3258 unsigned tgid;
b488893a 3259 struct pid_namespace *ns;
1da177e4 3260
801199ce
EB		 3261 result = proc_base_lookup(dir, dentry);
3262 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3263 goto out;
3264
1da177e4
LT		 3265 tgid = name_to_int(dentry);
3266 if (tgid == ~0U)
3267 goto out;
3268
b488893a 3269 ns = dentry->d_sb->s_fs_info;
de758734 3270 rcu_read_lock();
b488893a 3271 task = find_task_by_pid_ns(tgid, ns);
1da177e4
LT		 3272 if (task)
3273 get_task_struct(task);
de758734 3274 rcu_read_unlock();
1da177e4
LT		 3275 if (!task)
3276 goto out;
3277
444ceed8 3278 result = proc_pid_instantiate(dir, dentry, task, NULL);
1da177e4 3279 put_task_struct(task);
1da177e4 3280out:
cd6a3ce9 3281 return result;
1da177e4
LT		 3282}
3283
1da177e4 3284/*
0804ef4b 3285 * Find the first task with tgid >= tgid
0bc58a91 3286 *
1da177e4 3287 */
19fd4bb2
EB		 3288struct tgid_iter {
3289 unsigned int tgid;
0804ef4b 3290 struct task_struct *task;
19fd4bb2
EB		 3291};
3292static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3293{
0804ef4b 3294 struct pid *pid;
1da177e4 3295
19fd4bb2
EB		 3296 if (iter.task)
3297 put_task_struct(iter.task);
454cc105 3298 rcu_read_lock();
0804ef4b 3299retry:
19fd4bb2
EB		 3300 iter.task = NULL;
3301 pid = find_ge_pid(iter.tgid, ns);
0804ef4b 3302 if (pid) {
19fd4bb2
EB		 3303 iter.tgid = pid_nr_ns(pid, ns);
3304 iter.task = pid_task(pid, PIDTYPE_PID);
0804ef4b
EB		 3305 /* What we to know is if the pid we have find is the
3306 * pid of a thread_group_leader. Testing for task
3307 * being a thread_group_leader is the obvious thing
3308 * todo but there is a window when it fails, due to
3309 * the pid transfer logic in de_thread.
3310 *
3311 * So we perform the straight forward test of seeing
3312 * if the pid we have found is the pid of a thread
3313 * group leader, and don't worry if the task we have
3314 * found doesn't happen to be a thread group leader.
3315 * As we don't care in the case of readdir.
3316 */
19fd4bb2
EB		 3317 if (!iter.task || !has_group_leader_pid(iter.task)) {
3318 iter.tgid += 1;
0804ef4b 3319 goto retry;
19fd4bb2
EB		 3320 }
3321 get_task_struct(iter.task);
0bc58a91 3322 }
454cc105 3323 rcu_read_unlock();
19fd4bb2 3324 return iter;
1da177e4
LT		 3325}
3326
7bcd6b0e 3327#define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
0804ef4b 3328
61a28784 3329static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
19fd4bb2 3330 struct tgid_iter iter)
61a28784
EB		 3331{
3332 char name[PROC_NUMBUF];
19fd4bb2 3333 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
61a28784 3334 return proc_fill_cache(filp, dirent, filldir, name, len,
19fd4bb2 3335 proc_pid_instantiate, iter.task, NULL);
61a28784
EB		 3336}
3337
0499680a
VK		 3338static int fake_filldir(void *buf, const char *name, int namelen,
3339 loff_t offset, u64 ino, unsigned d_type)
3340{
3341 return 0;
3342}
3343
1da177e4
LT		 3344/* for the /proc/ directory itself, after non-process stuff has been done */
3345int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3346{
d8bdc59f
LT		 3347 unsigned int nr;
3348 struct task_struct *reaper;
19fd4bb2 3349 struct tgid_iter iter;
b488893a 3350 struct pid_namespace *ns;
0499680a 3351 filldir_t __filldir;
1da177e4 3352
d8bdc59f
LT		 3353 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3354 goto out_no_task;
3355 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3356
3357 reaper = get_proc_task(filp->f_path.dentry->d_inode);
61a28784
EB		 3358 if (!reaper)
3359 goto out_no_task;
3360
7bcd6b0e 3361 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
c5141e6d 3362 const struct pid_entry *p = &proc_base_stuff[nr];
61a28784 3363 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
801199ce 3364 goto out;
1da177e4
LT		 3365 }
3366
b488893a 3367 ns = filp->f_dentry->d_sb->s_fs_info;
19fd4bb2
EB		 3368 iter.task = NULL;
3369 iter.tgid = filp->f_pos - TGID_OFFSET;
3370 for (iter = next_tgid(ns, iter);
3371 iter.task;
3372 iter.tgid += 1, iter = next_tgid(ns, iter)) {
0499680a
VK		 3373 if (has_pid_permissions(ns, iter.task, 2))
3374 __filldir = filldir;
3375 else
3376 __filldir = fake_filldir;
3377
19fd4bb2 3378 filp->f_pos = iter.tgid + TGID_OFFSET;
0499680a 3379 if (proc_pid_fill_cache(filp, dirent, __filldir, iter) < 0) {
19fd4bb2 3380 put_task_struct(iter.task);
0804ef4b 3381 goto out;
1da177e4 3382 }
0bc58a91 3383 }
0804ef4b
EB		 3384 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3385out:
61a28784
EB		 3386 put_task_struct(reaper);
3387out_no_task:
0bc58a91
EB		 3388 return 0;
3389}
1da177e4 3390
28a6d671
EB		 3391/*
3392 * Tasks
3393 */
c5141e6d 3394static const struct pid_entry tid_base_stuff[] = {
631f9c18 3395 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3835541d 3396 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 3397 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
631f9c18
AD		 3398 REG("environ", S_IRUSR, proc_environ_operations),
3399 INF("auxv", S_IRUSR, proc_pid_auxv),
3400 ONE("status", S_IRUGO, proc_pid_status),
a9712bc1 3401 ONE("personality", S_IRUGO, proc_pid_personality),
3036e7b4 3402 INF("limits", S_IRUGO, proc_pid_limits),
43ae34cb 3403#ifdef CONFIG_SCHED_DEBUG
631f9c18 3404 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
ebcb6734 3405#endif
4614a696 3406 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 3407#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
a9712bc1 3408 INF("syscall", S_IRUGO, proc_pid_syscall),
43ae34cb 3409#endif
631f9c18
AD		 3410 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3411 ONE("stat", S_IRUGO, proc_tid_stat),
3412 ONE("statm", S_IRUGO, proc_pid_statm),
b7643757 3413 REG("maps", S_IRUGO, proc_tid_maps_operations),
81841161
CG		 3414#ifdef CONFIG_CHECKPOINT_RESTORE
3415 REG("children", S_IRUGO, proc_tid_children_operations),
3416#endif
28a6d671 3417#ifdef CONFIG_NUMA
b7643757 3418 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
28a6d671 3419#endif
631f9c18
AD		 3420 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3421 LNK("cwd", proc_cwd_link),
3422 LNK("root", proc_root_link),
3423 LNK("exe", proc_exe_link),
3424 REG("mounts", S_IRUGO, proc_mounts_operations),
3425 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
1e883281 3426#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18 3427 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
b7643757 3428 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
ca6b0bf0 3429 REG("pagemap", S_IRUGO, proc_pagemap_operations),
28a6d671
EB		 3430#endif
3431#ifdef CONFIG_SECURITY
631f9c18 3432 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 3433#endif
3434#ifdef CONFIG_KALLSYMS
631f9c18 3435 INF("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 3436#endif
2ec220e2 3437#ifdef CONFIG_STACKTRACE
a9712bc1 3438 ONE("stack", S_IRUGO, proc_pid_stack),
28a6d671
EB		 3439#endif
3440#ifdef CONFIG_SCHEDSTATS
631f9c18 3441 INF("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 3442#endif
9745512c 3443#ifdef CONFIG_LATENCYTOP
631f9c18 3444 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 3445#endif
8793d854 3446#ifdef CONFIG_PROC_PID_CPUSET
631f9c18 3447 REG("cpuset", S_IRUGO, proc_cpuset_operations),
a424316c
PM		 3448#endif
3449#ifdef CONFIG_CGROUPS
631f9c18 3450 REG("cgroup", S_IRUGO, proc_cgroup_operations),
28a6d671 3451#endif
631f9c18
AD		 3452 INF("oom_score", S_IRUGO, proc_oom_score),
3453 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
a63d83f4 3454 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 3455#ifdef CONFIG_AUDITSYSCALL
631f9c18 3456 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
26ec3c64 3457 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 3458#endif
f4f154fd 3459#ifdef CONFIG_FAULT_INJECTION
631f9c18 3460 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 3461#endif
297c5d92 3462#ifdef CONFIG_TASK_IO_ACCOUNTING
1d1221f3 3463 INF("io", S_IRUSR, proc_tid_io_accounting),
297c5d92 3464#endif
f133ecca
CM		 3465#ifdef CONFIG_HARDWALL
3466 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3467#endif
22d917d8
EB		 3468#ifdef CONFIG_USER_NS
3469 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3470 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3471#endif
28a6d671
EB		 3472};
3473
3474static int proc_tid_base_readdir(struct file * filp,
3475 void * dirent, filldir_t filldir)
3476{
3477 return proc_pident_readdir(filp,dirent,filldir,
3478 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3479}
3480
00cd8dd3
AV		 3481static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3482{
7bcd6b0e
EB		 3483 return proc_pident_lookup(dir, dentry,
3484 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
28a6d671
EB		 3485}
3486
00977a59 3487static const struct file_operations proc_tid_base_operations = {
28a6d671
EB		 3488 .read = generic_read_dir,
3489 .readdir = proc_tid_base_readdir,
6038f373 3490 .llseek = default_llseek,
28a6d671
EB		 3491};
3492
c5ef1c42 3493static const struct inode_operations proc_tid_base_inode_operations = {
28a6d671
EB		 3494 .lookup = proc_tid_base_lookup,
3495 .getattr = pid_getattr,
3496 .setattr = proc_setattr,
3497};
3498
444ceed8 3499static struct dentry *proc_task_instantiate(struct inode *dir,
c5141e6d 3500 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8
EB		 3501{
3502 struct dentry *error = ERR_PTR(-ENOENT);
3503 struct inode *inode;
61a28784 3504 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 3505
3506 if (!inode)
3507 goto out;
3508 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3509 inode->i_op = &proc_tid_base_inode_operations;
3510 inode->i_fop = &proc_tid_base_operations;
3511 inode->i_flags|=S_IMMUTABLE;
aed54175 3512
bfe86848
MS		 3513 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3514 ARRAY_SIZE(tid_base_stuff)));
444ceed8 3515
fb045adb 3516 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 3517
3518 d_add(dentry, inode);
3519 /* Close the race of the process dying before we return the dentry */
0b728e19 3520 if (pid_revalidate(dentry, 0))
444ceed8
EB		 3521 error = NULL;
3522out:
3523 return error;
3524}
3525
00cd8dd3 3526static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
28a6d671
EB		 3527{
3528 struct dentry *result = ERR_PTR(-ENOENT);
3529 struct task_struct *task;
3530 struct task_struct *leader = get_proc_task(dir);
28a6d671 3531 unsigned tid;
b488893a 3532 struct pid_namespace *ns;
28a6d671
EB		 3533
3534 if (!leader)
3535 goto out_no_task;
3536
3537 tid = name_to_int(dentry);
3538 if (tid == ~0U)
3539 goto out;
3540
b488893a 3541 ns = dentry->d_sb->s_fs_info;
28a6d671 3542 rcu_read_lock();
b488893a 3543 task = find_task_by_pid_ns(tid, ns);
28a6d671
EB		 3544 if (task)
3545 get_task_struct(task);
3546 rcu_read_unlock();
3547 if (!task)
3548 goto out;
bac0abd6 3549 if (!same_thread_group(leader, task))
28a6d671
EB		 3550 goto out_drop_task;
3551
444ceed8 3552 result = proc_task_instantiate(dir, dentry, task, NULL);
28a6d671
EB		 3553out_drop_task:
3554 put_task_struct(task);
3555out:
3556 put_task_struct(leader);
3557out_no_task:
3558 return result;
3559}
3560
0bc58a91
EB		 3561/*
3562 * Find the first tid of a thread group to return to user space.
3563 *
3564 * Usually this is just the thread group leader, but if the users
3565 * buffer was too small or there was a seek into the middle of the
3566 * directory we have more work todo.
3567 *
3568 * In the case of a short read we start with find_task_by_pid.
3569 *
3570 * In the case of a seek we start with the leader and walk nr
3571 * threads past it.
3572 */
cc288738 3573static struct task_struct *first_tid(struct task_struct *leader,
b488893a 3574 int tid, int nr, struct pid_namespace *ns)
0bc58a91 3575{
a872ff0c 3576 struct task_struct *pos;
1da177e4 3577
cc288738 3578 rcu_read_lock();
0bc58a91
EB		 3579 /* Attempt to start with the pid of a thread */
3580 if (tid && (nr > 0)) {
b488893a 3581 pos = find_task_by_pid_ns(tid, ns);
a872ff0c
ON		 3582 if (pos && (pos->group_leader == leader))
3583 goto found;
0bc58a91 3584 }
1da177e4 3585
0bc58a91 3586 /* If nr exceeds the number of threads there is nothing todo */
a872ff0c
ON		 3587 pos = NULL;
3588 if (nr && nr >= get_nr_threads(leader))
3589 goto out;
1da177e4 3590
a872ff0c
ON		 3591 /* If we haven't found our starting place yet start
3592 * with the leader and walk nr threads forward.
0bc58a91 3593 */
a872ff0c
ON		 3594 for (pos = leader; nr > 0; --nr) {
3595 pos = next_thread(pos);
3596 if (pos == leader) {
3597 pos = NULL;
3598 goto out;
3599 }
1da177e4 3600 }
a872ff0c
ON		 3601found:
3602 get_task_struct(pos);
3603out:
cc288738 3604 rcu_read_unlock();
0bc58a91
EB		 3605 return pos;
3606}
3607
3608/*
3609 * Find the next thread in the thread list.
3610 * Return NULL if there is an error or no next thread.
3611 *
3612 * The reference to the input task_struct is released.
3613 */
3614static struct task_struct *next_tid(struct task_struct *start)
3615{
c1df7fb8 3616 struct task_struct *pos = NULL;
cc288738 3617 rcu_read_lock();
c1df7fb8 3618 if (pid_alive(start)) {
0bc58a91 3619 pos = next_thread(start);
c1df7fb8
ON		 3620 if (thread_group_leader(pos))
3621 pos = NULL;
3622 else
3623 get_task_struct(pos);
3624 }
cc288738 3625 rcu_read_unlock();
0bc58a91
EB		 3626 put_task_struct(start);
3627 return pos;
1da177e4
LT		 3628}
3629
61a28784
EB		 3630static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3631 struct task_struct *task, int tid)
3632{
3633 char name[PROC_NUMBUF];
3634 int len = snprintf(name, sizeof(name), "%d", tid);
3635 return proc_fill_cache(filp, dirent, filldir, name, len,
3636 proc_task_instantiate, task, NULL);
3637}
3638
1da177e4
LT		 3639/* for the /proc/TGID/task/ directories */
3640static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3641{
2fddfeef 3642 struct dentry *dentry = filp->f_path.dentry;
1da177e4 3643 struct inode *inode = dentry->d_inode;
7d895244 3644 struct task_struct *leader = NULL;
0bc58a91 3645 struct task_struct *task;
1da177e4
LT		 3646 int retval = -ENOENT;
3647 ino_t ino;
0bc58a91 3648 int tid;
b488893a 3649 struct pid_namespace *ns;
1da177e4 3650
7d895244
GC		 3651 task = get_proc_task(inode);
3652 if (!task)
3653 goto out_no_task;
3654 rcu_read_lock();
3655 if (pid_alive(task)) {
3656 leader = task->group_leader;
3657 get_task_struct(leader);
3658 }
3659 rcu_read_unlock();
3660 put_task_struct(task);
99f89551
EB		 3661 if (!leader)
3662 goto out_no_task;
1da177e4
LT		 3663 retval = 0;
3664
ee568b25 3665 switch ((unsigned long)filp->f_pos) {
1da177e4
LT		 3666 case 0:
3667 ino = inode->i_ino;
ee6f779b 3668 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
1da177e4 3669 goto out;
ee6f779b 3670 filp->f_pos++;
1da177e4
LT		 3671 /* fall through */
3672 case 1:
3673 ino = parent_ino(dentry);
ee6f779b 3674 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
1da177e4 3675 goto out;
ee6f779b 3676 filp->f_pos++;
1da177e4
LT		 3677 /* fall through */
3678 }
3679
0bc58a91
EB		 3680 /* f_version caches the tgid value that the last readdir call couldn't
3681 * return. lseek aka telldir automagically resets f_version to 0.
3682 */
b488893a 3683 ns = filp->f_dentry->d_sb->s_fs_info;
2b47c361 3684 tid = (int)filp->f_version;
0bc58a91 3685 filp->f_version = 0;
ee6f779b 3686 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
0bc58a91 3687 task;
ee6f779b 3688 task = next_tid(task), filp->f_pos++) {
b488893a 3689 tid = task_pid_nr_ns(task, ns);
61a28784 3690 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
0bc58a91
EB		 3691 /* returning this tgid failed, save it as the first
3692 * pid for the next readir call */
2b47c361 3693 filp->f_version = (u64)tid;
0bc58a91 3694 put_task_struct(task);
1da177e4 3695 break;
0bc58a91 3696 }
1da177e4
LT		 3697 }
3698out:
99f89551
EB		 3699 put_task_struct(leader);
3700out_no_task:
1da177e4
LT		 3701 return retval;
3702}
6e66b52b
EB		 3703
3704static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3705{
3706 struct inode *inode = dentry->d_inode;
99f89551 3707 struct task_struct *p = get_proc_task(inode);
6e66b52b
EB		 3708 generic_fillattr(inode, stat);
3709
99f89551 3710 if (p) {
99f89551 3711 stat->nlink += get_nr_threads(p);
99f89551 3712 put_task_struct(p);
6e66b52b
EB		 3713 }
3714
3715 return 0;
3716}
28a6d671 3717
c5ef1c42 3718static const struct inode_operations proc_task_inode_operations = {
28a6d671
EB		 3719 .lookup = proc_task_lookup,
3720 .getattr = proc_task_getattr,
3721 .setattr = proc_setattr,
0499680a 3722 .permission = proc_pid_permission,
28a6d671
EB		 3723};
3724
00977a59 3725static const struct file_operations proc_task_operations = {
28a6d671
EB		 3726 .read = generic_read_dir,
3727 .readdir = proc_task_readdir,
6038f373 3728 .llseek = default_llseek,
28a6d671 3729};
Linux kernel - Deliverables
This page took 0.914396 seconds and 5 git commands to generate.