projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
procfs: Move /proc/pid/fd[info] handling code to fd.[ch]
[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 92#include "internal.h"
faf60af1 93#include "fd.h"
1da177e4 94
0f2fe20f
EB		 95/* NOTE:
96 * Implementing inode permission operations in /proc is almost
97 * certainly an error. Permission checks need to happen during
98 * each system call not at open time. The reason is that most of
99 * what we wish to check for permissions in /proc varies at runtime.
100 *
101 * The classic example of a problem is opening file descriptors
102 * in /proc for a task before it execs a suid executable.
103 */
104
1da177e4 105struct pid_entry {
1da177e4 106 char *name;
c5141e6d 107 int len;
d161a13f 108 umode_t mode;
c5ef1c42 109 const struct inode_operations *iop;
00977a59 110 const struct file_operations *fop;
20cdc894 111 union proc_op op;
1da177e4
LT		 112};
113
61a28784 114#define NOD(NAME, MODE, IOP, FOP, OP) { \
20cdc894 115 .name = (NAME), \
c5141e6d 116 .len = sizeof(NAME) - 1, \
20cdc894
EB		 117 .mode = MODE, \
118 .iop = IOP, \
119 .fop = FOP, \
120 .op = OP, \
121}
122
631f9c18
AD		 123#define DIR(NAME, MODE, iops, fops) \
124 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
125#define LNK(NAME, get_link) \
61a28784 126 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
20cdc894 127 &proc_pid_link_inode_operations, NULL, \
631f9c18
AD		 128 { .proc_get_link = get_link } )
129#define REG(NAME, MODE, fops) \
130 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
131#define INF(NAME, MODE, read) \
61a28784 132 NOD(NAME, (S_IFREG|(MODE)), \
20cdc894 133 NULL, &proc_info_file_operations, \
631f9c18
AD		 134 { .proc_read = read } )
135#define ONE(NAME, MODE, show) \
be614086
EB		 136 NOD(NAME, (S_IFREG|(MODE)), \
137 NULL, &proc_single_file_operations, \
631f9c18 138 { .proc_show = show } )
1da177e4 139
aed54175
VN		 140/*
141 * Count the number of hardlinks for the pid_entry table, excluding the .
142 * and .. links.
143 */
144static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
145 unsigned int n)
146{
147 unsigned int i;
148 unsigned int count;
149
150 count = 0;
151 for (i = 0; i < n; ++i) {
152 if (S_ISDIR(entries[i].mode))
153 ++count;
154 }
155
156 return count;
157}
158
f7ad3c6b 159static int get_task_root(struct task_struct *task, struct path *root)
1da177e4 160{
7c2c7d99
HD		 161 int result = -ENOENT;
162
0494f6ec 163 task_lock(task);
f7ad3c6b
MS		 164 if (task->fs) {
165 get_fs_root(task->fs, root);
7c2c7d99
HD		 166 result = 0;
167 }
0494f6ec 168 task_unlock(task);
7c2c7d99 169 return result;
0494f6ec
MS		 170}
171
7773fbc5 172static int proc_cwd_link(struct dentry *dentry, struct path *path)
0494f6ec 173{
7773fbc5 174 struct task_struct *task = get_proc_task(dentry->d_inode);
0494f6ec 175 int result = -ENOENT;
99f89551
EB		 176
177 if (task) {
f7ad3c6b
MS		 178 task_lock(task);
179 if (task->fs) {
180 get_fs_pwd(task->fs, path);
181 result = 0;
182 }
183 task_unlock(task);
99f89551
EB		 184 put_task_struct(task);
185 }
1da177e4
LT		 186 return result;
187}
188
7773fbc5 189static int proc_root_link(struct dentry *dentry, struct path *path)
1da177e4 190{
7773fbc5 191 struct task_struct *task = get_proc_task(dentry->d_inode);
1da177e4 192 int result = -ENOENT;
99f89551
EB		 193
194 if (task) {
f7ad3c6b 195 result = get_task_root(task, path);
99f89551
EB		 196 put_task_struct(task);
197 }
1da177e4
LT		 198 return result;
199}
200
1da177e4
LT		 201static int proc_pid_cmdline(struct task_struct *task, char * buffer)
202{
203 int res = 0;
204 unsigned int len;
205 struct mm_struct *mm = get_task_mm(task);
206 if (!mm)
207 goto out;
208 if (!mm->arg_end)
209 goto out_mm; /* Shh! No looking before we're done */
210
211 len = mm->arg_end - mm->arg_start;
212
213 if (len > PAGE_SIZE)
214 len = PAGE_SIZE;
215
216 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
217
218 // If the nul at the end of args has been overwritten, then
219 // assume application is using setproctitle(3).
220 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
221 len = strnlen(buffer, res);
222 if (len < res) {
223 res = len;
224 } else {
225 len = mm->env_end - mm->env_start;
226 if (len > PAGE_SIZE - res)
227 len = PAGE_SIZE - res;
228 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
229 res = strnlen(buffer, res);
230 }
231 }
232out_mm:
233 mmput(mm);
234out:
235 return res;
236}
237
238static int proc_pid_auxv(struct task_struct *task, char *buffer)
239{
e7dcd999 240 struct mm_struct *mm = mm_access(task, PTRACE_MODE_READ);
2fadaef4
AV		 241 int res = PTR_ERR(mm);
242 if (mm && !IS_ERR(mm)) {
1da177e4 243 unsigned int nwords = 0;
dfe6b7d9 244 do {
1da177e4 245 nwords += 2;
dfe6b7d9 246 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
1da177e4
LT		 247 res = nwords * sizeof(mm->saved_auxv[0]);
248 if (res > PAGE_SIZE)
249 res = PAGE_SIZE;
250 memcpy(buffer, mm->saved_auxv, res);
251 mmput(mm);
252 }
253 return res;
254}
255
256
257#ifdef CONFIG_KALLSYMS
258/*
259 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
260 * Returns the resolved symbol. If that fails, simply return the address.
261 */
262static int proc_pid_wchan(struct task_struct *task, char *buffer)
263{
ffb45122 264 unsigned long wchan;
9281acea 265 char symname[KSYM_NAME_LEN];
1da177e4
LT		 266
267 wchan = get_wchan(task);
268
9d65cb4a 269 if (lookup_symbol_name(wchan, symname) < 0)
f83ce3e6
JE		 270 if (!ptrace_may_access(task, PTRACE_MODE_READ))
271 return 0;
272 else
273 return sprintf(buffer, "%lu", wchan);
9d65cb4a
AD		 274 else
275 return sprintf(buffer, "%s", symname);
1da177e4
LT		 276}
277#endif /* CONFIG_KALLSYMS */
278
a9712bc1
AV		 279static int lock_trace(struct task_struct *task)
280{
281 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
282 if (err)
283 return err;
284 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
285 mutex_unlock(&task->signal->cred_guard_mutex);
286 return -EPERM;
287 }
288 return 0;
289}
290
291static void unlock_trace(struct task_struct *task)
292{
293 mutex_unlock(&task->signal->cred_guard_mutex);
294}
295
2ec220e2
KC		 296#ifdef CONFIG_STACKTRACE
297
298#define MAX_STACK_TRACE_DEPTH 64
299
300static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
301 struct pid *pid, struct task_struct *task)
302{
303 struct stack_trace trace;
304 unsigned long *entries;
a9712bc1 305 int err;
2ec220e2
KC		 306 int i;
307
308 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
309 if (!entries)
310 return -ENOMEM;
311
312 trace.nr_entries = 0;
313 trace.max_entries = MAX_STACK_TRACE_DEPTH;
314 trace.entries = entries;
315 trace.skip = 0;
2ec220e2 316
a9712bc1
AV		 317 err = lock_trace(task);
318 if (!err) {
319 save_stack_trace_tsk(task, &trace);
320
321 for (i = 0; i < trace.nr_entries; i++) {
b81a618d 322 seq_printf(m, "[<%pK>] %pS\n",
a9712bc1
AV		 323 (void *)entries[i], (void *)entries[i]);
324 }
325 unlock_trace(task);
2ec220e2
KC		 326 }
327 kfree(entries);
328
a9712bc1 329 return err;
2ec220e2
KC		 330}
331#endif
332
1da177e4
LT		 333#ifdef CONFIG_SCHEDSTATS
334/*
335 * Provides /proc/PID/schedstat
336 */
337static int proc_pid_schedstat(struct task_struct *task, char *buffer)
338{
172ba844 339 return sprintf(buffer, "%llu %llu %lu\n",
826e08b0
IM		 340 (unsigned long long)task->se.sum_exec_runtime,
341 (unsigned long long)task->sched_info.run_delay,
2d72376b 342 task->sched_info.pcount);
1da177e4
LT		 343}
344#endif
345
9745512c
AV		 346#ifdef CONFIG_LATENCYTOP
347static int lstats_show_proc(struct seq_file *m, void *v)
348{
349 int i;
13d77c37
HS		 350 struct inode *inode = m->private;
351 struct task_struct *task = get_proc_task(inode);
9745512c 352
13d77c37
HS		 353 if (!task)
354 return -ESRCH;
355 seq_puts(m, "Latency Top version : v0.1\n");
9745512c 356 for (i = 0; i < 32; i++) {
34e49d4f
JP		 357 struct latency_record *lr = &task->latency_record[i];
358 if (lr->backtrace[0]) {
9745512c 359 int q;
34e49d4f
JP		 360 seq_printf(m, "%i %li %li",
361 lr->count, lr->time, lr->max);
9745512c 362 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
34e49d4f
JP		 363 unsigned long bt = lr->backtrace[q];
364 if (!bt)
9745512c 365 break;
34e49d4f 366 if (bt == ULONG_MAX)
9745512c 367 break;
34e49d4f 368 seq_printf(m, " %ps", (void *)bt);
9745512c 369 }
9d6de12f 370 seq_putc(m, '\n');
9745512c
AV		 371 }
372
373 }
13d77c37 374 put_task_struct(task);
9745512c
AV		 375 return 0;
376}
377
378static int lstats_open(struct inode *inode, struct file *file)
379{
13d77c37 380 return single_open(file, lstats_show_proc, inode);
d6643d12
HS		 381}
382
9745512c
AV		 383static ssize_t lstats_write(struct file *file, const char __user *buf,
384 size_t count, loff_t *offs)
385{
13d77c37 386 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
9745512c 387
13d77c37
HS		 388 if (!task)
389 return -ESRCH;
9745512c 390 clear_all_latency_tracing(task);
13d77c37 391 put_task_struct(task);
9745512c
AV		 392
393 return count;
394}
395
396static const struct file_operations proc_lstats_operations = {
397 .open = lstats_open,
398 .read = seq_read,
399 .write = lstats_write,
400 .llseek = seq_lseek,
13d77c37 401 .release = single_release,
9745512c
AV		 402};
403
404#endif
405
1da177e4
LT		 406static int proc_oom_score(struct task_struct *task, char *buffer)
407{
a7f638f9 408 unsigned long totalpages = totalram_pages + total_swap_pages;
b95c35e7 409 unsigned long points = 0;
1da177e4 410
19c5d45a 411 read_lock(&tasklist_lock);
b95c35e7 412 if (pid_alive(task))
a7f638f9
DR		 413 points = oom_badness(task, NULL, NULL, totalpages) *
414 1000 / totalpages;
19c5d45a 415 read_unlock(&tasklist_lock);
1da177e4
LT		 416 return sprintf(buffer, "%lu\n", points);
417}
418
d85f50d5
NH		 419struct limit_names {
420 char *name;
421 char *unit;
422};
423
424static const struct limit_names lnames[RLIM_NLIMITS] = {
cff4edb5 425 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
d85f50d5
NH		 426 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
427 [RLIMIT_DATA] = {"Max data size", "bytes"},
428 [RLIMIT_STACK] = {"Max stack size", "bytes"},
429 [RLIMIT_CORE] = {"Max core file size", "bytes"},
430 [RLIMIT_RSS] = {"Max resident set", "bytes"},
431 [RLIMIT_NPROC] = {"Max processes", "processes"},
432 [RLIMIT_NOFILE] = {"Max open files", "files"},
433 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
434 [RLIMIT_AS] = {"Max address space", "bytes"},
435 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
436 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
437 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
438 [RLIMIT_NICE] = {"Max nice priority", NULL},
439 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
8808117c 440 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
d85f50d5
NH		 441};
442
443/* Display limits for a process */
444static int proc_pid_limits(struct task_struct *task, char *buffer)
445{
446 unsigned int i;
447 int count = 0;
448 unsigned long flags;
449 char *bufptr = buffer;
450
451 struct rlimit rlim[RLIM_NLIMITS];
452
a6bebbc8 453 if (!lock_task_sighand(task, &flags))
d85f50d5 454 return 0;
d85f50d5
NH		 455 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
456 unlock_task_sighand(task, &flags);
d85f50d5
NH		 457
458 /*
459 * print the file header
460 */
461 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
462 "Limit", "Soft Limit", "Hard Limit", "Units");
463
464 for (i = 0; i < RLIM_NLIMITS; i++) {
465 if (rlim[i].rlim_cur == RLIM_INFINITY)
466 count += sprintf(&bufptr[count], "%-25s %-20s ",
467 lnames[i].name, "unlimited");
468 else
469 count += sprintf(&bufptr[count], "%-25s %-20lu ",
470 lnames[i].name, rlim[i].rlim_cur);
471
472 if (rlim[i].rlim_max == RLIM_INFINITY)
473 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
474 else
475 count += sprintf(&bufptr[count], "%-20lu ",
476 rlim[i].rlim_max);
477
478 if (lnames[i].unit)
479 count += sprintf(&bufptr[count], "%-10s\n",
480 lnames[i].unit);
481 else
482 count += sprintf(&bufptr[count], "\n");
483 }
484
485 return count;
486}
487
ebcb6734
RM		 488#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
489static int proc_pid_syscall(struct task_struct *task, char *buffer)
490{
491 long nr;
492 unsigned long args[6], sp, pc;
a9712bc1
AV		 493 int res = lock_trace(task);
494 if (res)
495 return res;
ebcb6734
RM		 496
497 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
a9712bc1
AV		 498 res = sprintf(buffer, "running\n");
499 else if (nr < 0)
500 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
501 else
502 res = sprintf(buffer,
ebcb6734
RM		 503 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
504 nr,
505 args[0], args[1], args[2], args[3], args[4], args[5],
506 sp, pc);
a9712bc1
AV		 507 unlock_trace(task);
508 return res;
ebcb6734
RM		 509}
510#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
511
1da177e4
LT		 512/************************************************************************/
513/* Here the fs part begins */
514/************************************************************************/
515
516/* permission checks */
778c1144 517static int proc_fd_access_allowed(struct inode *inode)
1da177e4 518{
778c1144
EB		 519 struct task_struct *task;
520 int allowed = 0;
df26c40e
EB		 521 /* Allow access to a task's file descriptors if it is us or we
522 * may use ptrace attach to the process and find out that
523 * information.
778c1144
EB		 524 */
525 task = get_proc_task(inode);
df26c40e 526 if (task) {
006ebb40 527 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
778c1144 528 put_task_struct(task);
df26c40e 529 }
778c1144 530 return allowed;
1da177e4
LT		 531}
532
6b4e306a 533int proc_setattr(struct dentry *dentry, struct iattr *attr)
6d76fa58
LT		 534{
535 int error;
536 struct inode *inode = dentry->d_inode;
537
538 if (attr->ia_valid & ATTR_MODE)
539 return -EPERM;
540
541 error = inode_change_ok(inode, attr);
1025774c
CH		 542 if (error)
543 return error;
544
545 if ((attr->ia_valid & ATTR_SIZE) &&
546 attr->ia_size != i_size_read(inode)) {
547 error = vmtruncate(inode, attr->ia_size);
548 if (error)
549 return error;
550 }
551
552 setattr_copy(inode, attr);
553 mark_inode_dirty(inode);
554 return 0;
6d76fa58
LT		 555}
556
0499680a
VK		 557/*
558 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
559 * or euid/egid (for hide_pid_min=2)?
560 */
561static bool has_pid_permissions(struct pid_namespace *pid,
562 struct task_struct *task,
563 int hide_pid_min)
564{
565 if (pid->hide_pid < hide_pid_min)
566 return true;
567 if (in_group_p(pid->pid_gid))
568 return true;
569 return ptrace_may_access(task, PTRACE_MODE_READ);
570}
571
572
573static int proc_pid_permission(struct inode *inode, int mask)
574{
575 struct pid_namespace *pid = inode->i_sb->s_fs_info;
576 struct task_struct *task;
577 bool has_perms;
578
579 task = get_proc_task(inode);
a2ef990a
XF		 580 if (!task)
581 return -ESRCH;
0499680a
VK		 582 has_perms = has_pid_permissions(pid, task, 1);
583 put_task_struct(task);
584
585 if (!has_perms) {
586 if (pid->hide_pid == 2) {
587 /*
588 * Let's make getdents(), stat(), and open()
589 * consistent with each other. If a process
590 * may not stat() a file, it shouldn't be seen
591 * in procfs at all.
592 */
593 return -ENOENT;
594 }
595
596 return -EPERM;
597 }
598 return generic_permission(inode, mask);
599}
600
601
602
c5ef1c42 603static const struct inode_operations proc_def_inode_operations = {
6d76fa58
LT		 604 .setattr = proc_setattr,
605};
606
1da177e4
LT		 607#define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
608
609static ssize_t proc_info_read(struct file * file, char __user * buf,
610 size_t count, loff_t *ppos)
611{
2fddfeef 612 struct inode * inode = file->f_path.dentry->d_inode;
1da177e4
LT		 613 unsigned long page;
614 ssize_t length;
99f89551
EB		 615 struct task_struct *task = get_proc_task(inode);
616
617 length = -ESRCH;
618 if (!task)
619 goto out_no_task;
1da177e4
LT		 620
621 if (count > PROC_BLOCK_SIZE)
622 count = PROC_BLOCK_SIZE;
99f89551
EB		 623
624 length = -ENOMEM;
e12ba74d 625 if (!(page = __get_free_page(GFP_TEMPORARY)))
99f89551 626 goto out;
1da177e4
LT		 627
628 length = PROC_I(inode)->op.proc_read(task, (char*)page);
629
630 if (length >= 0)
631 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
632 free_page(page);
99f89551
EB		 633out:
634 put_task_struct(task);
635out_no_task:
1da177e4
LT		 636 return length;
637}
638
00977a59 639static const struct file_operations proc_info_file_operations = {
1da177e4 640 .read = proc_info_read,
87df8424 641 .llseek = generic_file_llseek,
1da177e4
LT		 642};
643
be614086
EB		 644static int proc_single_show(struct seq_file *m, void *v)
645{
646 struct inode *inode = m->private;
647 struct pid_namespace *ns;
648 struct pid *pid;
649 struct task_struct *task;
650 int ret;
651
652 ns = inode->i_sb->s_fs_info;
653 pid = proc_pid(inode);
654 task = get_pid_task(pid, PIDTYPE_PID);
655 if (!task)
656 return -ESRCH;
657
658 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
659
660 put_task_struct(task);
661 return ret;
662}
663
664static int proc_single_open(struct inode *inode, struct file *filp)
665{
c6a34058 666 return single_open(filp, proc_single_show, inode);
be614086
EB		 667}
668
669static const struct file_operations proc_single_file_operations = {
670 .open = proc_single_open,
671 .read = seq_read,
672 .llseek = seq_lseek,
673 .release = single_release,
674};
675
b409e578 676static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
1da177e4 677{
e268337d
LT		 678 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
679 struct mm_struct *mm;
680
681 if (!task)
682 return -ESRCH;
683
b409e578 684 mm = mm_access(task, mode);
e268337d
LT		 685 put_task_struct(task);
686
687 if (IS_ERR(mm))
688 return PTR_ERR(mm);
689
6d08f2c7
ON		 690 if (mm) {
691 /* ensure this mm_struct can't be freed */
692 atomic_inc(&mm->mm_count);
693 /* but do not pin its memory */
694 mmput(mm);
695 }
696
e268337d
LT		 697 file->private_data = mm;
698
1da177e4
LT		 699 return 0;
700}
701
b409e578
CW		 702static int mem_open(struct inode *inode, struct file *file)
703{
bc452b4b
DH		 704 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
705
706 /* OK to pass negative loff_t, we can catch out-of-range */
707 file->f_mode |= FMODE_UNSIGNED_OFFSET;
708
709 return ret;
b409e578
CW		 710}
711
572d34b9
ON		 712static ssize_t mem_rw(struct file *file, char __user *buf,
713 size_t count, loff_t *ppos, int write)
1da177e4 714{
e268337d 715 struct mm_struct *mm = file->private_data;
572d34b9
ON		 716 unsigned long addr = *ppos;
717 ssize_t copied;
1da177e4 718 char *page;
1da177e4 719
e268337d
LT		 720 if (!mm)
721 return 0;
99f89551 722
30cd8903
KM		 723 page = (char *)__get_free_page(GFP_TEMPORARY);
724 if (!page)
e268337d 725 return -ENOMEM;
1da177e4 726
f7ca54f4 727 copied = 0;
6d08f2c7
ON		 728 if (!atomic_inc_not_zero(&mm->mm_users))
729 goto free;
730
1da177e4 731 while (count > 0) {
572d34b9 732 int this_len = min_t(int, count, PAGE_SIZE);
1da177e4 733
572d34b9 734 if (write && copy_from_user(page, buf, this_len)) {
1da177e4
LT		 735 copied = -EFAULT;
736 break;
737 }
572d34b9
ON		 738
739 this_len = access_remote_vm(mm, addr, page, this_len, write);
740 if (!this_len) {
1da177e4
LT		 741 if (!copied)
742 copied = -EIO;
743 break;
744 }
572d34b9
ON		 745
746 if (!write && copy_to_user(buf, page, this_len)) {
747 copied = -EFAULT;
748 break;
749 }
750
751 buf += this_len;
752 addr += this_len;
753 copied += this_len;
754 count -= this_len;
1da177e4 755 }
572d34b9 756 *ppos = addr;
30cd8903 757
6d08f2c7
ON		 758 mmput(mm);
759free:
30cd8903 760 free_page((unsigned long) page);
1da177e4
LT		 761 return copied;
762}
1da177e4 763
572d34b9
ON		 764static ssize_t mem_read(struct file *file, char __user *buf,
765 size_t count, loff_t *ppos)
766{
767 return mem_rw(file, buf, count, ppos, 0);
768}
769
770static ssize_t mem_write(struct file *file, const char __user *buf,
771 size_t count, loff_t *ppos)
772{
773 return mem_rw(file, (char __user*)buf, count, ppos, 1);
774}
775
85863e47 776loff_t mem_lseek(struct file *file, loff_t offset, int orig)
1da177e4
LT		 777{
778 switch (orig) {
779 case 0:
780 file->f_pos = offset;
781 break;
782 case 1:
783 file->f_pos += offset;
784 break;
785 default:
786 return -EINVAL;
787 }
788 force_successful_syscall_return();
789 return file->f_pos;
790}
791
e268337d
LT		 792static int mem_release(struct inode *inode, struct file *file)
793{
794 struct mm_struct *mm = file->private_data;
71879d3c 795 if (mm)
6d08f2c7 796 mmdrop(mm);
e268337d
LT		 797 return 0;
798}
799
00977a59 800static const struct file_operations proc_mem_operations = {
1da177e4
LT		 801 .llseek = mem_lseek,
802 .read = mem_read,
803 .write = mem_write,
804 .open = mem_open,
e268337d 805 .release = mem_release,
1da177e4
LT		 806};
807
b409e578
CW		 808static int environ_open(struct inode *inode, struct file *file)
809{
810 return __mem_open(inode, file, PTRACE_MODE_READ);
811}
812
315e28c8
JP		 813static ssize_t environ_read(struct file *file, char __user *buf,
814 size_t count, loff_t *ppos)
815{
315e28c8
JP		 816 char *page;
817 unsigned long src = *ppos;
b409e578
CW		 818 int ret = 0;
819 struct mm_struct *mm = file->private_data;
315e28c8 820
b409e578
CW		 821 if (!mm)
822 return 0;
315e28c8 823
315e28c8
JP		 824 page = (char *)__get_free_page(GFP_TEMPORARY);
825 if (!page)
b409e578 826 return -ENOMEM;
315e28c8 827
d6f64b89 828 ret = 0;
b409e578
CW		 829 if (!atomic_inc_not_zero(&mm->mm_users))
830 goto free;
315e28c8 831 while (count > 0) {
e8905ec2
DH		 832 size_t this_len, max_len;
833 int retval;
315e28c8 834
e8905ec2 835 if (src >= (mm->env_end - mm->env_start))
315e28c8
JP		 836 break;
837
e8905ec2
DH		 838 this_len = mm->env_end - (mm->env_start + src);
839
840 max_len = min_t(size_t, PAGE_SIZE, count);
841 this_len = min(max_len, this_len);
315e28c8 842
b409e578 843 retval = access_remote_vm(mm, (mm->env_start + src),
315e28c8
JP		 844 page, this_len, 0);
845
846 if (retval <= 0) {
847 ret = retval;
848 break;
849 }
850
851 if (copy_to_user(buf, page, retval)) {
852 ret = -EFAULT;
853 break;
854 }
855
856 ret += retval;
857 src += retval;
858 buf += retval;
859 count -= retval;
860 }
861 *ppos = src;
315e28c8 862 mmput(mm);
b409e578
CW		 863
864free:
315e28c8 865 free_page((unsigned long) page);
315e28c8
JP		 866 return ret;
867}
868
869static const struct file_operations proc_environ_operations = {
b409e578 870 .open = environ_open,
315e28c8 871 .read = environ_read,
87df8424 872 .llseek = generic_file_llseek,
b409e578 873 .release = mem_release,
315e28c8
JP		 874};
875
1da177e4
LT		 876static ssize_t oom_adjust_read(struct file *file, char __user *buf,
877 size_t count, loff_t *ppos)
878{
2fddfeef 879 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
8578cea7 880 char buffer[PROC_NUMBUF];
1da177e4 881 size_t len;
28b83c51
KM		 882 int oom_adjust = OOM_DISABLE;
883 unsigned long flags;
1da177e4 884
99f89551
EB		 885 if (!task)
886 return -ESRCH;
28b83c51
KM		 887
888 if (lock_task_sighand(task, &flags)) {
889 oom_adjust = task->signal->oom_adj;
890 unlock_task_sighand(task, &flags);
891 }
892
99f89551
EB		 893 put_task_struct(task);
894
8578cea7 895 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
0c28f287
AM		 896
897 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1da177e4
LT		 898}
899
900static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
901 size_t count, loff_t *ppos)
902{
99f89551 903 struct task_struct *task;
5d863b89 904 char buffer[PROC_NUMBUF];
0a8cb8e3 905 int oom_adjust;
28b83c51 906 unsigned long flags;
5d863b89 907 int err;
1da177e4 908
8578cea7
EB		 909 memset(buffer, 0, sizeof(buffer));
910 if (count > sizeof(buffer) - 1)
911 count = sizeof(buffer) - 1;
723548bf
DR		 912 if (copy_from_user(buffer, buf, count)) {
913 err = -EFAULT;
914 goto out;
915 }
5d863b89 916
0a8cb8e3 917 err = kstrtoint(strstrip(buffer), 0, &oom_adjust);
5d863b89 918 if (err)
723548bf 919 goto out;
8ac773b4 920 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
723548bf
DR		 921 oom_adjust != OOM_DISABLE) {
922 err = -EINVAL;
923 goto out;
924 }
5d863b89 925
2fddfeef 926 task = get_proc_task(file->f_path.dentry->d_inode);
723548bf
DR		 927 if (!task) {
928 err = -ESRCH;
929 goto out;
930 }
d19d5476
DR		 931
932 task_lock(task);
933 if (!task->mm) {
934 err = -EINVAL;
935 goto err_task_lock;
936 }
937
28b83c51 938 if (!lock_task_sighand(task, &flags)) {
723548bf 939 err = -ESRCH;
d19d5476 940 goto err_task_lock;
28b83c51
KM		 941 }
942
943 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
723548bf
DR		 944 err = -EACCES;
945 goto err_sighand;
8fb4fc68 946 }
28b83c51 947
51b1bd2a
DR		 948 /*
949 * Warn that /proc/pid/oom_adj is deprecated, see
950 * Documentation/feature-removal-schedule.txt.
951 */
c2142704 952 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
be8f684d
DR		 953 current->comm, task_pid_nr(current), task_pid_nr(task),
954 task_pid_nr(task));
28b83c51 955 task->signal->oom_adj = oom_adjust;
a63d83f4
DR		 956 /*
957 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
958 * value is always attainable.
959 */
960 if (task->signal->oom_adj == OOM_ADJUST_MAX)
961 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
962 else
963 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
964 -OOM_DISABLE;
43d2b113 965 trace_oom_score_adj_update(task);
723548bf 966err_sighand:
28b83c51 967 unlock_task_sighand(task, &flags);
d19d5476
DR		 968err_task_lock:
969 task_unlock(task);
99f89551 970 put_task_struct(task);
723548bf
DR		 971out:
972 return err < 0 ? err : count;
1da177e4
LT		 973}
974
00977a59 975static const struct file_operations proc_oom_adjust_operations = {
1da177e4
LT		 976 .read = oom_adjust_read,
977 .write = oom_adjust_write,
87df8424 978 .llseek = generic_file_llseek,
1da177e4
LT		 979};
980
a63d83f4
DR		 981static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
982 size_t count, loff_t *ppos)
983{
984 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
985 char buffer[PROC_NUMBUF];
986 int oom_score_adj = OOM_SCORE_ADJ_MIN;
987 unsigned long flags;
988 size_t len;
989
990 if (!task)
991 return -ESRCH;
992 if (lock_task_sighand(task, &flags)) {
993 oom_score_adj = task->signal->oom_score_adj;
994 unlock_task_sighand(task, &flags);
995 }
996 put_task_struct(task);
997 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
998 return simple_read_from_buffer(buf, count, ppos, buffer, len);
999}
1000
1001static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1002 size_t count, loff_t *ppos)
1003{
1004 struct task_struct *task;
1005 char buffer[PROC_NUMBUF];
1006 unsigned long flags;
0a8cb8e3 1007 int oom_score_adj;
a63d83f4
DR		 1008 int err;
1009
1010 memset(buffer, 0, sizeof(buffer));
1011 if (count > sizeof(buffer) - 1)
1012 count = sizeof(buffer) - 1;
723548bf
DR		 1013 if (copy_from_user(buffer, buf, count)) {
1014 err = -EFAULT;
1015 goto out;
1016 }
a63d83f4 1017
0a8cb8e3 1018 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
a63d83f4 1019 if (err)
723548bf 1020 goto out;
a63d83f4 1021 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
723548bf
DR		 1022 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1023 err = -EINVAL;
1024 goto out;
1025 }
a63d83f4
DR		 1026
1027 task = get_proc_task(file->f_path.dentry->d_inode);
723548bf
DR		 1028 if (!task) {
1029 err = -ESRCH;
1030 goto out;
1031 }
d19d5476
DR		 1032
1033 task_lock(task);
1034 if (!task->mm) {
1035 err = -EINVAL;
1036 goto err_task_lock;
1037 }
1038
a63d83f4 1039 if (!lock_task_sighand(task, &flags)) {
723548bf 1040 err = -ESRCH;
d19d5476 1041 goto err_task_lock;
a63d83f4 1042 }
d19d5476 1043
dabb16f6 1044 if (oom_score_adj < task->signal->oom_score_adj_min &&
a63d83f4 1045 !capable(CAP_SYS_RESOURCE)) {
723548bf
DR		 1046 err = -EACCES;
1047 goto err_sighand;
a63d83f4
DR		 1048 }
1049
1050 task->signal->oom_score_adj = oom_score_adj;
dabb16f6
MSB		 1051 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1052 task->signal->oom_score_adj_min = oom_score_adj;
43d2b113 1053 trace_oom_score_adj_update(task);
a63d83f4
DR		 1054 /*
1055 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1056 * always attainable.
1057 */
1058 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1059 task->signal->oom_adj = OOM_DISABLE;
1060 else
1061 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1062 OOM_SCORE_ADJ_MAX;
723548bf 1063err_sighand:
a63d83f4 1064 unlock_task_sighand(task, &flags);
d19d5476
DR		 1065err_task_lock:
1066 task_unlock(task);
a63d83f4 1067 put_task_struct(task);
723548bf
DR		 1068out:
1069 return err < 0 ? err : count;
a63d83f4
DR		 1070}
1071
1072static const struct file_operations proc_oom_score_adj_operations = {
1073 .read = oom_score_adj_read,
1074 .write = oom_score_adj_write,
6038f373 1075 .llseek = default_llseek,
a63d83f4
DR		 1076};
1077
1da177e4
LT		 1078#ifdef CONFIG_AUDITSYSCALL
1079#define TMPBUFLEN 21
1080static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1081 size_t count, loff_t *ppos)
1082{
2fddfeef 1083 struct inode * inode = file->f_path.dentry->d_inode;
99f89551 1084 struct task_struct *task = get_proc_task(inode);
1da177e4
LT		 1085 ssize_t length;
1086 char tmpbuf[TMPBUFLEN];
1087
99f89551
EB		 1088 if (!task)
1089 return -ESRCH;
1da177e4 1090 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
0c11b942 1091 audit_get_loginuid(task));
99f89551 1092 put_task_struct(task);
1da177e4
LT		 1093 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1094}
1095
1096static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1097 size_t count, loff_t *ppos)
1098{
2fddfeef 1099 struct inode * inode = file->f_path.dentry->d_inode;
1da177e4
LT		 1100 char *page, *tmp;
1101 ssize_t length;
1da177e4
LT		 1102 uid_t loginuid;
1103
7dc52157
PM		 1104 rcu_read_lock();
1105 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1106 rcu_read_unlock();
1da177e4 1107 return -EPERM;
7dc52157
PM		 1108 }
1109 rcu_read_unlock();
1da177e4 1110
e0182909
AV		 1111 if (count >= PAGE_SIZE)
1112 count = PAGE_SIZE - 1;
1da177e4
LT		 1113
1114 if (*ppos != 0) {
1115 /* No partial writes. */
1116 return -EINVAL;
1117 }
e12ba74d 1118 page = (char*)__get_free_page(GFP_TEMPORARY);
1da177e4
LT		 1119 if (!page)
1120 return -ENOMEM;
1121 length = -EFAULT;
1122 if (copy_from_user(page, buf, count))
1123 goto out_free_page;
1124
e0182909 1125 page[count] = '\0';
1da177e4
LT		 1126 loginuid = simple_strtoul(page, &tmp, 10);
1127 if (tmp == page) {
1128 length = -EINVAL;
1129 goto out_free_page;
1130
1131 }
0a300be6 1132 length = audit_set_loginuid(loginuid);
1da177e4
LT		 1133 if (likely(length == 0))
1134 length = count;
1135
1136out_free_page:
1137 free_page((unsigned long) page);
1138 return length;
1139}
1140
00977a59 1141static const struct file_operations proc_loginuid_operations = {
1da177e4
LT		 1142 .read = proc_loginuid_read,
1143 .write = proc_loginuid_write,
87df8424 1144 .llseek = generic_file_llseek,
1da177e4 1145};
1e0bd755
EP		 1146
1147static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1148 size_t count, loff_t *ppos)
1149{
1150 struct inode * inode = file->f_path.dentry->d_inode;
1151 struct task_struct *task = get_proc_task(inode);
1152 ssize_t length;
1153 char tmpbuf[TMPBUFLEN];
1154
1155 if (!task)
1156 return -ESRCH;
1157 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1158 audit_get_sessionid(task));
1159 put_task_struct(task);
1160 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1161}
1162
1163static const struct file_operations proc_sessionid_operations = {
1164 .read = proc_sessionid_read,
87df8424 1165 .llseek = generic_file_llseek,
1e0bd755 1166};
1da177e4
LT		 1167#endif
1168
f4f154fd
AM		 1169#ifdef CONFIG_FAULT_INJECTION
1170static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1171 size_t count, loff_t *ppos)
1172{
1173 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1174 char buffer[PROC_NUMBUF];
1175 size_t len;
1176 int make_it_fail;
f4f154fd
AM		 1177
1178 if (!task)
1179 return -ESRCH;
1180 make_it_fail = task->make_it_fail;
1181 put_task_struct(task);
1182
1183 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
0c28f287
AM		 1184
1185 return simple_read_from_buffer(buf, count, ppos, buffer, len);
f4f154fd
AM		 1186}
1187
1188static ssize_t proc_fault_inject_write(struct file * file,
1189 const char __user * buf, size_t count, loff_t *ppos)
1190{
1191 struct task_struct *task;
1192 char buffer[PROC_NUMBUF], *end;
1193 int make_it_fail;
1194
1195 if (!capable(CAP_SYS_RESOURCE))
1196 return -EPERM;
1197 memset(buffer, 0, sizeof(buffer));
1198 if (count > sizeof(buffer) - 1)
1199 count = sizeof(buffer) - 1;
1200 if (copy_from_user(buffer, buf, count))
1201 return -EFAULT;
cba8aafe
VL		 1202 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1203 if (*end)
1204 return -EINVAL;
f4f154fd
AM		 1205 task = get_proc_task(file->f_dentry->d_inode);
1206 if (!task)
1207 return -ESRCH;
1208 task->make_it_fail = make_it_fail;
1209 put_task_struct(task);
cba8aafe
VL		 1210
1211 return count;
f4f154fd
AM		 1212}
1213
00977a59 1214static const struct file_operations proc_fault_inject_operations = {
f4f154fd
AM		 1215 .read = proc_fault_inject_read,
1216 .write = proc_fault_inject_write,
87df8424 1217 .llseek = generic_file_llseek,
f4f154fd
AM		 1218};
1219#endif
1220
9745512c 1221
43ae34cb
IM		 1222#ifdef CONFIG_SCHED_DEBUG
1223/*
1224 * Print out various scheduling related per-task fields:
1225 */
1226static int sched_show(struct seq_file *m, void *v)
1227{
1228 struct inode *inode = m->private;
1229 struct task_struct *p;
1230
43ae34cb
IM		 1231 p = get_proc_task(inode);
1232 if (!p)
1233 return -ESRCH;
1234 proc_sched_show_task(p, m);
1235
1236 put_task_struct(p);
1237
1238 return 0;
1239}
1240
1241static ssize_t
1242sched_write(struct file *file, const char __user *buf,
1243 size_t count, loff_t *offset)
1244{
1245 struct inode *inode = file->f_path.dentry->d_inode;
1246 struct task_struct *p;
1247
43ae34cb
IM		 1248 p = get_proc_task(inode);
1249 if (!p)
1250 return -ESRCH;
1251 proc_sched_set_task(p);
1252
1253 put_task_struct(p);
1254
1255 return count;
1256}
1257
1258static int sched_open(struct inode *inode, struct file *filp)
1259{
c6a34058 1260 return single_open(filp, sched_show, inode);
43ae34cb
IM		 1261}
1262
1263static const struct file_operations proc_pid_sched_operations = {
1264 .open = sched_open,
1265 .read = seq_read,
1266 .write = sched_write,
1267 .llseek = seq_lseek,
5ea473a1 1268 .release = single_release,
43ae34cb
IM		 1269};
1270
1271#endif
1272
5091faa4
MG		 1273#ifdef CONFIG_SCHED_AUTOGROUP
1274/*
1275 * Print out autogroup related information:
1276 */
1277static int sched_autogroup_show(struct seq_file *m, void *v)
1278{
1279 struct inode *inode = m->private;
1280 struct task_struct *p;
1281
1282 p = get_proc_task(inode);
1283 if (!p)
1284 return -ESRCH;
1285 proc_sched_autogroup_show_task(p, m);
1286
1287 put_task_struct(p);
1288
1289 return 0;
1290}
1291
1292static ssize_t
1293sched_autogroup_write(struct file *file, const char __user *buf,
1294 size_t count, loff_t *offset)
1295{
1296 struct inode *inode = file->f_path.dentry->d_inode;
1297 struct task_struct *p;
1298 char buffer[PROC_NUMBUF];
0a8cb8e3 1299 int nice;
5091faa4
MG		 1300 int err;
1301
1302 memset(buffer, 0, sizeof(buffer));
1303 if (count > sizeof(buffer) - 1)
1304 count = sizeof(buffer) - 1;
1305 if (copy_from_user(buffer, buf, count))
1306 return -EFAULT;
1307
0a8cb8e3
AD		 1308 err = kstrtoint(strstrip(buffer), 0, &nice);
1309 if (err < 0)
1310 return err;
5091faa4
MG		 1311
1312 p = get_proc_task(inode);
1313 if (!p)
1314 return -ESRCH;
1315
2e5b5b3a 1316 err = proc_sched_autogroup_set_nice(p, nice);
5091faa4
MG		 1317 if (err)
1318 count = err;
1319
1320 put_task_struct(p);
1321
1322 return count;
1323}
1324
1325static int sched_autogroup_open(struct inode *inode, struct file *filp)
1326{
1327 int ret;
1328
1329 ret = single_open(filp, sched_autogroup_show, NULL);
1330 if (!ret) {
1331 struct seq_file *m = filp->private_data;
1332
1333 m->private = inode;
1334 }
1335 return ret;
1336}
1337
1338static const struct file_operations proc_pid_sched_autogroup_operations = {
1339 .open = sched_autogroup_open,
1340 .read = seq_read,
1341 .write = sched_autogroup_write,
1342 .llseek = seq_lseek,
1343 .release = single_release,
1344};
1345
1346#endif /* CONFIG_SCHED_AUTOGROUP */
1347
4614a696 1348static ssize_t comm_write(struct file *file, const char __user *buf,
1349 size_t count, loff_t *offset)
1350{
1351 struct inode *inode = file->f_path.dentry->d_inode;
1352 struct task_struct *p;
1353 char buffer[TASK_COMM_LEN];
1354
1355 memset(buffer, 0, sizeof(buffer));
1356 if (count > sizeof(buffer) - 1)
1357 count = sizeof(buffer) - 1;
1358 if (copy_from_user(buffer, buf, count))
1359 return -EFAULT;
1360
1361 p = get_proc_task(inode);
1362 if (!p)
1363 return -ESRCH;
1364
1365 if (same_thread_group(current, p))
1366 set_task_comm(p, buffer);
1367 else
1368 count = -EINVAL;
1369
1370 put_task_struct(p);
1371
1372 return count;
1373}
1374
1375static int comm_show(struct seq_file *m, void *v)
1376{
1377 struct inode *inode = m->private;
1378 struct task_struct *p;
1379
1380 p = get_proc_task(inode);
1381 if (!p)
1382 return -ESRCH;
1383
1384 task_lock(p);
1385 seq_printf(m, "%s\n", p->comm);
1386 task_unlock(p);
1387
1388 put_task_struct(p);
1389
1390 return 0;
1391}
1392
1393static int comm_open(struct inode *inode, struct file *filp)
1394{
c6a34058 1395 return single_open(filp, comm_show, inode);
4614a696 1396}
1397
1398static const struct file_operations proc_pid_set_comm_operations = {
1399 .open = comm_open,
1400 .read = seq_read,
1401 .write = comm_write,
1402 .llseek = seq_lseek,
1403 .release = single_release,
1404};
1405
7773fbc5 1406static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
925d1c40
MH		 1407{
1408 struct task_struct *task;
1409 struct mm_struct *mm;
1410 struct file *exe_file;
1411
7773fbc5 1412 task = get_proc_task(dentry->d_inode);
925d1c40
MH		 1413 if (!task)
1414 return -ENOENT;
1415 mm = get_task_mm(task);
1416 put_task_struct(task);
1417 if (!mm)
1418 return -ENOENT;
1419 exe_file = get_mm_exe_file(mm);
1420 mmput(mm);
1421 if (exe_file) {
1422 *exe_path = exe_file->f_path;
1423 path_get(&exe_file->f_path);
1424 fput(exe_file);
1425 return 0;
1426 } else
1427 return -ENOENT;
1428}
1429
008b150a 1430static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1da177e4
LT		 1431{
1432 struct inode *inode = dentry->d_inode;
408ef013 1433 struct path path;
1da177e4
LT		 1434 int error = -EACCES;
1435
778c1144
EB		 1436 /* Are we allowed to snoop on the tasks file descriptors? */
1437 if (!proc_fd_access_allowed(inode))
1da177e4 1438 goto out;
1da177e4 1439
408ef013
CH		 1440 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1441 if (error)
1442 goto out;
1443
b5fb63c1 1444 nd_jump_link(nd, &path);
408ef013 1445 return NULL;
1da177e4 1446out:
008b150a 1447 return ERR_PTR(error);
1da177e4
LT		 1448}
1449
3dcd25f3 1450static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1da177e4 1451{
e12ba74d 1452 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
3dcd25f3 1453 char *pathname;
1da177e4
LT		 1454 int len;
1455
1456 if (!tmp)
1457 return -ENOMEM;
0c28f287 1458
7b2a69ba 1459 pathname = d_path(path, tmp, PAGE_SIZE);
3dcd25f3
JB		 1460 len = PTR_ERR(pathname);
1461 if (IS_ERR(pathname))
1da177e4 1462 goto out;
3dcd25f3 1463 len = tmp + PAGE_SIZE - 1 - pathname;
1da177e4
LT		 1464
1465 if (len > buflen)
1466 len = buflen;
3dcd25f3 1467 if (copy_to_user(buffer, pathname, len))
1da177e4
LT		 1468 len = -EFAULT;
1469 out:
1470 free_page((unsigned long)tmp);
1471 return len;
1472}
1473
1474static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1475{
1476 int error = -EACCES;
1477 struct inode *inode = dentry->d_inode;
3dcd25f3 1478 struct path path;
1da177e4 1479
778c1144
EB		 1480 /* Are we allowed to snoop on the tasks file descriptors? */
1481 if (!proc_fd_access_allowed(inode))
1da177e4 1482 goto out;
1da177e4 1483
7773fbc5 1484 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1da177e4
LT		 1485 if (error)
1486 goto out;
1487
3dcd25f3
JB		 1488 error = do_proc_readlink(&path, buffer, buflen);
1489 path_put(&path);
1da177e4 1490out:
1da177e4
LT		 1491 return error;
1492}
1493
faf60af1 1494const struct inode_operations proc_pid_link_inode_operations = {
1da177e4 1495 .readlink = proc_pid_readlink,
6d76fa58
LT		 1496 .follow_link = proc_pid_follow_link,
1497 .setattr = proc_setattr,
1da177e4
LT		 1498};
1499
28a6d671
EB		 1500
1501/* building an inode */
1502
6b4e306a 1503struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
28a6d671
EB		 1504{
1505 struct inode * inode;
1506 struct proc_inode *ei;
c69e8d9c 1507 const struct cred *cred;
1da177e4 1508
28a6d671 1509 /* We need a new inode */
1da177e4 1510
28a6d671
EB		 1511 inode = new_inode(sb);
1512 if (!inode)
1513 goto out;
1514
1515 /* Common stuff */
1516 ei = PROC_I(inode);
85fe4025 1517 inode->i_ino = get_next_ino();
28a6d671 1518 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
28a6d671
EB		 1519 inode->i_op = &proc_def_inode_operations;
1520
1521 /*
1522 * grab the reference to task.
1523 */
1a657f78 1524 ei->pid = get_task_pid(task, PIDTYPE_PID);
28a6d671
EB		 1525 if (!ei->pid)
1526 goto out_unlock;
1527
28a6d671 1528 if (task_dumpable(task)) {
c69e8d9c
DH		 1529 rcu_read_lock();
1530 cred = __task_cred(task);
1531 inode->i_uid = cred->euid;
1532 inode->i_gid = cred->egid;
1533 rcu_read_unlock();
1da177e4 1534 }
28a6d671
EB		 1535 security_task_to_inode(task, inode);
1536
1da177e4 1537out:
28a6d671
EB		 1538 return inode;
1539
1540out_unlock:
1541 iput(inode);
1542 return NULL;
1da177e4
LT		 1543}
1544
6b4e306a 1545int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1da177e4 1546{
1da177e4 1547 struct inode *inode = dentry->d_inode;
28a6d671 1548 struct task_struct *task;
c69e8d9c 1549 const struct cred *cred;
0499680a 1550 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
c69e8d9c 1551
28a6d671 1552 generic_fillattr(inode, stat);
1da177e4 1553
28a6d671 1554 rcu_read_lock();
dcb0f222
EB		 1555 stat->uid = GLOBAL_ROOT_UID;
1556 stat->gid = GLOBAL_ROOT_GID;
28a6d671
EB		 1557 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1558 if (task) {
0499680a
VK		 1559 if (!has_pid_permissions(pid, task, 2)) {
1560 rcu_read_unlock();
1561 /*
1562 * This doesn't prevent learning whether PID exists,
1563 * it only makes getattr() consistent with readdir().
1564 */
1565 return -ENOENT;
1566 }
28a6d671
EB		 1567 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1568 task_dumpable(task)) {
c69e8d9c
DH		 1569 cred = __task_cred(task);
1570 stat->uid = cred->euid;
1571 stat->gid = cred->egid;
1da177e4
LT		 1572 }
1573 }
28a6d671 1574 rcu_read_unlock();
d6e71144 1575 return 0;
1da177e4
LT		 1576}
1577
1da177e4
LT		 1578/* dentry stuff */
1579
1580/*
1581 * Exceptional case: normally we are not allowed to unhash a busy
1582 * directory. In this case, however, we can do it - no aliasing problems
1583 * due to the way we treat inodes.
1584 *
1585 * Rewrite the inode's ownerships here because the owning task may have
1586 * performed a setuid(), etc.
99f89551
EB		 1587 *
1588 * Before the /proc/pid/status file was created the only way to read
1589 * the effective uid of a /process was to stat /proc/pid. Reading
1590 * /proc/pid/status is slow enough that procps and other packages
1591 * kept stating /proc/pid. To keep the rules in /proc simple I have
1592 * made this apply to all per process world readable and executable
1593 * directories.
1da177e4 1594 */
0b728e19 1595int pid_revalidate(struct dentry *dentry, unsigned int flags)
1da177e4 1596{
34286d66
NP		 1597 struct inode *inode;
1598 struct task_struct *task;
c69e8d9c
DH		 1599 const struct cred *cred;
1600
0b728e19 1601 if (flags & LOOKUP_RCU)
34286d66
NP		 1602 return -ECHILD;
1603
1604 inode = dentry->d_inode;
1605 task = get_proc_task(inode);
1606
99f89551
EB		 1607 if (task) {
1608 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1609 task_dumpable(task)) {
c69e8d9c
DH		 1610 rcu_read_lock();
1611 cred = __task_cred(task);
1612 inode->i_uid = cred->euid;
1613 inode->i_gid = cred->egid;
1614 rcu_read_unlock();
1da177e4 1615 } else {
dcb0f222
EB		 1616 inode->i_uid = GLOBAL_ROOT_UID;
1617 inode->i_gid = GLOBAL_ROOT_GID;
1da177e4 1618 }
9ee8ab9f 1619 inode->i_mode &= ~(S_ISUID | S_ISGID);
1da177e4 1620 security_task_to_inode(task, inode);
99f89551 1621 put_task_struct(task);
1da177e4
LT		 1622 return 1;
1623 }
1624 d_drop(dentry);
1625 return 0;
1626}
1627
6b4e306a 1628const struct dentry_operations pid_dentry_operations =
28a6d671
EB		 1629{
1630 .d_revalidate = pid_revalidate,
1631 .d_delete = pid_delete_dentry,
1632};
1633
1634/* Lookups */
1635
1c0d04c9
EB		 1636/*
1637 * Fill a directory entry.
1638 *
1639 * If possible create the dcache entry and derive our inode number and
1640 * file type from dcache entry.
1641 *
1642 * Since all of the proc inode numbers are dynamically generated, the inode
1643 * numbers do not exist until the inode is cache. This means creating the
1644 * the dcache entry in readdir is necessary to keep the inode numbers
1645 * reported by readdir in sync with the inode numbers reported
1646 * by stat.
1647 */
6b4e306a
EB		 1648int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1649 const char *name, int len,
c5141e6d 1650 instantiate_t instantiate, struct task_struct *task, const void *ptr)
61a28784 1651{
2fddfeef 1652 struct dentry *child, *dir = filp->f_path.dentry;
61a28784
EB		 1653 struct inode *inode;
1654 struct qstr qname;
1655 ino_t ino = 0;
1656 unsigned type = DT_UNKNOWN;
1657
1658 qname.name = name;
1659 qname.len = len;
1660 qname.hash = full_name_hash(name, len);
1661
1662 child = d_lookup(dir, &qname);
1663 if (!child) {
1664 struct dentry *new;
1665 new = d_alloc(dir, &qname);
1666 if (new) {
1667 child = instantiate(dir->d_inode, new, task, ptr);
1668 if (child)
1669 dput(new);
1670 else
1671 child = new;
1672 }
1673 }
1674 if (!child || IS_ERR(child) || !child->d_inode)
1675 goto end_instantiate;
1676 inode = child->d_inode;
1677 if (inode) {
1678 ino = inode->i_ino;
1679 type = inode->i_mode >> 12;
1680 }
1681 dput(child);
1682end_instantiate:
1683 if (!ino)
1684 ino = find_inode_number(dir, &qname);
1685 if (!ino)
1686 ino = 1;
1687 return filldir(dirent, name, len, filp->f_pos, ino, type);
1688}
1689
640708a2
PE		 1690#ifdef CONFIG_CHECKPOINT_RESTORE
1691
1692/*
1693 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1694 * which represent vma start and end addresses.
1695 */
1696static int dname_to_vma_addr(struct dentry *dentry,
1697 unsigned long *start, unsigned long *end)
1698{
1699 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
1700 return -EINVAL;
1701
1702 return 0;
1703}
1704
0b728e19 1705static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
640708a2
PE		 1706{
1707 unsigned long vm_start, vm_end;
1708 bool exact_vma_exists = false;
1709 struct mm_struct *mm = NULL;
1710 struct task_struct *task;
1711 const struct cred *cred;
1712 struct inode *inode;
1713 int status = 0;
1714
0b728e19 1715 if (flags & LOOKUP_RCU)
640708a2
PE		 1716 return -ECHILD;
1717
1718 if (!capable(CAP_SYS_ADMIN)) {
1719 status = -EACCES;
1720 goto out_notask;
1721 }
1722
1723 inode = dentry->d_inode;
1724 task = get_proc_task(inode);
1725 if (!task)
1726 goto out_notask;
1727
2344bec7
CW		 1728 mm = mm_access(task, PTRACE_MODE_READ);
1729 if (IS_ERR_OR_NULL(mm))
640708a2
PE		 1730 goto out;
1731
1732 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1733 down_read(&mm->mmap_sem);
1734 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1735 up_read(&mm->mmap_sem);
1736 }
1737
1738 mmput(mm);
1739
1740 if (exact_vma_exists) {
1741 if (task_dumpable(task)) {
1742 rcu_read_lock();
1743 cred = __task_cred(task);
1744 inode->i_uid = cred->euid;
1745 inode->i_gid = cred->egid;
1746 rcu_read_unlock();
1747 } else {
dcb0f222
EB		 1748 inode->i_uid = GLOBAL_ROOT_UID;
1749 inode->i_gid = GLOBAL_ROOT_GID;
640708a2
PE		 1750 }
1751 security_task_to_inode(task, inode);
1752 status = 1;
1753 }
1754
1755out:
1756 put_task_struct(task);
1757
1758out_notask:
1759 if (status <= 0)
1760 d_drop(dentry);
1761
1762 return status;
1763}
1764
1765static const struct dentry_operations tid_map_files_dentry_operations = {
1766 .d_revalidate = map_files_d_revalidate,
1767 .d_delete = pid_delete_dentry,
1768};
1769
1770static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
1771{
1772 unsigned long vm_start, vm_end;
1773 struct vm_area_struct *vma;
1774 struct task_struct *task;
1775 struct mm_struct *mm;
1776 int rc;
1777
1778 rc = -ENOENT;
1779 task = get_proc_task(dentry->d_inode);
1780 if (!task)
1781 goto out;
1782
1783 mm = get_task_mm(task);
1784 put_task_struct(task);
1785 if (!mm)
1786 goto out;
1787
1788 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
1789 if (rc)
1790 goto out_mmput;
1791
1792 down_read(&mm->mmap_sem);
1793 vma = find_exact_vma(mm, vm_start, vm_end);
1794 if (vma && vma->vm_file) {
1795 *path = vma->vm_file->f_path;
1796 path_get(path);
1797 rc = 0;
1798 }
1799 up_read(&mm->mmap_sem);
1800
1801out_mmput:
1802 mmput(mm);
1803out:
1804 return rc;
1805}
1806
1807struct map_files_info {
1808 struct file *file;
1809 unsigned long len;
1810 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
1811};
1812
1813static struct dentry *
1814proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
1815 struct task_struct *task, const void *ptr)
1816{
1817 const struct file *file = ptr;
1818 struct proc_inode *ei;
1819 struct inode *inode;
1820
1821 if (!file)
1822 return ERR_PTR(-ENOENT);
1823
1824 inode = proc_pid_make_inode(dir->i_sb, task);
1825 if (!inode)
1826 return ERR_PTR(-ENOENT);
1827
1828 ei = PROC_I(inode);
1829 ei->op.proc_get_link = proc_map_files_get_link;
1830
1831 inode->i_op = &proc_pid_link_inode_operations;
1832 inode->i_size = 64;
1833 inode->i_mode = S_IFLNK;
1834
1835 if (file->f_mode & FMODE_READ)
1836 inode->i_mode |= S_IRUSR;
1837 if (file->f_mode & FMODE_WRITE)
1838 inode->i_mode |= S_IWUSR;
1839
1840 d_set_d_op(dentry, &tid_map_files_dentry_operations);
1841 d_add(dentry, inode);
1842
1843 return NULL;
1844}
1845
1846static struct dentry *proc_map_files_lookup(struct inode *dir,
00cd8dd3 1847 struct dentry *dentry, unsigned int flags)
640708a2
PE		 1848{
1849 unsigned long vm_start, vm_end;
1850 struct vm_area_struct *vma;
1851 struct task_struct *task;
1852 struct dentry *result;
1853 struct mm_struct *mm;
1854
1855 result = ERR_PTR(-EACCES);
1856 if (!capable(CAP_SYS_ADMIN))
1857 goto out;
1858
1859 result = ERR_PTR(-ENOENT);
1860 task = get_proc_task(dir);
1861 if (!task)
1862 goto out;
1863
1864 result = ERR_PTR(-EACCES);
eb94cd96 1865 if (!ptrace_may_access(task, PTRACE_MODE_READ))
640708a2
PE		 1866 goto out_put_task;
1867
1868 result = ERR_PTR(-ENOENT);
1869 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
eb94cd96 1870 goto out_put_task;
640708a2
PE		 1871
1872 mm = get_task_mm(task);
1873 if (!mm)
eb94cd96 1874 goto out_put_task;
640708a2
PE		 1875
1876 down_read(&mm->mmap_sem);
1877 vma = find_exact_vma(mm, vm_start, vm_end);
1878 if (!vma)
1879 goto out_no_vma;
1880
1881 result = proc_map_files_instantiate(dir, dentry, task, vma->vm_file);
1882
1883out_no_vma:
1884 up_read(&mm->mmap_sem);
1885 mmput(mm);
640708a2
PE		 1886out_put_task:
1887 put_task_struct(task);
1888out:
1889 return result;
1890}
1891
1892static const struct inode_operations proc_map_files_inode_operations = {
1893 .lookup = proc_map_files_lookup,
1894 .permission = proc_fd_permission,
1895 .setattr = proc_setattr,
1896};
1897
1898static int
1899proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
1900{
1901 struct dentry *dentry = filp->f_path.dentry;
1902 struct inode *inode = dentry->d_inode;
1903 struct vm_area_struct *vma;
1904 struct task_struct *task;
1905 struct mm_struct *mm;
1906 ino_t ino;
1907 int ret;
1908
1909 ret = -EACCES;
1910 if (!capable(CAP_SYS_ADMIN))
1911 goto out;
1912
1913 ret = -ENOENT;
1914 task = get_proc_task(inode);
1915 if (!task)
1916 goto out;
1917
1918 ret = -EACCES;
eb94cd96 1919 if (!ptrace_may_access(task, PTRACE_MODE_READ))
640708a2
PE		 1920 goto out_put_task;
1921
1922 ret = 0;
1923 switch (filp->f_pos) {
1924 case 0:
1925 ino = inode->i_ino;
1926 if (filldir(dirent, ".", 1, 0, ino, DT_DIR) < 0)
eb94cd96 1927 goto out_put_task;
640708a2
PE		 1928 filp->f_pos++;
1929 case 1:
1930 ino = parent_ino(dentry);
1931 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
eb94cd96 1932 goto out_put_task;
640708a2
PE		 1933 filp->f_pos++;
1934 default:
1935 {
1936 unsigned long nr_files, pos, i;
1937 struct flex_array *fa = NULL;
1938 struct map_files_info info;
1939 struct map_files_info *p;
1940
1941 mm = get_task_mm(task);
1942 if (!mm)
eb94cd96 1943 goto out_put_task;
640708a2
PE		 1944 down_read(&mm->mmap_sem);
1945
1946 nr_files = 0;
1947
1948 /*
1949 * We need two passes here:
1950 *
1951 * 1) Collect vmas of mapped files with mmap_sem taken
1952 * 2) Release mmap_sem and instantiate entries
1953 *
1954 * otherwise we get lockdep complained, since filldir()
1955 * routine might require mmap_sem taken in might_fault().
1956 */
1957
1958 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
1959 if (vma->vm_file && ++pos > filp->f_pos)
1960 nr_files++;
1961 }
1962
1963 if (nr_files) {
1964 fa = flex_array_alloc(sizeof(info), nr_files,
1965 GFP_KERNEL);
1966 if (!fa || flex_array_prealloc(fa, 0, nr_files,
1967 GFP_KERNEL)) {
1968 ret = -ENOMEM;
1969 if (fa)
1970 flex_array_free(fa);
1971 up_read(&mm->mmap_sem);
1972 mmput(mm);
eb94cd96 1973 goto out_put_task;
640708a2
PE		 1974 }
1975 for (i = 0, vma = mm->mmap, pos = 2; vma;
1976 vma = vma->vm_next) {
1977 if (!vma->vm_file)
1978 continue;
1979 if (++pos <= filp->f_pos)
1980 continue;
1981
1982 get_file(vma->vm_file);
1983 info.file = vma->vm_file;
1984 info.len = snprintf(info.name,
1985 sizeof(info.name), "%lx-%lx",
1986 vma->vm_start, vma->vm_end);
1987 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
1988 BUG();
1989 }
1990 }
1991 up_read(&mm->mmap_sem);
1992
1993 for (i = 0; i < nr_files; i++) {
1994 p = flex_array_get(fa, i);
1995 ret = proc_fill_cache(filp, dirent, filldir,
1996 p->name, p->len,
1997 proc_map_files_instantiate,
1998 task, p->file);
1999 if (ret)
2000 break;
2001 filp->f_pos++;
2002 fput(p->file);
2003 }
2004 for (; i < nr_files; i++) {
2005 /*
2006 * In case of error don't forget
2007 * to put rest of file refs.
2008 */
2009 p = flex_array_get(fa, i);
2010 fput(p->file);
2011 }
2012 if (fa)
2013 flex_array_free(fa);
2014 mmput(mm);
2015 }
2016 }
2017
640708a2
PE		 2018out_put_task:
2019 put_task_struct(task);
2020out:
2021 return ret;
2022}
2023
2024static const struct file_operations proc_map_files_operations = {
2025 .read = generic_read_dir,
2026 .readdir = proc_map_files_readdir,
2027 .llseek = default_llseek,
2028};
2029
2030#endif /* CONFIG_CHECKPOINT_RESTORE */
2031
444ceed8 2032static struct dentry *proc_pident_instantiate(struct inode *dir,
c5141e6d 2033 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8 2034{
c5141e6d 2035 const struct pid_entry *p = ptr;
444ceed8
EB		 2036 struct inode *inode;
2037 struct proc_inode *ei;
bd6daba9 2038 struct dentry *error = ERR_PTR(-ENOENT);
444ceed8 2039
61a28784 2040 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 2041 if (!inode)
2042 goto out;
2043
2044 ei = PROC_I(inode);
2045 inode->i_mode = p->mode;
2046 if (S_ISDIR(inode->i_mode))
bfe86848 2047 set_nlink(inode, 2); /* Use getattr to fix if necessary */
444ceed8
EB		 2048 if (p->iop)
2049 inode->i_op = p->iop;
2050 if (p->fop)
2051 inode->i_fop = p->fop;
2052 ei->op = p->op;
fb045adb 2053 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 2054 d_add(dentry, inode);
2055 /* Close the race of the process dying before we return the dentry */
0b728e19 2056 if (pid_revalidate(dentry, 0))
444ceed8
EB		 2057 error = NULL;
2058out:
2059 return error;
2060}
2061
1da177e4
LT		 2062static struct dentry *proc_pident_lookup(struct inode *dir,
2063 struct dentry *dentry,
c5141e6d 2064 const struct pid_entry *ents,
7bcd6b0e 2065 unsigned int nents)
1da177e4 2066{
cd6a3ce9 2067 struct dentry *error;
99f89551 2068 struct task_struct *task = get_proc_task(dir);
c5141e6d 2069 const struct pid_entry *p, *last;
1da177e4 2070
cd6a3ce9 2071 error = ERR_PTR(-ENOENT);
1da177e4 2072
99f89551
EB		 2073 if (!task)
2074 goto out_no_task;
1da177e4 2075
20cdc894
EB		 2076 /*
2077 * Yes, it does not scale. And it should not. Don't add
2078 * new entries into /proc/<tgid>/ without very good reasons.
2079 */
7bcd6b0e
EB		 2080 last = &ents[nents - 1];
2081 for (p = ents; p <= last; p++) {
1da177e4
LT		 2082 if (p->len != dentry->d_name.len)
2083 continue;
2084 if (!memcmp(dentry->d_name.name, p->name, p->len))
2085 break;
2086 }
7bcd6b0e 2087 if (p > last)
1da177e4
LT		 2088 goto out;
2089
444ceed8 2090 error = proc_pident_instantiate(dir, dentry, task, p);
1da177e4 2091out:
99f89551
EB		 2092 put_task_struct(task);
2093out_no_task:
cd6a3ce9 2094 return error;
1da177e4
LT		 2095}
2096
c5141e6d
ED		 2097static int proc_pident_fill_cache(struct file *filp, void *dirent,
2098 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
61a28784
EB		 2099{
2100 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2101 proc_pident_instantiate, task, p);
2102}
2103
28a6d671
EB		 2104static int proc_pident_readdir(struct file *filp,
2105 void *dirent, filldir_t filldir,
c5141e6d 2106 const struct pid_entry *ents, unsigned int nents)
28a6d671
EB		 2107{
2108 int i;
2fddfeef 2109 struct dentry *dentry = filp->f_path.dentry;
28a6d671
EB		 2110 struct inode *inode = dentry->d_inode;
2111 struct task_struct *task = get_proc_task(inode);
c5141e6d 2112 const struct pid_entry *p, *last;
28a6d671
EB		 2113 ino_t ino;
2114 int ret;
2115
2116 ret = -ENOENT;
2117 if (!task)
61a28784 2118 goto out_no_task;
28a6d671
EB		 2119
2120 ret = 0;
28a6d671
EB		 2121 i = filp->f_pos;
2122 switch (i) {
2123 case 0:
2124 ino = inode->i_ino;
2125 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2126 goto out;
2127 i++;
2128 filp->f_pos++;
2129 /* fall through */
2130 case 1:
2131 ino = parent_ino(dentry);
2132 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2133 goto out;
2134 i++;
2135 filp->f_pos++;
2136 /* fall through */
2137 default:
2138 i -= 2;
2139 if (i >= nents) {
2140 ret = 1;
2141 goto out;
2142 }
2143 p = ents + i;
7bcd6b0e
EB		 2144 last = &ents[nents - 1];
2145 while (p <= last) {
61a28784 2146 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
28a6d671
EB		 2147 goto out;
2148 filp->f_pos++;
2149 p++;
2150 }
2151 }
2152
2153 ret = 1;
2154out:
61a28784
EB		 2155 put_task_struct(task);
2156out_no_task:
28a6d671 2157 return ret;
1da177e4
LT		 2158}
2159
28a6d671
EB		 2160#ifdef CONFIG_SECURITY
2161static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2162 size_t count, loff_t *ppos)
2163{
2fddfeef 2164 struct inode * inode = file->f_path.dentry->d_inode;
04ff9708 2165 char *p = NULL;
28a6d671
EB		 2166 ssize_t length;
2167 struct task_struct *task = get_proc_task(inode);
2168
28a6d671 2169 if (!task)
04ff9708 2170 return -ESRCH;
28a6d671
EB		 2171
2172 length = security_getprocattr(task,
2fddfeef 2173 (char*)file->f_path.dentry->d_name.name,
04ff9708 2174 &p);
28a6d671 2175 put_task_struct(task);
04ff9708
AV		 2176 if (length > 0)
2177 length = simple_read_from_buffer(buf, count, ppos, p, length);
2178 kfree(p);
28a6d671 2179 return length;
1da177e4
LT		 2180}
2181
28a6d671
EB		 2182static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2183 size_t count, loff_t *ppos)
2184{
2fddfeef 2185 struct inode * inode = file->f_path.dentry->d_inode;
28a6d671
EB		 2186 char *page;
2187 ssize_t length;
2188 struct task_struct *task = get_proc_task(inode);
2189
2190 length = -ESRCH;
2191 if (!task)
2192 goto out_no_task;
2193 if (count > PAGE_SIZE)
2194 count = PAGE_SIZE;
2195
2196 /* No partial writes. */
2197 length = -EINVAL;
2198 if (*ppos != 0)
2199 goto out;
2200
2201 length = -ENOMEM;
e12ba74d 2202 page = (char*)__get_free_page(GFP_TEMPORARY);
28a6d671
EB		 2203 if (!page)
2204 goto out;
2205
2206 length = -EFAULT;
2207 if (copy_from_user(page, buf, count))
2208 goto out_free;
2209
107db7c7 2210 /* Guard against adverse ptrace interaction */
9b1bf12d 2211 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
107db7c7
DH		 2212 if (length < 0)
2213 goto out_free;
2214
28a6d671 2215 length = security_setprocattr(task,
2fddfeef 2216 (char*)file->f_path.dentry->d_name.name,
28a6d671 2217 (void*)page, count);
9b1bf12d 2218 mutex_unlock(&task->signal->cred_guard_mutex);
28a6d671
EB		 2219out_free:
2220 free_page((unsigned long) page);
2221out:
2222 put_task_struct(task);
2223out_no_task:
2224 return length;
2225}
2226
00977a59 2227static const struct file_operations proc_pid_attr_operations = {
28a6d671
EB		 2228 .read = proc_pid_attr_read,
2229 .write = proc_pid_attr_write,
87df8424 2230 .llseek = generic_file_llseek,
28a6d671
EB		 2231};
2232
c5141e6d 2233static const struct pid_entry attr_dir_stuff[] = {
631f9c18
AD		 2234 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2235 REG("prev", S_IRUGO, proc_pid_attr_operations),
2236 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2237 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2238 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2239 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
28a6d671
EB		 2240};
2241
72d9dcfc 2242static int proc_attr_dir_readdir(struct file * filp,
28a6d671
EB		 2243 void * dirent, filldir_t filldir)
2244{
2245 return proc_pident_readdir(filp,dirent,filldir,
72d9dcfc 2246 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2247}
2248
00977a59 2249static const struct file_operations proc_attr_dir_operations = {
1da177e4 2250 .read = generic_read_dir,
72d9dcfc 2251 .readdir = proc_attr_dir_readdir,
6038f373 2252 .llseek = default_llseek,
1da177e4
LT		 2253};
2254
72d9dcfc 2255static struct dentry *proc_attr_dir_lookup(struct inode *dir,
00cd8dd3 2256 struct dentry *dentry, unsigned int flags)
28a6d671 2257{
7bcd6b0e
EB		 2258 return proc_pident_lookup(dir, dentry,
2259 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2260}
2261
c5ef1c42 2262static const struct inode_operations proc_attr_dir_inode_operations = {
72d9dcfc 2263 .lookup = proc_attr_dir_lookup,
99f89551 2264 .getattr = pid_getattr,
6d76fa58 2265 .setattr = proc_setattr,
1da177e4
LT		 2266};
2267
28a6d671
EB		 2268#endif
2269
698ba7b5 2270#ifdef CONFIG_ELF_CORE
3cb4a0bb
KH		 2271static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2272 size_t count, loff_t *ppos)
2273{
2274 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2275 struct mm_struct *mm;
2276 char buffer[PROC_NUMBUF];
2277 size_t len;
2278 int ret;
2279
2280 if (!task)
2281 return -ESRCH;
2282
2283 ret = 0;
2284 mm = get_task_mm(task);
2285 if (mm) {
2286 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2287 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2288 MMF_DUMP_FILTER_SHIFT));
2289 mmput(mm);
2290 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2291 }
2292
2293 put_task_struct(task);
2294
2295 return ret;
2296}
2297
2298static ssize_t proc_coredump_filter_write(struct file *file,
2299 const char __user *buf,
2300 size_t count,
2301 loff_t *ppos)
2302{
2303 struct task_struct *task;
2304 struct mm_struct *mm;
2305 char buffer[PROC_NUMBUF], *end;
2306 unsigned int val;
2307 int ret;
2308 int i;
2309 unsigned long mask;
2310
2311 ret = -EFAULT;
2312 memset(buffer, 0, sizeof(buffer));
2313 if (count > sizeof(buffer) - 1)
2314 count = sizeof(buffer) - 1;
2315 if (copy_from_user(buffer, buf, count))
2316 goto out_no_task;
2317
2318 ret = -EINVAL;
2319 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2320 if (*end == '\n')
2321 end++;
2322 if (end - buffer == 0)
2323 goto out_no_task;
2324
2325 ret = -ESRCH;
2326 task = get_proc_task(file->f_dentry->d_inode);
2327 if (!task)
2328 goto out_no_task;
2329
2330 ret = end - buffer;
2331 mm = get_task_mm(task);
2332 if (!mm)
2333 goto out_no_mm;
2334
2335 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2336 if (val & mask)
2337 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2338 else
2339 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2340 }
2341
2342 mmput(mm);
2343 out_no_mm:
2344 put_task_struct(task);
2345 out_no_task:
2346 return ret;
2347}
2348
2349static const struct file_operations proc_coredump_filter_operations = {
2350 .read = proc_coredump_filter_read,
2351 .write = proc_coredump_filter_write,
87df8424 2352 .llseek = generic_file_llseek,
3cb4a0bb
KH		 2353};
2354#endif
2355
28a6d671
EB		 2356/*
2357 * /proc/self:
2358 */
2359static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2360 int buflen)
2361{
488e5bc4 2362 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
b55fcb22 2363 pid_t tgid = task_tgid_nr_ns(current, ns);
28a6d671 2364 char tmp[PROC_NUMBUF];
b55fcb22 2365 if (!tgid)
488e5bc4 2366 return -ENOENT;
b55fcb22 2367 sprintf(tmp, "%d", tgid);
28a6d671
EB		 2368 return vfs_readlink(dentry,buffer,buflen,tmp);
2369}
2370
2371static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2372{
488e5bc4 2373 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
b55fcb22 2374 pid_t tgid = task_tgid_nr_ns(current, ns);
7fee4868
AV		 2375 char *name = ERR_PTR(-ENOENT);
2376 if (tgid) {
2377 name = __getname();
2378 if (!name)
2379 name = ERR_PTR(-ENOMEM);
2380 else
2381 sprintf(name, "%d", tgid);
2382 }
2383 nd_set_link(nd, name);
2384 return NULL;
2385}
2386
2387static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2388 void *cookie)
2389{
2390 char *s = nd_get_link(nd);
2391 if (!IS_ERR(s))
2392 __putname(s);
28a6d671
EB		 2393}
2394
c5ef1c42 2395static const struct inode_operations proc_self_inode_operations = {
28a6d671
EB		 2396 .readlink = proc_self_readlink,
2397 .follow_link = proc_self_follow_link,
7fee4868 2398 .put_link = proc_self_put_link,
28a6d671
EB		 2399};
2400
801199ce
EB		 2401/*
2402 * proc base
2403 *
2404 * These are the directory entries in the root directory of /proc
2405 * that properly belong to the /proc filesystem, as they describe
2406 * describe something that is process related.
2407 */
c5141e6d 2408static const struct pid_entry proc_base_stuff[] = {
61a28784 2409 NOD("self", S_IFLNK|S_IRWXUGO,
801199ce 2410 &proc_self_inode_operations, NULL, {}),
801199ce
EB		 2411};
2412
444ceed8 2413static struct dentry *proc_base_instantiate(struct inode *dir,
c5141e6d 2414 struct dentry *dentry, struct task_struct *task, const void *ptr)
801199ce 2415{
c5141e6d 2416 const struct pid_entry *p = ptr;
801199ce 2417 struct inode *inode;
801199ce 2418 struct proc_inode *ei;
73d36460 2419 struct dentry *error;
801199ce
EB		 2420
2421 /* Allocate the inode */
2422 error = ERR_PTR(-ENOMEM);
2423 inode = new_inode(dir->i_sb);
2424 if (!inode)
2425 goto out;
2426
2427 /* Initialize the inode */
2428 ei = PROC_I(inode);
85fe4025 2429 inode->i_ino = get_next_ino();
801199ce 2430 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
801199ce
EB		 2431
2432 /*
2433 * grab the reference to the task.
2434 */
1a657f78 2435 ei->pid = get_task_pid(task, PIDTYPE_PID);
801199ce
EB		 2436 if (!ei->pid)
2437 goto out_iput;
2438
801199ce
EB		 2439 inode->i_mode = p->mode;
2440 if (S_ISDIR(inode->i_mode))
bfe86848 2441 set_nlink(inode, 2);
801199ce
EB		 2442 if (S_ISLNK(inode->i_mode))
2443 inode->i_size = 64;
2444 if (p->iop)
2445 inode->i_op = p->iop;
2446 if (p->fop)
2447 inode->i_fop = p->fop;
2448 ei->op = p->op;
801199ce
EB		 2449 d_add(dentry, inode);
2450 error = NULL;
2451out:
801199ce
EB		 2452 return error;
2453out_iput:
2454 iput(inode);
2455 goto out;
2456}
2457
444ceed8
EB		 2458static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2459{
2460 struct dentry *error;
2461 struct task_struct *task = get_proc_task(dir);
c5141e6d 2462 const struct pid_entry *p, *last;
444ceed8
EB		 2463
2464 error = ERR_PTR(-ENOENT);
2465
2466 if (!task)
2467 goto out_no_task;
2468
2469 /* Lookup the directory entry */
7bcd6b0e
EB		 2470 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2471 for (p = proc_base_stuff; p <= last; p++) {
444ceed8
EB		 2472 if (p->len != dentry->d_name.len)
2473 continue;
2474 if (!memcmp(dentry->d_name.name, p->name, p->len))
2475 break;
2476 }
7bcd6b0e 2477 if (p > last)
444ceed8
EB		 2478 goto out;
2479
2480 error = proc_base_instantiate(dir, dentry, task, p);
2481
2482out:
2483 put_task_struct(task);
2484out_no_task:
2485 return error;
2486}
2487
c5141e6d
ED		 2488static int proc_base_fill_cache(struct file *filp, void *dirent,
2489 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
61a28784
EB		 2490{
2491 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2492 proc_base_instantiate, task, p);
2493}
2494
aba76fdb 2495#ifdef CONFIG_TASK_IO_ACCOUNTING
297c5d92
AR		 2496static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2497{
940389b8 2498 struct task_io_accounting acct = task->ioac;
5995477a 2499 unsigned long flags;
293eb1e7 2500 int result;
5995477a 2501
293eb1e7
VK		 2502 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2503 if (result)
2504 return result;
2505
2506 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2507 result = -EACCES;
2508 goto out_unlock;
2509 }
1d1221f3 2510
5995477a
AR		 2511 if (whole && lock_task_sighand(task, &flags)) {
2512 struct task_struct *t = task;
2513
2514 task_io_accounting_add(&acct, &task->signal->ioac);
2515 while_each_thread(task, t)
2516 task_io_accounting_add(&acct, &t->ioac);
2517
2518 unlock_task_sighand(task, &flags);
297c5d92 2519 }
293eb1e7 2520 result = sprintf(buffer,
aba76fdb
AM		 2521 "rchar: %llu\n"
2522 "wchar: %llu\n"
2523 "syscr: %llu\n"
2524 "syscw: %llu\n"
2525 "read_bytes: %llu\n"
2526 "write_bytes: %llu\n"
2527 "cancelled_write_bytes: %llu\n",
7c44319d
AB		 2528 (unsigned long long)acct.rchar,
2529 (unsigned long long)acct.wchar,
2530 (unsigned long long)acct.syscr,
2531 (unsigned long long)acct.syscw,
2532 (unsigned long long)acct.read_bytes,
2533 (unsigned long long)acct.write_bytes,
2534 (unsigned long long)acct.cancelled_write_bytes);
293eb1e7
VK		 2535out_unlock:
2536 mutex_unlock(&task->signal->cred_guard_mutex);
2537 return result;
297c5d92
AR		 2538}
2539
2540static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2541{
2542 return do_io_accounting(task, buffer, 0);
aba76fdb 2543}
297c5d92
AR		 2544
2545static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2546{
2547 return do_io_accounting(task, buffer, 1);
2548}
2549#endif /* CONFIG_TASK_IO_ACCOUNTING */
aba76fdb 2550
22d917d8
EB		 2551#ifdef CONFIG_USER_NS
2552static int proc_id_map_open(struct inode *inode, struct file *file,
2553 struct seq_operations *seq_ops)
2554{
2555 struct user_namespace *ns = NULL;
2556 struct task_struct *task;
2557 struct seq_file *seq;
2558 int ret = -EINVAL;
2559
2560 task = get_proc_task(inode);
2561 if (task) {
2562 rcu_read_lock();
2563 ns = get_user_ns(task_cred_xxx(task, user_ns));
2564 rcu_read_unlock();
2565 put_task_struct(task);
2566 }
2567 if (!ns)
2568 goto err;
2569
2570 ret = seq_open(file, seq_ops);
2571 if (ret)
2572 goto err_put_ns;
2573
2574 seq = file->private_data;
2575 seq->private = ns;
2576
2577 return 0;
2578err_put_ns:
2579 put_user_ns(ns);
2580err:
2581 return ret;
2582}
2583
2584static int proc_id_map_release(struct inode *inode, struct file *file)
2585{
2586 struct seq_file *seq = file->private_data;
2587 struct user_namespace *ns = seq->private;
2588 put_user_ns(ns);
2589 return seq_release(inode, file);
2590}
2591
2592static int proc_uid_map_open(struct inode *inode, struct file *file)
2593{
2594 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2595}
2596
2597static int proc_gid_map_open(struct inode *inode, struct file *file)
2598{
2599 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2600}
2601
2602static const struct file_operations proc_uid_map_operations = {
2603 .open = proc_uid_map_open,
2604 .write = proc_uid_map_write,
2605 .read = seq_read,
2606 .llseek = seq_lseek,
2607 .release = proc_id_map_release,
2608};
2609
2610static const struct file_operations proc_gid_map_operations = {
2611 .open = proc_gid_map_open,
2612 .write = proc_gid_map_write,
2613 .read = seq_read,
2614 .llseek = seq_lseek,
2615 .release = proc_id_map_release,
2616};
2617#endif /* CONFIG_USER_NS */
2618
47830723
KC		 2619static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2620 struct pid *pid, struct task_struct *task)
2621{
a9712bc1
AV		 2622 int err = lock_trace(task);
2623 if (!err) {
2624 seq_printf(m, "%08x\n", task->personality);
2625 unlock_trace(task);
2626 }
2627 return err;
47830723
KC		 2628}
2629
28a6d671
EB		 2630/*
2631 * Thread groups
2632 */
00977a59 2633static const struct file_operations proc_task_operations;
c5ef1c42 2634static const struct inode_operations proc_task_inode_operations;
20cdc894 2635
c5141e6d 2636static const struct pid_entry tgid_base_stuff[] = {
631f9c18
AD		 2637 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2638 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
640708a2
PE		 2639#ifdef CONFIG_CHECKPOINT_RESTORE
2640 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
2641#endif
631f9c18 2642 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 2643 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
b2211a36 2644#ifdef CONFIG_NET
631f9c18 2645 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
b2211a36 2646#endif
631f9c18
AD		 2647 REG("environ", S_IRUSR, proc_environ_operations),
2648 INF("auxv", S_IRUSR, proc_pid_auxv),
2649 ONE("status", S_IRUGO, proc_pid_status),
a9712bc1 2650 ONE("personality", S_IRUGO, proc_pid_personality),
3036e7b4 2651 INF("limits", S_IRUGO, proc_pid_limits),
43ae34cb 2652#ifdef CONFIG_SCHED_DEBUG
631f9c18 2653 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
5091faa4
MG		 2654#endif
2655#ifdef CONFIG_SCHED_AUTOGROUP
2656 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
ebcb6734 2657#endif
4614a696 2658 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 2659#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
a9712bc1 2660 INF("syscall", S_IRUGO, proc_pid_syscall),
43ae34cb 2661#endif
631f9c18
AD		 2662 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2663 ONE("stat", S_IRUGO, proc_tgid_stat),
2664 ONE("statm", S_IRUGO, proc_pid_statm),
b7643757 2665 REG("maps", S_IRUGO, proc_pid_maps_operations),
28a6d671 2666#ifdef CONFIG_NUMA
b7643757 2667 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
28a6d671 2668#endif
631f9c18
AD		 2669 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2670 LNK("cwd", proc_cwd_link),
2671 LNK("root", proc_root_link),
2672 LNK("exe", proc_exe_link),
2673 REG("mounts", S_IRUGO, proc_mounts_operations),
2674 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2675 REG("mountstats", S_IRUSR, proc_mountstats_operations),
1e883281 2676#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18 2677 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
b7643757 2678 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
ca6b0bf0 2679 REG("pagemap", S_IRUGO, proc_pagemap_operations),
28a6d671
EB		 2680#endif
2681#ifdef CONFIG_SECURITY
631f9c18 2682 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 2683#endif
2684#ifdef CONFIG_KALLSYMS
631f9c18 2685 INF("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 2686#endif
2ec220e2 2687#ifdef CONFIG_STACKTRACE
a9712bc1 2688 ONE("stack", S_IRUGO, proc_pid_stack),
28a6d671
EB		 2689#endif
2690#ifdef CONFIG_SCHEDSTATS
631f9c18 2691 INF("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 2692#endif
9745512c 2693#ifdef CONFIG_LATENCYTOP
631f9c18 2694 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 2695#endif
8793d854 2696#ifdef CONFIG_PROC_PID_CPUSET
631f9c18 2697 REG("cpuset", S_IRUGO, proc_cpuset_operations),
a424316c
PM		 2698#endif
2699#ifdef CONFIG_CGROUPS
631f9c18 2700 REG("cgroup", S_IRUGO, proc_cgroup_operations),
28a6d671 2701#endif
631f9c18
AD		 2702 INF("oom_score", S_IRUGO, proc_oom_score),
2703 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
a63d83f4 2704 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 2705#ifdef CONFIG_AUDITSYSCALL
631f9c18
AD		 2706 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2707 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 2708#endif
f4f154fd 2709#ifdef CONFIG_FAULT_INJECTION
631f9c18 2710 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 2711#endif
698ba7b5 2712#ifdef CONFIG_ELF_CORE
631f9c18 2713 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3cb4a0bb 2714#endif
aba76fdb 2715#ifdef CONFIG_TASK_IO_ACCOUNTING
1d1221f3 2716 INF("io", S_IRUSR, proc_tgid_io_accounting),
aba76fdb 2717#endif
f133ecca
CM		 2718#ifdef CONFIG_HARDWALL
2719 INF("hardwall", S_IRUGO, proc_pid_hardwall),
2720#endif
22d917d8
EB		 2721#ifdef CONFIG_USER_NS
2722 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
2723 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
2724#endif
28a6d671 2725};
1da177e4 2726
28a6d671 2727static int proc_tgid_base_readdir(struct file * filp,
1da177e4
LT		 2728 void * dirent, filldir_t filldir)
2729{
2730 return proc_pident_readdir(filp,dirent,filldir,
28a6d671 2731 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 2732}
2733
00977a59 2734static const struct file_operations proc_tgid_base_operations = {
1da177e4 2735 .read = generic_read_dir,
28a6d671 2736 .readdir = proc_tgid_base_readdir,
6038f373 2737 .llseek = default_llseek,
1da177e4
LT		 2738};
2739
00cd8dd3
AV		 2740static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2741{
7bcd6b0e
EB		 2742 return proc_pident_lookup(dir, dentry,
2743 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 2744}
2745
c5ef1c42 2746static const struct inode_operations proc_tgid_base_inode_operations = {
28a6d671 2747 .lookup = proc_tgid_base_lookup,
99f89551 2748 .getattr = pid_getattr,
6d76fa58 2749 .setattr = proc_setattr,
0499680a 2750 .permission = proc_pid_permission,
1da177e4 2751};
1da177e4 2752
60347f67 2753static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
1da177e4 2754{
48e6484d 2755 struct dentry *dentry, *leader, *dir;
8578cea7 2756 char buf[PROC_NUMBUF];
48e6484d
EB		 2757 struct qstr name;
2758
2759 name.name = buf;
60347f67
PE		 2760 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2761 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d 2762 if (dentry) {
29f12ca3 2763 shrink_dcache_parent(dentry);
48e6484d
EB		 2764 d_drop(dentry);
2765 dput(dentry);
2766 }
1da177e4 2767
48e6484d 2768 name.name = buf;
60347f67
PE		 2769 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2770 leader = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d
EB		 2771 if (!leader)
2772 goto out;
1da177e4 2773
48e6484d
EB		 2774 name.name = "task";
2775 name.len = strlen(name.name);
2776 dir = d_hash_and_lookup(leader, &name);
2777 if (!dir)
2778 goto out_put_leader;
2779
2780 name.name = buf;
60347f67 2781 name.len = snprintf(buf, sizeof(buf), "%d", pid);
48e6484d
EB		 2782 dentry = d_hash_and_lookup(dir, &name);
2783 if (dentry) {
2784 shrink_dcache_parent(dentry);
2785 d_drop(dentry);
2786 dput(dentry);
1da177e4 2787 }
48e6484d
EB		 2788
2789 dput(dir);
2790out_put_leader:
2791 dput(leader);
2792out:
2793 return;
1da177e4
LT		 2794}
2795
0895e91d
RD		 2796/**
2797 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2798 * @task: task that should be flushed.
2799 *
2800 * When flushing dentries from proc, one needs to flush them from global
60347f67 2801 * proc (proc_mnt) and from all the namespaces' procs this task was seen
0895e91d
RD		 2802 * in. This call is supposed to do all of this job.
2803 *
2804 * Looks in the dcache for
2805 * /proc/@pid
2806 * /proc/@tgid/task/@pid
2807 * if either directory is present flushes it and all of it'ts children
2808 * from the dcache.
2809 *
2810 * It is safe and reasonable to cache /proc entries for a task until
2811 * that task exits. After that they just clog up the dcache with
2812 * useless entries, possibly causing useful dcache entries to be
2813 * flushed instead. This routine is proved to flush those useless
2814 * dcache entries at process exit time.
2815 *
2816 * NOTE: This routine is just an optimization so it does not guarantee
2817 * that no dcache entries will exist at process exit time it
2818 * just makes it very unlikely that any will persist.
60347f67
PE		 2819 */
2820
2821void proc_flush_task(struct task_struct *task)
2822{
9fcc2d15 2823 int i;
9b4d1cbe 2824 struct pid *pid, *tgid;
130f77ec
PE		 2825 struct upid *upid;
2826
130f77ec 2827 pid = task_pid(task);
9b4d1cbe 2828 tgid = task_tgid(task);
130f77ec 2829
9fcc2d15 2830 for (i = 0; i <= pid->level; i++) {
130f77ec
PE		 2831 upid = &pid->numbers[i];
2832 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
9b4d1cbe 2833 tgid->numbers[i].nr);
130f77ec 2834 }
6f4e6433
PE		 2835
2836 upid = &pid->numbers[pid->level];
2837 if (upid->nr == 1)
2838 pid_ns_release_proc(upid->ns);
60347f67
PE		 2839}
2840
9711ef99
AB		 2841static struct dentry *proc_pid_instantiate(struct inode *dir,
2842 struct dentry * dentry,
c5141e6d 2843 struct task_struct *task, const void *ptr)
444ceed8
EB		 2844{
2845 struct dentry *error = ERR_PTR(-ENOENT);
2846 struct inode *inode;
2847
61a28784 2848 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 2849 if (!inode)
2850 goto out;
2851
2852 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2853 inode->i_op = &proc_tgid_base_inode_operations;
2854 inode->i_fop = &proc_tgid_base_operations;
2855 inode->i_flags|=S_IMMUTABLE;
aed54175 2856
bfe86848
MS		 2857 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
2858 ARRAY_SIZE(tgid_base_stuff)));
444ceed8 2859
fb045adb 2860 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 2861
2862 d_add(dentry, inode);
2863 /* Close the race of the process dying before we return the dentry */
0b728e19 2864 if (pid_revalidate(dentry, 0))
444ceed8
EB		 2865 error = NULL;
2866out:
2867 return error;
2868}
2869
00cd8dd3 2870struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
1da177e4 2871{
73d36460 2872 struct dentry *result;
1da177e4 2873 struct task_struct *task;
1da177e4 2874 unsigned tgid;
b488893a 2875 struct pid_namespace *ns;
1da177e4 2876
801199ce
EB		 2877 result = proc_base_lookup(dir, dentry);
2878 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2879 goto out;
2880
1da177e4
LT		 2881 tgid = name_to_int(dentry);
2882 if (tgid == ~0U)
2883 goto out;
2884
b488893a 2885 ns = dentry->d_sb->s_fs_info;
de758734 2886 rcu_read_lock();
b488893a 2887 task = find_task_by_pid_ns(tgid, ns);
1da177e4
LT		 2888 if (task)
2889 get_task_struct(task);
de758734 2890 rcu_read_unlock();
1da177e4
LT		 2891 if (!task)
2892 goto out;
2893
444ceed8 2894 result = proc_pid_instantiate(dir, dentry, task, NULL);
1da177e4 2895 put_task_struct(task);
1da177e4 2896out:
cd6a3ce9 2897 return result;
1da177e4
LT		 2898}
2899
1da177e4 2900/*
0804ef4b 2901 * Find the first task with tgid >= tgid
0bc58a91 2902 *
1da177e4 2903 */
19fd4bb2
EB		 2904struct tgid_iter {
2905 unsigned int tgid;
0804ef4b 2906 struct task_struct *task;
19fd4bb2
EB		 2907};
2908static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2909{
0804ef4b 2910 struct pid *pid;
1da177e4 2911
19fd4bb2
EB		 2912 if (iter.task)
2913 put_task_struct(iter.task);
454cc105 2914 rcu_read_lock();
0804ef4b 2915retry:
19fd4bb2
EB		 2916 iter.task = NULL;
2917 pid = find_ge_pid(iter.tgid, ns);
0804ef4b 2918 if (pid) {
19fd4bb2
EB		 2919 iter.tgid = pid_nr_ns(pid, ns);
2920 iter.task = pid_task(pid, PIDTYPE_PID);
0804ef4b
EB		 2921 /* What we to know is if the pid we have find is the
2922 * pid of a thread_group_leader. Testing for task
2923 * being a thread_group_leader is the obvious thing
2924 * todo but there is a window when it fails, due to
2925 * the pid transfer logic in de_thread.
2926 *
2927 * So we perform the straight forward test of seeing
2928 * if the pid we have found is the pid of a thread
2929 * group leader, and don't worry if the task we have
2930 * found doesn't happen to be a thread group leader.
2931 * As we don't care in the case of readdir.
2932 */
19fd4bb2
EB		 2933 if (!iter.task || !has_group_leader_pid(iter.task)) {
2934 iter.tgid += 1;
0804ef4b 2935 goto retry;
19fd4bb2
EB		 2936 }
2937 get_task_struct(iter.task);
0bc58a91 2938 }
454cc105 2939 rcu_read_unlock();
19fd4bb2 2940 return iter;
1da177e4
LT		 2941}
2942
7bcd6b0e 2943#define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
0804ef4b 2944
61a28784 2945static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
19fd4bb2 2946 struct tgid_iter iter)
61a28784
EB		 2947{
2948 char name[PROC_NUMBUF];
19fd4bb2 2949 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
61a28784 2950 return proc_fill_cache(filp, dirent, filldir, name, len,
19fd4bb2 2951 proc_pid_instantiate, iter.task, NULL);
61a28784
EB		 2952}
2953
0499680a
VK		 2954static int fake_filldir(void *buf, const char *name, int namelen,
2955 loff_t offset, u64 ino, unsigned d_type)
2956{
2957 return 0;
2958}
2959
1da177e4
LT		 2960/* for the /proc/ directory itself, after non-process stuff has been done */
2961int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2962{
d8bdc59f
LT		 2963 unsigned int nr;
2964 struct task_struct *reaper;
19fd4bb2 2965 struct tgid_iter iter;
b488893a 2966 struct pid_namespace *ns;
0499680a 2967 filldir_t __filldir;
1da177e4 2968
d8bdc59f
LT		 2969 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
2970 goto out_no_task;
2971 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2972
2973 reaper = get_proc_task(filp->f_path.dentry->d_inode);
61a28784
EB		 2974 if (!reaper)
2975 goto out_no_task;
2976
7bcd6b0e 2977 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
c5141e6d 2978 const struct pid_entry *p = &proc_base_stuff[nr];
61a28784 2979 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
801199ce 2980 goto out;
1da177e4
LT		 2981 }
2982
b488893a 2983 ns = filp->f_dentry->d_sb->s_fs_info;
19fd4bb2
EB		 2984 iter.task = NULL;
2985 iter.tgid = filp->f_pos - TGID_OFFSET;
2986 for (iter = next_tgid(ns, iter);
2987 iter.task;
2988 iter.tgid += 1, iter = next_tgid(ns, iter)) {
0499680a
VK		 2989 if (has_pid_permissions(ns, iter.task, 2))
2990 __filldir = filldir;
2991 else
2992 __filldir = fake_filldir;
2993
19fd4bb2 2994 filp->f_pos = iter.tgid + TGID_OFFSET;
0499680a 2995 if (proc_pid_fill_cache(filp, dirent, __filldir, iter) < 0) {
19fd4bb2 2996 put_task_struct(iter.task);
0804ef4b 2997 goto out;
1da177e4 2998 }
0bc58a91 2999 }
0804ef4b
EB		 3000 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3001out:
61a28784
EB		 3002 put_task_struct(reaper);
3003out_no_task:
0bc58a91
EB		 3004 return 0;
3005}
1da177e4 3006
28a6d671
EB		 3007/*
3008 * Tasks
3009 */
c5141e6d 3010static const struct pid_entry tid_base_stuff[] = {
631f9c18 3011 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3835541d 3012 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 3013 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
631f9c18
AD		 3014 REG("environ", S_IRUSR, proc_environ_operations),
3015 INF("auxv", S_IRUSR, proc_pid_auxv),
3016 ONE("status", S_IRUGO, proc_pid_status),
a9712bc1 3017 ONE("personality", S_IRUGO, proc_pid_personality),
3036e7b4 3018 INF("limits", S_IRUGO, proc_pid_limits),
43ae34cb 3019#ifdef CONFIG_SCHED_DEBUG
631f9c18 3020 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
ebcb6734 3021#endif
4614a696 3022 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 3023#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
a9712bc1 3024 INF("syscall", S_IRUGO, proc_pid_syscall),
43ae34cb 3025#endif
631f9c18
AD		 3026 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3027 ONE("stat", S_IRUGO, proc_tid_stat),
3028 ONE("statm", S_IRUGO, proc_pid_statm),
b7643757 3029 REG("maps", S_IRUGO, proc_tid_maps_operations),
81841161
CG		 3030#ifdef CONFIG_CHECKPOINT_RESTORE
3031 REG("children", S_IRUGO, proc_tid_children_operations),
3032#endif
28a6d671 3033#ifdef CONFIG_NUMA
b7643757 3034 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
28a6d671 3035#endif
631f9c18
AD		 3036 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3037 LNK("cwd", proc_cwd_link),
3038 LNK("root", proc_root_link),
3039 LNK("exe", proc_exe_link),
3040 REG("mounts", S_IRUGO, proc_mounts_operations),
3041 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
1e883281 3042#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18 3043 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
b7643757 3044 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
ca6b0bf0 3045 REG("pagemap", S_IRUGO, proc_pagemap_operations),
28a6d671
EB		 3046#endif
3047#ifdef CONFIG_SECURITY
631f9c18 3048 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 3049#endif
3050#ifdef CONFIG_KALLSYMS
631f9c18 3051 INF("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 3052#endif
2ec220e2 3053#ifdef CONFIG_STACKTRACE
a9712bc1 3054 ONE("stack", S_IRUGO, proc_pid_stack),
28a6d671
EB		 3055#endif
3056#ifdef CONFIG_SCHEDSTATS
631f9c18 3057 INF("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 3058#endif
9745512c 3059#ifdef CONFIG_LATENCYTOP
631f9c18 3060 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 3061#endif
8793d854 3062#ifdef CONFIG_PROC_PID_CPUSET
631f9c18 3063 REG("cpuset", S_IRUGO, proc_cpuset_operations),
a424316c
PM		 3064#endif
3065#ifdef CONFIG_CGROUPS
631f9c18 3066 REG("cgroup", S_IRUGO, proc_cgroup_operations),
28a6d671 3067#endif
631f9c18
AD		 3068 INF("oom_score", S_IRUGO, proc_oom_score),
3069 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
a63d83f4 3070 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 3071#ifdef CONFIG_AUDITSYSCALL
631f9c18 3072 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
26ec3c64 3073 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 3074#endif
f4f154fd 3075#ifdef CONFIG_FAULT_INJECTION
631f9c18 3076 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 3077#endif
297c5d92 3078#ifdef CONFIG_TASK_IO_ACCOUNTING
1d1221f3 3079 INF("io", S_IRUSR, proc_tid_io_accounting),
297c5d92 3080#endif
f133ecca
CM		 3081#ifdef CONFIG_HARDWALL
3082 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3083#endif
22d917d8
EB		 3084#ifdef CONFIG_USER_NS
3085 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3086 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3087#endif
28a6d671
EB		 3088};
3089
3090static int proc_tid_base_readdir(struct file * filp,
3091 void * dirent, filldir_t filldir)
3092{
3093 return proc_pident_readdir(filp,dirent,filldir,
3094 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3095}
3096
00cd8dd3
AV		 3097static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3098{
7bcd6b0e
EB		 3099 return proc_pident_lookup(dir, dentry,
3100 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
28a6d671
EB		 3101}
3102
00977a59 3103static const struct file_operations proc_tid_base_operations = {
28a6d671
EB		 3104 .read = generic_read_dir,
3105 .readdir = proc_tid_base_readdir,
6038f373 3106 .llseek = default_llseek,
28a6d671
EB		 3107};
3108
c5ef1c42 3109static const struct inode_operations proc_tid_base_inode_operations = {
28a6d671
EB		 3110 .lookup = proc_tid_base_lookup,
3111 .getattr = pid_getattr,
3112 .setattr = proc_setattr,
3113};
3114
444ceed8 3115static struct dentry *proc_task_instantiate(struct inode *dir,
c5141e6d 3116 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8
EB		 3117{
3118 struct dentry *error = ERR_PTR(-ENOENT);
3119 struct inode *inode;
61a28784 3120 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 3121
3122 if (!inode)
3123 goto out;
3124 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3125 inode->i_op = &proc_tid_base_inode_operations;
3126 inode->i_fop = &proc_tid_base_operations;
3127 inode->i_flags|=S_IMMUTABLE;
aed54175 3128
bfe86848
MS		 3129 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3130 ARRAY_SIZE(tid_base_stuff)));
444ceed8 3131
fb045adb 3132 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 3133
3134 d_add(dentry, inode);
3135 /* Close the race of the process dying before we return the dentry */
0b728e19 3136 if (pid_revalidate(dentry, 0))
444ceed8
EB		 3137 error = NULL;
3138out:
3139 return error;
3140}
3141
00cd8dd3 3142static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
28a6d671
EB		 3143{
3144 struct dentry *result = ERR_PTR(-ENOENT);
3145 struct task_struct *task;
3146 struct task_struct *leader = get_proc_task(dir);
28a6d671 3147 unsigned tid;
b488893a 3148 struct pid_namespace *ns;
28a6d671
EB		 3149
3150 if (!leader)
3151 goto out_no_task;
3152
3153 tid = name_to_int(dentry);
3154 if (tid == ~0U)
3155 goto out;
3156
b488893a 3157 ns = dentry->d_sb->s_fs_info;
28a6d671 3158 rcu_read_lock();
b488893a 3159 task = find_task_by_pid_ns(tid, ns);
28a6d671
EB		 3160 if (task)
3161 get_task_struct(task);
3162 rcu_read_unlock();
3163 if (!task)
3164 goto out;
bac0abd6 3165 if (!same_thread_group(leader, task))
28a6d671
EB		 3166 goto out_drop_task;
3167
444ceed8 3168 result = proc_task_instantiate(dir, dentry, task, NULL);
28a6d671
EB		 3169out_drop_task:
3170 put_task_struct(task);
3171out:
3172 put_task_struct(leader);
3173out_no_task:
3174 return result;
3175}
3176
0bc58a91
EB		 3177/*
3178 * Find the first tid of a thread group to return to user space.
3179 *
3180 * Usually this is just the thread group leader, but if the users
3181 * buffer was too small or there was a seek into the middle of the
3182 * directory we have more work todo.
3183 *
3184 * In the case of a short read we start with find_task_by_pid.
3185 *
3186 * In the case of a seek we start with the leader and walk nr
3187 * threads past it.
3188 */
cc288738 3189static struct task_struct *first_tid(struct task_struct *leader,
b488893a 3190 int tid, int nr, struct pid_namespace *ns)
0bc58a91 3191{
a872ff0c 3192 struct task_struct *pos;
1da177e4 3193
cc288738 3194 rcu_read_lock();
0bc58a91
EB		 3195 /* Attempt to start with the pid of a thread */
3196 if (tid && (nr > 0)) {
b488893a 3197 pos = find_task_by_pid_ns(tid, ns);
a872ff0c
ON		 3198 if (pos && (pos->group_leader == leader))
3199 goto found;
0bc58a91 3200 }
1da177e4 3201
0bc58a91 3202 /* If nr exceeds the number of threads there is nothing todo */
a872ff0c
ON		 3203 pos = NULL;
3204 if (nr && nr >= get_nr_threads(leader))
3205 goto out;
1da177e4 3206
a872ff0c
ON		 3207 /* If we haven't found our starting place yet start
3208 * with the leader and walk nr threads forward.
0bc58a91 3209 */
a872ff0c
ON		 3210 for (pos = leader; nr > 0; --nr) {
3211 pos = next_thread(pos);
3212 if (pos == leader) {
3213 pos = NULL;
3214 goto out;
3215 }
1da177e4 3216 }
a872ff0c
ON		 3217found:
3218 get_task_struct(pos);
3219out:
cc288738 3220 rcu_read_unlock();
0bc58a91
EB		 3221 return pos;
3222}
3223
3224/*
3225 * Find the next thread in the thread list.
3226 * Return NULL if there is an error or no next thread.
3227 *
3228 * The reference to the input task_struct is released.
3229 */
3230static struct task_struct *next_tid(struct task_struct *start)
3231{
c1df7fb8 3232 struct task_struct *pos = NULL;
cc288738 3233 rcu_read_lock();
c1df7fb8 3234 if (pid_alive(start)) {
0bc58a91 3235 pos = next_thread(start);
c1df7fb8
ON		 3236 if (thread_group_leader(pos))
3237 pos = NULL;
3238 else
3239 get_task_struct(pos);
3240 }
cc288738 3241 rcu_read_unlock();
0bc58a91
EB		 3242 put_task_struct(start);
3243 return pos;
1da177e4
LT		 3244}
3245
61a28784
EB		 3246static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3247 struct task_struct *task, int tid)
3248{
3249 char name[PROC_NUMBUF];
3250 int len = snprintf(name, sizeof(name), "%d", tid);
3251 return proc_fill_cache(filp, dirent, filldir, name, len,
3252 proc_task_instantiate, task, NULL);
3253}
3254
1da177e4
LT		 3255/* for the /proc/TGID/task/ directories */
3256static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3257{
2fddfeef 3258 struct dentry *dentry = filp->f_path.dentry;
1da177e4 3259 struct inode *inode = dentry->d_inode;
7d895244 3260 struct task_struct *leader = NULL;
0bc58a91 3261 struct task_struct *task;
1da177e4
LT		 3262 int retval = -ENOENT;
3263 ino_t ino;
0bc58a91 3264 int tid;
b488893a 3265 struct pid_namespace *ns;
1da177e4 3266
7d895244
GC		 3267 task = get_proc_task(inode);
3268 if (!task)
3269 goto out_no_task;
3270 rcu_read_lock();
3271 if (pid_alive(task)) {
3272 leader = task->group_leader;
3273 get_task_struct(leader);
3274 }
3275 rcu_read_unlock();
3276 put_task_struct(task);
99f89551
EB		 3277 if (!leader)
3278 goto out_no_task;
1da177e4
LT		 3279 retval = 0;
3280
ee568b25 3281 switch ((unsigned long)filp->f_pos) {
1da177e4
LT		 3282 case 0:
3283 ino = inode->i_ino;
ee6f779b 3284 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
1da177e4 3285 goto out;
ee6f779b 3286 filp->f_pos++;
1da177e4
LT		 3287 /* fall through */
3288 case 1:
3289 ino = parent_ino(dentry);
ee6f779b 3290 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
1da177e4 3291 goto out;
ee6f779b 3292 filp->f_pos++;
1da177e4
LT		 3293 /* fall through */
3294 }
3295
0bc58a91
EB		 3296 /* f_version caches the tgid value that the last readdir call couldn't
3297 * return. lseek aka telldir automagically resets f_version to 0.
3298 */
b488893a 3299 ns = filp->f_dentry->d_sb->s_fs_info;
2b47c361 3300 tid = (int)filp->f_version;
0bc58a91 3301 filp->f_version = 0;
ee6f779b 3302 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
0bc58a91 3303 task;
ee6f779b 3304 task = next_tid(task), filp->f_pos++) {
b488893a 3305 tid = task_pid_nr_ns(task, ns);
61a28784 3306 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
0bc58a91
EB		 3307 /* returning this tgid failed, save it as the first
3308 * pid for the next readir call */
2b47c361 3309 filp->f_version = (u64)tid;
0bc58a91 3310 put_task_struct(task);
1da177e4 3311 break;
0bc58a91 3312 }
1da177e4
LT		 3313 }
3314out:
99f89551
EB		 3315 put_task_struct(leader);
3316out_no_task:
1da177e4
LT		 3317 return retval;
3318}
6e66b52b
EB		 3319
3320static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3321{
3322 struct inode *inode = dentry->d_inode;
99f89551 3323 struct task_struct *p = get_proc_task(inode);
6e66b52b
EB		 3324 generic_fillattr(inode, stat);
3325
99f89551 3326 if (p) {
99f89551 3327 stat->nlink += get_nr_threads(p);
99f89551 3328 put_task_struct(p);
6e66b52b
EB		 3329 }
3330
3331 return 0;
3332}
28a6d671 3333
c5ef1c42 3334static const struct inode_operations proc_task_inode_operations = {
28a6d671
EB		 3335 .lookup = proc_task_lookup,
3336 .getattr = proc_task_getattr,
3337 .setattr = proc_setattr,
0499680a 3338 .permission = proc_pid_permission,
28a6d671
EB		 3339};
3340
00977a59 3341static const struct file_operations proc_task_operations = {
28a6d671
EB		 3342 .read = generic_read_dir,
3343 .readdir = proc_task_readdir,
6038f373 3344 .llseek = default_llseek,
28a6d671 3345};
Linux kernel - Deliverables
This page took 0.977541 seconds and 5 git commands to generate.