projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
mm/vmscan: fix error number for failed kthread
[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",
e1760bd5
EB		 1091 from_kuid(file->f_cred->user_ns,
1092 audit_get_loginuid(task)));
99f89551 1093 put_task_struct(task);
1da177e4
LT		 1094 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1095}
1096
1097static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1098 size_t count, loff_t *ppos)
1099{
2fddfeef 1100 struct inode * inode = file->f_path.dentry->d_inode;
1da177e4
LT		 1101 char *page, *tmp;
1102 ssize_t length;
1da177e4 1103 uid_t loginuid;
e1760bd5 1104 kuid_t kloginuid;
1da177e4 1105
7dc52157
PM		 1106 rcu_read_lock();
1107 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1108 rcu_read_unlock();
1da177e4 1109 return -EPERM;
7dc52157
PM		 1110 }
1111 rcu_read_unlock();
1da177e4 1112
e0182909
AV		 1113 if (count >= PAGE_SIZE)
1114 count = PAGE_SIZE - 1;
1da177e4
LT		 1115
1116 if (*ppos != 0) {
1117 /* No partial writes. */
1118 return -EINVAL;
1119 }
e12ba74d 1120 page = (char*)__get_free_page(GFP_TEMPORARY);
1da177e4
LT		 1121 if (!page)
1122 return -ENOMEM;
1123 length = -EFAULT;
1124 if (copy_from_user(page, buf, count))
1125 goto out_free_page;
1126
e0182909 1127 page[count] = '\0';
1da177e4
LT		 1128 loginuid = simple_strtoul(page, &tmp, 10);
1129 if (tmp == page) {
1130 length = -EINVAL;
1131 goto out_free_page;
1132
1133 }
e1760bd5
EB		 1134 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1135 if (!uid_valid(kloginuid)) {
1136 length = -EINVAL;
1137 goto out_free_page;
1138 }
1139
1140 length = audit_set_loginuid(kloginuid);
1da177e4
LT		 1141 if (likely(length == 0))
1142 length = count;
1143
1144out_free_page:
1145 free_page((unsigned long) page);
1146 return length;
1147}
1148
00977a59 1149static const struct file_operations proc_loginuid_operations = {
1da177e4
LT		 1150 .read = proc_loginuid_read,
1151 .write = proc_loginuid_write,
87df8424 1152 .llseek = generic_file_llseek,
1da177e4 1153};
1e0bd755
EP		 1154
1155static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1156 size_t count, loff_t *ppos)
1157{
1158 struct inode * inode = file->f_path.dentry->d_inode;
1159 struct task_struct *task = get_proc_task(inode);
1160 ssize_t length;
1161 char tmpbuf[TMPBUFLEN];
1162
1163 if (!task)
1164 return -ESRCH;
1165 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1166 audit_get_sessionid(task));
1167 put_task_struct(task);
1168 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1169}
1170
1171static const struct file_operations proc_sessionid_operations = {
1172 .read = proc_sessionid_read,
87df8424 1173 .llseek = generic_file_llseek,
1e0bd755 1174};
1da177e4
LT		 1175#endif
1176
f4f154fd
AM		 1177#ifdef CONFIG_FAULT_INJECTION
1178static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1179 size_t count, loff_t *ppos)
1180{
1181 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1182 char buffer[PROC_NUMBUF];
1183 size_t len;
1184 int make_it_fail;
f4f154fd
AM		 1185
1186 if (!task)
1187 return -ESRCH;
1188 make_it_fail = task->make_it_fail;
1189 put_task_struct(task);
1190
1191 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
0c28f287
AM		 1192
1193 return simple_read_from_buffer(buf, count, ppos, buffer, len);
f4f154fd
AM		 1194}
1195
1196static ssize_t proc_fault_inject_write(struct file * file,
1197 const char __user * buf, size_t count, loff_t *ppos)
1198{
1199 struct task_struct *task;
1200 char buffer[PROC_NUMBUF], *end;
1201 int make_it_fail;
1202
1203 if (!capable(CAP_SYS_RESOURCE))
1204 return -EPERM;
1205 memset(buffer, 0, sizeof(buffer));
1206 if (count > sizeof(buffer) - 1)
1207 count = sizeof(buffer) - 1;
1208 if (copy_from_user(buffer, buf, count))
1209 return -EFAULT;
cba8aafe
VL		 1210 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1211 if (*end)
1212 return -EINVAL;
f4f154fd
AM		 1213 task = get_proc_task(file->f_dentry->d_inode);
1214 if (!task)
1215 return -ESRCH;
1216 task->make_it_fail = make_it_fail;
1217 put_task_struct(task);
cba8aafe
VL		 1218
1219 return count;
f4f154fd
AM		 1220}
1221
00977a59 1222static const struct file_operations proc_fault_inject_operations = {
f4f154fd
AM		 1223 .read = proc_fault_inject_read,
1224 .write = proc_fault_inject_write,
87df8424 1225 .llseek = generic_file_llseek,
f4f154fd
AM		 1226};
1227#endif
1228
9745512c 1229
43ae34cb
IM		 1230#ifdef CONFIG_SCHED_DEBUG
1231/*
1232 * Print out various scheduling related per-task fields:
1233 */
1234static int sched_show(struct seq_file *m, void *v)
1235{
1236 struct inode *inode = m->private;
1237 struct task_struct *p;
1238
43ae34cb
IM		 1239 p = get_proc_task(inode);
1240 if (!p)
1241 return -ESRCH;
1242 proc_sched_show_task(p, m);
1243
1244 put_task_struct(p);
1245
1246 return 0;
1247}
1248
1249static ssize_t
1250sched_write(struct file *file, const char __user *buf,
1251 size_t count, loff_t *offset)
1252{
1253 struct inode *inode = file->f_path.dentry->d_inode;
1254 struct task_struct *p;
1255
43ae34cb
IM		 1256 p = get_proc_task(inode);
1257 if (!p)
1258 return -ESRCH;
1259 proc_sched_set_task(p);
1260
1261 put_task_struct(p);
1262
1263 return count;
1264}
1265
1266static int sched_open(struct inode *inode, struct file *filp)
1267{
c6a34058 1268 return single_open(filp, sched_show, inode);
43ae34cb
IM		 1269}
1270
1271static const struct file_operations proc_pid_sched_operations = {
1272 .open = sched_open,
1273 .read = seq_read,
1274 .write = sched_write,
1275 .llseek = seq_lseek,
5ea473a1 1276 .release = single_release,
43ae34cb
IM		 1277};
1278
1279#endif
1280
5091faa4
MG		 1281#ifdef CONFIG_SCHED_AUTOGROUP
1282/*
1283 * Print out autogroup related information:
1284 */
1285static int sched_autogroup_show(struct seq_file *m, void *v)
1286{
1287 struct inode *inode = m->private;
1288 struct task_struct *p;
1289
1290 p = get_proc_task(inode);
1291 if (!p)
1292 return -ESRCH;
1293 proc_sched_autogroup_show_task(p, m);
1294
1295 put_task_struct(p);
1296
1297 return 0;
1298}
1299
1300static ssize_t
1301sched_autogroup_write(struct file *file, const char __user *buf,
1302 size_t count, loff_t *offset)
1303{
1304 struct inode *inode = file->f_path.dentry->d_inode;
1305 struct task_struct *p;
1306 char buffer[PROC_NUMBUF];
0a8cb8e3 1307 int nice;
5091faa4
MG		 1308 int err;
1309
1310 memset(buffer, 0, sizeof(buffer));
1311 if (count > sizeof(buffer) - 1)
1312 count = sizeof(buffer) - 1;
1313 if (copy_from_user(buffer, buf, count))
1314 return -EFAULT;
1315
0a8cb8e3
AD		 1316 err = kstrtoint(strstrip(buffer), 0, &nice);
1317 if (err < 0)
1318 return err;
5091faa4
MG		 1319
1320 p = get_proc_task(inode);
1321 if (!p)
1322 return -ESRCH;
1323
2e5b5b3a 1324 err = proc_sched_autogroup_set_nice(p, nice);
5091faa4
MG		 1325 if (err)
1326 count = err;
1327
1328 put_task_struct(p);
1329
1330 return count;
1331}
1332
1333static int sched_autogroup_open(struct inode *inode, struct file *filp)
1334{
1335 int ret;
1336
1337 ret = single_open(filp, sched_autogroup_show, NULL);
1338 if (!ret) {
1339 struct seq_file *m = filp->private_data;
1340
1341 m->private = inode;
1342 }
1343 return ret;
1344}
1345
1346static const struct file_operations proc_pid_sched_autogroup_operations = {
1347 .open = sched_autogroup_open,
1348 .read = seq_read,
1349 .write = sched_autogroup_write,
1350 .llseek = seq_lseek,
1351 .release = single_release,
1352};
1353
1354#endif /* CONFIG_SCHED_AUTOGROUP */
1355
4614a696 1356static ssize_t comm_write(struct file *file, const char __user *buf,
1357 size_t count, loff_t *offset)
1358{
1359 struct inode *inode = file->f_path.dentry->d_inode;
1360 struct task_struct *p;
1361 char buffer[TASK_COMM_LEN];
1362
1363 memset(buffer, 0, sizeof(buffer));
1364 if (count > sizeof(buffer) - 1)
1365 count = sizeof(buffer) - 1;
1366 if (copy_from_user(buffer, buf, count))
1367 return -EFAULT;
1368
1369 p = get_proc_task(inode);
1370 if (!p)
1371 return -ESRCH;
1372
1373 if (same_thread_group(current, p))
1374 set_task_comm(p, buffer);
1375 else
1376 count = -EINVAL;
1377
1378 put_task_struct(p);
1379
1380 return count;
1381}
1382
1383static int comm_show(struct seq_file *m, void *v)
1384{
1385 struct inode *inode = m->private;
1386 struct task_struct *p;
1387
1388 p = get_proc_task(inode);
1389 if (!p)
1390 return -ESRCH;
1391
1392 task_lock(p);
1393 seq_printf(m, "%s\n", p->comm);
1394 task_unlock(p);
1395
1396 put_task_struct(p);
1397
1398 return 0;
1399}
1400
1401static int comm_open(struct inode *inode, struct file *filp)
1402{
c6a34058 1403 return single_open(filp, comm_show, inode);
4614a696 1404}
1405
1406static const struct file_operations proc_pid_set_comm_operations = {
1407 .open = comm_open,
1408 .read = seq_read,
1409 .write = comm_write,
1410 .llseek = seq_lseek,
1411 .release = single_release,
1412};
1413
7773fbc5 1414static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
925d1c40
MH		 1415{
1416 struct task_struct *task;
1417 struct mm_struct *mm;
1418 struct file *exe_file;
1419
7773fbc5 1420 task = get_proc_task(dentry->d_inode);
925d1c40
MH		 1421 if (!task)
1422 return -ENOENT;
1423 mm = get_task_mm(task);
1424 put_task_struct(task);
1425 if (!mm)
1426 return -ENOENT;
1427 exe_file = get_mm_exe_file(mm);
1428 mmput(mm);
1429 if (exe_file) {
1430 *exe_path = exe_file->f_path;
1431 path_get(&exe_file->f_path);
1432 fput(exe_file);
1433 return 0;
1434 } else
1435 return -ENOENT;
1436}
1437
008b150a 1438static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1da177e4
LT		 1439{
1440 struct inode *inode = dentry->d_inode;
408ef013 1441 struct path path;
1da177e4
LT		 1442 int error = -EACCES;
1443
778c1144
EB		 1444 /* Are we allowed to snoop on the tasks file descriptors? */
1445 if (!proc_fd_access_allowed(inode))
1da177e4 1446 goto out;
1da177e4 1447
408ef013
CH		 1448 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1449 if (error)
1450 goto out;
1451
b5fb63c1 1452 nd_jump_link(nd, &path);
408ef013 1453 return NULL;
1da177e4 1454out:
008b150a 1455 return ERR_PTR(error);
1da177e4
LT		 1456}
1457
3dcd25f3 1458static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1da177e4 1459{
e12ba74d 1460 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
3dcd25f3 1461 char *pathname;
1da177e4
LT		 1462 int len;
1463
1464 if (!tmp)
1465 return -ENOMEM;
0c28f287 1466
7b2a69ba 1467 pathname = d_path(path, tmp, PAGE_SIZE);
3dcd25f3
JB		 1468 len = PTR_ERR(pathname);
1469 if (IS_ERR(pathname))
1da177e4 1470 goto out;
3dcd25f3 1471 len = tmp + PAGE_SIZE - 1 - pathname;
1da177e4
LT		 1472
1473 if (len > buflen)
1474 len = buflen;
3dcd25f3 1475 if (copy_to_user(buffer, pathname, len))
1da177e4
LT		 1476 len = -EFAULT;
1477 out:
1478 free_page((unsigned long)tmp);
1479 return len;
1480}
1481
1482static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1483{
1484 int error = -EACCES;
1485 struct inode *inode = dentry->d_inode;
3dcd25f3 1486 struct path path;
1da177e4 1487
778c1144
EB		 1488 /* Are we allowed to snoop on the tasks file descriptors? */
1489 if (!proc_fd_access_allowed(inode))
1da177e4 1490 goto out;
1da177e4 1491
7773fbc5 1492 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1da177e4
LT		 1493 if (error)
1494 goto out;
1495
3dcd25f3
JB		 1496 error = do_proc_readlink(&path, buffer, buflen);
1497 path_put(&path);
1da177e4 1498out:
1da177e4
LT		 1499 return error;
1500}
1501
faf60af1 1502const struct inode_operations proc_pid_link_inode_operations = {
1da177e4 1503 .readlink = proc_pid_readlink,
6d76fa58
LT		 1504 .follow_link = proc_pid_follow_link,
1505 .setattr = proc_setattr,
1da177e4
LT		 1506};
1507
28a6d671
EB		 1508
1509/* building an inode */
1510
6b4e306a 1511struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
28a6d671
EB		 1512{
1513 struct inode * inode;
1514 struct proc_inode *ei;
c69e8d9c 1515 const struct cred *cred;
1da177e4 1516
28a6d671 1517 /* We need a new inode */
1da177e4 1518
28a6d671
EB		 1519 inode = new_inode(sb);
1520 if (!inode)
1521 goto out;
1522
1523 /* Common stuff */
1524 ei = PROC_I(inode);
85fe4025 1525 inode->i_ino = get_next_ino();
28a6d671 1526 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
28a6d671
EB		 1527 inode->i_op = &proc_def_inode_operations;
1528
1529 /*
1530 * grab the reference to task.
1531 */
1a657f78 1532 ei->pid = get_task_pid(task, PIDTYPE_PID);
28a6d671
EB		 1533 if (!ei->pid)
1534 goto out_unlock;
1535
28a6d671 1536 if (task_dumpable(task)) {
c69e8d9c
DH		 1537 rcu_read_lock();
1538 cred = __task_cred(task);
1539 inode->i_uid = cred->euid;
1540 inode->i_gid = cred->egid;
1541 rcu_read_unlock();
1da177e4 1542 }
28a6d671
EB		 1543 security_task_to_inode(task, inode);
1544
1da177e4 1545out:
28a6d671
EB		 1546 return inode;
1547
1548out_unlock:
1549 iput(inode);
1550 return NULL;
1da177e4
LT		 1551}
1552
6b4e306a 1553int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1da177e4 1554{
1da177e4 1555 struct inode *inode = dentry->d_inode;
28a6d671 1556 struct task_struct *task;
c69e8d9c 1557 const struct cred *cred;
0499680a 1558 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
c69e8d9c 1559
28a6d671 1560 generic_fillattr(inode, stat);
1da177e4 1561
28a6d671 1562 rcu_read_lock();
dcb0f222
EB		 1563 stat->uid = GLOBAL_ROOT_UID;
1564 stat->gid = GLOBAL_ROOT_GID;
28a6d671
EB		 1565 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1566 if (task) {
0499680a
VK		 1567 if (!has_pid_permissions(pid, task, 2)) {
1568 rcu_read_unlock();
1569 /*
1570 * This doesn't prevent learning whether PID exists,
1571 * it only makes getattr() consistent with readdir().
1572 */
1573 return -ENOENT;
1574 }
28a6d671
EB		 1575 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1576 task_dumpable(task)) {
c69e8d9c
DH		 1577 cred = __task_cred(task);
1578 stat->uid = cred->euid;
1579 stat->gid = cred->egid;
1da177e4
LT		 1580 }
1581 }
28a6d671 1582 rcu_read_unlock();
d6e71144 1583 return 0;
1da177e4
LT		 1584}
1585
1da177e4
LT		 1586/* dentry stuff */
1587
1588/*
1589 * Exceptional case: normally we are not allowed to unhash a busy
1590 * directory. In this case, however, we can do it - no aliasing problems
1591 * due to the way we treat inodes.
1592 *
1593 * Rewrite the inode's ownerships here because the owning task may have
1594 * performed a setuid(), etc.
99f89551
EB		 1595 *
1596 * Before the /proc/pid/status file was created the only way to read
1597 * the effective uid of a /process was to stat /proc/pid. Reading
1598 * /proc/pid/status is slow enough that procps and other packages
1599 * kept stating /proc/pid. To keep the rules in /proc simple I have
1600 * made this apply to all per process world readable and executable
1601 * directories.
1da177e4 1602 */
0b728e19 1603int pid_revalidate(struct dentry *dentry, unsigned int flags)
1da177e4 1604{
34286d66
NP		 1605 struct inode *inode;
1606 struct task_struct *task;
c69e8d9c
DH		 1607 const struct cred *cred;
1608
0b728e19 1609 if (flags & LOOKUP_RCU)
34286d66
NP		 1610 return -ECHILD;
1611
1612 inode = dentry->d_inode;
1613 task = get_proc_task(inode);
1614
99f89551
EB		 1615 if (task) {
1616 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1617 task_dumpable(task)) {
c69e8d9c
DH		 1618 rcu_read_lock();
1619 cred = __task_cred(task);
1620 inode->i_uid = cred->euid;
1621 inode->i_gid = cred->egid;
1622 rcu_read_unlock();
1da177e4 1623 } else {
dcb0f222
EB		 1624 inode->i_uid = GLOBAL_ROOT_UID;
1625 inode->i_gid = GLOBAL_ROOT_GID;
1da177e4 1626 }
9ee8ab9f 1627 inode->i_mode &= ~(S_ISUID | S_ISGID);
1da177e4 1628 security_task_to_inode(task, inode);
99f89551 1629 put_task_struct(task);
1da177e4
LT		 1630 return 1;
1631 }
1632 d_drop(dentry);
1633 return 0;
1634}
1635
6b4e306a 1636const struct dentry_operations pid_dentry_operations =
28a6d671
EB		 1637{
1638 .d_revalidate = pid_revalidate,
1639 .d_delete = pid_delete_dentry,
1640};
1641
1642/* Lookups */
1643
1c0d04c9
EB		 1644/*
1645 * Fill a directory entry.
1646 *
1647 * If possible create the dcache entry and derive our inode number and
1648 * file type from dcache entry.
1649 *
1650 * Since all of the proc inode numbers are dynamically generated, the inode
1651 * numbers do not exist until the inode is cache. This means creating the
1652 * the dcache entry in readdir is necessary to keep the inode numbers
1653 * reported by readdir in sync with the inode numbers reported
1654 * by stat.
1655 */
6b4e306a
EB		 1656int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1657 const char *name, int len,
c5141e6d 1658 instantiate_t instantiate, struct task_struct *task, const void *ptr)
61a28784 1659{
2fddfeef 1660 struct dentry *child, *dir = filp->f_path.dentry;
61a28784
EB		 1661 struct inode *inode;
1662 struct qstr qname;
1663 ino_t ino = 0;
1664 unsigned type = DT_UNKNOWN;
1665
1666 qname.name = name;
1667 qname.len = len;
1668 qname.hash = full_name_hash(name, len);
1669
1670 child = d_lookup(dir, &qname);
1671 if (!child) {
1672 struct dentry *new;
1673 new = d_alloc(dir, &qname);
1674 if (new) {
1675 child = instantiate(dir->d_inode, new, task, ptr);
1676 if (child)
1677 dput(new);
1678 else
1679 child = new;
1680 }
1681 }
1682 if (!child || IS_ERR(child) || !child->d_inode)
1683 goto end_instantiate;
1684 inode = child->d_inode;
1685 if (inode) {
1686 ino = inode->i_ino;
1687 type = inode->i_mode >> 12;
1688 }
1689 dput(child);
1690end_instantiate:
1691 if (!ino)
1692 ino = find_inode_number(dir, &qname);
1693 if (!ino)
1694 ino = 1;
1695 return filldir(dirent, name, len, filp->f_pos, ino, type);
1696}
1697
640708a2
PE		 1698#ifdef CONFIG_CHECKPOINT_RESTORE
1699
1700/*
1701 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1702 * which represent vma start and end addresses.
1703 */
1704static int dname_to_vma_addr(struct dentry *dentry,
1705 unsigned long *start, unsigned long *end)
1706{
1707 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
1708 return -EINVAL;
1709
1710 return 0;
1711}
1712
0b728e19 1713static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
640708a2
PE		 1714{
1715 unsigned long vm_start, vm_end;
1716 bool exact_vma_exists = false;
1717 struct mm_struct *mm = NULL;
1718 struct task_struct *task;
1719 const struct cred *cred;
1720 struct inode *inode;
1721 int status = 0;
1722
0b728e19 1723 if (flags & LOOKUP_RCU)
640708a2
PE		 1724 return -ECHILD;
1725
1726 if (!capable(CAP_SYS_ADMIN)) {
1727 status = -EACCES;
1728 goto out_notask;
1729 }
1730
1731 inode = dentry->d_inode;
1732 task = get_proc_task(inode);
1733 if (!task)
1734 goto out_notask;
1735
2344bec7
CW		 1736 mm = mm_access(task, PTRACE_MODE_READ);
1737 if (IS_ERR_OR_NULL(mm))
640708a2
PE		 1738 goto out;
1739
1740 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1741 down_read(&mm->mmap_sem);
1742 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1743 up_read(&mm->mmap_sem);
1744 }
1745
1746 mmput(mm);
1747
1748 if (exact_vma_exists) {
1749 if (task_dumpable(task)) {
1750 rcu_read_lock();
1751 cred = __task_cred(task);
1752 inode->i_uid = cred->euid;
1753 inode->i_gid = cred->egid;
1754 rcu_read_unlock();
1755 } else {
dcb0f222
EB		 1756 inode->i_uid = GLOBAL_ROOT_UID;
1757 inode->i_gid = GLOBAL_ROOT_GID;
640708a2
PE		 1758 }
1759 security_task_to_inode(task, inode);
1760 status = 1;
1761 }
1762
1763out:
1764 put_task_struct(task);
1765
1766out_notask:
1767 if (status <= 0)
1768 d_drop(dentry);
1769
1770 return status;
1771}
1772
1773static const struct dentry_operations tid_map_files_dentry_operations = {
1774 .d_revalidate = map_files_d_revalidate,
1775 .d_delete = pid_delete_dentry,
1776};
1777
1778static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
1779{
1780 unsigned long vm_start, vm_end;
1781 struct vm_area_struct *vma;
1782 struct task_struct *task;
1783 struct mm_struct *mm;
1784 int rc;
1785
1786 rc = -ENOENT;
1787 task = get_proc_task(dentry->d_inode);
1788 if (!task)
1789 goto out;
1790
1791 mm = get_task_mm(task);
1792 put_task_struct(task);
1793 if (!mm)
1794 goto out;
1795
1796 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
1797 if (rc)
1798 goto out_mmput;
1799
1800 down_read(&mm->mmap_sem);
1801 vma = find_exact_vma(mm, vm_start, vm_end);
1802 if (vma && vma->vm_file) {
1803 *path = vma->vm_file->f_path;
1804 path_get(path);
1805 rc = 0;
1806 }
1807 up_read(&mm->mmap_sem);
1808
1809out_mmput:
1810 mmput(mm);
1811out:
1812 return rc;
1813}
1814
1815struct map_files_info {
7b540d06 1816 fmode_t mode;
640708a2
PE		 1817 unsigned long len;
1818 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
1819};
1820
1821static struct dentry *
1822proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
1823 struct task_struct *task, const void *ptr)
1824{
7b540d06 1825 fmode_t mode = (fmode_t)(unsigned long)ptr;
640708a2
PE		 1826 struct proc_inode *ei;
1827 struct inode *inode;
1828
640708a2
PE		 1829 inode = proc_pid_make_inode(dir->i_sb, task);
1830 if (!inode)
1831 return ERR_PTR(-ENOENT);
1832
1833 ei = PROC_I(inode);
1834 ei->op.proc_get_link = proc_map_files_get_link;
1835
1836 inode->i_op = &proc_pid_link_inode_operations;
1837 inode->i_size = 64;
1838 inode->i_mode = S_IFLNK;
1839
7b540d06 1840 if (mode & FMODE_READ)
640708a2 1841 inode->i_mode |= S_IRUSR;
7b540d06 1842 if (mode & FMODE_WRITE)
640708a2
PE		 1843 inode->i_mode |= S_IWUSR;
1844
1845 d_set_d_op(dentry, &tid_map_files_dentry_operations);
1846 d_add(dentry, inode);
1847
1848 return NULL;
1849}
1850
1851static struct dentry *proc_map_files_lookup(struct inode *dir,
00cd8dd3 1852 struct dentry *dentry, unsigned int flags)
640708a2
PE		 1853{
1854 unsigned long vm_start, vm_end;
1855 struct vm_area_struct *vma;
1856 struct task_struct *task;
1857 struct dentry *result;
1858 struct mm_struct *mm;
1859
1860 result = ERR_PTR(-EACCES);
1861 if (!capable(CAP_SYS_ADMIN))
1862 goto out;
1863
1864 result = ERR_PTR(-ENOENT);
1865 task = get_proc_task(dir);
1866 if (!task)
1867 goto out;
1868
1869 result = ERR_PTR(-EACCES);
eb94cd96 1870 if (!ptrace_may_access(task, PTRACE_MODE_READ))
640708a2
PE		 1871 goto out_put_task;
1872
1873 result = ERR_PTR(-ENOENT);
1874 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
eb94cd96 1875 goto out_put_task;
640708a2
PE		 1876
1877 mm = get_task_mm(task);
1878 if (!mm)
eb94cd96 1879 goto out_put_task;
640708a2
PE		 1880
1881 down_read(&mm->mmap_sem);
1882 vma = find_exact_vma(mm, vm_start, vm_end);
1883 if (!vma)
1884 goto out_no_vma;
1885
7b540d06
AV		 1886 result = proc_map_files_instantiate(dir, dentry, task,
1887 (void *)(unsigned long)vma->vm_file->f_mode);
640708a2
PE		 1888
1889out_no_vma:
1890 up_read(&mm->mmap_sem);
1891 mmput(mm);
640708a2
PE		 1892out_put_task:
1893 put_task_struct(task);
1894out:
1895 return result;
1896}
1897
1898static const struct inode_operations proc_map_files_inode_operations = {
1899 .lookup = proc_map_files_lookup,
1900 .permission = proc_fd_permission,
1901 .setattr = proc_setattr,
1902};
1903
1904static int
1905proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
1906{
1907 struct dentry *dentry = filp->f_path.dentry;
1908 struct inode *inode = dentry->d_inode;
1909 struct vm_area_struct *vma;
1910 struct task_struct *task;
1911 struct mm_struct *mm;
1912 ino_t ino;
1913 int ret;
1914
1915 ret = -EACCES;
1916 if (!capable(CAP_SYS_ADMIN))
1917 goto out;
1918
1919 ret = -ENOENT;
1920 task = get_proc_task(inode);
1921 if (!task)
1922 goto out;
1923
1924 ret = -EACCES;
eb94cd96 1925 if (!ptrace_may_access(task, PTRACE_MODE_READ))
640708a2
PE		 1926 goto out_put_task;
1927
1928 ret = 0;
1929 switch (filp->f_pos) {
1930 case 0:
1931 ino = inode->i_ino;
1932 if (filldir(dirent, ".", 1, 0, ino, DT_DIR) < 0)
eb94cd96 1933 goto out_put_task;
640708a2
PE		 1934 filp->f_pos++;
1935 case 1:
1936 ino = parent_ino(dentry);
1937 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
eb94cd96 1938 goto out_put_task;
640708a2
PE		 1939 filp->f_pos++;
1940 default:
1941 {
1942 unsigned long nr_files, pos, i;
1943 struct flex_array *fa = NULL;
1944 struct map_files_info info;
1945 struct map_files_info *p;
1946
1947 mm = get_task_mm(task);
1948 if (!mm)
eb94cd96 1949 goto out_put_task;
640708a2
PE		 1950 down_read(&mm->mmap_sem);
1951
1952 nr_files = 0;
1953
1954 /*
1955 * We need two passes here:
1956 *
1957 * 1) Collect vmas of mapped files with mmap_sem taken
1958 * 2) Release mmap_sem and instantiate entries
1959 *
1960 * otherwise we get lockdep complained, since filldir()
1961 * routine might require mmap_sem taken in might_fault().
1962 */
1963
1964 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
1965 if (vma->vm_file && ++pos > filp->f_pos)
1966 nr_files++;
1967 }
1968
1969 if (nr_files) {
1970 fa = flex_array_alloc(sizeof(info), nr_files,
1971 GFP_KERNEL);
1972 if (!fa || flex_array_prealloc(fa, 0, nr_files,
1973 GFP_KERNEL)) {
1974 ret = -ENOMEM;
1975 if (fa)
1976 flex_array_free(fa);
1977 up_read(&mm->mmap_sem);
1978 mmput(mm);
eb94cd96 1979 goto out_put_task;
640708a2
PE		 1980 }
1981 for (i = 0, vma = mm->mmap, pos = 2; vma;
1982 vma = vma->vm_next) {
1983 if (!vma->vm_file)
1984 continue;
1985 if (++pos <= filp->f_pos)
1986 continue;
1987
7b540d06 1988 info.mode = vma->vm_file->f_mode;
640708a2
PE		 1989 info.len = snprintf(info.name,
1990 sizeof(info.name), "%lx-%lx",
1991 vma->vm_start, vma->vm_end);
1992 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
1993 BUG();
1994 }
1995 }
1996 up_read(&mm->mmap_sem);
1997
1998 for (i = 0; i < nr_files; i++) {
1999 p = flex_array_get(fa, i);
2000 ret = proc_fill_cache(filp, dirent, filldir,
2001 p->name, p->len,
2002 proc_map_files_instantiate,
7b540d06
AV		 2003 task,
2004 (void *)(unsigned long)p->mode);
640708a2
PE		 2005 if (ret)
2006 break;
2007 filp->f_pos++;
640708a2
PE		 2008 }
2009 if (fa)
2010 flex_array_free(fa);
2011 mmput(mm);
2012 }
2013 }
2014
640708a2
PE		 2015out_put_task:
2016 put_task_struct(task);
2017out:
2018 return ret;
2019}
2020
2021static const struct file_operations proc_map_files_operations = {
2022 .read = generic_read_dir,
2023 .readdir = proc_map_files_readdir,
2024 .llseek = default_llseek,
2025};
2026
2027#endif /* CONFIG_CHECKPOINT_RESTORE */
2028
444ceed8 2029static struct dentry *proc_pident_instantiate(struct inode *dir,
c5141e6d 2030 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8 2031{
c5141e6d 2032 const struct pid_entry *p = ptr;
444ceed8
EB		 2033 struct inode *inode;
2034 struct proc_inode *ei;
bd6daba9 2035 struct dentry *error = ERR_PTR(-ENOENT);
444ceed8 2036
61a28784 2037 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 2038 if (!inode)
2039 goto out;
2040
2041 ei = PROC_I(inode);
2042 inode->i_mode = p->mode;
2043 if (S_ISDIR(inode->i_mode))
bfe86848 2044 set_nlink(inode, 2); /* Use getattr to fix if necessary */
444ceed8
EB		 2045 if (p->iop)
2046 inode->i_op = p->iop;
2047 if (p->fop)
2048 inode->i_fop = p->fop;
2049 ei->op = p->op;
fb045adb 2050 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 2051 d_add(dentry, inode);
2052 /* Close the race of the process dying before we return the dentry */
0b728e19 2053 if (pid_revalidate(dentry, 0))
444ceed8
EB		 2054 error = NULL;
2055out:
2056 return error;
2057}
2058
1da177e4
LT		 2059static struct dentry *proc_pident_lookup(struct inode *dir,
2060 struct dentry *dentry,
c5141e6d 2061 const struct pid_entry *ents,
7bcd6b0e 2062 unsigned int nents)
1da177e4 2063{
cd6a3ce9 2064 struct dentry *error;
99f89551 2065 struct task_struct *task = get_proc_task(dir);
c5141e6d 2066 const struct pid_entry *p, *last;
1da177e4 2067
cd6a3ce9 2068 error = ERR_PTR(-ENOENT);
1da177e4 2069
99f89551
EB		 2070 if (!task)
2071 goto out_no_task;
1da177e4 2072
20cdc894
EB		 2073 /*
2074 * Yes, it does not scale. And it should not. Don't add
2075 * new entries into /proc/<tgid>/ without very good reasons.
2076 */
7bcd6b0e
EB		 2077 last = &ents[nents - 1];
2078 for (p = ents; p <= last; p++) {
1da177e4
LT		 2079 if (p->len != dentry->d_name.len)
2080 continue;
2081 if (!memcmp(dentry->d_name.name, p->name, p->len))
2082 break;
2083 }
7bcd6b0e 2084 if (p > last)
1da177e4
LT		 2085 goto out;
2086
444ceed8 2087 error = proc_pident_instantiate(dir, dentry, task, p);
1da177e4 2088out:
99f89551
EB		 2089 put_task_struct(task);
2090out_no_task:
cd6a3ce9 2091 return error;
1da177e4
LT		 2092}
2093
c5141e6d
ED		 2094static int proc_pident_fill_cache(struct file *filp, void *dirent,
2095 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
61a28784
EB		 2096{
2097 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2098 proc_pident_instantiate, task, p);
2099}
2100
28a6d671
EB		 2101static int proc_pident_readdir(struct file *filp,
2102 void *dirent, filldir_t filldir,
c5141e6d 2103 const struct pid_entry *ents, unsigned int nents)
28a6d671
EB		 2104{
2105 int i;
2fddfeef 2106 struct dentry *dentry = filp->f_path.dentry;
28a6d671
EB		 2107 struct inode *inode = dentry->d_inode;
2108 struct task_struct *task = get_proc_task(inode);
c5141e6d 2109 const struct pid_entry *p, *last;
28a6d671
EB		 2110 ino_t ino;
2111 int ret;
2112
2113 ret = -ENOENT;
2114 if (!task)
61a28784 2115 goto out_no_task;
28a6d671
EB		 2116
2117 ret = 0;
28a6d671
EB		 2118 i = filp->f_pos;
2119 switch (i) {
2120 case 0:
2121 ino = inode->i_ino;
2122 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2123 goto out;
2124 i++;
2125 filp->f_pos++;
2126 /* fall through */
2127 case 1:
2128 ino = parent_ino(dentry);
2129 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2130 goto out;
2131 i++;
2132 filp->f_pos++;
2133 /* fall through */
2134 default:
2135 i -= 2;
2136 if (i >= nents) {
2137 ret = 1;
2138 goto out;
2139 }
2140 p = ents + i;
7bcd6b0e
EB		 2141 last = &ents[nents - 1];
2142 while (p <= last) {
61a28784 2143 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
28a6d671
EB		 2144 goto out;
2145 filp->f_pos++;
2146 p++;
2147 }
2148 }
2149
2150 ret = 1;
2151out:
61a28784
EB		 2152 put_task_struct(task);
2153out_no_task:
28a6d671 2154 return ret;
1da177e4
LT		 2155}
2156
28a6d671
EB		 2157#ifdef CONFIG_SECURITY
2158static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2159 size_t count, loff_t *ppos)
2160{
2fddfeef 2161 struct inode * inode = file->f_path.dentry->d_inode;
04ff9708 2162 char *p = NULL;
28a6d671
EB		 2163 ssize_t length;
2164 struct task_struct *task = get_proc_task(inode);
2165
28a6d671 2166 if (!task)
04ff9708 2167 return -ESRCH;
28a6d671
EB		 2168
2169 length = security_getprocattr(task,
2fddfeef 2170 (char*)file->f_path.dentry->d_name.name,
04ff9708 2171 &p);
28a6d671 2172 put_task_struct(task);
04ff9708
AV		 2173 if (length > 0)
2174 length = simple_read_from_buffer(buf, count, ppos, p, length);
2175 kfree(p);
28a6d671 2176 return length;
1da177e4
LT		 2177}
2178
28a6d671
EB		 2179static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2180 size_t count, loff_t *ppos)
2181{
2fddfeef 2182 struct inode * inode = file->f_path.dentry->d_inode;
28a6d671
EB		 2183 char *page;
2184 ssize_t length;
2185 struct task_struct *task = get_proc_task(inode);
2186
2187 length = -ESRCH;
2188 if (!task)
2189 goto out_no_task;
2190 if (count > PAGE_SIZE)
2191 count = PAGE_SIZE;
2192
2193 /* No partial writes. */
2194 length = -EINVAL;
2195 if (*ppos != 0)
2196 goto out;
2197
2198 length = -ENOMEM;
e12ba74d 2199 page = (char*)__get_free_page(GFP_TEMPORARY);
28a6d671
EB		 2200 if (!page)
2201 goto out;
2202
2203 length = -EFAULT;
2204 if (copy_from_user(page, buf, count))
2205 goto out_free;
2206
107db7c7 2207 /* Guard against adverse ptrace interaction */
9b1bf12d 2208 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
107db7c7
DH		 2209 if (length < 0)
2210 goto out_free;
2211
28a6d671 2212 length = security_setprocattr(task,
2fddfeef 2213 (char*)file->f_path.dentry->d_name.name,
28a6d671 2214 (void*)page, count);
9b1bf12d 2215 mutex_unlock(&task->signal->cred_guard_mutex);
28a6d671
EB		 2216out_free:
2217 free_page((unsigned long) page);
2218out:
2219 put_task_struct(task);
2220out_no_task:
2221 return length;
2222}
2223
00977a59 2224static const struct file_operations proc_pid_attr_operations = {
28a6d671
EB		 2225 .read = proc_pid_attr_read,
2226 .write = proc_pid_attr_write,
87df8424 2227 .llseek = generic_file_llseek,
28a6d671
EB		 2228};
2229
c5141e6d 2230static const struct pid_entry attr_dir_stuff[] = {
631f9c18
AD		 2231 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2232 REG("prev", S_IRUGO, proc_pid_attr_operations),
2233 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2234 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2235 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2236 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
28a6d671
EB		 2237};
2238
72d9dcfc 2239static int proc_attr_dir_readdir(struct file * filp,
28a6d671
EB		 2240 void * dirent, filldir_t filldir)
2241{
2242 return proc_pident_readdir(filp,dirent,filldir,
72d9dcfc 2243 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2244}
2245
00977a59 2246static const struct file_operations proc_attr_dir_operations = {
1da177e4 2247 .read = generic_read_dir,
72d9dcfc 2248 .readdir = proc_attr_dir_readdir,
6038f373 2249 .llseek = default_llseek,
1da177e4
LT		 2250};
2251
72d9dcfc 2252static struct dentry *proc_attr_dir_lookup(struct inode *dir,
00cd8dd3 2253 struct dentry *dentry, unsigned int flags)
28a6d671 2254{
7bcd6b0e
EB		 2255 return proc_pident_lookup(dir, dentry,
2256 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2257}
2258
c5ef1c42 2259static const struct inode_operations proc_attr_dir_inode_operations = {
72d9dcfc 2260 .lookup = proc_attr_dir_lookup,
99f89551 2261 .getattr = pid_getattr,
6d76fa58 2262 .setattr = proc_setattr,
1da177e4
LT		 2263};
2264
28a6d671
EB		 2265#endif
2266
698ba7b5 2267#ifdef CONFIG_ELF_CORE
3cb4a0bb
KH		 2268static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2269 size_t count, loff_t *ppos)
2270{
2271 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2272 struct mm_struct *mm;
2273 char buffer[PROC_NUMBUF];
2274 size_t len;
2275 int ret;
2276
2277 if (!task)
2278 return -ESRCH;
2279
2280 ret = 0;
2281 mm = get_task_mm(task);
2282 if (mm) {
2283 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2284 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2285 MMF_DUMP_FILTER_SHIFT));
2286 mmput(mm);
2287 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2288 }
2289
2290 put_task_struct(task);
2291
2292 return ret;
2293}
2294
2295static ssize_t proc_coredump_filter_write(struct file *file,
2296 const char __user *buf,
2297 size_t count,
2298 loff_t *ppos)
2299{
2300 struct task_struct *task;
2301 struct mm_struct *mm;
2302 char buffer[PROC_NUMBUF], *end;
2303 unsigned int val;
2304 int ret;
2305 int i;
2306 unsigned long mask;
2307
2308 ret = -EFAULT;
2309 memset(buffer, 0, sizeof(buffer));
2310 if (count > sizeof(buffer) - 1)
2311 count = sizeof(buffer) - 1;
2312 if (copy_from_user(buffer, buf, count))
2313 goto out_no_task;
2314
2315 ret = -EINVAL;
2316 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2317 if (*end == '\n')
2318 end++;
2319 if (end - buffer == 0)
2320 goto out_no_task;
2321
2322 ret = -ESRCH;
2323 task = get_proc_task(file->f_dentry->d_inode);
2324 if (!task)
2325 goto out_no_task;
2326
2327 ret = end - buffer;
2328 mm = get_task_mm(task);
2329 if (!mm)
2330 goto out_no_mm;
2331
2332 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2333 if (val & mask)
2334 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2335 else
2336 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2337 }
2338
2339 mmput(mm);
2340 out_no_mm:
2341 put_task_struct(task);
2342 out_no_task:
2343 return ret;
2344}
2345
2346static const struct file_operations proc_coredump_filter_operations = {
2347 .read = proc_coredump_filter_read,
2348 .write = proc_coredump_filter_write,
87df8424 2349 .llseek = generic_file_llseek,
3cb4a0bb
KH		 2350};
2351#endif
2352
28a6d671
EB		 2353/*
2354 * /proc/self:
2355 */
2356static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2357 int buflen)
2358{
488e5bc4 2359 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
b55fcb22 2360 pid_t tgid = task_tgid_nr_ns(current, ns);
28a6d671 2361 char tmp[PROC_NUMBUF];
b55fcb22 2362 if (!tgid)
488e5bc4 2363 return -ENOENT;
b55fcb22 2364 sprintf(tmp, "%d", tgid);
28a6d671
EB		 2365 return vfs_readlink(dentry,buffer,buflen,tmp);
2366}
2367
2368static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2369{
488e5bc4 2370 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
b55fcb22 2371 pid_t tgid = task_tgid_nr_ns(current, ns);
7fee4868
AV		 2372 char *name = ERR_PTR(-ENOENT);
2373 if (tgid) {
2374 name = __getname();
2375 if (!name)
2376 name = ERR_PTR(-ENOMEM);
2377 else
2378 sprintf(name, "%d", tgid);
2379 }
2380 nd_set_link(nd, name);
2381 return NULL;
2382}
2383
2384static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2385 void *cookie)
2386{
2387 char *s = nd_get_link(nd);
2388 if (!IS_ERR(s))
2389 __putname(s);
28a6d671
EB		 2390}
2391
c5ef1c42 2392static const struct inode_operations proc_self_inode_operations = {
28a6d671
EB		 2393 .readlink = proc_self_readlink,
2394 .follow_link = proc_self_follow_link,
7fee4868 2395 .put_link = proc_self_put_link,
28a6d671
EB		 2396};
2397
801199ce
EB		 2398/*
2399 * proc base
2400 *
2401 * These are the directory entries in the root directory of /proc
2402 * that properly belong to the /proc filesystem, as they describe
2403 * describe something that is process related.
2404 */
c5141e6d 2405static const struct pid_entry proc_base_stuff[] = {
61a28784 2406 NOD("self", S_IFLNK|S_IRWXUGO,
801199ce 2407 &proc_self_inode_operations, NULL, {}),
801199ce
EB		 2408};
2409
444ceed8 2410static struct dentry *proc_base_instantiate(struct inode *dir,
c5141e6d 2411 struct dentry *dentry, struct task_struct *task, const void *ptr)
801199ce 2412{
c5141e6d 2413 const struct pid_entry *p = ptr;
801199ce 2414 struct inode *inode;
801199ce 2415 struct proc_inode *ei;
73d36460 2416 struct dentry *error;
801199ce
EB		 2417
2418 /* Allocate the inode */
2419 error = ERR_PTR(-ENOMEM);
2420 inode = new_inode(dir->i_sb);
2421 if (!inode)
2422 goto out;
2423
2424 /* Initialize the inode */
2425 ei = PROC_I(inode);
85fe4025 2426 inode->i_ino = get_next_ino();
801199ce 2427 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
801199ce
EB		 2428
2429 /*
2430 * grab the reference to the task.
2431 */
1a657f78 2432 ei->pid = get_task_pid(task, PIDTYPE_PID);
801199ce
EB		 2433 if (!ei->pid)
2434 goto out_iput;
2435
801199ce
EB		 2436 inode->i_mode = p->mode;
2437 if (S_ISDIR(inode->i_mode))
bfe86848 2438 set_nlink(inode, 2);
801199ce
EB		 2439 if (S_ISLNK(inode->i_mode))
2440 inode->i_size = 64;
2441 if (p->iop)
2442 inode->i_op = p->iop;
2443 if (p->fop)
2444 inode->i_fop = p->fop;
2445 ei->op = p->op;
801199ce
EB		 2446 d_add(dentry, inode);
2447 error = NULL;
2448out:
801199ce
EB		 2449 return error;
2450out_iput:
2451 iput(inode);
2452 goto out;
2453}
2454
444ceed8
EB		 2455static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2456{
2457 struct dentry *error;
2458 struct task_struct *task = get_proc_task(dir);
c5141e6d 2459 const struct pid_entry *p, *last;
444ceed8
EB		 2460
2461 error = ERR_PTR(-ENOENT);
2462
2463 if (!task)
2464 goto out_no_task;
2465
2466 /* Lookup the directory entry */
7bcd6b0e
EB		 2467 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2468 for (p = proc_base_stuff; p <= last; p++) {
444ceed8
EB		 2469 if (p->len != dentry->d_name.len)
2470 continue;
2471 if (!memcmp(dentry->d_name.name, p->name, p->len))
2472 break;
2473 }
7bcd6b0e 2474 if (p > last)
444ceed8
EB		 2475 goto out;
2476
2477 error = proc_base_instantiate(dir, dentry, task, p);
2478
2479out:
2480 put_task_struct(task);
2481out_no_task:
2482 return error;
2483}
2484
c5141e6d
ED		 2485static int proc_base_fill_cache(struct file *filp, void *dirent,
2486 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
61a28784
EB		 2487{
2488 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2489 proc_base_instantiate, task, p);
2490}
2491
aba76fdb 2492#ifdef CONFIG_TASK_IO_ACCOUNTING
297c5d92
AR		 2493static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2494{
940389b8 2495 struct task_io_accounting acct = task->ioac;
5995477a 2496 unsigned long flags;
293eb1e7 2497 int result;
5995477a 2498
293eb1e7
VK		 2499 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2500 if (result)
2501 return result;
2502
2503 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2504 result = -EACCES;
2505 goto out_unlock;
2506 }
1d1221f3 2507
5995477a
AR		 2508 if (whole && lock_task_sighand(task, &flags)) {
2509 struct task_struct *t = task;
2510
2511 task_io_accounting_add(&acct, &task->signal->ioac);
2512 while_each_thread(task, t)
2513 task_io_accounting_add(&acct, &t->ioac);
2514
2515 unlock_task_sighand(task, &flags);
297c5d92 2516 }
293eb1e7 2517 result = sprintf(buffer,
aba76fdb
AM		 2518 "rchar: %llu\n"
2519 "wchar: %llu\n"
2520 "syscr: %llu\n"
2521 "syscw: %llu\n"
2522 "read_bytes: %llu\n"
2523 "write_bytes: %llu\n"
2524 "cancelled_write_bytes: %llu\n",
7c44319d
AB		 2525 (unsigned long long)acct.rchar,
2526 (unsigned long long)acct.wchar,
2527 (unsigned long long)acct.syscr,
2528 (unsigned long long)acct.syscw,
2529 (unsigned long long)acct.read_bytes,
2530 (unsigned long long)acct.write_bytes,
2531 (unsigned long long)acct.cancelled_write_bytes);
293eb1e7
VK		 2532out_unlock:
2533 mutex_unlock(&task->signal->cred_guard_mutex);
2534 return result;
297c5d92
AR		 2535}
2536
2537static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2538{
2539 return do_io_accounting(task, buffer, 0);
aba76fdb 2540}
297c5d92
AR		 2541
2542static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2543{
2544 return do_io_accounting(task, buffer, 1);
2545}
2546#endif /* CONFIG_TASK_IO_ACCOUNTING */
aba76fdb 2547
22d917d8
EB		 2548#ifdef CONFIG_USER_NS
2549static int proc_id_map_open(struct inode *inode, struct file *file,
2550 struct seq_operations *seq_ops)
2551{
2552 struct user_namespace *ns = NULL;
2553 struct task_struct *task;
2554 struct seq_file *seq;
2555 int ret = -EINVAL;
2556
2557 task = get_proc_task(inode);
2558 if (task) {
2559 rcu_read_lock();
2560 ns = get_user_ns(task_cred_xxx(task, user_ns));
2561 rcu_read_unlock();
2562 put_task_struct(task);
2563 }
2564 if (!ns)
2565 goto err;
2566
2567 ret = seq_open(file, seq_ops);
2568 if (ret)
2569 goto err_put_ns;
2570
2571 seq = file->private_data;
2572 seq->private = ns;
2573
2574 return 0;
2575err_put_ns:
2576 put_user_ns(ns);
2577err:
2578 return ret;
2579}
2580
2581static int proc_id_map_release(struct inode *inode, struct file *file)
2582{
2583 struct seq_file *seq = file->private_data;
2584 struct user_namespace *ns = seq->private;
2585 put_user_ns(ns);
2586 return seq_release(inode, file);
2587}
2588
2589static int proc_uid_map_open(struct inode *inode, struct file *file)
2590{
2591 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2592}
2593
2594static int proc_gid_map_open(struct inode *inode, struct file *file)
2595{
2596 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2597}
2598
f76d207a
EB		 2599static int proc_projid_map_open(struct inode *inode, struct file *file)
2600{
2601 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2602}
2603
22d917d8
EB		 2604static const struct file_operations proc_uid_map_operations = {
2605 .open = proc_uid_map_open,
2606 .write = proc_uid_map_write,
2607 .read = seq_read,
2608 .llseek = seq_lseek,
2609 .release = proc_id_map_release,
2610};
2611
2612static const struct file_operations proc_gid_map_operations = {
2613 .open = proc_gid_map_open,
2614 .write = proc_gid_map_write,
2615 .read = seq_read,
2616 .llseek = seq_lseek,
2617 .release = proc_id_map_release,
2618};
f76d207a
EB		 2619
2620static const struct file_operations proc_projid_map_operations = {
2621 .open = proc_projid_map_open,
2622 .write = proc_projid_map_write,
2623 .read = seq_read,
2624 .llseek = seq_lseek,
2625 .release = proc_id_map_release,
2626};
22d917d8
EB		 2627#endif /* CONFIG_USER_NS */
2628
47830723
KC		 2629static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2630 struct pid *pid, struct task_struct *task)
2631{
a9712bc1
AV		 2632 int err = lock_trace(task);
2633 if (!err) {
2634 seq_printf(m, "%08x\n", task->personality);
2635 unlock_trace(task);
2636 }
2637 return err;
47830723
KC		 2638}
2639
28a6d671
EB		 2640/*
2641 * Thread groups
2642 */
00977a59 2643static const struct file_operations proc_task_operations;
c5ef1c42 2644static const struct inode_operations proc_task_inode_operations;
20cdc894 2645
c5141e6d 2646static const struct pid_entry tgid_base_stuff[] = {
631f9c18
AD		 2647 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2648 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
640708a2
PE		 2649#ifdef CONFIG_CHECKPOINT_RESTORE
2650 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
2651#endif
631f9c18 2652 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 2653 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
b2211a36 2654#ifdef CONFIG_NET
631f9c18 2655 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
b2211a36 2656#endif
631f9c18
AD		 2657 REG("environ", S_IRUSR, proc_environ_operations),
2658 INF("auxv", S_IRUSR, proc_pid_auxv),
2659 ONE("status", S_IRUGO, proc_pid_status),
a9712bc1 2660 ONE("personality", S_IRUGO, proc_pid_personality),
3036e7b4 2661 INF("limits", S_IRUGO, proc_pid_limits),
43ae34cb 2662#ifdef CONFIG_SCHED_DEBUG
631f9c18 2663 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
5091faa4
MG		 2664#endif
2665#ifdef CONFIG_SCHED_AUTOGROUP
2666 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
ebcb6734 2667#endif
4614a696 2668 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 2669#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
a9712bc1 2670 INF("syscall", S_IRUGO, proc_pid_syscall),
43ae34cb 2671#endif
631f9c18
AD		 2672 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2673 ONE("stat", S_IRUGO, proc_tgid_stat),
2674 ONE("statm", S_IRUGO, proc_pid_statm),
b7643757 2675 REG("maps", S_IRUGO, proc_pid_maps_operations),
28a6d671 2676#ifdef CONFIG_NUMA
b7643757 2677 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
28a6d671 2678#endif
631f9c18
AD		 2679 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2680 LNK("cwd", proc_cwd_link),
2681 LNK("root", proc_root_link),
2682 LNK("exe", proc_exe_link),
2683 REG("mounts", S_IRUGO, proc_mounts_operations),
2684 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2685 REG("mountstats", S_IRUSR, proc_mountstats_operations),
1e883281 2686#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18 2687 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
b7643757 2688 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
ca6b0bf0 2689 REG("pagemap", S_IRUGO, proc_pagemap_operations),
28a6d671
EB		 2690#endif
2691#ifdef CONFIG_SECURITY
631f9c18 2692 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 2693#endif
2694#ifdef CONFIG_KALLSYMS
631f9c18 2695 INF("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 2696#endif
2ec220e2 2697#ifdef CONFIG_STACKTRACE
a9712bc1 2698 ONE("stack", S_IRUGO, proc_pid_stack),
28a6d671
EB		 2699#endif
2700#ifdef CONFIG_SCHEDSTATS
631f9c18 2701 INF("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 2702#endif
9745512c 2703#ifdef CONFIG_LATENCYTOP
631f9c18 2704 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 2705#endif
8793d854 2706#ifdef CONFIG_PROC_PID_CPUSET
631f9c18 2707 REG("cpuset", S_IRUGO, proc_cpuset_operations),
a424316c
PM		 2708#endif
2709#ifdef CONFIG_CGROUPS
631f9c18 2710 REG("cgroup", S_IRUGO, proc_cgroup_operations),
28a6d671 2711#endif
631f9c18
AD		 2712 INF("oom_score", S_IRUGO, proc_oom_score),
2713 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
a63d83f4 2714 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 2715#ifdef CONFIG_AUDITSYSCALL
631f9c18
AD		 2716 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2717 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 2718#endif
f4f154fd 2719#ifdef CONFIG_FAULT_INJECTION
631f9c18 2720 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 2721#endif
698ba7b5 2722#ifdef CONFIG_ELF_CORE
631f9c18 2723 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3cb4a0bb 2724#endif
aba76fdb 2725#ifdef CONFIG_TASK_IO_ACCOUNTING
1d1221f3 2726 INF("io", S_IRUSR, proc_tgid_io_accounting),
aba76fdb 2727#endif
f133ecca
CM		 2728#ifdef CONFIG_HARDWALL
2729 INF("hardwall", S_IRUGO, proc_pid_hardwall),
2730#endif
22d917d8
EB		 2731#ifdef CONFIG_USER_NS
2732 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
2733 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
f76d207a 2734 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
22d917d8 2735#endif
28a6d671 2736};
1da177e4 2737
28a6d671 2738static int proc_tgid_base_readdir(struct file * filp,
1da177e4
LT		 2739 void * dirent, filldir_t filldir)
2740{
2741 return proc_pident_readdir(filp,dirent,filldir,
28a6d671 2742 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 2743}
2744
00977a59 2745static const struct file_operations proc_tgid_base_operations = {
1da177e4 2746 .read = generic_read_dir,
28a6d671 2747 .readdir = proc_tgid_base_readdir,
6038f373 2748 .llseek = default_llseek,
1da177e4
LT		 2749};
2750
00cd8dd3
AV		 2751static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2752{
7bcd6b0e
EB		 2753 return proc_pident_lookup(dir, dentry,
2754 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 2755}
2756
c5ef1c42 2757static const struct inode_operations proc_tgid_base_inode_operations = {
28a6d671 2758 .lookup = proc_tgid_base_lookup,
99f89551 2759 .getattr = pid_getattr,
6d76fa58 2760 .setattr = proc_setattr,
0499680a 2761 .permission = proc_pid_permission,
1da177e4 2762};
1da177e4 2763
60347f67 2764static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
1da177e4 2765{
48e6484d 2766 struct dentry *dentry, *leader, *dir;
8578cea7 2767 char buf[PROC_NUMBUF];
48e6484d
EB		 2768 struct qstr name;
2769
2770 name.name = buf;
60347f67
PE		 2771 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2772 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d 2773 if (dentry) {
29f12ca3 2774 shrink_dcache_parent(dentry);
48e6484d
EB		 2775 d_drop(dentry);
2776 dput(dentry);
2777 }
1da177e4 2778
48e6484d 2779 name.name = buf;
60347f67
PE		 2780 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2781 leader = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d
EB		 2782 if (!leader)
2783 goto out;
1da177e4 2784
48e6484d
EB		 2785 name.name = "task";
2786 name.len = strlen(name.name);
2787 dir = d_hash_and_lookup(leader, &name);
2788 if (!dir)
2789 goto out_put_leader;
2790
2791 name.name = buf;
60347f67 2792 name.len = snprintf(buf, sizeof(buf), "%d", pid);
48e6484d
EB		 2793 dentry = d_hash_and_lookup(dir, &name);
2794 if (dentry) {
2795 shrink_dcache_parent(dentry);
2796 d_drop(dentry);
2797 dput(dentry);
1da177e4 2798 }
48e6484d
EB		 2799
2800 dput(dir);
2801out_put_leader:
2802 dput(leader);
2803out:
2804 return;
1da177e4
LT		 2805}
2806
0895e91d
RD		 2807/**
2808 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2809 * @task: task that should be flushed.
2810 *
2811 * When flushing dentries from proc, one needs to flush them from global
60347f67 2812 * proc (proc_mnt) and from all the namespaces' procs this task was seen
0895e91d
RD		 2813 * in. This call is supposed to do all of this job.
2814 *
2815 * Looks in the dcache for
2816 * /proc/@pid
2817 * /proc/@tgid/task/@pid
2818 * if either directory is present flushes it and all of it'ts children
2819 * from the dcache.
2820 *
2821 * It is safe and reasonable to cache /proc entries for a task until
2822 * that task exits. After that they just clog up the dcache with
2823 * useless entries, possibly causing useful dcache entries to be
2824 * flushed instead. This routine is proved to flush those useless
2825 * dcache entries at process exit time.
2826 *
2827 * NOTE: This routine is just an optimization so it does not guarantee
2828 * that no dcache entries will exist at process exit time it
2829 * just makes it very unlikely that any will persist.
60347f67
PE		 2830 */
2831
2832void proc_flush_task(struct task_struct *task)
2833{
9fcc2d15 2834 int i;
9b4d1cbe 2835 struct pid *pid, *tgid;
130f77ec
PE		 2836 struct upid *upid;
2837
130f77ec 2838 pid = task_pid(task);
9b4d1cbe 2839 tgid = task_tgid(task);
130f77ec 2840
9fcc2d15 2841 for (i = 0; i <= pid->level; i++) {
130f77ec
PE		 2842 upid = &pid->numbers[i];
2843 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
9b4d1cbe 2844 tgid->numbers[i].nr);
130f77ec 2845 }
6f4e6433
PE		 2846
2847 upid = &pid->numbers[pid->level];
2848 if (upid->nr == 1)
2849 pid_ns_release_proc(upid->ns);
60347f67
PE		 2850}
2851
9711ef99
AB		 2852static struct dentry *proc_pid_instantiate(struct inode *dir,
2853 struct dentry * dentry,
c5141e6d 2854 struct task_struct *task, const void *ptr)
444ceed8
EB		 2855{
2856 struct dentry *error = ERR_PTR(-ENOENT);
2857 struct inode *inode;
2858
61a28784 2859 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 2860 if (!inode)
2861 goto out;
2862
2863 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2864 inode->i_op = &proc_tgid_base_inode_operations;
2865 inode->i_fop = &proc_tgid_base_operations;
2866 inode->i_flags|=S_IMMUTABLE;
aed54175 2867
bfe86848
MS		 2868 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
2869 ARRAY_SIZE(tgid_base_stuff)));
444ceed8 2870
fb045adb 2871 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 2872
2873 d_add(dentry, inode);
2874 /* Close the race of the process dying before we return the dentry */
0b728e19 2875 if (pid_revalidate(dentry, 0))
444ceed8
EB		 2876 error = NULL;
2877out:
2878 return error;
2879}
2880
00cd8dd3 2881struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
1da177e4 2882{
73d36460 2883 struct dentry *result;
1da177e4 2884 struct task_struct *task;
1da177e4 2885 unsigned tgid;
b488893a 2886 struct pid_namespace *ns;
1da177e4 2887
801199ce
EB		 2888 result = proc_base_lookup(dir, dentry);
2889 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2890 goto out;
2891
1da177e4
LT		 2892 tgid = name_to_int(dentry);
2893 if (tgid == ~0U)
2894 goto out;
2895
b488893a 2896 ns = dentry->d_sb->s_fs_info;
de758734 2897 rcu_read_lock();
b488893a 2898 task = find_task_by_pid_ns(tgid, ns);
1da177e4
LT		 2899 if (task)
2900 get_task_struct(task);
de758734 2901 rcu_read_unlock();
1da177e4
LT		 2902 if (!task)
2903 goto out;
2904
444ceed8 2905 result = proc_pid_instantiate(dir, dentry, task, NULL);
1da177e4 2906 put_task_struct(task);
1da177e4 2907out:
cd6a3ce9 2908 return result;
1da177e4
LT		 2909}
2910
1da177e4 2911/*
0804ef4b 2912 * Find the first task with tgid >= tgid
0bc58a91 2913 *
1da177e4 2914 */
19fd4bb2
EB		 2915struct tgid_iter {
2916 unsigned int tgid;
0804ef4b 2917 struct task_struct *task;
19fd4bb2
EB		 2918};
2919static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2920{
0804ef4b 2921 struct pid *pid;
1da177e4 2922
19fd4bb2
EB		 2923 if (iter.task)
2924 put_task_struct(iter.task);
454cc105 2925 rcu_read_lock();
0804ef4b 2926retry:
19fd4bb2
EB		 2927 iter.task = NULL;
2928 pid = find_ge_pid(iter.tgid, ns);
0804ef4b 2929 if (pid) {
19fd4bb2
EB		 2930 iter.tgid = pid_nr_ns(pid, ns);
2931 iter.task = pid_task(pid, PIDTYPE_PID);
0804ef4b
EB		 2932 /* What we to know is if the pid we have find is the
2933 * pid of a thread_group_leader. Testing for task
2934 * being a thread_group_leader is the obvious thing
2935 * todo but there is a window when it fails, due to
2936 * the pid transfer logic in de_thread.
2937 *
2938 * So we perform the straight forward test of seeing
2939 * if the pid we have found is the pid of a thread
2940 * group leader, and don't worry if the task we have
2941 * found doesn't happen to be a thread group leader.
2942 * As we don't care in the case of readdir.
2943 */
19fd4bb2
EB		 2944 if (!iter.task || !has_group_leader_pid(iter.task)) {
2945 iter.tgid += 1;
0804ef4b 2946 goto retry;
19fd4bb2
EB		 2947 }
2948 get_task_struct(iter.task);
0bc58a91 2949 }
454cc105 2950 rcu_read_unlock();
19fd4bb2 2951 return iter;
1da177e4
LT		 2952}
2953
7bcd6b0e 2954#define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
0804ef4b 2955
61a28784 2956static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
19fd4bb2 2957 struct tgid_iter iter)
61a28784
EB		 2958{
2959 char name[PROC_NUMBUF];
19fd4bb2 2960 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
61a28784 2961 return proc_fill_cache(filp, dirent, filldir, name, len,
19fd4bb2 2962 proc_pid_instantiate, iter.task, NULL);
61a28784
EB		 2963}
2964
0499680a
VK		 2965static int fake_filldir(void *buf, const char *name, int namelen,
2966 loff_t offset, u64 ino, unsigned d_type)
2967{
2968 return 0;
2969}
2970
1da177e4
LT		 2971/* for the /proc/ directory itself, after non-process stuff has been done */
2972int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2973{
d8bdc59f
LT		 2974 unsigned int nr;
2975 struct task_struct *reaper;
19fd4bb2 2976 struct tgid_iter iter;
b488893a 2977 struct pid_namespace *ns;
0499680a 2978 filldir_t __filldir;
1da177e4 2979
d8bdc59f
LT		 2980 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
2981 goto out_no_task;
2982 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2983
2984 reaper = get_proc_task(filp->f_path.dentry->d_inode);
61a28784
EB		 2985 if (!reaper)
2986 goto out_no_task;
2987
7bcd6b0e 2988 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
c5141e6d 2989 const struct pid_entry *p = &proc_base_stuff[nr];
61a28784 2990 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
801199ce 2991 goto out;
1da177e4
LT		 2992 }
2993
b488893a 2994 ns = filp->f_dentry->d_sb->s_fs_info;
19fd4bb2
EB		 2995 iter.task = NULL;
2996 iter.tgid = filp->f_pos - TGID_OFFSET;
2997 for (iter = next_tgid(ns, iter);
2998 iter.task;
2999 iter.tgid += 1, iter = next_tgid(ns, iter)) {
0499680a
VK		 3000 if (has_pid_permissions(ns, iter.task, 2))
3001 __filldir = filldir;
3002 else
3003 __filldir = fake_filldir;
3004
19fd4bb2 3005 filp->f_pos = iter.tgid + TGID_OFFSET;
0499680a 3006 if (proc_pid_fill_cache(filp, dirent, __filldir, iter) < 0) {
19fd4bb2 3007 put_task_struct(iter.task);
0804ef4b 3008 goto out;
1da177e4 3009 }
0bc58a91 3010 }
0804ef4b
EB		 3011 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3012out:
61a28784
EB		 3013 put_task_struct(reaper);
3014out_no_task:
0bc58a91
EB		 3015 return 0;
3016}
1da177e4 3017
28a6d671
EB		 3018/*
3019 * Tasks
3020 */
c5141e6d 3021static const struct pid_entry tid_base_stuff[] = {
631f9c18 3022 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3835541d 3023 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 3024 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
631f9c18
AD		 3025 REG("environ", S_IRUSR, proc_environ_operations),
3026 INF("auxv", S_IRUSR, proc_pid_auxv),
3027 ONE("status", S_IRUGO, proc_pid_status),
a9712bc1 3028 ONE("personality", S_IRUGO, proc_pid_personality),
3036e7b4 3029 INF("limits", S_IRUGO, proc_pid_limits),
43ae34cb 3030#ifdef CONFIG_SCHED_DEBUG
631f9c18 3031 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
ebcb6734 3032#endif
4614a696 3033 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 3034#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
a9712bc1 3035 INF("syscall", S_IRUGO, proc_pid_syscall),
43ae34cb 3036#endif
631f9c18
AD		 3037 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3038 ONE("stat", S_IRUGO, proc_tid_stat),
3039 ONE("statm", S_IRUGO, proc_pid_statm),
b7643757 3040 REG("maps", S_IRUGO, proc_tid_maps_operations),
81841161
CG		 3041#ifdef CONFIG_CHECKPOINT_RESTORE
3042 REG("children", S_IRUGO, proc_tid_children_operations),
3043#endif
28a6d671 3044#ifdef CONFIG_NUMA
b7643757 3045 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
28a6d671 3046#endif
631f9c18
AD		 3047 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3048 LNK("cwd", proc_cwd_link),
3049 LNK("root", proc_root_link),
3050 LNK("exe", proc_exe_link),
3051 REG("mounts", S_IRUGO, proc_mounts_operations),
3052 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
1e883281 3053#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18 3054 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
b7643757 3055 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
ca6b0bf0 3056 REG("pagemap", S_IRUGO, proc_pagemap_operations),
28a6d671
EB		 3057#endif
3058#ifdef CONFIG_SECURITY
631f9c18 3059 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 3060#endif
3061#ifdef CONFIG_KALLSYMS
631f9c18 3062 INF("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 3063#endif
2ec220e2 3064#ifdef CONFIG_STACKTRACE
a9712bc1 3065 ONE("stack", S_IRUGO, proc_pid_stack),
28a6d671
EB		 3066#endif
3067#ifdef CONFIG_SCHEDSTATS
631f9c18 3068 INF("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 3069#endif
9745512c 3070#ifdef CONFIG_LATENCYTOP
631f9c18 3071 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 3072#endif
8793d854 3073#ifdef CONFIG_PROC_PID_CPUSET
631f9c18 3074 REG("cpuset", S_IRUGO, proc_cpuset_operations),
a424316c
PM		 3075#endif
3076#ifdef CONFIG_CGROUPS
631f9c18 3077 REG("cgroup", S_IRUGO, proc_cgroup_operations),
28a6d671 3078#endif
631f9c18
AD		 3079 INF("oom_score", S_IRUGO, proc_oom_score),
3080 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
a63d83f4 3081 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 3082#ifdef CONFIG_AUDITSYSCALL
631f9c18 3083 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
26ec3c64 3084 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 3085#endif
f4f154fd 3086#ifdef CONFIG_FAULT_INJECTION
631f9c18 3087 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 3088#endif
297c5d92 3089#ifdef CONFIG_TASK_IO_ACCOUNTING
1d1221f3 3090 INF("io", S_IRUSR, proc_tid_io_accounting),
297c5d92 3091#endif
f133ecca
CM		 3092#ifdef CONFIG_HARDWALL
3093 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3094#endif
22d917d8
EB		 3095#ifdef CONFIG_USER_NS
3096 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3097 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
f76d207a 3098 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
22d917d8 3099#endif
28a6d671
EB		 3100};
3101
3102static int proc_tid_base_readdir(struct file * filp,
3103 void * dirent, filldir_t filldir)
3104{
3105 return proc_pident_readdir(filp,dirent,filldir,
3106 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3107}
3108
00cd8dd3
AV		 3109static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3110{
7bcd6b0e
EB		 3111 return proc_pident_lookup(dir, dentry,
3112 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
28a6d671
EB		 3113}
3114
00977a59 3115static const struct file_operations proc_tid_base_operations = {
28a6d671
EB		 3116 .read = generic_read_dir,
3117 .readdir = proc_tid_base_readdir,
6038f373 3118 .llseek = default_llseek,
28a6d671
EB		 3119};
3120
c5ef1c42 3121static const struct inode_operations proc_tid_base_inode_operations = {
28a6d671
EB		 3122 .lookup = proc_tid_base_lookup,
3123 .getattr = pid_getattr,
3124 .setattr = proc_setattr,
3125};
3126
444ceed8 3127static struct dentry *proc_task_instantiate(struct inode *dir,
c5141e6d 3128 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8
EB		 3129{
3130 struct dentry *error = ERR_PTR(-ENOENT);
3131 struct inode *inode;
61a28784 3132 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 3133
3134 if (!inode)
3135 goto out;
3136 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3137 inode->i_op = &proc_tid_base_inode_operations;
3138 inode->i_fop = &proc_tid_base_operations;
3139 inode->i_flags|=S_IMMUTABLE;
aed54175 3140
bfe86848
MS		 3141 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3142 ARRAY_SIZE(tid_base_stuff)));
444ceed8 3143
fb045adb 3144 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB		 3145
3146 d_add(dentry, inode);
3147 /* Close the race of the process dying before we return the dentry */
0b728e19 3148 if (pid_revalidate(dentry, 0))
444ceed8
EB		 3149 error = NULL;
3150out:
3151 return error;
3152}
3153
00cd8dd3 3154static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
28a6d671
EB		 3155{
3156 struct dentry *result = ERR_PTR(-ENOENT);
3157 struct task_struct *task;
3158 struct task_struct *leader = get_proc_task(dir);
28a6d671 3159 unsigned tid;
b488893a 3160 struct pid_namespace *ns;
28a6d671
EB		 3161
3162 if (!leader)
3163 goto out_no_task;
3164
3165 tid = name_to_int(dentry);
3166 if (tid == ~0U)
3167 goto out;
3168
b488893a 3169 ns = dentry->d_sb->s_fs_info;
28a6d671 3170 rcu_read_lock();
b488893a 3171 task = find_task_by_pid_ns(tid, ns);
28a6d671
EB		 3172 if (task)
3173 get_task_struct(task);
3174 rcu_read_unlock();
3175 if (!task)
3176 goto out;
bac0abd6 3177 if (!same_thread_group(leader, task))
28a6d671
EB		 3178 goto out_drop_task;
3179
444ceed8 3180 result = proc_task_instantiate(dir, dentry, task, NULL);
28a6d671
EB		 3181out_drop_task:
3182 put_task_struct(task);
3183out:
3184 put_task_struct(leader);
3185out_no_task:
3186 return result;
3187}
3188
0bc58a91
EB		 3189/*
3190 * Find the first tid of a thread group to return to user space.
3191 *
3192 * Usually this is just the thread group leader, but if the users
3193 * buffer was too small or there was a seek into the middle of the
3194 * directory we have more work todo.
3195 *
3196 * In the case of a short read we start with find_task_by_pid.
3197 *
3198 * In the case of a seek we start with the leader and walk nr
3199 * threads past it.
3200 */
cc288738 3201static struct task_struct *first_tid(struct task_struct *leader,
b488893a 3202 int tid, int nr, struct pid_namespace *ns)
0bc58a91 3203{
a872ff0c 3204 struct task_struct *pos;
1da177e4 3205
cc288738 3206 rcu_read_lock();
0bc58a91
EB		 3207 /* Attempt to start with the pid of a thread */
3208 if (tid && (nr > 0)) {
b488893a 3209 pos = find_task_by_pid_ns(tid, ns);
a872ff0c
ON		 3210 if (pos && (pos->group_leader == leader))
3211 goto found;
0bc58a91 3212 }
1da177e4 3213
0bc58a91 3214 /* If nr exceeds the number of threads there is nothing todo */
a872ff0c
ON		 3215 pos = NULL;
3216 if (nr && nr >= get_nr_threads(leader))
3217 goto out;
1da177e4 3218
a872ff0c
ON		 3219 /* If we haven't found our starting place yet start
3220 * with the leader and walk nr threads forward.
0bc58a91 3221 */
a872ff0c
ON		 3222 for (pos = leader; nr > 0; --nr) {
3223 pos = next_thread(pos);
3224 if (pos == leader) {
3225 pos = NULL;
3226 goto out;
3227 }
1da177e4 3228 }
a872ff0c
ON		 3229found:
3230 get_task_struct(pos);
3231out:
cc288738 3232 rcu_read_unlock();
0bc58a91
EB		 3233 return pos;
3234}
3235
3236/*
3237 * Find the next thread in the thread list.
3238 * Return NULL if there is an error or no next thread.
3239 *
3240 * The reference to the input task_struct is released.
3241 */
3242static struct task_struct *next_tid(struct task_struct *start)
3243{
c1df7fb8 3244 struct task_struct *pos = NULL;
cc288738 3245 rcu_read_lock();
c1df7fb8 3246 if (pid_alive(start)) {
0bc58a91 3247 pos = next_thread(start);
c1df7fb8
ON		 3248 if (thread_group_leader(pos))
3249 pos = NULL;
3250 else
3251 get_task_struct(pos);
3252 }
cc288738 3253 rcu_read_unlock();
0bc58a91
EB		 3254 put_task_struct(start);
3255 return pos;
1da177e4
LT		 3256}
3257
61a28784
EB		 3258static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3259 struct task_struct *task, int tid)
3260{
3261 char name[PROC_NUMBUF];
3262 int len = snprintf(name, sizeof(name), "%d", tid);
3263 return proc_fill_cache(filp, dirent, filldir, name, len,
3264 proc_task_instantiate, task, NULL);
3265}
3266
1da177e4
LT		 3267/* for the /proc/TGID/task/ directories */
3268static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3269{
2fddfeef 3270 struct dentry *dentry = filp->f_path.dentry;
1da177e4 3271 struct inode *inode = dentry->d_inode;
7d895244 3272 struct task_struct *leader = NULL;
0bc58a91 3273 struct task_struct *task;
1da177e4
LT		 3274 int retval = -ENOENT;
3275 ino_t ino;
0bc58a91 3276 int tid;
b488893a 3277 struct pid_namespace *ns;
1da177e4 3278
7d895244
GC		 3279 task = get_proc_task(inode);
3280 if (!task)
3281 goto out_no_task;
3282 rcu_read_lock();
3283 if (pid_alive(task)) {
3284 leader = task->group_leader;
3285 get_task_struct(leader);
3286 }
3287 rcu_read_unlock();
3288 put_task_struct(task);
99f89551
EB		 3289 if (!leader)
3290 goto out_no_task;
1da177e4
LT		 3291 retval = 0;
3292
ee568b25 3293 switch ((unsigned long)filp->f_pos) {
1da177e4
LT		 3294 case 0:
3295 ino = inode->i_ino;
ee6f779b 3296 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
1da177e4 3297 goto out;
ee6f779b 3298 filp->f_pos++;
1da177e4
LT		 3299 /* fall through */
3300 case 1:
3301 ino = parent_ino(dentry);
ee6f779b 3302 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
1da177e4 3303 goto out;
ee6f779b 3304 filp->f_pos++;
1da177e4
LT		 3305 /* fall through */
3306 }
3307
0bc58a91
EB		 3308 /* f_version caches the tgid value that the last readdir call couldn't
3309 * return. lseek aka telldir automagically resets f_version to 0.
3310 */
b488893a 3311 ns = filp->f_dentry->d_sb->s_fs_info;
2b47c361 3312 tid = (int)filp->f_version;
0bc58a91 3313 filp->f_version = 0;
ee6f779b 3314 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
0bc58a91 3315 task;
ee6f779b 3316 task = next_tid(task), filp->f_pos++) {
b488893a 3317 tid = task_pid_nr_ns(task, ns);
61a28784 3318 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
0bc58a91
EB		 3319 /* returning this tgid failed, save it as the first
3320 * pid for the next readir call */
2b47c361 3321 filp->f_version = (u64)tid;
0bc58a91 3322 put_task_struct(task);
1da177e4 3323 break;
0bc58a91 3324 }
1da177e4
LT		 3325 }
3326out:
99f89551
EB		 3327 put_task_struct(leader);
3328out_no_task:
1da177e4
LT		 3329 return retval;
3330}
6e66b52b
EB		 3331
3332static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3333{
3334 struct inode *inode = dentry->d_inode;
99f89551 3335 struct task_struct *p = get_proc_task(inode);
6e66b52b
EB		 3336 generic_fillattr(inode, stat);
3337
99f89551 3338 if (p) {
99f89551 3339 stat->nlink += get_nr_threads(p);
99f89551 3340 put_task_struct(p);
6e66b52b
EB		 3341 }
3342
3343 return 0;
3344}
28a6d671 3345
c5ef1c42 3346static const struct inode_operations proc_task_inode_operations = {
28a6d671
EB		 3347 .lookup = proc_task_lookup,
3348 .getattr = proc_task_getattr,
3349 .setattr = proc_setattr,
0499680a 3350 .permission = proc_pid_permission,
28a6d671
EB		 3351};
3352
00977a59 3353static const struct file_operations proc_task_operations = {
28a6d671
EB		 3354 .read = generic_read_dir,
3355 .readdir = proc_task_readdir,
6038f373 3356 .llseek = default_llseek,
28a6d671 3357};
Linux kernel - Deliverables
This page took 1.083375 seconds and 5 git commands to generate.