projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
sched: Add 'autogroup' scheduling feature: automated per session task groups
[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>
5ad4e53b 84#include <linux/fs_struct.h>
5a0e3ad6 85#include <linux/slab.h>
1da177e4
LT		 86#include "internal.h"
87
0f2fe20f
EB		 88/* NOTE:
89 * Implementing inode permission operations in /proc is almost
90 * certainly an error. Permission checks need to happen during
91 * each system call not at open time. The reason is that most of
92 * what we wish to check for permissions in /proc varies at runtime.
93 *
94 * The classic example of a problem is opening file descriptors
95 * in /proc for a task before it execs a suid executable.
96 */
97
1da177e4 98struct pid_entry {
1da177e4 99 char *name;
c5141e6d 100 int len;
1da177e4 101 mode_t mode;
c5ef1c42 102 const struct inode_operations *iop;
00977a59 103 const struct file_operations *fop;
20cdc894 104 union proc_op op;
1da177e4
LT		 105};
106
61a28784 107#define NOD(NAME, MODE, IOP, FOP, OP) { \
20cdc894 108 .name = (NAME), \
c5141e6d 109 .len = sizeof(NAME) - 1, \
20cdc894
EB		 110 .mode = MODE, \
111 .iop = IOP, \
112 .fop = FOP, \
113 .op = OP, \
114}
115
631f9c18
AD		 116#define DIR(NAME, MODE, iops, fops) \
117 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
118#define LNK(NAME, get_link) \
61a28784 119 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
20cdc894 120 &proc_pid_link_inode_operations, NULL, \
631f9c18
AD		 121 { .proc_get_link = get_link } )
122#define REG(NAME, MODE, fops) \
123 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
124#define INF(NAME, MODE, read) \
61a28784 125 NOD(NAME, (S_IFREG|(MODE)), \
20cdc894 126 NULL, &proc_info_file_operations, \
631f9c18
AD		 127 { .proc_read = read } )
128#define ONE(NAME, MODE, show) \
be614086
EB		 129 NOD(NAME, (S_IFREG|(MODE)), \
130 NULL, &proc_single_file_operations, \
631f9c18 131 { .proc_show = show } )
1da177e4 132
aed54175
VN		 133/*
134 * Count the number of hardlinks for the pid_entry table, excluding the .
135 * and .. links.
136 */
137static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
138 unsigned int n)
139{
140 unsigned int i;
141 unsigned int count;
142
143 count = 0;
144 for (i = 0; i < n; ++i) {
145 if (S_ISDIR(entries[i].mode))
146 ++count;
147 }
148
149 return count;
150}
151
f7ad3c6b 152static int get_task_root(struct task_struct *task, struct path *root)
1da177e4 153{
7c2c7d99
HD		 154 int result = -ENOENT;
155
0494f6ec 156 task_lock(task);
f7ad3c6b
MS		 157 if (task->fs) {
158 get_fs_root(task->fs, root);
7c2c7d99
HD		 159 result = 0;
160 }
0494f6ec 161 task_unlock(task);
7c2c7d99 162 return result;
0494f6ec
MS		 163}
164
3dcd25f3 165static int proc_cwd_link(struct inode *inode, struct path *path)
0494f6ec 166{
99f89551 167 struct task_struct *task = get_proc_task(inode);
0494f6ec 168 int result = -ENOENT;
99f89551
EB		 169
170 if (task) {
f7ad3c6b
MS		 171 task_lock(task);
172 if (task->fs) {
173 get_fs_pwd(task->fs, path);
174 result = 0;
175 }
176 task_unlock(task);
99f89551
EB		 177 put_task_struct(task);
178 }
1da177e4
LT		 179 return result;
180}
181
3dcd25f3 182static int proc_root_link(struct inode *inode, struct path *path)
1da177e4 183{
99f89551 184 struct task_struct *task = get_proc_task(inode);
1da177e4 185 int result = -ENOENT;
99f89551
EB		 186
187 if (task) {
f7ad3c6b 188 result = get_task_root(task, path);
99f89551
EB		 189 put_task_struct(task);
190 }
1da177e4
LT		 191 return result;
192}
193
638fa202
RM		 194/*
195 * Return zero if current may access user memory in @task, -error if not.
196 */
197static int check_mem_permission(struct task_struct *task)
198{
199 /*
200 * A task can always look at itself, in case it chooses
201 * to use system calls instead of load instructions.
202 */
203 if (task == current)
204 return 0;
205
206 /*
207 * If current is actively ptrace'ing, and would also be
208 * permitted to freshly attach with ptrace now, permit it.
209 */
0d094efe
RM		 210 if (task_is_stopped_or_traced(task)) {
211 int match;
212 rcu_read_lock();
213 match = (tracehook_tracer_task(task) == current);
214 rcu_read_unlock();
215 if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH))
216 return 0;
217 }
638fa202
RM		 218
219 /*
220 * Noone else is allowed.
221 */
222 return -EPERM;
223}
1da177e4 224
831830b5
AV		 225struct mm_struct *mm_for_maps(struct task_struct *task)
226{
704b836c
ON		 227 struct mm_struct *mm;
228
9b1bf12d 229 if (mutex_lock_killable(&task->signal->cred_guard_mutex))
704b836c 230 return NULL;
00f89d21 231
704b836c
ON		 232 mm = get_task_mm(task);
233 if (mm && mm != current->mm &&
234 !ptrace_may_access(task, PTRACE_MODE_READ)) {
235 mmput(mm);
236 mm = NULL;
13f0feaf 237 }
9b1bf12d 238 mutex_unlock(&task->signal->cred_guard_mutex);
704b836c 239
831830b5 240 return mm;
831830b5
AV		 241}
242
1da177e4
LT		 243static int proc_pid_cmdline(struct task_struct *task, char * buffer)
244{
245 int res = 0;
246 unsigned int len;
247 struct mm_struct *mm = get_task_mm(task);
248 if (!mm)
249 goto out;
250 if (!mm->arg_end)
251 goto out_mm; /* Shh! No looking before we're done */
252
253 len = mm->arg_end - mm->arg_start;
254
255 if (len > PAGE_SIZE)
256 len = PAGE_SIZE;
257
258 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
259
260 // If the nul at the end of args has been overwritten, then
261 // assume application is using setproctitle(3).
262 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
263 len = strnlen(buffer, res);
264 if (len < res) {
265 res = len;
266 } else {
267 len = mm->env_end - mm->env_start;
268 if (len > PAGE_SIZE - res)
269 len = PAGE_SIZE - res;
270 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
271 res = strnlen(buffer, res);
272 }
273 }
274out_mm:
275 mmput(mm);
276out:
277 return res;
278}
279
280static int proc_pid_auxv(struct task_struct *task, char *buffer)
281{
282 int res = 0;
283 struct mm_struct *mm = get_task_mm(task);
284 if (mm) {
285 unsigned int nwords = 0;
dfe6b7d9 286 do {
1da177e4 287 nwords += 2;
dfe6b7d9 288 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
1da177e4
LT		 289 res = nwords * sizeof(mm->saved_auxv[0]);
290 if (res > PAGE_SIZE)
291 res = PAGE_SIZE;
292 memcpy(buffer, mm->saved_auxv, res);
293 mmput(mm);
294 }
295 return res;
296}
297
298
299#ifdef CONFIG_KALLSYMS
300/*
301 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
302 * Returns the resolved symbol. If that fails, simply return the address.
303 */
304static int proc_pid_wchan(struct task_struct *task, char *buffer)
305{
ffb45122 306 unsigned long wchan;
9281acea 307 char symname[KSYM_NAME_LEN];
1da177e4
LT		 308
309 wchan = get_wchan(task);
310
9d65cb4a 311 if (lookup_symbol_name(wchan, symname) < 0)
f83ce3e6
JE		 312 if (!ptrace_may_access(task, PTRACE_MODE_READ))
313 return 0;
314 else
315 return sprintf(buffer, "%lu", wchan);
9d65cb4a
AD		 316 else
317 return sprintf(buffer, "%s", symname);
1da177e4
LT		 318}
319#endif /* CONFIG_KALLSYMS */
320
2ec220e2
KC		 321#ifdef CONFIG_STACKTRACE
322
323#define MAX_STACK_TRACE_DEPTH 64
324
325static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
326 struct pid *pid, struct task_struct *task)
327{
328 struct stack_trace trace;
329 unsigned long *entries;
330 int i;
331
332 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
333 if (!entries)
334 return -ENOMEM;
335
336 trace.nr_entries = 0;
337 trace.max_entries = MAX_STACK_TRACE_DEPTH;
338 trace.entries = entries;
339 trace.skip = 0;
340 save_stack_trace_tsk(task, &trace);
341
342 for (i = 0; i < trace.nr_entries; i++) {
343 seq_printf(m, "[<%p>] %pS\n",
344 (void *)entries[i], (void *)entries[i]);
345 }
346 kfree(entries);
347
348 return 0;
349}
350#endif
351
1da177e4
LT		 352#ifdef CONFIG_SCHEDSTATS
353/*
354 * Provides /proc/PID/schedstat
355 */
356static int proc_pid_schedstat(struct task_struct *task, char *buffer)
357{
172ba844 358 return sprintf(buffer, "%llu %llu %lu\n",
826e08b0
IM		 359 (unsigned long long)task->se.sum_exec_runtime,
360 (unsigned long long)task->sched_info.run_delay,
2d72376b 361 task->sched_info.pcount);
1da177e4
LT		 362}
363#endif
364
9745512c
AV		 365#ifdef CONFIG_LATENCYTOP
366static int lstats_show_proc(struct seq_file *m, void *v)
367{
368 int i;
13d77c37
HS		 369 struct inode *inode = m->private;
370 struct task_struct *task = get_proc_task(inode);
9745512c 371
13d77c37
HS		 372 if (!task)
373 return -ESRCH;
374 seq_puts(m, "Latency Top version : v0.1\n");
9745512c
AV		 375 for (i = 0; i < 32; i++) {
376 if (task->latency_record[i].backtrace[0]) {
377 int q;
378 seq_printf(m, "%i %li %li ",
379 task->latency_record[i].count,
380 task->latency_record[i].time,
381 task->latency_record[i].max);
382 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
9c246247 383 char sym[KSYM_SYMBOL_LEN];
9745512c
AV		 384 char *c;
385 if (!task->latency_record[i].backtrace[q])
386 break;
387 if (task->latency_record[i].backtrace[q] == ULONG_MAX)
388 break;
389 sprint_symbol(sym, task->latency_record[i].backtrace[q]);
390 c = strchr(sym, '+');
391 if (c)
392 *c = 0;
393 seq_printf(m, "%s ", sym);
394 }
395 seq_printf(m, "\n");
396 }
397
398 }
13d77c37 399 put_task_struct(task);
9745512c
AV		 400 return 0;
401}
402
403static int lstats_open(struct inode *inode, struct file *file)
404{
13d77c37 405 return single_open(file, lstats_show_proc, inode);
d6643d12
HS		 406}
407
9745512c
AV		 408static ssize_t lstats_write(struct file *file, const char __user *buf,
409 size_t count, loff_t *offs)
410{
13d77c37 411 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
9745512c 412
13d77c37
HS		 413 if (!task)
414 return -ESRCH;
9745512c 415 clear_all_latency_tracing(task);
13d77c37 416 put_task_struct(task);
9745512c
AV		 417
418 return count;
419}
420
421static const struct file_operations proc_lstats_operations = {
422 .open = lstats_open,
423 .read = seq_read,
424 .write = lstats_write,
425 .llseek = seq_lseek,
13d77c37 426 .release = single_release,
9745512c
AV		 427};
428
429#endif
430
1da177e4
LT		 431static int proc_oom_score(struct task_struct *task, char *buffer)
432{
b95c35e7 433 unsigned long points = 0;
1da177e4 434
19c5d45a 435 read_lock(&tasklist_lock);
b95c35e7 436 if (pid_alive(task))
a63d83f4
DR		 437 points = oom_badness(task, NULL, NULL,
438 totalram_pages + total_swap_pages);
19c5d45a 439 read_unlock(&tasklist_lock);
1da177e4
LT		 440 return sprintf(buffer, "%lu\n", points);
441}
442
d85f50d5
NH		 443struct limit_names {
444 char *name;
445 char *unit;
446};
447
448static const struct limit_names lnames[RLIM_NLIMITS] = {
cff4edb5 449 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
d85f50d5
NH		 450 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
451 [RLIMIT_DATA] = {"Max data size", "bytes"},
452 [RLIMIT_STACK] = {"Max stack size", "bytes"},
453 [RLIMIT_CORE] = {"Max core file size", "bytes"},
454 [RLIMIT_RSS] = {"Max resident set", "bytes"},
455 [RLIMIT_NPROC] = {"Max processes", "processes"},
456 [RLIMIT_NOFILE] = {"Max open files", "files"},
457 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
458 [RLIMIT_AS] = {"Max address space", "bytes"},
459 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
460 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
461 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
462 [RLIMIT_NICE] = {"Max nice priority", NULL},
463 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
8808117c 464 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
d85f50d5
NH		 465};
466
467/* Display limits for a process */
468static int proc_pid_limits(struct task_struct *task, char *buffer)
469{
470 unsigned int i;
471 int count = 0;
472 unsigned long flags;
473 char *bufptr = buffer;
474
475 struct rlimit rlim[RLIM_NLIMITS];
476
a6bebbc8 477 if (!lock_task_sighand(task, &flags))
d85f50d5 478 return 0;
d85f50d5
NH		 479 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
480 unlock_task_sighand(task, &flags);
d85f50d5
NH		 481
482 /*
483 * print the file header
484 */
485 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
486 "Limit", "Soft Limit", "Hard Limit", "Units");
487
488 for (i = 0; i < RLIM_NLIMITS; i++) {
489 if (rlim[i].rlim_cur == RLIM_INFINITY)
490 count += sprintf(&bufptr[count], "%-25s %-20s ",
491 lnames[i].name, "unlimited");
492 else
493 count += sprintf(&bufptr[count], "%-25s %-20lu ",
494 lnames[i].name, rlim[i].rlim_cur);
495
496 if (rlim[i].rlim_max == RLIM_INFINITY)
497 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
498 else
499 count += sprintf(&bufptr[count], "%-20lu ",
500 rlim[i].rlim_max);
501
502 if (lnames[i].unit)
503 count += sprintf(&bufptr[count], "%-10s\n",
504 lnames[i].unit);
505 else
506 count += sprintf(&bufptr[count], "\n");
507 }
508
509 return count;
510}
511
ebcb6734
RM		 512#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
513static int proc_pid_syscall(struct task_struct *task, char *buffer)
514{
515 long nr;
516 unsigned long args[6], sp, pc;
517
518 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
519 return sprintf(buffer, "running\n");
520
521 if (nr < 0)
522 return sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
523
524 return sprintf(buffer,
525 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
526 nr,
527 args[0], args[1], args[2], args[3], args[4], args[5],
528 sp, pc);
529}
530#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
531
1da177e4
LT		 532/************************************************************************/
533/* Here the fs part begins */
534/************************************************************************/
535
536/* permission checks */
778c1144 537static int proc_fd_access_allowed(struct inode *inode)
1da177e4 538{
778c1144
EB		 539 struct task_struct *task;
540 int allowed = 0;
df26c40e
EB		 541 /* Allow access to a task's file descriptors if it is us or we
542 * may use ptrace attach to the process and find out that
543 * information.
778c1144
EB		 544 */
545 task = get_proc_task(inode);
df26c40e 546 if (task) {
006ebb40 547 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
778c1144 548 put_task_struct(task);
df26c40e 549 }
778c1144 550 return allowed;
1da177e4
LT		 551}
552
6d76fa58
LT		 553static int proc_setattr(struct dentry *dentry, struct iattr *attr)
554{
555 int error;
556 struct inode *inode = dentry->d_inode;
557
558 if (attr->ia_valid & ATTR_MODE)
559 return -EPERM;
560
561 error = inode_change_ok(inode, attr);
1025774c
CH		 562 if (error)
563 return error;
564
565 if ((attr->ia_valid & ATTR_SIZE) &&
566 attr->ia_size != i_size_read(inode)) {
567 error = vmtruncate(inode, attr->ia_size);
568 if (error)
569 return error;
570 }
571
572 setattr_copy(inode, attr);
573 mark_inode_dirty(inode);
574 return 0;
6d76fa58
LT		 575}
576
c5ef1c42 577static const struct inode_operations proc_def_inode_operations = {
6d76fa58
LT		 578 .setattr = proc_setattr,
579};
580
a1a2c409
MS		 581static int mounts_open_common(struct inode *inode, struct file *file,
582 const struct seq_operations *op)
1da177e4 583{
99f89551 584 struct task_struct *task = get_proc_task(inode);
cf7b708c 585 struct nsproxy *nsp;
6b3286ed 586 struct mnt_namespace *ns = NULL;
a1a2c409 587 struct path root;
5addc5dd
AV		 588 struct proc_mounts *p;
589 int ret = -EINVAL;
1da177e4 590
99f89551 591 if (task) {
cf7b708c
PE		 592 rcu_read_lock();
593 nsp = task_nsproxy(task);
594 if (nsp) {
595 ns = nsp->mnt_ns;
863c4702
AD		 596 if (ns)
597 get_mnt_ns(ns);
598 }
cf7b708c 599 rcu_read_unlock();
f7ad3c6b 600 if (ns && get_task_root(task, &root) == 0)
7c2c7d99 601 ret = 0;
99f89551
EB		 602 put_task_struct(task);
603 }
5addc5dd 604
a1a2c409
MS		 605 if (!ns)
606 goto err;
7c2c7d99 607 if (ret)
a1a2c409
MS		 608 goto err_put_ns;
609
a1a2c409
MS		 610 ret = -ENOMEM;
611 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
612 if (!p)
613 goto err_put_path;
614
615 file->private_data = &p->m;
616 ret = seq_open(file, op);
617 if (ret)
618 goto err_free;
619
620 p->m.private = p;
621 p->ns = ns;
622 p->root = root;
623 p->event = ns->event;
624
625 return 0;
626
627 err_free:
628 kfree(p);
629 err_put_path:
630 path_put(&root);
631 err_put_ns:
632 put_mnt_ns(ns);
633 err:
1da177e4
LT		 634 return ret;
635}
636
637static int mounts_release(struct inode *inode, struct file *file)
638{
a1a2c409
MS		 639 struct proc_mounts *p = file->private_data;
640 path_put(&p->root);
641 put_mnt_ns(p->ns);
1da177e4
LT		 642 return seq_release(inode, file);
643}
644
5addc5dd
AV		 645static unsigned mounts_poll(struct file *file, poll_table *wait)
646{
647 struct proc_mounts *p = file->private_data;
31b07093 648 unsigned res = POLLIN | POLLRDNORM;
5addc5dd 649
9f5596af
AV		 650 poll_wait(file, &p->ns->poll, wait);
651 if (mnt_had_events(p))
31b07093 652 res |= POLLERR | POLLPRI;
5addc5dd
AV		 653
654 return res;
655}
656
a1a2c409
MS		 657static int mounts_open(struct inode *inode, struct file *file)
658{
659 return mounts_open_common(inode, file, &mounts_op);
660}
661
00977a59 662static const struct file_operations proc_mounts_operations = {
1da177e4
LT		 663 .open = mounts_open,
664 .read = seq_read,
665 .llseek = seq_lseek,
666 .release = mounts_release,
5addc5dd 667 .poll = mounts_poll,
1da177e4
LT		 668};
669
2d4d4864
RP		 670static int mountinfo_open(struct inode *inode, struct file *file)
671{
672 return mounts_open_common(inode, file, &mountinfo_op);
673}
674
675static const struct file_operations proc_mountinfo_operations = {
676 .open = mountinfo_open,
677 .read = seq_read,
678 .llseek = seq_lseek,
679 .release = mounts_release,
680 .poll = mounts_poll,
681};
682
b4629fe2
CL		 683static int mountstats_open(struct inode *inode, struct file *file)
684{
a1a2c409 685 return mounts_open_common(inode, file, &mountstats_op);
b4629fe2
CL		 686}
687
00977a59 688static const struct file_operations proc_mountstats_operations = {
b4629fe2
CL		 689 .open = mountstats_open,
690 .read = seq_read,
691 .llseek = seq_lseek,
692 .release = mounts_release,
693};
694
1da177e4
LT		 695#define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
696
697static ssize_t proc_info_read(struct file * file, char __user * buf,
698 size_t count, loff_t *ppos)
699{
2fddfeef 700 struct inode * inode = file->f_path.dentry->d_inode;
1da177e4
LT		 701 unsigned long page;
702 ssize_t length;
99f89551
EB		 703 struct task_struct *task = get_proc_task(inode);
704
705 length = -ESRCH;
706 if (!task)
707 goto out_no_task;
1da177e4
LT		 708
709 if (count > PROC_BLOCK_SIZE)
710 count = PROC_BLOCK_SIZE;
99f89551
EB		 711
712 length = -ENOMEM;
e12ba74d 713 if (!(page = __get_free_page(GFP_TEMPORARY)))
99f89551 714 goto out;
1da177e4
LT		 715
716 length = PROC_I(inode)->op.proc_read(task, (char*)page);
717
718 if (length >= 0)
719 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
720 free_page(page);
99f89551
EB		 721out:
722 put_task_struct(task);
723out_no_task:
1da177e4
LT		 724 return length;
725}
726
00977a59 727static const struct file_operations proc_info_file_operations = {
1da177e4 728 .read = proc_info_read,
87df8424 729 .llseek = generic_file_llseek,
1da177e4
LT		 730};
731
be614086
EB		 732static int proc_single_show(struct seq_file *m, void *v)
733{
734 struct inode *inode = m->private;
735 struct pid_namespace *ns;
736 struct pid *pid;
737 struct task_struct *task;
738 int ret;
739
740 ns = inode->i_sb->s_fs_info;
741 pid = proc_pid(inode);
742 task = get_pid_task(pid, PIDTYPE_PID);
743 if (!task)
744 return -ESRCH;
745
746 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
747
748 put_task_struct(task);
749 return ret;
750}
751
752static int proc_single_open(struct inode *inode, struct file *filp)
753{
754 int ret;
755 ret = single_open(filp, proc_single_show, NULL);
756 if (!ret) {
757 struct seq_file *m = filp->private_data;
758
759 m->private = inode;
760 }
761 return ret;
762}
763
764static const struct file_operations proc_single_file_operations = {
765 .open = proc_single_open,
766 .read = seq_read,
767 .llseek = seq_lseek,
768 .release = single_release,
769};
770
1da177e4
LT		 771static int mem_open(struct inode* inode, struct file* file)
772{
773 file->private_data = (void*)((long)current->self_exec_id);
4a3956c7
KH		 774 /* OK to pass negative loff_t, we can catch out-of-range */
775 file->f_mode |= FMODE_UNSIGNED_OFFSET;
1da177e4
LT		 776 return 0;
777}
778
779static ssize_t mem_read(struct file * file, char __user * buf,
780 size_t count, loff_t *ppos)
781{
2fddfeef 782 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1da177e4
LT		 783 char *page;
784 unsigned long src = *ppos;
785 int ret = -ESRCH;
786 struct mm_struct *mm;
787
99f89551
EB		 788 if (!task)
789 goto out_no_task;
790
638fa202 791 if (check_mem_permission(task))
1da177e4
LT		 792 goto out;
793
794 ret = -ENOMEM;
e12ba74d 795 page = (char *)__get_free_page(GFP_TEMPORARY);
1da177e4
LT		 796 if (!page)
797 goto out;
798
799 ret = 0;
800
801 mm = get_task_mm(task);
802 if (!mm)
803 goto out_free;
804
805 ret = -EIO;
806
807 if (file->private_data != (void*)((long)current->self_exec_id))
808 goto out_put;
809
810 ret = 0;
811
812 while (count > 0) {
813 int this_len, retval;
814
815 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
816 retval = access_process_vm(task, src, page, this_len, 0);
638fa202 817 if (!retval || check_mem_permission(task)) {
1da177e4
LT		 818 if (!ret)
819 ret = -EIO;
820 break;
821 }
822
823 if (copy_to_user(buf, page, retval)) {
824 ret = -EFAULT;
825 break;
826 }
827
828 ret += retval;
829 src += retval;
830 buf += retval;
831 count -= retval;
832 }
833 *ppos = src;
834
835out_put:
836 mmput(mm);
837out_free:
838 free_page((unsigned long) page);
839out:
99f89551
EB		 840 put_task_struct(task);
841out_no_task:
1da177e4
LT		 842 return ret;
843}
844
845#define mem_write NULL
846
847#ifndef mem_write
848/* This is a security hazard */
63967fa9 849static ssize_t mem_write(struct file * file, const char __user *buf,
1da177e4
LT		 850 size_t count, loff_t *ppos)
851{
f7ca54f4 852 int copied;
1da177e4 853 char *page;
2fddfeef 854 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1da177e4
LT		 855 unsigned long dst = *ppos;
856
99f89551
EB		 857 copied = -ESRCH;
858 if (!task)
859 goto out_no_task;
860
638fa202 861 if (check_mem_permission(task))
99f89551 862 goto out;
1da177e4 863
99f89551 864 copied = -ENOMEM;
e12ba74d 865 page = (char *)__get_free_page(GFP_TEMPORARY);
1da177e4 866 if (!page)
99f89551 867 goto out;
1da177e4 868
f7ca54f4 869 copied = 0;
1da177e4
LT		 870 while (count > 0) {
871 int this_len, retval;
872
873 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
874 if (copy_from_user(page, buf, this_len)) {
875 copied = -EFAULT;
876 break;
877 }
878 retval = access_process_vm(task, dst, page, this_len, 1);
879 if (!retval) {
880 if (!copied)
881 copied = -EIO;
882 break;
883 }
884 copied += retval;
885 buf += retval;
886 dst += retval;
887 count -= retval;
888 }
889 *ppos = dst;
890 free_page((unsigned long) page);
99f89551
EB		 891out:
892 put_task_struct(task);
893out_no_task:
1da177e4
LT		 894 return copied;
895}
896#endif
897
85863e47 898loff_t mem_lseek(struct file *file, loff_t offset, int orig)
1da177e4
LT		 899{
900 switch (orig) {
901 case 0:
902 file->f_pos = offset;
903 break;
904 case 1:
905 file->f_pos += offset;
906 break;
907 default:
908 return -EINVAL;
909 }
910 force_successful_syscall_return();
911 return file->f_pos;
912}
913
00977a59 914static const struct file_operations proc_mem_operations = {
1da177e4
LT		 915 .llseek = mem_lseek,
916 .read = mem_read,
917 .write = mem_write,
918 .open = mem_open,
919};
920
315e28c8
JP		 921static ssize_t environ_read(struct file *file, char __user *buf,
922 size_t count, loff_t *ppos)
923{
924 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
925 char *page;
926 unsigned long src = *ppos;
927 int ret = -ESRCH;
928 struct mm_struct *mm;
929
930 if (!task)
931 goto out_no_task;
932
006ebb40 933 if (!ptrace_may_access(task, PTRACE_MODE_READ))
315e28c8
JP		 934 goto out;
935
936 ret = -ENOMEM;
937 page = (char *)__get_free_page(GFP_TEMPORARY);
938 if (!page)
939 goto out;
940
941 ret = 0;
942
943 mm = get_task_mm(task);
944 if (!mm)
945 goto out_free;
946
947 while (count > 0) {
948 int this_len, retval, max_len;
949
950 this_len = mm->env_end - (mm->env_start + src);
951
952 if (this_len <= 0)
953 break;
954
955 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
956 this_len = (this_len > max_len) ? max_len : this_len;
957
958 retval = access_process_vm(task, (mm->env_start + src),
959 page, this_len, 0);
960
961 if (retval <= 0) {
962 ret = retval;
963 break;
964 }
965
966 if (copy_to_user(buf, page, retval)) {
967 ret = -EFAULT;
968 break;
969 }
970
971 ret += retval;
972 src += retval;
973 buf += retval;
974 count -= retval;
975 }
976 *ppos = src;
977
978 mmput(mm);
979out_free:
980 free_page((unsigned long) page);
981out:
982 put_task_struct(task);
983out_no_task:
984 return ret;
985}
986
987static const struct file_operations proc_environ_operations = {
988 .read = environ_read,
87df8424 989 .llseek = generic_file_llseek,
315e28c8
JP		 990};
991
1da177e4
LT		 992static ssize_t oom_adjust_read(struct file *file, char __user *buf,
993 size_t count, loff_t *ppos)
994{
2fddfeef 995 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
8578cea7 996 char buffer[PROC_NUMBUF];
1da177e4 997 size_t len;
28b83c51
KM		 998 int oom_adjust = OOM_DISABLE;
999 unsigned long flags;
1da177e4 1000
99f89551
EB		 1001 if (!task)
1002 return -ESRCH;
28b83c51
KM		 1003
1004 if (lock_task_sighand(task, &flags)) {
1005 oom_adjust = task->signal->oom_adj;
1006 unlock_task_sighand(task, &flags);
1007 }
1008
99f89551
EB		 1009 put_task_struct(task);
1010
8578cea7 1011 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
0c28f287
AM		 1012
1013 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1da177e4
LT		 1014}
1015
1016static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1017 size_t count, loff_t *ppos)
1018{
99f89551 1019 struct task_struct *task;
5d863b89
KM		 1020 char buffer[PROC_NUMBUF];
1021 long oom_adjust;
28b83c51 1022 unsigned long flags;
5d863b89 1023 int err;
1da177e4 1024
8578cea7
EB		 1025 memset(buffer, 0, sizeof(buffer));
1026 if (count > sizeof(buffer) - 1)
1027 count = sizeof(buffer) - 1;
723548bf
DR		 1028 if (copy_from_user(buffer, buf, count)) {
1029 err = -EFAULT;
1030 goto out;
1031 }
5d863b89
KM		 1032
1033 err = strict_strtol(strstrip(buffer), 0, &oom_adjust);
1034 if (err)
723548bf 1035 goto out;
8ac773b4 1036 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
723548bf
DR		 1037 oom_adjust != OOM_DISABLE) {
1038 err = -EINVAL;
1039 goto out;
1040 }
5d863b89 1041
2fddfeef 1042 task = get_proc_task(file->f_path.dentry->d_inode);
723548bf
DR		 1043 if (!task) {
1044 err = -ESRCH;
1045 goto out;
1046 }
d19d5476
DR		 1047
1048 task_lock(task);
1049 if (!task->mm) {
1050 err = -EINVAL;
1051 goto err_task_lock;
1052 }
1053
28b83c51 1054 if (!lock_task_sighand(task, &flags)) {
723548bf 1055 err = -ESRCH;
d19d5476 1056 goto err_task_lock;
28b83c51
KM		 1057 }
1058
1059 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
723548bf
DR		 1060 err = -EACCES;
1061 goto err_sighand;
8fb4fc68 1062 }
28b83c51 1063
3d5992d2
YH		 1064 if (oom_adjust != task->signal->oom_adj) {
1065 if (oom_adjust == OOM_DISABLE)
1066 atomic_inc(&task->mm->oom_disable_count);
1067 if (task->signal->oom_adj == OOM_DISABLE)
1068 atomic_dec(&task->mm->oom_disable_count);
1069 }
1070
51b1bd2a
DR		 1071 /*
1072 * Warn that /proc/pid/oom_adj is deprecated, see
1073 * Documentation/feature-removal-schedule.txt.
1074 */
1075 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, "
1076 "please use /proc/%d/oom_score_adj instead.\n",
1077 current->comm, task_pid_nr(current),
1078 task_pid_nr(task), task_pid_nr(task));
28b83c51 1079 task->signal->oom_adj = oom_adjust;
a63d83f4
DR		 1080 /*
1081 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1082 * value is always attainable.
1083 */
1084 if (task->signal->oom_adj == OOM_ADJUST_MAX)
1085 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
1086 else
1087 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
1088 -OOM_DISABLE;
723548bf 1089err_sighand:
28b83c51 1090 unlock_task_sighand(task, &flags);
d19d5476
DR		 1091err_task_lock:
1092 task_unlock(task);
99f89551 1093 put_task_struct(task);
723548bf
DR		 1094out:
1095 return err < 0 ? err : count;
1da177e4
LT		 1096}
1097
00977a59 1098static const struct file_operations proc_oom_adjust_operations = {
1da177e4
LT		 1099 .read = oom_adjust_read,
1100 .write = oom_adjust_write,
87df8424 1101 .llseek = generic_file_llseek,
1da177e4
LT		 1102};
1103
a63d83f4
DR		 1104static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1105 size_t count, loff_t *ppos)
1106{
1107 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1108 char buffer[PROC_NUMBUF];
1109 int oom_score_adj = OOM_SCORE_ADJ_MIN;
1110 unsigned long flags;
1111 size_t len;
1112
1113 if (!task)
1114 return -ESRCH;
1115 if (lock_task_sighand(task, &flags)) {
1116 oom_score_adj = task->signal->oom_score_adj;
1117 unlock_task_sighand(task, &flags);
1118 }
1119 put_task_struct(task);
1120 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
1121 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1122}
1123
1124static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1125 size_t count, loff_t *ppos)
1126{
1127 struct task_struct *task;
1128 char buffer[PROC_NUMBUF];
1129 unsigned long flags;
1130 long oom_score_adj;
1131 int err;
1132
1133 memset(buffer, 0, sizeof(buffer));
1134 if (count > sizeof(buffer) - 1)
1135 count = sizeof(buffer) - 1;
723548bf
DR		 1136 if (copy_from_user(buffer, buf, count)) {
1137 err = -EFAULT;
1138 goto out;
1139 }
a63d83f4
DR		 1140
1141 err = strict_strtol(strstrip(buffer), 0, &oom_score_adj);
1142 if (err)
723548bf 1143 goto out;
a63d83f4 1144 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
723548bf
DR		 1145 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1146 err = -EINVAL;
1147 goto out;
1148 }
a63d83f4
DR		 1149
1150 task = get_proc_task(file->f_path.dentry->d_inode);
723548bf
DR		 1151 if (!task) {
1152 err = -ESRCH;
1153 goto out;
1154 }
d19d5476
DR		 1155
1156 task_lock(task);
1157 if (!task->mm) {
1158 err = -EINVAL;
1159 goto err_task_lock;
1160 }
1161
a63d83f4 1162 if (!lock_task_sighand(task, &flags)) {
723548bf 1163 err = -ESRCH;
d19d5476 1164 goto err_task_lock;
a63d83f4 1165 }
d19d5476 1166
a63d83f4
DR		 1167 if (oom_score_adj < task->signal->oom_score_adj &&
1168 !capable(CAP_SYS_RESOURCE)) {
723548bf
DR		 1169 err = -EACCES;
1170 goto err_sighand;
a63d83f4
DR		 1171 }
1172
3d5992d2
YH		 1173 if (oom_score_adj != task->signal->oom_score_adj) {
1174 if (oom_score_adj == OOM_SCORE_ADJ_MIN)
1175 atomic_inc(&task->mm->oom_disable_count);
1176 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1177 atomic_dec(&task->mm->oom_disable_count);
1178 }
a63d83f4
DR		 1179 task->signal->oom_score_adj = oom_score_adj;
1180 /*
1181 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1182 * always attainable.
1183 */
1184 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1185 task->signal->oom_adj = OOM_DISABLE;
1186 else
1187 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1188 OOM_SCORE_ADJ_MAX;
723548bf 1189err_sighand:
a63d83f4 1190 unlock_task_sighand(task, &flags);
d19d5476
DR		 1191err_task_lock:
1192 task_unlock(task);
a63d83f4 1193 put_task_struct(task);
723548bf
DR		 1194out:
1195 return err < 0 ? err : count;
a63d83f4
DR		 1196}
1197
1198static const struct file_operations proc_oom_score_adj_operations = {
1199 .read = oom_score_adj_read,
1200 .write = oom_score_adj_write,
6038f373 1201 .llseek = default_llseek,
a63d83f4
DR		 1202};
1203
1da177e4
LT		 1204#ifdef CONFIG_AUDITSYSCALL
1205#define TMPBUFLEN 21
1206static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1207 size_t count, loff_t *ppos)
1208{
2fddfeef 1209 struct inode * inode = file->f_path.dentry->d_inode;
99f89551 1210 struct task_struct *task = get_proc_task(inode);
1da177e4
LT		 1211 ssize_t length;
1212 char tmpbuf[TMPBUFLEN];
1213
99f89551
EB		 1214 if (!task)
1215 return -ESRCH;
1da177e4 1216 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
0c11b942 1217 audit_get_loginuid(task));
99f89551 1218 put_task_struct(task);
1da177e4
LT		 1219 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1220}
1221
1222static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1223 size_t count, loff_t *ppos)
1224{
2fddfeef 1225 struct inode * inode = file->f_path.dentry->d_inode;
1da177e4
LT		 1226 char *page, *tmp;
1227 ssize_t length;
1da177e4
LT		 1228 uid_t loginuid;
1229
1230 if (!capable(CAP_AUDIT_CONTROL))
1231 return -EPERM;
1232
7dc52157
PM		 1233 rcu_read_lock();
1234 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1235 rcu_read_unlock();
1da177e4 1236 return -EPERM;
7dc52157
PM		 1237 }
1238 rcu_read_unlock();
1da177e4 1239
e0182909
AV		 1240 if (count >= PAGE_SIZE)
1241 count = PAGE_SIZE - 1;
1da177e4
LT		 1242
1243 if (*ppos != 0) {
1244 /* No partial writes. */
1245 return -EINVAL;
1246 }
e12ba74d 1247 page = (char*)__get_free_page(GFP_TEMPORARY);
1da177e4
LT		 1248 if (!page)
1249 return -ENOMEM;
1250 length = -EFAULT;
1251 if (copy_from_user(page, buf, count))
1252 goto out_free_page;
1253
e0182909 1254 page[count] = '\0';
1da177e4
LT		 1255 loginuid = simple_strtoul(page, &tmp, 10);
1256 if (tmp == page) {
1257 length = -EINVAL;
1258 goto out_free_page;
1259
1260 }
99f89551 1261 length = audit_set_loginuid(current, loginuid);
1da177e4
LT		 1262 if (likely(length == 0))
1263 length = count;
1264
1265out_free_page:
1266 free_page((unsigned long) page);
1267 return length;
1268}
1269
00977a59 1270static const struct file_operations proc_loginuid_operations = {
1da177e4
LT		 1271 .read = proc_loginuid_read,
1272 .write = proc_loginuid_write,
87df8424 1273 .llseek = generic_file_llseek,
1da177e4 1274};
1e0bd755
EP		 1275
1276static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1277 size_t count, loff_t *ppos)
1278{
1279 struct inode * inode = file->f_path.dentry->d_inode;
1280 struct task_struct *task = get_proc_task(inode);
1281 ssize_t length;
1282 char tmpbuf[TMPBUFLEN];
1283
1284 if (!task)
1285 return -ESRCH;
1286 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1287 audit_get_sessionid(task));
1288 put_task_struct(task);
1289 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1290}
1291
1292static const struct file_operations proc_sessionid_operations = {
1293 .read = proc_sessionid_read,
87df8424 1294 .llseek = generic_file_llseek,
1e0bd755 1295};
1da177e4
LT		 1296#endif
1297
f4f154fd
AM		 1298#ifdef CONFIG_FAULT_INJECTION
1299static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1300 size_t count, loff_t *ppos)
1301{
1302 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1303 char buffer[PROC_NUMBUF];
1304 size_t len;
1305 int make_it_fail;
f4f154fd
AM		 1306
1307 if (!task)
1308 return -ESRCH;
1309 make_it_fail = task->make_it_fail;
1310 put_task_struct(task);
1311
1312 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
0c28f287
AM		 1313
1314 return simple_read_from_buffer(buf, count, ppos, buffer, len);
f4f154fd
AM		 1315}
1316
1317static ssize_t proc_fault_inject_write(struct file * file,
1318 const char __user * buf, size_t count, loff_t *ppos)
1319{
1320 struct task_struct *task;
1321 char buffer[PROC_NUMBUF], *end;
1322 int make_it_fail;
1323
1324 if (!capable(CAP_SYS_RESOURCE))
1325 return -EPERM;
1326 memset(buffer, 0, sizeof(buffer));
1327 if (count > sizeof(buffer) - 1)
1328 count = sizeof(buffer) - 1;
1329 if (copy_from_user(buffer, buf, count))
1330 return -EFAULT;
cba8aafe
VL		 1331 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1332 if (*end)
1333 return -EINVAL;
f4f154fd
AM		 1334 task = get_proc_task(file->f_dentry->d_inode);
1335 if (!task)
1336 return -ESRCH;
1337 task->make_it_fail = make_it_fail;
1338 put_task_struct(task);
cba8aafe
VL		 1339
1340 return count;
f4f154fd
AM		 1341}
1342
00977a59 1343static const struct file_operations proc_fault_inject_operations = {
f4f154fd
AM		 1344 .read = proc_fault_inject_read,
1345 .write = proc_fault_inject_write,
87df8424 1346 .llseek = generic_file_llseek,
f4f154fd
AM		 1347};
1348#endif
1349
9745512c 1350
43ae34cb
IM		 1351#ifdef CONFIG_SCHED_DEBUG
1352/*
1353 * Print out various scheduling related per-task fields:
1354 */
1355static int sched_show(struct seq_file *m, void *v)
1356{
1357 struct inode *inode = m->private;
1358 struct task_struct *p;
1359
43ae34cb
IM		 1360 p = get_proc_task(inode);
1361 if (!p)
1362 return -ESRCH;
1363 proc_sched_show_task(p, m);
1364
1365 put_task_struct(p);
1366
1367 return 0;
1368}
1369
1370static ssize_t
1371sched_write(struct file *file, const char __user *buf,
1372 size_t count, loff_t *offset)
1373{
1374 struct inode *inode = file->f_path.dentry->d_inode;
1375 struct task_struct *p;
1376
43ae34cb
IM		 1377 p = get_proc_task(inode);
1378 if (!p)
1379 return -ESRCH;
1380 proc_sched_set_task(p);
1381
1382 put_task_struct(p);
1383
1384 return count;
1385}
1386
1387static int sched_open(struct inode *inode, struct file *filp)
1388{
1389 int ret;
1390
1391 ret = single_open(filp, sched_show, NULL);
1392 if (!ret) {
1393 struct seq_file *m = filp->private_data;
1394
1395 m->private = inode;
1396 }
1397 return ret;
1398}
1399
1400static const struct file_operations proc_pid_sched_operations = {
1401 .open = sched_open,
1402 .read = seq_read,
1403 .write = sched_write,
1404 .llseek = seq_lseek,
5ea473a1 1405 .release = single_release,
43ae34cb
IM		 1406};
1407
1408#endif
1409
5091faa4
MG		 1410#ifdef CONFIG_SCHED_AUTOGROUP
1411/*
1412 * Print out autogroup related information:
1413 */
1414static int sched_autogroup_show(struct seq_file *m, void *v)
1415{
1416 struct inode *inode = m->private;
1417 struct task_struct *p;
1418
1419 p = get_proc_task(inode);
1420 if (!p)
1421 return -ESRCH;
1422 proc_sched_autogroup_show_task(p, m);
1423
1424 put_task_struct(p);
1425
1426 return 0;
1427}
1428
1429static ssize_t
1430sched_autogroup_write(struct file *file, const char __user *buf,
1431 size_t count, loff_t *offset)
1432{
1433 struct inode *inode = file->f_path.dentry->d_inode;
1434 struct task_struct *p;
1435 char buffer[PROC_NUMBUF];
1436 long nice;
1437 int err;
1438
1439 memset(buffer, 0, sizeof(buffer));
1440 if (count > sizeof(buffer) - 1)
1441 count = sizeof(buffer) - 1;
1442 if (copy_from_user(buffer, buf, count))
1443 return -EFAULT;
1444
1445 err = strict_strtol(strstrip(buffer), 0, &nice);
1446 if (err)
1447 return -EINVAL;
1448
1449 p = get_proc_task(inode);
1450 if (!p)
1451 return -ESRCH;
1452
1453 err = nice;
1454 err = proc_sched_autogroup_set_nice(p, &err);
1455 if (err)
1456 count = err;
1457
1458 put_task_struct(p);
1459
1460 return count;
1461}
1462
1463static int sched_autogroup_open(struct inode *inode, struct file *filp)
1464{
1465 int ret;
1466
1467 ret = single_open(filp, sched_autogroup_show, NULL);
1468 if (!ret) {
1469 struct seq_file *m = filp->private_data;
1470
1471 m->private = inode;
1472 }
1473 return ret;
1474}
1475
1476static const struct file_operations proc_pid_sched_autogroup_operations = {
1477 .open = sched_autogroup_open,
1478 .read = seq_read,
1479 .write = sched_autogroup_write,
1480 .llseek = seq_lseek,
1481 .release = single_release,
1482};
1483
1484#endif /* CONFIG_SCHED_AUTOGROUP */
1485
4614a696 1486static ssize_t comm_write(struct file *file, const char __user *buf,
1487 size_t count, loff_t *offset)
1488{
1489 struct inode *inode = file->f_path.dentry->d_inode;
1490 struct task_struct *p;
1491 char buffer[TASK_COMM_LEN];
1492
1493 memset(buffer, 0, sizeof(buffer));
1494 if (count > sizeof(buffer) - 1)
1495 count = sizeof(buffer) - 1;
1496 if (copy_from_user(buffer, buf, count))
1497 return -EFAULT;
1498
1499 p = get_proc_task(inode);
1500 if (!p)
1501 return -ESRCH;
1502
1503 if (same_thread_group(current, p))
1504 set_task_comm(p, buffer);
1505 else
1506 count = -EINVAL;
1507
1508 put_task_struct(p);
1509
1510 return count;
1511}
1512
1513static int comm_show(struct seq_file *m, void *v)
1514{
1515 struct inode *inode = m->private;
1516 struct task_struct *p;
1517
1518 p = get_proc_task(inode);
1519 if (!p)
1520 return -ESRCH;
1521
1522 task_lock(p);
1523 seq_printf(m, "%s\n", p->comm);
1524 task_unlock(p);
1525
1526 put_task_struct(p);
1527
1528 return 0;
1529}
1530
1531static int comm_open(struct inode *inode, struct file *filp)
1532{
1533 int ret;
1534
1535 ret = single_open(filp, comm_show, NULL);
1536 if (!ret) {
1537 struct seq_file *m = filp->private_data;
1538
1539 m->private = inode;
1540 }
1541 return ret;
1542}
1543
1544static const struct file_operations proc_pid_set_comm_operations = {
1545 .open = comm_open,
1546 .read = seq_read,
1547 .write = comm_write,
1548 .llseek = seq_lseek,
1549 .release = single_release,
1550};
1551
925d1c40
MH		 1552/*
1553 * We added or removed a vma mapping the executable. The vmas are only mapped
1554 * during exec and are not mapped with the mmap system call.
1555 * Callers must hold down_write() on the mm's mmap_sem for these
1556 */
1557void added_exe_file_vma(struct mm_struct *mm)
1558{
1559 mm->num_exe_file_vmas++;
1560}
1561
1562void removed_exe_file_vma(struct mm_struct *mm)
1563{
1564 mm->num_exe_file_vmas--;
1565 if ((mm->num_exe_file_vmas == 0) && mm->exe_file){
1566 fput(mm->exe_file);
1567 mm->exe_file = NULL;
1568 }
1569
1570}
1571
1572void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1573{
1574 if (new_exe_file)
1575 get_file(new_exe_file);
1576 if (mm->exe_file)
1577 fput(mm->exe_file);
1578 mm->exe_file = new_exe_file;
1579 mm->num_exe_file_vmas = 0;
1580}
1581
1582struct file *get_mm_exe_file(struct mm_struct *mm)
1583{
1584 struct file *exe_file;
1585
1586 /* We need mmap_sem to protect against races with removal of
1587 * VM_EXECUTABLE vmas */
1588 down_read(&mm->mmap_sem);
1589 exe_file = mm->exe_file;
1590 if (exe_file)
1591 get_file(exe_file);
1592 up_read(&mm->mmap_sem);
1593 return exe_file;
1594}
1595
1596void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
1597{
1598 /* It's safe to write the exe_file pointer without exe_file_lock because
1599 * this is called during fork when the task is not yet in /proc */
1600 newmm->exe_file = get_mm_exe_file(oldmm);
1601}
1602
1603static int proc_exe_link(struct inode *inode, struct path *exe_path)
1604{
1605 struct task_struct *task;
1606 struct mm_struct *mm;
1607 struct file *exe_file;
1608
1609 task = get_proc_task(inode);
1610 if (!task)
1611 return -ENOENT;
1612 mm = get_task_mm(task);
1613 put_task_struct(task);
1614 if (!mm)
1615 return -ENOENT;
1616 exe_file = get_mm_exe_file(mm);
1617 mmput(mm);
1618 if (exe_file) {
1619 *exe_path = exe_file->f_path;
1620 path_get(&exe_file->f_path);
1621 fput(exe_file);
1622 return 0;
1623 } else
1624 return -ENOENT;
1625}
1626
008b150a 1627static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1da177e4
LT		 1628{
1629 struct inode *inode = dentry->d_inode;
1630 int error = -EACCES;
1631
1632 /* We don't need a base pointer in the /proc filesystem */
1d957f9b 1633 path_put(&nd->path);
1da177e4 1634
778c1144
EB		 1635 /* Are we allowed to snoop on the tasks file descriptors? */
1636 if (!proc_fd_access_allowed(inode))
1da177e4 1637 goto out;
1da177e4 1638
3dcd25f3 1639 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
1da177e4 1640out:
008b150a 1641 return ERR_PTR(error);
1da177e4
LT		 1642}
1643
3dcd25f3 1644static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1da177e4 1645{
e12ba74d 1646 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
3dcd25f3 1647 char *pathname;
1da177e4
LT		 1648 int len;
1649
1650 if (!tmp)
1651 return -ENOMEM;
0c28f287 1652
8df9d1a4 1653 pathname = d_path_with_unreachable(path, tmp, PAGE_SIZE);
3dcd25f3
JB		 1654 len = PTR_ERR(pathname);
1655 if (IS_ERR(pathname))
1da177e4 1656 goto out;
3dcd25f3 1657 len = tmp + PAGE_SIZE - 1 - pathname;
1da177e4
LT		 1658
1659 if (len > buflen)
1660 len = buflen;
3dcd25f3 1661 if (copy_to_user(buffer, pathname, len))
1da177e4
LT		 1662 len = -EFAULT;
1663 out:
1664 free_page((unsigned long)tmp);
1665 return len;
1666}
1667
1668static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1669{
1670 int error = -EACCES;
1671 struct inode *inode = dentry->d_inode;
3dcd25f3 1672 struct path path;
1da177e4 1673
778c1144
EB		 1674 /* Are we allowed to snoop on the tasks file descriptors? */
1675 if (!proc_fd_access_allowed(inode))
1da177e4 1676 goto out;
1da177e4 1677
3dcd25f3 1678 error = PROC_I(inode)->op.proc_get_link(inode, &path);
1da177e4
LT		 1679 if (error)
1680 goto out;
1681
3dcd25f3
JB		 1682 error = do_proc_readlink(&path, buffer, buflen);
1683 path_put(&path);
1da177e4 1684out:
1da177e4
LT		 1685 return error;
1686}
1687
c5ef1c42 1688static const struct inode_operations proc_pid_link_inode_operations = {
1da177e4 1689 .readlink = proc_pid_readlink,
6d76fa58
LT		 1690 .follow_link = proc_pid_follow_link,
1691 .setattr = proc_setattr,
1da177e4
LT		 1692};
1693
28a6d671
EB		 1694
1695/* building an inode */
1696
1697static int task_dumpable(struct task_struct *task)
1da177e4 1698{
28a6d671
EB		 1699 int dumpable = 0;
1700 struct mm_struct *mm;
1da177e4 1701
28a6d671
EB		 1702 task_lock(task);
1703 mm = task->mm;
1704 if (mm)
6c5d5238 1705 dumpable = get_dumpable(mm);
28a6d671
EB		 1706 task_unlock(task);
1707 if(dumpable == 1)
1708 return 1;
1709 return 0;
1710}
1da177e4 1711
1da177e4 1712
61a28784 1713static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
28a6d671
EB		 1714{
1715 struct inode * inode;
1716 struct proc_inode *ei;
c69e8d9c 1717 const struct cred *cred;
1da177e4 1718
28a6d671 1719 /* We need a new inode */
1da177e4 1720
28a6d671
EB		 1721 inode = new_inode(sb);
1722 if (!inode)
1723 goto out;
1724
1725 /* Common stuff */
1726 ei = PROC_I(inode);
85fe4025 1727 inode->i_ino = get_next_ino();
28a6d671 1728 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
28a6d671
EB		 1729 inode->i_op = &proc_def_inode_operations;
1730
1731 /*
1732 * grab the reference to task.
1733 */
1a657f78 1734 ei->pid = get_task_pid(task, PIDTYPE_PID);
28a6d671
EB		 1735 if (!ei->pid)
1736 goto out_unlock;
1737
28a6d671 1738 if (task_dumpable(task)) {
c69e8d9c
DH		 1739 rcu_read_lock();
1740 cred = __task_cred(task);
1741 inode->i_uid = cred->euid;
1742 inode->i_gid = cred->egid;
1743 rcu_read_unlock();
1da177e4 1744 }
28a6d671
EB		 1745 security_task_to_inode(task, inode);
1746
1da177e4 1747out:
28a6d671
EB		 1748 return inode;
1749
1750out_unlock:
1751 iput(inode);
1752 return NULL;
1da177e4
LT		 1753}
1754
28a6d671 1755static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1da177e4 1756{
1da177e4 1757 struct inode *inode = dentry->d_inode;
28a6d671 1758 struct task_struct *task;
c69e8d9c
DH		 1759 const struct cred *cred;
1760
28a6d671 1761 generic_fillattr(inode, stat);
1da177e4 1762
28a6d671
EB		 1763 rcu_read_lock();
1764 stat->uid = 0;
1765 stat->gid = 0;
1766 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1767 if (task) {
1768 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1769 task_dumpable(task)) {
c69e8d9c
DH		 1770 cred = __task_cred(task);
1771 stat->uid = cred->euid;
1772 stat->gid = cred->egid;
1da177e4
LT		 1773 }
1774 }
28a6d671 1775 rcu_read_unlock();
d6e71144 1776 return 0;
1da177e4
LT		 1777}
1778
1da177e4
LT		 1779/* dentry stuff */
1780
1781/*
1782 * Exceptional case: normally we are not allowed to unhash a busy
1783 * directory. In this case, however, we can do it - no aliasing problems
1784 * due to the way we treat inodes.
1785 *
1786 * Rewrite the inode's ownerships here because the owning task may have
1787 * performed a setuid(), etc.
99f89551
EB		 1788 *
1789 * Before the /proc/pid/status file was created the only way to read
1790 * the effective uid of a /process was to stat /proc/pid. Reading
1791 * /proc/pid/status is slow enough that procps and other packages
1792 * kept stating /proc/pid. To keep the rules in /proc simple I have
1793 * made this apply to all per process world readable and executable
1794 * directories.
1da177e4
LT		 1795 */
1796static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1797{
1798 struct inode *inode = dentry->d_inode;
99f89551 1799 struct task_struct *task = get_proc_task(inode);
c69e8d9c
DH		 1800 const struct cred *cred;
1801
99f89551
EB		 1802 if (task) {
1803 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1804 task_dumpable(task)) {
c69e8d9c
DH		 1805 rcu_read_lock();
1806 cred = __task_cred(task);
1807 inode->i_uid = cred->euid;
1808 inode->i_gid = cred->egid;
1809 rcu_read_unlock();
1da177e4
LT		 1810 } else {
1811 inode->i_uid = 0;
1812 inode->i_gid = 0;
1813 }
9ee8ab9f 1814 inode->i_mode &= ~(S_ISUID | S_ISGID);
1da177e4 1815 security_task_to_inode(task, inode);
99f89551 1816 put_task_struct(task);
1da177e4
LT		 1817 return 1;
1818 }
1819 d_drop(dentry);
1820 return 0;
1821}
1822
28a6d671 1823static int pid_delete_dentry(struct dentry * dentry)
99f89551 1824{
28a6d671
EB		 1825 /* Is the task we represent dead?
1826 * If so, then don't put the dentry on the lru list,
1827 * kill it immediately.
1828 */
1829 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1830}
1831
d72f71eb 1832static const struct dentry_operations pid_dentry_operations =
28a6d671
EB		 1833{
1834 .d_revalidate = pid_revalidate,
1835 .d_delete = pid_delete_dentry,
1836};
1837
1838/* Lookups */
1839
c5141e6d
ED		 1840typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1841 struct task_struct *, const void *);
61a28784 1842
1c0d04c9
EB		 1843/*
1844 * Fill a directory entry.
1845 *
1846 * If possible create the dcache entry and derive our inode number and
1847 * file type from dcache entry.
1848 *
1849 * Since all of the proc inode numbers are dynamically generated, the inode
1850 * numbers do not exist until the inode is cache. This means creating the
1851 * the dcache entry in readdir is necessary to keep the inode numbers
1852 * reported by readdir in sync with the inode numbers reported
1853 * by stat.
1854 */
61a28784
EB		 1855static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1856 char *name, int len,
c5141e6d 1857 instantiate_t instantiate, struct task_struct *task, const void *ptr)
61a28784 1858{
2fddfeef 1859 struct dentry *child, *dir = filp->f_path.dentry;
61a28784
EB		 1860 struct inode *inode;
1861 struct qstr qname;
1862 ino_t ino = 0;
1863 unsigned type = DT_UNKNOWN;
1864
1865 qname.name = name;
1866 qname.len = len;
1867 qname.hash = full_name_hash(name, len);
1868
1869 child = d_lookup(dir, &qname);
1870 if (!child) {
1871 struct dentry *new;
1872 new = d_alloc(dir, &qname);
1873 if (new) {
1874 child = instantiate(dir->d_inode, new, task, ptr);
1875 if (child)
1876 dput(new);
1877 else
1878 child = new;
1879 }
1880 }
1881 if (!child || IS_ERR(child) || !child->d_inode)
1882 goto end_instantiate;
1883 inode = child->d_inode;
1884 if (inode) {
1885 ino = inode->i_ino;
1886 type = inode->i_mode >> 12;
1887 }
1888 dput(child);
1889end_instantiate:
1890 if (!ino)
1891 ino = find_inode_number(dir, &qname);
1892 if (!ino)
1893 ino = 1;
1894 return filldir(dirent, name, len, filp->f_pos, ino, type);
1895}
1896
28a6d671
EB		 1897static unsigned name_to_int(struct dentry *dentry)
1898{
1899 const char *name = dentry->d_name.name;
1900 int len = dentry->d_name.len;
1901 unsigned n = 0;
1902
1903 if (len > 1 && *name == '0')
1904 goto out;
1905 while (len-- > 0) {
1906 unsigned c = *name++ - '0';
1907 if (c > 9)
1908 goto out;
1909 if (n >= (~0U-9)/10)
1910 goto out;
1911 n *= 10;
1912 n += c;
1913 }
1914 return n;
1915out:
1916 return ~0U;
1917}
1918
27932742
MS		 1919#define PROC_FDINFO_MAX 64
1920
3dcd25f3 1921static int proc_fd_info(struct inode *inode, struct path *path, char *info)
28a6d671
EB		 1922{
1923 struct task_struct *task = get_proc_task(inode);
1924 struct files_struct *files = NULL;
1925 struct file *file;
1926 int fd = proc_fd(inode);
99f89551 1927
99f89551 1928 if (task) {
28a6d671
EB		 1929 files = get_files_struct(task);
1930 put_task_struct(task);
1931 }
1932 if (files) {
1933 /*
1934 * We are not taking a ref to the file structure, so we must
1935 * hold ->file_lock.
1936 */
1937 spin_lock(&files->file_lock);
1938 file = fcheck_files(files, fd);
1939 if (file) {
3dcd25f3
JB		 1940 if (path) {
1941 *path = file->f_path;
1942 path_get(&file->f_path);
1943 }
27932742
MS		 1944 if (info)
1945 snprintf(info, PROC_FDINFO_MAX,
1946 "pos:\t%lli\n"
1947 "flags:\t0%o\n",
1948 (long long) file->f_pos,
1949 file->f_flags);
28a6d671
EB		 1950 spin_unlock(&files->file_lock);
1951 put_files_struct(files);
1952 return 0;
99f89551 1953 }
28a6d671
EB		 1954 spin_unlock(&files->file_lock);
1955 put_files_struct(files);
99f89551 1956 }
28a6d671 1957 return -ENOENT;
99f89551
EB		 1958}
1959
3dcd25f3 1960static int proc_fd_link(struct inode *inode, struct path *path)
27932742 1961{
3dcd25f3 1962 return proc_fd_info(inode, path, NULL);
27932742
MS		 1963}
1964
1da177e4
LT		 1965static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1966{
1967 struct inode *inode = dentry->d_inode;
99f89551 1968 struct task_struct *task = get_proc_task(inode);
aed7a6c4 1969 int fd = proc_fd(inode);
1da177e4 1970 struct files_struct *files;
c69e8d9c 1971 const struct cred *cred;
1da177e4 1972
99f89551
EB		 1973 if (task) {
1974 files = get_files_struct(task);
1975 if (files) {
1976 rcu_read_lock();
1977 if (fcheck_files(files, fd)) {
1978 rcu_read_unlock();
1979 put_files_struct(files);
1980 if (task_dumpable(task)) {
c69e8d9c
DH		 1981 rcu_read_lock();
1982 cred = __task_cred(task);
1983 inode->i_uid = cred->euid;
1984 inode->i_gid = cred->egid;
1985 rcu_read_unlock();
99f89551
EB		 1986 } else {
1987 inode->i_uid = 0;
1988 inode->i_gid = 0;
1989 }
9ee8ab9f 1990 inode->i_mode &= ~(S_ISUID | S_ISGID);
99f89551
EB		 1991 security_task_to_inode(task, inode);
1992 put_task_struct(task);
1993 return 1;
1994 }
b835996f 1995 rcu_read_unlock();
1da177e4 1996 put_files_struct(files);
1da177e4 1997 }
99f89551 1998 put_task_struct(task);
1da177e4
LT		 1999 }
2000 d_drop(dentry);
2001 return 0;
2002}
2003
d72f71eb 2004static const struct dentry_operations tid_fd_dentry_operations =
1da177e4
LT		 2005{
2006 .d_revalidate = tid_fd_revalidate,
2007 .d_delete = pid_delete_dentry,
2008};
2009
444ceed8 2010static struct dentry *proc_fd_instantiate(struct inode *dir,
c5141e6d 2011 struct dentry *dentry, struct task_struct *task, const void *ptr)
1da177e4 2012{
c5141e6d 2013 unsigned fd = *(const unsigned *)ptr;
444ceed8
EB		 2014 struct file *file;
2015 struct files_struct *files;
2016 struct inode *inode;
2017 struct proc_inode *ei;
2018 struct dentry *error = ERR_PTR(-ENOENT);
1da177e4 2019
61a28784 2020 inode = proc_pid_make_inode(dir->i_sb, task);
1da177e4
LT		 2021 if (!inode)
2022 goto out;
2023 ei = PROC_I(inode);
aed7a6c4 2024 ei->fd = fd;
1da177e4
LT		 2025 files = get_files_struct(task);
2026 if (!files)
444ceed8 2027 goto out_iput;
1da177e4 2028 inode->i_mode = S_IFLNK;
ca99c1da
DS		 2029
2030 /*
2031 * We are not taking a ref to the file structure, so we must
2032 * hold ->file_lock.
2033 */
2034 spin_lock(&files->file_lock);
1da177e4
LT		 2035 file = fcheck_files(files, fd);
2036 if (!file)
444ceed8 2037 goto out_unlock;
aeb5d727 2038 if (file->f_mode & FMODE_READ)
1da177e4 2039 inode->i_mode |= S_IRUSR | S_IXUSR;
aeb5d727 2040 if (file->f_mode & FMODE_WRITE)
1da177e4 2041 inode->i_mode |= S_IWUSR | S_IXUSR;
ca99c1da 2042 spin_unlock(&files->file_lock);
1da177e4 2043 put_files_struct(files);
444ceed8 2044
1da177e4
LT		 2045 inode->i_op = &proc_pid_link_inode_operations;
2046 inode->i_size = 64;
2047 ei->op.proc_get_link = proc_fd_link;
2048 dentry->d_op = &tid_fd_dentry_operations;
2049 d_add(dentry, inode);
cd6a3ce9
EB		 2050 /* Close the race of the process dying before we return the dentry */
2051 if (tid_fd_revalidate(dentry, NULL))
444ceed8 2052 error = NULL;
1da177e4 2053
444ceed8
EB		 2054 out:
2055 return error;
2056out_unlock:
ca99c1da 2057 spin_unlock(&files->file_lock);
1da177e4 2058 put_files_struct(files);
444ceed8 2059out_iput:
1da177e4 2060 iput(inode);
cd6a3ce9 2061 goto out;
1da177e4
LT		 2062}
2063
27932742
MS		 2064static struct dentry *proc_lookupfd_common(struct inode *dir,
2065 struct dentry *dentry,
2066 instantiate_t instantiate)
444ceed8
EB		 2067{
2068 struct task_struct *task = get_proc_task(dir);
2069 unsigned fd = name_to_int(dentry);
2070 struct dentry *result = ERR_PTR(-ENOENT);
2071
2072 if (!task)
2073 goto out_no_task;
2074 if (fd == ~0U)
2075 goto out;
2076
27932742 2077 result = instantiate(dir, dentry, task, &fd);
444ceed8
EB		 2078out:
2079 put_task_struct(task);
2080out_no_task:
2081 return result;
2082}
2083
27932742
MS		 2084static int proc_readfd_common(struct file * filp, void * dirent,
2085 filldir_t filldir, instantiate_t instantiate)
28a6d671 2086{
2fddfeef 2087 struct dentry *dentry = filp->f_path.dentry;
28a6d671
EB		 2088 struct inode *inode = dentry->d_inode;
2089 struct task_struct *p = get_proc_task(inode);
457c2510 2090 unsigned int fd, ino;
28a6d671 2091 int retval;
28a6d671 2092 struct files_struct * files;
1da177e4 2093
28a6d671
EB		 2094 retval = -ENOENT;
2095 if (!p)
2096 goto out_no_task;
2097 retval = 0;
28a6d671
EB		 2098
2099 fd = filp->f_pos;
2100 switch (fd) {
2101 case 0:
2102 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
2103 goto out;
2104 filp->f_pos++;
2105 case 1:
2106 ino = parent_ino(dentry);
2107 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2108 goto out;
2109 filp->f_pos++;
2110 default:
2111 files = get_files_struct(p);
2112 if (!files)
2113 goto out;
2114 rcu_read_lock();
28a6d671 2115 for (fd = filp->f_pos-2;
9b4f526c 2116 fd < files_fdtable(files)->max_fds;
28a6d671 2117 fd++, filp->f_pos++) {
27932742
MS		 2118 char name[PROC_NUMBUF];
2119 int len;
28a6d671
EB		 2120
2121 if (!fcheck_files(files, fd))
2122 continue;
2123 rcu_read_unlock();
2124
27932742
MS		 2125 len = snprintf(name, sizeof(name), "%d", fd);
2126 if (proc_fill_cache(filp, dirent, filldir,
2127 name, len, instantiate,
2128 p, &fd) < 0) {
28a6d671
EB		 2129 rcu_read_lock();
2130 break;
2131 }
2132 rcu_read_lock();
2133 }
2134 rcu_read_unlock();
2135 put_files_struct(files);
2136 }
2137out:
2138 put_task_struct(p);
2139out_no_task:
2140 return retval;
2141}
2142
27932742
MS		 2143static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
2144 struct nameidata *nd)
2145{
2146 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
2147}
2148
2149static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
2150{
2151 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
2152}
2153
2154static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
2155 size_t len, loff_t *ppos)
2156{
2157 char tmp[PROC_FDINFO_MAX];
3dcd25f3 2158 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
27932742
MS		 2159 if (!err)
2160 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
2161 return err;
2162}
2163
2164static const struct file_operations proc_fdinfo_file_operations = {
87df8424 2165 .open = nonseekable_open,
27932742 2166 .read = proc_fdinfo_read,
6038f373 2167 .llseek = no_llseek,
27932742
MS		 2168};
2169
00977a59 2170static const struct file_operations proc_fd_operations = {
28a6d671
EB		 2171 .read = generic_read_dir,
2172 .readdir = proc_readfd,
6038f373 2173 .llseek = default_llseek,
1da177e4
LT		 2174};
2175
8948e11f
AD		 2176/*
2177 * /proc/pid/fd needs a special permission handler so that a process can still
2178 * access /proc/self/fd after it has executed a setuid().
2179 */
e6305c43 2180static int proc_fd_permission(struct inode *inode, int mask)
8948e11f
AD		 2181{
2182 int rv;
2183
2184 rv = generic_permission(inode, mask, NULL);
2185 if (rv == 0)
2186 return 0;
2187 if (task_pid(current) == proc_pid(inode))
2188 rv = 0;
2189 return rv;
2190}
2191
1da177e4
LT		 2192/*
2193 * proc directories can do almost nothing..
2194 */
c5ef1c42 2195static const struct inode_operations proc_fd_inode_operations = {
1da177e4 2196 .lookup = proc_lookupfd,
8948e11f 2197 .permission = proc_fd_permission,
6d76fa58 2198 .setattr = proc_setattr,
1da177e4
LT		 2199};
2200
27932742
MS		 2201static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2202 struct dentry *dentry, struct task_struct *task, const void *ptr)
2203{
2204 unsigned fd = *(unsigned *)ptr;
2205 struct inode *inode;
2206 struct proc_inode *ei;
2207 struct dentry *error = ERR_PTR(-ENOENT);
2208
2209 inode = proc_pid_make_inode(dir->i_sb, task);
2210 if (!inode)
2211 goto out;
2212 ei = PROC_I(inode);
2213 ei->fd = fd;
2214 inode->i_mode = S_IFREG | S_IRUSR;
2215 inode->i_fop = &proc_fdinfo_file_operations;
2216 dentry->d_op = &tid_fd_dentry_operations;
2217 d_add(dentry, inode);
2218 /* Close the race of the process dying before we return the dentry */
2219 if (tid_fd_revalidate(dentry, NULL))
2220 error = NULL;
2221
2222 out:
2223 return error;
2224}
2225
2226static struct dentry *proc_lookupfdinfo(struct inode *dir,
2227 struct dentry *dentry,
2228 struct nameidata *nd)
2229{
2230 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2231}
2232
2233static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2234{
2235 return proc_readfd_common(filp, dirent, filldir,
2236 proc_fdinfo_instantiate);
2237}
2238
2239static const struct file_operations proc_fdinfo_operations = {
2240 .read = generic_read_dir,
2241 .readdir = proc_readfdinfo,
6038f373 2242 .llseek = default_llseek,
27932742
MS		 2243};
2244
2245/*
2246 * proc directories can do almost nothing..
2247 */
2248static const struct inode_operations proc_fdinfo_inode_operations = {
2249 .lookup = proc_lookupfdinfo,
2250 .setattr = proc_setattr,
2251};
2252
2253
444ceed8 2254static struct dentry *proc_pident_instantiate(struct inode *dir,
c5141e6d 2255 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8 2256{
c5141e6d 2257 const struct pid_entry *p = ptr;
444ceed8
EB		 2258 struct inode *inode;
2259 struct proc_inode *ei;
bd6daba9 2260 struct dentry *error = ERR_PTR(-ENOENT);
444ceed8 2261
61a28784 2262 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 2263 if (!inode)
2264 goto out;
2265
2266 ei = PROC_I(inode);
2267 inode->i_mode = p->mode;
2268 if (S_ISDIR(inode->i_mode))
2269 inode->i_nlink = 2; /* Use getattr to fix if necessary */
2270 if (p->iop)
2271 inode->i_op = p->iop;
2272 if (p->fop)
2273 inode->i_fop = p->fop;
2274 ei->op = p->op;
2275 dentry->d_op = &pid_dentry_operations;
2276 d_add(dentry, inode);
2277 /* Close the race of the process dying before we return the dentry */
2278 if (pid_revalidate(dentry, NULL))
2279 error = NULL;
2280out:
2281 return error;
2282}
2283
1da177e4
LT		 2284static struct dentry *proc_pident_lookup(struct inode *dir,
2285 struct dentry *dentry,
c5141e6d 2286 const struct pid_entry *ents,
7bcd6b0e 2287 unsigned int nents)
1da177e4 2288{
cd6a3ce9 2289 struct dentry *error;
99f89551 2290 struct task_struct *task = get_proc_task(dir);
c5141e6d 2291 const struct pid_entry *p, *last;
1da177e4 2292
cd6a3ce9 2293 error = ERR_PTR(-ENOENT);
1da177e4 2294
99f89551
EB		 2295 if (!task)
2296 goto out_no_task;
1da177e4 2297
20cdc894
EB		 2298 /*
2299 * Yes, it does not scale. And it should not. Don't add
2300 * new entries into /proc/<tgid>/ without very good reasons.
2301 */
7bcd6b0e
EB		 2302 last = &ents[nents - 1];
2303 for (p = ents; p <= last; p++) {
1da177e4
LT		 2304 if (p->len != dentry->d_name.len)
2305 continue;
2306 if (!memcmp(dentry->d_name.name, p->name, p->len))
2307 break;
2308 }
7bcd6b0e 2309 if (p > last)
1da177e4
LT		 2310 goto out;
2311
444ceed8 2312 error = proc_pident_instantiate(dir, dentry, task, p);
1da177e4 2313out:
99f89551
EB		 2314 put_task_struct(task);
2315out_no_task:
cd6a3ce9 2316 return error;
1da177e4
LT		 2317}
2318
c5141e6d
ED		 2319static int proc_pident_fill_cache(struct file *filp, void *dirent,
2320 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
61a28784
EB		 2321{
2322 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2323 proc_pident_instantiate, task, p);
2324}
2325
28a6d671
EB		 2326static int proc_pident_readdir(struct file *filp,
2327 void *dirent, filldir_t filldir,
c5141e6d 2328 const struct pid_entry *ents, unsigned int nents)
28a6d671
EB		 2329{
2330 int i;
2fddfeef 2331 struct dentry *dentry = filp->f_path.dentry;
28a6d671
EB		 2332 struct inode *inode = dentry->d_inode;
2333 struct task_struct *task = get_proc_task(inode);
c5141e6d 2334 const struct pid_entry *p, *last;
28a6d671
EB		 2335 ino_t ino;
2336 int ret;
2337
2338 ret = -ENOENT;
2339 if (!task)
61a28784 2340 goto out_no_task;
28a6d671
EB		 2341
2342 ret = 0;
28a6d671
EB		 2343 i = filp->f_pos;
2344 switch (i) {
2345 case 0:
2346 ino = inode->i_ino;
2347 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2348 goto out;
2349 i++;
2350 filp->f_pos++;
2351 /* fall through */
2352 case 1:
2353 ino = parent_ino(dentry);
2354 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2355 goto out;
2356 i++;
2357 filp->f_pos++;
2358 /* fall through */
2359 default:
2360 i -= 2;
2361 if (i >= nents) {
2362 ret = 1;
2363 goto out;
2364 }
2365 p = ents + i;
7bcd6b0e
EB		 2366 last = &ents[nents - 1];
2367 while (p <= last) {
61a28784 2368 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
28a6d671
EB		 2369 goto out;
2370 filp->f_pos++;
2371 p++;
2372 }
2373 }
2374
2375 ret = 1;
2376out:
61a28784
EB		 2377 put_task_struct(task);
2378out_no_task:
28a6d671 2379 return ret;
1da177e4
LT		 2380}
2381
28a6d671
EB		 2382#ifdef CONFIG_SECURITY
2383static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2384 size_t count, loff_t *ppos)
2385{
2fddfeef 2386 struct inode * inode = file->f_path.dentry->d_inode;
04ff9708 2387 char *p = NULL;
28a6d671
EB		 2388 ssize_t length;
2389 struct task_struct *task = get_proc_task(inode);
2390
28a6d671 2391 if (!task)
04ff9708 2392 return -ESRCH;
28a6d671
EB		 2393
2394 length = security_getprocattr(task,
2fddfeef 2395 (char*)file->f_path.dentry->d_name.name,
04ff9708 2396 &p);
28a6d671 2397 put_task_struct(task);
04ff9708
AV		 2398 if (length > 0)
2399 length = simple_read_from_buffer(buf, count, ppos, p, length);
2400 kfree(p);
28a6d671 2401 return length;
1da177e4
LT		 2402}
2403
28a6d671
EB		 2404static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2405 size_t count, loff_t *ppos)
2406{
2fddfeef 2407 struct inode * inode = file->f_path.dentry->d_inode;
28a6d671
EB		 2408 char *page;
2409 ssize_t length;
2410 struct task_struct *task = get_proc_task(inode);
2411
2412 length = -ESRCH;
2413 if (!task)
2414 goto out_no_task;
2415 if (count > PAGE_SIZE)
2416 count = PAGE_SIZE;
2417
2418 /* No partial writes. */
2419 length = -EINVAL;
2420 if (*ppos != 0)
2421 goto out;
2422
2423 length = -ENOMEM;
e12ba74d 2424 page = (char*)__get_free_page(GFP_TEMPORARY);
28a6d671
EB		 2425 if (!page)
2426 goto out;
2427
2428 length = -EFAULT;
2429 if (copy_from_user(page, buf, count))
2430 goto out_free;
2431
107db7c7 2432 /* Guard against adverse ptrace interaction */
9b1bf12d 2433 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
107db7c7
DH		 2434 if (length < 0)
2435 goto out_free;
2436
28a6d671 2437 length = security_setprocattr(task,
2fddfeef 2438 (char*)file->f_path.dentry->d_name.name,
28a6d671 2439 (void*)page, count);
9b1bf12d 2440 mutex_unlock(&task->signal->cred_guard_mutex);
28a6d671
EB		 2441out_free:
2442 free_page((unsigned long) page);
2443out:
2444 put_task_struct(task);
2445out_no_task:
2446 return length;
2447}
2448
00977a59 2449static const struct file_operations proc_pid_attr_operations = {
28a6d671
EB		 2450 .read = proc_pid_attr_read,
2451 .write = proc_pid_attr_write,
87df8424 2452 .llseek = generic_file_llseek,
28a6d671
EB		 2453};
2454
c5141e6d 2455static const struct pid_entry attr_dir_stuff[] = {
631f9c18
AD		 2456 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2457 REG("prev", S_IRUGO, proc_pid_attr_operations),
2458 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2459 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2460 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2461 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
28a6d671
EB		 2462};
2463
72d9dcfc 2464static int proc_attr_dir_readdir(struct file * filp,
28a6d671
EB		 2465 void * dirent, filldir_t filldir)
2466{
2467 return proc_pident_readdir(filp,dirent,filldir,
72d9dcfc 2468 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2469}
2470
00977a59 2471static const struct file_operations proc_attr_dir_operations = {
1da177e4 2472 .read = generic_read_dir,
72d9dcfc 2473 .readdir = proc_attr_dir_readdir,
6038f373 2474 .llseek = default_llseek,
1da177e4
LT		 2475};
2476
72d9dcfc 2477static struct dentry *proc_attr_dir_lookup(struct inode *dir,
28a6d671
EB		 2478 struct dentry *dentry, struct nameidata *nd)
2479{
7bcd6b0e
EB		 2480 return proc_pident_lookup(dir, dentry,
2481 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2482}
2483
c5ef1c42 2484static const struct inode_operations proc_attr_dir_inode_operations = {
72d9dcfc 2485 .lookup = proc_attr_dir_lookup,
99f89551 2486 .getattr = pid_getattr,
6d76fa58 2487 .setattr = proc_setattr,
1da177e4
LT		 2488};
2489
28a6d671
EB		 2490#endif
2491
698ba7b5 2492#ifdef CONFIG_ELF_CORE
3cb4a0bb
KH		 2493static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2494 size_t count, loff_t *ppos)
2495{
2496 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2497 struct mm_struct *mm;
2498 char buffer[PROC_NUMBUF];
2499 size_t len;
2500 int ret;
2501
2502 if (!task)
2503 return -ESRCH;
2504
2505 ret = 0;
2506 mm = get_task_mm(task);
2507 if (mm) {
2508 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2509 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2510 MMF_DUMP_FILTER_SHIFT));
2511 mmput(mm);
2512 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2513 }
2514
2515 put_task_struct(task);
2516
2517 return ret;
2518}
2519
2520static ssize_t proc_coredump_filter_write(struct file *file,
2521 const char __user *buf,
2522 size_t count,
2523 loff_t *ppos)
2524{
2525 struct task_struct *task;
2526 struct mm_struct *mm;
2527 char buffer[PROC_NUMBUF], *end;
2528 unsigned int val;
2529 int ret;
2530 int i;
2531 unsigned long mask;
2532
2533 ret = -EFAULT;
2534 memset(buffer, 0, sizeof(buffer));
2535 if (count > sizeof(buffer) - 1)
2536 count = sizeof(buffer) - 1;
2537 if (copy_from_user(buffer, buf, count))
2538 goto out_no_task;
2539
2540 ret = -EINVAL;
2541 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2542 if (*end == '\n')
2543 end++;
2544 if (end - buffer == 0)
2545 goto out_no_task;
2546
2547 ret = -ESRCH;
2548 task = get_proc_task(file->f_dentry->d_inode);
2549 if (!task)
2550 goto out_no_task;
2551
2552 ret = end - buffer;
2553 mm = get_task_mm(task);
2554 if (!mm)
2555 goto out_no_mm;
2556
2557 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2558 if (val & mask)
2559 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2560 else
2561 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2562 }
2563
2564 mmput(mm);
2565 out_no_mm:
2566 put_task_struct(task);
2567 out_no_task:
2568 return ret;
2569}
2570
2571static const struct file_operations proc_coredump_filter_operations = {
2572 .read = proc_coredump_filter_read,
2573 .write = proc_coredump_filter_write,
87df8424 2574 .llseek = generic_file_llseek,
3cb4a0bb
KH		 2575};
2576#endif
2577
28a6d671
EB		 2578/*
2579 * /proc/self:
2580 */
2581static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2582 int buflen)
2583{
488e5bc4 2584 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
b55fcb22 2585 pid_t tgid = task_tgid_nr_ns(current, ns);
28a6d671 2586 char tmp[PROC_NUMBUF];
b55fcb22 2587 if (!tgid)
488e5bc4 2588 return -ENOENT;
b55fcb22 2589 sprintf(tmp, "%d", tgid);
28a6d671
EB		 2590 return vfs_readlink(dentry,buffer,buflen,tmp);
2591}
2592
2593static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2594{
488e5bc4 2595 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
b55fcb22 2596 pid_t tgid = task_tgid_nr_ns(current, ns);
7fee4868
AV		 2597 char *name = ERR_PTR(-ENOENT);
2598 if (tgid) {
2599 name = __getname();
2600 if (!name)
2601 name = ERR_PTR(-ENOMEM);
2602 else
2603 sprintf(name, "%d", tgid);
2604 }
2605 nd_set_link(nd, name);
2606 return NULL;
2607}
2608
2609static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2610 void *cookie)
2611{
2612 char *s = nd_get_link(nd);
2613 if (!IS_ERR(s))
2614 __putname(s);
28a6d671
EB		 2615}
2616
c5ef1c42 2617static const struct inode_operations proc_self_inode_operations = {
28a6d671
EB		 2618 .readlink = proc_self_readlink,
2619 .follow_link = proc_self_follow_link,
7fee4868 2620 .put_link = proc_self_put_link,
28a6d671
EB		 2621};
2622
801199ce
EB		 2623/*
2624 * proc base
2625 *
2626 * These are the directory entries in the root directory of /proc
2627 * that properly belong to the /proc filesystem, as they describe
2628 * describe something that is process related.
2629 */
c5141e6d 2630static const struct pid_entry proc_base_stuff[] = {
61a28784 2631 NOD("self", S_IFLNK|S_IRWXUGO,
801199ce 2632 &proc_self_inode_operations, NULL, {}),
801199ce
EB		 2633};
2634
2635/*
2636 * Exceptional case: normally we are not allowed to unhash a busy
2637 * directory. In this case, however, we can do it - no aliasing problems
2638 * due to the way we treat inodes.
2639 */
2640static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
2641{
2642 struct inode *inode = dentry->d_inode;
2643 struct task_struct *task = get_proc_task(inode);
2644 if (task) {
2645 put_task_struct(task);
2646 return 1;
2647 }
2648 d_drop(dentry);
2649 return 0;
2650}
2651
d72f71eb 2652static const struct dentry_operations proc_base_dentry_operations =
801199ce
EB		 2653{
2654 .d_revalidate = proc_base_revalidate,
2655 .d_delete = pid_delete_dentry,
2656};
2657
444ceed8 2658static struct dentry *proc_base_instantiate(struct inode *dir,
c5141e6d 2659 struct dentry *dentry, struct task_struct *task, const void *ptr)
801199ce 2660{
c5141e6d 2661 const struct pid_entry *p = ptr;
801199ce 2662 struct inode *inode;
801199ce 2663 struct proc_inode *ei;
73d36460 2664 struct dentry *error;
801199ce
EB		 2665
2666 /* Allocate the inode */
2667 error = ERR_PTR(-ENOMEM);
2668 inode = new_inode(dir->i_sb);
2669 if (!inode)
2670 goto out;
2671
2672 /* Initialize the inode */
2673 ei = PROC_I(inode);
85fe4025 2674 inode->i_ino = get_next_ino();
801199ce 2675 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
801199ce
EB		 2676
2677 /*
2678 * grab the reference to the task.
2679 */
1a657f78 2680 ei->pid = get_task_pid(task, PIDTYPE_PID);
801199ce
EB		 2681 if (!ei->pid)
2682 goto out_iput;
2683
801199ce
EB		 2684 inode->i_mode = p->mode;
2685 if (S_ISDIR(inode->i_mode))
2686 inode->i_nlink = 2;
2687 if (S_ISLNK(inode->i_mode))
2688 inode->i_size = 64;
2689 if (p->iop)
2690 inode->i_op = p->iop;
2691 if (p->fop)
2692 inode->i_fop = p->fop;
2693 ei->op = p->op;
2694 dentry->d_op = &proc_base_dentry_operations;
2695 d_add(dentry, inode);
2696 error = NULL;
2697out:
801199ce
EB		 2698 return error;
2699out_iput:
2700 iput(inode);
2701 goto out;
2702}
2703
444ceed8
EB		 2704static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2705{
2706 struct dentry *error;
2707 struct task_struct *task = get_proc_task(dir);
c5141e6d 2708 const struct pid_entry *p, *last;
444ceed8
EB		 2709
2710 error = ERR_PTR(-ENOENT);
2711
2712 if (!task)
2713 goto out_no_task;
2714
2715 /* Lookup the directory entry */
7bcd6b0e
EB		 2716 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2717 for (p = proc_base_stuff; p <= last; p++) {
444ceed8
EB		 2718 if (p->len != dentry->d_name.len)
2719 continue;
2720 if (!memcmp(dentry->d_name.name, p->name, p->len))
2721 break;
2722 }
7bcd6b0e 2723 if (p > last)
444ceed8
EB		 2724 goto out;
2725
2726 error = proc_base_instantiate(dir, dentry, task, p);
2727
2728out:
2729 put_task_struct(task);
2730out_no_task:
2731 return error;
2732}
2733
c5141e6d
ED		 2734static int proc_base_fill_cache(struct file *filp, void *dirent,
2735 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
61a28784
EB		 2736{
2737 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2738 proc_base_instantiate, task, p);
2739}
2740
aba76fdb 2741#ifdef CONFIG_TASK_IO_ACCOUNTING
297c5d92
AR		 2742static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2743{
940389b8 2744 struct task_io_accounting acct = task->ioac;
5995477a
AR		 2745 unsigned long flags;
2746
2747 if (whole && lock_task_sighand(task, &flags)) {
2748 struct task_struct *t = task;
2749
2750 task_io_accounting_add(&acct, &task->signal->ioac);
2751 while_each_thread(task, t)
2752 task_io_accounting_add(&acct, &t->ioac);
2753
2754 unlock_task_sighand(task, &flags);
297c5d92 2755 }
aba76fdb
AM		 2756 return sprintf(buffer,
2757 "rchar: %llu\n"
2758 "wchar: %llu\n"
2759 "syscr: %llu\n"
2760 "syscw: %llu\n"
2761 "read_bytes: %llu\n"
2762 "write_bytes: %llu\n"
2763 "cancelled_write_bytes: %llu\n",
7c44319d
AB		 2764 (unsigned long long)acct.rchar,
2765 (unsigned long long)acct.wchar,
2766 (unsigned long long)acct.syscr,
2767 (unsigned long long)acct.syscw,
2768 (unsigned long long)acct.read_bytes,
2769 (unsigned long long)acct.write_bytes,
2770 (unsigned long long)acct.cancelled_write_bytes);
297c5d92
AR		 2771}
2772
2773static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2774{
2775 return do_io_accounting(task, buffer, 0);
aba76fdb 2776}
297c5d92
AR		 2777
2778static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2779{
2780 return do_io_accounting(task, buffer, 1);
2781}
2782#endif /* CONFIG_TASK_IO_ACCOUNTING */
aba76fdb 2783
47830723
KC		 2784static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2785 struct pid *pid, struct task_struct *task)
2786{
2787 seq_printf(m, "%08x\n", task->personality);
2788 return 0;
2789}
2790
28a6d671
EB		 2791/*
2792 * Thread groups
2793 */
00977a59 2794static const struct file_operations proc_task_operations;
c5ef1c42 2795static const struct inode_operations proc_task_inode_operations;
20cdc894 2796
c5141e6d 2797static const struct pid_entry tgid_base_stuff[] = {
631f9c18
AD		 2798 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2799 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2800 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
b2211a36 2801#ifdef CONFIG_NET
631f9c18 2802 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
b2211a36 2803#endif
631f9c18
AD		 2804 REG("environ", S_IRUSR, proc_environ_operations),
2805 INF("auxv", S_IRUSR, proc_pid_auxv),
2806 ONE("status", S_IRUGO, proc_pid_status),
2807 ONE("personality", S_IRUSR, proc_pid_personality),
3036e7b4 2808 INF("limits", S_IRUGO, proc_pid_limits),
43ae34cb 2809#ifdef CONFIG_SCHED_DEBUG
631f9c18 2810 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
5091faa4
MG		 2811#endif
2812#ifdef CONFIG_SCHED_AUTOGROUP
2813 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
ebcb6734 2814#endif
4614a696 2815 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 2816#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
631f9c18 2817 INF("syscall", S_IRUSR, proc_pid_syscall),
43ae34cb 2818#endif
631f9c18
AD		 2819 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2820 ONE("stat", S_IRUGO, proc_tgid_stat),
2821 ONE("statm", S_IRUGO, proc_pid_statm),
2822 REG("maps", S_IRUGO, proc_maps_operations),
28a6d671 2823#ifdef CONFIG_NUMA
631f9c18 2824 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
28a6d671 2825#endif
631f9c18
AD		 2826 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2827 LNK("cwd", proc_cwd_link),
2828 LNK("root", proc_root_link),
2829 LNK("exe", proc_exe_link),
2830 REG("mounts", S_IRUGO, proc_mounts_operations),
2831 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2832 REG("mountstats", S_IRUSR, proc_mountstats_operations),
1e883281 2833#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18
AD		 2834 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2835 REG("smaps", S_IRUGO, proc_smaps_operations),
2836 REG("pagemap", S_IRUSR, proc_pagemap_operations),
28a6d671
EB		 2837#endif
2838#ifdef CONFIG_SECURITY
631f9c18 2839 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 2840#endif
2841#ifdef CONFIG_KALLSYMS
631f9c18 2842 INF("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 2843#endif
2ec220e2
KC		 2844#ifdef CONFIG_STACKTRACE
2845 ONE("stack", S_IRUSR, proc_pid_stack),
28a6d671
EB		 2846#endif
2847#ifdef CONFIG_SCHEDSTATS
631f9c18 2848 INF("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 2849#endif
9745512c 2850#ifdef CONFIG_LATENCYTOP
631f9c18 2851 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 2852#endif
8793d854 2853#ifdef CONFIG_PROC_PID_CPUSET
631f9c18 2854 REG("cpuset", S_IRUGO, proc_cpuset_operations),
a424316c
PM		 2855#endif
2856#ifdef CONFIG_CGROUPS
631f9c18 2857 REG("cgroup", S_IRUGO, proc_cgroup_operations),
28a6d671 2858#endif
631f9c18
AD		 2859 INF("oom_score", S_IRUGO, proc_oom_score),
2860 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
a63d83f4 2861 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 2862#ifdef CONFIG_AUDITSYSCALL
631f9c18
AD		 2863 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2864 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 2865#endif
f4f154fd 2866#ifdef CONFIG_FAULT_INJECTION
631f9c18 2867 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 2868#endif
698ba7b5 2869#ifdef CONFIG_ELF_CORE
631f9c18 2870 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3cb4a0bb 2871#endif
aba76fdb 2872#ifdef CONFIG_TASK_IO_ACCOUNTING
631f9c18 2873 INF("io", S_IRUGO, proc_tgid_io_accounting),
aba76fdb 2874#endif
28a6d671 2875};
1da177e4 2876
28a6d671 2877static int proc_tgid_base_readdir(struct file * filp,
1da177e4
LT		 2878 void * dirent, filldir_t filldir)
2879{
2880 return proc_pident_readdir(filp,dirent,filldir,
28a6d671 2881 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 2882}
2883
00977a59 2884static const struct file_operations proc_tgid_base_operations = {
1da177e4 2885 .read = generic_read_dir,
28a6d671 2886 .readdir = proc_tgid_base_readdir,
6038f373 2887 .llseek = default_llseek,
1da177e4
LT		 2888};
2889
28a6d671 2890static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
7bcd6b0e
EB		 2891 return proc_pident_lookup(dir, dentry,
2892 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 2893}
2894
c5ef1c42 2895static const struct inode_operations proc_tgid_base_inode_operations = {
28a6d671 2896 .lookup = proc_tgid_base_lookup,
99f89551 2897 .getattr = pid_getattr,
6d76fa58 2898 .setattr = proc_setattr,
1da177e4 2899};
1da177e4 2900
60347f67 2901static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
1da177e4 2902{
48e6484d 2903 struct dentry *dentry, *leader, *dir;
8578cea7 2904 char buf[PROC_NUMBUF];
48e6484d
EB		 2905 struct qstr name;
2906
2907 name.name = buf;
60347f67
PE		 2908 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2909 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d 2910 if (dentry) {
29f12ca3 2911 shrink_dcache_parent(dentry);
48e6484d
EB		 2912 d_drop(dentry);
2913 dput(dentry);
2914 }
1da177e4 2915
48e6484d 2916 name.name = buf;
60347f67
PE		 2917 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2918 leader = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d
EB		 2919 if (!leader)
2920 goto out;
1da177e4 2921
48e6484d
EB		 2922 name.name = "task";
2923 name.len = strlen(name.name);
2924 dir = d_hash_and_lookup(leader, &name);
2925 if (!dir)
2926 goto out_put_leader;
2927
2928 name.name = buf;
60347f67 2929 name.len = snprintf(buf, sizeof(buf), "%d", pid);
48e6484d
EB		 2930 dentry = d_hash_and_lookup(dir, &name);
2931 if (dentry) {
2932 shrink_dcache_parent(dentry);
2933 d_drop(dentry);
2934 dput(dentry);
1da177e4 2935 }
48e6484d
EB		 2936
2937 dput(dir);
2938out_put_leader:
2939 dput(leader);
2940out:
2941 return;
1da177e4
LT		 2942}
2943
0895e91d
RD		 2944/**
2945 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2946 * @task: task that should be flushed.
2947 *
2948 * When flushing dentries from proc, one needs to flush them from global
60347f67 2949 * proc (proc_mnt) and from all the namespaces' procs this task was seen
0895e91d
RD		 2950 * in. This call is supposed to do all of this job.
2951 *
2952 * Looks in the dcache for
2953 * /proc/@pid
2954 * /proc/@tgid/task/@pid
2955 * if either directory is present flushes it and all of it'ts children
2956 * from the dcache.
2957 *
2958 * It is safe and reasonable to cache /proc entries for a task until
2959 * that task exits. After that they just clog up the dcache with
2960 * useless entries, possibly causing useful dcache entries to be
2961 * flushed instead. This routine is proved to flush those useless
2962 * dcache entries at process exit time.
2963 *
2964 * NOTE: This routine is just an optimization so it does not guarantee
2965 * that no dcache entries will exist at process exit time it
2966 * just makes it very unlikely that any will persist.
60347f67
PE		 2967 */
2968
2969void proc_flush_task(struct task_struct *task)
2970{
9fcc2d15 2971 int i;
9b4d1cbe 2972 struct pid *pid, *tgid;
130f77ec
PE		 2973 struct upid *upid;
2974
130f77ec 2975 pid = task_pid(task);
9b4d1cbe 2976 tgid = task_tgid(task);
130f77ec 2977
9fcc2d15 2978 for (i = 0; i <= pid->level; i++) {
130f77ec
PE		 2979 upid = &pid->numbers[i];
2980 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
9b4d1cbe 2981 tgid->numbers[i].nr);
130f77ec 2982 }
6f4e6433
PE		 2983
2984 upid = &pid->numbers[pid->level];
2985 if (upid->nr == 1)
2986 pid_ns_release_proc(upid->ns);
60347f67
PE		 2987}
2988
9711ef99
AB		 2989static struct dentry *proc_pid_instantiate(struct inode *dir,
2990 struct dentry * dentry,
c5141e6d 2991 struct task_struct *task, const void *ptr)
444ceed8
EB		 2992{
2993 struct dentry *error = ERR_PTR(-ENOENT);
2994 struct inode *inode;
2995
61a28784 2996 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 2997 if (!inode)
2998 goto out;
2999
3000 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3001 inode->i_op = &proc_tgid_base_inode_operations;
3002 inode->i_fop = &proc_tgid_base_operations;
3003 inode->i_flags|=S_IMMUTABLE;
aed54175
VN		 3004
3005 inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
3006 ARRAY_SIZE(tgid_base_stuff));
444ceed8
EB		 3007
3008 dentry->d_op = &pid_dentry_operations;
3009
3010 d_add(dentry, inode);
3011 /* Close the race of the process dying before we return the dentry */
3012 if (pid_revalidate(dentry, NULL))
3013 error = NULL;
3014out:
3015 return error;
3016}
3017
1da177e4
LT		 3018struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3019{
73d36460 3020 struct dentry *result;
1da177e4 3021 struct task_struct *task;
1da177e4 3022 unsigned tgid;
b488893a 3023 struct pid_namespace *ns;
1da177e4 3024
801199ce
EB		 3025 result = proc_base_lookup(dir, dentry);
3026 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3027 goto out;
3028
1da177e4
LT		 3029 tgid = name_to_int(dentry);
3030 if (tgid == ~0U)
3031 goto out;
3032
b488893a 3033 ns = dentry->d_sb->s_fs_info;
de758734 3034 rcu_read_lock();
b488893a 3035 task = find_task_by_pid_ns(tgid, ns);
1da177e4
LT		 3036 if (task)
3037 get_task_struct(task);
de758734 3038 rcu_read_unlock();
1da177e4
LT		 3039 if (!task)
3040 goto out;
3041
444ceed8 3042 result = proc_pid_instantiate(dir, dentry, task, NULL);
1da177e4 3043 put_task_struct(task);
1da177e4 3044out:
cd6a3ce9 3045 return result;
1da177e4
LT		 3046}
3047
1da177e4 3048/*
0804ef4b 3049 * Find the first task with tgid >= tgid
0bc58a91 3050 *
1da177e4 3051 */
19fd4bb2
EB		 3052struct tgid_iter {
3053 unsigned int tgid;
0804ef4b 3054 struct task_struct *task;
19fd4bb2
EB		 3055};
3056static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3057{
0804ef4b 3058 struct pid *pid;
1da177e4 3059
19fd4bb2
EB		 3060 if (iter.task)
3061 put_task_struct(iter.task);
454cc105 3062 rcu_read_lock();
0804ef4b 3063retry:
19fd4bb2
EB		 3064 iter.task = NULL;
3065 pid = find_ge_pid(iter.tgid, ns);
0804ef4b 3066 if (pid) {
19fd4bb2
EB		 3067 iter.tgid = pid_nr_ns(pid, ns);
3068 iter.task = pid_task(pid, PIDTYPE_PID);
0804ef4b
EB		 3069 /* What we to know is if the pid we have find is the
3070 * pid of a thread_group_leader. Testing for task
3071 * being a thread_group_leader is the obvious thing
3072 * todo but there is a window when it fails, due to
3073 * the pid transfer logic in de_thread.
3074 *
3075 * So we perform the straight forward test of seeing
3076 * if the pid we have found is the pid of a thread
3077 * group leader, and don't worry if the task we have
3078 * found doesn't happen to be a thread group leader.
3079 * As we don't care in the case of readdir.
3080 */
19fd4bb2
EB		 3081 if (!iter.task || !has_group_leader_pid(iter.task)) {
3082 iter.tgid += 1;
0804ef4b 3083 goto retry;
19fd4bb2
EB		 3084 }
3085 get_task_struct(iter.task);
0bc58a91 3086 }
454cc105 3087 rcu_read_unlock();
19fd4bb2 3088 return iter;
1da177e4
LT		 3089}
3090
7bcd6b0e 3091#define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
0804ef4b 3092
61a28784 3093static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
19fd4bb2 3094 struct tgid_iter iter)
61a28784
EB		 3095{
3096 char name[PROC_NUMBUF];
19fd4bb2 3097 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
61a28784 3098 return proc_fill_cache(filp, dirent, filldir, name, len,
19fd4bb2 3099 proc_pid_instantiate, iter.task, NULL);
61a28784
EB		 3100}
3101
1da177e4
LT		 3102/* for the /proc/ directory itself, after non-process stuff has been done */
3103int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3104{
1da177e4 3105 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2fddfeef 3106 struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
19fd4bb2 3107 struct tgid_iter iter;
b488893a 3108 struct pid_namespace *ns;
1da177e4 3109
61a28784
EB		 3110 if (!reaper)
3111 goto out_no_task;
3112
7bcd6b0e 3113 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
c5141e6d 3114 const struct pid_entry *p = &proc_base_stuff[nr];
61a28784 3115 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
801199ce 3116 goto out;
1da177e4
LT		 3117 }
3118
b488893a 3119 ns = filp->f_dentry->d_sb->s_fs_info;
19fd4bb2
EB		 3120 iter.task = NULL;
3121 iter.tgid = filp->f_pos - TGID_OFFSET;
3122 for (iter = next_tgid(ns, iter);
3123 iter.task;
3124 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3125 filp->f_pos = iter.tgid + TGID_OFFSET;
3126 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
3127 put_task_struct(iter.task);
0804ef4b 3128 goto out;
1da177e4 3129 }
0bc58a91 3130 }
0804ef4b
EB		 3131 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3132out:
61a28784
EB		 3133 put_task_struct(reaper);
3134out_no_task:
0bc58a91
EB		 3135 return 0;
3136}
1da177e4 3137
28a6d671
EB		 3138/*
3139 * Tasks
3140 */
c5141e6d 3141static const struct pid_entry tid_base_stuff[] = {
631f9c18 3142 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3835541d 3143 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
631f9c18
AD		 3144 REG("environ", S_IRUSR, proc_environ_operations),
3145 INF("auxv", S_IRUSR, proc_pid_auxv),
3146 ONE("status", S_IRUGO, proc_pid_status),
3147 ONE("personality", S_IRUSR, proc_pid_personality),
3036e7b4 3148 INF("limits", S_IRUGO, proc_pid_limits),
43ae34cb 3149#ifdef CONFIG_SCHED_DEBUG
631f9c18 3150 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
ebcb6734 3151#endif
4614a696 3152 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 3153#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
631f9c18 3154 INF("syscall", S_IRUSR, proc_pid_syscall),
43ae34cb 3155#endif
631f9c18
AD		 3156 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3157 ONE("stat", S_IRUGO, proc_tid_stat),
3158 ONE("statm", S_IRUGO, proc_pid_statm),
3159 REG("maps", S_IRUGO, proc_maps_operations),
28a6d671 3160#ifdef CONFIG_NUMA
631f9c18 3161 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
28a6d671 3162#endif
631f9c18
AD		 3163 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3164 LNK("cwd", proc_cwd_link),
3165 LNK("root", proc_root_link),
3166 LNK("exe", proc_exe_link),
3167 REG("mounts", S_IRUGO, proc_mounts_operations),
3168 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
1e883281 3169#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18
AD		 3170 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3171 REG("smaps", S_IRUGO, proc_smaps_operations),
3172 REG("pagemap", S_IRUSR, proc_pagemap_operations),
28a6d671
EB		 3173#endif
3174#ifdef CONFIG_SECURITY
631f9c18 3175 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 3176#endif
3177#ifdef CONFIG_KALLSYMS
631f9c18 3178 INF("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 3179#endif
2ec220e2
KC		 3180#ifdef CONFIG_STACKTRACE
3181 ONE("stack", S_IRUSR, proc_pid_stack),
28a6d671
EB		 3182#endif
3183#ifdef CONFIG_SCHEDSTATS
631f9c18 3184 INF("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 3185#endif
9745512c 3186#ifdef CONFIG_LATENCYTOP
631f9c18 3187 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 3188#endif
8793d854 3189#ifdef CONFIG_PROC_PID_CPUSET
631f9c18 3190 REG("cpuset", S_IRUGO, proc_cpuset_operations),
a424316c
PM		 3191#endif
3192#ifdef CONFIG_CGROUPS
631f9c18 3193 REG("cgroup", S_IRUGO, proc_cgroup_operations),
28a6d671 3194#endif
631f9c18
AD		 3195 INF("oom_score", S_IRUGO, proc_oom_score),
3196 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
a63d83f4 3197 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 3198#ifdef CONFIG_AUDITSYSCALL
631f9c18
AD		 3199 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3200 REG("sessionid", S_IRUSR, proc_sessionid_operations),
28a6d671 3201#endif
f4f154fd 3202#ifdef CONFIG_FAULT_INJECTION
631f9c18 3203 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 3204#endif
297c5d92 3205#ifdef CONFIG_TASK_IO_ACCOUNTING
631f9c18 3206 INF("io", S_IRUGO, proc_tid_io_accounting),
297c5d92 3207#endif
28a6d671
EB		 3208};
3209
3210static int proc_tid_base_readdir(struct file * filp,
3211 void * dirent, filldir_t filldir)
3212{
3213 return proc_pident_readdir(filp,dirent,filldir,
3214 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3215}
3216
3217static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
7bcd6b0e
EB		 3218 return proc_pident_lookup(dir, dentry,
3219 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
28a6d671
EB		 3220}
3221
00977a59 3222static const struct file_operations proc_tid_base_operations = {
28a6d671
EB		 3223 .read = generic_read_dir,
3224 .readdir = proc_tid_base_readdir,
6038f373 3225 .llseek = default_llseek,
28a6d671
EB		 3226};
3227
c5ef1c42 3228static const struct inode_operations proc_tid_base_inode_operations = {
28a6d671
EB		 3229 .lookup = proc_tid_base_lookup,
3230 .getattr = pid_getattr,
3231 .setattr = proc_setattr,
3232};
3233
444ceed8 3234static struct dentry *proc_task_instantiate(struct inode *dir,
c5141e6d 3235 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8
EB		 3236{
3237 struct dentry *error = ERR_PTR(-ENOENT);
3238 struct inode *inode;
61a28784 3239 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 3240
3241 if (!inode)
3242 goto out;
3243 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3244 inode->i_op = &proc_tid_base_inode_operations;
3245 inode->i_fop = &proc_tid_base_operations;
3246 inode->i_flags|=S_IMMUTABLE;
aed54175
VN		 3247
3248 inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
3249 ARRAY_SIZE(tid_base_stuff));
444ceed8
EB		 3250
3251 dentry->d_op = &pid_dentry_operations;
3252
3253 d_add(dentry, inode);
3254 /* Close the race of the process dying before we return the dentry */
3255 if (pid_revalidate(dentry, NULL))
3256 error = NULL;
3257out:
3258 return error;
3259}
3260
28a6d671
EB		 3261static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3262{
3263 struct dentry *result = ERR_PTR(-ENOENT);
3264 struct task_struct *task;
3265 struct task_struct *leader = get_proc_task(dir);
28a6d671 3266 unsigned tid;
b488893a 3267 struct pid_namespace *ns;
28a6d671
EB		 3268
3269 if (!leader)
3270 goto out_no_task;
3271
3272 tid = name_to_int(dentry);
3273 if (tid == ~0U)
3274 goto out;
3275
b488893a 3276 ns = dentry->d_sb->s_fs_info;
28a6d671 3277 rcu_read_lock();
b488893a 3278 task = find_task_by_pid_ns(tid, ns);
28a6d671
EB		 3279 if (task)
3280 get_task_struct(task);
3281 rcu_read_unlock();
3282 if (!task)
3283 goto out;
bac0abd6 3284 if (!same_thread_group(leader, task))
28a6d671
EB		 3285 goto out_drop_task;
3286
444ceed8 3287 result = proc_task_instantiate(dir, dentry, task, NULL);
28a6d671
EB		 3288out_drop_task:
3289 put_task_struct(task);
3290out:
3291 put_task_struct(leader);
3292out_no_task:
3293 return result;
3294}
3295
0bc58a91
EB		 3296/*
3297 * Find the first tid of a thread group to return to user space.
3298 *
3299 * Usually this is just the thread group leader, but if the users
3300 * buffer was too small or there was a seek into the middle of the
3301 * directory we have more work todo.
3302 *
3303 * In the case of a short read we start with find_task_by_pid.
3304 *
3305 * In the case of a seek we start with the leader and walk nr
3306 * threads past it.
3307 */
cc288738 3308static struct task_struct *first_tid(struct task_struct *leader,
b488893a 3309 int tid, int nr, struct pid_namespace *ns)
0bc58a91 3310{
a872ff0c 3311 struct task_struct *pos;
1da177e4 3312
cc288738 3313 rcu_read_lock();
0bc58a91
EB		 3314 /* Attempt to start with the pid of a thread */
3315 if (tid && (nr > 0)) {
b488893a 3316 pos = find_task_by_pid_ns(tid, ns);
a872ff0c
ON		 3317 if (pos && (pos->group_leader == leader))
3318 goto found;
0bc58a91 3319 }
1da177e4 3320
0bc58a91 3321 /* If nr exceeds the number of threads there is nothing todo */
a872ff0c
ON		 3322 pos = NULL;
3323 if (nr && nr >= get_nr_threads(leader))
3324 goto out;
1da177e4 3325
a872ff0c
ON		 3326 /* If we haven't found our starting place yet start
3327 * with the leader and walk nr threads forward.
0bc58a91 3328 */
a872ff0c
ON		 3329 for (pos = leader; nr > 0; --nr) {
3330 pos = next_thread(pos);
3331 if (pos == leader) {
3332 pos = NULL;
3333 goto out;
3334 }
1da177e4 3335 }
a872ff0c
ON		 3336found:
3337 get_task_struct(pos);
3338out:
cc288738 3339 rcu_read_unlock();
0bc58a91
EB		 3340 return pos;
3341}
3342
3343/*
3344 * Find the next thread in the thread list.
3345 * Return NULL if there is an error or no next thread.
3346 *
3347 * The reference to the input task_struct is released.
3348 */
3349static struct task_struct *next_tid(struct task_struct *start)
3350{
c1df7fb8 3351 struct task_struct *pos = NULL;
cc288738 3352 rcu_read_lock();
c1df7fb8 3353 if (pid_alive(start)) {
0bc58a91 3354 pos = next_thread(start);
c1df7fb8
ON		 3355 if (thread_group_leader(pos))
3356 pos = NULL;
3357 else
3358 get_task_struct(pos);
3359 }
cc288738 3360 rcu_read_unlock();
0bc58a91
EB		 3361 put_task_struct(start);
3362 return pos;
1da177e4
LT		 3363}
3364
61a28784
EB		 3365static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3366 struct task_struct *task, int tid)
3367{
3368 char name[PROC_NUMBUF];
3369 int len = snprintf(name, sizeof(name), "%d", tid);
3370 return proc_fill_cache(filp, dirent, filldir, name, len,
3371 proc_task_instantiate, task, NULL);
3372}
3373
1da177e4
LT		 3374/* for the /proc/TGID/task/ directories */
3375static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3376{
2fddfeef 3377 struct dentry *dentry = filp->f_path.dentry;
1da177e4 3378 struct inode *inode = dentry->d_inode;
7d895244 3379 struct task_struct *leader = NULL;
0bc58a91 3380 struct task_struct *task;
1da177e4
LT		 3381 int retval = -ENOENT;
3382 ino_t ino;
0bc58a91 3383 int tid;
b488893a 3384 struct pid_namespace *ns;
1da177e4 3385
7d895244
GC		 3386 task = get_proc_task(inode);
3387 if (!task)
3388 goto out_no_task;
3389 rcu_read_lock();
3390 if (pid_alive(task)) {
3391 leader = task->group_leader;
3392 get_task_struct(leader);
3393 }
3394 rcu_read_unlock();
3395 put_task_struct(task);
99f89551
EB		 3396 if (!leader)
3397 goto out_no_task;
1da177e4
LT		 3398 retval = 0;
3399
ee568b25 3400 switch ((unsigned long)filp->f_pos) {
1da177e4
LT		 3401 case 0:
3402 ino = inode->i_ino;
ee6f779b 3403 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
1da177e4 3404 goto out;
ee6f779b 3405 filp->f_pos++;
1da177e4
LT		 3406 /* fall through */
3407 case 1:
3408 ino = parent_ino(dentry);
ee6f779b 3409 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
1da177e4 3410 goto out;
ee6f779b 3411 filp->f_pos++;
1da177e4
LT		 3412 /* fall through */
3413 }
3414
0bc58a91
EB		 3415 /* f_version caches the tgid value that the last readdir call couldn't
3416 * return. lseek aka telldir automagically resets f_version to 0.
3417 */
b488893a 3418 ns = filp->f_dentry->d_sb->s_fs_info;
2b47c361 3419 tid = (int)filp->f_version;
0bc58a91 3420 filp->f_version = 0;
ee6f779b 3421 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
0bc58a91 3422 task;
ee6f779b 3423 task = next_tid(task), filp->f_pos++) {
b488893a 3424 tid = task_pid_nr_ns(task, ns);
61a28784 3425 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
0bc58a91
EB		 3426 /* returning this tgid failed, save it as the first
3427 * pid for the next readir call */
2b47c361 3428 filp->f_version = (u64)tid;
0bc58a91 3429 put_task_struct(task);
1da177e4 3430 break;
0bc58a91 3431 }
1da177e4
LT		 3432 }
3433out:
99f89551
EB		 3434 put_task_struct(leader);
3435out_no_task:
1da177e4
LT		 3436 return retval;
3437}
6e66b52b
EB		 3438
3439static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3440{
3441 struct inode *inode = dentry->d_inode;
99f89551 3442 struct task_struct *p = get_proc_task(inode);
6e66b52b
EB		 3443 generic_fillattr(inode, stat);
3444
99f89551 3445 if (p) {
99f89551 3446 stat->nlink += get_nr_threads(p);
99f89551 3447 put_task_struct(p);
6e66b52b
EB		 3448 }
3449
3450 return 0;
3451}
28a6d671 3452
c5ef1c42 3453static const struct inode_operations proc_task_inode_operations = {
28a6d671
EB		 3454 .lookup = proc_task_lookup,
3455 .getattr = proc_task_getattr,
3456 .setattr = proc_setattr,
3457};
3458
00977a59 3459static const struct file_operations proc_task_operations = {
28a6d671
EB		 3460 .read = generic_read_dir,
3461 .readdir = proc_task_readdir,
6038f373 3462 .llseek = default_llseek,
28a6d671 3463};
Linux kernel - Deliverables
This page took 0.772221 seconds and 5 git commands to generate.