]> git.ipfire.org Git - thirdparty/kernel/stable.git/blame - fs/proc/base.c
projects / thirdparty / kernel / stable.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
oom: kill all threads sharing oom killed task's mm
[thirdparty/kernel/stable.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
229 if (mutex_lock_killable(&task->cred_guard_mutex))
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 }
704b836c
ON		 238 mutex_unlock(&task->cred_guard_mutex);
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);
774 return 0;
775}
776
777static ssize_t mem_read(struct file * file, char __user * buf,
778 size_t count, loff_t *ppos)
779{
2fddfeef 780 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1da177e4
LT		 781 char *page;
782 unsigned long src = *ppos;
783 int ret = -ESRCH;
784 struct mm_struct *mm;
785
99f89551
EB		 786 if (!task)
787 goto out_no_task;
788
638fa202 789 if (check_mem_permission(task))
1da177e4
LT		 790 goto out;
791
792 ret = -ENOMEM;
e12ba74d 793 page = (char *)__get_free_page(GFP_TEMPORARY);
1da177e4
LT		 794 if (!page)
795 goto out;
796
797 ret = 0;
798
799 mm = get_task_mm(task);
800 if (!mm)
801 goto out_free;
802
803 ret = -EIO;
804
805 if (file->private_data != (void*)((long)current->self_exec_id))
806 goto out_put;
807
808 ret = 0;
809
810 while (count > 0) {
811 int this_len, retval;
812
813 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
814 retval = access_process_vm(task, src, page, this_len, 0);
638fa202 815 if (!retval || check_mem_permission(task)) {
1da177e4
LT		 816 if (!ret)
817 ret = -EIO;
818 break;
819 }
820
821 if (copy_to_user(buf, page, retval)) {
822 ret = -EFAULT;
823 break;
824 }
825
826 ret += retval;
827 src += retval;
828 buf += retval;
829 count -= retval;
830 }
831 *ppos = src;
832
833out_put:
834 mmput(mm);
835out_free:
836 free_page((unsigned long) page);
837out:
99f89551
EB		 838 put_task_struct(task);
839out_no_task:
1da177e4
LT		 840 return ret;
841}
842
843#define mem_write NULL
844
845#ifndef mem_write
846/* This is a security hazard */
63967fa9 847static ssize_t mem_write(struct file * file, const char __user *buf,
1da177e4
LT		 848 size_t count, loff_t *ppos)
849{
f7ca54f4 850 int copied;
1da177e4 851 char *page;
2fddfeef 852 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1da177e4
LT		 853 unsigned long dst = *ppos;
854
99f89551
EB		 855 copied = -ESRCH;
856 if (!task)
857 goto out_no_task;
858
638fa202 859 if (check_mem_permission(task))
99f89551 860 goto out;
1da177e4 861
99f89551 862 copied = -ENOMEM;
e12ba74d 863 page = (char *)__get_free_page(GFP_TEMPORARY);
1da177e4 864 if (!page)
99f89551 865 goto out;
1da177e4 866
f7ca54f4 867 copied = 0;
1da177e4
LT		 868 while (count > 0) {
869 int this_len, retval;
870
871 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
872 if (copy_from_user(page, buf, this_len)) {
873 copied = -EFAULT;
874 break;
875 }
876 retval = access_process_vm(task, dst, page, this_len, 1);
877 if (!retval) {
878 if (!copied)
879 copied = -EIO;
880 break;
881 }
882 copied += retval;
883 buf += retval;
884 dst += retval;
885 count -= retval;
886 }
887 *ppos = dst;
888 free_page((unsigned long) page);
99f89551
EB		 889out:
890 put_task_struct(task);
891out_no_task:
1da177e4
LT		 892 return copied;
893}
894#endif
895
85863e47 896loff_t mem_lseek(struct file *file, loff_t offset, int orig)
1da177e4
LT		 897{
898 switch (orig) {
899 case 0:
900 file->f_pos = offset;
901 break;
902 case 1:
903 file->f_pos += offset;
904 break;
905 default:
906 return -EINVAL;
907 }
908 force_successful_syscall_return();
909 return file->f_pos;
910}
911
00977a59 912static const struct file_operations proc_mem_operations = {
1da177e4
LT		 913 .llseek = mem_lseek,
914 .read = mem_read,
915 .write = mem_write,
916 .open = mem_open,
917};
918
315e28c8
JP		 919static ssize_t environ_read(struct file *file, char __user *buf,
920 size_t count, loff_t *ppos)
921{
922 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
923 char *page;
924 unsigned long src = *ppos;
925 int ret = -ESRCH;
926 struct mm_struct *mm;
927
928 if (!task)
929 goto out_no_task;
930
006ebb40 931 if (!ptrace_may_access(task, PTRACE_MODE_READ))
315e28c8
JP		 932 goto out;
933
934 ret = -ENOMEM;
935 page = (char *)__get_free_page(GFP_TEMPORARY);
936 if (!page)
937 goto out;
938
939 ret = 0;
940
941 mm = get_task_mm(task);
942 if (!mm)
943 goto out_free;
944
945 while (count > 0) {
946 int this_len, retval, max_len;
947
948 this_len = mm->env_end - (mm->env_start + src);
949
950 if (this_len <= 0)
951 break;
952
953 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
954 this_len = (this_len > max_len) ? max_len : this_len;
955
956 retval = access_process_vm(task, (mm->env_start + src),
957 page, this_len, 0);
958
959 if (retval <= 0) {
960 ret = retval;
961 break;
962 }
963
964 if (copy_to_user(buf, page, retval)) {
965 ret = -EFAULT;
966 break;
967 }
968
969 ret += retval;
970 src += retval;
971 buf += retval;
972 count -= retval;
973 }
974 *ppos = src;
975
976 mmput(mm);
977out_free:
978 free_page((unsigned long) page);
979out:
980 put_task_struct(task);
981out_no_task:
982 return ret;
983}
984
985static const struct file_operations proc_environ_operations = {
986 .read = environ_read,
87df8424 987 .llseek = generic_file_llseek,
315e28c8
JP		 988};
989
1da177e4
LT		 990static ssize_t oom_adjust_read(struct file *file, char __user *buf,
991 size_t count, loff_t *ppos)
992{
2fddfeef 993 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
8578cea7 994 char buffer[PROC_NUMBUF];
1da177e4 995 size_t len;
28b83c51
KM		 996 int oom_adjust = OOM_DISABLE;
997 unsigned long flags;
1da177e4 998
99f89551
EB		 999 if (!task)
1000 return -ESRCH;
28b83c51
KM		 1001
1002 if (lock_task_sighand(task, &flags)) {
1003 oom_adjust = task->signal->oom_adj;
1004 unlock_task_sighand(task, &flags);
1005 }
1006
99f89551
EB		 1007 put_task_struct(task);
1008
8578cea7 1009 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
0c28f287
AM		 1010
1011 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1da177e4
LT		 1012}
1013
1014static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1015 size_t count, loff_t *ppos)
1016{
99f89551 1017 struct task_struct *task;
5d863b89
KM		 1018 char buffer[PROC_NUMBUF];
1019 long oom_adjust;
28b83c51 1020 unsigned long flags;
5d863b89 1021 int err;
1da177e4 1022
8578cea7
EB		 1023 memset(buffer, 0, sizeof(buffer));
1024 if (count > sizeof(buffer) - 1)
1025 count = sizeof(buffer) - 1;
1da177e4
LT		 1026 if (copy_from_user(buffer, buf, count))
1027 return -EFAULT;
5d863b89
KM		 1028
1029 err = strict_strtol(strstrip(buffer), 0, &oom_adjust);
1030 if (err)
1031 return -EINVAL;
8ac773b4
AD		 1032 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
1033 oom_adjust != OOM_DISABLE)
1da177e4 1034 return -EINVAL;
5d863b89 1035
2fddfeef 1036 task = get_proc_task(file->f_path.dentry->d_inode);
99f89551
EB		 1037 if (!task)
1038 return -ESRCH;
28b83c51
KM		 1039 if (!lock_task_sighand(task, &flags)) {
1040 put_task_struct(task);
1041 return -ESRCH;
1042 }
1043
1044 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
1045 unlock_task_sighand(task, &flags);
8fb4fc68
GJ		 1046 put_task_struct(task);
1047 return -EACCES;
1048 }
28b83c51 1049
3d5992d2
YH		 1050 task_lock(task);
1051 if (!task->mm) {
1052 task_unlock(task);
1053 unlock_task_sighand(task, &flags);
1054 put_task_struct(task);
1055 return -EINVAL;
1056 }
1057
1058 if (oom_adjust != task->signal->oom_adj) {
1059 if (oom_adjust == OOM_DISABLE)
1060 atomic_inc(&task->mm->oom_disable_count);
1061 if (task->signal->oom_adj == OOM_DISABLE)
1062 atomic_dec(&task->mm->oom_disable_count);
1063 }
1064
51b1bd2a
DR		 1065 /*
1066 * Warn that /proc/pid/oom_adj is deprecated, see
1067 * Documentation/feature-removal-schedule.txt.
1068 */
1069 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, "
1070 "please use /proc/%d/oom_score_adj instead.\n",
1071 current->comm, task_pid_nr(current),
1072 task_pid_nr(task), task_pid_nr(task));
28b83c51 1073 task->signal->oom_adj = oom_adjust;
a63d83f4
DR		 1074 /*
1075 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1076 * value is always attainable.
1077 */
1078 if (task->signal->oom_adj == OOM_ADJUST_MAX)
1079 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
1080 else
1081 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
1082 -OOM_DISABLE;
3d5992d2 1083 task_unlock(task);
28b83c51 1084 unlock_task_sighand(task, &flags);
99f89551 1085 put_task_struct(task);
5d863b89
KM		 1086
1087 return count;
1da177e4
LT		 1088}
1089
00977a59 1090static const struct file_operations proc_oom_adjust_operations = {
1da177e4
LT		 1091 .read = oom_adjust_read,
1092 .write = oom_adjust_write,
87df8424 1093 .llseek = generic_file_llseek,
1da177e4
LT		 1094};
1095
a63d83f4
DR		 1096static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1097 size_t count, loff_t *ppos)
1098{
1099 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1100 char buffer[PROC_NUMBUF];
1101 int oom_score_adj = OOM_SCORE_ADJ_MIN;
1102 unsigned long flags;
1103 size_t len;
1104
1105 if (!task)
1106 return -ESRCH;
1107 if (lock_task_sighand(task, &flags)) {
1108 oom_score_adj = task->signal->oom_score_adj;
1109 unlock_task_sighand(task, &flags);
1110 }
1111 put_task_struct(task);
1112 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
1113 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1114}
1115
1116static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1117 size_t count, loff_t *ppos)
1118{
1119 struct task_struct *task;
1120 char buffer[PROC_NUMBUF];
1121 unsigned long flags;
1122 long oom_score_adj;
1123 int err;
1124
1125 memset(buffer, 0, sizeof(buffer));
1126 if (count > sizeof(buffer) - 1)
1127 count = sizeof(buffer) - 1;
1128 if (copy_from_user(buffer, buf, count))
1129 return -EFAULT;
1130
1131 err = strict_strtol(strstrip(buffer), 0, &oom_score_adj);
1132 if (err)
1133 return -EINVAL;
1134 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1135 oom_score_adj > OOM_SCORE_ADJ_MAX)
1136 return -EINVAL;
1137
1138 task = get_proc_task(file->f_path.dentry->d_inode);
1139 if (!task)
1140 return -ESRCH;
1141 if (!lock_task_sighand(task, &flags)) {
1142 put_task_struct(task);
1143 return -ESRCH;
1144 }
1145 if (oom_score_adj < task->signal->oom_score_adj &&
1146 !capable(CAP_SYS_RESOURCE)) {
1147 unlock_task_sighand(task, &flags);
1148 put_task_struct(task);
1149 return -EACCES;
1150 }
1151
3d5992d2
YH		 1152 task_lock(task);
1153 if (!task->mm) {
1154 task_unlock(task);
1155 unlock_task_sighand(task, &flags);
1156 put_task_struct(task);
1157 return -EINVAL;
1158 }
1159 if (oom_score_adj != task->signal->oom_score_adj) {
1160 if (oom_score_adj == OOM_SCORE_ADJ_MIN)
1161 atomic_inc(&task->mm->oom_disable_count);
1162 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1163 atomic_dec(&task->mm->oom_disable_count);
1164 }
a63d83f4
DR		 1165 task->signal->oom_score_adj = oom_score_adj;
1166 /*
1167 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1168 * always attainable.
1169 */
1170 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1171 task->signal->oom_adj = OOM_DISABLE;
1172 else
1173 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1174 OOM_SCORE_ADJ_MAX;
3d5992d2 1175 task_unlock(task);
a63d83f4
DR		 1176 unlock_task_sighand(task, &flags);
1177 put_task_struct(task);
1178 return count;
1179}
1180
1181static const struct file_operations proc_oom_score_adj_operations = {
1182 .read = oom_score_adj_read,
1183 .write = oom_score_adj_write,
6038f373 1184 .llseek = default_llseek,
a63d83f4
DR		 1185};
1186
1da177e4
LT		 1187#ifdef CONFIG_AUDITSYSCALL
1188#define TMPBUFLEN 21
1189static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1190 size_t count, loff_t *ppos)
1191{
2fddfeef 1192 struct inode * inode = file->f_path.dentry->d_inode;
99f89551 1193 struct task_struct *task = get_proc_task(inode);
1da177e4
LT		 1194 ssize_t length;
1195 char tmpbuf[TMPBUFLEN];
1196
99f89551
EB		 1197 if (!task)
1198 return -ESRCH;
1da177e4 1199 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
0c11b942 1200 audit_get_loginuid(task));
99f89551 1201 put_task_struct(task);
1da177e4
LT		 1202 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1203}
1204
1205static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1206 size_t count, loff_t *ppos)
1207{
2fddfeef 1208 struct inode * inode = file->f_path.dentry->d_inode;
1da177e4
LT		 1209 char *page, *tmp;
1210 ssize_t length;
1da177e4
LT		 1211 uid_t loginuid;
1212
1213 if (!capable(CAP_AUDIT_CONTROL))
1214 return -EPERM;
1215
7dc52157
PM		 1216 rcu_read_lock();
1217 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1218 rcu_read_unlock();
1da177e4 1219 return -EPERM;
7dc52157
PM		 1220 }
1221 rcu_read_unlock();
1da177e4 1222
e0182909
AV		 1223 if (count >= PAGE_SIZE)
1224 count = PAGE_SIZE - 1;
1da177e4
LT		 1225
1226 if (*ppos != 0) {
1227 /* No partial writes. */
1228 return -EINVAL;
1229 }
e12ba74d 1230 page = (char*)__get_free_page(GFP_TEMPORARY);
1da177e4
LT		 1231 if (!page)
1232 return -ENOMEM;
1233 length = -EFAULT;
1234 if (copy_from_user(page, buf, count))
1235 goto out_free_page;
1236
e0182909 1237 page[count] = '\0';
1da177e4
LT		 1238 loginuid = simple_strtoul(page, &tmp, 10);
1239 if (tmp == page) {
1240 length = -EINVAL;
1241 goto out_free_page;
1242
1243 }
99f89551 1244 length = audit_set_loginuid(current, loginuid);
1da177e4
LT		 1245 if (likely(length == 0))
1246 length = count;
1247
1248out_free_page:
1249 free_page((unsigned long) page);
1250 return length;
1251}
1252
00977a59 1253static const struct file_operations proc_loginuid_operations = {
1da177e4
LT		 1254 .read = proc_loginuid_read,
1255 .write = proc_loginuid_write,
87df8424 1256 .llseek = generic_file_llseek,
1da177e4 1257};
1e0bd755
EP		 1258
1259static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1260 size_t count, loff_t *ppos)
1261{
1262 struct inode * inode = file->f_path.dentry->d_inode;
1263 struct task_struct *task = get_proc_task(inode);
1264 ssize_t length;
1265 char tmpbuf[TMPBUFLEN];
1266
1267 if (!task)
1268 return -ESRCH;
1269 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1270 audit_get_sessionid(task));
1271 put_task_struct(task);
1272 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1273}
1274
1275static const struct file_operations proc_sessionid_operations = {
1276 .read = proc_sessionid_read,
87df8424 1277 .llseek = generic_file_llseek,
1e0bd755 1278};
1da177e4
LT		 1279#endif
1280
f4f154fd
AM		 1281#ifdef CONFIG_FAULT_INJECTION
1282static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1283 size_t count, loff_t *ppos)
1284{
1285 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1286 char buffer[PROC_NUMBUF];
1287 size_t len;
1288 int make_it_fail;
f4f154fd
AM		 1289
1290 if (!task)
1291 return -ESRCH;
1292 make_it_fail = task->make_it_fail;
1293 put_task_struct(task);
1294
1295 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
0c28f287
AM		 1296
1297 return simple_read_from_buffer(buf, count, ppos, buffer, len);
f4f154fd
AM		 1298}
1299
1300static ssize_t proc_fault_inject_write(struct file * file,
1301 const char __user * buf, size_t count, loff_t *ppos)
1302{
1303 struct task_struct *task;
1304 char buffer[PROC_NUMBUF], *end;
1305 int make_it_fail;
1306
1307 if (!capable(CAP_SYS_RESOURCE))
1308 return -EPERM;
1309 memset(buffer, 0, sizeof(buffer));
1310 if (count > sizeof(buffer) - 1)
1311 count = sizeof(buffer) - 1;
1312 if (copy_from_user(buffer, buf, count))
1313 return -EFAULT;
cba8aafe
VL		 1314 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1315 if (*end)
1316 return -EINVAL;
f4f154fd
AM		 1317 task = get_proc_task(file->f_dentry->d_inode);
1318 if (!task)
1319 return -ESRCH;
1320 task->make_it_fail = make_it_fail;
1321 put_task_struct(task);
cba8aafe
VL		 1322
1323 return count;
f4f154fd
AM		 1324}
1325
00977a59 1326static const struct file_operations proc_fault_inject_operations = {
f4f154fd
AM		 1327 .read = proc_fault_inject_read,
1328 .write = proc_fault_inject_write,
87df8424 1329 .llseek = generic_file_llseek,
f4f154fd
AM		 1330};
1331#endif
1332
9745512c 1333
43ae34cb
IM		 1334#ifdef CONFIG_SCHED_DEBUG
1335/*
1336 * Print out various scheduling related per-task fields:
1337 */
1338static int sched_show(struct seq_file *m, void *v)
1339{
1340 struct inode *inode = m->private;
1341 struct task_struct *p;
1342
43ae34cb
IM		 1343 p = get_proc_task(inode);
1344 if (!p)
1345 return -ESRCH;
1346 proc_sched_show_task(p, m);
1347
1348 put_task_struct(p);
1349
1350 return 0;
1351}
1352
1353static ssize_t
1354sched_write(struct file *file, const char __user *buf,
1355 size_t count, loff_t *offset)
1356{
1357 struct inode *inode = file->f_path.dentry->d_inode;
1358 struct task_struct *p;
1359
43ae34cb
IM		 1360 p = get_proc_task(inode);
1361 if (!p)
1362 return -ESRCH;
1363 proc_sched_set_task(p);
1364
1365 put_task_struct(p);
1366
1367 return count;
1368}
1369
1370static int sched_open(struct inode *inode, struct file *filp)
1371{
1372 int ret;
1373
1374 ret = single_open(filp, sched_show, NULL);
1375 if (!ret) {
1376 struct seq_file *m = filp->private_data;
1377
1378 m->private = inode;
1379 }
1380 return ret;
1381}
1382
1383static const struct file_operations proc_pid_sched_operations = {
1384 .open = sched_open,
1385 .read = seq_read,
1386 .write = sched_write,
1387 .llseek = seq_lseek,
5ea473a1 1388 .release = single_release,
43ae34cb
IM		 1389};
1390
1391#endif
1392
4614a696
JS		 1393static ssize_t comm_write(struct file *file, const char __user *buf,
1394 size_t count, loff_t *offset)
1395{
1396 struct inode *inode = file->f_path.dentry->d_inode;
1397 struct task_struct *p;
1398 char buffer[TASK_COMM_LEN];
1399
1400 memset(buffer, 0, sizeof(buffer));
1401 if (count > sizeof(buffer) - 1)
1402 count = sizeof(buffer) - 1;
1403 if (copy_from_user(buffer, buf, count))
1404 return -EFAULT;
1405
1406 p = get_proc_task(inode);
1407 if (!p)
1408 return -ESRCH;
1409
1410 if (same_thread_group(current, p))
1411 set_task_comm(p, buffer);
1412 else
1413 count = -EINVAL;
1414
1415 put_task_struct(p);
1416
1417 return count;
1418}
1419
1420static int comm_show(struct seq_file *m, void *v)
1421{
1422 struct inode *inode = m->private;
1423 struct task_struct *p;
1424
1425 p = get_proc_task(inode);
1426 if (!p)
1427 return -ESRCH;
1428
1429 task_lock(p);
1430 seq_printf(m, "%s\n", p->comm);
1431 task_unlock(p);
1432
1433 put_task_struct(p);
1434
1435 return 0;
1436}
1437
1438static int comm_open(struct inode *inode, struct file *filp)
1439{
1440 int ret;
1441
1442 ret = single_open(filp, comm_show, NULL);
1443 if (!ret) {
1444 struct seq_file *m = filp->private_data;
1445
1446 m->private = inode;
1447 }
1448 return ret;
1449}
1450
1451static const struct file_operations proc_pid_set_comm_operations = {
1452 .open = comm_open,
1453 .read = seq_read,
1454 .write = comm_write,
1455 .llseek = seq_lseek,
1456 .release = single_release,
1457};
1458
925d1c40
MH		 1459/*
1460 * We added or removed a vma mapping the executable. The vmas are only mapped
1461 * during exec and are not mapped with the mmap system call.
1462 * Callers must hold down_write() on the mm's mmap_sem for these
1463 */
1464void added_exe_file_vma(struct mm_struct *mm)
1465{
1466 mm->num_exe_file_vmas++;
1467}
1468
1469void removed_exe_file_vma(struct mm_struct *mm)
1470{
1471 mm->num_exe_file_vmas--;
1472 if ((mm->num_exe_file_vmas == 0) && mm->exe_file){
1473 fput(mm->exe_file);
1474 mm->exe_file = NULL;
1475 }
1476
1477}
1478
1479void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1480{
1481 if (new_exe_file)
1482 get_file(new_exe_file);
1483 if (mm->exe_file)
1484 fput(mm->exe_file);
1485 mm->exe_file = new_exe_file;
1486 mm->num_exe_file_vmas = 0;
1487}
1488
1489struct file *get_mm_exe_file(struct mm_struct *mm)
1490{
1491 struct file *exe_file;
1492
1493 /* We need mmap_sem to protect against races with removal of
1494 * VM_EXECUTABLE vmas */
1495 down_read(&mm->mmap_sem);
1496 exe_file = mm->exe_file;
1497 if (exe_file)
1498 get_file(exe_file);
1499 up_read(&mm->mmap_sem);
1500 return exe_file;
1501}
1502
1503void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
1504{
1505 /* It's safe to write the exe_file pointer without exe_file_lock because
1506 * this is called during fork when the task is not yet in /proc */
1507 newmm->exe_file = get_mm_exe_file(oldmm);
1508}
1509
1510static int proc_exe_link(struct inode *inode, struct path *exe_path)
1511{
1512 struct task_struct *task;
1513 struct mm_struct *mm;
1514 struct file *exe_file;
1515
1516 task = get_proc_task(inode);
1517 if (!task)
1518 return -ENOENT;
1519 mm = get_task_mm(task);
1520 put_task_struct(task);
1521 if (!mm)
1522 return -ENOENT;
1523 exe_file = get_mm_exe_file(mm);
1524 mmput(mm);
1525 if (exe_file) {
1526 *exe_path = exe_file->f_path;
1527 path_get(&exe_file->f_path);
1528 fput(exe_file);
1529 return 0;
1530 } else
1531 return -ENOENT;
1532}
1533
008b150a 1534static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1da177e4
LT		 1535{
1536 struct inode *inode = dentry->d_inode;
1537 int error = -EACCES;
1538
1539 /* We don't need a base pointer in the /proc filesystem */
1d957f9b 1540 path_put(&nd->path);
1da177e4 1541
778c1144
EB		 1542 /* Are we allowed to snoop on the tasks file descriptors? */
1543 if (!proc_fd_access_allowed(inode))
1da177e4 1544 goto out;
1da177e4 1545
3dcd25f3 1546 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
1da177e4 1547out:
008b150a 1548 return ERR_PTR(error);
1da177e4
LT		 1549}
1550
3dcd25f3 1551static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1da177e4 1552{
e12ba74d 1553 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
3dcd25f3 1554 char *pathname;
1da177e4
LT		 1555 int len;
1556
1557 if (!tmp)
1558 return -ENOMEM;
0c28f287 1559
8df9d1a4 1560 pathname = d_path_with_unreachable(path, tmp, PAGE_SIZE);
3dcd25f3
JB		 1561 len = PTR_ERR(pathname);
1562 if (IS_ERR(pathname))
1da177e4 1563 goto out;
3dcd25f3 1564 len = tmp + PAGE_SIZE - 1 - pathname;
1da177e4
LT		 1565
1566 if (len > buflen)
1567 len = buflen;
3dcd25f3 1568 if (copy_to_user(buffer, pathname, len))
1da177e4
LT		 1569 len = -EFAULT;
1570 out:
1571 free_page((unsigned long)tmp);
1572 return len;
1573}
1574
1575static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1576{
1577 int error = -EACCES;
1578 struct inode *inode = dentry->d_inode;
3dcd25f3 1579 struct path path;
1da177e4 1580
778c1144
EB		 1581 /* Are we allowed to snoop on the tasks file descriptors? */
1582 if (!proc_fd_access_allowed(inode))
1da177e4 1583 goto out;
1da177e4 1584
3dcd25f3 1585 error = PROC_I(inode)->op.proc_get_link(inode, &path);
1da177e4
LT		 1586 if (error)
1587 goto out;
1588
3dcd25f3
JB		 1589 error = do_proc_readlink(&path, buffer, buflen);
1590 path_put(&path);
1da177e4 1591out:
1da177e4
LT		 1592 return error;
1593}
1594
c5ef1c42 1595static const struct inode_operations proc_pid_link_inode_operations = {
1da177e4 1596 .readlink = proc_pid_readlink,
6d76fa58
LT		 1597 .follow_link = proc_pid_follow_link,
1598 .setattr = proc_setattr,
1da177e4
LT		 1599};
1600
28a6d671
EB		 1601
1602/* building an inode */
1603
1604static int task_dumpable(struct task_struct *task)
1da177e4 1605{
28a6d671
EB		 1606 int dumpable = 0;
1607 struct mm_struct *mm;
1da177e4 1608
28a6d671
EB		 1609 task_lock(task);
1610 mm = task->mm;
1611 if (mm)
6c5d5238 1612 dumpable = get_dumpable(mm);
28a6d671
EB		 1613 task_unlock(task);
1614 if(dumpable == 1)
1615 return 1;
1616 return 0;
1617}
1da177e4 1618
1da177e4 1619
61a28784 1620static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
28a6d671
EB		 1621{
1622 struct inode * inode;
1623 struct proc_inode *ei;
c69e8d9c 1624 const struct cred *cred;
1da177e4 1625
28a6d671 1626 /* We need a new inode */
1da177e4 1627
28a6d671
EB		 1628 inode = new_inode(sb);
1629 if (!inode)
1630 goto out;
1631
1632 /* Common stuff */
1633 ei = PROC_I(inode);
1634 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
28a6d671
EB		 1635 inode->i_op = &proc_def_inode_operations;
1636
1637 /*
1638 * grab the reference to task.
1639 */
1a657f78 1640 ei->pid = get_task_pid(task, PIDTYPE_PID);
28a6d671
EB		 1641 if (!ei->pid)
1642 goto out_unlock;
1643
28a6d671 1644 if (task_dumpable(task)) {
c69e8d9c
DH		 1645 rcu_read_lock();
1646 cred = __task_cred(task);
1647 inode->i_uid = cred->euid;
1648 inode->i_gid = cred->egid;
1649 rcu_read_unlock();
1da177e4 1650 }
28a6d671
EB		 1651 security_task_to_inode(task, inode);
1652
1da177e4 1653out:
28a6d671
EB		 1654 return inode;
1655
1656out_unlock:
1657 iput(inode);
1658 return NULL;
1da177e4
LT		 1659}
1660
28a6d671 1661static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1da177e4 1662{
1da177e4 1663 struct inode *inode = dentry->d_inode;
28a6d671 1664 struct task_struct *task;
c69e8d9c
DH		 1665 const struct cred *cred;
1666
28a6d671 1667 generic_fillattr(inode, stat);
1da177e4 1668
28a6d671
EB		 1669 rcu_read_lock();
1670 stat->uid = 0;
1671 stat->gid = 0;
1672 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1673 if (task) {
1674 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1675 task_dumpable(task)) {
c69e8d9c
DH		 1676 cred = __task_cred(task);
1677 stat->uid = cred->euid;
1678 stat->gid = cred->egid;
1da177e4
LT		 1679 }
1680 }
28a6d671 1681 rcu_read_unlock();
d6e71144 1682 return 0;
1da177e4
LT		 1683}
1684
1da177e4
LT		 1685/* dentry stuff */
1686
1687/*
1688 * Exceptional case: normally we are not allowed to unhash a busy
1689 * directory. In this case, however, we can do it - no aliasing problems
1690 * due to the way we treat inodes.
1691 *
1692 * Rewrite the inode's ownerships here because the owning task may have
1693 * performed a setuid(), etc.
99f89551
EB		 1694 *
1695 * Before the /proc/pid/status file was created the only way to read
1696 * the effective uid of a /process was to stat /proc/pid. Reading
1697 * /proc/pid/status is slow enough that procps and other packages
1698 * kept stating /proc/pid. To keep the rules in /proc simple I have
1699 * made this apply to all per process world readable and executable
1700 * directories.
1da177e4
LT		 1701 */
1702static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1703{
1704 struct inode *inode = dentry->d_inode;
99f89551 1705 struct task_struct *task = get_proc_task(inode);
c69e8d9c
DH		 1706 const struct cred *cred;
1707
99f89551
EB		 1708 if (task) {
1709 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1710 task_dumpable(task)) {
c69e8d9c
DH		 1711 rcu_read_lock();
1712 cred = __task_cred(task);
1713 inode->i_uid = cred->euid;
1714 inode->i_gid = cred->egid;
1715 rcu_read_unlock();
1da177e4
LT		 1716 } else {
1717 inode->i_uid = 0;
1718 inode->i_gid = 0;
1719 }
9ee8ab9f 1720 inode->i_mode &= ~(S_ISUID | S_ISGID);
1da177e4 1721 security_task_to_inode(task, inode);
99f89551 1722 put_task_struct(task);
1da177e4
LT		 1723 return 1;
1724 }
1725 d_drop(dentry);
1726 return 0;
1727}
1728
28a6d671 1729static int pid_delete_dentry(struct dentry * dentry)
99f89551 1730{
28a6d671
EB		 1731 /* Is the task we represent dead?
1732 * If so, then don't put the dentry on the lru list,
1733 * kill it immediately.
1734 */
1735 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1736}
1737
d72f71eb 1738static const struct dentry_operations pid_dentry_operations =
28a6d671
EB		 1739{
1740 .d_revalidate = pid_revalidate,
1741 .d_delete = pid_delete_dentry,
1742};
1743
1744/* Lookups */
1745
c5141e6d
ED		 1746typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1747 struct task_struct *, const void *);
61a28784 1748
1c0d04c9
EB		 1749/*
1750 * Fill a directory entry.
1751 *
1752 * If possible create the dcache entry and derive our inode number and
1753 * file type from dcache entry.
1754 *
1755 * Since all of the proc inode numbers are dynamically generated, the inode
1756 * numbers do not exist until the inode is cache. This means creating the
1757 * the dcache entry in readdir is necessary to keep the inode numbers
1758 * reported by readdir in sync with the inode numbers reported
1759 * by stat.
1760 */
61a28784
EB		 1761static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1762 char *name, int len,
c5141e6d 1763 instantiate_t instantiate, struct task_struct *task, const void *ptr)
61a28784 1764{
2fddfeef 1765 struct dentry *child, *dir = filp->f_path.dentry;
61a28784
EB		 1766 struct inode *inode;
1767 struct qstr qname;
1768 ino_t ino = 0;
1769 unsigned type = DT_UNKNOWN;
1770
1771 qname.name = name;
1772 qname.len = len;
1773 qname.hash = full_name_hash(name, len);
1774
1775 child = d_lookup(dir, &qname);
1776 if (!child) {
1777 struct dentry *new;
1778 new = d_alloc(dir, &qname);
1779 if (new) {
1780 child = instantiate(dir->d_inode, new, task, ptr);
1781 if (child)
1782 dput(new);
1783 else
1784 child = new;
1785 }
1786 }
1787 if (!child || IS_ERR(child) || !child->d_inode)
1788 goto end_instantiate;
1789 inode = child->d_inode;
1790 if (inode) {
1791 ino = inode->i_ino;
1792 type = inode->i_mode >> 12;
1793 }
1794 dput(child);
1795end_instantiate:
1796 if (!ino)
1797 ino = find_inode_number(dir, &qname);
1798 if (!ino)
1799 ino = 1;
1800 return filldir(dirent, name, len, filp->f_pos, ino, type);
1801}
1802
28a6d671
EB		 1803static unsigned name_to_int(struct dentry *dentry)
1804{
1805 const char *name = dentry->d_name.name;
1806 int len = dentry->d_name.len;
1807 unsigned n = 0;
1808
1809 if (len > 1 && *name == '0')
1810 goto out;
1811 while (len-- > 0) {
1812 unsigned c = *name++ - '0';
1813 if (c > 9)
1814 goto out;
1815 if (n >= (~0U-9)/10)
1816 goto out;
1817 n *= 10;
1818 n += c;
1819 }
1820 return n;
1821out:
1822 return ~0U;
1823}
1824
27932742
MS		 1825#define PROC_FDINFO_MAX 64
1826
3dcd25f3 1827static int proc_fd_info(struct inode *inode, struct path *path, char *info)
28a6d671
EB		 1828{
1829 struct task_struct *task = get_proc_task(inode);
1830 struct files_struct *files = NULL;
1831 struct file *file;
1832 int fd = proc_fd(inode);
99f89551 1833
99f89551 1834 if (task) {
28a6d671
EB		 1835 files = get_files_struct(task);
1836 put_task_struct(task);
1837 }
1838 if (files) {
1839 /*
1840 * We are not taking a ref to the file structure, so we must
1841 * hold ->file_lock.
1842 */
1843 spin_lock(&files->file_lock);
1844 file = fcheck_files(files, fd);
1845 if (file) {
3dcd25f3
JB		 1846 if (path) {
1847 *path = file->f_path;
1848 path_get(&file->f_path);
1849 }
27932742
MS		 1850 if (info)
1851 snprintf(info, PROC_FDINFO_MAX,
1852 "pos:\t%lli\n"
1853 "flags:\t0%o\n",
1854 (long long) file->f_pos,
1855 file->f_flags);
28a6d671
EB		 1856 spin_unlock(&files->file_lock);
1857 put_files_struct(files);
1858 return 0;
99f89551 1859 }
28a6d671
EB		 1860 spin_unlock(&files->file_lock);
1861 put_files_struct(files);
99f89551 1862 }
28a6d671 1863 return -ENOENT;
99f89551
EB		 1864}
1865
3dcd25f3 1866static int proc_fd_link(struct inode *inode, struct path *path)
27932742 1867{
3dcd25f3 1868 return proc_fd_info(inode, path, NULL);
27932742
MS		 1869}
1870
1da177e4
LT		 1871static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1872{
1873 struct inode *inode = dentry->d_inode;
99f89551 1874 struct task_struct *task = get_proc_task(inode);
aed7a6c4 1875 int fd = proc_fd(inode);
1da177e4 1876 struct files_struct *files;
c69e8d9c 1877 const struct cred *cred;
1da177e4 1878
99f89551
EB		 1879 if (task) {
1880 files = get_files_struct(task);
1881 if (files) {
1882 rcu_read_lock();
1883 if (fcheck_files(files, fd)) {
1884 rcu_read_unlock();
1885 put_files_struct(files);
1886 if (task_dumpable(task)) {
c69e8d9c
DH		 1887 rcu_read_lock();
1888 cred = __task_cred(task);
1889 inode->i_uid = cred->euid;
1890 inode->i_gid = cred->egid;
1891 rcu_read_unlock();
99f89551
EB		 1892 } else {
1893 inode->i_uid = 0;
1894 inode->i_gid = 0;
1895 }
9ee8ab9f 1896 inode->i_mode &= ~(S_ISUID | S_ISGID);
99f89551
EB		 1897 security_task_to_inode(task, inode);
1898 put_task_struct(task);
1899 return 1;
1900 }
b835996f 1901 rcu_read_unlock();
1da177e4 1902 put_files_struct(files);
1da177e4 1903 }
99f89551 1904 put_task_struct(task);
1da177e4
LT		 1905 }
1906 d_drop(dentry);
1907 return 0;
1908}
1909
d72f71eb 1910static const struct dentry_operations tid_fd_dentry_operations =
1da177e4
LT		 1911{
1912 .d_revalidate = tid_fd_revalidate,
1913 .d_delete = pid_delete_dentry,
1914};
1915
444ceed8 1916static struct dentry *proc_fd_instantiate(struct inode *dir,
c5141e6d 1917 struct dentry *dentry, struct task_struct *task, const void *ptr)
1da177e4 1918{
c5141e6d 1919 unsigned fd = *(const unsigned *)ptr;
444ceed8
EB		 1920 struct file *file;
1921 struct files_struct *files;
1922 struct inode *inode;
1923 struct proc_inode *ei;
1924 struct dentry *error = ERR_PTR(-ENOENT);
1da177e4 1925
61a28784 1926 inode = proc_pid_make_inode(dir->i_sb, task);
1da177e4
LT		 1927 if (!inode)
1928 goto out;
1929 ei = PROC_I(inode);
aed7a6c4 1930 ei->fd = fd;
1da177e4
LT		 1931 files = get_files_struct(task);
1932 if (!files)
444ceed8 1933 goto out_iput;
1da177e4 1934 inode->i_mode = S_IFLNK;
ca99c1da
DS		 1935
1936 /*
1937 * We are not taking a ref to the file structure, so we must
1938 * hold ->file_lock.
1939 */
1940 spin_lock(&files->file_lock);
1da177e4
LT		 1941 file = fcheck_files(files, fd);
1942 if (!file)
444ceed8 1943 goto out_unlock;
aeb5d727 1944 if (file->f_mode & FMODE_READ)
1da177e4 1945 inode->i_mode |= S_IRUSR | S_IXUSR;
aeb5d727 1946 if (file->f_mode & FMODE_WRITE)
1da177e4 1947 inode->i_mode |= S_IWUSR | S_IXUSR;
ca99c1da 1948 spin_unlock(&files->file_lock);
1da177e4 1949 put_files_struct(files);
444ceed8 1950
1da177e4
LT		 1951 inode->i_op = &proc_pid_link_inode_operations;
1952 inode->i_size = 64;
1953 ei->op.proc_get_link = proc_fd_link;
1954 dentry->d_op = &tid_fd_dentry_operations;
1955 d_add(dentry, inode);
cd6a3ce9
EB		 1956 /* Close the race of the process dying before we return the dentry */
1957 if (tid_fd_revalidate(dentry, NULL))
444ceed8 1958 error = NULL;
1da177e4 1959
444ceed8
EB		 1960 out:
1961 return error;
1962out_unlock:
ca99c1da 1963 spin_unlock(&files->file_lock);
1da177e4 1964 put_files_struct(files);
444ceed8 1965out_iput:
1da177e4 1966 iput(inode);
cd6a3ce9 1967 goto out;
1da177e4
LT		 1968}
1969
27932742
MS		 1970static struct dentry *proc_lookupfd_common(struct inode *dir,
1971 struct dentry *dentry,
1972 instantiate_t instantiate)
444ceed8
EB		 1973{
1974 struct task_struct *task = get_proc_task(dir);
1975 unsigned fd = name_to_int(dentry);
1976 struct dentry *result = ERR_PTR(-ENOENT);
1977
1978 if (!task)
1979 goto out_no_task;
1980 if (fd == ~0U)
1981 goto out;
1982
27932742 1983 result = instantiate(dir, dentry, task, &fd);
444ceed8
EB		 1984out:
1985 put_task_struct(task);
1986out_no_task:
1987 return result;
1988}
1989
27932742
MS		 1990static int proc_readfd_common(struct file * filp, void * dirent,
1991 filldir_t filldir, instantiate_t instantiate)
28a6d671 1992{
2fddfeef 1993 struct dentry *dentry = filp->f_path.dentry;
28a6d671
EB		 1994 struct inode *inode = dentry->d_inode;
1995 struct task_struct *p = get_proc_task(inode);
457c2510 1996 unsigned int fd, ino;
28a6d671 1997 int retval;
28a6d671 1998 struct files_struct * files;
1da177e4 1999
28a6d671
EB		 2000 retval = -ENOENT;
2001 if (!p)
2002 goto out_no_task;
2003 retval = 0;
28a6d671
EB		 2004
2005 fd = filp->f_pos;
2006 switch (fd) {
2007 case 0:
2008 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
2009 goto out;
2010 filp->f_pos++;
2011 case 1:
2012 ino = parent_ino(dentry);
2013 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2014 goto out;
2015 filp->f_pos++;
2016 default:
2017 files = get_files_struct(p);
2018 if (!files)
2019 goto out;
2020 rcu_read_lock();
28a6d671 2021 for (fd = filp->f_pos-2;
9b4f526c 2022 fd < files_fdtable(files)->max_fds;
28a6d671 2023 fd++, filp->f_pos++) {
27932742
MS		 2024 char name[PROC_NUMBUF];
2025 int len;
28a6d671
EB		 2026
2027 if (!fcheck_files(files, fd))
2028 continue;
2029 rcu_read_unlock();
2030
27932742
MS		 2031 len = snprintf(name, sizeof(name), "%d", fd);
2032 if (proc_fill_cache(filp, dirent, filldir,
2033 name, len, instantiate,
2034 p, &fd) < 0) {
28a6d671
EB		 2035 rcu_read_lock();
2036 break;
2037 }
2038 rcu_read_lock();
2039 }
2040 rcu_read_unlock();
2041 put_files_struct(files);
2042 }
2043out:
2044 put_task_struct(p);
2045out_no_task:
2046 return retval;
2047}
2048
27932742
MS		 2049static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
2050 struct nameidata *nd)
2051{
2052 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
2053}
2054
2055static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
2056{
2057 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
2058}
2059
2060static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
2061 size_t len, loff_t *ppos)
2062{
2063 char tmp[PROC_FDINFO_MAX];
3dcd25f3 2064 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
27932742
MS		 2065 if (!err)
2066 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
2067 return err;
2068}
2069
2070static const struct file_operations proc_fdinfo_file_operations = {
87df8424 2071 .open = nonseekable_open,
27932742 2072 .read = proc_fdinfo_read,
6038f373 2073 .llseek = no_llseek,
27932742
MS		 2074};
2075
00977a59 2076static const struct file_operations proc_fd_operations = {
28a6d671
EB		 2077 .read = generic_read_dir,
2078 .readdir = proc_readfd,
6038f373 2079 .llseek = default_llseek,
1da177e4
LT		 2080};
2081
8948e11f
AD		 2082/*
2083 * /proc/pid/fd needs a special permission handler so that a process can still
2084 * access /proc/self/fd after it has executed a setuid().
2085 */
e6305c43 2086static int proc_fd_permission(struct inode *inode, int mask)
8948e11f
AD		 2087{
2088 int rv;
2089
2090 rv = generic_permission(inode, mask, NULL);
2091 if (rv == 0)
2092 return 0;
2093 if (task_pid(current) == proc_pid(inode))
2094 rv = 0;
2095 return rv;
2096}
2097
1da177e4
LT		 2098/*
2099 * proc directories can do almost nothing..
2100 */
c5ef1c42 2101static const struct inode_operations proc_fd_inode_operations = {
1da177e4 2102 .lookup = proc_lookupfd,
8948e11f 2103 .permission = proc_fd_permission,
6d76fa58 2104 .setattr = proc_setattr,
1da177e4
LT		 2105};
2106
27932742
MS		 2107static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2108 struct dentry *dentry, struct task_struct *task, const void *ptr)
2109{
2110 unsigned fd = *(unsigned *)ptr;
2111 struct inode *inode;
2112 struct proc_inode *ei;
2113 struct dentry *error = ERR_PTR(-ENOENT);
2114
2115 inode = proc_pid_make_inode(dir->i_sb, task);
2116 if (!inode)
2117 goto out;
2118 ei = PROC_I(inode);
2119 ei->fd = fd;
2120 inode->i_mode = S_IFREG | S_IRUSR;
2121 inode->i_fop = &proc_fdinfo_file_operations;
2122 dentry->d_op = &tid_fd_dentry_operations;
2123 d_add(dentry, inode);
2124 /* Close the race of the process dying before we return the dentry */
2125 if (tid_fd_revalidate(dentry, NULL))
2126 error = NULL;
2127
2128 out:
2129 return error;
2130}
2131
2132static struct dentry *proc_lookupfdinfo(struct inode *dir,
2133 struct dentry *dentry,
2134 struct nameidata *nd)
2135{
2136 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2137}
2138
2139static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2140{
2141 return proc_readfd_common(filp, dirent, filldir,
2142 proc_fdinfo_instantiate);
2143}
2144
2145static const struct file_operations proc_fdinfo_operations = {
2146 .read = generic_read_dir,
2147 .readdir = proc_readfdinfo,
6038f373 2148 .llseek = default_llseek,
27932742
MS		 2149};
2150
2151/*
2152 * proc directories can do almost nothing..
2153 */
2154static const struct inode_operations proc_fdinfo_inode_operations = {
2155 .lookup = proc_lookupfdinfo,
2156 .setattr = proc_setattr,
2157};
2158
2159
444ceed8 2160static struct dentry *proc_pident_instantiate(struct inode *dir,
c5141e6d 2161 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8 2162{
c5141e6d 2163 const struct pid_entry *p = ptr;
444ceed8
EB		 2164 struct inode *inode;
2165 struct proc_inode *ei;
bd6daba9 2166 struct dentry *error = ERR_PTR(-ENOENT);
444ceed8 2167
61a28784 2168 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 2169 if (!inode)
2170 goto out;
2171
2172 ei = PROC_I(inode);
2173 inode->i_mode = p->mode;
2174 if (S_ISDIR(inode->i_mode))
2175 inode->i_nlink = 2; /* Use getattr to fix if necessary */
2176 if (p->iop)
2177 inode->i_op = p->iop;
2178 if (p->fop)
2179 inode->i_fop = p->fop;
2180 ei->op = p->op;
2181 dentry->d_op = &pid_dentry_operations;
2182 d_add(dentry, inode);
2183 /* Close the race of the process dying before we return the dentry */
2184 if (pid_revalidate(dentry, NULL))
2185 error = NULL;
2186out:
2187 return error;
2188}
2189
1da177e4
LT		 2190static struct dentry *proc_pident_lookup(struct inode *dir,
2191 struct dentry *dentry,
c5141e6d 2192 const struct pid_entry *ents,
7bcd6b0e 2193 unsigned int nents)
1da177e4 2194{
cd6a3ce9 2195 struct dentry *error;
99f89551 2196 struct task_struct *task = get_proc_task(dir);
c5141e6d 2197 const struct pid_entry *p, *last;
1da177e4 2198
cd6a3ce9 2199 error = ERR_PTR(-ENOENT);
1da177e4 2200
99f89551
EB		 2201 if (!task)
2202 goto out_no_task;
1da177e4 2203
20cdc894
EB		 2204 /*
2205 * Yes, it does not scale. And it should not. Don't add
2206 * new entries into /proc/<tgid>/ without very good reasons.
2207 */
7bcd6b0e
EB		 2208 last = &ents[nents - 1];
2209 for (p = ents; p <= last; p++) {
1da177e4
LT		 2210 if (p->len != dentry->d_name.len)
2211 continue;
2212 if (!memcmp(dentry->d_name.name, p->name, p->len))
2213 break;
2214 }
7bcd6b0e 2215 if (p > last)
1da177e4
LT		 2216 goto out;
2217
444ceed8 2218 error = proc_pident_instantiate(dir, dentry, task, p);
1da177e4 2219out:
99f89551
EB		 2220 put_task_struct(task);
2221out_no_task:
cd6a3ce9 2222 return error;
1da177e4
LT		 2223}
2224
c5141e6d
ED		 2225static int proc_pident_fill_cache(struct file *filp, void *dirent,
2226 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
61a28784
EB		 2227{
2228 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2229 proc_pident_instantiate, task, p);
2230}
2231
28a6d671
EB		 2232static int proc_pident_readdir(struct file *filp,
2233 void *dirent, filldir_t filldir,
c5141e6d 2234 const struct pid_entry *ents, unsigned int nents)
28a6d671
EB		 2235{
2236 int i;
2fddfeef 2237 struct dentry *dentry = filp->f_path.dentry;
28a6d671
EB		 2238 struct inode *inode = dentry->d_inode;
2239 struct task_struct *task = get_proc_task(inode);
c5141e6d 2240 const struct pid_entry *p, *last;
28a6d671
EB		 2241 ino_t ino;
2242 int ret;
2243
2244 ret = -ENOENT;
2245 if (!task)
61a28784 2246 goto out_no_task;
28a6d671
EB		 2247
2248 ret = 0;
28a6d671
EB		 2249 i = filp->f_pos;
2250 switch (i) {
2251 case 0:
2252 ino = inode->i_ino;
2253 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2254 goto out;
2255 i++;
2256 filp->f_pos++;
2257 /* fall through */
2258 case 1:
2259 ino = parent_ino(dentry);
2260 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2261 goto out;
2262 i++;
2263 filp->f_pos++;
2264 /* fall through */
2265 default:
2266 i -= 2;
2267 if (i >= nents) {
2268 ret = 1;
2269 goto out;
2270 }
2271 p = ents + i;
7bcd6b0e
EB		 2272 last = &ents[nents - 1];
2273 while (p <= last) {
61a28784 2274 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
28a6d671
EB		 2275 goto out;
2276 filp->f_pos++;
2277 p++;
2278 }
2279 }
2280
2281 ret = 1;
2282out:
61a28784
EB		 2283 put_task_struct(task);
2284out_no_task:
28a6d671 2285 return ret;
1da177e4
LT		 2286}
2287
28a6d671
EB		 2288#ifdef CONFIG_SECURITY
2289static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2290 size_t count, loff_t *ppos)
2291{
2fddfeef 2292 struct inode * inode = file->f_path.dentry->d_inode;
04ff9708 2293 char *p = NULL;
28a6d671
EB		 2294 ssize_t length;
2295 struct task_struct *task = get_proc_task(inode);
2296
28a6d671 2297 if (!task)
04ff9708 2298 return -ESRCH;
28a6d671
EB		 2299
2300 length = security_getprocattr(task,
2fddfeef 2301 (char*)file->f_path.dentry->d_name.name,
04ff9708 2302 &p);
28a6d671 2303 put_task_struct(task);
04ff9708
AV		 2304 if (length > 0)
2305 length = simple_read_from_buffer(buf, count, ppos, p, length);
2306 kfree(p);
28a6d671 2307 return length;
1da177e4
LT		 2308}
2309
28a6d671
EB		 2310static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2311 size_t count, loff_t *ppos)
2312{
2fddfeef 2313 struct inode * inode = file->f_path.dentry->d_inode;
28a6d671
EB		 2314 char *page;
2315 ssize_t length;
2316 struct task_struct *task = get_proc_task(inode);
2317
2318 length = -ESRCH;
2319 if (!task)
2320 goto out_no_task;
2321 if (count > PAGE_SIZE)
2322 count = PAGE_SIZE;
2323
2324 /* No partial writes. */
2325 length = -EINVAL;
2326 if (*ppos != 0)
2327 goto out;
2328
2329 length = -ENOMEM;
e12ba74d 2330 page = (char*)__get_free_page(GFP_TEMPORARY);
28a6d671
EB		 2331 if (!page)
2332 goto out;
2333
2334 length = -EFAULT;
2335 if (copy_from_user(page, buf, count))
2336 goto out_free;
2337
107db7c7
DH		 2338 /* Guard against adverse ptrace interaction */
2339 length = mutex_lock_interruptible(&task->cred_guard_mutex);
2340 if (length < 0)
2341 goto out_free;
2342
28a6d671 2343 length = security_setprocattr(task,
2fddfeef 2344 (char*)file->f_path.dentry->d_name.name,
28a6d671 2345 (void*)page, count);
107db7c7 2346 mutex_unlock(&task->cred_guard_mutex);
28a6d671
EB		 2347out_free:
2348 free_page((unsigned long) page);
2349out:
2350 put_task_struct(task);
2351out_no_task:
2352 return length;
2353}
2354
00977a59 2355static const struct file_operations proc_pid_attr_operations = {
28a6d671
EB		 2356 .read = proc_pid_attr_read,
2357 .write = proc_pid_attr_write,
87df8424 2358 .llseek = generic_file_llseek,
28a6d671
EB		 2359};
2360
c5141e6d 2361static const struct pid_entry attr_dir_stuff[] = {
631f9c18
AD		 2362 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2363 REG("prev", S_IRUGO, proc_pid_attr_operations),
2364 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2365 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2366 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2367 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
28a6d671
EB		 2368};
2369
72d9dcfc 2370static int proc_attr_dir_readdir(struct file * filp,
28a6d671
EB		 2371 void * dirent, filldir_t filldir)
2372{
2373 return proc_pident_readdir(filp,dirent,filldir,
72d9dcfc 2374 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2375}
2376
00977a59 2377static const struct file_operations proc_attr_dir_operations = {
1da177e4 2378 .read = generic_read_dir,
72d9dcfc 2379 .readdir = proc_attr_dir_readdir,
6038f373 2380 .llseek = default_llseek,
1da177e4
LT		 2381};
2382
72d9dcfc 2383static struct dentry *proc_attr_dir_lookup(struct inode *dir,
28a6d671
EB		 2384 struct dentry *dentry, struct nameidata *nd)
2385{
7bcd6b0e
EB		 2386 return proc_pident_lookup(dir, dentry,
2387 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2388}
2389
c5ef1c42 2390static const struct inode_operations proc_attr_dir_inode_operations = {
72d9dcfc 2391 .lookup = proc_attr_dir_lookup,
99f89551 2392 .getattr = pid_getattr,
6d76fa58 2393 .setattr = proc_setattr,
1da177e4
LT		 2394};
2395
28a6d671
EB		 2396#endif
2397
698ba7b5 2398#ifdef CONFIG_ELF_CORE
3cb4a0bb
KH		 2399static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2400 size_t count, loff_t *ppos)
2401{
2402 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2403 struct mm_struct *mm;
2404 char buffer[PROC_NUMBUF];
2405 size_t len;
2406 int ret;
2407
2408 if (!task)
2409 return -ESRCH;
2410
2411 ret = 0;
2412 mm = get_task_mm(task);
2413 if (mm) {
2414 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2415 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2416 MMF_DUMP_FILTER_SHIFT));
2417 mmput(mm);
2418 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2419 }
2420
2421 put_task_struct(task);
2422
2423 return ret;
2424}
2425
2426static ssize_t proc_coredump_filter_write(struct file *file,
2427 const char __user *buf,
2428 size_t count,
2429 loff_t *ppos)
2430{
2431 struct task_struct *task;
2432 struct mm_struct *mm;
2433 char buffer[PROC_NUMBUF], *end;
2434 unsigned int val;
2435 int ret;
2436 int i;
2437 unsigned long mask;
2438
2439 ret = -EFAULT;
2440 memset(buffer, 0, sizeof(buffer));
2441 if (count > sizeof(buffer) - 1)
2442 count = sizeof(buffer) - 1;
2443 if (copy_from_user(buffer, buf, count))
2444 goto out_no_task;
2445
2446 ret = -EINVAL;
2447 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2448 if (*end == '\n')
2449 end++;
2450 if (end - buffer == 0)
2451 goto out_no_task;
2452
2453 ret = -ESRCH;
2454 task = get_proc_task(file->f_dentry->d_inode);
2455 if (!task)
2456 goto out_no_task;
2457
2458 ret = end - buffer;
2459 mm = get_task_mm(task);
2460 if (!mm)
2461 goto out_no_mm;
2462
2463 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2464 if (val & mask)
2465 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2466 else
2467 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2468 }
2469
2470 mmput(mm);
2471 out_no_mm:
2472 put_task_struct(task);
2473 out_no_task:
2474 return ret;
2475}
2476
2477static const struct file_operations proc_coredump_filter_operations = {
2478 .read = proc_coredump_filter_read,
2479 .write = proc_coredump_filter_write,
87df8424 2480 .llseek = generic_file_llseek,
3cb4a0bb
KH		 2481};
2482#endif
2483
28a6d671
EB		 2484/*
2485 * /proc/self:
2486 */
2487static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2488 int buflen)
2489{
488e5bc4 2490 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
b55fcb22 2491 pid_t tgid = task_tgid_nr_ns(current, ns);
28a6d671 2492 char tmp[PROC_NUMBUF];
b55fcb22 2493 if (!tgid)
488e5bc4 2494 return -ENOENT;
b55fcb22 2495 sprintf(tmp, "%d", tgid);
28a6d671
EB		 2496 return vfs_readlink(dentry,buffer,buflen,tmp);
2497}
2498
2499static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2500{
488e5bc4 2501 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
b55fcb22 2502 pid_t tgid = task_tgid_nr_ns(current, ns);
7fee4868
AV		 2503 char *name = ERR_PTR(-ENOENT);
2504 if (tgid) {
2505 name = __getname();
2506 if (!name)
2507 name = ERR_PTR(-ENOMEM);
2508 else
2509 sprintf(name, "%d", tgid);
2510 }
2511 nd_set_link(nd, name);
2512 return NULL;
2513}
2514
2515static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2516 void *cookie)
2517{
2518 char *s = nd_get_link(nd);
2519 if (!IS_ERR(s))
2520 __putname(s);
28a6d671
EB		 2521}
2522
c5ef1c42 2523static const struct inode_operations proc_self_inode_operations = {
28a6d671
EB		 2524 .readlink = proc_self_readlink,
2525 .follow_link = proc_self_follow_link,
7fee4868 2526 .put_link = proc_self_put_link,
28a6d671
EB		 2527};
2528
801199ce
EB		 2529/*
2530 * proc base
2531 *
2532 * These are the directory entries in the root directory of /proc
2533 * that properly belong to the /proc filesystem, as they describe
2534 * describe something that is process related.
2535 */
c5141e6d 2536static const struct pid_entry proc_base_stuff[] = {
61a28784 2537 NOD("self", S_IFLNK|S_IRWXUGO,
801199ce 2538 &proc_self_inode_operations, NULL, {}),
801199ce
EB		 2539};
2540
2541/*
2542 * Exceptional case: normally we are not allowed to unhash a busy
2543 * directory. In this case, however, we can do it - no aliasing problems
2544 * due to the way we treat inodes.
2545 */
2546static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
2547{
2548 struct inode *inode = dentry->d_inode;
2549 struct task_struct *task = get_proc_task(inode);
2550 if (task) {
2551 put_task_struct(task);
2552 return 1;
2553 }
2554 d_drop(dentry);
2555 return 0;
2556}
2557
d72f71eb 2558static const struct dentry_operations proc_base_dentry_operations =
801199ce
EB		 2559{
2560 .d_revalidate = proc_base_revalidate,
2561 .d_delete = pid_delete_dentry,
2562};
2563
444ceed8 2564static struct dentry *proc_base_instantiate(struct inode *dir,
c5141e6d 2565 struct dentry *dentry, struct task_struct *task, const void *ptr)
801199ce 2566{
c5141e6d 2567 const struct pid_entry *p = ptr;
801199ce 2568 struct inode *inode;
801199ce 2569 struct proc_inode *ei;
73d36460 2570 struct dentry *error;
801199ce
EB		 2571
2572 /* Allocate the inode */
2573 error = ERR_PTR(-ENOMEM);
2574 inode = new_inode(dir->i_sb);
2575 if (!inode)
2576 goto out;
2577
2578 /* Initialize the inode */
2579 ei = PROC_I(inode);
2580 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
801199ce
EB		 2581
2582 /*
2583 * grab the reference to the task.
2584 */
1a657f78 2585 ei->pid = get_task_pid(task, PIDTYPE_PID);
801199ce
EB		 2586 if (!ei->pid)
2587 goto out_iput;
2588
801199ce
EB		 2589 inode->i_mode = p->mode;
2590 if (S_ISDIR(inode->i_mode))
2591 inode->i_nlink = 2;
2592 if (S_ISLNK(inode->i_mode))
2593 inode->i_size = 64;
2594 if (p->iop)
2595 inode->i_op = p->iop;
2596 if (p->fop)
2597 inode->i_fop = p->fop;
2598 ei->op = p->op;
2599 dentry->d_op = &proc_base_dentry_operations;
2600 d_add(dentry, inode);
2601 error = NULL;
2602out:
801199ce
EB		 2603 return error;
2604out_iput:
2605 iput(inode);
2606 goto out;
2607}
2608
444ceed8
EB		 2609static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2610{
2611 struct dentry *error;
2612 struct task_struct *task = get_proc_task(dir);
c5141e6d 2613 const struct pid_entry *p, *last;
444ceed8
EB		 2614
2615 error = ERR_PTR(-ENOENT);
2616
2617 if (!task)
2618 goto out_no_task;
2619
2620 /* Lookup the directory entry */
7bcd6b0e
EB		 2621 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2622 for (p = proc_base_stuff; p <= last; p++) {
444ceed8
EB		 2623 if (p->len != dentry->d_name.len)
2624 continue;
2625 if (!memcmp(dentry->d_name.name, p->name, p->len))
2626 break;
2627 }
7bcd6b0e 2628 if (p > last)
444ceed8
EB		 2629 goto out;
2630
2631 error = proc_base_instantiate(dir, dentry, task, p);
2632
2633out:
2634 put_task_struct(task);
2635out_no_task:
2636 return error;
2637}
2638
c5141e6d
ED		 2639static int proc_base_fill_cache(struct file *filp, void *dirent,
2640 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
61a28784
EB		 2641{
2642 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2643 proc_base_instantiate, task, p);
2644}
2645
aba76fdb 2646#ifdef CONFIG_TASK_IO_ACCOUNTING
297c5d92
AR		 2647static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2648{
940389b8 2649 struct task_io_accounting acct = task->ioac;
5995477a
AR		 2650 unsigned long flags;
2651
2652 if (whole && lock_task_sighand(task, &flags)) {
2653 struct task_struct *t = task;
2654
2655 task_io_accounting_add(&acct, &task->signal->ioac);
2656 while_each_thread(task, t)
2657 task_io_accounting_add(&acct, &t->ioac);
2658
2659 unlock_task_sighand(task, &flags);
297c5d92 2660 }
aba76fdb
AM		 2661 return sprintf(buffer,
2662 "rchar: %llu\n"
2663 "wchar: %llu\n"
2664 "syscr: %llu\n"
2665 "syscw: %llu\n"
2666 "read_bytes: %llu\n"
2667 "write_bytes: %llu\n"
2668 "cancelled_write_bytes: %llu\n",
7c44319d
AB		 2669 (unsigned long long)acct.rchar,
2670 (unsigned long long)acct.wchar,
2671 (unsigned long long)acct.syscr,
2672 (unsigned long long)acct.syscw,
2673 (unsigned long long)acct.read_bytes,
2674 (unsigned long long)acct.write_bytes,
2675 (unsigned long long)acct.cancelled_write_bytes);
297c5d92
AR		 2676}
2677
2678static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2679{
2680 return do_io_accounting(task, buffer, 0);
aba76fdb 2681}
297c5d92
AR		 2682
2683static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2684{
2685 return do_io_accounting(task, buffer, 1);
2686}
2687#endif /* CONFIG_TASK_IO_ACCOUNTING */
aba76fdb 2688
47830723
KC		 2689static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2690 struct pid *pid, struct task_struct *task)
2691{
2692 seq_printf(m, "%08x\n", task->personality);
2693 return 0;
2694}
2695
28a6d671
EB		 2696/*
2697 * Thread groups
2698 */
00977a59 2699static const struct file_operations proc_task_operations;
c5ef1c42 2700static const struct inode_operations proc_task_inode_operations;
20cdc894 2701
c5141e6d 2702static const struct pid_entry tgid_base_stuff[] = {
631f9c18
AD		 2703 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2704 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2705 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
b2211a36 2706#ifdef CONFIG_NET
631f9c18 2707 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
b2211a36 2708#endif
631f9c18
AD		 2709 REG("environ", S_IRUSR, proc_environ_operations),
2710 INF("auxv", S_IRUSR, proc_pid_auxv),
2711 ONE("status", S_IRUGO, proc_pid_status),
2712 ONE("personality", S_IRUSR, proc_pid_personality),
3036e7b4 2713 INF("limits", S_IRUGO, proc_pid_limits),
43ae34cb 2714#ifdef CONFIG_SCHED_DEBUG
631f9c18 2715 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
ebcb6734 2716#endif
4614a696 2717 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 2718#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
631f9c18 2719 INF("syscall", S_IRUSR, proc_pid_syscall),
43ae34cb 2720#endif
631f9c18
AD		 2721 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2722 ONE("stat", S_IRUGO, proc_tgid_stat),
2723 ONE("statm", S_IRUGO, proc_pid_statm),
2724 REG("maps", S_IRUGO, proc_maps_operations),
28a6d671 2725#ifdef CONFIG_NUMA
631f9c18 2726 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
28a6d671 2727#endif
631f9c18
AD		 2728 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2729 LNK("cwd", proc_cwd_link),
2730 LNK("root", proc_root_link),
2731 LNK("exe", proc_exe_link),
2732 REG("mounts", S_IRUGO, proc_mounts_operations),
2733 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2734 REG("mountstats", S_IRUSR, proc_mountstats_operations),
1e883281 2735#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18
AD		 2736 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2737 REG("smaps", S_IRUGO, proc_smaps_operations),
2738 REG("pagemap", S_IRUSR, proc_pagemap_operations),
28a6d671
EB		 2739#endif
2740#ifdef CONFIG_SECURITY
631f9c18 2741 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 2742#endif
2743#ifdef CONFIG_KALLSYMS
631f9c18 2744 INF("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 2745#endif
2ec220e2
KC		 2746#ifdef CONFIG_STACKTRACE
2747 ONE("stack", S_IRUSR, proc_pid_stack),
28a6d671
EB		 2748#endif
2749#ifdef CONFIG_SCHEDSTATS
631f9c18 2750 INF("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 2751#endif
9745512c 2752#ifdef CONFIG_LATENCYTOP
631f9c18 2753 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 2754#endif
8793d854 2755#ifdef CONFIG_PROC_PID_CPUSET
631f9c18 2756 REG("cpuset", S_IRUGO, proc_cpuset_operations),
a424316c
PM		 2757#endif
2758#ifdef CONFIG_CGROUPS
631f9c18 2759 REG("cgroup", S_IRUGO, proc_cgroup_operations),
28a6d671 2760#endif
631f9c18
AD		 2761 INF("oom_score", S_IRUGO, proc_oom_score),
2762 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
a63d83f4 2763 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 2764#ifdef CONFIG_AUDITSYSCALL
631f9c18
AD		 2765 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2766 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 2767#endif
f4f154fd 2768#ifdef CONFIG_FAULT_INJECTION
631f9c18 2769 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 2770#endif
698ba7b5 2771#ifdef CONFIG_ELF_CORE
631f9c18 2772 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3cb4a0bb 2773#endif
aba76fdb 2774#ifdef CONFIG_TASK_IO_ACCOUNTING
631f9c18 2775 INF("io", S_IRUGO, proc_tgid_io_accounting),
aba76fdb 2776#endif
28a6d671 2777};
1da177e4 2778
28a6d671 2779static int proc_tgid_base_readdir(struct file * filp,
1da177e4
LT		 2780 void * dirent, filldir_t filldir)
2781{
2782 return proc_pident_readdir(filp,dirent,filldir,
28a6d671 2783 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 2784}
2785
00977a59 2786static const struct file_operations proc_tgid_base_operations = {
1da177e4 2787 .read = generic_read_dir,
28a6d671 2788 .readdir = proc_tgid_base_readdir,
6038f373 2789 .llseek = default_llseek,
1da177e4
LT		 2790};
2791
28a6d671 2792static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
7bcd6b0e
EB		 2793 return proc_pident_lookup(dir, dentry,
2794 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 2795}
2796
c5ef1c42 2797static const struct inode_operations proc_tgid_base_inode_operations = {
28a6d671 2798 .lookup = proc_tgid_base_lookup,
99f89551 2799 .getattr = pid_getattr,
6d76fa58 2800 .setattr = proc_setattr,
1da177e4 2801};
1da177e4 2802
60347f67 2803static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
1da177e4 2804{
48e6484d 2805 struct dentry *dentry, *leader, *dir;
8578cea7 2806 char buf[PROC_NUMBUF];
48e6484d
EB		 2807 struct qstr name;
2808
2809 name.name = buf;
60347f67
PE		 2810 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2811 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d 2812 if (dentry) {
29f12ca3 2813 shrink_dcache_parent(dentry);
48e6484d
EB		 2814 d_drop(dentry);
2815 dput(dentry);
2816 }
1da177e4 2817
48e6484d 2818 name.name = buf;
60347f67
PE		 2819 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2820 leader = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d
EB		 2821 if (!leader)
2822 goto out;
1da177e4 2823
48e6484d
EB		 2824 name.name = "task";
2825 name.len = strlen(name.name);
2826 dir = d_hash_and_lookup(leader, &name);
2827 if (!dir)
2828 goto out_put_leader;
2829
2830 name.name = buf;
60347f67 2831 name.len = snprintf(buf, sizeof(buf), "%d", pid);
48e6484d
EB		 2832 dentry = d_hash_and_lookup(dir, &name);
2833 if (dentry) {
2834 shrink_dcache_parent(dentry);
2835 d_drop(dentry);
2836 dput(dentry);
1da177e4 2837 }
48e6484d
EB		 2838
2839 dput(dir);
2840out_put_leader:
2841 dput(leader);
2842out:
2843 return;
1da177e4
LT		 2844}
2845
0895e91d
RD		 2846/**
2847 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2848 * @task: task that should be flushed.
2849 *
2850 * When flushing dentries from proc, one needs to flush them from global
60347f67 2851 * proc (proc_mnt) and from all the namespaces' procs this task was seen
0895e91d
RD		 2852 * in. This call is supposed to do all of this job.
2853 *
2854 * Looks in the dcache for
2855 * /proc/@pid
2856 * /proc/@tgid/task/@pid
2857 * if either directory is present flushes it and all of it'ts children
2858 * from the dcache.
2859 *
2860 * It is safe and reasonable to cache /proc entries for a task until
2861 * that task exits. After that they just clog up the dcache with
2862 * useless entries, possibly causing useful dcache entries to be
2863 * flushed instead. This routine is proved to flush those useless
2864 * dcache entries at process exit time.
2865 *
2866 * NOTE: This routine is just an optimization so it does not guarantee
2867 * that no dcache entries will exist at process exit time it
2868 * just makes it very unlikely that any will persist.
60347f67
PE		 2869 */
2870
2871void proc_flush_task(struct task_struct *task)
2872{
9fcc2d15 2873 int i;
9b4d1cbe 2874 struct pid *pid, *tgid;
130f77ec
PE		 2875 struct upid *upid;
2876
130f77ec 2877 pid = task_pid(task);
9b4d1cbe 2878 tgid = task_tgid(task);
130f77ec 2879
9fcc2d15 2880 for (i = 0; i <= pid->level; i++) {
130f77ec
PE		 2881 upid = &pid->numbers[i];
2882 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
9b4d1cbe 2883 tgid->numbers[i].nr);
130f77ec 2884 }
6f4e6433
PE		 2885
2886 upid = &pid->numbers[pid->level];
2887 if (upid->nr == 1)
2888 pid_ns_release_proc(upid->ns);
60347f67
PE		 2889}
2890
9711ef99
AB		 2891static struct dentry *proc_pid_instantiate(struct inode *dir,
2892 struct dentry * dentry,
c5141e6d 2893 struct task_struct *task, const void *ptr)
444ceed8
EB		 2894{
2895 struct dentry *error = ERR_PTR(-ENOENT);
2896 struct inode *inode;
2897
61a28784 2898 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 2899 if (!inode)
2900 goto out;
2901
2902 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2903 inode->i_op = &proc_tgid_base_inode_operations;
2904 inode->i_fop = &proc_tgid_base_operations;
2905 inode->i_flags|=S_IMMUTABLE;
aed54175
VN		 2906
2907 inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
2908 ARRAY_SIZE(tgid_base_stuff));
444ceed8
EB		 2909
2910 dentry->d_op = &pid_dentry_operations;
2911
2912 d_add(dentry, inode);
2913 /* Close the race of the process dying before we return the dentry */
2914 if (pid_revalidate(dentry, NULL))
2915 error = NULL;
2916out:
2917 return error;
2918}
2919
1da177e4
LT		 2920struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2921{
73d36460 2922 struct dentry *result;
1da177e4 2923 struct task_struct *task;
1da177e4 2924 unsigned tgid;
b488893a 2925 struct pid_namespace *ns;
1da177e4 2926
801199ce
EB		 2927 result = proc_base_lookup(dir, dentry);
2928 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2929 goto out;
2930
1da177e4
LT		 2931 tgid = name_to_int(dentry);
2932 if (tgid == ~0U)
2933 goto out;
2934
b488893a 2935 ns = dentry->d_sb->s_fs_info;
de758734 2936 rcu_read_lock();
b488893a 2937 task = find_task_by_pid_ns(tgid, ns);
1da177e4
LT		 2938 if (task)
2939 get_task_struct(task);
de758734 2940 rcu_read_unlock();
1da177e4
LT		 2941 if (!task)
2942 goto out;
2943
444ceed8 2944 result = proc_pid_instantiate(dir, dentry, task, NULL);
1da177e4 2945 put_task_struct(task);
1da177e4 2946out:
cd6a3ce9 2947 return result;
1da177e4
LT		 2948}
2949
1da177e4 2950/*
0804ef4b 2951 * Find the first task with tgid >= tgid
0bc58a91 2952 *
1da177e4 2953 */
19fd4bb2
EB		 2954struct tgid_iter {
2955 unsigned int tgid;
0804ef4b 2956 struct task_struct *task;
19fd4bb2
EB		 2957};
2958static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2959{
0804ef4b 2960 struct pid *pid;
1da177e4 2961
19fd4bb2
EB		 2962 if (iter.task)
2963 put_task_struct(iter.task);
454cc105 2964 rcu_read_lock();
0804ef4b 2965retry:
19fd4bb2
EB		 2966 iter.task = NULL;
2967 pid = find_ge_pid(iter.tgid, ns);
0804ef4b 2968 if (pid) {
19fd4bb2
EB		 2969 iter.tgid = pid_nr_ns(pid, ns);
2970 iter.task = pid_task(pid, PIDTYPE_PID);
0804ef4b
EB		 2971 /* What we to know is if the pid we have find is the
2972 * pid of a thread_group_leader. Testing for task
2973 * being a thread_group_leader is the obvious thing
2974 * todo but there is a window when it fails, due to
2975 * the pid transfer logic in de_thread.
2976 *
2977 * So we perform the straight forward test of seeing
2978 * if the pid we have found is the pid of a thread
2979 * group leader, and don't worry if the task we have
2980 * found doesn't happen to be a thread group leader.
2981 * As we don't care in the case of readdir.
2982 */
19fd4bb2
EB		 2983 if (!iter.task || !has_group_leader_pid(iter.task)) {
2984 iter.tgid += 1;
0804ef4b 2985 goto retry;
19fd4bb2
EB		 2986 }
2987 get_task_struct(iter.task);
0bc58a91 2988 }
454cc105 2989 rcu_read_unlock();
19fd4bb2 2990 return iter;
1da177e4
LT		 2991}
2992
7bcd6b0e 2993#define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
0804ef4b 2994
61a28784 2995static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
19fd4bb2 2996 struct tgid_iter iter)
61a28784
EB		 2997{
2998 char name[PROC_NUMBUF];
19fd4bb2 2999 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
61a28784 3000 return proc_fill_cache(filp, dirent, filldir, name, len,
19fd4bb2 3001 proc_pid_instantiate, iter.task, NULL);
61a28784
EB		 3002}
3003
1da177e4
LT		 3004/* for the /proc/ directory itself, after non-process stuff has been done */
3005int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3006{
1da177e4 3007 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2fddfeef 3008 struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
19fd4bb2 3009 struct tgid_iter iter;
b488893a 3010 struct pid_namespace *ns;
1da177e4 3011
61a28784
EB		 3012 if (!reaper)
3013 goto out_no_task;
3014
7bcd6b0e 3015 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
c5141e6d 3016 const struct pid_entry *p = &proc_base_stuff[nr];
61a28784 3017 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
801199ce 3018 goto out;
1da177e4
LT		 3019 }
3020
b488893a 3021 ns = filp->f_dentry->d_sb->s_fs_info;
19fd4bb2
EB		 3022 iter.task = NULL;
3023 iter.tgid = filp->f_pos - TGID_OFFSET;
3024 for (iter = next_tgid(ns, iter);
3025 iter.task;
3026 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3027 filp->f_pos = iter.tgid + TGID_OFFSET;
3028 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
3029 put_task_struct(iter.task);
0804ef4b 3030 goto out;
1da177e4 3031 }
0bc58a91 3032 }
0804ef4b
EB		 3033 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3034out:
61a28784
EB		 3035 put_task_struct(reaper);
3036out_no_task:
0bc58a91
EB		 3037 return 0;
3038}
1da177e4 3039
28a6d671
EB		 3040/*
3041 * Tasks
3042 */
c5141e6d 3043static const struct pid_entry tid_base_stuff[] = {
631f9c18 3044 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3835541d 3045 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
631f9c18
AD		 3046 REG("environ", S_IRUSR, proc_environ_operations),
3047 INF("auxv", S_IRUSR, proc_pid_auxv),
3048 ONE("status", S_IRUGO, proc_pid_status),
3049 ONE("personality", S_IRUSR, proc_pid_personality),
3036e7b4 3050 INF("limits", S_IRUGO, proc_pid_limits),
43ae34cb 3051#ifdef CONFIG_SCHED_DEBUG
631f9c18 3052 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
ebcb6734 3053#endif
4614a696 3054 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 3055#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
631f9c18 3056 INF("syscall", S_IRUSR, proc_pid_syscall),
43ae34cb 3057#endif
631f9c18
AD		 3058 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3059 ONE("stat", S_IRUGO, proc_tid_stat),
3060 ONE("statm", S_IRUGO, proc_pid_statm),
3061 REG("maps", S_IRUGO, proc_maps_operations),
28a6d671 3062#ifdef CONFIG_NUMA
631f9c18 3063 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
28a6d671 3064#endif
631f9c18
AD		 3065 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3066 LNK("cwd", proc_cwd_link),
3067 LNK("root", proc_root_link),
3068 LNK("exe", proc_exe_link),
3069 REG("mounts", S_IRUGO, proc_mounts_operations),
3070 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
1e883281 3071#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18
AD		 3072 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3073 REG("smaps", S_IRUGO, proc_smaps_operations),
3074 REG("pagemap", S_IRUSR, proc_pagemap_operations),
28a6d671
EB		 3075#endif
3076#ifdef CONFIG_SECURITY
631f9c18 3077 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 3078#endif
3079#ifdef CONFIG_KALLSYMS
631f9c18 3080 INF("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 3081#endif
2ec220e2
KC		 3082#ifdef CONFIG_STACKTRACE
3083 ONE("stack", S_IRUSR, proc_pid_stack),
28a6d671
EB		 3084#endif
3085#ifdef CONFIG_SCHEDSTATS
631f9c18 3086 INF("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 3087#endif
9745512c 3088#ifdef CONFIG_LATENCYTOP
631f9c18 3089 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 3090#endif
8793d854 3091#ifdef CONFIG_PROC_PID_CPUSET
631f9c18 3092 REG("cpuset", S_IRUGO, proc_cpuset_operations),
a424316c
PM		 3093#endif
3094#ifdef CONFIG_CGROUPS
631f9c18 3095 REG("cgroup", S_IRUGO, proc_cgroup_operations),
28a6d671 3096#endif
631f9c18
AD		 3097 INF("oom_score", S_IRUGO, proc_oom_score),
3098 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
a63d83f4 3099 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 3100#ifdef CONFIG_AUDITSYSCALL
631f9c18
AD		 3101 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3102 REG("sessionid", S_IRUSR, proc_sessionid_operations),
28a6d671 3103#endif
f4f154fd 3104#ifdef CONFIG_FAULT_INJECTION
631f9c18 3105 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 3106#endif
297c5d92 3107#ifdef CONFIG_TASK_IO_ACCOUNTING
631f9c18 3108 INF("io", S_IRUGO, proc_tid_io_accounting),
297c5d92 3109#endif
28a6d671
EB		 3110};
3111
3112static int proc_tid_base_readdir(struct file * filp,
3113 void * dirent, filldir_t filldir)
3114{
3115 return proc_pident_readdir(filp,dirent,filldir,
3116 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3117}
3118
3119static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
7bcd6b0e
EB		 3120 return proc_pident_lookup(dir, dentry,
3121 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
28a6d671
EB		 3122}
3123
00977a59 3124static const struct file_operations proc_tid_base_operations = {
28a6d671
EB		 3125 .read = generic_read_dir,
3126 .readdir = proc_tid_base_readdir,
6038f373 3127 .llseek = default_llseek,
28a6d671
EB		 3128};
3129
c5ef1c42 3130static const struct inode_operations proc_tid_base_inode_operations = {
28a6d671
EB		 3131 .lookup = proc_tid_base_lookup,
3132 .getattr = pid_getattr,
3133 .setattr = proc_setattr,
3134};
3135
444ceed8 3136static struct dentry *proc_task_instantiate(struct inode *dir,
c5141e6d 3137 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8
EB		 3138{
3139 struct dentry *error = ERR_PTR(-ENOENT);
3140 struct inode *inode;
61a28784 3141 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 3142
3143 if (!inode)
3144 goto out;
3145 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3146 inode->i_op = &proc_tid_base_inode_operations;
3147 inode->i_fop = &proc_tid_base_operations;
3148 inode->i_flags|=S_IMMUTABLE;
aed54175
VN		 3149
3150 inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
3151 ARRAY_SIZE(tid_base_stuff));
444ceed8
EB		 3152
3153 dentry->d_op = &pid_dentry_operations;
3154
3155 d_add(dentry, inode);
3156 /* Close the race of the process dying before we return the dentry */
3157 if (pid_revalidate(dentry, NULL))
3158 error = NULL;
3159out:
3160 return error;
3161}
3162
28a6d671
EB		 3163static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3164{
3165 struct dentry *result = ERR_PTR(-ENOENT);
3166 struct task_struct *task;
3167 struct task_struct *leader = get_proc_task(dir);
28a6d671 3168 unsigned tid;
b488893a 3169 struct pid_namespace *ns;
28a6d671
EB		 3170
3171 if (!leader)
3172 goto out_no_task;
3173
3174 tid = name_to_int(dentry);
3175 if (tid == ~0U)
3176 goto out;
3177
b488893a 3178 ns = dentry->d_sb->s_fs_info;
28a6d671 3179 rcu_read_lock();
b488893a 3180 task = find_task_by_pid_ns(tid, ns);
28a6d671
EB		 3181 if (task)
3182 get_task_struct(task);
3183 rcu_read_unlock();
3184 if (!task)
3185 goto out;
bac0abd6 3186 if (!same_thread_group(leader, task))
28a6d671
EB		 3187 goto out_drop_task;
3188
444ceed8 3189 result = proc_task_instantiate(dir, dentry, task, NULL);
28a6d671
EB		 3190out_drop_task:
3191 put_task_struct(task);
3192out:
3193 put_task_struct(leader);
3194out_no_task:
3195 return result;
3196}
3197
0bc58a91
EB		 3198/*
3199 * Find the first tid of a thread group to return to user space.
3200 *
3201 * Usually this is just the thread group leader, but if the users
3202 * buffer was too small or there was a seek into the middle of the
3203 * directory we have more work todo.
3204 *
3205 * In the case of a short read we start with find_task_by_pid.
3206 *
3207 * In the case of a seek we start with the leader and walk nr
3208 * threads past it.
3209 */
cc288738 3210static struct task_struct *first_tid(struct task_struct *leader,
b488893a 3211 int tid, int nr, struct pid_namespace *ns)
0bc58a91 3212{
a872ff0c 3213 struct task_struct *pos;
1da177e4 3214
cc288738 3215 rcu_read_lock();
0bc58a91
EB		 3216 /* Attempt to start with the pid of a thread */
3217 if (tid && (nr > 0)) {
b488893a 3218 pos = find_task_by_pid_ns(tid, ns);
a872ff0c
ON		 3219 if (pos && (pos->group_leader == leader))
3220 goto found;
0bc58a91 3221 }
1da177e4 3222
0bc58a91 3223 /* If nr exceeds the number of threads there is nothing todo */
a872ff0c
ON		 3224 pos = NULL;
3225 if (nr && nr >= get_nr_threads(leader))
3226 goto out;
1da177e4 3227
a872ff0c
ON		 3228 /* If we haven't found our starting place yet start
3229 * with the leader and walk nr threads forward.
0bc58a91 3230 */
a872ff0c
ON		 3231 for (pos = leader; nr > 0; --nr) {
3232 pos = next_thread(pos);
3233 if (pos == leader) {
3234 pos = NULL;
3235 goto out;
3236 }
1da177e4 3237 }
a872ff0c
ON		 3238found:
3239 get_task_struct(pos);
3240out:
cc288738 3241 rcu_read_unlock();
0bc58a91
EB		 3242 return pos;
3243}
3244
3245/*
3246 * Find the next thread in the thread list.
3247 * Return NULL if there is an error or no next thread.
3248 *
3249 * The reference to the input task_struct is released.
3250 */
3251static struct task_struct *next_tid(struct task_struct *start)
3252{
c1df7fb8 3253 struct task_struct *pos = NULL;
cc288738 3254 rcu_read_lock();
c1df7fb8 3255 if (pid_alive(start)) {
0bc58a91 3256 pos = next_thread(start);
c1df7fb8
ON		 3257 if (thread_group_leader(pos))
3258 pos = NULL;
3259 else
3260 get_task_struct(pos);
3261 }
cc288738 3262 rcu_read_unlock();
0bc58a91
EB		 3263 put_task_struct(start);
3264 return pos;
1da177e4
LT		 3265}
3266
61a28784
EB		 3267static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3268 struct task_struct *task, int tid)
3269{
3270 char name[PROC_NUMBUF];
3271 int len = snprintf(name, sizeof(name), "%d", tid);
3272 return proc_fill_cache(filp, dirent, filldir, name, len,
3273 proc_task_instantiate, task, NULL);
3274}
3275
1da177e4
LT		 3276/* for the /proc/TGID/task/ directories */
3277static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3278{
2fddfeef 3279 struct dentry *dentry = filp->f_path.dentry;
1da177e4 3280 struct inode *inode = dentry->d_inode;
7d895244 3281 struct task_struct *leader = NULL;
0bc58a91 3282 struct task_struct *task;
1da177e4
LT		 3283 int retval = -ENOENT;
3284 ino_t ino;
0bc58a91 3285 int tid;
b488893a 3286 struct pid_namespace *ns;
1da177e4 3287
7d895244
GC		 3288 task = get_proc_task(inode);
3289 if (!task)
3290 goto out_no_task;
3291 rcu_read_lock();
3292 if (pid_alive(task)) {
3293 leader = task->group_leader;
3294 get_task_struct(leader);
3295 }
3296 rcu_read_unlock();
3297 put_task_struct(task);
99f89551
EB		 3298 if (!leader)
3299 goto out_no_task;
1da177e4
LT		 3300 retval = 0;
3301
ee568b25 3302 switch ((unsigned long)filp->f_pos) {
1da177e4
LT		 3303 case 0:
3304 ino = inode->i_ino;
ee6f779b 3305 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
1da177e4 3306 goto out;
ee6f779b 3307 filp->f_pos++;
1da177e4
LT		 3308 /* fall through */
3309 case 1:
3310 ino = parent_ino(dentry);
ee6f779b 3311 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
1da177e4 3312 goto out;
ee6f779b 3313 filp->f_pos++;
1da177e4
LT		 3314 /* fall through */
3315 }
3316
0bc58a91
EB		 3317 /* f_version caches the tgid value that the last readdir call couldn't
3318 * return. lseek aka telldir automagically resets f_version to 0.
3319 */
b488893a 3320 ns = filp->f_dentry->d_sb->s_fs_info;
2b47c361 3321 tid = (int)filp->f_version;
0bc58a91 3322 filp->f_version = 0;
ee6f779b 3323 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
0bc58a91 3324 task;
ee6f779b 3325 task = next_tid(task), filp->f_pos++) {
b488893a 3326 tid = task_pid_nr_ns(task, ns);
61a28784 3327 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
0bc58a91
EB		 3328 /* returning this tgid failed, save it as the first
3329 * pid for the next readir call */
2b47c361 3330 filp->f_version = (u64)tid;
0bc58a91 3331 put_task_struct(task);
1da177e4 3332 break;
0bc58a91 3333 }
1da177e4
LT		 3334 }
3335out:
99f89551
EB		 3336 put_task_struct(leader);
3337out_no_task:
1da177e4
LT		 3338 return retval;
3339}
6e66b52b
EB		 3340
3341static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3342{
3343 struct inode *inode = dentry->d_inode;
99f89551 3344 struct task_struct *p = get_proc_task(inode);
6e66b52b
EB		 3345 generic_fillattr(inode, stat);
3346
99f89551 3347 if (p) {
99f89551 3348 stat->nlink += get_nr_threads(p);
99f89551 3349 put_task_struct(p);
6e66b52b
EB		 3350 }
3351
3352 return 0;
3353}
28a6d671 3354
c5ef1c42 3355static const struct inode_operations proc_task_inode_operations = {
28a6d671
EB		 3356 .lookup = proc_task_lookup,
3357 .getattr = proc_task_getattr,
3358 .setattr = proc_setattr,
3359};
3360
00977a59 3361static const struct file_operations proc_task_operations = {
28a6d671
EB		 3362 .read = generic_read_dir,
3363 .readdir = proc_task_readdir,
6038f373 3364 .llseek = default_llseek,
28a6d671 3365};
Mirror https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git