]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - fs/proc/base.c
projects / mirror_ubuntu-bionic-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
fs: introduce a per-cpu last_ino allocator
[mirror_ubuntu-bionic-kernel.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);
4a3956c7
KH		 774 /* OK to pass negative loff_t, we can catch out-of-range */
775 file->f_mode |= FMODE_UNSIGNED_OFFSET;
1da177e4
LT		 776 return 0;
777}
778
779static ssize_t mem_read(struct file * file, char __user * buf,
780 size_t count, loff_t *ppos)
781{
2fddfeef 782 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1da177e4
LT		 783 char *page;
784 unsigned long src = *ppos;
785 int ret = -ESRCH;
786 struct mm_struct *mm;
787
99f89551
EB		 788 if (!task)
789 goto out_no_task;
790
638fa202 791 if (check_mem_permission(task))
1da177e4
LT		 792 goto out;
793
794 ret = -ENOMEM;
e12ba74d 795 page = (char *)__get_free_page(GFP_TEMPORARY);
1da177e4
LT		 796 if (!page)
797 goto out;
798
799 ret = 0;
800
801 mm = get_task_mm(task);
802 if (!mm)
803 goto out_free;
804
805 ret = -EIO;
806
807 if (file->private_data != (void*)((long)current->self_exec_id))
808 goto out_put;
809
810 ret = 0;
811
812 while (count > 0) {
813 int this_len, retval;
814
815 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
816 retval = access_process_vm(task, src, page, this_len, 0);
638fa202 817 if (!retval || check_mem_permission(task)) {
1da177e4
LT		 818 if (!ret)
819 ret = -EIO;
820 break;
821 }
822
823 if (copy_to_user(buf, page, retval)) {
824 ret = -EFAULT;
825 break;
826 }
827
828 ret += retval;
829 src += retval;
830 buf += retval;
831 count -= retval;
832 }
833 *ppos = src;
834
835out_put:
836 mmput(mm);
837out_free:
838 free_page((unsigned long) page);
839out:
99f89551
EB		 840 put_task_struct(task);
841out_no_task:
1da177e4
LT		 842 return ret;
843}
844
845#define mem_write NULL
846
847#ifndef mem_write
848/* This is a security hazard */
63967fa9 849static ssize_t mem_write(struct file * file, const char __user *buf,
1da177e4
LT		 850 size_t count, loff_t *ppos)
851{
f7ca54f4 852 int copied;
1da177e4 853 char *page;
2fddfeef 854 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1da177e4
LT		 855 unsigned long dst = *ppos;
856
99f89551
EB		 857 copied = -ESRCH;
858 if (!task)
859 goto out_no_task;
860
638fa202 861 if (check_mem_permission(task))
99f89551 862 goto out;
1da177e4 863
99f89551 864 copied = -ENOMEM;
e12ba74d 865 page = (char *)__get_free_page(GFP_TEMPORARY);
1da177e4 866 if (!page)
99f89551 867 goto out;
1da177e4 868
f7ca54f4 869 copied = 0;
1da177e4
LT		 870 while (count > 0) {
871 int this_len, retval;
872
873 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
874 if (copy_from_user(page, buf, this_len)) {
875 copied = -EFAULT;
876 break;
877 }
878 retval = access_process_vm(task, dst, page, this_len, 1);
879 if (!retval) {
880 if (!copied)
881 copied = -EIO;
882 break;
883 }
884 copied += retval;
885 buf += retval;
886 dst += retval;
887 count -= retval;
888 }
889 *ppos = dst;
890 free_page((unsigned long) page);
99f89551
EB		 891out:
892 put_task_struct(task);
893out_no_task:
1da177e4
LT		 894 return copied;
895}
896#endif
897
85863e47 898loff_t mem_lseek(struct file *file, loff_t offset, int orig)
1da177e4
LT		 899{
900 switch (orig) {
901 case 0:
902 file->f_pos = offset;
903 break;
904 case 1:
905 file->f_pos += offset;
906 break;
907 default:
908 return -EINVAL;
909 }
910 force_successful_syscall_return();
911 return file->f_pos;
912}
913
00977a59 914static const struct file_operations proc_mem_operations = {
1da177e4
LT		 915 .llseek = mem_lseek,
916 .read = mem_read,
917 .write = mem_write,
918 .open = mem_open,
919};
920
315e28c8
JP		 921static ssize_t environ_read(struct file *file, char __user *buf,
922 size_t count, loff_t *ppos)
923{
924 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
925 char *page;
926 unsigned long src = *ppos;
927 int ret = -ESRCH;
928 struct mm_struct *mm;
929
930 if (!task)
931 goto out_no_task;
932
006ebb40 933 if (!ptrace_may_access(task, PTRACE_MODE_READ))
315e28c8
JP		 934 goto out;
935
936 ret = -ENOMEM;
937 page = (char *)__get_free_page(GFP_TEMPORARY);
938 if (!page)
939 goto out;
940
941 ret = 0;
942
943 mm = get_task_mm(task);
944 if (!mm)
945 goto out_free;
946
947 while (count > 0) {
948 int this_len, retval, max_len;
949
950 this_len = mm->env_end - (mm->env_start + src);
951
952 if (this_len <= 0)
953 break;
954
955 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
956 this_len = (this_len > max_len) ? max_len : this_len;
957
958 retval = access_process_vm(task, (mm->env_start + src),
959 page, this_len, 0);
960
961 if (retval <= 0) {
962 ret = retval;
963 break;
964 }
965
966 if (copy_to_user(buf, page, retval)) {
967 ret = -EFAULT;
968 break;
969 }
970
971 ret += retval;
972 src += retval;
973 buf += retval;
974 count -= retval;
975 }
976 *ppos = src;
977
978 mmput(mm);
979out_free:
980 free_page((unsigned long) page);
981out:
982 put_task_struct(task);
983out_no_task:
984 return ret;
985}
986
987static const struct file_operations proc_environ_operations = {
988 .read = environ_read,
87df8424 989 .llseek = generic_file_llseek,
315e28c8
JP		 990};
991
1da177e4
LT		 992static ssize_t oom_adjust_read(struct file *file, char __user *buf,
993 size_t count, loff_t *ppos)
994{
2fddfeef 995 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
8578cea7 996 char buffer[PROC_NUMBUF];
1da177e4 997 size_t len;
28b83c51
KM		 998 int oom_adjust = OOM_DISABLE;
999 unsigned long flags;
1da177e4 1000
99f89551
EB		 1001 if (!task)
1002 return -ESRCH;
28b83c51
KM		 1003
1004 if (lock_task_sighand(task, &flags)) {
1005 oom_adjust = task->signal->oom_adj;
1006 unlock_task_sighand(task, &flags);
1007 }
1008
99f89551
EB		 1009 put_task_struct(task);
1010
8578cea7 1011 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
0c28f287
AM		 1012
1013 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1da177e4
LT		 1014}
1015
1016static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1017 size_t count, loff_t *ppos)
1018{
99f89551 1019 struct task_struct *task;
5d863b89
KM		 1020 char buffer[PROC_NUMBUF];
1021 long oom_adjust;
28b83c51 1022 unsigned long flags;
5d863b89 1023 int err;
1da177e4 1024
8578cea7
EB		 1025 memset(buffer, 0, sizeof(buffer));
1026 if (count > sizeof(buffer) - 1)
1027 count = sizeof(buffer) - 1;
1da177e4
LT		 1028 if (copy_from_user(buffer, buf, count))
1029 return -EFAULT;
5d863b89
KM		 1030
1031 err = strict_strtol(strstrip(buffer), 0, &oom_adjust);
1032 if (err)
1033 return -EINVAL;
8ac773b4
AD		 1034 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
1035 oom_adjust != OOM_DISABLE)
1da177e4 1036 return -EINVAL;
5d863b89 1037
2fddfeef 1038 task = get_proc_task(file->f_path.dentry->d_inode);
99f89551
EB		 1039 if (!task)
1040 return -ESRCH;
28b83c51
KM		 1041 if (!lock_task_sighand(task, &flags)) {
1042 put_task_struct(task);
1043 return -ESRCH;
1044 }
1045
1046 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
1047 unlock_task_sighand(task, &flags);
8fb4fc68
GJ		 1048 put_task_struct(task);
1049 return -EACCES;
1050 }
28b83c51 1051
51b1bd2a
DR		 1052 /*
1053 * Warn that /proc/pid/oom_adj is deprecated, see
1054 * Documentation/feature-removal-schedule.txt.
1055 */
1056 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, "
1057 "please use /proc/%d/oom_score_adj instead.\n",
1058 current->comm, task_pid_nr(current),
1059 task_pid_nr(task), task_pid_nr(task));
28b83c51 1060 task->signal->oom_adj = oom_adjust;
a63d83f4
DR		 1061 /*
1062 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1063 * value is always attainable.
1064 */
1065 if (task->signal->oom_adj == OOM_ADJUST_MAX)
1066 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
1067 else
1068 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
1069 -OOM_DISABLE;
28b83c51 1070 unlock_task_sighand(task, &flags);
99f89551 1071 put_task_struct(task);
5d863b89
KM		 1072
1073 return count;
1da177e4
LT		 1074}
1075
00977a59 1076static const struct file_operations proc_oom_adjust_operations = {
1da177e4
LT		 1077 .read = oom_adjust_read,
1078 .write = oom_adjust_write,
87df8424 1079 .llseek = generic_file_llseek,
1da177e4
LT		 1080};
1081
a63d83f4
DR		 1082static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1083 size_t count, loff_t *ppos)
1084{
1085 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1086 char buffer[PROC_NUMBUF];
1087 int oom_score_adj = OOM_SCORE_ADJ_MIN;
1088 unsigned long flags;
1089 size_t len;
1090
1091 if (!task)
1092 return -ESRCH;
1093 if (lock_task_sighand(task, &flags)) {
1094 oom_score_adj = task->signal->oom_score_adj;
1095 unlock_task_sighand(task, &flags);
1096 }
1097 put_task_struct(task);
1098 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
1099 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1100}
1101
1102static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1103 size_t count, loff_t *ppos)
1104{
1105 struct task_struct *task;
1106 char buffer[PROC_NUMBUF];
1107 unsigned long flags;
1108 long oom_score_adj;
1109 int err;
1110
1111 memset(buffer, 0, sizeof(buffer));
1112 if (count > sizeof(buffer) - 1)
1113 count = sizeof(buffer) - 1;
1114 if (copy_from_user(buffer, buf, count))
1115 return -EFAULT;
1116
1117 err = strict_strtol(strstrip(buffer), 0, &oom_score_adj);
1118 if (err)
1119 return -EINVAL;
1120 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1121 oom_score_adj > OOM_SCORE_ADJ_MAX)
1122 return -EINVAL;
1123
1124 task = get_proc_task(file->f_path.dentry->d_inode);
1125 if (!task)
1126 return -ESRCH;
1127 if (!lock_task_sighand(task, &flags)) {
1128 put_task_struct(task);
1129 return -ESRCH;
1130 }
1131 if (oom_score_adj < task->signal->oom_score_adj &&
1132 !capable(CAP_SYS_RESOURCE)) {
1133 unlock_task_sighand(task, &flags);
1134 put_task_struct(task);
1135 return -EACCES;
1136 }
1137
1138 task->signal->oom_score_adj = oom_score_adj;
1139 /*
1140 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1141 * always attainable.
1142 */
1143 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1144 task->signal->oom_adj = OOM_DISABLE;
1145 else
1146 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1147 OOM_SCORE_ADJ_MAX;
1148 unlock_task_sighand(task, &flags);
1149 put_task_struct(task);
1150 return count;
1151}
1152
1153static const struct file_operations proc_oom_score_adj_operations = {
1154 .read = oom_score_adj_read,
1155 .write = oom_score_adj_write,
6038f373 1156 .llseek = default_llseek,
a63d83f4
DR		 1157};
1158
1da177e4
LT		 1159#ifdef CONFIG_AUDITSYSCALL
1160#define TMPBUFLEN 21
1161static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1162 size_t count, loff_t *ppos)
1163{
2fddfeef 1164 struct inode * inode = file->f_path.dentry->d_inode;
99f89551 1165 struct task_struct *task = get_proc_task(inode);
1da177e4
LT		 1166 ssize_t length;
1167 char tmpbuf[TMPBUFLEN];
1168
99f89551
EB		 1169 if (!task)
1170 return -ESRCH;
1da177e4 1171 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
0c11b942 1172 audit_get_loginuid(task));
99f89551 1173 put_task_struct(task);
1da177e4
LT		 1174 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1175}
1176
1177static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1178 size_t count, loff_t *ppos)
1179{
2fddfeef 1180 struct inode * inode = file->f_path.dentry->d_inode;
1da177e4
LT		 1181 char *page, *tmp;
1182 ssize_t length;
1da177e4
LT		 1183 uid_t loginuid;
1184
1185 if (!capable(CAP_AUDIT_CONTROL))
1186 return -EPERM;
1187
7dc52157
PM		 1188 rcu_read_lock();
1189 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1190 rcu_read_unlock();
1da177e4 1191 return -EPERM;
7dc52157
PM		 1192 }
1193 rcu_read_unlock();
1da177e4 1194
e0182909
AV		 1195 if (count >= PAGE_SIZE)
1196 count = PAGE_SIZE - 1;
1da177e4
LT		 1197
1198 if (*ppos != 0) {
1199 /* No partial writes. */
1200 return -EINVAL;
1201 }
e12ba74d 1202 page = (char*)__get_free_page(GFP_TEMPORARY);
1da177e4
LT		 1203 if (!page)
1204 return -ENOMEM;
1205 length = -EFAULT;
1206 if (copy_from_user(page, buf, count))
1207 goto out_free_page;
1208
e0182909 1209 page[count] = '\0';
1da177e4
LT		 1210 loginuid = simple_strtoul(page, &tmp, 10);
1211 if (tmp == page) {
1212 length = -EINVAL;
1213 goto out_free_page;
1214
1215 }
99f89551 1216 length = audit_set_loginuid(current, loginuid);
1da177e4
LT		 1217 if (likely(length == 0))
1218 length = count;
1219
1220out_free_page:
1221 free_page((unsigned long) page);
1222 return length;
1223}
1224
00977a59 1225static const struct file_operations proc_loginuid_operations = {
1da177e4
LT		 1226 .read = proc_loginuid_read,
1227 .write = proc_loginuid_write,
87df8424 1228 .llseek = generic_file_llseek,
1da177e4 1229};
1e0bd755
EP		 1230
1231static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1232 size_t count, loff_t *ppos)
1233{
1234 struct inode * inode = file->f_path.dentry->d_inode;
1235 struct task_struct *task = get_proc_task(inode);
1236 ssize_t length;
1237 char tmpbuf[TMPBUFLEN];
1238
1239 if (!task)
1240 return -ESRCH;
1241 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1242 audit_get_sessionid(task));
1243 put_task_struct(task);
1244 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1245}
1246
1247static const struct file_operations proc_sessionid_operations = {
1248 .read = proc_sessionid_read,
87df8424 1249 .llseek = generic_file_llseek,
1e0bd755 1250};
1da177e4
LT		 1251#endif
1252
f4f154fd
AM		 1253#ifdef CONFIG_FAULT_INJECTION
1254static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1255 size_t count, loff_t *ppos)
1256{
1257 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1258 char buffer[PROC_NUMBUF];
1259 size_t len;
1260 int make_it_fail;
f4f154fd
AM		 1261
1262 if (!task)
1263 return -ESRCH;
1264 make_it_fail = task->make_it_fail;
1265 put_task_struct(task);
1266
1267 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
0c28f287
AM		 1268
1269 return simple_read_from_buffer(buf, count, ppos, buffer, len);
f4f154fd
AM		 1270}
1271
1272static ssize_t proc_fault_inject_write(struct file * file,
1273 const char __user * buf, size_t count, loff_t *ppos)
1274{
1275 struct task_struct *task;
1276 char buffer[PROC_NUMBUF], *end;
1277 int make_it_fail;
1278
1279 if (!capable(CAP_SYS_RESOURCE))
1280 return -EPERM;
1281 memset(buffer, 0, sizeof(buffer));
1282 if (count > sizeof(buffer) - 1)
1283 count = sizeof(buffer) - 1;
1284 if (copy_from_user(buffer, buf, count))
1285 return -EFAULT;
cba8aafe
VL		 1286 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1287 if (*end)
1288 return -EINVAL;
f4f154fd
AM		 1289 task = get_proc_task(file->f_dentry->d_inode);
1290 if (!task)
1291 return -ESRCH;
1292 task->make_it_fail = make_it_fail;
1293 put_task_struct(task);
cba8aafe
VL		 1294
1295 return count;
f4f154fd
AM		 1296}
1297
00977a59 1298static const struct file_operations proc_fault_inject_operations = {
f4f154fd
AM		 1299 .read = proc_fault_inject_read,
1300 .write = proc_fault_inject_write,
87df8424 1301 .llseek = generic_file_llseek,
f4f154fd
AM		 1302};
1303#endif
1304
9745512c 1305
43ae34cb
IM		 1306#ifdef CONFIG_SCHED_DEBUG
1307/*
1308 * Print out various scheduling related per-task fields:
1309 */
1310static int sched_show(struct seq_file *m, void *v)
1311{
1312 struct inode *inode = m->private;
1313 struct task_struct *p;
1314
43ae34cb
IM		 1315 p = get_proc_task(inode);
1316 if (!p)
1317 return -ESRCH;
1318 proc_sched_show_task(p, m);
1319
1320 put_task_struct(p);
1321
1322 return 0;
1323}
1324
1325static ssize_t
1326sched_write(struct file *file, const char __user *buf,
1327 size_t count, loff_t *offset)
1328{
1329 struct inode *inode = file->f_path.dentry->d_inode;
1330 struct task_struct *p;
1331
43ae34cb
IM		 1332 p = get_proc_task(inode);
1333 if (!p)
1334 return -ESRCH;
1335 proc_sched_set_task(p);
1336
1337 put_task_struct(p);
1338
1339 return count;
1340}
1341
1342static int sched_open(struct inode *inode, struct file *filp)
1343{
1344 int ret;
1345
1346 ret = single_open(filp, sched_show, NULL);
1347 if (!ret) {
1348 struct seq_file *m = filp->private_data;
1349
1350 m->private = inode;
1351 }
1352 return ret;
1353}
1354
1355static const struct file_operations proc_pid_sched_operations = {
1356 .open = sched_open,
1357 .read = seq_read,
1358 .write = sched_write,
1359 .llseek = seq_lseek,
5ea473a1 1360 .release = single_release,
43ae34cb
IM		 1361};
1362
1363#endif
1364
4614a696
JS		 1365static ssize_t comm_write(struct file *file, const char __user *buf,
1366 size_t count, loff_t *offset)
1367{
1368 struct inode *inode = file->f_path.dentry->d_inode;
1369 struct task_struct *p;
1370 char buffer[TASK_COMM_LEN];
1371
1372 memset(buffer, 0, sizeof(buffer));
1373 if (count > sizeof(buffer) - 1)
1374 count = sizeof(buffer) - 1;
1375 if (copy_from_user(buffer, buf, count))
1376 return -EFAULT;
1377
1378 p = get_proc_task(inode);
1379 if (!p)
1380 return -ESRCH;
1381
1382 if (same_thread_group(current, p))
1383 set_task_comm(p, buffer);
1384 else
1385 count = -EINVAL;
1386
1387 put_task_struct(p);
1388
1389 return count;
1390}
1391
1392static int comm_show(struct seq_file *m, void *v)
1393{
1394 struct inode *inode = m->private;
1395 struct task_struct *p;
1396
1397 p = get_proc_task(inode);
1398 if (!p)
1399 return -ESRCH;
1400
1401 task_lock(p);
1402 seq_printf(m, "%s\n", p->comm);
1403 task_unlock(p);
1404
1405 put_task_struct(p);
1406
1407 return 0;
1408}
1409
1410static int comm_open(struct inode *inode, struct file *filp)
1411{
1412 int ret;
1413
1414 ret = single_open(filp, comm_show, NULL);
1415 if (!ret) {
1416 struct seq_file *m = filp->private_data;
1417
1418 m->private = inode;
1419 }
1420 return ret;
1421}
1422
1423static const struct file_operations proc_pid_set_comm_operations = {
1424 .open = comm_open,
1425 .read = seq_read,
1426 .write = comm_write,
1427 .llseek = seq_lseek,
1428 .release = single_release,
1429};
1430
925d1c40
MH		 1431/*
1432 * We added or removed a vma mapping the executable. The vmas are only mapped
1433 * during exec and are not mapped with the mmap system call.
1434 * Callers must hold down_write() on the mm's mmap_sem for these
1435 */
1436void added_exe_file_vma(struct mm_struct *mm)
1437{
1438 mm->num_exe_file_vmas++;
1439}
1440
1441void removed_exe_file_vma(struct mm_struct *mm)
1442{
1443 mm->num_exe_file_vmas--;
1444 if ((mm->num_exe_file_vmas == 0) && mm->exe_file){
1445 fput(mm->exe_file);
1446 mm->exe_file = NULL;
1447 }
1448
1449}
1450
1451void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1452{
1453 if (new_exe_file)
1454 get_file(new_exe_file);
1455 if (mm->exe_file)
1456 fput(mm->exe_file);
1457 mm->exe_file = new_exe_file;
1458 mm->num_exe_file_vmas = 0;
1459}
1460
1461struct file *get_mm_exe_file(struct mm_struct *mm)
1462{
1463 struct file *exe_file;
1464
1465 /* We need mmap_sem to protect against races with removal of
1466 * VM_EXECUTABLE vmas */
1467 down_read(&mm->mmap_sem);
1468 exe_file = mm->exe_file;
1469 if (exe_file)
1470 get_file(exe_file);
1471 up_read(&mm->mmap_sem);
1472 return exe_file;
1473}
1474
1475void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
1476{
1477 /* It's safe to write the exe_file pointer without exe_file_lock because
1478 * this is called during fork when the task is not yet in /proc */
1479 newmm->exe_file = get_mm_exe_file(oldmm);
1480}
1481
1482static int proc_exe_link(struct inode *inode, struct path *exe_path)
1483{
1484 struct task_struct *task;
1485 struct mm_struct *mm;
1486 struct file *exe_file;
1487
1488 task = get_proc_task(inode);
1489 if (!task)
1490 return -ENOENT;
1491 mm = get_task_mm(task);
1492 put_task_struct(task);
1493 if (!mm)
1494 return -ENOENT;
1495 exe_file = get_mm_exe_file(mm);
1496 mmput(mm);
1497 if (exe_file) {
1498 *exe_path = exe_file->f_path;
1499 path_get(&exe_file->f_path);
1500 fput(exe_file);
1501 return 0;
1502 } else
1503 return -ENOENT;
1504}
1505
008b150a 1506static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1da177e4
LT		 1507{
1508 struct inode *inode = dentry->d_inode;
1509 int error = -EACCES;
1510
1511 /* We don't need a base pointer in the /proc filesystem */
1d957f9b 1512 path_put(&nd->path);
1da177e4 1513
778c1144
EB		 1514 /* Are we allowed to snoop on the tasks file descriptors? */
1515 if (!proc_fd_access_allowed(inode))
1da177e4 1516 goto out;
1da177e4 1517
3dcd25f3 1518 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
1da177e4 1519out:
008b150a 1520 return ERR_PTR(error);
1da177e4
LT		 1521}
1522
3dcd25f3 1523static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1da177e4 1524{
e12ba74d 1525 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
3dcd25f3 1526 char *pathname;
1da177e4
LT		 1527 int len;
1528
1529 if (!tmp)
1530 return -ENOMEM;
0c28f287 1531
8df9d1a4 1532 pathname = d_path_with_unreachable(path, tmp, PAGE_SIZE);
3dcd25f3
JB		 1533 len = PTR_ERR(pathname);
1534 if (IS_ERR(pathname))
1da177e4 1535 goto out;
3dcd25f3 1536 len = tmp + PAGE_SIZE - 1 - pathname;
1da177e4
LT		 1537
1538 if (len > buflen)
1539 len = buflen;
3dcd25f3 1540 if (copy_to_user(buffer, pathname, len))
1da177e4
LT		 1541 len = -EFAULT;
1542 out:
1543 free_page((unsigned long)tmp);
1544 return len;
1545}
1546
1547static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1548{
1549 int error = -EACCES;
1550 struct inode *inode = dentry->d_inode;
3dcd25f3 1551 struct path path;
1da177e4 1552
778c1144
EB		 1553 /* Are we allowed to snoop on the tasks file descriptors? */
1554 if (!proc_fd_access_allowed(inode))
1da177e4 1555 goto out;
1da177e4 1556
3dcd25f3 1557 error = PROC_I(inode)->op.proc_get_link(inode, &path);
1da177e4
LT		 1558 if (error)
1559 goto out;
1560
3dcd25f3
JB		 1561 error = do_proc_readlink(&path, buffer, buflen);
1562 path_put(&path);
1da177e4 1563out:
1da177e4
LT		 1564 return error;
1565}
1566
c5ef1c42 1567static const struct inode_operations proc_pid_link_inode_operations = {
1da177e4 1568 .readlink = proc_pid_readlink,
6d76fa58
LT		 1569 .follow_link = proc_pid_follow_link,
1570 .setattr = proc_setattr,
1da177e4
LT		 1571};
1572
28a6d671
EB		 1573
1574/* building an inode */
1575
1576static int task_dumpable(struct task_struct *task)
1da177e4 1577{
28a6d671
EB		 1578 int dumpable = 0;
1579 struct mm_struct *mm;
1da177e4 1580
28a6d671
EB		 1581 task_lock(task);
1582 mm = task->mm;
1583 if (mm)
6c5d5238 1584 dumpable = get_dumpable(mm);
28a6d671
EB		 1585 task_unlock(task);
1586 if(dumpable == 1)
1587 return 1;
1588 return 0;
1589}
1da177e4 1590
1da177e4 1591
61a28784 1592static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
28a6d671
EB		 1593{
1594 struct inode * inode;
1595 struct proc_inode *ei;
c69e8d9c 1596 const struct cred *cred;
1da177e4 1597
28a6d671 1598 /* We need a new inode */
1da177e4 1599
28a6d671
EB		 1600 inode = new_inode(sb);
1601 if (!inode)
1602 goto out;
1603
1604 /* Common stuff */
1605 ei = PROC_I(inode);
1606 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
28a6d671
EB		 1607 inode->i_op = &proc_def_inode_operations;
1608
1609 /*
1610 * grab the reference to task.
1611 */
1a657f78 1612 ei->pid = get_task_pid(task, PIDTYPE_PID);
28a6d671
EB		 1613 if (!ei->pid)
1614 goto out_unlock;
1615
28a6d671 1616 if (task_dumpable(task)) {
c69e8d9c
DH		 1617 rcu_read_lock();
1618 cred = __task_cred(task);
1619 inode->i_uid = cred->euid;
1620 inode->i_gid = cred->egid;
1621 rcu_read_unlock();
1da177e4 1622 }
28a6d671
EB		 1623 security_task_to_inode(task, inode);
1624
1da177e4 1625out:
28a6d671
EB		 1626 return inode;
1627
1628out_unlock:
1629 iput(inode);
1630 return NULL;
1da177e4
LT		 1631}
1632
28a6d671 1633static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1da177e4 1634{
1da177e4 1635 struct inode *inode = dentry->d_inode;
28a6d671 1636 struct task_struct *task;
c69e8d9c
DH		 1637 const struct cred *cred;
1638
28a6d671 1639 generic_fillattr(inode, stat);
1da177e4 1640
28a6d671
EB		 1641 rcu_read_lock();
1642 stat->uid = 0;
1643 stat->gid = 0;
1644 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1645 if (task) {
1646 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1647 task_dumpable(task)) {
c69e8d9c
DH		 1648 cred = __task_cred(task);
1649 stat->uid = cred->euid;
1650 stat->gid = cred->egid;
1da177e4
LT		 1651 }
1652 }
28a6d671 1653 rcu_read_unlock();
d6e71144 1654 return 0;
1da177e4
LT		 1655}
1656
1da177e4
LT		 1657/* dentry stuff */
1658
1659/*
1660 * Exceptional case: normally we are not allowed to unhash a busy
1661 * directory. In this case, however, we can do it - no aliasing problems
1662 * due to the way we treat inodes.
1663 *
1664 * Rewrite the inode's ownerships here because the owning task may have
1665 * performed a setuid(), etc.
99f89551
EB		 1666 *
1667 * Before the /proc/pid/status file was created the only way to read
1668 * the effective uid of a /process was to stat /proc/pid. Reading
1669 * /proc/pid/status is slow enough that procps and other packages
1670 * kept stating /proc/pid. To keep the rules in /proc simple I have
1671 * made this apply to all per process world readable and executable
1672 * directories.
1da177e4
LT		 1673 */
1674static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1675{
1676 struct inode *inode = dentry->d_inode;
99f89551 1677 struct task_struct *task = get_proc_task(inode);
c69e8d9c
DH		 1678 const struct cred *cred;
1679
99f89551
EB		 1680 if (task) {
1681 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1682 task_dumpable(task)) {
c69e8d9c
DH		 1683 rcu_read_lock();
1684 cred = __task_cred(task);
1685 inode->i_uid = cred->euid;
1686 inode->i_gid = cred->egid;
1687 rcu_read_unlock();
1da177e4
LT		 1688 } else {
1689 inode->i_uid = 0;
1690 inode->i_gid = 0;
1691 }
9ee8ab9f 1692 inode->i_mode &= ~(S_ISUID | S_ISGID);
1da177e4 1693 security_task_to_inode(task, inode);
99f89551 1694 put_task_struct(task);
1da177e4
LT		 1695 return 1;
1696 }
1697 d_drop(dentry);
1698 return 0;
1699}
1700
28a6d671 1701static int pid_delete_dentry(struct dentry * dentry)
99f89551 1702{
28a6d671
EB		 1703 /* Is the task we represent dead?
1704 * If so, then don't put the dentry on the lru list,
1705 * kill it immediately.
1706 */
1707 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1708}
1709
d72f71eb 1710static const struct dentry_operations pid_dentry_operations =
28a6d671
EB		 1711{
1712 .d_revalidate = pid_revalidate,
1713 .d_delete = pid_delete_dentry,
1714};
1715
1716/* Lookups */
1717
c5141e6d
ED		 1718typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1719 struct task_struct *, const void *);
61a28784 1720
1c0d04c9
EB		 1721/*
1722 * Fill a directory entry.
1723 *
1724 * If possible create the dcache entry and derive our inode number and
1725 * file type from dcache entry.
1726 *
1727 * Since all of the proc inode numbers are dynamically generated, the inode
1728 * numbers do not exist until the inode is cache. This means creating the
1729 * the dcache entry in readdir is necessary to keep the inode numbers
1730 * reported by readdir in sync with the inode numbers reported
1731 * by stat.
1732 */
61a28784
EB		 1733static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1734 char *name, int len,
c5141e6d 1735 instantiate_t instantiate, struct task_struct *task, const void *ptr)
61a28784 1736{
2fddfeef 1737 struct dentry *child, *dir = filp->f_path.dentry;
61a28784
EB		 1738 struct inode *inode;
1739 struct qstr qname;
1740 ino_t ino = 0;
1741 unsigned type = DT_UNKNOWN;
1742
1743 qname.name = name;
1744 qname.len = len;
1745 qname.hash = full_name_hash(name, len);
1746
1747 child = d_lookup(dir, &qname);
1748 if (!child) {
1749 struct dentry *new;
1750 new = d_alloc(dir, &qname);
1751 if (new) {
1752 child = instantiate(dir->d_inode, new, task, ptr);
1753 if (child)
1754 dput(new);
1755 else
1756 child = new;
1757 }
1758 }
1759 if (!child || IS_ERR(child) || !child->d_inode)
1760 goto end_instantiate;
1761 inode = child->d_inode;
1762 if (inode) {
1763 ino = inode->i_ino;
1764 type = inode->i_mode >> 12;
1765 }
1766 dput(child);
1767end_instantiate:
1768 if (!ino)
1769 ino = find_inode_number(dir, &qname);
1770 if (!ino)
1771 ino = 1;
1772 return filldir(dirent, name, len, filp->f_pos, ino, type);
1773}
1774
28a6d671
EB		 1775static unsigned name_to_int(struct dentry *dentry)
1776{
1777 const char *name = dentry->d_name.name;
1778 int len = dentry->d_name.len;
1779 unsigned n = 0;
1780
1781 if (len > 1 && *name == '0')
1782 goto out;
1783 while (len-- > 0) {
1784 unsigned c = *name++ - '0';
1785 if (c > 9)
1786 goto out;
1787 if (n >= (~0U-9)/10)
1788 goto out;
1789 n *= 10;
1790 n += c;
1791 }
1792 return n;
1793out:
1794 return ~0U;
1795}
1796
27932742
MS		 1797#define PROC_FDINFO_MAX 64
1798
3dcd25f3 1799static int proc_fd_info(struct inode *inode, struct path *path, char *info)
28a6d671
EB		 1800{
1801 struct task_struct *task = get_proc_task(inode);
1802 struct files_struct *files = NULL;
1803 struct file *file;
1804 int fd = proc_fd(inode);
99f89551 1805
99f89551 1806 if (task) {
28a6d671
EB		 1807 files = get_files_struct(task);
1808 put_task_struct(task);
1809 }
1810 if (files) {
1811 /*
1812 * We are not taking a ref to the file structure, so we must
1813 * hold ->file_lock.
1814 */
1815 spin_lock(&files->file_lock);
1816 file = fcheck_files(files, fd);
1817 if (file) {
3dcd25f3
JB		 1818 if (path) {
1819 *path = file->f_path;
1820 path_get(&file->f_path);
1821 }
27932742
MS		 1822 if (info)
1823 snprintf(info, PROC_FDINFO_MAX,
1824 "pos:\t%lli\n"
1825 "flags:\t0%o\n",
1826 (long long) file->f_pos,
1827 file->f_flags);
28a6d671
EB		 1828 spin_unlock(&files->file_lock);
1829 put_files_struct(files);
1830 return 0;
99f89551 1831 }
28a6d671
EB		 1832 spin_unlock(&files->file_lock);
1833 put_files_struct(files);
99f89551 1834 }
28a6d671 1835 return -ENOENT;
99f89551
EB		 1836}
1837
3dcd25f3 1838static int proc_fd_link(struct inode *inode, struct path *path)
27932742 1839{
3dcd25f3 1840 return proc_fd_info(inode, path, NULL);
27932742
MS		 1841}
1842
1da177e4
LT		 1843static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1844{
1845 struct inode *inode = dentry->d_inode;
99f89551 1846 struct task_struct *task = get_proc_task(inode);
aed7a6c4 1847 int fd = proc_fd(inode);
1da177e4 1848 struct files_struct *files;
c69e8d9c 1849 const struct cred *cred;
1da177e4 1850
99f89551
EB		 1851 if (task) {
1852 files = get_files_struct(task);
1853 if (files) {
1854 rcu_read_lock();
1855 if (fcheck_files(files, fd)) {
1856 rcu_read_unlock();
1857 put_files_struct(files);
1858 if (task_dumpable(task)) {
c69e8d9c
DH		 1859 rcu_read_lock();
1860 cred = __task_cred(task);
1861 inode->i_uid = cred->euid;
1862 inode->i_gid = cred->egid;
1863 rcu_read_unlock();
99f89551
EB		 1864 } else {
1865 inode->i_uid = 0;
1866 inode->i_gid = 0;
1867 }
9ee8ab9f 1868 inode->i_mode &= ~(S_ISUID | S_ISGID);
99f89551
EB		 1869 security_task_to_inode(task, inode);
1870 put_task_struct(task);
1871 return 1;
1872 }
b835996f 1873 rcu_read_unlock();
1da177e4 1874 put_files_struct(files);
1da177e4 1875 }
99f89551 1876 put_task_struct(task);
1da177e4
LT		 1877 }
1878 d_drop(dentry);
1879 return 0;
1880}
1881
d72f71eb 1882static const struct dentry_operations tid_fd_dentry_operations =
1da177e4
LT		 1883{
1884 .d_revalidate = tid_fd_revalidate,
1885 .d_delete = pid_delete_dentry,
1886};
1887
444ceed8 1888static struct dentry *proc_fd_instantiate(struct inode *dir,
c5141e6d 1889 struct dentry *dentry, struct task_struct *task, const void *ptr)
1da177e4 1890{
c5141e6d 1891 unsigned fd = *(const unsigned *)ptr;
444ceed8
EB		 1892 struct file *file;
1893 struct files_struct *files;
1894 struct inode *inode;
1895 struct proc_inode *ei;
1896 struct dentry *error = ERR_PTR(-ENOENT);
1da177e4 1897
61a28784 1898 inode = proc_pid_make_inode(dir->i_sb, task);
1da177e4
LT		 1899 if (!inode)
1900 goto out;
1901 ei = PROC_I(inode);
aed7a6c4 1902 ei->fd = fd;
1da177e4
LT		 1903 files = get_files_struct(task);
1904 if (!files)
444ceed8 1905 goto out_iput;
1da177e4 1906 inode->i_mode = S_IFLNK;
ca99c1da
DS		 1907
1908 /*
1909 * We are not taking a ref to the file structure, so we must
1910 * hold ->file_lock.
1911 */
1912 spin_lock(&files->file_lock);
1da177e4
LT		 1913 file = fcheck_files(files, fd);
1914 if (!file)
444ceed8 1915 goto out_unlock;
aeb5d727 1916 if (file->f_mode & FMODE_READ)
1da177e4 1917 inode->i_mode |= S_IRUSR | S_IXUSR;
aeb5d727 1918 if (file->f_mode & FMODE_WRITE)
1da177e4 1919 inode->i_mode |= S_IWUSR | S_IXUSR;
ca99c1da 1920 spin_unlock(&files->file_lock);
1da177e4 1921 put_files_struct(files);
444ceed8 1922
1da177e4
LT		 1923 inode->i_op = &proc_pid_link_inode_operations;
1924 inode->i_size = 64;
1925 ei->op.proc_get_link = proc_fd_link;
1926 dentry->d_op = &tid_fd_dentry_operations;
1927 d_add(dentry, inode);
cd6a3ce9
EB		 1928 /* Close the race of the process dying before we return the dentry */
1929 if (tid_fd_revalidate(dentry, NULL))
444ceed8 1930 error = NULL;
1da177e4 1931
444ceed8
EB		 1932 out:
1933 return error;
1934out_unlock:
ca99c1da 1935 spin_unlock(&files->file_lock);
1da177e4 1936 put_files_struct(files);
444ceed8 1937out_iput:
1da177e4 1938 iput(inode);
cd6a3ce9 1939 goto out;
1da177e4
LT		 1940}
1941
27932742
MS		 1942static struct dentry *proc_lookupfd_common(struct inode *dir,
1943 struct dentry *dentry,
1944 instantiate_t instantiate)
444ceed8
EB		 1945{
1946 struct task_struct *task = get_proc_task(dir);
1947 unsigned fd = name_to_int(dentry);
1948 struct dentry *result = ERR_PTR(-ENOENT);
1949
1950 if (!task)
1951 goto out_no_task;
1952 if (fd == ~0U)
1953 goto out;
1954
27932742 1955 result = instantiate(dir, dentry, task, &fd);
444ceed8
EB		 1956out:
1957 put_task_struct(task);
1958out_no_task:
1959 return result;
1960}
1961
27932742
MS		 1962static int proc_readfd_common(struct file * filp, void * dirent,
1963 filldir_t filldir, instantiate_t instantiate)
28a6d671 1964{
2fddfeef 1965 struct dentry *dentry = filp->f_path.dentry;
28a6d671
EB		 1966 struct inode *inode = dentry->d_inode;
1967 struct task_struct *p = get_proc_task(inode);
457c2510 1968 unsigned int fd, ino;
28a6d671 1969 int retval;
28a6d671 1970 struct files_struct * files;
1da177e4 1971
28a6d671
EB		 1972 retval = -ENOENT;
1973 if (!p)
1974 goto out_no_task;
1975 retval = 0;
28a6d671
EB		 1976
1977 fd = filp->f_pos;
1978 switch (fd) {
1979 case 0:
1980 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1981 goto out;
1982 filp->f_pos++;
1983 case 1:
1984 ino = parent_ino(dentry);
1985 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1986 goto out;
1987 filp->f_pos++;
1988 default:
1989 files = get_files_struct(p);
1990 if (!files)
1991 goto out;
1992 rcu_read_lock();
28a6d671 1993 for (fd = filp->f_pos-2;
9b4f526c 1994 fd < files_fdtable(files)->max_fds;
28a6d671 1995 fd++, filp->f_pos++) {
27932742
MS		 1996 char name[PROC_NUMBUF];
1997 int len;
28a6d671
EB		 1998
1999 if (!fcheck_files(files, fd))
2000 continue;
2001 rcu_read_unlock();
2002
27932742
MS		 2003 len = snprintf(name, sizeof(name), "%d", fd);
2004 if (proc_fill_cache(filp, dirent, filldir,
2005 name, len, instantiate,
2006 p, &fd) < 0) {
28a6d671
EB		 2007 rcu_read_lock();
2008 break;
2009 }
2010 rcu_read_lock();
2011 }
2012 rcu_read_unlock();
2013 put_files_struct(files);
2014 }
2015out:
2016 put_task_struct(p);
2017out_no_task:
2018 return retval;
2019}
2020
27932742
MS		 2021static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
2022 struct nameidata *nd)
2023{
2024 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
2025}
2026
2027static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
2028{
2029 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
2030}
2031
2032static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
2033 size_t len, loff_t *ppos)
2034{
2035 char tmp[PROC_FDINFO_MAX];
3dcd25f3 2036 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
27932742
MS		 2037 if (!err)
2038 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
2039 return err;
2040}
2041
2042static const struct file_operations proc_fdinfo_file_operations = {
87df8424 2043 .open = nonseekable_open,
27932742 2044 .read = proc_fdinfo_read,
6038f373 2045 .llseek = no_llseek,
27932742
MS		 2046};
2047
00977a59 2048static const struct file_operations proc_fd_operations = {
28a6d671
EB		 2049 .read = generic_read_dir,
2050 .readdir = proc_readfd,
6038f373 2051 .llseek = default_llseek,
1da177e4
LT		 2052};
2053
8948e11f
AD		 2054/*
2055 * /proc/pid/fd needs a special permission handler so that a process can still
2056 * access /proc/self/fd after it has executed a setuid().
2057 */
e6305c43 2058static int proc_fd_permission(struct inode *inode, int mask)
8948e11f
AD		 2059{
2060 int rv;
2061
2062 rv = generic_permission(inode, mask, NULL);
2063 if (rv == 0)
2064 return 0;
2065 if (task_pid(current) == proc_pid(inode))
2066 rv = 0;
2067 return rv;
2068}
2069
1da177e4
LT		 2070/*
2071 * proc directories can do almost nothing..
2072 */
c5ef1c42 2073static const struct inode_operations proc_fd_inode_operations = {
1da177e4 2074 .lookup = proc_lookupfd,
8948e11f 2075 .permission = proc_fd_permission,
6d76fa58 2076 .setattr = proc_setattr,
1da177e4
LT		 2077};
2078
27932742
MS		 2079static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2080 struct dentry *dentry, struct task_struct *task, const void *ptr)
2081{
2082 unsigned fd = *(unsigned *)ptr;
2083 struct inode *inode;
2084 struct proc_inode *ei;
2085 struct dentry *error = ERR_PTR(-ENOENT);
2086
2087 inode = proc_pid_make_inode(dir->i_sb, task);
2088 if (!inode)
2089 goto out;
2090 ei = PROC_I(inode);
2091 ei->fd = fd;
2092 inode->i_mode = S_IFREG | S_IRUSR;
2093 inode->i_fop = &proc_fdinfo_file_operations;
2094 dentry->d_op = &tid_fd_dentry_operations;
2095 d_add(dentry, inode);
2096 /* Close the race of the process dying before we return the dentry */
2097 if (tid_fd_revalidate(dentry, NULL))
2098 error = NULL;
2099
2100 out:
2101 return error;
2102}
2103
2104static struct dentry *proc_lookupfdinfo(struct inode *dir,
2105 struct dentry *dentry,
2106 struct nameidata *nd)
2107{
2108 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2109}
2110
2111static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2112{
2113 return proc_readfd_common(filp, dirent, filldir,
2114 proc_fdinfo_instantiate);
2115}
2116
2117static const struct file_operations proc_fdinfo_operations = {
2118 .read = generic_read_dir,
2119 .readdir = proc_readfdinfo,
6038f373 2120 .llseek = default_llseek,
27932742
MS		 2121};
2122
2123/*
2124 * proc directories can do almost nothing..
2125 */
2126static const struct inode_operations proc_fdinfo_inode_operations = {
2127 .lookup = proc_lookupfdinfo,
2128 .setattr = proc_setattr,
2129};
2130
2131
444ceed8 2132static struct dentry *proc_pident_instantiate(struct inode *dir,
c5141e6d 2133 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8 2134{
c5141e6d 2135 const struct pid_entry *p = ptr;
444ceed8
EB		 2136 struct inode *inode;
2137 struct proc_inode *ei;
bd6daba9 2138 struct dentry *error = ERR_PTR(-ENOENT);
444ceed8 2139
61a28784 2140 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 2141 if (!inode)
2142 goto out;
2143
2144 ei = PROC_I(inode);
2145 inode->i_mode = p->mode;
2146 if (S_ISDIR(inode->i_mode))
2147 inode->i_nlink = 2; /* Use getattr to fix if necessary */
2148 if (p->iop)
2149 inode->i_op = p->iop;
2150 if (p->fop)
2151 inode->i_fop = p->fop;
2152 ei->op = p->op;
2153 dentry->d_op = &pid_dentry_operations;
2154 d_add(dentry, inode);
2155 /* Close the race of the process dying before we return the dentry */
2156 if (pid_revalidate(dentry, NULL))
2157 error = NULL;
2158out:
2159 return error;
2160}
2161
1da177e4
LT		 2162static struct dentry *proc_pident_lookup(struct inode *dir,
2163 struct dentry *dentry,
c5141e6d 2164 const struct pid_entry *ents,
7bcd6b0e 2165 unsigned int nents)
1da177e4 2166{
cd6a3ce9 2167 struct dentry *error;
99f89551 2168 struct task_struct *task = get_proc_task(dir);
c5141e6d 2169 const struct pid_entry *p, *last;
1da177e4 2170
cd6a3ce9 2171 error = ERR_PTR(-ENOENT);
1da177e4 2172
99f89551
EB		 2173 if (!task)
2174 goto out_no_task;
1da177e4 2175
20cdc894
EB		 2176 /*
2177 * Yes, it does not scale. And it should not. Don't add
2178 * new entries into /proc/<tgid>/ without very good reasons.
2179 */
7bcd6b0e
EB		 2180 last = &ents[nents - 1];
2181 for (p = ents; p <= last; p++) {
1da177e4
LT		 2182 if (p->len != dentry->d_name.len)
2183 continue;
2184 if (!memcmp(dentry->d_name.name, p->name, p->len))
2185 break;
2186 }
7bcd6b0e 2187 if (p > last)
1da177e4
LT		 2188 goto out;
2189
444ceed8 2190 error = proc_pident_instantiate(dir, dentry, task, p);
1da177e4 2191out:
99f89551
EB		 2192 put_task_struct(task);
2193out_no_task:
cd6a3ce9 2194 return error;
1da177e4
LT		 2195}
2196
c5141e6d
ED		 2197static int proc_pident_fill_cache(struct file *filp, void *dirent,
2198 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
61a28784
EB		 2199{
2200 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2201 proc_pident_instantiate, task, p);
2202}
2203
28a6d671
EB		 2204static int proc_pident_readdir(struct file *filp,
2205 void *dirent, filldir_t filldir,
c5141e6d 2206 const struct pid_entry *ents, unsigned int nents)
28a6d671
EB		 2207{
2208 int i;
2fddfeef 2209 struct dentry *dentry = filp->f_path.dentry;
28a6d671
EB		 2210 struct inode *inode = dentry->d_inode;
2211 struct task_struct *task = get_proc_task(inode);
c5141e6d 2212 const struct pid_entry *p, *last;
28a6d671
EB		 2213 ino_t ino;
2214 int ret;
2215
2216 ret = -ENOENT;
2217 if (!task)
61a28784 2218 goto out_no_task;
28a6d671
EB		 2219
2220 ret = 0;
28a6d671
EB		 2221 i = filp->f_pos;
2222 switch (i) {
2223 case 0:
2224 ino = inode->i_ino;
2225 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2226 goto out;
2227 i++;
2228 filp->f_pos++;
2229 /* fall through */
2230 case 1:
2231 ino = parent_ino(dentry);
2232 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2233 goto out;
2234 i++;
2235 filp->f_pos++;
2236 /* fall through */
2237 default:
2238 i -= 2;
2239 if (i >= nents) {
2240 ret = 1;
2241 goto out;
2242 }
2243 p = ents + i;
7bcd6b0e
EB		 2244 last = &ents[nents - 1];
2245 while (p <= last) {
61a28784 2246 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
28a6d671
EB		 2247 goto out;
2248 filp->f_pos++;
2249 p++;
2250 }
2251 }
2252
2253 ret = 1;
2254out:
61a28784
EB		 2255 put_task_struct(task);
2256out_no_task:
28a6d671 2257 return ret;
1da177e4
LT		 2258}
2259
28a6d671
EB		 2260#ifdef CONFIG_SECURITY
2261static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2262 size_t count, loff_t *ppos)
2263{
2fddfeef 2264 struct inode * inode = file->f_path.dentry->d_inode;
04ff9708 2265 char *p = NULL;
28a6d671
EB		 2266 ssize_t length;
2267 struct task_struct *task = get_proc_task(inode);
2268
28a6d671 2269 if (!task)
04ff9708 2270 return -ESRCH;
28a6d671
EB		 2271
2272 length = security_getprocattr(task,
2fddfeef 2273 (char*)file->f_path.dentry->d_name.name,
04ff9708 2274 &p);
28a6d671 2275 put_task_struct(task);
04ff9708
AV		 2276 if (length > 0)
2277 length = simple_read_from_buffer(buf, count, ppos, p, length);
2278 kfree(p);
28a6d671 2279 return length;
1da177e4
LT		 2280}
2281
28a6d671
EB		 2282static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2283 size_t count, loff_t *ppos)
2284{
2fddfeef 2285 struct inode * inode = file->f_path.dentry->d_inode;
28a6d671
EB		 2286 char *page;
2287 ssize_t length;
2288 struct task_struct *task = get_proc_task(inode);
2289
2290 length = -ESRCH;
2291 if (!task)
2292 goto out_no_task;
2293 if (count > PAGE_SIZE)
2294 count = PAGE_SIZE;
2295
2296 /* No partial writes. */
2297 length = -EINVAL;
2298 if (*ppos != 0)
2299 goto out;
2300
2301 length = -ENOMEM;
e12ba74d 2302 page = (char*)__get_free_page(GFP_TEMPORARY);
28a6d671
EB		 2303 if (!page)
2304 goto out;
2305
2306 length = -EFAULT;
2307 if (copy_from_user(page, buf, count))
2308 goto out_free;
2309
107db7c7
DH		 2310 /* Guard against adverse ptrace interaction */
2311 length = mutex_lock_interruptible(&task->cred_guard_mutex);
2312 if (length < 0)
2313 goto out_free;
2314
28a6d671 2315 length = security_setprocattr(task,
2fddfeef 2316 (char*)file->f_path.dentry->d_name.name,
28a6d671 2317 (void*)page, count);
107db7c7 2318 mutex_unlock(&task->cred_guard_mutex);
28a6d671
EB		 2319out_free:
2320 free_page((unsigned long) page);
2321out:
2322 put_task_struct(task);
2323out_no_task:
2324 return length;
2325}
2326
00977a59 2327static const struct file_operations proc_pid_attr_operations = {
28a6d671
EB		 2328 .read = proc_pid_attr_read,
2329 .write = proc_pid_attr_write,
87df8424 2330 .llseek = generic_file_llseek,
28a6d671
EB		 2331};
2332
c5141e6d 2333static const struct pid_entry attr_dir_stuff[] = {
631f9c18
AD		 2334 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2335 REG("prev", S_IRUGO, proc_pid_attr_operations),
2336 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2337 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2338 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2339 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
28a6d671
EB		 2340};
2341
72d9dcfc 2342static int proc_attr_dir_readdir(struct file * filp,
28a6d671
EB		 2343 void * dirent, filldir_t filldir)
2344{
2345 return proc_pident_readdir(filp,dirent,filldir,
72d9dcfc 2346 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2347}
2348
00977a59 2349static const struct file_operations proc_attr_dir_operations = {
1da177e4 2350 .read = generic_read_dir,
72d9dcfc 2351 .readdir = proc_attr_dir_readdir,
6038f373 2352 .llseek = default_llseek,
1da177e4
LT		 2353};
2354
72d9dcfc 2355static struct dentry *proc_attr_dir_lookup(struct inode *dir,
28a6d671
EB		 2356 struct dentry *dentry, struct nameidata *nd)
2357{
7bcd6b0e
EB		 2358 return proc_pident_lookup(dir, dentry,
2359 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB		 2360}
2361
c5ef1c42 2362static const struct inode_operations proc_attr_dir_inode_operations = {
72d9dcfc 2363 .lookup = proc_attr_dir_lookup,
99f89551 2364 .getattr = pid_getattr,
6d76fa58 2365 .setattr = proc_setattr,
1da177e4
LT		 2366};
2367
28a6d671
EB		 2368#endif
2369
698ba7b5 2370#ifdef CONFIG_ELF_CORE
3cb4a0bb
KH		 2371static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2372 size_t count, loff_t *ppos)
2373{
2374 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2375 struct mm_struct *mm;
2376 char buffer[PROC_NUMBUF];
2377 size_t len;
2378 int ret;
2379
2380 if (!task)
2381 return -ESRCH;
2382
2383 ret = 0;
2384 mm = get_task_mm(task);
2385 if (mm) {
2386 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2387 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2388 MMF_DUMP_FILTER_SHIFT));
2389 mmput(mm);
2390 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2391 }
2392
2393 put_task_struct(task);
2394
2395 return ret;
2396}
2397
2398static ssize_t proc_coredump_filter_write(struct file *file,
2399 const char __user *buf,
2400 size_t count,
2401 loff_t *ppos)
2402{
2403 struct task_struct *task;
2404 struct mm_struct *mm;
2405 char buffer[PROC_NUMBUF], *end;
2406 unsigned int val;
2407 int ret;
2408 int i;
2409 unsigned long mask;
2410
2411 ret = -EFAULT;
2412 memset(buffer, 0, sizeof(buffer));
2413 if (count > sizeof(buffer) - 1)
2414 count = sizeof(buffer) - 1;
2415 if (copy_from_user(buffer, buf, count))
2416 goto out_no_task;
2417
2418 ret = -EINVAL;
2419 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2420 if (*end == '\n')
2421 end++;
2422 if (end - buffer == 0)
2423 goto out_no_task;
2424
2425 ret = -ESRCH;
2426 task = get_proc_task(file->f_dentry->d_inode);
2427 if (!task)
2428 goto out_no_task;
2429
2430 ret = end - buffer;
2431 mm = get_task_mm(task);
2432 if (!mm)
2433 goto out_no_mm;
2434
2435 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2436 if (val & mask)
2437 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2438 else
2439 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2440 }
2441
2442 mmput(mm);
2443 out_no_mm:
2444 put_task_struct(task);
2445 out_no_task:
2446 return ret;
2447}
2448
2449static const struct file_operations proc_coredump_filter_operations = {
2450 .read = proc_coredump_filter_read,
2451 .write = proc_coredump_filter_write,
87df8424 2452 .llseek = generic_file_llseek,
3cb4a0bb
KH		 2453};
2454#endif
2455
28a6d671
EB		 2456/*
2457 * /proc/self:
2458 */
2459static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2460 int buflen)
2461{
488e5bc4 2462 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
b55fcb22 2463 pid_t tgid = task_tgid_nr_ns(current, ns);
28a6d671 2464 char tmp[PROC_NUMBUF];
b55fcb22 2465 if (!tgid)
488e5bc4 2466 return -ENOENT;
b55fcb22 2467 sprintf(tmp, "%d", tgid);
28a6d671
EB		 2468 return vfs_readlink(dentry,buffer,buflen,tmp);
2469}
2470
2471static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2472{
488e5bc4 2473 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
b55fcb22 2474 pid_t tgid = task_tgid_nr_ns(current, ns);
7fee4868
AV		 2475 char *name = ERR_PTR(-ENOENT);
2476 if (tgid) {
2477 name = __getname();
2478 if (!name)
2479 name = ERR_PTR(-ENOMEM);
2480 else
2481 sprintf(name, "%d", tgid);
2482 }
2483 nd_set_link(nd, name);
2484 return NULL;
2485}
2486
2487static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2488 void *cookie)
2489{
2490 char *s = nd_get_link(nd);
2491 if (!IS_ERR(s))
2492 __putname(s);
28a6d671
EB		 2493}
2494
c5ef1c42 2495static const struct inode_operations proc_self_inode_operations = {
28a6d671
EB		 2496 .readlink = proc_self_readlink,
2497 .follow_link = proc_self_follow_link,
7fee4868 2498 .put_link = proc_self_put_link,
28a6d671
EB		 2499};
2500
801199ce
EB		 2501/*
2502 * proc base
2503 *
2504 * These are the directory entries in the root directory of /proc
2505 * that properly belong to the /proc filesystem, as they describe
2506 * describe something that is process related.
2507 */
c5141e6d 2508static const struct pid_entry proc_base_stuff[] = {
61a28784 2509 NOD("self", S_IFLNK|S_IRWXUGO,
801199ce 2510 &proc_self_inode_operations, NULL, {}),
801199ce
EB		 2511};
2512
2513/*
2514 * Exceptional case: normally we are not allowed to unhash a busy
2515 * directory. In this case, however, we can do it - no aliasing problems
2516 * due to the way we treat inodes.
2517 */
2518static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
2519{
2520 struct inode *inode = dentry->d_inode;
2521 struct task_struct *task = get_proc_task(inode);
2522 if (task) {
2523 put_task_struct(task);
2524 return 1;
2525 }
2526 d_drop(dentry);
2527 return 0;
2528}
2529
d72f71eb 2530static const struct dentry_operations proc_base_dentry_operations =
801199ce
EB		 2531{
2532 .d_revalidate = proc_base_revalidate,
2533 .d_delete = pid_delete_dentry,
2534};
2535
444ceed8 2536static struct dentry *proc_base_instantiate(struct inode *dir,
c5141e6d 2537 struct dentry *dentry, struct task_struct *task, const void *ptr)
801199ce 2538{
c5141e6d 2539 const struct pid_entry *p = ptr;
801199ce 2540 struct inode *inode;
801199ce 2541 struct proc_inode *ei;
73d36460 2542 struct dentry *error;
801199ce
EB		 2543
2544 /* Allocate the inode */
2545 error = ERR_PTR(-ENOMEM);
2546 inode = new_inode(dir->i_sb);
2547 if (!inode)
2548 goto out;
2549
2550 /* Initialize the inode */
2551 ei = PROC_I(inode);
2552 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
801199ce
EB		 2553
2554 /*
2555 * grab the reference to the task.
2556 */
1a657f78 2557 ei->pid = get_task_pid(task, PIDTYPE_PID);
801199ce
EB		 2558 if (!ei->pid)
2559 goto out_iput;
2560
801199ce
EB		 2561 inode->i_mode = p->mode;
2562 if (S_ISDIR(inode->i_mode))
2563 inode->i_nlink = 2;
2564 if (S_ISLNK(inode->i_mode))
2565 inode->i_size = 64;
2566 if (p->iop)
2567 inode->i_op = p->iop;
2568 if (p->fop)
2569 inode->i_fop = p->fop;
2570 ei->op = p->op;
2571 dentry->d_op = &proc_base_dentry_operations;
2572 d_add(dentry, inode);
2573 error = NULL;
2574out:
801199ce
EB		 2575 return error;
2576out_iput:
2577 iput(inode);
2578 goto out;
2579}
2580
444ceed8
EB		 2581static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2582{
2583 struct dentry *error;
2584 struct task_struct *task = get_proc_task(dir);
c5141e6d 2585 const struct pid_entry *p, *last;
444ceed8
EB		 2586
2587 error = ERR_PTR(-ENOENT);
2588
2589 if (!task)
2590 goto out_no_task;
2591
2592 /* Lookup the directory entry */
7bcd6b0e
EB		 2593 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2594 for (p = proc_base_stuff; p <= last; p++) {
444ceed8
EB		 2595 if (p->len != dentry->d_name.len)
2596 continue;
2597 if (!memcmp(dentry->d_name.name, p->name, p->len))
2598 break;
2599 }
7bcd6b0e 2600 if (p > last)
444ceed8
EB		 2601 goto out;
2602
2603 error = proc_base_instantiate(dir, dentry, task, p);
2604
2605out:
2606 put_task_struct(task);
2607out_no_task:
2608 return error;
2609}
2610
c5141e6d
ED		 2611static int proc_base_fill_cache(struct file *filp, void *dirent,
2612 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
61a28784
EB		 2613{
2614 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2615 proc_base_instantiate, task, p);
2616}
2617
aba76fdb 2618#ifdef CONFIG_TASK_IO_ACCOUNTING
297c5d92
AR		 2619static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2620{
940389b8 2621 struct task_io_accounting acct = task->ioac;
5995477a
AR		 2622 unsigned long flags;
2623
2624 if (whole && lock_task_sighand(task, &flags)) {
2625 struct task_struct *t = task;
2626
2627 task_io_accounting_add(&acct, &task->signal->ioac);
2628 while_each_thread(task, t)
2629 task_io_accounting_add(&acct, &t->ioac);
2630
2631 unlock_task_sighand(task, &flags);
297c5d92 2632 }
aba76fdb
AM		 2633 return sprintf(buffer,
2634 "rchar: %llu\n"
2635 "wchar: %llu\n"
2636 "syscr: %llu\n"
2637 "syscw: %llu\n"
2638 "read_bytes: %llu\n"
2639 "write_bytes: %llu\n"
2640 "cancelled_write_bytes: %llu\n",
7c44319d
AB		 2641 (unsigned long long)acct.rchar,
2642 (unsigned long long)acct.wchar,
2643 (unsigned long long)acct.syscr,
2644 (unsigned long long)acct.syscw,
2645 (unsigned long long)acct.read_bytes,
2646 (unsigned long long)acct.write_bytes,
2647 (unsigned long long)acct.cancelled_write_bytes);
297c5d92
AR		 2648}
2649
2650static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2651{
2652 return do_io_accounting(task, buffer, 0);
aba76fdb 2653}
297c5d92
AR		 2654
2655static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2656{
2657 return do_io_accounting(task, buffer, 1);
2658}
2659#endif /* CONFIG_TASK_IO_ACCOUNTING */
aba76fdb 2660
47830723
KC		 2661static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2662 struct pid *pid, struct task_struct *task)
2663{
2664 seq_printf(m, "%08x\n", task->personality);
2665 return 0;
2666}
2667
28a6d671
EB		 2668/*
2669 * Thread groups
2670 */
00977a59 2671static const struct file_operations proc_task_operations;
c5ef1c42 2672static const struct inode_operations proc_task_inode_operations;
20cdc894 2673
c5141e6d 2674static const struct pid_entry tgid_base_stuff[] = {
631f9c18
AD		 2675 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2676 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2677 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
b2211a36 2678#ifdef CONFIG_NET
631f9c18 2679 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
b2211a36 2680#endif
631f9c18
AD		 2681 REG("environ", S_IRUSR, proc_environ_operations),
2682 INF("auxv", S_IRUSR, proc_pid_auxv),
2683 ONE("status", S_IRUGO, proc_pid_status),
2684 ONE("personality", S_IRUSR, proc_pid_personality),
3036e7b4 2685 INF("limits", S_IRUGO, proc_pid_limits),
43ae34cb 2686#ifdef CONFIG_SCHED_DEBUG
631f9c18 2687 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
ebcb6734 2688#endif
4614a696 2689 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 2690#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
631f9c18 2691 INF("syscall", S_IRUSR, proc_pid_syscall),
43ae34cb 2692#endif
631f9c18
AD		 2693 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2694 ONE("stat", S_IRUGO, proc_tgid_stat),
2695 ONE("statm", S_IRUGO, proc_pid_statm),
2696 REG("maps", S_IRUGO, proc_maps_operations),
28a6d671 2697#ifdef CONFIG_NUMA
631f9c18 2698 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
28a6d671 2699#endif
631f9c18
AD		 2700 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2701 LNK("cwd", proc_cwd_link),
2702 LNK("root", proc_root_link),
2703 LNK("exe", proc_exe_link),
2704 REG("mounts", S_IRUGO, proc_mounts_operations),
2705 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2706 REG("mountstats", S_IRUSR, proc_mountstats_operations),
1e883281 2707#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18
AD		 2708 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2709 REG("smaps", S_IRUGO, proc_smaps_operations),
2710 REG("pagemap", S_IRUSR, proc_pagemap_operations),
28a6d671
EB		 2711#endif
2712#ifdef CONFIG_SECURITY
631f9c18 2713 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 2714#endif
2715#ifdef CONFIG_KALLSYMS
631f9c18 2716 INF("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 2717#endif
2ec220e2
KC		 2718#ifdef CONFIG_STACKTRACE
2719 ONE("stack", S_IRUSR, proc_pid_stack),
28a6d671
EB		 2720#endif
2721#ifdef CONFIG_SCHEDSTATS
631f9c18 2722 INF("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 2723#endif
9745512c 2724#ifdef CONFIG_LATENCYTOP
631f9c18 2725 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 2726#endif
8793d854 2727#ifdef CONFIG_PROC_PID_CPUSET
631f9c18 2728 REG("cpuset", S_IRUGO, proc_cpuset_operations),
a424316c
PM		 2729#endif
2730#ifdef CONFIG_CGROUPS
631f9c18 2731 REG("cgroup", S_IRUGO, proc_cgroup_operations),
28a6d671 2732#endif
631f9c18
AD		 2733 INF("oom_score", S_IRUGO, proc_oom_score),
2734 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
a63d83f4 2735 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 2736#ifdef CONFIG_AUDITSYSCALL
631f9c18
AD		 2737 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2738 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 2739#endif
f4f154fd 2740#ifdef CONFIG_FAULT_INJECTION
631f9c18 2741 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 2742#endif
698ba7b5 2743#ifdef CONFIG_ELF_CORE
631f9c18 2744 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3cb4a0bb 2745#endif
aba76fdb 2746#ifdef CONFIG_TASK_IO_ACCOUNTING
631f9c18 2747 INF("io", S_IRUGO, proc_tgid_io_accounting),
aba76fdb 2748#endif
28a6d671 2749};
1da177e4 2750
28a6d671 2751static int proc_tgid_base_readdir(struct file * filp,
1da177e4
LT		 2752 void * dirent, filldir_t filldir)
2753{
2754 return proc_pident_readdir(filp,dirent,filldir,
28a6d671 2755 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 2756}
2757
00977a59 2758static const struct file_operations proc_tgid_base_operations = {
1da177e4 2759 .read = generic_read_dir,
28a6d671 2760 .readdir = proc_tgid_base_readdir,
6038f373 2761 .llseek = default_llseek,
1da177e4
LT		 2762};
2763
28a6d671 2764static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
7bcd6b0e
EB		 2765 return proc_pident_lookup(dir, dentry,
2766 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT		 2767}
2768
c5ef1c42 2769static const struct inode_operations proc_tgid_base_inode_operations = {
28a6d671 2770 .lookup = proc_tgid_base_lookup,
99f89551 2771 .getattr = pid_getattr,
6d76fa58 2772 .setattr = proc_setattr,
1da177e4 2773};
1da177e4 2774
60347f67 2775static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
1da177e4 2776{
48e6484d 2777 struct dentry *dentry, *leader, *dir;
8578cea7 2778 char buf[PROC_NUMBUF];
48e6484d
EB		 2779 struct qstr name;
2780
2781 name.name = buf;
60347f67
PE		 2782 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2783 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d 2784 if (dentry) {
29f12ca3 2785 shrink_dcache_parent(dentry);
48e6484d
EB		 2786 d_drop(dentry);
2787 dput(dentry);
2788 }
1da177e4 2789
48e6484d 2790 name.name = buf;
60347f67
PE		 2791 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2792 leader = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d
EB		 2793 if (!leader)
2794 goto out;
1da177e4 2795
48e6484d
EB		 2796 name.name = "task";
2797 name.len = strlen(name.name);
2798 dir = d_hash_and_lookup(leader, &name);
2799 if (!dir)
2800 goto out_put_leader;
2801
2802 name.name = buf;
60347f67 2803 name.len = snprintf(buf, sizeof(buf), "%d", pid);
48e6484d
EB		 2804 dentry = d_hash_and_lookup(dir, &name);
2805 if (dentry) {
2806 shrink_dcache_parent(dentry);
2807 d_drop(dentry);
2808 dput(dentry);
1da177e4 2809 }
48e6484d
EB		 2810
2811 dput(dir);
2812out_put_leader:
2813 dput(leader);
2814out:
2815 return;
1da177e4
LT		 2816}
2817
0895e91d
RD		 2818/**
2819 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2820 * @task: task that should be flushed.
2821 *
2822 * When flushing dentries from proc, one needs to flush them from global
60347f67 2823 * proc (proc_mnt) and from all the namespaces' procs this task was seen
0895e91d
RD		 2824 * in. This call is supposed to do all of this job.
2825 *
2826 * Looks in the dcache for
2827 * /proc/@pid
2828 * /proc/@tgid/task/@pid
2829 * if either directory is present flushes it and all of it'ts children
2830 * from the dcache.
2831 *
2832 * It is safe and reasonable to cache /proc entries for a task until
2833 * that task exits. After that they just clog up the dcache with
2834 * useless entries, possibly causing useful dcache entries to be
2835 * flushed instead. This routine is proved to flush those useless
2836 * dcache entries at process exit time.
2837 *
2838 * NOTE: This routine is just an optimization so it does not guarantee
2839 * that no dcache entries will exist at process exit time it
2840 * just makes it very unlikely that any will persist.
60347f67
PE		 2841 */
2842
2843void proc_flush_task(struct task_struct *task)
2844{
9fcc2d15 2845 int i;
9b4d1cbe 2846 struct pid *pid, *tgid;
130f77ec
PE		 2847 struct upid *upid;
2848
130f77ec 2849 pid = task_pid(task);
9b4d1cbe 2850 tgid = task_tgid(task);
130f77ec 2851
9fcc2d15 2852 for (i = 0; i <= pid->level; i++) {
130f77ec
PE		 2853 upid = &pid->numbers[i];
2854 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
9b4d1cbe 2855 tgid->numbers[i].nr);
130f77ec 2856 }
6f4e6433
PE		 2857
2858 upid = &pid->numbers[pid->level];
2859 if (upid->nr == 1)
2860 pid_ns_release_proc(upid->ns);
60347f67
PE		 2861}
2862
9711ef99
AB		 2863static struct dentry *proc_pid_instantiate(struct inode *dir,
2864 struct dentry * dentry,
c5141e6d 2865 struct task_struct *task, const void *ptr)
444ceed8
EB		 2866{
2867 struct dentry *error = ERR_PTR(-ENOENT);
2868 struct inode *inode;
2869
61a28784 2870 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 2871 if (!inode)
2872 goto out;
2873
2874 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2875 inode->i_op = &proc_tgid_base_inode_operations;
2876 inode->i_fop = &proc_tgid_base_operations;
2877 inode->i_flags|=S_IMMUTABLE;
aed54175
VN		 2878
2879 inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
2880 ARRAY_SIZE(tgid_base_stuff));
444ceed8
EB		 2881
2882 dentry->d_op = &pid_dentry_operations;
2883
2884 d_add(dentry, inode);
2885 /* Close the race of the process dying before we return the dentry */
2886 if (pid_revalidate(dentry, NULL))
2887 error = NULL;
2888out:
2889 return error;
2890}
2891
1da177e4
LT		 2892struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2893{
73d36460 2894 struct dentry *result;
1da177e4 2895 struct task_struct *task;
1da177e4 2896 unsigned tgid;
b488893a 2897 struct pid_namespace *ns;
1da177e4 2898
801199ce
EB		 2899 result = proc_base_lookup(dir, dentry);
2900 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2901 goto out;
2902
1da177e4
LT		 2903 tgid = name_to_int(dentry);
2904 if (tgid == ~0U)
2905 goto out;
2906
b488893a 2907 ns = dentry->d_sb->s_fs_info;
de758734 2908 rcu_read_lock();
b488893a 2909 task = find_task_by_pid_ns(tgid, ns);
1da177e4
LT		 2910 if (task)
2911 get_task_struct(task);
de758734 2912 rcu_read_unlock();
1da177e4
LT		 2913 if (!task)
2914 goto out;
2915
444ceed8 2916 result = proc_pid_instantiate(dir, dentry, task, NULL);
1da177e4 2917 put_task_struct(task);
1da177e4 2918out:
cd6a3ce9 2919 return result;
1da177e4
LT		 2920}
2921
1da177e4 2922/*
0804ef4b 2923 * Find the first task with tgid >= tgid
0bc58a91 2924 *
1da177e4 2925 */
19fd4bb2
EB		 2926struct tgid_iter {
2927 unsigned int tgid;
0804ef4b 2928 struct task_struct *task;
19fd4bb2
EB		 2929};
2930static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2931{
0804ef4b 2932 struct pid *pid;
1da177e4 2933
19fd4bb2
EB		 2934 if (iter.task)
2935 put_task_struct(iter.task);
454cc105 2936 rcu_read_lock();
0804ef4b 2937retry:
19fd4bb2
EB		 2938 iter.task = NULL;
2939 pid = find_ge_pid(iter.tgid, ns);
0804ef4b 2940 if (pid) {
19fd4bb2
EB		 2941 iter.tgid = pid_nr_ns(pid, ns);
2942 iter.task = pid_task(pid, PIDTYPE_PID);
0804ef4b
EB		 2943 /* What we to know is if the pid we have find is the
2944 * pid of a thread_group_leader. Testing for task
2945 * being a thread_group_leader is the obvious thing
2946 * todo but there is a window when it fails, due to
2947 * the pid transfer logic in de_thread.
2948 *
2949 * So we perform the straight forward test of seeing
2950 * if the pid we have found is the pid of a thread
2951 * group leader, and don't worry if the task we have
2952 * found doesn't happen to be a thread group leader.
2953 * As we don't care in the case of readdir.
2954 */
19fd4bb2
EB		 2955 if (!iter.task || !has_group_leader_pid(iter.task)) {
2956 iter.tgid += 1;
0804ef4b 2957 goto retry;
19fd4bb2
EB		 2958 }
2959 get_task_struct(iter.task);
0bc58a91 2960 }
454cc105 2961 rcu_read_unlock();
19fd4bb2 2962 return iter;
1da177e4
LT		 2963}
2964
7bcd6b0e 2965#define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
0804ef4b 2966
61a28784 2967static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
19fd4bb2 2968 struct tgid_iter iter)
61a28784
EB		 2969{
2970 char name[PROC_NUMBUF];
19fd4bb2 2971 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
61a28784 2972 return proc_fill_cache(filp, dirent, filldir, name, len,
19fd4bb2 2973 proc_pid_instantiate, iter.task, NULL);
61a28784
EB		 2974}
2975
1da177e4
LT		 2976/* for the /proc/ directory itself, after non-process stuff has been done */
2977int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2978{
1da177e4 2979 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2fddfeef 2980 struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
19fd4bb2 2981 struct tgid_iter iter;
b488893a 2982 struct pid_namespace *ns;
1da177e4 2983
61a28784
EB		 2984 if (!reaper)
2985 goto out_no_task;
2986
7bcd6b0e 2987 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
c5141e6d 2988 const struct pid_entry *p = &proc_base_stuff[nr];
61a28784 2989 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
801199ce 2990 goto out;
1da177e4
LT		 2991 }
2992
b488893a 2993 ns = filp->f_dentry->d_sb->s_fs_info;
19fd4bb2
EB		 2994 iter.task = NULL;
2995 iter.tgid = filp->f_pos - TGID_OFFSET;
2996 for (iter = next_tgid(ns, iter);
2997 iter.task;
2998 iter.tgid += 1, iter = next_tgid(ns, iter)) {
2999 filp->f_pos = iter.tgid + TGID_OFFSET;
3000 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
3001 put_task_struct(iter.task);
0804ef4b 3002 goto out;
1da177e4 3003 }
0bc58a91 3004 }
0804ef4b
EB		 3005 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3006out:
61a28784
EB		 3007 put_task_struct(reaper);
3008out_no_task:
0bc58a91
EB		 3009 return 0;
3010}
1da177e4 3011
28a6d671
EB		 3012/*
3013 * Tasks
3014 */
c5141e6d 3015static const struct pid_entry tid_base_stuff[] = {
631f9c18 3016 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3835541d 3017 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
631f9c18
AD		 3018 REG("environ", S_IRUSR, proc_environ_operations),
3019 INF("auxv", S_IRUSR, proc_pid_auxv),
3020 ONE("status", S_IRUGO, proc_pid_status),
3021 ONE("personality", S_IRUSR, proc_pid_personality),
3036e7b4 3022 INF("limits", S_IRUGO, proc_pid_limits),
43ae34cb 3023#ifdef CONFIG_SCHED_DEBUG
631f9c18 3024 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
ebcb6734 3025#endif
4614a696 3026 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 3027#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
631f9c18 3028 INF("syscall", S_IRUSR, proc_pid_syscall),
43ae34cb 3029#endif
631f9c18
AD		 3030 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3031 ONE("stat", S_IRUGO, proc_tid_stat),
3032 ONE("statm", S_IRUGO, proc_pid_statm),
3033 REG("maps", S_IRUGO, proc_maps_operations),
28a6d671 3034#ifdef CONFIG_NUMA
631f9c18 3035 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
28a6d671 3036#endif
631f9c18
AD		 3037 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3038 LNK("cwd", proc_cwd_link),
3039 LNK("root", proc_root_link),
3040 LNK("exe", proc_exe_link),
3041 REG("mounts", S_IRUGO, proc_mounts_operations),
3042 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
1e883281 3043#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18
AD		 3044 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3045 REG("smaps", S_IRUGO, proc_smaps_operations),
3046 REG("pagemap", S_IRUSR, proc_pagemap_operations),
28a6d671
EB		 3047#endif
3048#ifdef CONFIG_SECURITY
631f9c18 3049 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB		 3050#endif
3051#ifdef CONFIG_KALLSYMS
631f9c18 3052 INF("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 3053#endif
2ec220e2
KC		 3054#ifdef CONFIG_STACKTRACE
3055 ONE("stack", S_IRUSR, proc_pid_stack),
28a6d671
EB		 3056#endif
3057#ifdef CONFIG_SCHEDSTATS
631f9c18 3058 INF("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 3059#endif
9745512c 3060#ifdef CONFIG_LATENCYTOP
631f9c18 3061 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 3062#endif
8793d854 3063#ifdef CONFIG_PROC_PID_CPUSET
631f9c18 3064 REG("cpuset", S_IRUGO, proc_cpuset_operations),
a424316c
PM		 3065#endif
3066#ifdef CONFIG_CGROUPS
631f9c18 3067 REG("cgroup", S_IRUGO, proc_cgroup_operations),
28a6d671 3068#endif
631f9c18
AD		 3069 INF("oom_score", S_IRUGO, proc_oom_score),
3070 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
a63d83f4 3071 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 3072#ifdef CONFIG_AUDITSYSCALL
631f9c18
AD		 3073 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3074 REG("sessionid", S_IRUSR, proc_sessionid_operations),
28a6d671 3075#endif
f4f154fd 3076#ifdef CONFIG_FAULT_INJECTION
631f9c18 3077 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 3078#endif
297c5d92 3079#ifdef CONFIG_TASK_IO_ACCOUNTING
631f9c18 3080 INF("io", S_IRUGO, proc_tid_io_accounting),
297c5d92 3081#endif
28a6d671
EB		 3082};
3083
3084static int proc_tid_base_readdir(struct file * filp,
3085 void * dirent, filldir_t filldir)
3086{
3087 return proc_pident_readdir(filp,dirent,filldir,
3088 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3089}
3090
3091static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
7bcd6b0e
EB		 3092 return proc_pident_lookup(dir, dentry,
3093 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
28a6d671
EB		 3094}
3095
00977a59 3096static const struct file_operations proc_tid_base_operations = {
28a6d671
EB		 3097 .read = generic_read_dir,
3098 .readdir = proc_tid_base_readdir,
6038f373 3099 .llseek = default_llseek,
28a6d671
EB		 3100};
3101
c5ef1c42 3102static const struct inode_operations proc_tid_base_inode_operations = {
28a6d671
EB		 3103 .lookup = proc_tid_base_lookup,
3104 .getattr = pid_getattr,
3105 .setattr = proc_setattr,
3106};
3107
444ceed8 3108static struct dentry *proc_task_instantiate(struct inode *dir,
c5141e6d 3109 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8
EB		 3110{
3111 struct dentry *error = ERR_PTR(-ENOENT);
3112 struct inode *inode;
61a28784 3113 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB		 3114
3115 if (!inode)
3116 goto out;
3117 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3118 inode->i_op = &proc_tid_base_inode_operations;
3119 inode->i_fop = &proc_tid_base_operations;
3120 inode->i_flags|=S_IMMUTABLE;
aed54175
VN		 3121
3122 inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
3123 ARRAY_SIZE(tid_base_stuff));
444ceed8
EB		 3124
3125 dentry->d_op = &pid_dentry_operations;
3126
3127 d_add(dentry, inode);
3128 /* Close the race of the process dying before we return the dentry */
3129 if (pid_revalidate(dentry, NULL))
3130 error = NULL;
3131out:
3132 return error;
3133}
3134
28a6d671
EB		 3135static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3136{
3137 struct dentry *result = ERR_PTR(-ENOENT);
3138 struct task_struct *task;
3139 struct task_struct *leader = get_proc_task(dir);
28a6d671 3140 unsigned tid;
b488893a 3141 struct pid_namespace *ns;
28a6d671
EB		 3142
3143 if (!leader)
3144 goto out_no_task;
3145
3146 tid = name_to_int(dentry);
3147 if (tid == ~0U)
3148 goto out;
3149
b488893a 3150 ns = dentry->d_sb->s_fs_info;
28a6d671 3151 rcu_read_lock();
b488893a 3152 task = find_task_by_pid_ns(tid, ns);
28a6d671
EB		 3153 if (task)
3154 get_task_struct(task);
3155 rcu_read_unlock();
3156 if (!task)
3157 goto out;
bac0abd6 3158 if (!same_thread_group(leader, task))
28a6d671
EB		 3159 goto out_drop_task;
3160
444ceed8 3161 result = proc_task_instantiate(dir, dentry, task, NULL);
28a6d671
EB		 3162out_drop_task:
3163 put_task_struct(task);
3164out:
3165 put_task_struct(leader);
3166out_no_task:
3167 return result;
3168}
3169
0bc58a91
EB		 3170/*
3171 * Find the first tid of a thread group to return to user space.
3172 *
3173 * Usually this is just the thread group leader, but if the users
3174 * buffer was too small or there was a seek into the middle of the
3175 * directory we have more work todo.
3176 *
3177 * In the case of a short read we start with find_task_by_pid.
3178 *
3179 * In the case of a seek we start with the leader and walk nr
3180 * threads past it.
3181 */
cc288738 3182static struct task_struct *first_tid(struct task_struct *leader,
b488893a 3183 int tid, int nr, struct pid_namespace *ns)
0bc58a91 3184{
a872ff0c 3185 struct task_struct *pos;
1da177e4 3186
cc288738 3187 rcu_read_lock();
0bc58a91
EB		 3188 /* Attempt to start with the pid of a thread */
3189 if (tid && (nr > 0)) {
b488893a 3190 pos = find_task_by_pid_ns(tid, ns);
a872ff0c
ON		 3191 if (pos && (pos->group_leader == leader))
3192 goto found;
0bc58a91 3193 }
1da177e4 3194
0bc58a91 3195 /* If nr exceeds the number of threads there is nothing todo */
a872ff0c
ON		 3196 pos = NULL;
3197 if (nr && nr >= get_nr_threads(leader))
3198 goto out;
1da177e4 3199
a872ff0c
ON		 3200 /* If we haven't found our starting place yet start
3201 * with the leader and walk nr threads forward.
0bc58a91 3202 */
a872ff0c
ON		 3203 for (pos = leader; nr > 0; --nr) {
3204 pos = next_thread(pos);
3205 if (pos == leader) {
3206 pos = NULL;
3207 goto out;
3208 }
1da177e4 3209 }
a872ff0c
ON		 3210found:
3211 get_task_struct(pos);
3212out:
cc288738 3213 rcu_read_unlock();
0bc58a91
EB		 3214 return pos;
3215}
3216
3217/*
3218 * Find the next thread in the thread list.
3219 * Return NULL if there is an error or no next thread.
3220 *
3221 * The reference to the input task_struct is released.
3222 */
3223static struct task_struct *next_tid(struct task_struct *start)
3224{
c1df7fb8 3225 struct task_struct *pos = NULL;
cc288738 3226 rcu_read_lock();
c1df7fb8 3227 if (pid_alive(start)) {
0bc58a91 3228 pos = next_thread(start);
c1df7fb8
ON		 3229 if (thread_group_leader(pos))
3230 pos = NULL;
3231 else
3232 get_task_struct(pos);
3233 }
cc288738 3234 rcu_read_unlock();
0bc58a91
EB		 3235 put_task_struct(start);
3236 return pos;
1da177e4
LT		 3237}
3238
61a28784
EB		 3239static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3240 struct task_struct *task, int tid)
3241{
3242 char name[PROC_NUMBUF];
3243 int len = snprintf(name, sizeof(name), "%d", tid);
3244 return proc_fill_cache(filp, dirent, filldir, name, len,
3245 proc_task_instantiate, task, NULL);
3246}
3247
1da177e4
LT		 3248/* for the /proc/TGID/task/ directories */
3249static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3250{
2fddfeef 3251 struct dentry *dentry = filp->f_path.dentry;
1da177e4 3252 struct inode *inode = dentry->d_inode;
7d895244 3253 struct task_struct *leader = NULL;
0bc58a91 3254 struct task_struct *task;
1da177e4
LT		 3255 int retval = -ENOENT;
3256 ino_t ino;
0bc58a91 3257 int tid;
b488893a 3258 struct pid_namespace *ns;
1da177e4 3259
7d895244
GC		 3260 task = get_proc_task(inode);
3261 if (!task)
3262 goto out_no_task;
3263 rcu_read_lock();
3264 if (pid_alive(task)) {
3265 leader = task->group_leader;
3266 get_task_struct(leader);
3267 }
3268 rcu_read_unlock();
3269 put_task_struct(task);
99f89551
EB		 3270 if (!leader)
3271 goto out_no_task;
1da177e4
LT		 3272 retval = 0;
3273
ee568b25 3274 switch ((unsigned long)filp->f_pos) {
1da177e4
LT		 3275 case 0:
3276 ino = inode->i_ino;
ee6f779b 3277 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
1da177e4 3278 goto out;
ee6f779b 3279 filp->f_pos++;
1da177e4
LT		 3280 /* fall through */
3281 case 1:
3282 ino = parent_ino(dentry);
ee6f779b 3283 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
1da177e4 3284 goto out;
ee6f779b 3285 filp->f_pos++;
1da177e4
LT		 3286 /* fall through */
3287 }
3288
0bc58a91
EB		 3289 /* f_version caches the tgid value that the last readdir call couldn't
3290 * return. lseek aka telldir automagically resets f_version to 0.
3291 */
b488893a 3292 ns = filp->f_dentry->d_sb->s_fs_info;
2b47c361 3293 tid = (int)filp->f_version;
0bc58a91 3294 filp->f_version = 0;
ee6f779b 3295 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
0bc58a91 3296 task;
ee6f779b 3297 task = next_tid(task), filp->f_pos++) {
b488893a 3298 tid = task_pid_nr_ns(task, ns);
61a28784 3299 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
0bc58a91
EB		 3300 /* returning this tgid failed, save it as the first
3301 * pid for the next readir call */
2b47c361 3302 filp->f_version = (u64)tid;
0bc58a91 3303 put_task_struct(task);
1da177e4 3304 break;
0bc58a91 3305 }
1da177e4
LT		 3306 }
3307out:
99f89551
EB		 3308 put_task_struct(leader);
3309out_no_task:
1da177e4
LT		 3310 return retval;
3311}
6e66b52b
EB		 3312
3313static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3314{
3315 struct inode *inode = dentry->d_inode;
99f89551 3316 struct task_struct *p = get_proc_task(inode);
6e66b52b
EB		 3317 generic_fillattr(inode, stat);
3318
99f89551 3319 if (p) {
99f89551 3320 stat->nlink += get_nr_threads(p);
99f89551 3321 put_task_struct(p);
6e66b52b
EB		 3322 }
3323
3324 return 0;
3325}
28a6d671 3326
c5ef1c42 3327static const struct inode_operations proc_task_inode_operations = {
28a6d671
EB		 3328 .lookup = proc_task_lookup,
3329 .getattr = proc_task_getattr,
3330 .setattr = proc_setattr,
3331};
3332
00977a59 3333static const struct file_operations proc_task_operations = {
28a6d671
EB		 3334 .read = generic_read_dir,
3335 .readdir = proc_task_readdir,
6038f373 3336 .llseek = default_llseek,
28a6d671 3337};
ubuntu-bionic-kernel mirror