]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - fs/proc/base.c
proc: remove mm_for_maps()
[mirror_ubuntu-artful-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>
22d917d8 84#include <linux/user_namespace.h>
5ad4e53b 85#include <linux/fs_struct.h>
5a0e3ad6 86#include <linux/slab.h>
640708a2 87#include <linux/flex_array.h>
f133ecca
CM
88#ifdef CONFIG_HARDWALL
89#include <asm/hardwall.h>
90#endif
43d2b113 91#include <trace/events/oom.h>
1da177e4
LT
92#include "internal.h"
93
0f2fe20f
EB
94/* NOTE:
95 * Implementing inode permission operations in /proc is almost
96 * certainly an error. Permission checks need to happen during
97 * each system call not at open time. The reason is that most of
98 * what we wish to check for permissions in /proc varies at runtime.
99 *
100 * The classic example of a problem is opening file descriptors
101 * in /proc for a task before it execs a suid executable.
102 */
103
1da177e4 104struct pid_entry {
1da177e4 105 char *name;
c5141e6d 106 int len;
d161a13f 107 umode_t mode;
c5ef1c42 108 const struct inode_operations *iop;
00977a59 109 const struct file_operations *fop;
20cdc894 110 union proc_op op;
1da177e4
LT
111};
112
61a28784 113#define NOD(NAME, MODE, IOP, FOP, OP) { \
20cdc894 114 .name = (NAME), \
c5141e6d 115 .len = sizeof(NAME) - 1, \
20cdc894
EB
116 .mode = MODE, \
117 .iop = IOP, \
118 .fop = FOP, \
119 .op = OP, \
120}
121
631f9c18
AD
122#define DIR(NAME, MODE, iops, fops) \
123 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
124#define LNK(NAME, get_link) \
61a28784 125 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
20cdc894 126 &proc_pid_link_inode_operations, NULL, \
631f9c18
AD
127 { .proc_get_link = get_link } )
128#define REG(NAME, MODE, fops) \
129 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
130#define INF(NAME, MODE, read) \
61a28784 131 NOD(NAME, (S_IFREG|(MODE)), \
20cdc894 132 NULL, &proc_info_file_operations, \
631f9c18
AD
133 { .proc_read = read } )
134#define ONE(NAME, MODE, show) \
be614086
EB
135 NOD(NAME, (S_IFREG|(MODE)), \
136 NULL, &proc_single_file_operations, \
631f9c18 137 { .proc_show = show } )
1da177e4 138
640708a2
PE
139static int proc_fd_permission(struct inode *inode, int mask);
140
aed54175
VN
141/*
142 * Count the number of hardlinks for the pid_entry table, excluding the .
143 * and .. links.
144 */
145static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
146 unsigned int n)
147{
148 unsigned int i;
149 unsigned int count;
150
151 count = 0;
152 for (i = 0; i < n; ++i) {
153 if (S_ISDIR(entries[i].mode))
154 ++count;
155 }
156
157 return count;
158}
159
f7ad3c6b 160static int get_task_root(struct task_struct *task, struct path *root)
1da177e4 161{
7c2c7d99
HD
162 int result = -ENOENT;
163
0494f6ec 164 task_lock(task);
f7ad3c6b
MS
165 if (task->fs) {
166 get_fs_root(task->fs, root);
7c2c7d99
HD
167 result = 0;
168 }
0494f6ec 169 task_unlock(task);
7c2c7d99 170 return result;
0494f6ec
MS
171}
172
7773fbc5 173static int proc_cwd_link(struct dentry *dentry, struct path *path)
0494f6ec 174{
7773fbc5 175 struct task_struct *task = get_proc_task(dentry->d_inode);
0494f6ec 176 int result = -ENOENT;
99f89551
EB
177
178 if (task) {
f7ad3c6b
MS
179 task_lock(task);
180 if (task->fs) {
181 get_fs_pwd(task->fs, path);
182 result = 0;
183 }
184 task_unlock(task);
99f89551
EB
185 put_task_struct(task);
186 }
1da177e4
LT
187 return result;
188}
189
7773fbc5 190static int proc_root_link(struct dentry *dentry, struct path *path)
1da177e4 191{
7773fbc5 192 struct task_struct *task = get_proc_task(dentry->d_inode);
1da177e4 193 int result = -ENOENT;
99f89551
EB
194
195 if (task) {
f7ad3c6b 196 result = get_task_root(task, path);
99f89551
EB
197 put_task_struct(task);
198 }
1da177e4
LT
199 return result;
200}
201
1da177e4
LT
202static int proc_pid_cmdline(struct task_struct *task, char * buffer)
203{
204 int res = 0;
205 unsigned int len;
206 struct mm_struct *mm = get_task_mm(task);
207 if (!mm)
208 goto out;
209 if (!mm->arg_end)
210 goto out_mm; /* Shh! No looking before we're done */
211
212 len = mm->arg_end - mm->arg_start;
213
214 if (len > PAGE_SIZE)
215 len = PAGE_SIZE;
216
217 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
218
219 // If the nul at the end of args has been overwritten, then
220 // assume application is using setproctitle(3).
221 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
222 len = strnlen(buffer, res);
223 if (len < res) {
224 res = len;
225 } else {
226 len = mm->env_end - mm->env_start;
227 if (len > PAGE_SIZE - res)
228 len = PAGE_SIZE - res;
229 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
230 res = strnlen(buffer, res);
231 }
232 }
233out_mm:
234 mmput(mm);
235out:
236 return res;
237}
238
239static int proc_pid_auxv(struct task_struct *task, char *buffer)
240{
e7dcd999 241 struct mm_struct *mm = mm_access(task, PTRACE_MODE_READ);
2fadaef4
AV
242 int res = PTR_ERR(mm);
243 if (mm && !IS_ERR(mm)) {
1da177e4 244 unsigned int nwords = 0;
dfe6b7d9 245 do {
1da177e4 246 nwords += 2;
dfe6b7d9 247 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
1da177e4
LT
248 res = nwords * sizeof(mm->saved_auxv[0]);
249 if (res > PAGE_SIZE)
250 res = PAGE_SIZE;
251 memcpy(buffer, mm->saved_auxv, res);
252 mmput(mm);
253 }
254 return res;
255}
256
257
258#ifdef CONFIG_KALLSYMS
259/*
260 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
261 * Returns the resolved symbol. If that fails, simply return the address.
262 */
263static int proc_pid_wchan(struct task_struct *task, char *buffer)
264{
ffb45122 265 unsigned long wchan;
9281acea 266 char symname[KSYM_NAME_LEN];
1da177e4
LT
267
268 wchan = get_wchan(task);
269
9d65cb4a 270 if (lookup_symbol_name(wchan, symname) < 0)
f83ce3e6
JE
271 if (!ptrace_may_access(task, PTRACE_MODE_READ))
272 return 0;
273 else
274 return sprintf(buffer, "%lu", wchan);
9d65cb4a
AD
275 else
276 return sprintf(buffer, "%s", symname);
1da177e4
LT
277}
278#endif /* CONFIG_KALLSYMS */
279
a9712bc1
AV
280static int lock_trace(struct task_struct *task)
281{
282 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
283 if (err)
284 return err;
285 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
286 mutex_unlock(&task->signal->cred_guard_mutex);
287 return -EPERM;
288 }
289 return 0;
290}
291
292static void unlock_trace(struct task_struct *task)
293{
294 mutex_unlock(&task->signal->cred_guard_mutex);
295}
296
2ec220e2
KC
297#ifdef CONFIG_STACKTRACE
298
299#define MAX_STACK_TRACE_DEPTH 64
300
301static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
302 struct pid *pid, struct task_struct *task)
303{
304 struct stack_trace trace;
305 unsigned long *entries;
a9712bc1 306 int err;
2ec220e2
KC
307 int i;
308
309 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
310 if (!entries)
311 return -ENOMEM;
312
313 trace.nr_entries = 0;
314 trace.max_entries = MAX_STACK_TRACE_DEPTH;
315 trace.entries = entries;
316 trace.skip = 0;
2ec220e2 317
a9712bc1
AV
318 err = lock_trace(task);
319 if (!err) {
320 save_stack_trace_tsk(task, &trace);
321
322 for (i = 0; i < trace.nr_entries; i++) {
b81a618d 323 seq_printf(m, "[<%pK>] %pS\n",
a9712bc1
AV
324 (void *)entries[i], (void *)entries[i]);
325 }
326 unlock_trace(task);
2ec220e2
KC
327 }
328 kfree(entries);
329
a9712bc1 330 return err;
2ec220e2
KC
331}
332#endif
333
1da177e4
LT
334#ifdef CONFIG_SCHEDSTATS
335/*
336 * Provides /proc/PID/schedstat
337 */
338static int proc_pid_schedstat(struct task_struct *task, char *buffer)
339{
172ba844 340 return sprintf(buffer, "%llu %llu %lu\n",
826e08b0
IM
341 (unsigned long long)task->se.sum_exec_runtime,
342 (unsigned long long)task->sched_info.run_delay,
2d72376b 343 task->sched_info.pcount);
1da177e4
LT
344}
345#endif
346
9745512c
AV
347#ifdef CONFIG_LATENCYTOP
348static int lstats_show_proc(struct seq_file *m, void *v)
349{
350 int i;
13d77c37
HS
351 struct inode *inode = m->private;
352 struct task_struct *task = get_proc_task(inode);
9745512c 353
13d77c37
HS
354 if (!task)
355 return -ESRCH;
356 seq_puts(m, "Latency Top version : v0.1\n");
9745512c 357 for (i = 0; i < 32; i++) {
34e49d4f
JP
358 struct latency_record *lr = &task->latency_record[i];
359 if (lr->backtrace[0]) {
9745512c 360 int q;
34e49d4f
JP
361 seq_printf(m, "%i %li %li",
362 lr->count, lr->time, lr->max);
9745512c 363 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
34e49d4f
JP
364 unsigned long bt = lr->backtrace[q];
365 if (!bt)
9745512c 366 break;
34e49d4f 367 if (bt == ULONG_MAX)
9745512c 368 break;
34e49d4f 369 seq_printf(m, " %ps", (void *)bt);
9745512c 370 }
9d6de12f 371 seq_putc(m, '\n');
9745512c
AV
372 }
373
374 }
13d77c37 375 put_task_struct(task);
9745512c
AV
376 return 0;
377}
378
379static int lstats_open(struct inode *inode, struct file *file)
380{
13d77c37 381 return single_open(file, lstats_show_proc, inode);
d6643d12
HS
382}
383
9745512c
AV
384static ssize_t lstats_write(struct file *file, const char __user *buf,
385 size_t count, loff_t *offs)
386{
13d77c37 387 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
9745512c 388
13d77c37
HS
389 if (!task)
390 return -ESRCH;
9745512c 391 clear_all_latency_tracing(task);
13d77c37 392 put_task_struct(task);
9745512c
AV
393
394 return count;
395}
396
397static const struct file_operations proc_lstats_operations = {
398 .open = lstats_open,
399 .read = seq_read,
400 .write = lstats_write,
401 .llseek = seq_lseek,
13d77c37 402 .release = single_release,
9745512c
AV
403};
404
405#endif
406
1da177e4
LT
407static int proc_oom_score(struct task_struct *task, char *buffer)
408{
a7f638f9 409 unsigned long totalpages = totalram_pages + total_swap_pages;
b95c35e7 410 unsigned long points = 0;
1da177e4 411
19c5d45a 412 read_lock(&tasklist_lock);
b95c35e7 413 if (pid_alive(task))
a7f638f9
DR
414 points = oom_badness(task, NULL, NULL, totalpages) *
415 1000 / totalpages;
19c5d45a 416 read_unlock(&tasklist_lock);
1da177e4
LT
417 return sprintf(buffer, "%lu\n", points);
418}
419
d85f50d5
NH
420struct limit_names {
421 char *name;
422 char *unit;
423};
424
425static const struct limit_names lnames[RLIM_NLIMITS] = {
cff4edb5 426 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
d85f50d5
NH
427 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
428 [RLIMIT_DATA] = {"Max data size", "bytes"},
429 [RLIMIT_STACK] = {"Max stack size", "bytes"},
430 [RLIMIT_CORE] = {"Max core file size", "bytes"},
431 [RLIMIT_RSS] = {"Max resident set", "bytes"},
432 [RLIMIT_NPROC] = {"Max processes", "processes"},
433 [RLIMIT_NOFILE] = {"Max open files", "files"},
434 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
435 [RLIMIT_AS] = {"Max address space", "bytes"},
436 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
437 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
438 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
439 [RLIMIT_NICE] = {"Max nice priority", NULL},
440 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
8808117c 441 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
d85f50d5
NH
442};
443
444/* Display limits for a process */
445static int proc_pid_limits(struct task_struct *task, char *buffer)
446{
447 unsigned int i;
448 int count = 0;
449 unsigned long flags;
450 char *bufptr = buffer;
451
452 struct rlimit rlim[RLIM_NLIMITS];
453
a6bebbc8 454 if (!lock_task_sighand(task, &flags))
d85f50d5 455 return 0;
d85f50d5
NH
456 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
457 unlock_task_sighand(task, &flags);
d85f50d5
NH
458
459 /*
460 * print the file header
461 */
462 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
463 "Limit", "Soft Limit", "Hard Limit", "Units");
464
465 for (i = 0; i < RLIM_NLIMITS; i++) {
466 if (rlim[i].rlim_cur == RLIM_INFINITY)
467 count += sprintf(&bufptr[count], "%-25s %-20s ",
468 lnames[i].name, "unlimited");
469 else
470 count += sprintf(&bufptr[count], "%-25s %-20lu ",
471 lnames[i].name, rlim[i].rlim_cur);
472
473 if (rlim[i].rlim_max == RLIM_INFINITY)
474 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
475 else
476 count += sprintf(&bufptr[count], "%-20lu ",
477 rlim[i].rlim_max);
478
479 if (lnames[i].unit)
480 count += sprintf(&bufptr[count], "%-10s\n",
481 lnames[i].unit);
482 else
483 count += sprintf(&bufptr[count], "\n");
484 }
485
486 return count;
487}
488
ebcb6734
RM
489#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
490static int proc_pid_syscall(struct task_struct *task, char *buffer)
491{
492 long nr;
493 unsigned long args[6], sp, pc;
a9712bc1
AV
494 int res = lock_trace(task);
495 if (res)
496 return res;
ebcb6734
RM
497
498 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
a9712bc1
AV
499 res = sprintf(buffer, "running\n");
500 else if (nr < 0)
501 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
502 else
503 res = sprintf(buffer,
ebcb6734
RM
504 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
505 nr,
506 args[0], args[1], args[2], args[3], args[4], args[5],
507 sp, pc);
a9712bc1
AV
508 unlock_trace(task);
509 return res;
ebcb6734
RM
510}
511#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
512
1da177e4
LT
513/************************************************************************/
514/* Here the fs part begins */
515/************************************************************************/
516
517/* permission checks */
778c1144 518static int proc_fd_access_allowed(struct inode *inode)
1da177e4 519{
778c1144
EB
520 struct task_struct *task;
521 int allowed = 0;
df26c40e
EB
522 /* Allow access to a task's file descriptors if it is us or we
523 * may use ptrace attach to the process and find out that
524 * information.
778c1144
EB
525 */
526 task = get_proc_task(inode);
df26c40e 527 if (task) {
006ebb40 528 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
778c1144 529 put_task_struct(task);
df26c40e 530 }
778c1144 531 return allowed;
1da177e4
LT
532}
533
6b4e306a 534int proc_setattr(struct dentry *dentry, struct iattr *attr)
6d76fa58
LT
535{
536 int error;
537 struct inode *inode = dentry->d_inode;
538
539 if (attr->ia_valid & ATTR_MODE)
540 return -EPERM;
541
542 error = inode_change_ok(inode, attr);
1025774c
CH
543 if (error)
544 return error;
545
546 if ((attr->ia_valid & ATTR_SIZE) &&
547 attr->ia_size != i_size_read(inode)) {
548 error = vmtruncate(inode, attr->ia_size);
549 if (error)
550 return error;
551 }
552
553 setattr_copy(inode, attr);
554 mark_inode_dirty(inode);
555 return 0;
6d76fa58
LT
556}
557
0499680a
VK
558/*
559 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
560 * or euid/egid (for hide_pid_min=2)?
561 */
562static bool has_pid_permissions(struct pid_namespace *pid,
563 struct task_struct *task,
564 int hide_pid_min)
565{
566 if (pid->hide_pid < hide_pid_min)
567 return true;
568 if (in_group_p(pid->pid_gid))
569 return true;
570 return ptrace_may_access(task, PTRACE_MODE_READ);
571}
572
573
574static int proc_pid_permission(struct inode *inode, int mask)
575{
576 struct pid_namespace *pid = inode->i_sb->s_fs_info;
577 struct task_struct *task;
578 bool has_perms;
579
580 task = get_proc_task(inode);
a2ef990a
XF
581 if (!task)
582 return -ESRCH;
0499680a
VK
583 has_perms = has_pid_permissions(pid, task, 1);
584 put_task_struct(task);
585
586 if (!has_perms) {
587 if (pid->hide_pid == 2) {
588 /*
589 * Let's make getdents(), stat(), and open()
590 * consistent with each other. If a process
591 * may not stat() a file, it shouldn't be seen
592 * in procfs at all.
593 */
594 return -ENOENT;
595 }
596
597 return -EPERM;
598 }
599 return generic_permission(inode, mask);
600}
601
602
603
c5ef1c42 604static const struct inode_operations proc_def_inode_operations = {
6d76fa58
LT
605 .setattr = proc_setattr,
606};
607
1da177e4
LT
608#define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
609
610static ssize_t proc_info_read(struct file * file, char __user * buf,
611 size_t count, loff_t *ppos)
612{
2fddfeef 613 struct inode * inode = file->f_path.dentry->d_inode;
1da177e4
LT
614 unsigned long page;
615 ssize_t length;
99f89551
EB
616 struct task_struct *task = get_proc_task(inode);
617
618 length = -ESRCH;
619 if (!task)
620 goto out_no_task;
1da177e4
LT
621
622 if (count > PROC_BLOCK_SIZE)
623 count = PROC_BLOCK_SIZE;
99f89551
EB
624
625 length = -ENOMEM;
e12ba74d 626 if (!(page = __get_free_page(GFP_TEMPORARY)))
99f89551 627 goto out;
1da177e4
LT
628
629 length = PROC_I(inode)->op.proc_read(task, (char*)page);
630
631 if (length >= 0)
632 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
633 free_page(page);
99f89551
EB
634out:
635 put_task_struct(task);
636out_no_task:
1da177e4
LT
637 return length;
638}
639
00977a59 640static const struct file_operations proc_info_file_operations = {
1da177e4 641 .read = proc_info_read,
87df8424 642 .llseek = generic_file_llseek,
1da177e4
LT
643};
644
be614086
EB
645static int proc_single_show(struct seq_file *m, void *v)
646{
647 struct inode *inode = m->private;
648 struct pid_namespace *ns;
649 struct pid *pid;
650 struct task_struct *task;
651 int ret;
652
653 ns = inode->i_sb->s_fs_info;
654 pid = proc_pid(inode);
655 task = get_pid_task(pid, PIDTYPE_PID);
656 if (!task)
657 return -ESRCH;
658
659 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
660
661 put_task_struct(task);
662 return ret;
663}
664
665static int proc_single_open(struct inode *inode, struct file *filp)
666{
c6a34058 667 return single_open(filp, proc_single_show, inode);
be614086
EB
668}
669
670static const struct file_operations proc_single_file_operations = {
671 .open = proc_single_open,
672 .read = seq_read,
673 .llseek = seq_lseek,
674 .release = single_release,
675};
676
b409e578 677static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
1da177e4 678{
e268337d
LT
679 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
680 struct mm_struct *mm;
681
682 if (!task)
683 return -ESRCH;
684
b409e578 685 mm = mm_access(task, mode);
e268337d
LT
686 put_task_struct(task);
687
688 if (IS_ERR(mm))
689 return PTR_ERR(mm);
690
6d08f2c7
ON
691 if (mm) {
692 /* ensure this mm_struct can't be freed */
693 atomic_inc(&mm->mm_count);
694 /* but do not pin its memory */
695 mmput(mm);
696 }
697
4a3956c7
KH
698 /* OK to pass negative loff_t, we can catch out-of-range */
699 file->f_mode |= FMODE_UNSIGNED_OFFSET;
e268337d
LT
700 file->private_data = mm;
701
1da177e4
LT
702 return 0;
703}
704
b409e578
CW
705static int mem_open(struct inode *inode, struct file *file)
706{
707 return __mem_open(inode, file, PTRACE_MODE_ATTACH);
708}
709
572d34b9
ON
710static ssize_t mem_rw(struct file *file, char __user *buf,
711 size_t count, loff_t *ppos, int write)
1da177e4 712{
e268337d 713 struct mm_struct *mm = file->private_data;
572d34b9
ON
714 unsigned long addr = *ppos;
715 ssize_t copied;
1da177e4 716 char *page;
1da177e4 717
e268337d
LT
718 if (!mm)
719 return 0;
99f89551 720
30cd8903
KM
721 page = (char *)__get_free_page(GFP_TEMPORARY);
722 if (!page)
e268337d 723 return -ENOMEM;
1da177e4 724
f7ca54f4 725 copied = 0;
6d08f2c7
ON
726 if (!atomic_inc_not_zero(&mm->mm_users))
727 goto free;
728
1da177e4 729 while (count > 0) {
572d34b9 730 int this_len = min_t(int, count, PAGE_SIZE);
1da177e4 731
572d34b9 732 if (write && copy_from_user(page, buf, this_len)) {
1da177e4
LT
733 copied = -EFAULT;
734 break;
735 }
572d34b9
ON
736
737 this_len = access_remote_vm(mm, addr, page, this_len, write);
738 if (!this_len) {
1da177e4
LT
739 if (!copied)
740 copied = -EIO;
741 break;
742 }
572d34b9
ON
743
744 if (!write && copy_to_user(buf, page, this_len)) {
745 copied = -EFAULT;
746 break;
747 }
748
749 buf += this_len;
750 addr += this_len;
751 copied += this_len;
752 count -= this_len;
1da177e4 753 }
572d34b9 754 *ppos = addr;
30cd8903 755
6d08f2c7
ON
756 mmput(mm);
757free:
30cd8903 758 free_page((unsigned long) page);
1da177e4
LT
759 return copied;
760}
1da177e4 761
572d34b9
ON
762static ssize_t mem_read(struct file *file, char __user *buf,
763 size_t count, loff_t *ppos)
764{
765 return mem_rw(file, buf, count, ppos, 0);
766}
767
768static ssize_t mem_write(struct file *file, const char __user *buf,
769 size_t count, loff_t *ppos)
770{
771 return mem_rw(file, (char __user*)buf, count, ppos, 1);
772}
773
85863e47 774loff_t mem_lseek(struct file *file, loff_t offset, int orig)
1da177e4
LT
775{
776 switch (orig) {
777 case 0:
778 file->f_pos = offset;
779 break;
780 case 1:
781 file->f_pos += offset;
782 break;
783 default:
784 return -EINVAL;
785 }
786 force_successful_syscall_return();
787 return file->f_pos;
788}
789
e268337d
LT
790static int mem_release(struct inode *inode, struct file *file)
791{
792 struct mm_struct *mm = file->private_data;
71879d3c 793 if (mm)
6d08f2c7 794 mmdrop(mm);
e268337d
LT
795 return 0;
796}
797
00977a59 798static const struct file_operations proc_mem_operations = {
1da177e4
LT
799 .llseek = mem_lseek,
800 .read = mem_read,
801 .write = mem_write,
802 .open = mem_open,
e268337d 803 .release = mem_release,
1da177e4
LT
804};
805
b409e578
CW
806static int environ_open(struct inode *inode, struct file *file)
807{
808 return __mem_open(inode, file, PTRACE_MODE_READ);
809}
810
315e28c8
JP
811static ssize_t environ_read(struct file *file, char __user *buf,
812 size_t count, loff_t *ppos)
813{
315e28c8
JP
814 char *page;
815 unsigned long src = *ppos;
b409e578
CW
816 int ret = 0;
817 struct mm_struct *mm = file->private_data;
315e28c8 818
b409e578
CW
819 if (!mm)
820 return 0;
315e28c8 821
315e28c8
JP
822 page = (char *)__get_free_page(GFP_TEMPORARY);
823 if (!page)
b409e578 824 return -ENOMEM;
315e28c8 825
d6f64b89 826 ret = 0;
b409e578
CW
827 if (!atomic_inc_not_zero(&mm->mm_users))
828 goto free;
315e28c8
JP
829 while (count > 0) {
830 int this_len, retval, max_len;
831
832 this_len = mm->env_end - (mm->env_start + src);
833
834 if (this_len <= 0)
835 break;
836
837 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
838 this_len = (this_len > max_len) ? max_len : this_len;
839
b409e578 840 retval = access_remote_vm(mm, (mm->env_start + src),
315e28c8
JP
841 page, this_len, 0);
842
843 if (retval <= 0) {
844 ret = retval;
845 break;
846 }
847
848 if (copy_to_user(buf, page, retval)) {
849 ret = -EFAULT;
850 break;
851 }
852
853 ret += retval;
854 src += retval;
855 buf += retval;
856 count -= retval;
857 }
858 *ppos = src;
315e28c8 859 mmput(mm);
b409e578
CW
860
861free:
315e28c8 862 free_page((unsigned long) page);
315e28c8
JP
863 return ret;
864}
865
866static const struct file_operations proc_environ_operations = {
b409e578 867 .open = environ_open,
315e28c8 868 .read = environ_read,
87df8424 869 .llseek = generic_file_llseek,
b409e578 870 .release = mem_release,
315e28c8
JP
871};
872
1da177e4
LT
873static ssize_t oom_adjust_read(struct file *file, char __user *buf,
874 size_t count, loff_t *ppos)
875{
2fddfeef 876 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
8578cea7 877 char buffer[PROC_NUMBUF];
1da177e4 878 size_t len;
28b83c51
KM
879 int oom_adjust = OOM_DISABLE;
880 unsigned long flags;
1da177e4 881
99f89551
EB
882 if (!task)
883 return -ESRCH;
28b83c51
KM
884
885 if (lock_task_sighand(task, &flags)) {
886 oom_adjust = task->signal->oom_adj;
887 unlock_task_sighand(task, &flags);
888 }
889
99f89551
EB
890 put_task_struct(task);
891
8578cea7 892 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
0c28f287
AM
893
894 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1da177e4
LT
895}
896
897static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
898 size_t count, loff_t *ppos)
899{
99f89551 900 struct task_struct *task;
5d863b89 901 char buffer[PROC_NUMBUF];
0a8cb8e3 902 int oom_adjust;
28b83c51 903 unsigned long flags;
5d863b89 904 int err;
1da177e4 905
8578cea7
EB
906 memset(buffer, 0, sizeof(buffer));
907 if (count > sizeof(buffer) - 1)
908 count = sizeof(buffer) - 1;
723548bf
DR
909 if (copy_from_user(buffer, buf, count)) {
910 err = -EFAULT;
911 goto out;
912 }
5d863b89 913
0a8cb8e3 914 err = kstrtoint(strstrip(buffer), 0, &oom_adjust);
5d863b89 915 if (err)
723548bf 916 goto out;
8ac773b4 917 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
723548bf
DR
918 oom_adjust != OOM_DISABLE) {
919 err = -EINVAL;
920 goto out;
921 }
5d863b89 922
2fddfeef 923 task = get_proc_task(file->f_path.dentry->d_inode);
723548bf
DR
924 if (!task) {
925 err = -ESRCH;
926 goto out;
927 }
d19d5476
DR
928
929 task_lock(task);
930 if (!task->mm) {
931 err = -EINVAL;
932 goto err_task_lock;
933 }
934
28b83c51 935 if (!lock_task_sighand(task, &flags)) {
723548bf 936 err = -ESRCH;
d19d5476 937 goto err_task_lock;
28b83c51
KM
938 }
939
940 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
723548bf
DR
941 err = -EACCES;
942 goto err_sighand;
8fb4fc68 943 }
28b83c51 944
51b1bd2a
DR
945 /*
946 * Warn that /proc/pid/oom_adj is deprecated, see
947 * Documentation/feature-removal-schedule.txt.
948 */
c2142704 949 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
be8f684d
DR
950 current->comm, task_pid_nr(current), task_pid_nr(task),
951 task_pid_nr(task));
28b83c51 952 task->signal->oom_adj = oom_adjust;
a63d83f4
DR
953 /*
954 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
955 * value is always attainable.
956 */
957 if (task->signal->oom_adj == OOM_ADJUST_MAX)
958 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
959 else
960 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
961 -OOM_DISABLE;
43d2b113 962 trace_oom_score_adj_update(task);
723548bf 963err_sighand:
28b83c51 964 unlock_task_sighand(task, &flags);
d19d5476
DR
965err_task_lock:
966 task_unlock(task);
99f89551 967 put_task_struct(task);
723548bf
DR
968out:
969 return err < 0 ? err : count;
1da177e4
LT
970}
971
00977a59 972static const struct file_operations proc_oom_adjust_operations = {
1da177e4
LT
973 .read = oom_adjust_read,
974 .write = oom_adjust_write,
87df8424 975 .llseek = generic_file_llseek,
1da177e4
LT
976};
977
a63d83f4
DR
978static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
979 size_t count, loff_t *ppos)
980{
981 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
982 char buffer[PROC_NUMBUF];
983 int oom_score_adj = OOM_SCORE_ADJ_MIN;
984 unsigned long flags;
985 size_t len;
986
987 if (!task)
988 return -ESRCH;
989 if (lock_task_sighand(task, &flags)) {
990 oom_score_adj = task->signal->oom_score_adj;
991 unlock_task_sighand(task, &flags);
992 }
993 put_task_struct(task);
994 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
995 return simple_read_from_buffer(buf, count, ppos, buffer, len);
996}
997
998static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
999 size_t count, loff_t *ppos)
1000{
1001 struct task_struct *task;
1002 char buffer[PROC_NUMBUF];
1003 unsigned long flags;
0a8cb8e3 1004 int oom_score_adj;
a63d83f4
DR
1005 int err;
1006
1007 memset(buffer, 0, sizeof(buffer));
1008 if (count > sizeof(buffer) - 1)
1009 count = sizeof(buffer) - 1;
723548bf
DR
1010 if (copy_from_user(buffer, buf, count)) {
1011 err = -EFAULT;
1012 goto out;
1013 }
a63d83f4 1014
0a8cb8e3 1015 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
a63d83f4 1016 if (err)
723548bf 1017 goto out;
a63d83f4 1018 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
723548bf
DR
1019 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1020 err = -EINVAL;
1021 goto out;
1022 }
a63d83f4
DR
1023
1024 task = get_proc_task(file->f_path.dentry->d_inode);
723548bf
DR
1025 if (!task) {
1026 err = -ESRCH;
1027 goto out;
1028 }
d19d5476
DR
1029
1030 task_lock(task);
1031 if (!task->mm) {
1032 err = -EINVAL;
1033 goto err_task_lock;
1034 }
1035
a63d83f4 1036 if (!lock_task_sighand(task, &flags)) {
723548bf 1037 err = -ESRCH;
d19d5476 1038 goto err_task_lock;
a63d83f4 1039 }
d19d5476 1040
dabb16f6 1041 if (oom_score_adj < task->signal->oom_score_adj_min &&
a63d83f4 1042 !capable(CAP_SYS_RESOURCE)) {
723548bf
DR
1043 err = -EACCES;
1044 goto err_sighand;
a63d83f4
DR
1045 }
1046
1047 task->signal->oom_score_adj = oom_score_adj;
dabb16f6
MSB
1048 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1049 task->signal->oom_score_adj_min = oom_score_adj;
43d2b113 1050 trace_oom_score_adj_update(task);
a63d83f4
DR
1051 /*
1052 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1053 * always attainable.
1054 */
1055 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1056 task->signal->oom_adj = OOM_DISABLE;
1057 else
1058 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1059 OOM_SCORE_ADJ_MAX;
723548bf 1060err_sighand:
a63d83f4 1061 unlock_task_sighand(task, &flags);
d19d5476
DR
1062err_task_lock:
1063 task_unlock(task);
a63d83f4 1064 put_task_struct(task);
723548bf
DR
1065out:
1066 return err < 0 ? err : count;
a63d83f4
DR
1067}
1068
1069static const struct file_operations proc_oom_score_adj_operations = {
1070 .read = oom_score_adj_read,
1071 .write = oom_score_adj_write,
6038f373 1072 .llseek = default_llseek,
a63d83f4
DR
1073};
1074
1da177e4
LT
1075#ifdef CONFIG_AUDITSYSCALL
1076#define TMPBUFLEN 21
1077static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1078 size_t count, loff_t *ppos)
1079{
2fddfeef 1080 struct inode * inode = file->f_path.dentry->d_inode;
99f89551 1081 struct task_struct *task = get_proc_task(inode);
1da177e4
LT
1082 ssize_t length;
1083 char tmpbuf[TMPBUFLEN];
1084
99f89551
EB
1085 if (!task)
1086 return -ESRCH;
1da177e4 1087 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
0c11b942 1088 audit_get_loginuid(task));
99f89551 1089 put_task_struct(task);
1da177e4
LT
1090 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1091}
1092
1093static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1094 size_t count, loff_t *ppos)
1095{
2fddfeef 1096 struct inode * inode = file->f_path.dentry->d_inode;
1da177e4
LT
1097 char *page, *tmp;
1098 ssize_t length;
1da177e4
LT
1099 uid_t loginuid;
1100
7dc52157
PM
1101 rcu_read_lock();
1102 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1103 rcu_read_unlock();
1da177e4 1104 return -EPERM;
7dc52157
PM
1105 }
1106 rcu_read_unlock();
1da177e4 1107
e0182909
AV
1108 if (count >= PAGE_SIZE)
1109 count = PAGE_SIZE - 1;
1da177e4
LT
1110
1111 if (*ppos != 0) {
1112 /* No partial writes. */
1113 return -EINVAL;
1114 }
e12ba74d 1115 page = (char*)__get_free_page(GFP_TEMPORARY);
1da177e4
LT
1116 if (!page)
1117 return -ENOMEM;
1118 length = -EFAULT;
1119 if (copy_from_user(page, buf, count))
1120 goto out_free_page;
1121
e0182909 1122 page[count] = '\0';
1da177e4
LT
1123 loginuid = simple_strtoul(page, &tmp, 10);
1124 if (tmp == page) {
1125 length = -EINVAL;
1126 goto out_free_page;
1127
1128 }
0a300be6 1129 length = audit_set_loginuid(loginuid);
1da177e4
LT
1130 if (likely(length == 0))
1131 length = count;
1132
1133out_free_page:
1134 free_page((unsigned long) page);
1135 return length;
1136}
1137
00977a59 1138static const struct file_operations proc_loginuid_operations = {
1da177e4
LT
1139 .read = proc_loginuid_read,
1140 .write = proc_loginuid_write,
87df8424 1141 .llseek = generic_file_llseek,
1da177e4 1142};
1e0bd755
EP
1143
1144static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1145 size_t count, loff_t *ppos)
1146{
1147 struct inode * inode = file->f_path.dentry->d_inode;
1148 struct task_struct *task = get_proc_task(inode);
1149 ssize_t length;
1150 char tmpbuf[TMPBUFLEN];
1151
1152 if (!task)
1153 return -ESRCH;
1154 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1155 audit_get_sessionid(task));
1156 put_task_struct(task);
1157 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1158}
1159
1160static const struct file_operations proc_sessionid_operations = {
1161 .read = proc_sessionid_read,
87df8424 1162 .llseek = generic_file_llseek,
1e0bd755 1163};
1da177e4
LT
1164#endif
1165
f4f154fd
AM
1166#ifdef CONFIG_FAULT_INJECTION
1167static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1168 size_t count, loff_t *ppos)
1169{
1170 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1171 char buffer[PROC_NUMBUF];
1172 size_t len;
1173 int make_it_fail;
f4f154fd
AM
1174
1175 if (!task)
1176 return -ESRCH;
1177 make_it_fail = task->make_it_fail;
1178 put_task_struct(task);
1179
1180 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
0c28f287
AM
1181
1182 return simple_read_from_buffer(buf, count, ppos, buffer, len);
f4f154fd
AM
1183}
1184
1185static ssize_t proc_fault_inject_write(struct file * file,
1186 const char __user * buf, size_t count, loff_t *ppos)
1187{
1188 struct task_struct *task;
1189 char buffer[PROC_NUMBUF], *end;
1190 int make_it_fail;
1191
1192 if (!capable(CAP_SYS_RESOURCE))
1193 return -EPERM;
1194 memset(buffer, 0, sizeof(buffer));
1195 if (count > sizeof(buffer) - 1)
1196 count = sizeof(buffer) - 1;
1197 if (copy_from_user(buffer, buf, count))
1198 return -EFAULT;
cba8aafe
VL
1199 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1200 if (*end)
1201 return -EINVAL;
f4f154fd
AM
1202 task = get_proc_task(file->f_dentry->d_inode);
1203 if (!task)
1204 return -ESRCH;
1205 task->make_it_fail = make_it_fail;
1206 put_task_struct(task);
cba8aafe
VL
1207
1208 return count;
f4f154fd
AM
1209}
1210
00977a59 1211static const struct file_operations proc_fault_inject_operations = {
f4f154fd
AM
1212 .read = proc_fault_inject_read,
1213 .write = proc_fault_inject_write,
87df8424 1214 .llseek = generic_file_llseek,
f4f154fd
AM
1215};
1216#endif
1217
9745512c 1218
43ae34cb
IM
1219#ifdef CONFIG_SCHED_DEBUG
1220/*
1221 * Print out various scheduling related per-task fields:
1222 */
1223static int sched_show(struct seq_file *m, void *v)
1224{
1225 struct inode *inode = m->private;
1226 struct task_struct *p;
1227
43ae34cb
IM
1228 p = get_proc_task(inode);
1229 if (!p)
1230 return -ESRCH;
1231 proc_sched_show_task(p, m);
1232
1233 put_task_struct(p);
1234
1235 return 0;
1236}
1237
1238static ssize_t
1239sched_write(struct file *file, const char __user *buf,
1240 size_t count, loff_t *offset)
1241{
1242 struct inode *inode = file->f_path.dentry->d_inode;
1243 struct task_struct *p;
1244
43ae34cb
IM
1245 p = get_proc_task(inode);
1246 if (!p)
1247 return -ESRCH;
1248 proc_sched_set_task(p);
1249
1250 put_task_struct(p);
1251
1252 return count;
1253}
1254
1255static int sched_open(struct inode *inode, struct file *filp)
1256{
c6a34058 1257 return single_open(filp, sched_show, inode);
43ae34cb
IM
1258}
1259
1260static const struct file_operations proc_pid_sched_operations = {
1261 .open = sched_open,
1262 .read = seq_read,
1263 .write = sched_write,
1264 .llseek = seq_lseek,
5ea473a1 1265 .release = single_release,
43ae34cb
IM
1266};
1267
1268#endif
1269
5091faa4
MG
1270#ifdef CONFIG_SCHED_AUTOGROUP
1271/*
1272 * Print out autogroup related information:
1273 */
1274static int sched_autogroup_show(struct seq_file *m, void *v)
1275{
1276 struct inode *inode = m->private;
1277 struct task_struct *p;
1278
1279 p = get_proc_task(inode);
1280 if (!p)
1281 return -ESRCH;
1282 proc_sched_autogroup_show_task(p, m);
1283
1284 put_task_struct(p);
1285
1286 return 0;
1287}
1288
1289static ssize_t
1290sched_autogroup_write(struct file *file, const char __user *buf,
1291 size_t count, loff_t *offset)
1292{
1293 struct inode *inode = file->f_path.dentry->d_inode;
1294 struct task_struct *p;
1295 char buffer[PROC_NUMBUF];
0a8cb8e3 1296 int nice;
5091faa4
MG
1297 int err;
1298
1299 memset(buffer, 0, sizeof(buffer));
1300 if (count > sizeof(buffer) - 1)
1301 count = sizeof(buffer) - 1;
1302 if (copy_from_user(buffer, buf, count))
1303 return -EFAULT;
1304
0a8cb8e3
AD
1305 err = kstrtoint(strstrip(buffer), 0, &nice);
1306 if (err < 0)
1307 return err;
5091faa4
MG
1308
1309 p = get_proc_task(inode);
1310 if (!p)
1311 return -ESRCH;
1312
2e5b5b3a 1313 err = proc_sched_autogroup_set_nice(p, nice);
5091faa4
MG
1314 if (err)
1315 count = err;
1316
1317 put_task_struct(p);
1318
1319 return count;
1320}
1321
1322static int sched_autogroup_open(struct inode *inode, struct file *filp)
1323{
1324 int ret;
1325
1326 ret = single_open(filp, sched_autogroup_show, NULL);
1327 if (!ret) {
1328 struct seq_file *m = filp->private_data;
1329
1330 m->private = inode;
1331 }
1332 return ret;
1333}
1334
1335static const struct file_operations proc_pid_sched_autogroup_operations = {
1336 .open = sched_autogroup_open,
1337 .read = seq_read,
1338 .write = sched_autogroup_write,
1339 .llseek = seq_lseek,
1340 .release = single_release,
1341};
1342
1343#endif /* CONFIG_SCHED_AUTOGROUP */
1344
4614a696
JS
1345static ssize_t comm_write(struct file *file, const char __user *buf,
1346 size_t count, loff_t *offset)
1347{
1348 struct inode *inode = file->f_path.dentry->d_inode;
1349 struct task_struct *p;
1350 char buffer[TASK_COMM_LEN];
1351
1352 memset(buffer, 0, sizeof(buffer));
1353 if (count > sizeof(buffer) - 1)
1354 count = sizeof(buffer) - 1;
1355 if (copy_from_user(buffer, buf, count))
1356 return -EFAULT;
1357
1358 p = get_proc_task(inode);
1359 if (!p)
1360 return -ESRCH;
1361
1362 if (same_thread_group(current, p))
1363 set_task_comm(p, buffer);
1364 else
1365 count = -EINVAL;
1366
1367 put_task_struct(p);
1368
1369 return count;
1370}
1371
1372static int comm_show(struct seq_file *m, void *v)
1373{
1374 struct inode *inode = m->private;
1375 struct task_struct *p;
1376
1377 p = get_proc_task(inode);
1378 if (!p)
1379 return -ESRCH;
1380
1381 task_lock(p);
1382 seq_printf(m, "%s\n", p->comm);
1383 task_unlock(p);
1384
1385 put_task_struct(p);
1386
1387 return 0;
1388}
1389
1390static int comm_open(struct inode *inode, struct file *filp)
1391{
c6a34058 1392 return single_open(filp, comm_show, inode);
4614a696
JS
1393}
1394
1395static const struct file_operations proc_pid_set_comm_operations = {
1396 .open = comm_open,
1397 .read = seq_read,
1398 .write = comm_write,
1399 .llseek = seq_lseek,
1400 .release = single_release,
1401};
1402
7773fbc5 1403static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
925d1c40
MH
1404{
1405 struct task_struct *task;
1406 struct mm_struct *mm;
1407 struct file *exe_file;
1408
7773fbc5 1409 task = get_proc_task(dentry->d_inode);
925d1c40
MH
1410 if (!task)
1411 return -ENOENT;
1412 mm = get_task_mm(task);
1413 put_task_struct(task);
1414 if (!mm)
1415 return -ENOENT;
1416 exe_file = get_mm_exe_file(mm);
1417 mmput(mm);
1418 if (exe_file) {
1419 *exe_path = exe_file->f_path;
1420 path_get(&exe_file->f_path);
1421 fput(exe_file);
1422 return 0;
1423 } else
1424 return -ENOENT;
1425}
1426
008b150a 1427static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1da177e4
LT
1428{
1429 struct inode *inode = dentry->d_inode;
1430 int error = -EACCES;
1431
1432 /* We don't need a base pointer in the /proc filesystem */
1d957f9b 1433 path_put(&nd->path);
1da177e4 1434
778c1144
EB
1435 /* Are we allowed to snoop on the tasks file descriptors? */
1436 if (!proc_fd_access_allowed(inode))
1da177e4 1437 goto out;
1da177e4 1438
7773fbc5 1439 error = PROC_I(inode)->op.proc_get_link(dentry, &nd->path);
1da177e4 1440out:
008b150a 1441 return ERR_PTR(error);
1da177e4
LT
1442}
1443
3dcd25f3 1444static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1da177e4 1445{
e12ba74d 1446 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
3dcd25f3 1447 char *pathname;
1da177e4
LT
1448 int len;
1449
1450 if (!tmp)
1451 return -ENOMEM;
0c28f287 1452
7b2a69ba 1453 pathname = d_path(path, tmp, PAGE_SIZE);
3dcd25f3
JB
1454 len = PTR_ERR(pathname);
1455 if (IS_ERR(pathname))
1da177e4 1456 goto out;
3dcd25f3 1457 len = tmp + PAGE_SIZE - 1 - pathname;
1da177e4
LT
1458
1459 if (len > buflen)
1460 len = buflen;
3dcd25f3 1461 if (copy_to_user(buffer, pathname, len))
1da177e4
LT
1462 len = -EFAULT;
1463 out:
1464 free_page((unsigned long)tmp);
1465 return len;
1466}
1467
1468static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1469{
1470 int error = -EACCES;
1471 struct inode *inode = dentry->d_inode;
3dcd25f3 1472 struct path path;
1da177e4 1473
778c1144
EB
1474 /* Are we allowed to snoop on the tasks file descriptors? */
1475 if (!proc_fd_access_allowed(inode))
1da177e4 1476 goto out;
1da177e4 1477
7773fbc5 1478 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1da177e4
LT
1479 if (error)
1480 goto out;
1481
3dcd25f3
JB
1482 error = do_proc_readlink(&path, buffer, buflen);
1483 path_put(&path);
1da177e4 1484out:
1da177e4
LT
1485 return error;
1486}
1487
c5ef1c42 1488static const struct inode_operations proc_pid_link_inode_operations = {
1da177e4 1489 .readlink = proc_pid_readlink,
6d76fa58
LT
1490 .follow_link = proc_pid_follow_link,
1491 .setattr = proc_setattr,
1da177e4
LT
1492};
1493
28a6d671
EB
1494
1495/* building an inode */
1496
1497static int task_dumpable(struct task_struct *task)
1da177e4 1498{
28a6d671
EB
1499 int dumpable = 0;
1500 struct mm_struct *mm;
1da177e4 1501
28a6d671
EB
1502 task_lock(task);
1503 mm = task->mm;
1504 if (mm)
6c5d5238 1505 dumpable = get_dumpable(mm);
28a6d671
EB
1506 task_unlock(task);
1507 if(dumpable == 1)
1508 return 1;
1509 return 0;
1510}
1da177e4 1511
6b4e306a 1512struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
28a6d671
EB
1513{
1514 struct inode * inode;
1515 struct proc_inode *ei;
c69e8d9c 1516 const struct cred *cred;
1da177e4 1517
28a6d671 1518 /* We need a new inode */
1da177e4 1519
28a6d671
EB
1520 inode = new_inode(sb);
1521 if (!inode)
1522 goto out;
1523
1524 /* Common stuff */
1525 ei = PROC_I(inode);
85fe4025 1526 inode->i_ino = get_next_ino();
28a6d671 1527 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
28a6d671
EB
1528 inode->i_op = &proc_def_inode_operations;
1529
1530 /*
1531 * grab the reference to task.
1532 */
1a657f78 1533 ei->pid = get_task_pid(task, PIDTYPE_PID);
28a6d671
EB
1534 if (!ei->pid)
1535 goto out_unlock;
1536
28a6d671 1537 if (task_dumpable(task)) {
c69e8d9c
DH
1538 rcu_read_lock();
1539 cred = __task_cred(task);
1540 inode->i_uid = cred->euid;
1541 inode->i_gid = cred->egid;
1542 rcu_read_unlock();
1da177e4 1543 }
28a6d671
EB
1544 security_task_to_inode(task, inode);
1545
1da177e4 1546out:
28a6d671
EB
1547 return inode;
1548
1549out_unlock:
1550 iput(inode);
1551 return NULL;
1da177e4
LT
1552}
1553
6b4e306a 1554int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1da177e4 1555{
1da177e4 1556 struct inode *inode = dentry->d_inode;
28a6d671 1557 struct task_struct *task;
c69e8d9c 1558 const struct cred *cred;
0499680a 1559 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
c69e8d9c 1560
28a6d671 1561 generic_fillattr(inode, stat);
1da177e4 1562
28a6d671 1563 rcu_read_lock();
dcb0f222
EB
1564 stat->uid = GLOBAL_ROOT_UID;
1565 stat->gid = GLOBAL_ROOT_GID;
28a6d671
EB
1566 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1567 if (task) {
0499680a
VK
1568 if (!has_pid_permissions(pid, task, 2)) {
1569 rcu_read_unlock();
1570 /*
1571 * This doesn't prevent learning whether PID exists,
1572 * it only makes getattr() consistent with readdir().
1573 */
1574 return -ENOENT;
1575 }
28a6d671
EB
1576 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1577 task_dumpable(task)) {
c69e8d9c
DH
1578 cred = __task_cred(task);
1579 stat->uid = cred->euid;
1580 stat->gid = cred->egid;
1da177e4
LT
1581 }
1582 }
28a6d671 1583 rcu_read_unlock();
d6e71144 1584 return 0;
1da177e4
LT
1585}
1586
1da177e4
LT
1587/* dentry stuff */
1588
1589/*
1590 * Exceptional case: normally we are not allowed to unhash a busy
1591 * directory. In this case, however, we can do it - no aliasing problems
1592 * due to the way we treat inodes.
1593 *
1594 * Rewrite the inode's ownerships here because the owning task may have
1595 * performed a setuid(), etc.
99f89551
EB
1596 *
1597 * Before the /proc/pid/status file was created the only way to read
1598 * the effective uid of a /process was to stat /proc/pid. Reading
1599 * /proc/pid/status is slow enough that procps and other packages
1600 * kept stating /proc/pid. To keep the rules in /proc simple I have
1601 * made this apply to all per process world readable and executable
1602 * directories.
1da177e4 1603 */
6b4e306a 1604int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1da177e4 1605{
34286d66
NP
1606 struct inode *inode;
1607 struct task_struct *task;
c69e8d9c
DH
1608 const struct cred *cred;
1609
34286d66
NP
1610 if (nd && nd->flags & LOOKUP_RCU)
1611 return -ECHILD;
1612
1613 inode = dentry->d_inode;
1614 task = get_proc_task(inode);
1615
99f89551
EB
1616 if (task) {
1617 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1618 task_dumpable(task)) {
c69e8d9c
DH
1619 rcu_read_lock();
1620 cred = __task_cred(task);
1621 inode->i_uid = cred->euid;
1622 inode->i_gid = cred->egid;
1623 rcu_read_unlock();
1da177e4 1624 } else {
dcb0f222
EB
1625 inode->i_uid = GLOBAL_ROOT_UID;
1626 inode->i_gid = GLOBAL_ROOT_GID;
1da177e4 1627 }
9ee8ab9f 1628 inode->i_mode &= ~(S_ISUID | S_ISGID);
1da177e4 1629 security_task_to_inode(task, inode);
99f89551 1630 put_task_struct(task);
1da177e4
LT
1631 return 1;
1632 }
1633 d_drop(dentry);
1634 return 0;
1635}
1636
fe15ce44 1637static int pid_delete_dentry(const struct dentry * dentry)
99f89551 1638{
28a6d671
EB
1639 /* Is the task we represent dead?
1640 * If so, then don't put the dentry on the lru list,
1641 * kill it immediately.
1642 */
1643 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1644}
1645
6b4e306a 1646const struct dentry_operations pid_dentry_operations =
28a6d671
EB
1647{
1648 .d_revalidate = pid_revalidate,
1649 .d_delete = pid_delete_dentry,
1650};
1651
1652/* Lookups */
1653
1c0d04c9
EB
1654/*
1655 * Fill a directory entry.
1656 *
1657 * If possible create the dcache entry and derive our inode number and
1658 * file type from dcache entry.
1659 *
1660 * Since all of the proc inode numbers are dynamically generated, the inode
1661 * numbers do not exist until the inode is cache. This means creating the
1662 * the dcache entry in readdir is necessary to keep the inode numbers
1663 * reported by readdir in sync with the inode numbers reported
1664 * by stat.
1665 */
6b4e306a
EB
1666int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1667 const char *name, int len,
c5141e6d 1668 instantiate_t instantiate, struct task_struct *task, const void *ptr)
61a28784 1669{
2fddfeef 1670 struct dentry *child, *dir = filp->f_path.dentry;
61a28784
EB
1671 struct inode *inode;
1672 struct qstr qname;
1673 ino_t ino = 0;
1674 unsigned type = DT_UNKNOWN;
1675
1676 qname.name = name;
1677 qname.len = len;
1678 qname.hash = full_name_hash(name, len);
1679
1680 child = d_lookup(dir, &qname);
1681 if (!child) {
1682 struct dentry *new;
1683 new = d_alloc(dir, &qname);
1684 if (new) {
1685 child = instantiate(dir->d_inode, new, task, ptr);
1686 if (child)
1687 dput(new);
1688 else
1689 child = new;
1690 }
1691 }
1692 if (!child || IS_ERR(child) || !child->d_inode)
1693 goto end_instantiate;
1694 inode = child->d_inode;
1695 if (inode) {
1696 ino = inode->i_ino;
1697 type = inode->i_mode >> 12;
1698 }
1699 dput(child);
1700end_instantiate:
1701 if (!ino)
1702 ino = find_inode_number(dir, &qname);
1703 if (!ino)
1704 ino = 1;
1705 return filldir(dirent, name, len, filp->f_pos, ino, type);
1706}
1707
28a6d671
EB
1708static unsigned name_to_int(struct dentry *dentry)
1709{
1710 const char *name = dentry->d_name.name;
1711 int len = dentry->d_name.len;
1712 unsigned n = 0;
1713
1714 if (len > 1 && *name == '0')
1715 goto out;
1716 while (len-- > 0) {
1717 unsigned c = *name++ - '0';
1718 if (c > 9)
1719 goto out;
1720 if (n >= (~0U-9)/10)
1721 goto out;
1722 n *= 10;
1723 n += c;
1724 }
1725 return n;
1726out:
1727 return ~0U;
1728}
1729
27932742
MS
1730#define PROC_FDINFO_MAX 64
1731
3dcd25f3 1732static int proc_fd_info(struct inode *inode, struct path *path, char *info)
28a6d671 1733{
5e442a49
LT
1734 struct task_struct *task = get_proc_task(inode);
1735 struct files_struct *files = NULL;
28a6d671
EB
1736 struct file *file;
1737 int fd = proc_fd(inode);
aa6afca5 1738
5e442a49
LT
1739 if (task) {
1740 files = get_files_struct(task);
1741 put_task_struct(task);
1742 }
1743 if (files) {
1744 /*
1745 * We are not taking a ref to the file structure, so we must
1746 * hold ->file_lock.
1747 */
1748 spin_lock(&files->file_lock);
1749 file = fcheck_files(files, fd);
1750 if (file) {
1751 unsigned int f_flags;
1752 struct fdtable *fdt;
1753
1754 fdt = files_fdtable(files);
1755 f_flags = file->f_flags & ~O_CLOEXEC;
1dce27c5 1756 if (close_on_exec(fd, fdt))
5e442a49
LT
1757 f_flags |= O_CLOEXEC;
1758
1759 if (path) {
1760 *path = file->f_path;
1761 path_get(&file->f_path);
1762 }
1763 if (info)
1764 snprintf(info, PROC_FDINFO_MAX,
1765 "pos:\t%lli\n"
1766 "flags:\t0%o\n",
1767 (long long) file->f_pos,
1768 f_flags);
1769 spin_unlock(&files->file_lock);
1770 put_files_struct(files);
1771 return 0;
99f89551 1772 }
5e442a49
LT
1773 spin_unlock(&files->file_lock);
1774 put_files_struct(files);
1775 }
1776 return -ENOENT;
99f89551
EB
1777}
1778
7773fbc5 1779static int proc_fd_link(struct dentry *dentry, struct path *path)
27932742 1780{
7773fbc5 1781 return proc_fd_info(dentry->d_inode, path, NULL);
27932742
MS
1782}
1783
1da177e4
LT
1784static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1785{
34286d66
NP
1786 struct inode *inode;
1787 struct task_struct *task;
1788 int fd;
1da177e4 1789 struct files_struct *files;
c69e8d9c 1790 const struct cred *cred;
1da177e4 1791
34286d66
NP
1792 if (nd && nd->flags & LOOKUP_RCU)
1793 return -ECHILD;
1794
1795 inode = dentry->d_inode;
1796 task = get_proc_task(inode);
1797 fd = proc_fd(inode);
1798
99f89551
EB
1799 if (task) {
1800 files = get_files_struct(task);
1801 if (files) {
30a08bf2 1802 struct file *file;
99f89551 1803 rcu_read_lock();
30a08bf2
LT
1804 file = fcheck_files(files, fd);
1805 if (file) {
1806 unsigned i_mode, f_mode = file->f_mode;
1807
99f89551
EB
1808 rcu_read_unlock();
1809 put_files_struct(files);
30a08bf2 1810
99f89551 1811 if (task_dumpable(task)) {
c69e8d9c
DH
1812 rcu_read_lock();
1813 cred = __task_cred(task);
1814 inode->i_uid = cred->euid;
1815 inode->i_gid = cred->egid;
1816 rcu_read_unlock();
99f89551 1817 } else {
dcb0f222
EB
1818 inode->i_uid = GLOBAL_ROOT_UID;
1819 inode->i_gid = GLOBAL_ROOT_GID;
99f89551 1820 }
30a08bf2
LT
1821
1822 i_mode = S_IFLNK;
1823 if (f_mode & FMODE_READ)
1824 i_mode |= S_IRUSR | S_IXUSR;
1825 if (f_mode & FMODE_WRITE)
1826 i_mode |= S_IWUSR | S_IXUSR;
1827 inode->i_mode = i_mode;
1828
99f89551
EB
1829 security_task_to_inode(task, inode);
1830 put_task_struct(task);
1831 return 1;
1832 }
b835996f 1833 rcu_read_unlock();
1da177e4 1834 put_files_struct(files);
1da177e4 1835 }
99f89551 1836 put_task_struct(task);
1da177e4
LT
1837 }
1838 d_drop(dentry);
1839 return 0;
1840}
1841
d72f71eb 1842static const struct dentry_operations tid_fd_dentry_operations =
1da177e4
LT
1843{
1844 .d_revalidate = tid_fd_revalidate,
1845 .d_delete = pid_delete_dentry,
1846};
1847
444ceed8 1848static struct dentry *proc_fd_instantiate(struct inode *dir,
c5141e6d 1849 struct dentry *dentry, struct task_struct *task, const void *ptr)
1da177e4 1850{
c5141e6d 1851 unsigned fd = *(const unsigned *)ptr;
444ceed8
EB
1852 struct inode *inode;
1853 struct proc_inode *ei;
1854 struct dentry *error = ERR_PTR(-ENOENT);
1da177e4 1855
61a28784 1856 inode = proc_pid_make_inode(dir->i_sb, task);
1da177e4
LT
1857 if (!inode)
1858 goto out;
1859 ei = PROC_I(inode);
aed7a6c4 1860 ei->fd = fd;
444ceed8 1861
5e442a49 1862 inode->i_op = &proc_pid_link_inode_operations;
1da177e4
LT
1863 inode->i_size = 64;
1864 ei->op.proc_get_link = proc_fd_link;
fb045adb 1865 d_set_d_op(dentry, &tid_fd_dentry_operations);
1da177e4 1866 d_add(dentry, inode);
cd6a3ce9
EB
1867 /* Close the race of the process dying before we return the dentry */
1868 if (tid_fd_revalidate(dentry, NULL))
444ceed8 1869 error = NULL;
1da177e4 1870
444ceed8
EB
1871 out:
1872 return error;
1da177e4
LT
1873}
1874
27932742
MS
1875static struct dentry *proc_lookupfd_common(struct inode *dir,
1876 struct dentry *dentry,
1877 instantiate_t instantiate)
444ceed8
EB
1878{
1879 struct task_struct *task = get_proc_task(dir);
1880 unsigned fd = name_to_int(dentry);
1881 struct dentry *result = ERR_PTR(-ENOENT);
1882
1883 if (!task)
1884 goto out_no_task;
1885 if (fd == ~0U)
1886 goto out;
1887
27932742 1888 result = instantiate(dir, dentry, task, &fd);
444ceed8
EB
1889out:
1890 put_task_struct(task);
1891out_no_task:
1892 return result;
1893}
1894
27932742
MS
1895static int proc_readfd_common(struct file * filp, void * dirent,
1896 filldir_t filldir, instantiate_t instantiate)
28a6d671 1897{
2fddfeef 1898 struct dentry *dentry = filp->f_path.dentry;
28a6d671
EB
1899 struct inode *inode = dentry->d_inode;
1900 struct task_struct *p = get_proc_task(inode);
457c2510 1901 unsigned int fd, ino;
28a6d671 1902 int retval;
28a6d671 1903 struct files_struct * files;
1da177e4 1904
28a6d671
EB
1905 retval = -ENOENT;
1906 if (!p)
1907 goto out_no_task;
1908 retval = 0;
28a6d671
EB
1909
1910 fd = filp->f_pos;
1911 switch (fd) {
1912 case 0:
1913 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
5e442a49 1914 goto out;
28a6d671
EB
1915 filp->f_pos++;
1916 case 1:
1917 ino = parent_ino(dentry);
1918 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
5e442a49 1919 goto out;
28a6d671
EB
1920 filp->f_pos++;
1921 default:
1922 files = get_files_struct(p);
1923 if (!files)
5e442a49 1924 goto out;
28a6d671 1925 rcu_read_lock();
28a6d671 1926 for (fd = filp->f_pos-2;
9b4f526c 1927 fd < files_fdtable(files)->max_fds;
28a6d671 1928 fd++, filp->f_pos++) {
27932742
MS
1929 char name[PROC_NUMBUF];
1930 int len;
28a6d671
EB
1931
1932 if (!fcheck_files(files, fd))
1933 continue;
1934 rcu_read_unlock();
1935
27932742
MS
1936 len = snprintf(name, sizeof(name), "%d", fd);
1937 if (proc_fill_cache(filp, dirent, filldir,
1938 name, len, instantiate,
1939 p, &fd) < 0) {
28a6d671
EB
1940 rcu_read_lock();
1941 break;
1942 }
1943 rcu_read_lock();
1944 }
1945 rcu_read_unlock();
1946 put_files_struct(files);
1947 }
1948out:
1949 put_task_struct(p);
1950out_no_task:
1951 return retval;
1952}
1953
27932742
MS
1954static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1955 struct nameidata *nd)
1956{
1957 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1958}
1959
1960static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1961{
1962 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1963}
1964
1965static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1966 size_t len, loff_t *ppos)
1967{
1968 char tmp[PROC_FDINFO_MAX];
3dcd25f3 1969 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
27932742
MS
1970 if (!err)
1971 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1972 return err;
1973}
1974
1975static const struct file_operations proc_fdinfo_file_operations = {
87df8424 1976 .open = nonseekable_open,
27932742 1977 .read = proc_fdinfo_read,
6038f373 1978 .llseek = no_llseek,
27932742
MS
1979};
1980
00977a59 1981static const struct file_operations proc_fd_operations = {
28a6d671
EB
1982 .read = generic_read_dir,
1983 .readdir = proc_readfd,
6038f373 1984 .llseek = default_llseek,
1da177e4
LT
1985};
1986
640708a2
PE
1987#ifdef CONFIG_CHECKPOINT_RESTORE
1988
1989/*
1990 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1991 * which represent vma start and end addresses.
1992 */
1993static int dname_to_vma_addr(struct dentry *dentry,
1994 unsigned long *start, unsigned long *end)
1995{
1996 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
1997 return -EINVAL;
1998
1999 return 0;
2000}
2001
2002static int map_files_d_revalidate(struct dentry *dentry, struct nameidata *nd)
2003{
2004 unsigned long vm_start, vm_end;
2005 bool exact_vma_exists = false;
2006 struct mm_struct *mm = NULL;
2007 struct task_struct *task;
2008 const struct cred *cred;
2009 struct inode *inode;
2010 int status = 0;
2011
2012 if (nd && nd->flags & LOOKUP_RCU)
2013 return -ECHILD;
2014
2015 if (!capable(CAP_SYS_ADMIN)) {
2016 status = -EACCES;
2017 goto out_notask;
2018 }
2019
2020 inode = dentry->d_inode;
2021 task = get_proc_task(inode);
2022 if (!task)
2023 goto out_notask;
2024
2025 if (!ptrace_may_access(task, PTRACE_MODE_READ))
2026 goto out;
2027
2028 mm = get_task_mm(task);
2029 if (!mm)
2030 goto out;
2031
2032 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
2033 down_read(&mm->mmap_sem);
2034 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
2035 up_read(&mm->mmap_sem);
2036 }
2037
2038 mmput(mm);
2039
2040 if (exact_vma_exists) {
2041 if (task_dumpable(task)) {
2042 rcu_read_lock();
2043 cred = __task_cred(task);
2044 inode->i_uid = cred->euid;
2045 inode->i_gid = cred->egid;
2046 rcu_read_unlock();
2047 } else {
dcb0f222
EB
2048 inode->i_uid = GLOBAL_ROOT_UID;
2049 inode->i_gid = GLOBAL_ROOT_GID;
640708a2
PE
2050 }
2051 security_task_to_inode(task, inode);
2052 status = 1;
2053 }
2054
2055out:
2056 put_task_struct(task);
2057
2058out_notask:
2059 if (status <= 0)
2060 d_drop(dentry);
2061
2062 return status;
2063}
2064
2065static const struct dentry_operations tid_map_files_dentry_operations = {
2066 .d_revalidate = map_files_d_revalidate,
2067 .d_delete = pid_delete_dentry,
2068};
2069
2070static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
2071{
2072 unsigned long vm_start, vm_end;
2073 struct vm_area_struct *vma;
2074 struct task_struct *task;
2075 struct mm_struct *mm;
2076 int rc;
2077
2078 rc = -ENOENT;
2079 task = get_proc_task(dentry->d_inode);
2080 if (!task)
2081 goto out;
2082
2083 mm = get_task_mm(task);
2084 put_task_struct(task);
2085 if (!mm)
2086 goto out;
2087
2088 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2089 if (rc)
2090 goto out_mmput;
2091
2092 down_read(&mm->mmap_sem);
2093 vma = find_exact_vma(mm, vm_start, vm_end);
2094 if (vma && vma->vm_file) {
2095 *path = vma->vm_file->f_path;
2096 path_get(path);
2097 rc = 0;
2098 }
2099 up_read(&mm->mmap_sem);
2100
2101out_mmput:
2102 mmput(mm);
2103out:
2104 return rc;
2105}
2106
2107struct map_files_info {
2108 struct file *file;
2109 unsigned long len;
2110 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
2111};
2112
2113static struct dentry *
2114proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
2115 struct task_struct *task, const void *ptr)
2116{
2117 const struct file *file = ptr;
2118 struct proc_inode *ei;
2119 struct inode *inode;
2120
2121 if (!file)
2122 return ERR_PTR(-ENOENT);
2123
2124 inode = proc_pid_make_inode(dir->i_sb, task);
2125 if (!inode)
2126 return ERR_PTR(-ENOENT);
2127
2128 ei = PROC_I(inode);
2129 ei->op.proc_get_link = proc_map_files_get_link;
2130
2131 inode->i_op = &proc_pid_link_inode_operations;
2132 inode->i_size = 64;
2133 inode->i_mode = S_IFLNK;
2134
2135 if (file->f_mode & FMODE_READ)
2136 inode->i_mode |= S_IRUSR;
2137 if (file->f_mode & FMODE_WRITE)
2138 inode->i_mode |= S_IWUSR;
2139
2140 d_set_d_op(dentry, &tid_map_files_dentry_operations);
2141 d_add(dentry, inode);
2142
2143 return NULL;
2144}
2145
2146static struct dentry *proc_map_files_lookup(struct inode *dir,
2147 struct dentry *dentry, struct nameidata *nd)
2148{
2149 unsigned long vm_start, vm_end;
2150 struct vm_area_struct *vma;
2151 struct task_struct *task;
2152 struct dentry *result;
2153 struct mm_struct *mm;
2154
2155 result = ERR_PTR(-EACCES);
2156 if (!capable(CAP_SYS_ADMIN))
2157 goto out;
2158
2159 result = ERR_PTR(-ENOENT);
2160 task = get_proc_task(dir);
2161 if (!task)
2162 goto out;
2163
2164 result = ERR_PTR(-EACCES);
eb94cd96 2165 if (!ptrace_may_access(task, PTRACE_MODE_READ))
640708a2
PE
2166 goto out_put_task;
2167
2168 result = ERR_PTR(-ENOENT);
2169 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
eb94cd96 2170 goto out_put_task;
640708a2
PE
2171
2172 mm = get_task_mm(task);
2173 if (!mm)
eb94cd96 2174 goto out_put_task;
640708a2
PE
2175
2176 down_read(&mm->mmap_sem);
2177 vma = find_exact_vma(mm, vm_start, vm_end);
2178 if (!vma)
2179 goto out_no_vma;
2180
2181 result = proc_map_files_instantiate(dir, dentry, task, vma->vm_file);
2182
2183out_no_vma:
2184 up_read(&mm->mmap_sem);
2185 mmput(mm);
640708a2
PE
2186out_put_task:
2187 put_task_struct(task);
2188out:
2189 return result;
2190}
2191
2192static const struct inode_operations proc_map_files_inode_operations = {
2193 .lookup = proc_map_files_lookup,
2194 .permission = proc_fd_permission,
2195 .setattr = proc_setattr,
2196};
2197
2198static int
2199proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
2200{
2201 struct dentry *dentry = filp->f_path.dentry;
2202 struct inode *inode = dentry->d_inode;
2203 struct vm_area_struct *vma;
2204 struct task_struct *task;
2205 struct mm_struct *mm;
2206 ino_t ino;
2207 int ret;
2208
2209 ret = -EACCES;
2210 if (!capable(CAP_SYS_ADMIN))
2211 goto out;
2212
2213 ret = -ENOENT;
2214 task = get_proc_task(inode);
2215 if (!task)
2216 goto out;
2217
2218 ret = -EACCES;
eb94cd96 2219 if (!ptrace_may_access(task, PTRACE_MODE_READ))
640708a2
PE
2220 goto out_put_task;
2221
2222 ret = 0;
2223 switch (filp->f_pos) {
2224 case 0:
2225 ino = inode->i_ino;
2226 if (filldir(dirent, ".", 1, 0, ino, DT_DIR) < 0)
eb94cd96 2227 goto out_put_task;
640708a2
PE
2228 filp->f_pos++;
2229 case 1:
2230 ino = parent_ino(dentry);
2231 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
eb94cd96 2232 goto out_put_task;
640708a2
PE
2233 filp->f_pos++;
2234 default:
2235 {
2236 unsigned long nr_files, pos, i;
2237 struct flex_array *fa = NULL;
2238 struct map_files_info info;
2239 struct map_files_info *p;
2240
2241 mm = get_task_mm(task);
2242 if (!mm)
eb94cd96 2243 goto out_put_task;
640708a2
PE
2244 down_read(&mm->mmap_sem);
2245
2246 nr_files = 0;
2247
2248 /*
2249 * We need two passes here:
2250 *
2251 * 1) Collect vmas of mapped files with mmap_sem taken
2252 * 2) Release mmap_sem and instantiate entries
2253 *
2254 * otherwise we get lockdep complained, since filldir()
2255 * routine might require mmap_sem taken in might_fault().
2256 */
2257
2258 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2259 if (vma->vm_file && ++pos > filp->f_pos)
2260 nr_files++;
2261 }
2262
2263 if (nr_files) {
2264 fa = flex_array_alloc(sizeof(info), nr_files,
2265 GFP_KERNEL);
2266 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2267 GFP_KERNEL)) {
2268 ret = -ENOMEM;
2269 if (fa)
2270 flex_array_free(fa);
2271 up_read(&mm->mmap_sem);
2272 mmput(mm);
eb94cd96 2273 goto out_put_task;
640708a2
PE
2274 }
2275 for (i = 0, vma = mm->mmap, pos = 2; vma;
2276 vma = vma->vm_next) {
2277 if (!vma->vm_file)
2278 continue;
2279 if (++pos <= filp->f_pos)
2280 continue;
2281
2282 get_file(vma->vm_file);
2283 info.file = vma->vm_file;
2284 info.len = snprintf(info.name,
2285 sizeof(info.name), "%lx-%lx",
2286 vma->vm_start, vma->vm_end);
2287 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2288 BUG();
2289 }
2290 }
2291 up_read(&mm->mmap_sem);
2292
2293 for (i = 0; i < nr_files; i++) {
2294 p = flex_array_get(fa, i);
2295 ret = proc_fill_cache(filp, dirent, filldir,
2296 p->name, p->len,
2297 proc_map_files_instantiate,
2298 task, p->file);
2299 if (ret)
2300 break;
2301 filp->f_pos++;
2302 fput(p->file);
2303 }
2304 for (; i < nr_files; i++) {
2305 /*
2306 * In case of error don't forget
2307 * to put rest of file refs.
2308 */
2309 p = flex_array_get(fa, i);
2310 fput(p->file);
2311 }
2312 if (fa)
2313 flex_array_free(fa);
2314 mmput(mm);
2315 }
2316 }
2317
640708a2
PE
2318out_put_task:
2319 put_task_struct(task);
2320out:
2321 return ret;
2322}
2323
2324static const struct file_operations proc_map_files_operations = {
2325 .read = generic_read_dir,
2326 .readdir = proc_map_files_readdir,
2327 .llseek = default_llseek,
2328};
2329
2330#endif /* CONFIG_CHECKPOINT_RESTORE */
2331
8948e11f
AD
2332/*
2333 * /proc/pid/fd needs a special permission handler so that a process can still
2334 * access /proc/self/fd after it has executed a setuid().
2335 */
10556cb2 2336static int proc_fd_permission(struct inode *inode, int mask)
8948e11f 2337{
2830ba7f 2338 int rv = generic_permission(inode, mask);
8948e11f
AD
2339 if (rv == 0)
2340 return 0;
2341 if (task_pid(current) == proc_pid(inode))
2342 rv = 0;
2343 return rv;
2344}
2345
1da177e4
LT
2346/*
2347 * proc directories can do almost nothing..
2348 */
c5ef1c42 2349static const struct inode_operations proc_fd_inode_operations = {
1da177e4 2350 .lookup = proc_lookupfd,
8948e11f 2351 .permission = proc_fd_permission,
6d76fa58 2352 .setattr = proc_setattr,
1da177e4
LT
2353};
2354
27932742
MS
2355static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2356 struct dentry *dentry, struct task_struct *task, const void *ptr)
2357{
2358 unsigned fd = *(unsigned *)ptr;
2359 struct inode *inode;
2360 struct proc_inode *ei;
2361 struct dentry *error = ERR_PTR(-ENOENT);
2362
2363 inode = proc_pid_make_inode(dir->i_sb, task);
2364 if (!inode)
2365 goto out;
2366 ei = PROC_I(inode);
2367 ei->fd = fd;
2368 inode->i_mode = S_IFREG | S_IRUSR;
2369 inode->i_fop = &proc_fdinfo_file_operations;
fb045adb 2370 d_set_d_op(dentry, &tid_fd_dentry_operations);
27932742
MS
2371 d_add(dentry, inode);
2372 /* Close the race of the process dying before we return the dentry */
2373 if (tid_fd_revalidate(dentry, NULL))
2374 error = NULL;
2375
2376 out:
2377 return error;
2378}
2379
2380static struct dentry *proc_lookupfdinfo(struct inode *dir,
2381 struct dentry *dentry,
2382 struct nameidata *nd)
2383{
2384 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2385}
2386
2387static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2388{
2389 return proc_readfd_common(filp, dirent, filldir,
2390 proc_fdinfo_instantiate);
2391}
2392
2393static const struct file_operations proc_fdinfo_operations = {
2394 .read = generic_read_dir,
2395 .readdir = proc_readfdinfo,
6038f373 2396 .llseek = default_llseek,
27932742
MS
2397};
2398
2399/*
2400 * proc directories can do almost nothing..
2401 */
2402static const struct inode_operations proc_fdinfo_inode_operations = {
2403 .lookup = proc_lookupfdinfo,
2404 .setattr = proc_setattr,
2405};
2406
2407
444ceed8 2408static struct dentry *proc_pident_instantiate(struct inode *dir,
c5141e6d 2409 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8 2410{
c5141e6d 2411 const struct pid_entry *p = ptr;
444ceed8
EB
2412 struct inode *inode;
2413 struct proc_inode *ei;
bd6daba9 2414 struct dentry *error = ERR_PTR(-ENOENT);
444ceed8 2415
61a28784 2416 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB
2417 if (!inode)
2418 goto out;
2419
2420 ei = PROC_I(inode);
2421 inode->i_mode = p->mode;
2422 if (S_ISDIR(inode->i_mode))
bfe86848 2423 set_nlink(inode, 2); /* Use getattr to fix if necessary */
444ceed8
EB
2424 if (p->iop)
2425 inode->i_op = p->iop;
2426 if (p->fop)
2427 inode->i_fop = p->fop;
2428 ei->op = p->op;
fb045adb 2429 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB
2430 d_add(dentry, inode);
2431 /* Close the race of the process dying before we return the dentry */
2432 if (pid_revalidate(dentry, NULL))
2433 error = NULL;
2434out:
2435 return error;
2436}
2437
1da177e4
LT
2438static struct dentry *proc_pident_lookup(struct inode *dir,
2439 struct dentry *dentry,
c5141e6d 2440 const struct pid_entry *ents,
7bcd6b0e 2441 unsigned int nents)
1da177e4 2442{
cd6a3ce9 2443 struct dentry *error;
99f89551 2444 struct task_struct *task = get_proc_task(dir);
c5141e6d 2445 const struct pid_entry *p, *last;
1da177e4 2446
cd6a3ce9 2447 error = ERR_PTR(-ENOENT);
1da177e4 2448
99f89551
EB
2449 if (!task)
2450 goto out_no_task;
1da177e4 2451
20cdc894
EB
2452 /*
2453 * Yes, it does not scale. And it should not. Don't add
2454 * new entries into /proc/<tgid>/ without very good reasons.
2455 */
7bcd6b0e
EB
2456 last = &ents[nents - 1];
2457 for (p = ents; p <= last; p++) {
1da177e4
LT
2458 if (p->len != dentry->d_name.len)
2459 continue;
2460 if (!memcmp(dentry->d_name.name, p->name, p->len))
2461 break;
2462 }
7bcd6b0e 2463 if (p > last)
1da177e4
LT
2464 goto out;
2465
444ceed8 2466 error = proc_pident_instantiate(dir, dentry, task, p);
1da177e4 2467out:
99f89551
EB
2468 put_task_struct(task);
2469out_no_task:
cd6a3ce9 2470 return error;
1da177e4
LT
2471}
2472
c5141e6d
ED
2473static int proc_pident_fill_cache(struct file *filp, void *dirent,
2474 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
61a28784
EB
2475{
2476 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2477 proc_pident_instantiate, task, p);
2478}
2479
28a6d671
EB
2480static int proc_pident_readdir(struct file *filp,
2481 void *dirent, filldir_t filldir,
c5141e6d 2482 const struct pid_entry *ents, unsigned int nents)
28a6d671
EB
2483{
2484 int i;
2fddfeef 2485 struct dentry *dentry = filp->f_path.dentry;
28a6d671
EB
2486 struct inode *inode = dentry->d_inode;
2487 struct task_struct *task = get_proc_task(inode);
c5141e6d 2488 const struct pid_entry *p, *last;
28a6d671
EB
2489 ino_t ino;
2490 int ret;
2491
2492 ret = -ENOENT;
2493 if (!task)
61a28784 2494 goto out_no_task;
28a6d671
EB
2495
2496 ret = 0;
28a6d671
EB
2497 i = filp->f_pos;
2498 switch (i) {
2499 case 0:
2500 ino = inode->i_ino;
2501 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2502 goto out;
2503 i++;
2504 filp->f_pos++;
2505 /* fall through */
2506 case 1:
2507 ino = parent_ino(dentry);
2508 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2509 goto out;
2510 i++;
2511 filp->f_pos++;
2512 /* fall through */
2513 default:
2514 i -= 2;
2515 if (i >= nents) {
2516 ret = 1;
2517 goto out;
2518 }
2519 p = ents + i;
7bcd6b0e
EB
2520 last = &ents[nents - 1];
2521 while (p <= last) {
61a28784 2522 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
28a6d671
EB
2523 goto out;
2524 filp->f_pos++;
2525 p++;
2526 }
2527 }
2528
2529 ret = 1;
2530out:
61a28784
EB
2531 put_task_struct(task);
2532out_no_task:
28a6d671 2533 return ret;
1da177e4
LT
2534}
2535
28a6d671
EB
2536#ifdef CONFIG_SECURITY
2537static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2538 size_t count, loff_t *ppos)
2539{
2fddfeef 2540 struct inode * inode = file->f_path.dentry->d_inode;
04ff9708 2541 char *p = NULL;
28a6d671
EB
2542 ssize_t length;
2543 struct task_struct *task = get_proc_task(inode);
2544
28a6d671 2545 if (!task)
04ff9708 2546 return -ESRCH;
28a6d671
EB
2547
2548 length = security_getprocattr(task,
2fddfeef 2549 (char*)file->f_path.dentry->d_name.name,
04ff9708 2550 &p);
28a6d671 2551 put_task_struct(task);
04ff9708
AV
2552 if (length > 0)
2553 length = simple_read_from_buffer(buf, count, ppos, p, length);
2554 kfree(p);
28a6d671 2555 return length;
1da177e4
LT
2556}
2557
28a6d671
EB
2558static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2559 size_t count, loff_t *ppos)
2560{
2fddfeef 2561 struct inode * inode = file->f_path.dentry->d_inode;
28a6d671
EB
2562 char *page;
2563 ssize_t length;
2564 struct task_struct *task = get_proc_task(inode);
2565
2566 length = -ESRCH;
2567 if (!task)
2568 goto out_no_task;
2569 if (count > PAGE_SIZE)
2570 count = PAGE_SIZE;
2571
2572 /* No partial writes. */
2573 length = -EINVAL;
2574 if (*ppos != 0)
2575 goto out;
2576
2577 length = -ENOMEM;
e12ba74d 2578 page = (char*)__get_free_page(GFP_TEMPORARY);
28a6d671
EB
2579 if (!page)
2580 goto out;
2581
2582 length = -EFAULT;
2583 if (copy_from_user(page, buf, count))
2584 goto out_free;
2585
107db7c7 2586 /* Guard against adverse ptrace interaction */
9b1bf12d 2587 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
107db7c7
DH
2588 if (length < 0)
2589 goto out_free;
2590
28a6d671 2591 length = security_setprocattr(task,
2fddfeef 2592 (char*)file->f_path.dentry->d_name.name,
28a6d671 2593 (void*)page, count);
9b1bf12d 2594 mutex_unlock(&task->signal->cred_guard_mutex);
28a6d671
EB
2595out_free:
2596 free_page((unsigned long) page);
2597out:
2598 put_task_struct(task);
2599out_no_task:
2600 return length;
2601}
2602
00977a59 2603static const struct file_operations proc_pid_attr_operations = {
28a6d671
EB
2604 .read = proc_pid_attr_read,
2605 .write = proc_pid_attr_write,
87df8424 2606 .llseek = generic_file_llseek,
28a6d671
EB
2607};
2608
c5141e6d 2609static const struct pid_entry attr_dir_stuff[] = {
631f9c18
AD
2610 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2611 REG("prev", S_IRUGO, proc_pid_attr_operations),
2612 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2613 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2614 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2615 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
28a6d671
EB
2616};
2617
72d9dcfc 2618static int proc_attr_dir_readdir(struct file * filp,
28a6d671
EB
2619 void * dirent, filldir_t filldir)
2620{
2621 return proc_pident_readdir(filp,dirent,filldir,
72d9dcfc 2622 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB
2623}
2624
00977a59 2625static const struct file_operations proc_attr_dir_operations = {
1da177e4 2626 .read = generic_read_dir,
72d9dcfc 2627 .readdir = proc_attr_dir_readdir,
6038f373 2628 .llseek = default_llseek,
1da177e4
LT
2629};
2630
72d9dcfc 2631static struct dentry *proc_attr_dir_lookup(struct inode *dir,
28a6d671
EB
2632 struct dentry *dentry, struct nameidata *nd)
2633{
7bcd6b0e
EB
2634 return proc_pident_lookup(dir, dentry,
2635 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
28a6d671
EB
2636}
2637
c5ef1c42 2638static const struct inode_operations proc_attr_dir_inode_operations = {
72d9dcfc 2639 .lookup = proc_attr_dir_lookup,
99f89551 2640 .getattr = pid_getattr,
6d76fa58 2641 .setattr = proc_setattr,
1da177e4
LT
2642};
2643
28a6d671
EB
2644#endif
2645
698ba7b5 2646#ifdef CONFIG_ELF_CORE
3cb4a0bb
KH
2647static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2648 size_t count, loff_t *ppos)
2649{
2650 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2651 struct mm_struct *mm;
2652 char buffer[PROC_NUMBUF];
2653 size_t len;
2654 int ret;
2655
2656 if (!task)
2657 return -ESRCH;
2658
2659 ret = 0;
2660 mm = get_task_mm(task);
2661 if (mm) {
2662 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2663 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2664 MMF_DUMP_FILTER_SHIFT));
2665 mmput(mm);
2666 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2667 }
2668
2669 put_task_struct(task);
2670
2671 return ret;
2672}
2673
2674static ssize_t proc_coredump_filter_write(struct file *file,
2675 const char __user *buf,
2676 size_t count,
2677 loff_t *ppos)
2678{
2679 struct task_struct *task;
2680 struct mm_struct *mm;
2681 char buffer[PROC_NUMBUF], *end;
2682 unsigned int val;
2683 int ret;
2684 int i;
2685 unsigned long mask;
2686
2687 ret = -EFAULT;
2688 memset(buffer, 0, sizeof(buffer));
2689 if (count > sizeof(buffer) - 1)
2690 count = sizeof(buffer) - 1;
2691 if (copy_from_user(buffer, buf, count))
2692 goto out_no_task;
2693
2694 ret = -EINVAL;
2695 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2696 if (*end == '\n')
2697 end++;
2698 if (end - buffer == 0)
2699 goto out_no_task;
2700
2701 ret = -ESRCH;
2702 task = get_proc_task(file->f_dentry->d_inode);
2703 if (!task)
2704 goto out_no_task;
2705
2706 ret = end - buffer;
2707 mm = get_task_mm(task);
2708 if (!mm)
2709 goto out_no_mm;
2710
2711 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2712 if (val & mask)
2713 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2714 else
2715 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2716 }
2717
2718 mmput(mm);
2719 out_no_mm:
2720 put_task_struct(task);
2721 out_no_task:
2722 return ret;
2723}
2724
2725static const struct file_operations proc_coredump_filter_operations = {
2726 .read = proc_coredump_filter_read,
2727 .write = proc_coredump_filter_write,
87df8424 2728 .llseek = generic_file_llseek,
3cb4a0bb
KH
2729};
2730#endif
2731
28a6d671
EB
2732/*
2733 * /proc/self:
2734 */
2735static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2736 int buflen)
2737{
488e5bc4 2738 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
b55fcb22 2739 pid_t tgid = task_tgid_nr_ns(current, ns);
28a6d671 2740 char tmp[PROC_NUMBUF];
b55fcb22 2741 if (!tgid)
488e5bc4 2742 return -ENOENT;
b55fcb22 2743 sprintf(tmp, "%d", tgid);
28a6d671
EB
2744 return vfs_readlink(dentry,buffer,buflen,tmp);
2745}
2746
2747static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2748{
488e5bc4 2749 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
b55fcb22 2750 pid_t tgid = task_tgid_nr_ns(current, ns);
7fee4868
AV
2751 char *name = ERR_PTR(-ENOENT);
2752 if (tgid) {
2753 name = __getname();
2754 if (!name)
2755 name = ERR_PTR(-ENOMEM);
2756 else
2757 sprintf(name, "%d", tgid);
2758 }
2759 nd_set_link(nd, name);
2760 return NULL;
2761}
2762
2763static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2764 void *cookie)
2765{
2766 char *s = nd_get_link(nd);
2767 if (!IS_ERR(s))
2768 __putname(s);
28a6d671
EB
2769}
2770
c5ef1c42 2771static const struct inode_operations proc_self_inode_operations = {
28a6d671
EB
2772 .readlink = proc_self_readlink,
2773 .follow_link = proc_self_follow_link,
7fee4868 2774 .put_link = proc_self_put_link,
28a6d671
EB
2775};
2776
801199ce
EB
2777/*
2778 * proc base
2779 *
2780 * These are the directory entries in the root directory of /proc
2781 * that properly belong to the /proc filesystem, as they describe
2782 * describe something that is process related.
2783 */
c5141e6d 2784static const struct pid_entry proc_base_stuff[] = {
61a28784 2785 NOD("self", S_IFLNK|S_IRWXUGO,
801199ce 2786 &proc_self_inode_operations, NULL, {}),
801199ce
EB
2787};
2788
444ceed8 2789static struct dentry *proc_base_instantiate(struct inode *dir,
c5141e6d 2790 struct dentry *dentry, struct task_struct *task, const void *ptr)
801199ce 2791{
c5141e6d 2792 const struct pid_entry *p = ptr;
801199ce 2793 struct inode *inode;
801199ce 2794 struct proc_inode *ei;
73d36460 2795 struct dentry *error;
801199ce
EB
2796
2797 /* Allocate the inode */
2798 error = ERR_PTR(-ENOMEM);
2799 inode = new_inode(dir->i_sb);
2800 if (!inode)
2801 goto out;
2802
2803 /* Initialize the inode */
2804 ei = PROC_I(inode);
85fe4025 2805 inode->i_ino = get_next_ino();
801199ce 2806 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
801199ce
EB
2807
2808 /*
2809 * grab the reference to the task.
2810 */
1a657f78 2811 ei->pid = get_task_pid(task, PIDTYPE_PID);
801199ce
EB
2812 if (!ei->pid)
2813 goto out_iput;
2814
801199ce
EB
2815 inode->i_mode = p->mode;
2816 if (S_ISDIR(inode->i_mode))
bfe86848 2817 set_nlink(inode, 2);
801199ce
EB
2818 if (S_ISLNK(inode->i_mode))
2819 inode->i_size = 64;
2820 if (p->iop)
2821 inode->i_op = p->iop;
2822 if (p->fop)
2823 inode->i_fop = p->fop;
2824 ei->op = p->op;
801199ce
EB
2825 d_add(dentry, inode);
2826 error = NULL;
2827out:
801199ce
EB
2828 return error;
2829out_iput:
2830 iput(inode);
2831 goto out;
2832}
2833
444ceed8
EB
2834static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2835{
2836 struct dentry *error;
2837 struct task_struct *task = get_proc_task(dir);
c5141e6d 2838 const struct pid_entry *p, *last;
444ceed8
EB
2839
2840 error = ERR_PTR(-ENOENT);
2841
2842 if (!task)
2843 goto out_no_task;
2844
2845 /* Lookup the directory entry */
7bcd6b0e
EB
2846 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2847 for (p = proc_base_stuff; p <= last; p++) {
444ceed8
EB
2848 if (p->len != dentry->d_name.len)
2849 continue;
2850 if (!memcmp(dentry->d_name.name, p->name, p->len))
2851 break;
2852 }
7bcd6b0e 2853 if (p > last)
444ceed8
EB
2854 goto out;
2855
2856 error = proc_base_instantiate(dir, dentry, task, p);
2857
2858out:
2859 put_task_struct(task);
2860out_no_task:
2861 return error;
2862}
2863
c5141e6d
ED
2864static int proc_base_fill_cache(struct file *filp, void *dirent,
2865 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
61a28784
EB
2866{
2867 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2868 proc_base_instantiate, task, p);
2869}
2870
aba76fdb 2871#ifdef CONFIG_TASK_IO_ACCOUNTING
297c5d92
AR
2872static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2873{
940389b8 2874 struct task_io_accounting acct = task->ioac;
5995477a 2875 unsigned long flags;
293eb1e7 2876 int result;
5995477a 2877
293eb1e7
VK
2878 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2879 if (result)
2880 return result;
2881
2882 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2883 result = -EACCES;
2884 goto out_unlock;
2885 }
1d1221f3 2886
5995477a
AR
2887 if (whole && lock_task_sighand(task, &flags)) {
2888 struct task_struct *t = task;
2889
2890 task_io_accounting_add(&acct, &task->signal->ioac);
2891 while_each_thread(task, t)
2892 task_io_accounting_add(&acct, &t->ioac);
2893
2894 unlock_task_sighand(task, &flags);
297c5d92 2895 }
293eb1e7 2896 result = sprintf(buffer,
aba76fdb
AM
2897 "rchar: %llu\n"
2898 "wchar: %llu\n"
2899 "syscr: %llu\n"
2900 "syscw: %llu\n"
2901 "read_bytes: %llu\n"
2902 "write_bytes: %llu\n"
2903 "cancelled_write_bytes: %llu\n",
7c44319d
AB
2904 (unsigned long long)acct.rchar,
2905 (unsigned long long)acct.wchar,
2906 (unsigned long long)acct.syscr,
2907 (unsigned long long)acct.syscw,
2908 (unsigned long long)acct.read_bytes,
2909 (unsigned long long)acct.write_bytes,
2910 (unsigned long long)acct.cancelled_write_bytes);
293eb1e7
VK
2911out_unlock:
2912 mutex_unlock(&task->signal->cred_guard_mutex);
2913 return result;
297c5d92
AR
2914}
2915
2916static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2917{
2918 return do_io_accounting(task, buffer, 0);
aba76fdb 2919}
297c5d92
AR
2920
2921static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2922{
2923 return do_io_accounting(task, buffer, 1);
2924}
2925#endif /* CONFIG_TASK_IO_ACCOUNTING */
aba76fdb 2926
22d917d8
EB
2927#ifdef CONFIG_USER_NS
2928static int proc_id_map_open(struct inode *inode, struct file *file,
2929 struct seq_operations *seq_ops)
2930{
2931 struct user_namespace *ns = NULL;
2932 struct task_struct *task;
2933 struct seq_file *seq;
2934 int ret = -EINVAL;
2935
2936 task = get_proc_task(inode);
2937 if (task) {
2938 rcu_read_lock();
2939 ns = get_user_ns(task_cred_xxx(task, user_ns));
2940 rcu_read_unlock();
2941 put_task_struct(task);
2942 }
2943 if (!ns)
2944 goto err;
2945
2946 ret = seq_open(file, seq_ops);
2947 if (ret)
2948 goto err_put_ns;
2949
2950 seq = file->private_data;
2951 seq->private = ns;
2952
2953 return 0;
2954err_put_ns:
2955 put_user_ns(ns);
2956err:
2957 return ret;
2958}
2959
2960static int proc_id_map_release(struct inode *inode, struct file *file)
2961{
2962 struct seq_file *seq = file->private_data;
2963 struct user_namespace *ns = seq->private;
2964 put_user_ns(ns);
2965 return seq_release(inode, file);
2966}
2967
2968static int proc_uid_map_open(struct inode *inode, struct file *file)
2969{
2970 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2971}
2972
2973static int proc_gid_map_open(struct inode *inode, struct file *file)
2974{
2975 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2976}
2977
2978static const struct file_operations proc_uid_map_operations = {
2979 .open = proc_uid_map_open,
2980 .write = proc_uid_map_write,
2981 .read = seq_read,
2982 .llseek = seq_lseek,
2983 .release = proc_id_map_release,
2984};
2985
2986static const struct file_operations proc_gid_map_operations = {
2987 .open = proc_gid_map_open,
2988 .write = proc_gid_map_write,
2989 .read = seq_read,
2990 .llseek = seq_lseek,
2991 .release = proc_id_map_release,
2992};
2993#endif /* CONFIG_USER_NS */
2994
47830723
KC
2995static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2996 struct pid *pid, struct task_struct *task)
2997{
a9712bc1
AV
2998 int err = lock_trace(task);
2999 if (!err) {
3000 seq_printf(m, "%08x\n", task->personality);
3001 unlock_trace(task);
3002 }
3003 return err;
47830723
KC
3004}
3005
28a6d671
EB
3006/*
3007 * Thread groups
3008 */
00977a59 3009static const struct file_operations proc_task_operations;
c5ef1c42 3010static const struct inode_operations proc_task_inode_operations;
20cdc894 3011
c5141e6d 3012static const struct pid_entry tgid_base_stuff[] = {
631f9c18
AD
3013 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
3014 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
640708a2
PE
3015#ifdef CONFIG_CHECKPOINT_RESTORE
3016 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
3017#endif
631f9c18 3018 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 3019 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
b2211a36 3020#ifdef CONFIG_NET
631f9c18 3021 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
b2211a36 3022#endif
631f9c18
AD
3023 REG("environ", S_IRUSR, proc_environ_operations),
3024 INF("auxv", S_IRUSR, proc_pid_auxv),
3025 ONE("status", S_IRUGO, proc_pid_status),
a9712bc1 3026 ONE("personality", S_IRUGO, proc_pid_personality),
3036e7b4 3027 INF("limits", S_IRUGO, proc_pid_limits),
43ae34cb 3028#ifdef CONFIG_SCHED_DEBUG
631f9c18 3029 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
5091faa4
MG
3030#endif
3031#ifdef CONFIG_SCHED_AUTOGROUP
3032 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
ebcb6734 3033#endif
4614a696 3034 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 3035#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
a9712bc1 3036 INF("syscall", S_IRUGO, proc_pid_syscall),
43ae34cb 3037#endif
631f9c18
AD
3038 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3039 ONE("stat", S_IRUGO, proc_tgid_stat),
3040 ONE("statm", S_IRUGO, proc_pid_statm),
b7643757 3041 REG("maps", S_IRUGO, proc_pid_maps_operations),
28a6d671 3042#ifdef CONFIG_NUMA
b7643757 3043 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
28a6d671 3044#endif
631f9c18
AD
3045 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3046 LNK("cwd", proc_cwd_link),
3047 LNK("root", proc_root_link),
3048 LNK("exe", proc_exe_link),
3049 REG("mounts", S_IRUGO, proc_mounts_operations),
3050 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3051 REG("mountstats", S_IRUSR, proc_mountstats_operations),
1e883281 3052#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18 3053 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
b7643757 3054 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
ca6b0bf0 3055 REG("pagemap", S_IRUGO, proc_pagemap_operations),
28a6d671
EB
3056#endif
3057#ifdef CONFIG_SECURITY
631f9c18 3058 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB
3059#endif
3060#ifdef CONFIG_KALLSYMS
631f9c18 3061 INF("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 3062#endif
2ec220e2 3063#ifdef CONFIG_STACKTRACE
a9712bc1 3064 ONE("stack", S_IRUGO, proc_pid_stack),
28a6d671
EB
3065#endif
3066#ifdef CONFIG_SCHEDSTATS
631f9c18 3067 INF("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 3068#endif
9745512c 3069#ifdef CONFIG_LATENCYTOP
631f9c18 3070 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 3071#endif
8793d854 3072#ifdef CONFIG_PROC_PID_CPUSET
631f9c18 3073 REG("cpuset", S_IRUGO, proc_cpuset_operations),
a424316c
PM
3074#endif
3075#ifdef CONFIG_CGROUPS
631f9c18 3076 REG("cgroup", S_IRUGO, proc_cgroup_operations),
28a6d671 3077#endif
631f9c18
AD
3078 INF("oom_score", S_IRUGO, proc_oom_score),
3079 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
a63d83f4 3080 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 3081#ifdef CONFIG_AUDITSYSCALL
631f9c18
AD
3082 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3083 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 3084#endif
f4f154fd 3085#ifdef CONFIG_FAULT_INJECTION
631f9c18 3086 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 3087#endif
698ba7b5 3088#ifdef CONFIG_ELF_CORE
631f9c18 3089 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3cb4a0bb 3090#endif
aba76fdb 3091#ifdef CONFIG_TASK_IO_ACCOUNTING
1d1221f3 3092 INF("io", S_IRUSR, proc_tgid_io_accounting),
aba76fdb 3093#endif
f133ecca
CM
3094#ifdef CONFIG_HARDWALL
3095 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3096#endif
22d917d8
EB
3097#ifdef CONFIG_USER_NS
3098 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3099 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3100#endif
28a6d671 3101};
1da177e4 3102
28a6d671 3103static int proc_tgid_base_readdir(struct file * filp,
1da177e4
LT
3104 void * dirent, filldir_t filldir)
3105{
3106 return proc_pident_readdir(filp,dirent,filldir,
28a6d671 3107 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT
3108}
3109
00977a59 3110static const struct file_operations proc_tgid_base_operations = {
1da177e4 3111 .read = generic_read_dir,
28a6d671 3112 .readdir = proc_tgid_base_readdir,
6038f373 3113 .llseek = default_llseek,
1da177e4
LT
3114};
3115
28a6d671 3116static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
7bcd6b0e
EB
3117 return proc_pident_lookup(dir, dentry,
3118 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1da177e4
LT
3119}
3120
c5ef1c42 3121static const struct inode_operations proc_tgid_base_inode_operations = {
28a6d671 3122 .lookup = proc_tgid_base_lookup,
99f89551 3123 .getattr = pid_getattr,
6d76fa58 3124 .setattr = proc_setattr,
0499680a 3125 .permission = proc_pid_permission,
1da177e4 3126};
1da177e4 3127
60347f67 3128static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
1da177e4 3129{
48e6484d 3130 struct dentry *dentry, *leader, *dir;
8578cea7 3131 char buf[PROC_NUMBUF];
48e6484d
EB
3132 struct qstr name;
3133
3134 name.name = buf;
60347f67
PE
3135 name.len = snprintf(buf, sizeof(buf), "%d", pid);
3136 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d 3137 if (dentry) {
29f12ca3 3138 shrink_dcache_parent(dentry);
48e6484d
EB
3139 d_drop(dentry);
3140 dput(dentry);
3141 }
1da177e4 3142
48e6484d 3143 name.name = buf;
60347f67
PE
3144 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
3145 leader = d_hash_and_lookup(mnt->mnt_root, &name);
48e6484d
EB
3146 if (!leader)
3147 goto out;
1da177e4 3148
48e6484d
EB
3149 name.name = "task";
3150 name.len = strlen(name.name);
3151 dir = d_hash_and_lookup(leader, &name);
3152 if (!dir)
3153 goto out_put_leader;
3154
3155 name.name = buf;
60347f67 3156 name.len = snprintf(buf, sizeof(buf), "%d", pid);
48e6484d
EB
3157 dentry = d_hash_and_lookup(dir, &name);
3158 if (dentry) {
3159 shrink_dcache_parent(dentry);
3160 d_drop(dentry);
3161 dput(dentry);
1da177e4 3162 }
48e6484d
EB
3163
3164 dput(dir);
3165out_put_leader:
3166 dput(leader);
3167out:
3168 return;
1da177e4
LT
3169}
3170
0895e91d
RD
3171/**
3172 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3173 * @task: task that should be flushed.
3174 *
3175 * When flushing dentries from proc, one needs to flush them from global
60347f67 3176 * proc (proc_mnt) and from all the namespaces' procs this task was seen
0895e91d
RD
3177 * in. This call is supposed to do all of this job.
3178 *
3179 * Looks in the dcache for
3180 * /proc/@pid
3181 * /proc/@tgid/task/@pid
3182 * if either directory is present flushes it and all of it'ts children
3183 * from the dcache.
3184 *
3185 * It is safe and reasonable to cache /proc entries for a task until
3186 * that task exits. After that they just clog up the dcache with
3187 * useless entries, possibly causing useful dcache entries to be
3188 * flushed instead. This routine is proved to flush those useless
3189 * dcache entries at process exit time.
3190 *
3191 * NOTE: This routine is just an optimization so it does not guarantee
3192 * that no dcache entries will exist at process exit time it
3193 * just makes it very unlikely that any will persist.
60347f67
PE
3194 */
3195
3196void proc_flush_task(struct task_struct *task)
3197{
9fcc2d15 3198 int i;
9b4d1cbe 3199 struct pid *pid, *tgid;
130f77ec
PE
3200 struct upid *upid;
3201
130f77ec 3202 pid = task_pid(task);
9b4d1cbe 3203 tgid = task_tgid(task);
130f77ec 3204
9fcc2d15 3205 for (i = 0; i <= pid->level; i++) {
130f77ec
PE
3206 upid = &pid->numbers[i];
3207 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
9b4d1cbe 3208 tgid->numbers[i].nr);
130f77ec 3209 }
6f4e6433
PE
3210
3211 upid = &pid->numbers[pid->level];
3212 if (upid->nr == 1)
3213 pid_ns_release_proc(upid->ns);
60347f67
PE
3214}
3215
9711ef99
AB
3216static struct dentry *proc_pid_instantiate(struct inode *dir,
3217 struct dentry * dentry,
c5141e6d 3218 struct task_struct *task, const void *ptr)
444ceed8
EB
3219{
3220 struct dentry *error = ERR_PTR(-ENOENT);
3221 struct inode *inode;
3222
61a28784 3223 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB
3224 if (!inode)
3225 goto out;
3226
3227 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3228 inode->i_op = &proc_tgid_base_inode_operations;
3229 inode->i_fop = &proc_tgid_base_operations;
3230 inode->i_flags|=S_IMMUTABLE;
aed54175 3231
bfe86848
MS
3232 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
3233 ARRAY_SIZE(tgid_base_stuff)));
444ceed8 3234
fb045adb 3235 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB
3236
3237 d_add(dentry, inode);
3238 /* Close the race of the process dying before we return the dentry */
3239 if (pid_revalidate(dentry, NULL))
3240 error = NULL;
3241out:
3242 return error;
3243}
3244
1da177e4
LT
3245struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3246{
73d36460 3247 struct dentry *result;
1da177e4 3248 struct task_struct *task;
1da177e4 3249 unsigned tgid;
b488893a 3250 struct pid_namespace *ns;
1da177e4 3251
801199ce
EB
3252 result = proc_base_lookup(dir, dentry);
3253 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3254 goto out;
3255
1da177e4
LT
3256 tgid = name_to_int(dentry);
3257 if (tgid == ~0U)
3258 goto out;
3259
b488893a 3260 ns = dentry->d_sb->s_fs_info;
de758734 3261 rcu_read_lock();
b488893a 3262 task = find_task_by_pid_ns(tgid, ns);
1da177e4
LT
3263 if (task)
3264 get_task_struct(task);
de758734 3265 rcu_read_unlock();
1da177e4
LT
3266 if (!task)
3267 goto out;
3268
444ceed8 3269 result = proc_pid_instantiate(dir, dentry, task, NULL);
1da177e4 3270 put_task_struct(task);
1da177e4 3271out:
cd6a3ce9 3272 return result;
1da177e4
LT
3273}
3274
1da177e4 3275/*
0804ef4b 3276 * Find the first task with tgid >= tgid
0bc58a91 3277 *
1da177e4 3278 */
19fd4bb2
EB
3279struct tgid_iter {
3280 unsigned int tgid;
0804ef4b 3281 struct task_struct *task;
19fd4bb2
EB
3282};
3283static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3284{
0804ef4b 3285 struct pid *pid;
1da177e4 3286
19fd4bb2
EB
3287 if (iter.task)
3288 put_task_struct(iter.task);
454cc105 3289 rcu_read_lock();
0804ef4b 3290retry:
19fd4bb2
EB
3291 iter.task = NULL;
3292 pid = find_ge_pid(iter.tgid, ns);
0804ef4b 3293 if (pid) {
19fd4bb2
EB
3294 iter.tgid = pid_nr_ns(pid, ns);
3295 iter.task = pid_task(pid, PIDTYPE_PID);
0804ef4b
EB
3296 /* What we to know is if the pid we have find is the
3297 * pid of a thread_group_leader. Testing for task
3298 * being a thread_group_leader is the obvious thing
3299 * todo but there is a window when it fails, due to
3300 * the pid transfer logic in de_thread.
3301 *
3302 * So we perform the straight forward test of seeing
3303 * if the pid we have found is the pid of a thread
3304 * group leader, and don't worry if the task we have
3305 * found doesn't happen to be a thread group leader.
3306 * As we don't care in the case of readdir.
3307 */
19fd4bb2
EB
3308 if (!iter.task || !has_group_leader_pid(iter.task)) {
3309 iter.tgid += 1;
0804ef4b 3310 goto retry;
19fd4bb2
EB
3311 }
3312 get_task_struct(iter.task);
0bc58a91 3313 }
454cc105 3314 rcu_read_unlock();
19fd4bb2 3315 return iter;
1da177e4
LT
3316}
3317
7bcd6b0e 3318#define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
0804ef4b 3319
61a28784 3320static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
19fd4bb2 3321 struct tgid_iter iter)
61a28784
EB
3322{
3323 char name[PROC_NUMBUF];
19fd4bb2 3324 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
61a28784 3325 return proc_fill_cache(filp, dirent, filldir, name, len,
19fd4bb2 3326 proc_pid_instantiate, iter.task, NULL);
61a28784
EB
3327}
3328
0499680a
VK
3329static int fake_filldir(void *buf, const char *name, int namelen,
3330 loff_t offset, u64 ino, unsigned d_type)
3331{
3332 return 0;
3333}
3334
1da177e4
LT
3335/* for the /proc/ directory itself, after non-process stuff has been done */
3336int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3337{
d8bdc59f
LT
3338 unsigned int nr;
3339 struct task_struct *reaper;
19fd4bb2 3340 struct tgid_iter iter;
b488893a 3341 struct pid_namespace *ns;
0499680a 3342 filldir_t __filldir;
1da177e4 3343
d8bdc59f
LT
3344 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3345 goto out_no_task;
3346 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3347
3348 reaper = get_proc_task(filp->f_path.dentry->d_inode);
61a28784
EB
3349 if (!reaper)
3350 goto out_no_task;
3351
7bcd6b0e 3352 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
c5141e6d 3353 const struct pid_entry *p = &proc_base_stuff[nr];
61a28784 3354 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
801199ce 3355 goto out;
1da177e4
LT
3356 }
3357
b488893a 3358 ns = filp->f_dentry->d_sb->s_fs_info;
19fd4bb2
EB
3359 iter.task = NULL;
3360 iter.tgid = filp->f_pos - TGID_OFFSET;
3361 for (iter = next_tgid(ns, iter);
3362 iter.task;
3363 iter.tgid += 1, iter = next_tgid(ns, iter)) {
0499680a
VK
3364 if (has_pid_permissions(ns, iter.task, 2))
3365 __filldir = filldir;
3366 else
3367 __filldir = fake_filldir;
3368
19fd4bb2 3369 filp->f_pos = iter.tgid + TGID_OFFSET;
0499680a 3370 if (proc_pid_fill_cache(filp, dirent, __filldir, iter) < 0) {
19fd4bb2 3371 put_task_struct(iter.task);
0804ef4b 3372 goto out;
1da177e4 3373 }
0bc58a91 3374 }
0804ef4b
EB
3375 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3376out:
61a28784
EB
3377 put_task_struct(reaper);
3378out_no_task:
0bc58a91
EB
3379 return 0;
3380}
1da177e4 3381
28a6d671
EB
3382/*
3383 * Tasks
3384 */
c5141e6d 3385static const struct pid_entry tid_base_stuff[] = {
631f9c18 3386 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3835541d 3387 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
6b4e306a 3388 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
631f9c18
AD
3389 REG("environ", S_IRUSR, proc_environ_operations),
3390 INF("auxv", S_IRUSR, proc_pid_auxv),
3391 ONE("status", S_IRUGO, proc_pid_status),
a9712bc1 3392 ONE("personality", S_IRUGO, proc_pid_personality),
3036e7b4 3393 INF("limits", S_IRUGO, proc_pid_limits),
43ae34cb 3394#ifdef CONFIG_SCHED_DEBUG
631f9c18 3395 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
ebcb6734 3396#endif
4614a696 3397 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
ebcb6734 3398#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
a9712bc1 3399 INF("syscall", S_IRUGO, proc_pid_syscall),
43ae34cb 3400#endif
631f9c18
AD
3401 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3402 ONE("stat", S_IRUGO, proc_tid_stat),
3403 ONE("statm", S_IRUGO, proc_pid_statm),
b7643757 3404 REG("maps", S_IRUGO, proc_tid_maps_operations),
28a6d671 3405#ifdef CONFIG_NUMA
b7643757 3406 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
28a6d671 3407#endif
631f9c18
AD
3408 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3409 LNK("cwd", proc_cwd_link),
3410 LNK("root", proc_root_link),
3411 LNK("exe", proc_exe_link),
3412 REG("mounts", S_IRUGO, proc_mounts_operations),
3413 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
1e883281 3414#ifdef CONFIG_PROC_PAGE_MONITOR
631f9c18 3415 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
b7643757 3416 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
ca6b0bf0 3417 REG("pagemap", S_IRUGO, proc_pagemap_operations),
28a6d671
EB
3418#endif
3419#ifdef CONFIG_SECURITY
631f9c18 3420 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
28a6d671
EB
3421#endif
3422#ifdef CONFIG_KALLSYMS
631f9c18 3423 INF("wchan", S_IRUGO, proc_pid_wchan),
28a6d671 3424#endif
2ec220e2 3425#ifdef CONFIG_STACKTRACE
a9712bc1 3426 ONE("stack", S_IRUGO, proc_pid_stack),
28a6d671
EB
3427#endif
3428#ifdef CONFIG_SCHEDSTATS
631f9c18 3429 INF("schedstat", S_IRUGO, proc_pid_schedstat),
28a6d671 3430#endif
9745512c 3431#ifdef CONFIG_LATENCYTOP
631f9c18 3432 REG("latency", S_IRUGO, proc_lstats_operations),
9745512c 3433#endif
8793d854 3434#ifdef CONFIG_PROC_PID_CPUSET
631f9c18 3435 REG("cpuset", S_IRUGO, proc_cpuset_operations),
a424316c
PM
3436#endif
3437#ifdef CONFIG_CGROUPS
631f9c18 3438 REG("cgroup", S_IRUGO, proc_cgroup_operations),
28a6d671 3439#endif
631f9c18
AD
3440 INF("oom_score", S_IRUGO, proc_oom_score),
3441 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
a63d83f4 3442 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
28a6d671 3443#ifdef CONFIG_AUDITSYSCALL
631f9c18 3444 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
26ec3c64 3445 REG("sessionid", S_IRUGO, proc_sessionid_operations),
28a6d671 3446#endif
f4f154fd 3447#ifdef CONFIG_FAULT_INJECTION
631f9c18 3448 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
f4f154fd 3449#endif
297c5d92 3450#ifdef CONFIG_TASK_IO_ACCOUNTING
1d1221f3 3451 INF("io", S_IRUSR, proc_tid_io_accounting),
297c5d92 3452#endif
f133ecca
CM
3453#ifdef CONFIG_HARDWALL
3454 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3455#endif
22d917d8
EB
3456#ifdef CONFIG_USER_NS
3457 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3458 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3459#endif
28a6d671
EB
3460};
3461
3462static int proc_tid_base_readdir(struct file * filp,
3463 void * dirent, filldir_t filldir)
3464{
3465 return proc_pident_readdir(filp,dirent,filldir,
3466 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3467}
3468
3469static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
7bcd6b0e
EB
3470 return proc_pident_lookup(dir, dentry,
3471 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
28a6d671
EB
3472}
3473
00977a59 3474static const struct file_operations proc_tid_base_operations = {
28a6d671
EB
3475 .read = generic_read_dir,
3476 .readdir = proc_tid_base_readdir,
6038f373 3477 .llseek = default_llseek,
28a6d671
EB
3478};
3479
c5ef1c42 3480static const struct inode_operations proc_tid_base_inode_operations = {
28a6d671
EB
3481 .lookup = proc_tid_base_lookup,
3482 .getattr = pid_getattr,
3483 .setattr = proc_setattr,
3484};
3485
444ceed8 3486static struct dentry *proc_task_instantiate(struct inode *dir,
c5141e6d 3487 struct dentry *dentry, struct task_struct *task, const void *ptr)
444ceed8
EB
3488{
3489 struct dentry *error = ERR_PTR(-ENOENT);
3490 struct inode *inode;
61a28784 3491 inode = proc_pid_make_inode(dir->i_sb, task);
444ceed8
EB
3492
3493 if (!inode)
3494 goto out;
3495 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3496 inode->i_op = &proc_tid_base_inode_operations;
3497 inode->i_fop = &proc_tid_base_operations;
3498 inode->i_flags|=S_IMMUTABLE;
aed54175 3499
bfe86848
MS
3500 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3501 ARRAY_SIZE(tid_base_stuff)));
444ceed8 3502
fb045adb 3503 d_set_d_op(dentry, &pid_dentry_operations);
444ceed8
EB
3504
3505 d_add(dentry, inode);
3506 /* Close the race of the process dying before we return the dentry */
3507 if (pid_revalidate(dentry, NULL))
3508 error = NULL;
3509out:
3510 return error;
3511}
3512
28a6d671
EB
3513static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3514{
3515 struct dentry *result = ERR_PTR(-ENOENT);
3516 struct task_struct *task;
3517 struct task_struct *leader = get_proc_task(dir);
28a6d671 3518 unsigned tid;
b488893a 3519 struct pid_namespace *ns;
28a6d671
EB
3520
3521 if (!leader)
3522 goto out_no_task;
3523
3524 tid = name_to_int(dentry);
3525 if (tid == ~0U)
3526 goto out;
3527
b488893a 3528 ns = dentry->d_sb->s_fs_info;
28a6d671 3529 rcu_read_lock();
b488893a 3530 task = find_task_by_pid_ns(tid, ns);
28a6d671
EB
3531 if (task)
3532 get_task_struct(task);
3533 rcu_read_unlock();
3534 if (!task)
3535 goto out;
bac0abd6 3536 if (!same_thread_group(leader, task))
28a6d671
EB
3537 goto out_drop_task;
3538
444ceed8 3539 result = proc_task_instantiate(dir, dentry, task, NULL);
28a6d671
EB
3540out_drop_task:
3541 put_task_struct(task);
3542out:
3543 put_task_struct(leader);
3544out_no_task:
3545 return result;
3546}
3547
0bc58a91
EB
3548/*
3549 * Find the first tid of a thread group to return to user space.
3550 *
3551 * Usually this is just the thread group leader, but if the users
3552 * buffer was too small or there was a seek into the middle of the
3553 * directory we have more work todo.
3554 *
3555 * In the case of a short read we start with find_task_by_pid.
3556 *
3557 * In the case of a seek we start with the leader and walk nr
3558 * threads past it.
3559 */
cc288738 3560static struct task_struct *first_tid(struct task_struct *leader,
b488893a 3561 int tid, int nr, struct pid_namespace *ns)
0bc58a91 3562{
a872ff0c 3563 struct task_struct *pos;
1da177e4 3564
cc288738 3565 rcu_read_lock();
0bc58a91
EB
3566 /* Attempt to start with the pid of a thread */
3567 if (tid && (nr > 0)) {
b488893a 3568 pos = find_task_by_pid_ns(tid, ns);
a872ff0c
ON
3569 if (pos && (pos->group_leader == leader))
3570 goto found;
0bc58a91 3571 }
1da177e4 3572
0bc58a91 3573 /* If nr exceeds the number of threads there is nothing todo */
a872ff0c
ON
3574 pos = NULL;
3575 if (nr && nr >= get_nr_threads(leader))
3576 goto out;
1da177e4 3577
a872ff0c
ON
3578 /* If we haven't found our starting place yet start
3579 * with the leader and walk nr threads forward.
0bc58a91 3580 */
a872ff0c
ON
3581 for (pos = leader; nr > 0; --nr) {
3582 pos = next_thread(pos);
3583 if (pos == leader) {
3584 pos = NULL;
3585 goto out;
3586 }
1da177e4 3587 }
a872ff0c
ON
3588found:
3589 get_task_struct(pos);
3590out:
cc288738 3591 rcu_read_unlock();
0bc58a91
EB
3592 return pos;
3593}
3594
3595/*
3596 * Find the next thread in the thread list.
3597 * Return NULL if there is an error or no next thread.
3598 *
3599 * The reference to the input task_struct is released.
3600 */
3601static struct task_struct *next_tid(struct task_struct *start)
3602{
c1df7fb8 3603 struct task_struct *pos = NULL;
cc288738 3604 rcu_read_lock();
c1df7fb8 3605 if (pid_alive(start)) {
0bc58a91 3606 pos = next_thread(start);
c1df7fb8
ON
3607 if (thread_group_leader(pos))
3608 pos = NULL;
3609 else
3610 get_task_struct(pos);
3611 }
cc288738 3612 rcu_read_unlock();
0bc58a91
EB
3613 put_task_struct(start);
3614 return pos;
1da177e4
LT
3615}
3616
61a28784
EB
3617static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3618 struct task_struct *task, int tid)
3619{
3620 char name[PROC_NUMBUF];
3621 int len = snprintf(name, sizeof(name), "%d", tid);
3622 return proc_fill_cache(filp, dirent, filldir, name, len,
3623 proc_task_instantiate, task, NULL);
3624}
3625
1da177e4
LT
3626/* for the /proc/TGID/task/ directories */
3627static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3628{
2fddfeef 3629 struct dentry *dentry = filp->f_path.dentry;
1da177e4 3630 struct inode *inode = dentry->d_inode;
7d895244 3631 struct task_struct *leader = NULL;
0bc58a91 3632 struct task_struct *task;
1da177e4
LT
3633 int retval = -ENOENT;
3634 ino_t ino;
0bc58a91 3635 int tid;
b488893a 3636 struct pid_namespace *ns;
1da177e4 3637
7d895244
GC
3638 task = get_proc_task(inode);
3639 if (!task)
3640 goto out_no_task;
3641 rcu_read_lock();
3642 if (pid_alive(task)) {
3643 leader = task->group_leader;
3644 get_task_struct(leader);
3645 }
3646 rcu_read_unlock();
3647 put_task_struct(task);
99f89551
EB
3648 if (!leader)
3649 goto out_no_task;
1da177e4
LT
3650 retval = 0;
3651
ee568b25 3652 switch ((unsigned long)filp->f_pos) {
1da177e4
LT
3653 case 0:
3654 ino = inode->i_ino;
ee6f779b 3655 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
1da177e4 3656 goto out;
ee6f779b 3657 filp->f_pos++;
1da177e4
LT
3658 /* fall through */
3659 case 1:
3660 ino = parent_ino(dentry);
ee6f779b 3661 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
1da177e4 3662 goto out;
ee6f779b 3663 filp->f_pos++;
1da177e4
LT
3664 /* fall through */
3665 }
3666
0bc58a91
EB
3667 /* f_version caches the tgid value that the last readdir call couldn't
3668 * return. lseek aka telldir automagically resets f_version to 0.
3669 */
b488893a 3670 ns = filp->f_dentry->d_sb->s_fs_info;
2b47c361 3671 tid = (int)filp->f_version;
0bc58a91 3672 filp->f_version = 0;
ee6f779b 3673 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
0bc58a91 3674 task;
ee6f779b 3675 task = next_tid(task), filp->f_pos++) {
b488893a 3676 tid = task_pid_nr_ns(task, ns);
61a28784 3677 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
0bc58a91
EB
3678 /* returning this tgid failed, save it as the first
3679 * pid for the next readir call */
2b47c361 3680 filp->f_version = (u64)tid;
0bc58a91 3681 put_task_struct(task);
1da177e4 3682 break;
0bc58a91 3683 }
1da177e4
LT
3684 }
3685out:
99f89551
EB
3686 put_task_struct(leader);
3687out_no_task:
1da177e4
LT
3688 return retval;
3689}
6e66b52b
EB
3690
3691static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3692{
3693 struct inode *inode = dentry->d_inode;
99f89551 3694 struct task_struct *p = get_proc_task(inode);
6e66b52b
EB
3695 generic_fillattr(inode, stat);
3696
99f89551 3697 if (p) {
99f89551 3698 stat->nlink += get_nr_threads(p);
99f89551 3699 put_task_struct(p);
6e66b52b
EB
3700 }
3701
3702 return 0;
3703}
28a6d671 3704
c5ef1c42 3705static const struct inode_operations proc_task_inode_operations = {
28a6d671
EB
3706 .lookup = proc_task_lookup,
3707 .getattr = proc_task_getattr,
3708 .setattr = proc_setattr,
0499680a 3709 .permission = proc_pid_permission,
28a6d671
EB
3710};
3711
00977a59 3712static const struct file_operations proc_task_operations = {
28a6d671
EB
3713 .read = generic_read_dir,
3714 .readdir = proc_task_readdir,
6038f373 3715 .llseek = default_llseek,
28a6d671 3716};