blob: dad8ea4e00a03efae08302c41e8951eb84eb811e [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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.
14 */
15
16#include <asm/uaccess.h>
17
18#include <linux/config.h>
19#include <linux/errno.h>
20#include <linux/time.h>
21#include <linux/proc_fs.h>
22#include <linux/stat.h>
23#include <linux/init.h>
24#include <linux/file.h>
25#include <linux/string.h>
26#include <linux/seq_file.h>
27#include <linux/namei.h>
28#include <linux/namespace.h>
29#include <linux/mm.h>
30#include <linux/smp_lock.h>
31#include <linux/kallsyms.h>
32#include <linux/mount.h>
33#include <linux/security.h>
34#include <linux/ptrace.h>
35#include <linux/seccomp.h>
36#include <linux/cpuset.h>
37#include <linux/audit.h>
38#include "internal.h"
39
40/*
41 * For hysterical raisins we keep the same inumbers as in the old procfs.
42 * Feel free to change the macro below - just keep the range distinct from
43 * inumbers of the rest of procfs (currently those are in 0x0000--0xffff).
44 * As soon as we'll get a separate superblock we will be able to forget
45 * about magical ranges too.
46 */
47
48#define fake_ino(pid,ino) (((pid)<<16)|(ino))
49
50enum pid_directory_inos {
51 PROC_TGID_INO = 2,
52 PROC_TGID_TASK,
53 PROC_TGID_STATUS,
54 PROC_TGID_MEM,
55#ifdef CONFIG_SECCOMP
56 PROC_TGID_SECCOMP,
57#endif
58 PROC_TGID_CWD,
59 PROC_TGID_ROOT,
60 PROC_TGID_EXE,
61 PROC_TGID_FD,
62 PROC_TGID_ENVIRON,
63 PROC_TGID_AUXV,
64 PROC_TGID_CMDLINE,
65 PROC_TGID_STAT,
66 PROC_TGID_STATM,
67 PROC_TGID_MAPS,
68 PROC_TGID_MOUNTS,
69 PROC_TGID_WCHAN,
70#ifdef CONFIG_SCHEDSTATS
71 PROC_TGID_SCHEDSTAT,
72#endif
73#ifdef CONFIG_CPUSETS
74 PROC_TGID_CPUSET,
75#endif
76#ifdef CONFIG_SECURITY
77 PROC_TGID_ATTR,
78 PROC_TGID_ATTR_CURRENT,
79 PROC_TGID_ATTR_PREV,
80 PROC_TGID_ATTR_EXEC,
81 PROC_TGID_ATTR_FSCREATE,
82#endif
83#ifdef CONFIG_AUDITSYSCALL
84 PROC_TGID_LOGINUID,
85#endif
86 PROC_TGID_FD_DIR,
87 PROC_TGID_OOM_SCORE,
88 PROC_TGID_OOM_ADJUST,
89 PROC_TID_INO,
90 PROC_TID_STATUS,
91 PROC_TID_MEM,
92#ifdef CONFIG_SECCOMP
93 PROC_TID_SECCOMP,
94#endif
95 PROC_TID_CWD,
96 PROC_TID_ROOT,
97 PROC_TID_EXE,
98 PROC_TID_FD,
99 PROC_TID_ENVIRON,
100 PROC_TID_AUXV,
101 PROC_TID_CMDLINE,
102 PROC_TID_STAT,
103 PROC_TID_STATM,
104 PROC_TID_MAPS,
105 PROC_TID_MOUNTS,
106 PROC_TID_WCHAN,
107#ifdef CONFIG_SCHEDSTATS
108 PROC_TID_SCHEDSTAT,
109#endif
110#ifdef CONFIG_CPUSETS
111 PROC_TID_CPUSET,
112#endif
113#ifdef CONFIG_SECURITY
114 PROC_TID_ATTR,
115 PROC_TID_ATTR_CURRENT,
116 PROC_TID_ATTR_PREV,
117 PROC_TID_ATTR_EXEC,
118 PROC_TID_ATTR_FSCREATE,
119#endif
120#ifdef CONFIG_AUDITSYSCALL
121 PROC_TID_LOGINUID,
122#endif
123 PROC_TID_FD_DIR = 0x8000, /* 0x8000-0xffff */
124 PROC_TID_OOM_SCORE,
125 PROC_TID_OOM_ADJUST,
126};
127
128struct pid_entry {
129 int type;
130 int len;
131 char *name;
132 mode_t mode;
133};
134
135#define E(type,name,mode) {(type),sizeof(name)-1,(name),(mode)}
136
137static struct pid_entry tgid_base_stuff[] = {
138 E(PROC_TGID_TASK, "task", S_IFDIR|S_IRUGO|S_IXUGO),
139 E(PROC_TGID_FD, "fd", S_IFDIR|S_IRUSR|S_IXUSR),
140 E(PROC_TGID_ENVIRON, "environ", S_IFREG|S_IRUSR),
141 E(PROC_TGID_AUXV, "auxv", S_IFREG|S_IRUSR),
142 E(PROC_TGID_STATUS, "status", S_IFREG|S_IRUGO),
143 E(PROC_TGID_CMDLINE, "cmdline", S_IFREG|S_IRUGO),
144 E(PROC_TGID_STAT, "stat", S_IFREG|S_IRUGO),
145 E(PROC_TGID_STATM, "statm", S_IFREG|S_IRUGO),
146 E(PROC_TGID_MAPS, "maps", S_IFREG|S_IRUGO),
147 E(PROC_TGID_MEM, "mem", S_IFREG|S_IRUSR|S_IWUSR),
148#ifdef CONFIG_SECCOMP
149 E(PROC_TGID_SECCOMP, "seccomp", S_IFREG|S_IRUSR|S_IWUSR),
150#endif
151 E(PROC_TGID_CWD, "cwd", S_IFLNK|S_IRWXUGO),
152 E(PROC_TGID_ROOT, "root", S_IFLNK|S_IRWXUGO),
153 E(PROC_TGID_EXE, "exe", S_IFLNK|S_IRWXUGO),
154 E(PROC_TGID_MOUNTS, "mounts", S_IFREG|S_IRUGO),
155#ifdef CONFIG_SECURITY
156 E(PROC_TGID_ATTR, "attr", S_IFDIR|S_IRUGO|S_IXUGO),
157#endif
158#ifdef CONFIG_KALLSYMS
159 E(PROC_TGID_WCHAN, "wchan", S_IFREG|S_IRUGO),
160#endif
161#ifdef CONFIG_SCHEDSTATS
162 E(PROC_TGID_SCHEDSTAT, "schedstat", S_IFREG|S_IRUGO),
163#endif
164#ifdef CONFIG_CPUSETS
165 E(PROC_TGID_CPUSET, "cpuset", S_IFREG|S_IRUGO),
166#endif
167 E(PROC_TGID_OOM_SCORE, "oom_score",S_IFREG|S_IRUGO),
168 E(PROC_TGID_OOM_ADJUST,"oom_adj", S_IFREG|S_IRUGO|S_IWUSR),
169#ifdef CONFIG_AUDITSYSCALL
170 E(PROC_TGID_LOGINUID, "loginuid", S_IFREG|S_IWUSR|S_IRUGO),
171#endif
172 {0,0,NULL,0}
173};
174static struct pid_entry tid_base_stuff[] = {
175 E(PROC_TID_FD, "fd", S_IFDIR|S_IRUSR|S_IXUSR),
176 E(PROC_TID_ENVIRON, "environ", S_IFREG|S_IRUSR),
177 E(PROC_TID_AUXV, "auxv", S_IFREG|S_IRUSR),
178 E(PROC_TID_STATUS, "status", S_IFREG|S_IRUGO),
179 E(PROC_TID_CMDLINE, "cmdline", S_IFREG|S_IRUGO),
180 E(PROC_TID_STAT, "stat", S_IFREG|S_IRUGO),
181 E(PROC_TID_STATM, "statm", S_IFREG|S_IRUGO),
182 E(PROC_TID_MAPS, "maps", S_IFREG|S_IRUGO),
183 E(PROC_TID_MEM, "mem", S_IFREG|S_IRUSR|S_IWUSR),
184#ifdef CONFIG_SECCOMP
185 E(PROC_TID_SECCOMP, "seccomp", S_IFREG|S_IRUSR|S_IWUSR),
186#endif
187 E(PROC_TID_CWD, "cwd", S_IFLNK|S_IRWXUGO),
188 E(PROC_TID_ROOT, "root", S_IFLNK|S_IRWXUGO),
189 E(PROC_TID_EXE, "exe", S_IFLNK|S_IRWXUGO),
190 E(PROC_TID_MOUNTS, "mounts", S_IFREG|S_IRUGO),
191#ifdef CONFIG_SECURITY
192 E(PROC_TID_ATTR, "attr", S_IFDIR|S_IRUGO|S_IXUGO),
193#endif
194#ifdef CONFIG_KALLSYMS
195 E(PROC_TID_WCHAN, "wchan", S_IFREG|S_IRUGO),
196#endif
197#ifdef CONFIG_SCHEDSTATS
198 E(PROC_TID_SCHEDSTAT, "schedstat",S_IFREG|S_IRUGO),
199#endif
200#ifdef CONFIG_CPUSETS
201 E(PROC_TID_CPUSET, "cpuset", S_IFREG|S_IRUGO),
202#endif
203 E(PROC_TID_OOM_SCORE, "oom_score",S_IFREG|S_IRUGO),
204 E(PROC_TID_OOM_ADJUST, "oom_adj", S_IFREG|S_IRUGO|S_IWUSR),
205#ifdef CONFIG_AUDITSYSCALL
206 E(PROC_TID_LOGINUID, "loginuid", S_IFREG|S_IWUSR|S_IRUGO),
207#endif
208 {0,0,NULL,0}
209};
210
211#ifdef CONFIG_SECURITY
212static struct pid_entry tgid_attr_stuff[] = {
213 E(PROC_TGID_ATTR_CURRENT, "current", S_IFREG|S_IRUGO|S_IWUGO),
214 E(PROC_TGID_ATTR_PREV, "prev", S_IFREG|S_IRUGO),
215 E(PROC_TGID_ATTR_EXEC, "exec", S_IFREG|S_IRUGO|S_IWUGO),
216 E(PROC_TGID_ATTR_FSCREATE, "fscreate", S_IFREG|S_IRUGO|S_IWUGO),
217 {0,0,NULL,0}
218};
219static struct pid_entry tid_attr_stuff[] = {
220 E(PROC_TID_ATTR_CURRENT, "current", S_IFREG|S_IRUGO|S_IWUGO),
221 E(PROC_TID_ATTR_PREV, "prev", S_IFREG|S_IRUGO),
222 E(PROC_TID_ATTR_EXEC, "exec", S_IFREG|S_IRUGO|S_IWUGO),
223 E(PROC_TID_ATTR_FSCREATE, "fscreate", S_IFREG|S_IRUGO|S_IWUGO),
224 {0,0,NULL,0}
225};
226#endif
227
228#undef E
229
230static int proc_fd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
231{
232 struct task_struct *task = proc_task(inode);
233 struct files_struct *files;
234 struct file *file;
235 int fd = proc_type(inode) - PROC_TID_FD_DIR;
236
237 files = get_files_struct(task);
238 if (files) {
239 spin_lock(&files->file_lock);
240 file = fcheck_files(files, fd);
241 if (file) {
242 *mnt = mntget(file->f_vfsmnt);
243 *dentry = dget(file->f_dentry);
244 spin_unlock(&files->file_lock);
245 put_files_struct(files);
246 return 0;
247 }
248 spin_unlock(&files->file_lock);
249 put_files_struct(files);
250 }
251 return -ENOENT;
252}
253
254static int proc_cwd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
255{
256 struct fs_struct *fs;
257 int result = -ENOENT;
258 task_lock(proc_task(inode));
259 fs = proc_task(inode)->fs;
260 if(fs)
261 atomic_inc(&fs->count);
262 task_unlock(proc_task(inode));
263 if (fs) {
264 read_lock(&fs->lock);
265 *mnt = mntget(fs->pwdmnt);
266 *dentry = dget(fs->pwd);
267 read_unlock(&fs->lock);
268 result = 0;
269 put_fs_struct(fs);
270 }
271 return result;
272}
273
274static int proc_root_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
275{
276 struct fs_struct *fs;
277 int result = -ENOENT;
278 task_lock(proc_task(inode));
279 fs = proc_task(inode)->fs;
280 if(fs)
281 atomic_inc(&fs->count);
282 task_unlock(proc_task(inode));
283 if (fs) {
284 read_lock(&fs->lock);
285 *mnt = mntget(fs->rootmnt);
286 *dentry = dget(fs->root);
287 read_unlock(&fs->lock);
288 result = 0;
289 put_fs_struct(fs);
290 }
291 return result;
292}
293
294#define MAY_PTRACE(task) \
295 (task == current || \
296 (task->parent == current && \
297 (task->ptrace & PT_PTRACED) && \
298 (task->state == TASK_STOPPED || task->state == TASK_TRACED) && \
299 security_ptrace(current,task) == 0))
300
301static int may_ptrace_attach(struct task_struct *task)
302{
303 int retval = 0;
304
305 task_lock(task);
306
307 if (!task->mm)
308 goto out;
309 if (((current->uid != task->euid) ||
310 (current->uid != task->suid) ||
311 (current->uid != task->uid) ||
312 (current->gid != task->egid) ||
313 (current->gid != task->sgid) ||
314 (current->gid != task->gid)) && !capable(CAP_SYS_PTRACE))
315 goto out;
316 rmb();
317 if (!task->mm->dumpable && !capable(CAP_SYS_PTRACE))
318 goto out;
319 if (security_ptrace(current, task))
320 goto out;
321
322 retval = 1;
323out:
324 task_unlock(task);
325 return retval;
326}
327
328static int proc_pid_environ(struct task_struct *task, char * buffer)
329{
330 int res = 0;
331 struct mm_struct *mm = get_task_mm(task);
332 if (mm) {
333 unsigned int len = mm->env_end - mm->env_start;
334 if (len > PAGE_SIZE)
335 len = PAGE_SIZE;
336 res = access_process_vm(task, mm->env_start, buffer, len, 0);
337 if (!may_ptrace_attach(task))
338 res = -ESRCH;
339 mmput(mm);
340 }
341 return res;
342}
343
344static int proc_pid_cmdline(struct task_struct *task, char * buffer)
345{
346 int res = 0;
347 unsigned int len;
348 struct mm_struct *mm = get_task_mm(task);
349 if (!mm)
350 goto out;
351 if (!mm->arg_end)
352 goto out_mm; /* Shh! No looking before we're done */
353
354 len = mm->arg_end - mm->arg_start;
355
356 if (len > PAGE_SIZE)
357 len = PAGE_SIZE;
358
359 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
360
361 // If the nul at the end of args has been overwritten, then
362 // assume application is using setproctitle(3).
363 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
364 len = strnlen(buffer, res);
365 if (len < res) {
366 res = len;
367 } else {
368 len = mm->env_end - mm->env_start;
369 if (len > PAGE_SIZE - res)
370 len = PAGE_SIZE - res;
371 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
372 res = strnlen(buffer, res);
373 }
374 }
375out_mm:
376 mmput(mm);
377out:
378 return res;
379}
380
381static int proc_pid_auxv(struct task_struct *task, char *buffer)
382{
383 int res = 0;
384 struct mm_struct *mm = get_task_mm(task);
385 if (mm) {
386 unsigned int nwords = 0;
387 do
388 nwords += 2;
389 while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
390 res = nwords * sizeof(mm->saved_auxv[0]);
391 if (res > PAGE_SIZE)
392 res = PAGE_SIZE;
393 memcpy(buffer, mm->saved_auxv, res);
394 mmput(mm);
395 }
396 return res;
397}
398
399
400#ifdef CONFIG_KALLSYMS
401/*
402 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
403 * Returns the resolved symbol. If that fails, simply return the address.
404 */
405static int proc_pid_wchan(struct task_struct *task, char *buffer)
406{
407 char *modname;
408 const char *sym_name;
409 unsigned long wchan, size, offset;
410 char namebuf[KSYM_NAME_LEN+1];
411
412 wchan = get_wchan(task);
413
414 sym_name = kallsyms_lookup(wchan, &size, &offset, &modname, namebuf);
415 if (sym_name)
416 return sprintf(buffer, "%s", sym_name);
417 return sprintf(buffer, "%lu", wchan);
418}
419#endif /* CONFIG_KALLSYMS */
420
421#ifdef CONFIG_SCHEDSTATS
422/*
423 * Provides /proc/PID/schedstat
424 */
425static int proc_pid_schedstat(struct task_struct *task, char *buffer)
426{
427 return sprintf(buffer, "%lu %lu %lu\n",
428 task->sched_info.cpu_time,
429 task->sched_info.run_delay,
430 task->sched_info.pcnt);
431}
432#endif
433
434/* The badness from the OOM killer */
435unsigned long badness(struct task_struct *p, unsigned long uptime);
436static int proc_oom_score(struct task_struct *task, char *buffer)
437{
438 unsigned long points;
439 struct timespec uptime;
440
441 do_posix_clock_monotonic_gettime(&uptime);
442 points = badness(task, uptime.tv_sec);
443 return sprintf(buffer, "%lu\n", points);
444}
445
446/************************************************************************/
447/* Here the fs part begins */
448/************************************************************************/
449
450/* permission checks */
451
452static int proc_check_root(struct inode *inode)
453{
454 struct dentry *de, *base, *root;
455 struct vfsmount *our_vfsmnt, *vfsmnt, *mnt;
456 int res = 0;
457
458 if (proc_root_link(inode, &root, &vfsmnt)) /* Ewww... */
459 return -ENOENT;
460 read_lock(&current->fs->lock);
461 our_vfsmnt = mntget(current->fs->rootmnt);
462 base = dget(current->fs->root);
463 read_unlock(&current->fs->lock);
464
465 spin_lock(&vfsmount_lock);
466 de = root;
467 mnt = vfsmnt;
468
469 while (vfsmnt != our_vfsmnt) {
470 if (vfsmnt == vfsmnt->mnt_parent)
471 goto out;
472 de = vfsmnt->mnt_mountpoint;
473 vfsmnt = vfsmnt->mnt_parent;
474 }
475
476 if (!is_subdir(de, base))
477 goto out;
478 spin_unlock(&vfsmount_lock);
479
480exit:
481 dput(base);
482 mntput(our_vfsmnt);
483 dput(root);
484 mntput(mnt);
485 return res;
486out:
487 spin_unlock(&vfsmount_lock);
488 res = -EACCES;
489 goto exit;
490}
491
492static int proc_permission(struct inode *inode, int mask, struct nameidata *nd)
493{
494 if (generic_permission(inode, mask, NULL) != 0)
495 return -EACCES;
496 return proc_check_root(inode);
497}
498
499extern struct seq_operations proc_pid_maps_op;
500static int maps_open(struct inode *inode, struct file *file)
501{
502 struct task_struct *task = proc_task(inode);
503 int ret = seq_open(file, &proc_pid_maps_op);
504 if (!ret) {
505 struct seq_file *m = file->private_data;
506 m->private = task;
507 }
508 return ret;
509}
510
511static struct file_operations proc_maps_operations = {
512 .open = maps_open,
513 .read = seq_read,
514 .llseek = seq_lseek,
515 .release = seq_release,
516};
517
518extern struct seq_operations mounts_op;
519static int mounts_open(struct inode *inode, struct file *file)
520{
521 struct task_struct *task = proc_task(inode);
522 int ret = seq_open(file, &mounts_op);
523
524 if (!ret) {
525 struct seq_file *m = file->private_data;
526 struct namespace *namespace;
527 task_lock(task);
528 namespace = task->namespace;
529 if (namespace)
530 get_namespace(namespace);
531 task_unlock(task);
532
533 if (namespace)
534 m->private = namespace;
535 else {
536 seq_release(inode, file);
537 ret = -EINVAL;
538 }
539 }
540 return ret;
541}
542
543static int mounts_release(struct inode *inode, struct file *file)
544{
545 struct seq_file *m = file->private_data;
546 struct namespace *namespace = m->private;
547 put_namespace(namespace);
548 return seq_release(inode, file);
549}
550
551static struct file_operations proc_mounts_operations = {
552 .open = mounts_open,
553 .read = seq_read,
554 .llseek = seq_lseek,
555 .release = mounts_release,
556};
557
558#define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
559
560static ssize_t proc_info_read(struct file * file, char __user * buf,
561 size_t count, loff_t *ppos)
562{
563 struct inode * inode = file->f_dentry->d_inode;
564 unsigned long page;
565 ssize_t length;
566 struct task_struct *task = proc_task(inode);
567
568 if (count > PROC_BLOCK_SIZE)
569 count = PROC_BLOCK_SIZE;
570 if (!(page = __get_free_page(GFP_KERNEL)))
571 return -ENOMEM;
572
573 length = PROC_I(inode)->op.proc_read(task, (char*)page);
574
575 if (length >= 0)
576 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
577 free_page(page);
578 return length;
579}
580
581static struct file_operations proc_info_file_operations = {
582 .read = proc_info_read,
583};
584
585static int mem_open(struct inode* inode, struct file* file)
586{
587 file->private_data = (void*)((long)current->self_exec_id);
588 return 0;
589}
590
591static ssize_t mem_read(struct file * file, char __user * buf,
592 size_t count, loff_t *ppos)
593{
594 struct task_struct *task = proc_task(file->f_dentry->d_inode);
595 char *page;
596 unsigned long src = *ppos;
597 int ret = -ESRCH;
598 struct mm_struct *mm;
599
600 if (!MAY_PTRACE(task) || !may_ptrace_attach(task))
601 goto out;
602
603 ret = -ENOMEM;
604 page = (char *)__get_free_page(GFP_USER);
605 if (!page)
606 goto out;
607
608 ret = 0;
609
610 mm = get_task_mm(task);
611 if (!mm)
612 goto out_free;
613
614 ret = -EIO;
615
616 if (file->private_data != (void*)((long)current->self_exec_id))
617 goto out_put;
618
619 ret = 0;
620
621 while (count > 0) {
622 int this_len, retval;
623
624 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
625 retval = access_process_vm(task, src, page, this_len, 0);
626 if (!retval || !MAY_PTRACE(task) || !may_ptrace_attach(task)) {
627 if (!ret)
628 ret = -EIO;
629 break;
630 }
631
632 if (copy_to_user(buf, page, retval)) {
633 ret = -EFAULT;
634 break;
635 }
636
637 ret += retval;
638 src += retval;
639 buf += retval;
640 count -= retval;
641 }
642 *ppos = src;
643
644out_put:
645 mmput(mm);
646out_free:
647 free_page((unsigned long) page);
648out:
649 return ret;
650}
651
652#define mem_write NULL
653
654#ifndef mem_write
655/* This is a security hazard */
656static ssize_t mem_write(struct file * file, const char * buf,
657 size_t count, loff_t *ppos)
658{
659 int copied = 0;
660 char *page;
661 struct task_struct *task = proc_task(file->f_dentry->d_inode);
662 unsigned long dst = *ppos;
663
664 if (!MAY_PTRACE(task) || !may_ptrace_attach(task))
665 return -ESRCH;
666
667 page = (char *)__get_free_page(GFP_USER);
668 if (!page)
669 return -ENOMEM;
670
671 while (count > 0) {
672 int this_len, retval;
673
674 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
675 if (copy_from_user(page, buf, this_len)) {
676 copied = -EFAULT;
677 break;
678 }
679 retval = access_process_vm(task, dst, page, this_len, 1);
680 if (!retval) {
681 if (!copied)
682 copied = -EIO;
683 break;
684 }
685 copied += retval;
686 buf += retval;
687 dst += retval;
688 count -= retval;
689 }
690 *ppos = dst;
691 free_page((unsigned long) page);
692 return copied;
693}
694#endif
695
696static loff_t mem_lseek(struct file * file, loff_t offset, int orig)
697{
698 switch (orig) {
699 case 0:
700 file->f_pos = offset;
701 break;
702 case 1:
703 file->f_pos += offset;
704 break;
705 default:
706 return -EINVAL;
707 }
708 force_successful_syscall_return();
709 return file->f_pos;
710}
711
712static struct file_operations proc_mem_operations = {
713 .llseek = mem_lseek,
714 .read = mem_read,
715 .write = mem_write,
716 .open = mem_open,
717};
718
719static ssize_t oom_adjust_read(struct file *file, char __user *buf,
720 size_t count, loff_t *ppos)
721{
722 struct task_struct *task = proc_task(file->f_dentry->d_inode);
723 char buffer[8];
724 size_t len;
725 int oom_adjust = task->oomkilladj;
726 loff_t __ppos = *ppos;
727
728 len = sprintf(buffer, "%i\n", oom_adjust);
729 if (__ppos >= len)
730 return 0;
731 if (count > len-__ppos)
732 count = len-__ppos;
733 if (copy_to_user(buf, buffer + __ppos, count))
734 return -EFAULT;
735 *ppos = __ppos + count;
736 return count;
737}
738
739static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
740 size_t count, loff_t *ppos)
741{
742 struct task_struct *task = proc_task(file->f_dentry->d_inode);
743 char buffer[8], *end;
744 int oom_adjust;
745
746 if (!capable(CAP_SYS_RESOURCE))
747 return -EPERM;
748 memset(buffer, 0, 8);
749 if (count > 6)
750 count = 6;
751 if (copy_from_user(buffer, buf, count))
752 return -EFAULT;
753 oom_adjust = simple_strtol(buffer, &end, 0);
754 if (oom_adjust < -16 || oom_adjust > 15)
755 return -EINVAL;
756 if (*end == '\n')
757 end++;
758 task->oomkilladj = oom_adjust;
759 if (end - buffer == 0)
760 return -EIO;
761 return end - buffer;
762}
763
764static struct file_operations proc_oom_adjust_operations = {
765 .read = oom_adjust_read,
766 .write = oom_adjust_write,
767};
768
769static struct inode_operations proc_mem_inode_operations = {
770 .permission = proc_permission,
771};
772
773#ifdef CONFIG_AUDITSYSCALL
774#define TMPBUFLEN 21
775static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
776 size_t count, loff_t *ppos)
777{
778 struct inode * inode = file->f_dentry->d_inode;
779 struct task_struct *task = proc_task(inode);
780 ssize_t length;
781 char tmpbuf[TMPBUFLEN];
782
783 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
784 audit_get_loginuid(task->audit_context));
785 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
786}
787
788static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
789 size_t count, loff_t *ppos)
790{
791 struct inode * inode = file->f_dentry->d_inode;
792 char *page, *tmp;
793 ssize_t length;
794 struct task_struct *task = proc_task(inode);
795 uid_t loginuid;
796
797 if (!capable(CAP_AUDIT_CONTROL))
798 return -EPERM;
799
800 if (current != task)
801 return -EPERM;
802
803 if (count > PAGE_SIZE)
804 count = PAGE_SIZE;
805
806 if (*ppos != 0) {
807 /* No partial writes. */
808 return -EINVAL;
809 }
810 page = (char*)__get_free_page(GFP_USER);
811 if (!page)
812 return -ENOMEM;
813 length = -EFAULT;
814 if (copy_from_user(page, buf, count))
815 goto out_free_page;
816
817 loginuid = simple_strtoul(page, &tmp, 10);
818 if (tmp == page) {
819 length = -EINVAL;
820 goto out_free_page;
821
822 }
823 length = audit_set_loginuid(task->audit_context, loginuid);
824 if (likely(length == 0))
825 length = count;
826
827out_free_page:
828 free_page((unsigned long) page);
829 return length;
830}
831
832static struct file_operations proc_loginuid_operations = {
833 .read = proc_loginuid_read,
834 .write = proc_loginuid_write,
835};
836#endif
837
838#ifdef CONFIG_SECCOMP
839static ssize_t seccomp_read(struct file *file, char __user *buf,
840 size_t count, loff_t *ppos)
841{
842 struct task_struct *tsk = proc_task(file->f_dentry->d_inode);
843 char __buf[20];
844 loff_t __ppos = *ppos;
845 size_t len;
846
847 /* no need to print the trailing zero, so use only len */
848 len = sprintf(__buf, "%u\n", tsk->seccomp.mode);
849 if (__ppos >= len)
850 return 0;
851 if (count > len - __ppos)
852 count = len - __ppos;
853 if (copy_to_user(buf, __buf + __ppos, count))
854 return -EFAULT;
855 *ppos = __ppos + count;
856 return count;
857}
858
859static ssize_t seccomp_write(struct file *file, const char __user *buf,
860 size_t count, loff_t *ppos)
861{
862 struct task_struct *tsk = proc_task(file->f_dentry->d_inode);
863 char __buf[20], *end;
864 unsigned int seccomp_mode;
865
866 /* can set it only once to be even more secure */
867 if (unlikely(tsk->seccomp.mode))
868 return -EPERM;
869
870 memset(__buf, 0, sizeof(__buf));
871 count = min(count, sizeof(__buf) - 1);
872 if (copy_from_user(__buf, buf, count))
873 return -EFAULT;
874 seccomp_mode = simple_strtoul(__buf, &end, 0);
875 if (*end == '\n')
876 end++;
877 if (seccomp_mode && seccomp_mode <= NR_SECCOMP_MODES) {
878 tsk->seccomp.mode = seccomp_mode;
879 set_tsk_thread_flag(tsk, TIF_SECCOMP);
880 } else
881 return -EINVAL;
882 if (unlikely(!(end - __buf)))
883 return -EIO;
884 return end - __buf;
885}
886
887static struct file_operations proc_seccomp_operations = {
888 .read = seccomp_read,
889 .write = seccomp_write,
890};
891#endif /* CONFIG_SECCOMP */
892
893static int proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
894{
895 struct inode *inode = dentry->d_inode;
896 int error = -EACCES;
897
898 /* We don't need a base pointer in the /proc filesystem */
899 path_release(nd);
900
901 if (current->fsuid != inode->i_uid && !capable(CAP_DAC_OVERRIDE))
902 goto out;
903 error = proc_check_root(inode);
904 if (error)
905 goto out;
906
907 error = PROC_I(inode)->op.proc_get_link(inode, &nd->dentry, &nd->mnt);
908 nd->last_type = LAST_BIND;
909out:
910 return error;
911}
912
913static int do_proc_readlink(struct dentry *dentry, struct vfsmount *mnt,
914 char __user *buffer, int buflen)
915{
916 struct inode * inode;
917 char *tmp = (char*)__get_free_page(GFP_KERNEL), *path;
918 int len;
919
920 if (!tmp)
921 return -ENOMEM;
922
923 inode = dentry->d_inode;
924 path = d_path(dentry, mnt, tmp, PAGE_SIZE);
925 len = PTR_ERR(path);
926 if (IS_ERR(path))
927 goto out;
928 len = tmp + PAGE_SIZE - 1 - path;
929
930 if (len > buflen)
931 len = buflen;
932 if (copy_to_user(buffer, path, len))
933 len = -EFAULT;
934 out:
935 free_page((unsigned long)tmp);
936 return len;
937}
938
939static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
940{
941 int error = -EACCES;
942 struct inode *inode = dentry->d_inode;
943 struct dentry *de;
944 struct vfsmount *mnt = NULL;
945
946 lock_kernel();
947
948 if (current->fsuid != inode->i_uid && !capable(CAP_DAC_OVERRIDE))
949 goto out;
950 error = proc_check_root(inode);
951 if (error)
952 goto out;
953
954 error = PROC_I(inode)->op.proc_get_link(inode, &de, &mnt);
955 if (error)
956 goto out;
957
958 error = do_proc_readlink(de, mnt, buffer, buflen);
959 dput(de);
960 mntput(mnt);
961out:
962 unlock_kernel();
963 return error;
964}
965
966static struct inode_operations proc_pid_link_inode_operations = {
967 .readlink = proc_pid_readlink,
968 .follow_link = proc_pid_follow_link
969};
970
971#define NUMBUF 10
972
973static int proc_readfd(struct file * filp, void * dirent, filldir_t filldir)
974{
975 struct inode *inode = filp->f_dentry->d_inode;
976 struct task_struct *p = proc_task(inode);
977 unsigned int fd, tid, ino;
978 int retval;
979 char buf[NUMBUF];
980 struct files_struct * files;
981
982 retval = -ENOENT;
983 if (!pid_alive(p))
984 goto out;
985 retval = 0;
986 tid = p->pid;
987
988 fd = filp->f_pos;
989 switch (fd) {
990 case 0:
991 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
992 goto out;
993 filp->f_pos++;
994 case 1:
995 ino = fake_ino(tid, PROC_TID_INO);
996 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
997 goto out;
998 filp->f_pos++;
999 default:
1000 files = get_files_struct(p);
1001 if (!files)
1002 goto out;
1003 spin_lock(&files->file_lock);
1004 for (fd = filp->f_pos-2;
1005 fd < files->max_fds;
1006 fd++, filp->f_pos++) {
1007 unsigned int i,j;
1008
1009 if (!fcheck_files(files, fd))
1010 continue;
1011 spin_unlock(&files->file_lock);
1012
1013 j = NUMBUF;
1014 i = fd;
1015 do {
1016 j--;
1017 buf[j] = '0' + (i % 10);
1018 i /= 10;
1019 } while (i);
1020
1021 ino = fake_ino(tid, PROC_TID_FD_DIR + fd);
1022 if (filldir(dirent, buf+j, NUMBUF-j, fd+2, ino, DT_LNK) < 0) {
1023 spin_lock(&files->file_lock);
1024 break;
1025 }
1026 spin_lock(&files->file_lock);
1027 }
1028 spin_unlock(&files->file_lock);
1029 put_files_struct(files);
1030 }
1031out:
1032 return retval;
1033}
1034
1035static int proc_pident_readdir(struct file *filp,
1036 void *dirent, filldir_t filldir,
1037 struct pid_entry *ents, unsigned int nents)
1038{
1039 int i;
1040 int pid;
1041 struct dentry *dentry = filp->f_dentry;
1042 struct inode *inode = dentry->d_inode;
1043 struct pid_entry *p;
1044 ino_t ino;
1045 int ret;
1046
1047 ret = -ENOENT;
1048 if (!pid_alive(proc_task(inode)))
1049 goto out;
1050
1051 ret = 0;
1052 pid = proc_task(inode)->pid;
1053 i = filp->f_pos;
1054 switch (i) {
1055 case 0:
1056 ino = inode->i_ino;
1057 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
1058 goto out;
1059 i++;
1060 filp->f_pos++;
1061 /* fall through */
1062 case 1:
1063 ino = parent_ino(dentry);
1064 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
1065 goto out;
1066 i++;
1067 filp->f_pos++;
1068 /* fall through */
1069 default:
1070 i -= 2;
1071 if (i >= nents) {
1072 ret = 1;
1073 goto out;
1074 }
1075 p = ents + i;
1076 while (p->name) {
1077 if (filldir(dirent, p->name, p->len, filp->f_pos,
1078 fake_ino(pid, p->type), p->mode >> 12) < 0)
1079 goto out;
1080 filp->f_pos++;
1081 p++;
1082 }
1083 }
1084
1085 ret = 1;
1086out:
1087 return ret;
1088}
1089
1090static int proc_tgid_base_readdir(struct file * filp,
1091 void * dirent, filldir_t filldir)
1092{
1093 return proc_pident_readdir(filp,dirent,filldir,
1094 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
1095}
1096
1097static int proc_tid_base_readdir(struct file * filp,
1098 void * dirent, filldir_t filldir)
1099{
1100 return proc_pident_readdir(filp,dirent,filldir,
1101 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
1102}
1103
1104/* building an inode */
1105
1106static int task_dumpable(struct task_struct *task)
1107{
1108 int dumpable = 0;
1109 struct mm_struct *mm;
1110
1111 task_lock(task);
1112 mm = task->mm;
1113 if (mm)
1114 dumpable = mm->dumpable;
1115 task_unlock(task);
1116 return dumpable;
1117}
1118
1119
1120static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task, int ino)
1121{
1122 struct inode * inode;
1123 struct proc_inode *ei;
1124
1125 /* We need a new inode */
1126
1127 inode = new_inode(sb);
1128 if (!inode)
1129 goto out;
1130
1131 /* Common stuff */
1132 ei = PROC_I(inode);
1133 ei->task = NULL;
1134 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1135 inode->i_ino = fake_ino(task->pid, ino);
1136
1137 if (!pid_alive(task))
1138 goto out_unlock;
1139
1140 /*
1141 * grab the reference to task.
1142 */
1143 get_task_struct(task);
1144 ei->task = task;
1145 ei->type = ino;
1146 inode->i_uid = 0;
1147 inode->i_gid = 0;
1148 if (ino == PROC_TGID_INO || ino == PROC_TID_INO || task_dumpable(task)) {
1149 inode->i_uid = task->euid;
1150 inode->i_gid = task->egid;
1151 }
1152 security_task_to_inode(task, inode);
1153
1154out:
1155 return inode;
1156
1157out_unlock:
1158 ei->pde = NULL;
1159 iput(inode);
1160 return NULL;
1161}
1162
1163/* dentry stuff */
1164
1165/*
1166 * Exceptional case: normally we are not allowed to unhash a busy
1167 * directory. In this case, however, we can do it - no aliasing problems
1168 * due to the way we treat inodes.
1169 *
1170 * Rewrite the inode's ownerships here because the owning task may have
1171 * performed a setuid(), etc.
1172 */
1173static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1174{
1175 struct inode *inode = dentry->d_inode;
1176 struct task_struct *task = proc_task(inode);
1177 if (pid_alive(task)) {
1178 if (proc_type(inode) == PROC_TGID_INO || proc_type(inode) == PROC_TID_INO || task_dumpable(task)) {
1179 inode->i_uid = task->euid;
1180 inode->i_gid = task->egid;
1181 } else {
1182 inode->i_uid = 0;
1183 inode->i_gid = 0;
1184 }
1185 security_task_to_inode(task, inode);
1186 return 1;
1187 }
1188 d_drop(dentry);
1189 return 0;
1190}
1191
1192static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1193{
1194 struct inode *inode = dentry->d_inode;
1195 struct task_struct *task = proc_task(inode);
1196 int fd = proc_type(inode) - PROC_TID_FD_DIR;
1197 struct files_struct *files;
1198
1199 files = get_files_struct(task);
1200 if (files) {
1201 spin_lock(&files->file_lock);
1202 if (fcheck_files(files, fd)) {
1203 spin_unlock(&files->file_lock);
1204 put_files_struct(files);
1205 if (task_dumpable(task)) {
1206 inode->i_uid = task->euid;
1207 inode->i_gid = task->egid;
1208 } else {
1209 inode->i_uid = 0;
1210 inode->i_gid = 0;
1211 }
1212 security_task_to_inode(task, inode);
1213 return 1;
1214 }
1215 spin_unlock(&files->file_lock);
1216 put_files_struct(files);
1217 }
1218 d_drop(dentry);
1219 return 0;
1220}
1221
1222static void pid_base_iput(struct dentry *dentry, struct inode *inode)
1223{
1224 struct task_struct *task = proc_task(inode);
1225 spin_lock(&task->proc_lock);
1226 if (task->proc_dentry == dentry)
1227 task->proc_dentry = NULL;
1228 spin_unlock(&task->proc_lock);
1229 iput(inode);
1230}
1231
1232static int pid_delete_dentry(struct dentry * dentry)
1233{
1234 /* Is the task we represent dead?
1235 * If so, then don't put the dentry on the lru list,
1236 * kill it immediately.
1237 */
1238 return !pid_alive(proc_task(dentry->d_inode));
1239}
1240
1241static struct dentry_operations tid_fd_dentry_operations =
1242{
1243 .d_revalidate = tid_fd_revalidate,
1244 .d_delete = pid_delete_dentry,
1245};
1246
1247static struct dentry_operations pid_dentry_operations =
1248{
1249 .d_revalidate = pid_revalidate,
1250 .d_delete = pid_delete_dentry,
1251};
1252
1253static struct dentry_operations pid_base_dentry_operations =
1254{
1255 .d_revalidate = pid_revalidate,
1256 .d_iput = pid_base_iput,
1257 .d_delete = pid_delete_dentry,
1258};
1259
1260/* Lookups */
1261
1262static unsigned name_to_int(struct dentry *dentry)
1263{
1264 const char *name = dentry->d_name.name;
1265 int len = dentry->d_name.len;
1266 unsigned n = 0;
1267
1268 if (len > 1 && *name == '0')
1269 goto out;
1270 while (len-- > 0) {
1271 unsigned c = *name++ - '0';
1272 if (c > 9)
1273 goto out;
1274 if (n >= (~0U-9)/10)
1275 goto out;
1276 n *= 10;
1277 n += c;
1278 }
1279 return n;
1280out:
1281 return ~0U;
1282}
1283
1284/* SMP-safe */
1285static struct dentry *proc_lookupfd(struct inode * dir, struct dentry * dentry, struct nameidata *nd)
1286{
1287 struct task_struct *task = proc_task(dir);
1288 unsigned fd = name_to_int(dentry);
1289 struct file * file;
1290 struct files_struct * files;
1291 struct inode *inode;
1292 struct proc_inode *ei;
1293
1294 if (fd == ~0U)
1295 goto out;
1296 if (!pid_alive(task))
1297 goto out;
1298
1299 inode = proc_pid_make_inode(dir->i_sb, task, PROC_TID_FD_DIR+fd);
1300 if (!inode)
1301 goto out;
1302 ei = PROC_I(inode);
1303 files = get_files_struct(task);
1304 if (!files)
1305 goto out_unlock;
1306 inode->i_mode = S_IFLNK;
1307 spin_lock(&files->file_lock);
1308 file = fcheck_files(files, fd);
1309 if (!file)
1310 goto out_unlock2;
1311 if (file->f_mode & 1)
1312 inode->i_mode |= S_IRUSR | S_IXUSR;
1313 if (file->f_mode & 2)
1314 inode->i_mode |= S_IWUSR | S_IXUSR;
1315 spin_unlock(&files->file_lock);
1316 put_files_struct(files);
1317 inode->i_op = &proc_pid_link_inode_operations;
1318 inode->i_size = 64;
1319 ei->op.proc_get_link = proc_fd_link;
1320 dentry->d_op = &tid_fd_dentry_operations;
1321 d_add(dentry, inode);
1322 return NULL;
1323
1324out_unlock2:
1325 spin_unlock(&files->file_lock);
1326 put_files_struct(files);
1327out_unlock:
1328 iput(inode);
1329out:
1330 return ERR_PTR(-ENOENT);
1331}
1332
1333static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir);
1334static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd);
1335
1336static struct file_operations proc_fd_operations = {
1337 .read = generic_read_dir,
1338 .readdir = proc_readfd,
1339};
1340
1341static struct file_operations proc_task_operations = {
1342 .read = generic_read_dir,
1343 .readdir = proc_task_readdir,
1344};
1345
1346/*
1347 * proc directories can do almost nothing..
1348 */
1349static struct inode_operations proc_fd_inode_operations = {
1350 .lookup = proc_lookupfd,
1351 .permission = proc_permission,
1352};
1353
1354static struct inode_operations proc_task_inode_operations = {
1355 .lookup = proc_task_lookup,
1356 .permission = proc_permission,
1357};
1358
1359#ifdef CONFIG_SECURITY
1360static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
1361 size_t count, loff_t *ppos)
1362{
1363 struct inode * inode = file->f_dentry->d_inode;
1364 unsigned long page;
1365 ssize_t length;
1366 struct task_struct *task = proc_task(inode);
1367
1368 if (count > PAGE_SIZE)
1369 count = PAGE_SIZE;
1370 if (!(page = __get_free_page(GFP_KERNEL)))
1371 return -ENOMEM;
1372
1373 length = security_getprocattr(task,
1374 (char*)file->f_dentry->d_name.name,
1375 (void*)page, count);
1376 if (length >= 0)
1377 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
1378 free_page(page);
1379 return length;
1380}
1381
1382static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
1383 size_t count, loff_t *ppos)
1384{
1385 struct inode * inode = file->f_dentry->d_inode;
1386 char *page;
1387 ssize_t length;
1388 struct task_struct *task = proc_task(inode);
1389
1390 if (count > PAGE_SIZE)
1391 count = PAGE_SIZE;
1392 if (*ppos != 0) {
1393 /* No partial writes. */
1394 return -EINVAL;
1395 }
1396 page = (char*)__get_free_page(GFP_USER);
1397 if (!page)
1398 return -ENOMEM;
1399 length = -EFAULT;
1400 if (copy_from_user(page, buf, count))
1401 goto out;
1402
1403 length = security_setprocattr(task,
1404 (char*)file->f_dentry->d_name.name,
1405 (void*)page, count);
1406out:
1407 free_page((unsigned long) page);
1408 return length;
1409}
1410
1411static struct file_operations proc_pid_attr_operations = {
1412 .read = proc_pid_attr_read,
1413 .write = proc_pid_attr_write,
1414};
1415
1416static struct file_operations proc_tid_attr_operations;
1417static struct inode_operations proc_tid_attr_inode_operations;
1418static struct file_operations proc_tgid_attr_operations;
1419static struct inode_operations proc_tgid_attr_inode_operations;
1420#endif
1421
1422/* SMP-safe */
1423static struct dentry *proc_pident_lookup(struct inode *dir,
1424 struct dentry *dentry,
1425 struct pid_entry *ents)
1426{
1427 struct inode *inode;
1428 int error;
1429 struct task_struct *task = proc_task(dir);
1430 struct pid_entry *p;
1431 struct proc_inode *ei;
1432
1433 error = -ENOENT;
1434 inode = NULL;
1435
1436 if (!pid_alive(task))
1437 goto out;
1438
1439 for (p = ents; p->name; p++) {
1440 if (p->len != dentry->d_name.len)
1441 continue;
1442 if (!memcmp(dentry->d_name.name, p->name, p->len))
1443 break;
1444 }
1445 if (!p->name)
1446 goto out;
1447
1448 error = -EINVAL;
1449 inode = proc_pid_make_inode(dir->i_sb, task, p->type);
1450 if (!inode)
1451 goto out;
1452
1453 ei = PROC_I(inode);
1454 inode->i_mode = p->mode;
1455 /*
1456 * Yes, it does not scale. And it should not. Don't add
1457 * new entries into /proc/<tgid>/ without very good reasons.
1458 */
1459 switch(p->type) {
1460 case PROC_TGID_TASK:
1461 inode->i_nlink = 3;
1462 inode->i_op = &proc_task_inode_operations;
1463 inode->i_fop = &proc_task_operations;
1464 break;
1465 case PROC_TID_FD:
1466 case PROC_TGID_FD:
1467 inode->i_nlink = 2;
1468 inode->i_op = &proc_fd_inode_operations;
1469 inode->i_fop = &proc_fd_operations;
1470 break;
1471 case PROC_TID_EXE:
1472 case PROC_TGID_EXE:
1473 inode->i_op = &proc_pid_link_inode_operations;
1474 ei->op.proc_get_link = proc_exe_link;
1475 break;
1476 case PROC_TID_CWD:
1477 case PROC_TGID_CWD:
1478 inode->i_op = &proc_pid_link_inode_operations;
1479 ei->op.proc_get_link = proc_cwd_link;
1480 break;
1481 case PROC_TID_ROOT:
1482 case PROC_TGID_ROOT:
1483 inode->i_op = &proc_pid_link_inode_operations;
1484 ei->op.proc_get_link = proc_root_link;
1485 break;
1486 case PROC_TID_ENVIRON:
1487 case PROC_TGID_ENVIRON:
1488 inode->i_fop = &proc_info_file_operations;
1489 ei->op.proc_read = proc_pid_environ;
1490 break;
1491 case PROC_TID_AUXV:
1492 case PROC_TGID_AUXV:
1493 inode->i_fop = &proc_info_file_operations;
1494 ei->op.proc_read = proc_pid_auxv;
1495 break;
1496 case PROC_TID_STATUS:
1497 case PROC_TGID_STATUS:
1498 inode->i_fop = &proc_info_file_operations;
1499 ei->op.proc_read = proc_pid_status;
1500 break;
1501 case PROC_TID_STAT:
1502 inode->i_fop = &proc_info_file_operations;
1503 ei->op.proc_read = proc_tid_stat;
1504 break;
1505 case PROC_TGID_STAT:
1506 inode->i_fop = &proc_info_file_operations;
1507 ei->op.proc_read = proc_tgid_stat;
1508 break;
1509 case PROC_TID_CMDLINE:
1510 case PROC_TGID_CMDLINE:
1511 inode->i_fop = &proc_info_file_operations;
1512 ei->op.proc_read = proc_pid_cmdline;
1513 break;
1514 case PROC_TID_STATM:
1515 case PROC_TGID_STATM:
1516 inode->i_fop = &proc_info_file_operations;
1517 ei->op.proc_read = proc_pid_statm;
1518 break;
1519 case PROC_TID_MAPS:
1520 case PROC_TGID_MAPS:
1521 inode->i_fop = &proc_maps_operations;
1522 break;
1523 case PROC_TID_MEM:
1524 case PROC_TGID_MEM:
1525 inode->i_op = &proc_mem_inode_operations;
1526 inode->i_fop = &proc_mem_operations;
1527 break;
1528#ifdef CONFIG_SECCOMP
1529 case PROC_TID_SECCOMP:
1530 case PROC_TGID_SECCOMP:
1531 inode->i_fop = &proc_seccomp_operations;
1532 break;
1533#endif /* CONFIG_SECCOMP */
1534 case PROC_TID_MOUNTS:
1535 case PROC_TGID_MOUNTS:
1536 inode->i_fop = &proc_mounts_operations;
1537 break;
1538#ifdef CONFIG_SECURITY
1539 case PROC_TID_ATTR:
1540 inode->i_nlink = 2;
1541 inode->i_op = &proc_tid_attr_inode_operations;
1542 inode->i_fop = &proc_tid_attr_operations;
1543 break;
1544 case PROC_TGID_ATTR:
1545 inode->i_nlink = 2;
1546 inode->i_op = &proc_tgid_attr_inode_operations;
1547 inode->i_fop = &proc_tgid_attr_operations;
1548 break;
1549 case PROC_TID_ATTR_CURRENT:
1550 case PROC_TGID_ATTR_CURRENT:
1551 case PROC_TID_ATTR_PREV:
1552 case PROC_TGID_ATTR_PREV:
1553 case PROC_TID_ATTR_EXEC:
1554 case PROC_TGID_ATTR_EXEC:
1555 case PROC_TID_ATTR_FSCREATE:
1556 case PROC_TGID_ATTR_FSCREATE:
1557 inode->i_fop = &proc_pid_attr_operations;
1558 break;
1559#endif
1560#ifdef CONFIG_KALLSYMS
1561 case PROC_TID_WCHAN:
1562 case PROC_TGID_WCHAN:
1563 inode->i_fop = &proc_info_file_operations;
1564 ei->op.proc_read = proc_pid_wchan;
1565 break;
1566#endif
1567#ifdef CONFIG_SCHEDSTATS
1568 case PROC_TID_SCHEDSTAT:
1569 case PROC_TGID_SCHEDSTAT:
1570 inode->i_fop = &proc_info_file_operations;
1571 ei->op.proc_read = proc_pid_schedstat;
1572 break;
1573#endif
1574#ifdef CONFIG_CPUSETS
1575 case PROC_TID_CPUSET:
1576 case PROC_TGID_CPUSET:
1577 inode->i_fop = &proc_cpuset_operations;
1578 break;
1579#endif
1580 case PROC_TID_OOM_SCORE:
1581 case PROC_TGID_OOM_SCORE:
1582 inode->i_fop = &proc_info_file_operations;
1583 ei->op.proc_read = proc_oom_score;
1584 break;
1585 case PROC_TID_OOM_ADJUST:
1586 case PROC_TGID_OOM_ADJUST:
1587 inode->i_fop = &proc_oom_adjust_operations;
1588 break;
1589#ifdef CONFIG_AUDITSYSCALL
1590 case PROC_TID_LOGINUID:
1591 case PROC_TGID_LOGINUID:
1592 inode->i_fop = &proc_loginuid_operations;
1593 break;
1594#endif
1595 default:
1596 printk("procfs: impossible type (%d)",p->type);
1597 iput(inode);
1598 return ERR_PTR(-EINVAL);
1599 }
1600 dentry->d_op = &pid_dentry_operations;
1601 d_add(dentry, inode);
1602 return NULL;
1603
1604out:
1605 return ERR_PTR(error);
1606}
1607
1608static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
1609 return proc_pident_lookup(dir, dentry, tgid_base_stuff);
1610}
1611
1612static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
1613 return proc_pident_lookup(dir, dentry, tid_base_stuff);
1614}
1615
1616static struct file_operations proc_tgid_base_operations = {
1617 .read = generic_read_dir,
1618 .readdir = proc_tgid_base_readdir,
1619};
1620
1621static struct file_operations proc_tid_base_operations = {
1622 .read = generic_read_dir,
1623 .readdir = proc_tid_base_readdir,
1624};
1625
1626static struct inode_operations proc_tgid_base_inode_operations = {
1627 .lookup = proc_tgid_base_lookup,
1628};
1629
1630static struct inode_operations proc_tid_base_inode_operations = {
1631 .lookup = proc_tid_base_lookup,
1632};
1633
1634#ifdef CONFIG_SECURITY
1635static int proc_tgid_attr_readdir(struct file * filp,
1636 void * dirent, filldir_t filldir)
1637{
1638 return proc_pident_readdir(filp,dirent,filldir,
1639 tgid_attr_stuff,ARRAY_SIZE(tgid_attr_stuff));
1640}
1641
1642static int proc_tid_attr_readdir(struct file * filp,
1643 void * dirent, filldir_t filldir)
1644{
1645 return proc_pident_readdir(filp,dirent,filldir,
1646 tid_attr_stuff,ARRAY_SIZE(tid_attr_stuff));
1647}
1648
1649static struct file_operations proc_tgid_attr_operations = {
1650 .read = generic_read_dir,
1651 .readdir = proc_tgid_attr_readdir,
1652};
1653
1654static struct file_operations proc_tid_attr_operations = {
1655 .read = generic_read_dir,
1656 .readdir = proc_tid_attr_readdir,
1657};
1658
1659static struct dentry *proc_tgid_attr_lookup(struct inode *dir,
1660 struct dentry *dentry, struct nameidata *nd)
1661{
1662 return proc_pident_lookup(dir, dentry, tgid_attr_stuff);
1663}
1664
1665static struct dentry *proc_tid_attr_lookup(struct inode *dir,
1666 struct dentry *dentry, struct nameidata *nd)
1667{
1668 return proc_pident_lookup(dir, dentry, tid_attr_stuff);
1669}
1670
1671static struct inode_operations proc_tgid_attr_inode_operations = {
1672 .lookup = proc_tgid_attr_lookup,
1673};
1674
1675static struct inode_operations proc_tid_attr_inode_operations = {
1676 .lookup = proc_tid_attr_lookup,
1677};
1678#endif
1679
1680/*
1681 * /proc/self:
1682 */
1683static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
1684 int buflen)
1685{
1686 char tmp[30];
1687 sprintf(tmp, "%d", current->tgid);
1688 return vfs_readlink(dentry,buffer,buflen,tmp);
1689}
1690
1691static int proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
1692{
1693 char tmp[30];
1694 sprintf(tmp, "%d", current->tgid);
1695 return vfs_follow_link(nd,tmp);
1696}
1697
1698static struct inode_operations proc_self_inode_operations = {
1699 .readlink = proc_self_readlink,
1700 .follow_link = proc_self_follow_link,
1701};
1702
1703/**
1704 * proc_pid_unhash - Unhash /proc/<pid> entry from the dcache.
1705 * @p: task that should be flushed.
1706 *
1707 * Drops the /proc/<pid> dcache entry from the hash chains.
1708 *
1709 * Dropping /proc/<pid> entries and detach_pid must be synchroneous,
1710 * otherwise e.g. /proc/<pid>/exe might point to the wrong executable,
1711 * if the pid value is immediately reused. This is enforced by
1712 * - caller must acquire spin_lock(p->proc_lock)
1713 * - must be called before detach_pid()
1714 * - proc_pid_lookup acquires proc_lock, and checks that
1715 * the target is not dead by looking at the attach count
1716 * of PIDTYPE_PID.
1717 */
1718
1719struct dentry *proc_pid_unhash(struct task_struct *p)
1720{
1721 struct dentry *proc_dentry;
1722
1723 proc_dentry = p->proc_dentry;
1724 if (proc_dentry != NULL) {
1725
1726 spin_lock(&dcache_lock);
1727 spin_lock(&proc_dentry->d_lock);
1728 if (!d_unhashed(proc_dentry)) {
1729 dget_locked(proc_dentry);
1730 __d_drop(proc_dentry);
1731 spin_unlock(&proc_dentry->d_lock);
1732 } else {
1733 spin_unlock(&proc_dentry->d_lock);
1734 proc_dentry = NULL;
1735 }
1736 spin_unlock(&dcache_lock);
1737 }
1738 return proc_dentry;
1739}
1740
1741/**
1742 * proc_pid_flush - recover memory used by stale /proc/<pid>/x entries
1743 * @proc_entry: directoy to prune.
1744 *
1745 * Shrink the /proc directory that was used by the just killed thread.
1746 */
1747
1748void proc_pid_flush(struct dentry *proc_dentry)
1749{
1750 might_sleep();
1751 if(proc_dentry != NULL) {
1752 shrink_dcache_parent(proc_dentry);
1753 dput(proc_dentry);
1754 }
1755}
1756
1757/* SMP-safe */
1758struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
1759{
1760 struct task_struct *task;
1761 struct inode *inode;
1762 struct proc_inode *ei;
1763 unsigned tgid;
1764 int died;
1765
1766 if (dentry->d_name.len == 4 && !memcmp(dentry->d_name.name,"self",4)) {
1767 inode = new_inode(dir->i_sb);
1768 if (!inode)
1769 return ERR_PTR(-ENOMEM);
1770 ei = PROC_I(inode);
1771 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1772 inode->i_ino = fake_ino(0, PROC_TGID_INO);
1773 ei->pde = NULL;
1774 inode->i_mode = S_IFLNK|S_IRWXUGO;
1775 inode->i_uid = inode->i_gid = 0;
1776 inode->i_size = 64;
1777 inode->i_op = &proc_self_inode_operations;
1778 d_add(dentry, inode);
1779 return NULL;
1780 }
1781 tgid = name_to_int(dentry);
1782 if (tgid == ~0U)
1783 goto out;
1784
1785 read_lock(&tasklist_lock);
1786 task = find_task_by_pid(tgid);
1787 if (task)
1788 get_task_struct(task);
1789 read_unlock(&tasklist_lock);
1790 if (!task)
1791 goto out;
1792
1793 inode = proc_pid_make_inode(dir->i_sb, task, PROC_TGID_INO);
1794
1795
1796 if (!inode) {
1797 put_task_struct(task);
1798 goto out;
1799 }
1800 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
1801 inode->i_op = &proc_tgid_base_inode_operations;
1802 inode->i_fop = &proc_tgid_base_operations;
1803 inode->i_nlink = 3;
1804 inode->i_flags|=S_IMMUTABLE;
1805
1806 dentry->d_op = &pid_base_dentry_operations;
1807
1808 died = 0;
1809 d_add(dentry, inode);
1810 spin_lock(&task->proc_lock);
1811 task->proc_dentry = dentry;
1812 if (!pid_alive(task)) {
1813 dentry = proc_pid_unhash(task);
1814 died = 1;
1815 }
1816 spin_unlock(&task->proc_lock);
1817
1818 put_task_struct(task);
1819 if (died) {
1820 proc_pid_flush(dentry);
1821 goto out;
1822 }
1823 return NULL;
1824out:
1825 return ERR_PTR(-ENOENT);
1826}
1827
1828/* SMP-safe */
1829static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
1830{
1831 struct task_struct *task;
1832 struct task_struct *leader = proc_task(dir);
1833 struct inode *inode;
1834 unsigned tid;
1835
1836 tid = name_to_int(dentry);
1837 if (tid == ~0U)
1838 goto out;
1839
1840 read_lock(&tasklist_lock);
1841 task = find_task_by_pid(tid);
1842 if (task)
1843 get_task_struct(task);
1844 read_unlock(&tasklist_lock);
1845 if (!task)
1846 goto out;
1847 if (leader->tgid != task->tgid)
1848 goto out_drop_task;
1849
1850 inode = proc_pid_make_inode(dir->i_sb, task, PROC_TID_INO);
1851
1852
1853 if (!inode)
1854 goto out_drop_task;
1855 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
1856 inode->i_op = &proc_tid_base_inode_operations;
1857 inode->i_fop = &proc_tid_base_operations;
1858 inode->i_nlink = 3;
1859 inode->i_flags|=S_IMMUTABLE;
1860
1861 dentry->d_op = &pid_base_dentry_operations;
1862
1863 d_add(dentry, inode);
1864
1865 put_task_struct(task);
1866 return NULL;
1867out_drop_task:
1868 put_task_struct(task);
1869out:
1870 return ERR_PTR(-ENOENT);
1871}
1872
1873#define PROC_NUMBUF 10
1874#define PROC_MAXPIDS 20
1875
1876/*
1877 * Get a few tgid's to return for filldir - we need to hold the
1878 * tasklist lock while doing this, and we must release it before
1879 * we actually do the filldir itself, so we use a temp buffer..
1880 */
1881static int get_tgid_list(int index, unsigned long version, unsigned int *tgids)
1882{
1883 struct task_struct *p;
1884 int nr_tgids = 0;
1885
1886 index--;
1887 read_lock(&tasklist_lock);
1888 p = NULL;
1889 if (version) {
1890 p = find_task_by_pid(version);
1891 if (p && !thread_group_leader(p))
1892 p = NULL;
1893 }
1894
1895 if (p)
1896 index = 0;
1897 else
1898 p = next_task(&init_task);
1899
1900 for ( ; p != &init_task; p = next_task(p)) {
1901 int tgid = p->pid;
1902 if (!pid_alive(p))
1903 continue;
1904 if (--index >= 0)
1905 continue;
1906 tgids[nr_tgids] = tgid;
1907 nr_tgids++;
1908 if (nr_tgids >= PROC_MAXPIDS)
1909 break;
1910 }
1911 read_unlock(&tasklist_lock);
1912 return nr_tgids;
1913}
1914
1915/*
1916 * Get a few tid's to return for filldir - we need to hold the
1917 * tasklist lock while doing this, and we must release it before
1918 * we actually do the filldir itself, so we use a temp buffer..
1919 */
1920static int get_tid_list(int index, unsigned int *tids, struct inode *dir)
1921{
1922 struct task_struct *leader_task = proc_task(dir);
1923 struct task_struct *task = leader_task;
1924 int nr_tids = 0;
1925
1926 index -= 2;
1927 read_lock(&tasklist_lock);
1928 /*
1929 * The starting point task (leader_task) might be an already
1930 * unlinked task, which cannot be used to access the task-list
1931 * via next_thread().
1932 */
1933 if (pid_alive(task)) do {
1934 int tid = task->pid;
1935
1936 if (--index >= 0)
1937 continue;
1938 tids[nr_tids] = tid;
1939 nr_tids++;
1940 if (nr_tids >= PROC_MAXPIDS)
1941 break;
1942 } while ((task = next_thread(task)) != leader_task);
1943 read_unlock(&tasklist_lock);
1944 return nr_tids;
1945}
1946
1947/* for the /proc/ directory itself, after non-process stuff has been done */
1948int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
1949{
1950 unsigned int tgid_array[PROC_MAXPIDS];
1951 char buf[PROC_NUMBUF];
1952 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
1953 unsigned int nr_tgids, i;
1954 int next_tgid;
1955
1956 if (!nr) {
1957 ino_t ino = fake_ino(0,PROC_TGID_INO);
1958 if (filldir(dirent, "self", 4, filp->f_pos, ino, DT_LNK) < 0)
1959 return 0;
1960 filp->f_pos++;
1961 nr++;
1962 }
1963
1964 /* f_version caches the tgid value that the last readdir call couldn't
1965 * return. lseek aka telldir automagically resets f_version to 0.
1966 */
1967 next_tgid = filp->f_version;
1968 filp->f_version = 0;
1969 for (;;) {
1970 nr_tgids = get_tgid_list(nr, next_tgid, tgid_array);
1971 if (!nr_tgids) {
1972 /* no more entries ! */
1973 break;
1974 }
1975 next_tgid = 0;
1976
1977 /* do not use the last found pid, reserve it for next_tgid */
1978 if (nr_tgids == PROC_MAXPIDS) {
1979 nr_tgids--;
1980 next_tgid = tgid_array[nr_tgids];
1981 }
1982
1983 for (i=0;i<nr_tgids;i++) {
1984 int tgid = tgid_array[i];
1985 ino_t ino = fake_ino(tgid,PROC_TGID_INO);
1986 unsigned long j = PROC_NUMBUF;
1987
1988 do
1989 buf[--j] = '0' + (tgid % 10);
1990 while ((tgid /= 10) != 0);
1991
1992 if (filldir(dirent, buf+j, PROC_NUMBUF-j, filp->f_pos, ino, DT_DIR) < 0) {
1993 /* returning this tgid failed, save it as the first
1994 * pid for the next readir call */
1995 filp->f_version = tgid_array[i];
1996 goto out;
1997 }
1998 filp->f_pos++;
1999 nr++;
2000 }
2001 }
2002out:
2003 return 0;
2004}
2005
2006/* for the /proc/TGID/task/ directories */
2007static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
2008{
2009 unsigned int tid_array[PROC_MAXPIDS];
2010 char buf[PROC_NUMBUF];
2011 unsigned int nr_tids, i;
2012 struct dentry *dentry = filp->f_dentry;
2013 struct inode *inode = dentry->d_inode;
2014 int retval = -ENOENT;
2015 ino_t ino;
2016 unsigned long pos = filp->f_pos; /* avoiding "long long" filp->f_pos */
2017
2018 if (!pid_alive(proc_task(inode)))
2019 goto out;
2020 retval = 0;
2021
2022 switch (pos) {
2023 case 0:
2024 ino = inode->i_ino;
2025 if (filldir(dirent, ".", 1, pos, ino, DT_DIR) < 0)
2026 goto out;
2027 pos++;
2028 /* fall through */
2029 case 1:
2030 ino = parent_ino(dentry);
2031 if (filldir(dirent, "..", 2, pos, ino, DT_DIR) < 0)
2032 goto out;
2033 pos++;
2034 /* fall through */
2035 }
2036
2037 nr_tids = get_tid_list(pos, tid_array, inode);
2038
2039 for (i = 0; i < nr_tids; i++) {
2040 unsigned long j = PROC_NUMBUF;
2041 int tid = tid_array[i];
2042
2043 ino = fake_ino(tid,PROC_TID_INO);
2044
2045 do
2046 buf[--j] = '0' + (tid % 10);
2047 while ((tid /= 10) != 0);
2048
2049 if (filldir(dirent, buf+j, PROC_NUMBUF-j, pos, ino, DT_DIR) < 0)
2050 break;
2051 pos++;
2052 }
2053out:
2054 filp->f_pos = pos;
2055 return retval;
2056}