?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Remove redundant check from proc_setattr()
[linux-2.6/cjktty.git] / fs / proc / base.c
blobff9c131e6b3062d2d1f4fd800557023a0936bf36
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.
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.
48 */
49
50 #include <asm/uaccess.h>
51
52 #include <linux/errno.h>
53 #include <linux/time.h>
54 #include <linux/proc_fs.h>
55 #include <linux/stat.h>
56 #include <linux/init.h>
57 #include <linux/capability.h>
58 #include <linux/file.h>
59 #include <linux/string.h>
60 #include <linux/seq_file.h>
61 #include <linux/namei.h>
62 #include <linux/mnt_namespace.h>
63 #include <linux/mm.h>
64 #include <linux/rcupdate.h>
65 #include <linux/kallsyms.h>
66 #include <linux/module.h>
67 #include <linux/mount.h>
68 #include <linux/security.h>
69 #include <linux/ptrace.h>
70 #include <linux/seccomp.h>
71 #include <linux/cpuset.h>
72 #include <linux/audit.h>
73 #include <linux/poll.h>
74 #include <linux/nsproxy.h>
75 #include <linux/oom.h>
76 #include "internal.h"
77
78 /* NOTE:
79 * Implementing inode permission operations in /proc is almost
80 * certainly an error. Permission checks need to happen during
81 * each system call not at open time. The reason is that most of
82 * what we wish to check for permissions in /proc varies at runtime.
83 *
84 * The classic example of a problem is opening file descriptors
85 * in /proc for a task before it execs a suid executable.
86 */
87
88
89 /* Worst case buffer size needed for holding an integer. */
90 #define PROC_NUMBUF 13
91
92 struct pid_entry {
93 char *name;
94 int len;
95 mode_t mode;
96 const struct inode_operations *iop;
97 const struct file_operations *fop;
98 union proc_op op;
99 };
100
101 #define NOD(NAME, MODE, IOP, FOP, OP) { \
102 .name = (NAME), \
103 .len = sizeof(NAME) - 1, \
104 .mode = MODE, \
105 .iop = IOP, \
106 .fop = FOP, \
107 .op = OP, \
108 }
109
110 #define DIR(NAME, MODE, OTYPE) \
111 NOD(NAME, (S_IFDIR|(MODE)), \
112 &proc_##OTYPE##_inode_operations, &proc_##OTYPE##_operations, \
113 {} )
114 #define LNK(NAME, OTYPE) \
115 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
116 &proc_pid_link_inode_operations, NULL, \
117 { .proc_get_link = &proc_##OTYPE##_link } )
118 #define REG(NAME, MODE, OTYPE) \
119 NOD(NAME, (S_IFREG|(MODE)), NULL, \
120 &proc_##OTYPE##_operations, {})
121 #define INF(NAME, MODE, OTYPE) \
122 NOD(NAME, (S_IFREG|(MODE)), \
123 NULL, &proc_info_file_operations, \
124 { .proc_read = &proc_##OTYPE } )
125
126 int maps_protect;
127 EXPORT_SYMBOL(maps_protect);
128
129 static struct fs_struct *get_fs_struct(struct task_struct *task)
130 {
131 struct fs_struct *fs;
132 task_lock(task);
133 fs = task->fs;
134 if(fs)
135 atomic_inc(&fs->count);
136 task_unlock(task);
137 return fs;
138 }
139
140 static int get_nr_threads(struct task_struct *tsk)
141 {
142 /* Must be called with the rcu_read_lock held */
143 unsigned long flags;
144 int count = 0;
145
146 if (lock_task_sighand(tsk, &flags)) {
147 count = atomic_read(&tsk->signal->count);
148 unlock_task_sighand(tsk, &flags);
149 }
150 return count;
151 }
152
153 static int proc_cwd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
154 {
155 struct task_struct *task = get_proc_task(inode);
156 struct fs_struct *fs = NULL;
157 int result = -ENOENT;
158
159 if (task) {
160 fs = get_fs_struct(task);
161 put_task_struct(task);
162 }
163 if (fs) {
164 read_lock(&fs->lock);
165 *mnt = mntget(fs->pwdmnt);
166 *dentry = dget(fs->pwd);
167 read_unlock(&fs->lock);
168 result = 0;
169 put_fs_struct(fs);
170 }
171 return result;
172 }
173
174 static int proc_root_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
175 {
176 struct task_struct *task = get_proc_task(inode);
177 struct fs_struct *fs = NULL;
178 int result = -ENOENT;
179
180 if (task) {
181 fs = get_fs_struct(task);
182 put_task_struct(task);
183 }
184 if (fs) {
185 read_lock(&fs->lock);
186 *mnt = mntget(fs->rootmnt);
187 *dentry = dget(fs->root);
188 read_unlock(&fs->lock);
189 result = 0;
190 put_fs_struct(fs);
191 }
192 return result;
193 }
194
195 #define MAY_PTRACE(task) \
196 (task == current || \
197 (task->parent == current && \
198 (task->ptrace & PT_PTRACED) && \
199 (task->state == TASK_STOPPED || task->state == TASK_TRACED) && \
200 security_ptrace(current,task) == 0))
201
202 static int proc_pid_environ(struct task_struct *task, char * buffer)
203 {
204 int res = 0;
205 struct mm_struct *mm = get_task_mm(task);
206 if (mm) {
207 unsigned int len = mm->env_end - mm->env_start;
208 if (len > PAGE_SIZE)
209 len = PAGE_SIZE;
210 res = access_process_vm(task, mm->env_start, buffer, len, 0);
211 if (!ptrace_may_attach(task))
212 res = -ESRCH;
213 mmput(mm);
214 }
215 return res;
216 }
217
218 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
219 {
220 int res = 0;
221 unsigned int len;
222 struct mm_struct *mm = get_task_mm(task);
223 if (!mm)
224 goto out;
225 if (!mm->arg_end)
226 goto out_mm; /* Shh! No looking before we're done */
227
228 len = mm->arg_end - mm->arg_start;
229
230 if (len > PAGE_SIZE)
231 len = PAGE_SIZE;
232
233 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
234
235 // If the nul at the end of args has been overwritten, then
236 // assume application is using setproctitle(3).
237 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
238 len = strnlen(buffer, res);
239 if (len < res) {
240 res = len;
241 } else {
242 len = mm->env_end - mm->env_start;
243 if (len > PAGE_SIZE - res)
244 len = PAGE_SIZE - res;
245 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
246 res = strnlen(buffer, res);
247 }
248 }
249 out_mm:
250 mmput(mm);
251 out:
252 return res;
253 }
254
255 static int proc_pid_auxv(struct task_struct *task, char *buffer)
256 {
257 int res = 0;
258 struct mm_struct *mm = get_task_mm(task);
259 if (mm) {
260 unsigned int nwords = 0;
261 do
262 nwords += 2;
263 while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
264 res = nwords * sizeof(mm->saved_auxv[0]);
265 if (res > PAGE_SIZE)
266 res = PAGE_SIZE;
267 memcpy(buffer, mm->saved_auxv, res);
268 mmput(mm);
269 }
270 return res;
271 }
272
273
274 #ifdef CONFIG_KALLSYMS
275 /*
276 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
277 * Returns the resolved symbol. If that fails, simply return the address.
278 */
279 static int proc_pid_wchan(struct task_struct *task, char *buffer)
280 {
281 unsigned long wchan;
282 char symname[KSYM_NAME_LEN+1];
283
284 wchan = get_wchan(task);
285
286 if (lookup_symbol_name(wchan, symname) < 0)
287 return sprintf(buffer, "%lu", wchan);
288 else
289 return sprintf(buffer, "%s", symname);
290 }
291 #endif /* CONFIG_KALLSYMS */
292
293 #ifdef CONFIG_SCHEDSTATS
294 /*
295 * Provides /proc/PID/schedstat
296 */
297 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
298 {
299 return sprintf(buffer, "%lu %lu %lu\n",
300 task->sched_info.cpu_time,
301 task->sched_info.run_delay,
302 task->sched_info.pcnt);
303 }
304 #endif
305
306 /* The badness from the OOM killer */
307 unsigned long badness(struct task_struct *p, unsigned long uptime);
308 static int proc_oom_score(struct task_struct *task, char *buffer)
309 {
310 unsigned long points;
311 struct timespec uptime;
312
313 do_posix_clock_monotonic_gettime(&uptime);
314 read_lock(&tasklist_lock);
315 points = badness(task, uptime.tv_sec);
316 read_unlock(&tasklist_lock);
317 return sprintf(buffer, "%lu\n", points);
318 }
319
320 /************************************************************************/
321 /* Here the fs part begins */
322 /************************************************************************/
323
324 /* permission checks */
325 static int proc_fd_access_allowed(struct inode *inode)
326 {
327 struct task_struct *task;
328 int allowed = 0;
329 /* Allow access to a task's file descriptors if it is us or we
330 * may use ptrace attach to the process and find out that
331 * information.
332 */
333 task = get_proc_task(inode);
334 if (task) {
335 allowed = ptrace_may_attach(task);
336 put_task_struct(task);
337 }
338 return allowed;
339 }
340
341 static int proc_setattr(struct dentry *dentry, struct iattr *attr)
342 {
343 int error;
344 struct inode *inode = dentry->d_inode;
345
346 if (attr->ia_valid & ATTR_MODE)
347 return -EPERM;
348
349 error = inode_change_ok(inode, attr);
350 if (!error)
351 error = inode_setattr(inode, attr);
352 return error;
353 }
354
355 static const struct inode_operations proc_def_inode_operations = {
356 .setattr = proc_setattr,
357 };
358
359 extern struct seq_operations mounts_op;
360 struct proc_mounts {
361 struct seq_file m;
362 int event;
363 };
364
365 static int mounts_open(struct inode *inode, struct file *file)
366 {
367 struct task_struct *task = get_proc_task(inode);
368 struct mnt_namespace *ns = NULL;
369 struct proc_mounts *p;
370 int ret = -EINVAL;
371
372 if (task) {
373 task_lock(task);
374 if (task->nsproxy) {
375 ns = task->nsproxy->mnt_ns;
376 if (ns)
377 get_mnt_ns(ns);
378 }
379 task_unlock(task);
380 put_task_struct(task);
381 }
382
383 if (ns) {
384 ret = -ENOMEM;
385 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
386 if (p) {
387 file->private_data = &p->m;
388 ret = seq_open(file, &mounts_op);
389 if (!ret) {
390 p->m.private = ns;
391 p->event = ns->event;
392 return 0;
393 }
394 kfree(p);
395 }
396 put_mnt_ns(ns);
397 }
398 return ret;
399 }
400
401 static int mounts_release(struct inode *inode, struct file *file)
402 {
403 struct seq_file *m = file->private_data;
404 struct mnt_namespace *ns = m->private;
405 put_mnt_ns(ns);
406 return seq_release(inode, file);
407 }
408
409 static unsigned mounts_poll(struct file *file, poll_table *wait)
410 {
411 struct proc_mounts *p = file->private_data;
412 struct mnt_namespace *ns = p->m.private;
413 unsigned res = 0;
414
415 poll_wait(file, &ns->poll, wait);
416
417 spin_lock(&vfsmount_lock);
418 if (p->event != ns->event) {
419 p->event = ns->event;
420 res = POLLERR;
421 }
422 spin_unlock(&vfsmount_lock);
423
424 return res;
425 }
426
427 static const struct file_operations proc_mounts_operations = {
428 .open = mounts_open,
429 .read = seq_read,
430 .llseek = seq_lseek,
431 .release = mounts_release,
432 .poll = mounts_poll,
433 };
434
435 extern struct seq_operations mountstats_op;
436 static int mountstats_open(struct inode *inode, struct file *file)
437 {
438 int ret = seq_open(file, &mountstats_op);
439
440 if (!ret) {
441 struct seq_file *m = file->private_data;
442 struct mnt_namespace *mnt_ns = NULL;
443 struct task_struct *task = get_proc_task(inode);
444
445 if (task) {
446 task_lock(task);
447 if (task->nsproxy)
448 mnt_ns = task->nsproxy->mnt_ns;
449 if (mnt_ns)
450 get_mnt_ns(mnt_ns);
451 task_unlock(task);
452 put_task_struct(task);
453 }
454
455 if (mnt_ns)
456 m->private = mnt_ns;
457 else {
458 seq_release(inode, file);
459 ret = -EINVAL;
460 }
461 }
462 return ret;
463 }
464
465 static const struct file_operations proc_mountstats_operations = {
466 .open = mountstats_open,
467 .read = seq_read,
468 .llseek = seq_lseek,
469 .release = mounts_release,
470 };
471
472 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
473
474 static ssize_t proc_info_read(struct file * file, char __user * buf,
475 size_t count, loff_t *ppos)
476 {
477 struct inode * inode = file->f_path.dentry->d_inode;
478 unsigned long page;
479 ssize_t length;
480 struct task_struct *task = get_proc_task(inode);
481
482 length = -ESRCH;
483 if (!task)
484 goto out_no_task;
485
486 if (count > PROC_BLOCK_SIZE)
487 count = PROC_BLOCK_SIZE;
488
489 length = -ENOMEM;
490 if (!(page = __get_free_page(GFP_KERNEL)))
491 goto out;
492
493 length = PROC_I(inode)->op.proc_read(task, (char*)page);
494
495 if (length >= 0)
496 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
497 free_page(page);
498 out:
499 put_task_struct(task);
500 out_no_task:
501 return length;
502 }
503
504 static const struct file_operations proc_info_file_operations = {
505 .read = proc_info_read,
506 };
507
508 static int mem_open(struct inode* inode, struct file* file)
509 {
510 file->private_data = (void*)((long)current->self_exec_id);
511 return 0;
512 }
513
514 static ssize_t mem_read(struct file * file, char __user * buf,
515 size_t count, loff_t *ppos)
516 {
517 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
518 char *page;
519 unsigned long src = *ppos;
520 int ret = -ESRCH;
521 struct mm_struct *mm;
522
523 if (!task)
524 goto out_no_task;
525
526 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
527 goto out;
528
529 ret = -ENOMEM;
530 page = (char *)__get_free_page(GFP_USER);
531 if (!page)
532 goto out;
533
534 ret = 0;
535
536 mm = get_task_mm(task);
537 if (!mm)
538 goto out_free;
539
540 ret = -EIO;
541
542 if (file->private_data != (void*)((long)current->self_exec_id))
543 goto out_put;
544
545 ret = 0;
546
547 while (count > 0) {
548 int this_len, retval;
549
550 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
551 retval = access_process_vm(task, src, page, this_len, 0);
552 if (!retval || !MAY_PTRACE(task) || !ptrace_may_attach(task)) {
553 if (!ret)
554 ret = -EIO;
555 break;
556 }
557
558 if (copy_to_user(buf, page, retval)) {
559 ret = -EFAULT;
560 break;
561 }
562
563 ret += retval;
564 src += retval;
565 buf += retval;
566 count -= retval;
567 }
568 *ppos = src;
569
570 out_put:
571 mmput(mm);
572 out_free:
573 free_page((unsigned long) page);
574 out:
575 put_task_struct(task);
576 out_no_task:
577 return ret;
578 }
579
580 #define mem_write NULL
581
582 #ifndef mem_write
583 /* This is a security hazard */
584 static ssize_t mem_write(struct file * file, const char __user *buf,
585 size_t count, loff_t *ppos)
586 {
587 int copied;
588 char *page;
589 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
590 unsigned long dst = *ppos;
591
592 copied = -ESRCH;
593 if (!task)
594 goto out_no_task;
595
596 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
597 goto out;
598
599 copied = -ENOMEM;
600 page = (char *)__get_free_page(GFP_USER);
601 if (!page)
602 goto out;
603
604 copied = 0;
605 while (count > 0) {
606 int this_len, retval;
607
608 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
609 if (copy_from_user(page, buf, this_len)) {
610 copied = -EFAULT;
611 break;
612 }
613 retval = access_process_vm(task, dst, page, this_len, 1);
614 if (!retval) {
615 if (!copied)
616 copied = -EIO;
617 break;
618 }
619 copied += retval;
620 buf += retval;
621 dst += retval;
622 count -= retval;
623 }
624 *ppos = dst;
625 free_page((unsigned long) page);
626 out:
627 put_task_struct(task);
628 out_no_task:
629 return copied;
630 }
631 #endif
632
633 static loff_t mem_lseek(struct file * file, loff_t offset, int orig)
634 {
635 switch (orig) {
636 case 0:
637 file->f_pos = offset;
638 break;
639 case 1:
640 file->f_pos += offset;
641 break;
642 default:
643 return -EINVAL;
644 }
645 force_successful_syscall_return();
646 return file->f_pos;
647 }
648
649 static const struct file_operations proc_mem_operations = {
650 .llseek = mem_lseek,
651 .read = mem_read,
652 .write = mem_write,
653 .open = mem_open,
654 };
655
656 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
657 size_t count, loff_t *ppos)
658 {
659 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
660 char buffer[PROC_NUMBUF];
661 size_t len;
662 int oom_adjust;
663 loff_t __ppos = *ppos;
664
665 if (!task)
666 return -ESRCH;
667 oom_adjust = task->oomkilladj;
668 put_task_struct(task);
669
670 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
671 if (__ppos >= len)
672 return 0;
673 if (count > len-__ppos)
674 count = len-__ppos;
675 if (copy_to_user(buf, buffer + __ppos, count))
676 return -EFAULT;
677 *ppos = __ppos + count;
678 return count;
679 }
680
681 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
682 size_t count, loff_t *ppos)
683 {
684 struct task_struct *task;
685 char buffer[PROC_NUMBUF], *end;
686 int oom_adjust;
687
688 memset(buffer, 0, sizeof(buffer));
689 if (count > sizeof(buffer) - 1)
690 count = sizeof(buffer) - 1;
691 if (copy_from_user(buffer, buf, count))
692 return -EFAULT;
693 oom_adjust = simple_strtol(buffer, &end, 0);
694 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
695 oom_adjust != OOM_DISABLE)
696 return -EINVAL;
697 if (*end == '\n')
698 end++;
699 task = get_proc_task(file->f_path.dentry->d_inode);
700 if (!task)
701 return -ESRCH;
702 if (oom_adjust < task->oomkilladj && !capable(CAP_SYS_RESOURCE)) {
703 put_task_struct(task);
704 return -EACCES;
705 }
706 task->oomkilladj = oom_adjust;
707 put_task_struct(task);
708 if (end - buffer == 0)
709 return -EIO;
710 return end - buffer;
711 }
712
713 static const struct file_operations proc_oom_adjust_operations = {
714 .read = oom_adjust_read,
715 .write = oom_adjust_write,
716 };
717
718 static ssize_t clear_refs_write(struct file *file, const char __user *buf,
719 size_t count, loff_t *ppos)
720 {
721 struct task_struct *task;
722 char buffer[PROC_NUMBUF], *end;
723 struct mm_struct *mm;
724
725 memset(buffer, 0, sizeof(buffer));
726 if (count > sizeof(buffer) - 1)
727 count = sizeof(buffer) - 1;
728 if (copy_from_user(buffer, buf, count))
729 return -EFAULT;
730 if (!simple_strtol(buffer, &end, 0))
731 return -EINVAL;
732 if (*end == '\n')
733 end++;
734 task = get_proc_task(file->f_path.dentry->d_inode);
735 if (!task)
736 return -ESRCH;
737 mm = get_task_mm(task);
738 if (mm) {
739 clear_refs_smap(mm);
740 mmput(mm);
741 }
742 put_task_struct(task);
743 if (end - buffer == 0)
744 return -EIO;
745 return end - buffer;
746 }
747
748 static struct file_operations proc_clear_refs_operations = {
749 .write = clear_refs_write,
750 };
751
752 #ifdef CONFIG_AUDITSYSCALL
753 #define TMPBUFLEN 21
754 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
755 size_t count, loff_t *ppos)
756 {
757 struct inode * inode = file->f_path.dentry->d_inode;
758 struct task_struct *task = get_proc_task(inode);
759 ssize_t length;
760 char tmpbuf[TMPBUFLEN];
761
762 if (!task)
763 return -ESRCH;
764 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
765 audit_get_loginuid(task->audit_context));
766 put_task_struct(task);
767 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
768 }
769
770 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
771 size_t count, loff_t *ppos)
772 {
773 struct inode * inode = file->f_path.dentry->d_inode;
774 char *page, *tmp;
775 ssize_t length;
776 uid_t loginuid;
777
778 if (!capable(CAP_AUDIT_CONTROL))
779 return -EPERM;
780
781 if (current != pid_task(proc_pid(inode), PIDTYPE_PID))
782 return -EPERM;
783
784 if (count >= PAGE_SIZE)
785 count = PAGE_SIZE - 1;
786
787 if (*ppos != 0) {
788 /* No partial writes. */
789 return -EINVAL;
790 }
791 page = (char*)__get_free_page(GFP_USER);
792 if (!page)
793 return -ENOMEM;
794 length = -EFAULT;
795 if (copy_from_user(page, buf, count))
796 goto out_free_page;
797
798 page[count] = '\0';
799 loginuid = simple_strtoul(page, &tmp, 10);
800 if (tmp == page) {
801 length = -EINVAL;
802 goto out_free_page;
803
804 }
805 length = audit_set_loginuid(current, loginuid);
806 if (likely(length == 0))
807 length = count;
808
809 out_free_page:
810 free_page((unsigned long) page);
811 return length;
812 }
813
814 static const struct file_operations proc_loginuid_operations = {
815 .read = proc_loginuid_read,
816 .write = proc_loginuid_write,
817 };
818 #endif
819
820 #ifdef CONFIG_SECCOMP
821 static ssize_t seccomp_read(struct file *file, char __user *buf,
822 size_t count, loff_t *ppos)
823 {
824 struct task_struct *tsk = get_proc_task(file->f_dentry->d_inode);
825 char __buf[20];
826 loff_t __ppos = *ppos;
827 size_t len;
828
829 if (!tsk)
830 return -ESRCH;
831 /* no need to print the trailing zero, so use only len */
832 len = sprintf(__buf, "%u\n", tsk->seccomp.mode);
833 put_task_struct(tsk);
834 if (__ppos >= len)
835 return 0;
836 if (count > len - __ppos)
837 count = len - __ppos;
838 if (copy_to_user(buf, __buf + __ppos, count))
839 return -EFAULT;
840 *ppos = __ppos + count;
841 return count;
842 }
843
844 static ssize_t seccomp_write(struct file *file, const char __user *buf,
845 size_t count, loff_t *ppos)
846 {
847 struct task_struct *tsk = get_proc_task(file->f_dentry->d_inode);
848 char __buf[20], *end;
849 unsigned int seccomp_mode;
850 ssize_t result;
851
852 result = -ESRCH;
853 if (!tsk)
854 goto out_no_task;
855
856 /* can set it only once to be even more secure */
857 result = -EPERM;
858 if (unlikely(tsk->seccomp.mode))
859 goto out;
860
861 result = -EFAULT;
862 memset(__buf, 0, sizeof(__buf));
863 count = min(count, sizeof(__buf) - 1);
864 if (copy_from_user(__buf, buf, count))
865 goto out;
866
867 seccomp_mode = simple_strtoul(__buf, &end, 0);
868 if (*end == '\n')
869 end++;
870 result = -EINVAL;
871 if (seccomp_mode && seccomp_mode <= NR_SECCOMP_MODES) {
872 tsk->seccomp.mode = seccomp_mode;
873 set_tsk_thread_flag(tsk, TIF_SECCOMP);
874 } else
875 goto out;
876 result = -EIO;
877 if (unlikely(!(end - __buf)))
878 goto out;
879 result = end - __buf;
880 out:
881 put_task_struct(tsk);
882 out_no_task:
883 return result;
884 }
885
886 static const struct file_operations proc_seccomp_operations = {
887 .read = seccomp_read,
888 .write = seccomp_write,
889 };
890 #endif /* CONFIG_SECCOMP */
891
892 #ifdef CONFIG_FAULT_INJECTION
893 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
894 size_t count, loff_t *ppos)
895 {
896 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
897 char buffer[PROC_NUMBUF];
898 size_t len;
899 int make_it_fail;
900 loff_t __ppos = *ppos;
901
902 if (!task)
903 return -ESRCH;
904 make_it_fail = task->make_it_fail;
905 put_task_struct(task);
906
907 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
908 if (__ppos >= len)
909 return 0;
910 if (count > len-__ppos)
911 count = len-__ppos;
912 if (copy_to_user(buf, buffer + __ppos, count))
913 return -EFAULT;
914 *ppos = __ppos + count;
915 return count;
916 }
917
918 static ssize_t proc_fault_inject_write(struct file * file,
919 const char __user * buf, size_t count, loff_t *ppos)
920 {
921 struct task_struct *task;
922 char buffer[PROC_NUMBUF], *end;
923 int make_it_fail;
924
925 if (!capable(CAP_SYS_RESOURCE))
926 return -EPERM;
927 memset(buffer, 0, sizeof(buffer));
928 if (count > sizeof(buffer) - 1)
929 count = sizeof(buffer) - 1;
930 if (copy_from_user(buffer, buf, count))
931 return -EFAULT;
932 make_it_fail = simple_strtol(buffer, &end, 0);
933 if (*end == '\n')
934 end++;
935 task = get_proc_task(file->f_dentry->d_inode);
936 if (!task)
937 return -ESRCH;
938 task->make_it_fail = make_it_fail;
939 put_task_struct(task);
940 if (end - buffer == 0)
941 return -EIO;
942 return end - buffer;
943 }
944
945 static const struct file_operations proc_fault_inject_operations = {
946 .read = proc_fault_inject_read,
947 .write = proc_fault_inject_write,
948 };
949 #endif
950
951 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
952 {
953 struct inode *inode = dentry->d_inode;
954 int error = -EACCES;
955
956 /* We don't need a base pointer in the /proc filesystem */
957 path_release(nd);
958
959 /* Are we allowed to snoop on the tasks file descriptors? */
960 if (!proc_fd_access_allowed(inode))
961 goto out;
962
963 error = PROC_I(inode)->op.proc_get_link(inode, &nd->dentry, &nd->mnt);
964 nd->last_type = LAST_BIND;
965 out:
966 return ERR_PTR(error);
967 }
968
969 static int do_proc_readlink(struct dentry *dentry, struct vfsmount *mnt,
970 char __user *buffer, int buflen)
971 {
972 struct inode * inode;
973 char *tmp = (char*)__get_free_page(GFP_KERNEL), *path;
974 int len;
975
976 if (!tmp)
977 return -ENOMEM;
978
979 inode = dentry->d_inode;
980 path = d_path(dentry, mnt, tmp, PAGE_SIZE);
981 len = PTR_ERR(path);
982 if (IS_ERR(path))
983 goto out;
984 len = tmp + PAGE_SIZE - 1 - path;
985
986 if (len > buflen)
987 len = buflen;
988 if (copy_to_user(buffer, path, len))
989 len = -EFAULT;
990 out:
991 free_page((unsigned long)tmp);
992 return len;
993 }
994
995 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
996 {
997 int error = -EACCES;
998 struct inode *inode = dentry->d_inode;
999 struct dentry *de;
1000 struct vfsmount *mnt = NULL;
1001
1002 /* Are we allowed to snoop on the tasks file descriptors? */
1003 if (!proc_fd_access_allowed(inode))
1004 goto out;
1005
1006 error = PROC_I(inode)->op.proc_get_link(inode, &de, &mnt);
1007 if (error)
1008 goto out;
1009
1010 error = do_proc_readlink(de, mnt, buffer, buflen);
1011 dput(de);
1012 mntput(mnt);
1013 out:
1014 return error;
1015 }
1016
1017 static const struct inode_operations proc_pid_link_inode_operations = {
1018 .readlink = proc_pid_readlink,
1019 .follow_link = proc_pid_follow_link,
1020 .setattr = proc_setattr,
1021 };
1022
1023
1024 /* building an inode */
1025
1026 static int task_dumpable(struct task_struct *task)
1027 {
1028 int dumpable = 0;
1029 struct mm_struct *mm;
1030
1031 task_lock(task);
1032 mm = task->mm;
1033 if (mm)
1034 dumpable = mm->dumpable;
1035 task_unlock(task);
1036 if(dumpable == 1)
1037 return 1;
1038 return 0;
1039 }
1040
1041
1042 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1043 {
1044 struct inode * inode;
1045 struct proc_inode *ei;
1046
1047 /* We need a new inode */
1048
1049 inode = new_inode(sb);
1050 if (!inode)
1051 goto out;
1052
1053 /* Common stuff */
1054 ei = PROC_I(inode);
1055 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1056 inode->i_op = &proc_def_inode_operations;
1057
1058 /*
1059 * grab the reference to task.
1060 */
1061 ei->pid = get_task_pid(task, PIDTYPE_PID);
1062 if (!ei->pid)
1063 goto out_unlock;
1064
1065 inode->i_uid = 0;
1066 inode->i_gid = 0;
1067 if (task_dumpable(task)) {
1068 inode->i_uid = task->euid;
1069 inode->i_gid = task->egid;
1070 }
1071 security_task_to_inode(task, inode);
1072
1073 out:
1074 return inode;
1075
1076 out_unlock:
1077 iput(inode);
1078 return NULL;
1079 }
1080
1081 static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1082 {
1083 struct inode *inode = dentry->d_inode;
1084 struct task_struct *task;
1085 generic_fillattr(inode, stat);
1086
1087 rcu_read_lock();
1088 stat->uid = 0;
1089 stat->gid = 0;
1090 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1091 if (task) {
1092 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1093 task_dumpable(task)) {
1094 stat->uid = task->euid;
1095 stat->gid = task->egid;
1096 }
1097 }
1098 rcu_read_unlock();
1099 return 0;
1100 }
1101
1102 /* dentry stuff */
1103
1104 /*
1105 * Exceptional case: normally we are not allowed to unhash a busy
1106 * directory. In this case, however, we can do it - no aliasing problems
1107 * due to the way we treat inodes.
1108 *
1109 * Rewrite the inode's ownerships here because the owning task may have
1110 * performed a setuid(), etc.
1111 *
1112 * Before the /proc/pid/status file was created the only way to read
1113 * the effective uid of a /process was to stat /proc/pid. Reading
1114 * /proc/pid/status is slow enough that procps and other packages
1115 * kept stating /proc/pid. To keep the rules in /proc simple I have
1116 * made this apply to all per process world readable and executable
1117 * directories.
1118 */
1119 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1120 {
1121 struct inode *inode = dentry->d_inode;
1122 struct task_struct *task = get_proc_task(inode);
1123 if (task) {
1124 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1125 task_dumpable(task)) {
1126 inode->i_uid = task->euid;
1127 inode->i_gid = task->egid;
1128 } else {
1129 inode->i_uid = 0;
1130 inode->i_gid = 0;
1131 }
1132 inode->i_mode &= ~(S_ISUID | S_ISGID);
1133 security_task_to_inode(task, inode);
1134 put_task_struct(task);
1135 return 1;
1136 }
1137 d_drop(dentry);
1138 return 0;
1139 }
1140
1141 static int pid_delete_dentry(struct dentry * dentry)
1142 {
1143 /* Is the task we represent dead?
1144 * If so, then don't put the dentry on the lru list,
1145 * kill it immediately.
1146 */
1147 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1148 }
1149
1150 static struct dentry_operations pid_dentry_operations =
1151 {
1152 .d_revalidate = pid_revalidate,
1153 .d_delete = pid_delete_dentry,
1154 };
1155
1156 /* Lookups */
1157
1158 typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1159 struct task_struct *, const void *);
1160
1161 /*
1162 * Fill a directory entry.
1163 *
1164 * If possible create the dcache entry and derive our inode number and
1165 * file type from dcache entry.
1166 *
1167 * Since all of the proc inode numbers are dynamically generated, the inode
1168 * numbers do not exist until the inode is cache. This means creating the
1169 * the dcache entry in readdir is necessary to keep the inode numbers
1170 * reported by readdir in sync with the inode numbers reported
1171 * by stat.
1172 */
1173 static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1174 char *name, int len,
1175 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1176 {
1177 struct dentry *child, *dir = filp->f_path.dentry;
1178 struct inode *inode;
1179 struct qstr qname;
1180 ino_t ino = 0;
1181 unsigned type = DT_UNKNOWN;
1182
1183 qname.name = name;
1184 qname.len = len;
1185 qname.hash = full_name_hash(name, len);
1186
1187 child = d_lookup(dir, &qname);
1188 if (!child) {
1189 struct dentry *new;
1190 new = d_alloc(dir, &qname);
1191 if (new) {
1192 child = instantiate(dir->d_inode, new, task, ptr);
1193 if (child)
1194 dput(new);
1195 else
1196 child = new;
1197 }
1198 }
1199 if (!child || IS_ERR(child) || !child->d_inode)
1200 goto end_instantiate;
1201 inode = child->d_inode;
1202 if (inode) {
1203 ino = inode->i_ino;
1204 type = inode->i_mode >> 12;
1205 }
1206 dput(child);
1207 end_instantiate:
1208 if (!ino)
1209 ino = find_inode_number(dir, &qname);
1210 if (!ino)
1211 ino = 1;
1212 return filldir(dirent, name, len, filp->f_pos, ino, type);
1213 }
1214
1215 static unsigned name_to_int(struct dentry *dentry)
1216 {
1217 const char *name = dentry->d_name.name;
1218 int len = dentry->d_name.len;
1219 unsigned n = 0;
1220
1221 if (len > 1 && *name == '0')
1222 goto out;
1223 while (len-- > 0) {
1224 unsigned c = *name++ - '0';
1225 if (c > 9)
1226 goto out;
1227 if (n >= (~0U-9)/10)
1228 goto out;
1229 n *= 10;
1230 n += c;
1231 }
1232 return n;
1233 out:
1234 return ~0U;
1235 }
1236
1237 #define PROC_FDINFO_MAX 64
1238
1239 static int proc_fd_info(struct inode *inode, struct dentry **dentry,
1240 struct vfsmount **mnt, char *info)
1241 {
1242 struct task_struct *task = get_proc_task(inode);
1243 struct files_struct *files = NULL;
1244 struct file *file;
1245 int fd = proc_fd(inode);
1246
1247 if (task) {
1248 files = get_files_struct(task);
1249 put_task_struct(task);
1250 }
1251 if (files) {
1252 /*
1253 * We are not taking a ref to the file structure, so we must
1254 * hold ->file_lock.
1255 */
1256 spin_lock(&files->file_lock);
1257 file = fcheck_files(files, fd);
1258 if (file) {
1259 if (mnt)
1260 *mnt = mntget(file->f_path.mnt);
1261 if (dentry)
1262 *dentry = dget(file->f_path.dentry);
1263 if (info)
1264 snprintf(info, PROC_FDINFO_MAX,
1265 "pos:\t%lli\n"
1266 "flags:\t0%o\n",
1267 (long long) file->f_pos,
1268 file->f_flags);
1269 spin_unlock(&files->file_lock);
1270 put_files_struct(files);
1271 return 0;
1272 }
1273 spin_unlock(&files->file_lock);
1274 put_files_struct(files);
1275 }
1276 return -ENOENT;
1277 }
1278
1279 static int proc_fd_link(struct inode *inode, struct dentry **dentry,
1280 struct vfsmount **mnt)
1281 {
1282 return proc_fd_info(inode, dentry, mnt, NULL);
1283 }
1284
1285 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1286 {
1287 struct inode *inode = dentry->d_inode;
1288 struct task_struct *task = get_proc_task(inode);
1289 int fd = proc_fd(inode);
1290 struct files_struct *files;
1291
1292 if (task) {
1293 files = get_files_struct(task);
1294 if (files) {
1295 rcu_read_lock();
1296 if (fcheck_files(files, fd)) {
1297 rcu_read_unlock();
1298 put_files_struct(files);
1299 if (task_dumpable(task)) {
1300 inode->i_uid = task->euid;
1301 inode->i_gid = task->egid;
1302 } else {
1303 inode->i_uid = 0;
1304 inode->i_gid = 0;
1305 }
1306 inode->i_mode &= ~(S_ISUID | S_ISGID);
1307 security_task_to_inode(task, inode);
1308 put_task_struct(task);
1309 return 1;
1310 }
1311 rcu_read_unlock();
1312 put_files_struct(files);
1313 }
1314 put_task_struct(task);
1315 }
1316 d_drop(dentry);
1317 return 0;
1318 }
1319
1320 static struct dentry_operations tid_fd_dentry_operations =
1321 {
1322 .d_revalidate = tid_fd_revalidate,
1323 .d_delete = pid_delete_dentry,
1324 };
1325
1326 static struct dentry *proc_fd_instantiate(struct inode *dir,
1327 struct dentry *dentry, struct task_struct *task, const void *ptr)
1328 {
1329 unsigned fd = *(const unsigned *)ptr;
1330 struct file *file;
1331 struct files_struct *files;
1332 struct inode *inode;
1333 struct proc_inode *ei;
1334 struct dentry *error = ERR_PTR(-ENOENT);
1335
1336 inode = proc_pid_make_inode(dir->i_sb, task);
1337 if (!inode)
1338 goto out;
1339 ei = PROC_I(inode);
1340 ei->fd = fd;
1341 files = get_files_struct(task);
1342 if (!files)
1343 goto out_iput;
1344 inode->i_mode = S_IFLNK;
1345
1346 /*
1347 * We are not taking a ref to the file structure, so we must
1348 * hold ->file_lock.
1349 */
1350 spin_lock(&files->file_lock);
1351 file = fcheck_files(files, fd);
1352 if (!file)
1353 goto out_unlock;
1354 if (file->f_mode & 1)
1355 inode->i_mode |= S_IRUSR | S_IXUSR;
1356 if (file->f_mode & 2)
1357 inode->i_mode |= S_IWUSR | S_IXUSR;
1358 spin_unlock(&files->file_lock);
1359 put_files_struct(files);
1360
1361 inode->i_op = &proc_pid_link_inode_operations;
1362 inode->i_size = 64;
1363 ei->op.proc_get_link = proc_fd_link;
1364 dentry->d_op = &tid_fd_dentry_operations;
1365 d_add(dentry, inode);
1366 /* Close the race of the process dying before we return the dentry */
1367 if (tid_fd_revalidate(dentry, NULL))
1368 error = NULL;
1369
1370 out:
1371 return error;
1372 out_unlock:
1373 spin_unlock(&files->file_lock);
1374 put_files_struct(files);
1375 out_iput:
1376 iput(inode);
1377 goto out;
1378 }
1379
1380 static struct dentry *proc_lookupfd_common(struct inode *dir,
1381 struct dentry *dentry,
1382 instantiate_t instantiate)
1383 {
1384 struct task_struct *task = get_proc_task(dir);
1385 unsigned fd = name_to_int(dentry);
1386 struct dentry *result = ERR_PTR(-ENOENT);
1387
1388 if (!task)
1389 goto out_no_task;
1390 if (fd == ~0U)
1391 goto out;
1392
1393 result = instantiate(dir, dentry, task, &fd);
1394 out:
1395 put_task_struct(task);
1396 out_no_task:
1397 return result;
1398 }
1399
1400 static int proc_readfd_common(struct file * filp, void * dirent,
1401 filldir_t filldir, instantiate_t instantiate)
1402 {
1403 struct dentry *dentry = filp->f_path.dentry;
1404 struct inode *inode = dentry->d_inode;
1405 struct task_struct *p = get_proc_task(inode);
1406 unsigned int fd, tid, ino;
1407 int retval;
1408 struct files_struct * files;
1409 struct fdtable *fdt;
1410
1411 retval = -ENOENT;
1412 if (!p)
1413 goto out_no_task;
1414 retval = 0;
1415 tid = p->pid;
1416
1417 fd = filp->f_pos;
1418 switch (fd) {
1419 case 0:
1420 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1421 goto out;
1422 filp->f_pos++;
1423 case 1:
1424 ino = parent_ino(dentry);
1425 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1426 goto out;
1427 filp->f_pos++;
1428 default:
1429 files = get_files_struct(p);
1430 if (!files)
1431 goto out;
1432 rcu_read_lock();
1433 fdt = files_fdtable(files);
1434 for (fd = filp->f_pos-2;
1435 fd < fdt->max_fds;
1436 fd++, filp->f_pos++) {
1437 char name[PROC_NUMBUF];
1438 int len;
1439
1440 if (!fcheck_files(files, fd))
1441 continue;
1442 rcu_read_unlock();
1443
1444 len = snprintf(name, sizeof(name), "%d", fd);
1445 if (proc_fill_cache(filp, dirent, filldir,
1446 name, len, instantiate,
1447 p, &fd) < 0) {
1448 rcu_read_lock();
1449 break;
1450 }
1451 rcu_read_lock();
1452 }
1453 rcu_read_unlock();
1454 put_files_struct(files);
1455 }
1456 out:
1457 put_task_struct(p);
1458 out_no_task:
1459 return retval;
1460 }
1461
1462 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1463 struct nameidata *nd)
1464 {
1465 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1466 }
1467
1468 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1469 {
1470 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1471 }
1472
1473 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1474 size_t len, loff_t *ppos)
1475 {
1476 char tmp[PROC_FDINFO_MAX];
1477 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, NULL, tmp);
1478 if (!err)
1479 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1480 return err;
1481 }
1482
1483 static const struct file_operations proc_fdinfo_file_operations = {
1484 .open = nonseekable_open,
1485 .read = proc_fdinfo_read,
1486 };
1487
1488 static const struct file_operations proc_fd_operations = {
1489 .read = generic_read_dir,
1490 .readdir = proc_readfd,
1491 };
1492
1493 /*
1494 * /proc/pid/fd needs a special permission handler so that a process can still
1495 * access /proc/self/fd after it has executed a setuid().
1496 */
1497 static int proc_fd_permission(struct inode *inode, int mask,
1498 struct nameidata *nd)
1499 {
1500 int rv;
1501
1502 rv = generic_permission(inode, mask, NULL);
1503 if (rv == 0)
1504 return 0;
1505 if (task_pid(current) == proc_pid(inode))
1506 rv = 0;
1507 return rv;
1508 }
1509
1510 /*
1511 * proc directories can do almost nothing..
1512 */
1513 static const struct inode_operations proc_fd_inode_operations = {
1514 .lookup = proc_lookupfd,
1515 .permission = proc_fd_permission,
1516 .setattr = proc_setattr,
1517 };
1518
1519 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
1520 struct dentry *dentry, struct task_struct *task, const void *ptr)
1521 {
1522 unsigned fd = *(unsigned *)ptr;
1523 struct inode *inode;
1524 struct proc_inode *ei;
1525 struct dentry *error = ERR_PTR(-ENOENT);
1526
1527 inode = proc_pid_make_inode(dir->i_sb, task);
1528 if (!inode)
1529 goto out;
1530 ei = PROC_I(inode);
1531 ei->fd = fd;
1532 inode->i_mode = S_IFREG | S_IRUSR;
1533 inode->i_fop = &proc_fdinfo_file_operations;
1534 dentry->d_op = &tid_fd_dentry_operations;
1535 d_add(dentry, inode);
1536 /* Close the race of the process dying before we return the dentry */
1537 if (tid_fd_revalidate(dentry, NULL))
1538 error = NULL;
1539
1540 out:
1541 return error;
1542 }
1543
1544 static struct dentry *proc_lookupfdinfo(struct inode *dir,
1545 struct dentry *dentry,
1546 struct nameidata *nd)
1547 {
1548 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
1549 }
1550
1551 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
1552 {
1553 return proc_readfd_common(filp, dirent, filldir,
1554 proc_fdinfo_instantiate);
1555 }
1556
1557 static const struct file_operations proc_fdinfo_operations = {
1558 .read = generic_read_dir,
1559 .readdir = proc_readfdinfo,
1560 };
1561
1562 /*
1563 * proc directories can do almost nothing..
1564 */
1565 static const struct inode_operations proc_fdinfo_inode_operations = {
1566 .lookup = proc_lookupfdinfo,
1567 .setattr = proc_setattr,
1568 };
1569
1570
1571 static struct dentry *proc_pident_instantiate(struct inode *dir,
1572 struct dentry *dentry, struct task_struct *task, const void *ptr)
1573 {
1574 const struct pid_entry *p = ptr;
1575 struct inode *inode;
1576 struct proc_inode *ei;
1577 struct dentry *error = ERR_PTR(-EINVAL);
1578
1579 inode = proc_pid_make_inode(dir->i_sb, task);
1580 if (!inode)
1581 goto out;
1582
1583 ei = PROC_I(inode);
1584 inode->i_mode = p->mode;
1585 if (S_ISDIR(inode->i_mode))
1586 inode->i_nlink = 2; /* Use getattr to fix if necessary */
1587 if (p->iop)
1588 inode->i_op = p->iop;
1589 if (p->fop)
1590 inode->i_fop = p->fop;
1591 ei->op = p->op;
1592 dentry->d_op = &pid_dentry_operations;
1593 d_add(dentry, inode);
1594 /* Close the race of the process dying before we return the dentry */
1595 if (pid_revalidate(dentry, NULL))
1596 error = NULL;
1597 out:
1598 return error;
1599 }
1600
1601 static struct dentry *proc_pident_lookup(struct inode *dir,
1602 struct dentry *dentry,
1603 const struct pid_entry *ents,
1604 unsigned int nents)
1605 {
1606 struct inode *inode;
1607 struct dentry *error;
1608 struct task_struct *task = get_proc_task(dir);
1609 const struct pid_entry *p, *last;
1610
1611 error = ERR_PTR(-ENOENT);
1612 inode = NULL;
1613
1614 if (!task)
1615 goto out_no_task;
1616
1617 /*
1618 * Yes, it does not scale. And it should not. Don't add
1619 * new entries into /proc/<tgid>/ without very good reasons.
1620 */
1621 last = &ents[nents - 1];
1622 for (p = ents; p <= last; p++) {
1623 if (p->len != dentry->d_name.len)
1624 continue;
1625 if (!memcmp(dentry->d_name.name, p->name, p->len))
1626 break;
1627 }
1628 if (p > last)
1629 goto out;
1630
1631 error = proc_pident_instantiate(dir, dentry, task, p);
1632 out:
1633 put_task_struct(task);
1634 out_no_task:
1635 return error;
1636 }
1637
1638 static int proc_pident_fill_cache(struct file *filp, void *dirent,
1639 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
1640 {
1641 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
1642 proc_pident_instantiate, task, p);
1643 }
1644
1645 static int proc_pident_readdir(struct file *filp,
1646 void *dirent, filldir_t filldir,
1647 const struct pid_entry *ents, unsigned int nents)
1648 {
1649 int i;
1650 int pid;
1651 struct dentry *dentry = filp->f_path.dentry;
1652 struct inode *inode = dentry->d_inode;
1653 struct task_struct *task = get_proc_task(inode);
1654 const struct pid_entry *p, *last;
1655 ino_t ino;
1656 int ret;
1657
1658 ret = -ENOENT;
1659 if (!task)
1660 goto out_no_task;
1661
1662 ret = 0;
1663 pid = task->pid;
1664 i = filp->f_pos;
1665 switch (i) {
1666 case 0:
1667 ino = inode->i_ino;
1668 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
1669 goto out;
1670 i++;
1671 filp->f_pos++;
1672 /* fall through */
1673 case 1:
1674 ino = parent_ino(dentry);
1675 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
1676 goto out;
1677 i++;
1678 filp->f_pos++;
1679 /* fall through */
1680 default:
1681 i -= 2;
1682 if (i >= nents) {
1683 ret = 1;
1684 goto out;
1685 }
1686 p = ents + i;
1687 last = &ents[nents - 1];
1688 while (p <= last) {
1689 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
1690 goto out;
1691 filp->f_pos++;
1692 p++;
1693 }
1694 }
1695
1696 ret = 1;
1697 out:
1698 put_task_struct(task);
1699 out_no_task:
1700 return ret;
1701 }
1702
1703 #ifdef CONFIG_SECURITY
1704 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
1705 size_t count, loff_t *ppos)
1706 {
1707 struct inode * inode = file->f_path.dentry->d_inode;
1708 char *p = NULL;
1709 ssize_t length;
1710 struct task_struct *task = get_proc_task(inode);
1711
1712 if (!task)
1713 return -ESRCH;
1714
1715 length = security_getprocattr(task,
1716 (char*)file->f_path.dentry->d_name.name,
1717 &p);
1718 put_task_struct(task);
1719 if (length > 0)
1720 length = simple_read_from_buffer(buf, count, ppos, p, length);
1721 kfree(p);
1722 return length;
1723 }
1724
1725 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
1726 size_t count, loff_t *ppos)
1727 {
1728 struct inode * inode = file->f_path.dentry->d_inode;
1729 char *page;
1730 ssize_t length;
1731 struct task_struct *task = get_proc_task(inode);
1732
1733 length = -ESRCH;
1734 if (!task)
1735 goto out_no_task;
1736 if (count > PAGE_SIZE)
1737 count = PAGE_SIZE;
1738
1739 /* No partial writes. */
1740 length = -EINVAL;
1741 if (*ppos != 0)
1742 goto out;
1743
1744 length = -ENOMEM;
1745 page = (char*)__get_free_page(GFP_USER);
1746 if (!page)
1747 goto out;
1748
1749 length = -EFAULT;
1750 if (copy_from_user(page, buf, count))
1751 goto out_free;
1752
1753 length = security_setprocattr(task,
1754 (char*)file->f_path.dentry->d_name.name,
1755 (void*)page, count);
1756 out_free:
1757 free_page((unsigned long) page);
1758 out:
1759 put_task_struct(task);
1760 out_no_task:
1761 return length;
1762 }
1763
1764 static const struct file_operations proc_pid_attr_operations = {
1765 .read = proc_pid_attr_read,
1766 .write = proc_pid_attr_write,
1767 };
1768
1769 static const struct pid_entry attr_dir_stuff[] = {
1770 REG("current", S_IRUGO|S_IWUGO, pid_attr),
1771 REG("prev", S_IRUGO, pid_attr),
1772 REG("exec", S_IRUGO|S_IWUGO, pid_attr),
1773 REG("fscreate", S_IRUGO|S_IWUGO, pid_attr),
1774 REG("keycreate", S_IRUGO|S_IWUGO, pid_attr),
1775 REG("sockcreate", S_IRUGO|S_IWUGO, pid_attr),
1776 };
1777
1778 static int proc_attr_dir_readdir(struct file * filp,
1779 void * dirent, filldir_t filldir)
1780 {
1781 return proc_pident_readdir(filp,dirent,filldir,
1782 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
1783 }
1784
1785 static const struct file_operations proc_attr_dir_operations = {
1786 .read = generic_read_dir,
1787 .readdir = proc_attr_dir_readdir,
1788 };
1789
1790 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
1791 struct dentry *dentry, struct nameidata *nd)
1792 {
1793 return proc_pident_lookup(dir, dentry,
1794 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
1795 }
1796
1797 static const struct inode_operations proc_attr_dir_inode_operations = {
1798 .lookup = proc_attr_dir_lookup,
1799 .getattr = pid_getattr,
1800 .setattr = proc_setattr,
1801 };
1802
1803 #endif
1804
1805 /*
1806 * /proc/self:
1807 */
1808 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
1809 int buflen)
1810 {
1811 char tmp[PROC_NUMBUF];
1812 sprintf(tmp, "%d", current->tgid);
1813 return vfs_readlink(dentry,buffer,buflen,tmp);
1814 }
1815
1816 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
1817 {
1818 char tmp[PROC_NUMBUF];
1819 sprintf(tmp, "%d", current->tgid);
1820 return ERR_PTR(vfs_follow_link(nd,tmp));
1821 }
1822
1823 static const struct inode_operations proc_self_inode_operations = {
1824 .readlink = proc_self_readlink,
1825 .follow_link = proc_self_follow_link,
1826 };
1827
1828 /*
1829 * proc base
1830 *
1831 * These are the directory entries in the root directory of /proc
1832 * that properly belong to the /proc filesystem, as they describe
1833 * describe something that is process related.
1834 */
1835 static const struct pid_entry proc_base_stuff[] = {
1836 NOD("self", S_IFLNK|S_IRWXUGO,
1837 &proc_self_inode_operations, NULL, {}),
1838 };
1839
1840 /*
1841 * Exceptional case: normally we are not allowed to unhash a busy
1842 * directory. In this case, however, we can do it - no aliasing problems
1843 * due to the way we treat inodes.
1844 */
1845 static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
1846 {
1847 struct inode *inode = dentry->d_inode;
1848 struct task_struct *task = get_proc_task(inode);
1849 if (task) {
1850 put_task_struct(task);
1851 return 1;
1852 }
1853 d_drop(dentry);
1854 return 0;
1855 }
1856
1857 static struct dentry_operations proc_base_dentry_operations =
1858 {
1859 .d_revalidate = proc_base_revalidate,
1860 .d_delete = pid_delete_dentry,
1861 };
1862
1863 static struct dentry *proc_base_instantiate(struct inode *dir,
1864 struct dentry *dentry, struct task_struct *task, const void *ptr)
1865 {
1866 const struct pid_entry *p = ptr;
1867 struct inode *inode;
1868 struct proc_inode *ei;
1869 struct dentry *error = ERR_PTR(-EINVAL);
1870
1871 /* Allocate the inode */
1872 error = ERR_PTR(-ENOMEM);
1873 inode = new_inode(dir->i_sb);
1874 if (!inode)
1875 goto out;
1876
1877 /* Initialize the inode */
1878 ei = PROC_I(inode);
1879 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1880
1881 /*
1882 * grab the reference to the task.
1883 */
1884 ei->pid = get_task_pid(task, PIDTYPE_PID);
1885 if (!ei->pid)
1886 goto out_iput;
1887
1888 inode->i_uid = 0;
1889 inode->i_gid = 0;
1890 inode->i_mode = p->mode;
1891 if (S_ISDIR(inode->i_mode))
1892 inode->i_nlink = 2;
1893 if (S_ISLNK(inode->i_mode))
1894 inode->i_size = 64;
1895 if (p->iop)
1896 inode->i_op = p->iop;
1897 if (p->fop)
1898 inode->i_fop = p->fop;
1899 ei->op = p->op;
1900 dentry->d_op = &proc_base_dentry_operations;
1901 d_add(dentry, inode);
1902 error = NULL;
1903 out:
1904 return error;
1905 out_iput:
1906 iput(inode);
1907 goto out;
1908 }
1909
1910 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
1911 {
1912 struct dentry *error;
1913 struct task_struct *task = get_proc_task(dir);
1914 const struct pid_entry *p, *last;
1915
1916 error = ERR_PTR(-ENOENT);
1917
1918 if (!task)
1919 goto out_no_task;
1920
1921 /* Lookup the directory entry */
1922 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
1923 for (p = proc_base_stuff; p <= last; p++) {
1924 if (p->len != dentry->d_name.len)
1925 continue;
1926 if (!memcmp(dentry->d_name.name, p->name, p->len))
1927 break;
1928 }
1929 if (p > last)
1930 goto out;
1931
1932 error = proc_base_instantiate(dir, dentry, task, p);
1933
1934 out:
1935 put_task_struct(task);
1936 out_no_task:
1937 return error;
1938 }
1939
1940 static int proc_base_fill_cache(struct file *filp, void *dirent,
1941 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
1942 {
1943 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
1944 proc_base_instantiate, task, p);
1945 }
1946
1947 #ifdef CONFIG_TASK_IO_ACCOUNTING
1948 static int proc_pid_io_accounting(struct task_struct *task, char *buffer)
1949 {
1950 return sprintf(buffer,
1951 #ifdef CONFIG_TASK_XACCT
1952 "rchar: %llu\n"
1953 "wchar: %llu\n"
1954 "syscr: %llu\n"
1955 "syscw: %llu\n"
1956 #endif
1957 "read_bytes: %llu\n"
1958 "write_bytes: %llu\n"
1959 "cancelled_write_bytes: %llu\n",
1960 #ifdef CONFIG_TASK_XACCT
1961 (unsigned long long)task->rchar,
1962 (unsigned long long)task->wchar,
1963 (unsigned long long)task->syscr,
1964 (unsigned long long)task->syscw,
1965 #endif
1966 (unsigned long long)task->ioac.read_bytes,
1967 (unsigned long long)task->ioac.write_bytes,
1968 (unsigned long long)task->ioac.cancelled_write_bytes);
1969 }
1970 #endif
1971
1972 /*
1973 * Thread groups
1974 */
1975 static const struct file_operations proc_task_operations;
1976 static const struct inode_operations proc_task_inode_operations;
1977
1978 static const struct pid_entry tgid_base_stuff[] = {
1979 DIR("task", S_IRUGO|S_IXUGO, task),
1980 DIR("fd", S_IRUSR|S_IXUSR, fd),
1981 DIR("fdinfo", S_IRUSR|S_IXUSR, fdinfo),
1982 INF("environ", S_IRUSR, pid_environ),
1983 INF("auxv", S_IRUSR, pid_auxv),
1984 INF("status", S_IRUGO, pid_status),
1985 INF("cmdline", S_IRUGO, pid_cmdline),
1986 INF("stat", S_IRUGO, tgid_stat),
1987 INF("statm", S_IRUGO, pid_statm),
1988 REG("maps", S_IRUGO, maps),
1989 #ifdef CONFIG_NUMA
1990 REG("numa_maps", S_IRUGO, numa_maps),
1991 #endif
1992 REG("mem", S_IRUSR|S_IWUSR, mem),
1993 #ifdef CONFIG_SECCOMP
1994 REG("seccomp", S_IRUSR|S_IWUSR, seccomp),
1995 #endif
1996 LNK("cwd", cwd),
1997 LNK("root", root),
1998 LNK("exe", exe),
1999 REG("mounts", S_IRUGO, mounts),
2000 REG("mountstats", S_IRUSR, mountstats),
2001 #ifdef CONFIG_MMU
2002 REG("clear_refs", S_IWUSR, clear_refs),
2003 REG("smaps", S_IRUGO, smaps),
2004 #endif
2005 #ifdef CONFIG_SECURITY
2006 DIR("attr", S_IRUGO|S_IXUGO, attr_dir),
2007 #endif
2008 #ifdef CONFIG_KALLSYMS
2009 INF("wchan", S_IRUGO, pid_wchan),
2010 #endif
2011 #ifdef CONFIG_SCHEDSTATS
2012 INF("schedstat", S_IRUGO, pid_schedstat),
2013 #endif
2014 #ifdef CONFIG_CPUSETS
2015 REG("cpuset", S_IRUGO, cpuset),
2016 #endif
2017 INF("oom_score", S_IRUGO, oom_score),
2018 REG("oom_adj", S_IRUGO|S_IWUSR, oom_adjust),
2019 #ifdef CONFIG_AUDITSYSCALL
2020 REG("loginuid", S_IWUSR|S_IRUGO, loginuid),
2021 #endif
2022 #ifdef CONFIG_FAULT_INJECTION
2023 REG("make-it-fail", S_IRUGO|S_IWUSR, fault_inject),
2024 #endif
2025 #ifdef CONFIG_TASK_IO_ACCOUNTING
2026 INF("io", S_IRUGO, pid_io_accounting),
2027 #endif
2028 };
2029
2030 static int proc_tgid_base_readdir(struct file * filp,
2031 void * dirent, filldir_t filldir)
2032 {
2033 return proc_pident_readdir(filp,dirent,filldir,
2034 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2035 }
2036
2037 static const struct file_operations proc_tgid_base_operations = {
2038 .read = generic_read_dir,
2039 .readdir = proc_tgid_base_readdir,
2040 };
2041
2042 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2043 return proc_pident_lookup(dir, dentry,
2044 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2045 }
2046
2047 static const struct inode_operations proc_tgid_base_inode_operations = {
2048 .lookup = proc_tgid_base_lookup,
2049 .getattr = pid_getattr,
2050 .setattr = proc_setattr,
2051 };
2052
2053 /**
2054 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2055 *
2056 * @task: task that should be flushed.
2057 *
2058 * Looks in the dcache for
2059 * /proc/@pid
2060 * /proc/@tgid/task/@pid
2061 * if either directory is present flushes it and all of it'ts children
2062 * from the dcache.
2063 *
2064 * It is safe and reasonable to cache /proc entries for a task until
2065 * that task exits. After that they just clog up the dcache with
2066 * useless entries, possibly causing useful dcache entries to be
2067 * flushed instead. This routine is proved to flush those useless
2068 * dcache entries at process exit time.
2069 *
2070 * NOTE: This routine is just an optimization so it does not guarantee
2071 * that no dcache entries will exist at process exit time it
2072 * just makes it very unlikely that any will persist.
2073 */
2074 void proc_flush_task(struct task_struct *task)
2075 {
2076 struct dentry *dentry, *leader, *dir;
2077 char buf[PROC_NUMBUF];
2078 struct qstr name;
2079
2080 name.name = buf;
2081 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
2082 dentry = d_hash_and_lookup(proc_mnt->mnt_root, &name);
2083 if (dentry) {
2084 shrink_dcache_parent(dentry);
2085 d_drop(dentry);
2086 dput(dentry);
2087 }
2088
2089 if (thread_group_leader(task))
2090 goto out;
2091
2092 name.name = buf;
2093 name.len = snprintf(buf, sizeof(buf), "%d", task->tgid);
2094 leader = d_hash_and_lookup(proc_mnt->mnt_root, &name);
2095 if (!leader)
2096 goto out;
2097
2098 name.name = "task";
2099 name.len = strlen(name.name);
2100 dir = d_hash_and_lookup(leader, &name);
2101 if (!dir)
2102 goto out_put_leader;
2103
2104 name.name = buf;
2105 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
2106 dentry = d_hash_and_lookup(dir, &name);
2107 if (dentry) {
2108 shrink_dcache_parent(dentry);
2109 d_drop(dentry);
2110 dput(dentry);
2111 }
2112
2113 dput(dir);
2114 out_put_leader:
2115 dput(leader);
2116 out:
2117 return;
2118 }
2119
2120 static struct dentry *proc_pid_instantiate(struct inode *dir,
2121 struct dentry * dentry,
2122 struct task_struct *task, const void *ptr)
2123 {
2124 struct dentry *error = ERR_PTR(-ENOENT);
2125 struct inode *inode;
2126
2127 inode = proc_pid_make_inode(dir->i_sb, task);
2128 if (!inode)
2129 goto out;
2130
2131 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2132 inode->i_op = &proc_tgid_base_inode_operations;
2133 inode->i_fop = &proc_tgid_base_operations;
2134 inode->i_flags|=S_IMMUTABLE;
2135 inode->i_nlink = 5;
2136 #ifdef CONFIG_SECURITY
2137 inode->i_nlink += 1;
2138 #endif
2139
2140 dentry->d_op = &pid_dentry_operations;
2141
2142 d_add(dentry, inode);
2143 /* Close the race of the process dying before we return the dentry */
2144 if (pid_revalidate(dentry, NULL))
2145 error = NULL;
2146 out:
2147 return error;
2148 }
2149
2150 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2151 {
2152 struct dentry *result = ERR_PTR(-ENOENT);
2153 struct task_struct *task;
2154 unsigned tgid;
2155
2156 result = proc_base_lookup(dir, dentry);
2157 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2158 goto out;
2159
2160 tgid = name_to_int(dentry);
2161 if (tgid == ~0U)
2162 goto out;
2163
2164 rcu_read_lock();
2165 task = find_task_by_pid(tgid);
2166 if (task)
2167 get_task_struct(task);
2168 rcu_read_unlock();
2169 if (!task)
2170 goto out;
2171
2172 result = proc_pid_instantiate(dir, dentry, task, NULL);
2173 put_task_struct(task);
2174 out:
2175 return result;
2176 }
2177
2178 /*
2179 * Find the first task with tgid >= tgid
2180 *
2181 */
2182 static struct task_struct *next_tgid(unsigned int tgid)
2183 {
2184 struct task_struct *task;
2185 struct pid *pid;
2186
2187 rcu_read_lock();
2188 retry:
2189 task = NULL;
2190 pid = find_ge_pid(tgid);
2191 if (pid) {
2192 tgid = pid->nr + 1;
2193 task = pid_task(pid, PIDTYPE_PID);
2194 /* What we to know is if the pid we have find is the
2195 * pid of a thread_group_leader. Testing for task
2196 * being a thread_group_leader is the obvious thing
2197 * todo but there is a window when it fails, due to
2198 * the pid transfer logic in de_thread.
2199 *
2200 * So we perform the straight forward test of seeing
2201 * if the pid we have found is the pid of a thread
2202 * group leader, and don't worry if the task we have
2203 * found doesn't happen to be a thread group leader.
2204 * As we don't care in the case of readdir.
2205 */
2206 if (!task || !has_group_leader_pid(task))
2207 goto retry;
2208 get_task_struct(task);
2209 }
2210 rcu_read_unlock();
2211 return task;
2212 }
2213
2214 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
2215
2216 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2217 struct task_struct *task, int tgid)
2218 {
2219 char name[PROC_NUMBUF];
2220 int len = snprintf(name, sizeof(name), "%d", tgid);
2221 return proc_fill_cache(filp, dirent, filldir, name, len,
2222 proc_pid_instantiate, task, NULL);
2223 }
2224
2225 /* for the /proc/ directory itself, after non-process stuff has been done */
2226 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2227 {
2228 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2229 struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
2230 struct task_struct *task;
2231 int tgid;
2232
2233 if (!reaper)
2234 goto out_no_task;
2235
2236 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
2237 const struct pid_entry *p = &proc_base_stuff[nr];
2238 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
2239 goto out;
2240 }
2241
2242 tgid = filp->f_pos - TGID_OFFSET;
2243 for (task = next_tgid(tgid);
2244 task;
2245 put_task_struct(task), task = next_tgid(tgid + 1)) {
2246 tgid = task->pid;
2247 filp->f_pos = tgid + TGID_OFFSET;
2248 if (proc_pid_fill_cache(filp, dirent, filldir, task, tgid) < 0) {
2249 put_task_struct(task);
2250 goto out;
2251 }
2252 }
2253 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
2254 out:
2255 put_task_struct(reaper);
2256 out_no_task:
2257 return 0;
2258 }
2259
2260 /*
2261 * Tasks
2262 */
2263 static const struct pid_entry tid_base_stuff[] = {
2264 DIR("fd", S_IRUSR|S_IXUSR, fd),
2265 DIR("fdinfo", S_IRUSR|S_IXUSR, fdinfo),
2266 INF("environ", S_IRUSR, pid_environ),
2267 INF("auxv", S_IRUSR, pid_auxv),
2268 INF("status", S_IRUGO, pid_status),
2269 INF("cmdline", S_IRUGO, pid_cmdline),
2270 INF("stat", S_IRUGO, tid_stat),
2271 INF("statm", S_IRUGO, pid_statm),
2272 REG("maps", S_IRUGO, maps),
2273 #ifdef CONFIG_NUMA
2274 REG("numa_maps", S_IRUGO, numa_maps),
2275 #endif
2276 REG("mem", S_IRUSR|S_IWUSR, mem),
2277 #ifdef CONFIG_SECCOMP
2278 REG("seccomp", S_IRUSR|S_IWUSR, seccomp),
2279 #endif
2280 LNK("cwd", cwd),
2281 LNK("root", root),
2282 LNK("exe", exe),
2283 REG("mounts", S_IRUGO, mounts),
2284 #ifdef CONFIG_MMU
2285 REG("clear_refs", S_IWUSR, clear_refs),
2286 REG("smaps", S_IRUGO, smaps),
2287 #endif
2288 #ifdef CONFIG_SECURITY
2289 DIR("attr", S_IRUGO|S_IXUGO, attr_dir),
2290 #endif
2291 #ifdef CONFIG_KALLSYMS
2292 INF("wchan", S_IRUGO, pid_wchan),
2293 #endif
2294 #ifdef CONFIG_SCHEDSTATS
2295 INF("schedstat", S_IRUGO, pid_schedstat),
2296 #endif
2297 #ifdef CONFIG_CPUSETS
2298 REG("cpuset", S_IRUGO, cpuset),
2299 #endif
2300 INF("oom_score", S_IRUGO, oom_score),
2301 REG("oom_adj", S_IRUGO|S_IWUSR, oom_adjust),
2302 #ifdef CONFIG_AUDITSYSCALL
2303 REG("loginuid", S_IWUSR|S_IRUGO, loginuid),
2304 #endif
2305 #ifdef CONFIG_FAULT_INJECTION
2306 REG("make-it-fail", S_IRUGO|S_IWUSR, fault_inject),
2307 #endif
2308 };
2309
2310 static int proc_tid_base_readdir(struct file * filp,
2311 void * dirent, filldir_t filldir)
2312 {
2313 return proc_pident_readdir(filp,dirent,filldir,
2314 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
2315 }
2316
2317 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2318 return proc_pident_lookup(dir, dentry,
2319 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
2320 }
2321
2322 static const struct file_operations proc_tid_base_operations = {
2323 .read = generic_read_dir,
2324 .readdir = proc_tid_base_readdir,
2325 };
2326
2327 static const struct inode_operations proc_tid_base_inode_operations = {
2328 .lookup = proc_tid_base_lookup,
2329 .getattr = pid_getattr,
2330 .setattr = proc_setattr,
2331 };
2332
2333 static struct dentry *proc_task_instantiate(struct inode *dir,
2334 struct dentry *dentry, struct task_struct *task, const void *ptr)
2335 {
2336 struct dentry *error = ERR_PTR(-ENOENT);
2337 struct inode *inode;
2338 inode = proc_pid_make_inode(dir->i_sb, task);
2339
2340 if (!inode)
2341 goto out;
2342 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2343 inode->i_op = &proc_tid_base_inode_operations;
2344 inode->i_fop = &proc_tid_base_operations;
2345 inode->i_flags|=S_IMMUTABLE;
2346 inode->i_nlink = 4;
2347 #ifdef CONFIG_SECURITY
2348 inode->i_nlink += 1;
2349 #endif
2350
2351 dentry->d_op = &pid_dentry_operations;
2352
2353 d_add(dentry, inode);
2354 /* Close the race of the process dying before we return the dentry */
2355 if (pid_revalidate(dentry, NULL))
2356 error = NULL;
2357 out:
2358 return error;
2359 }
2360
2361 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2362 {
2363 struct dentry *result = ERR_PTR(-ENOENT);
2364 struct task_struct *task;
2365 struct task_struct *leader = get_proc_task(dir);
2366 unsigned tid;
2367
2368 if (!leader)
2369 goto out_no_task;
2370
2371 tid = name_to_int(dentry);
2372 if (tid == ~0U)
2373 goto out;
2374
2375 rcu_read_lock();
2376 task = find_task_by_pid(tid);
2377 if (task)
2378 get_task_struct(task);
2379 rcu_read_unlock();
2380 if (!task)
2381 goto out;
2382 if (leader->tgid != task->tgid)
2383 goto out_drop_task;
2384
2385 result = proc_task_instantiate(dir, dentry, task, NULL);
2386 out_drop_task:
2387 put_task_struct(task);
2388 out:
2389 put_task_struct(leader);
2390 out_no_task:
2391 return result;
2392 }
2393
2394 /*
2395 * Find the first tid of a thread group to return to user space.
2396 *
2397 * Usually this is just the thread group leader, but if the users
2398 * buffer was too small or there was a seek into the middle of the
2399 * directory we have more work todo.
2400 *
2401 * In the case of a short read we start with find_task_by_pid.
2402 *
2403 * In the case of a seek we start with the leader and walk nr
2404 * threads past it.
2405 */
2406 static struct task_struct *first_tid(struct task_struct *leader,
2407 int tid, int nr)
2408 {
2409 struct task_struct *pos;
2410
2411 rcu_read_lock();
2412 /* Attempt to start with the pid of a thread */
2413 if (tid && (nr > 0)) {
2414 pos = find_task_by_pid(tid);
2415 if (pos && (pos->group_leader == leader))
2416 goto found;
2417 }
2418
2419 /* If nr exceeds the number of threads there is nothing todo */
2420 pos = NULL;
2421 if (nr && nr >= get_nr_threads(leader))
2422 goto out;
2423
2424 /* If we haven't found our starting place yet start
2425 * with the leader and walk nr threads forward.
2426 */
2427 for (pos = leader; nr > 0; --nr) {
2428 pos = next_thread(pos);
2429 if (pos == leader) {
2430 pos = NULL;
2431 goto out;
2432 }
2433 }
2434 found:
2435 get_task_struct(pos);
2436 out:
2437 rcu_read_unlock();
2438 return pos;
2439 }
2440
2441 /*
2442 * Find the next thread in the thread list.
2443 * Return NULL if there is an error or no next thread.
2444 *
2445 * The reference to the input task_struct is released.
2446 */
2447 static struct task_struct *next_tid(struct task_struct *start)
2448 {
2449 struct task_struct *pos = NULL;
2450 rcu_read_lock();
2451 if (pid_alive(start)) {
2452 pos = next_thread(start);
2453 if (thread_group_leader(pos))
2454 pos = NULL;
2455 else
2456 get_task_struct(pos);
2457 }
2458 rcu_read_unlock();
2459 put_task_struct(start);
2460 return pos;
2461 }
2462
2463 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2464 struct task_struct *task, int tid)
2465 {
2466 char name[PROC_NUMBUF];
2467 int len = snprintf(name, sizeof(name), "%d", tid);
2468 return proc_fill_cache(filp, dirent, filldir, name, len,
2469 proc_task_instantiate, task, NULL);
2470 }
2471
2472 /* for the /proc/TGID/task/ directories */
2473 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
2474 {
2475 struct dentry *dentry = filp->f_path.dentry;
2476 struct inode *inode = dentry->d_inode;
2477 struct task_struct *leader = NULL;
2478 struct task_struct *task;
2479 int retval = -ENOENT;
2480 ino_t ino;
2481 int tid;
2482 unsigned long pos = filp->f_pos; /* avoiding "long long" filp->f_pos */
2483
2484 task = get_proc_task(inode);
2485 if (!task)
2486 goto out_no_task;
2487 rcu_read_lock();
2488 if (pid_alive(task)) {
2489 leader = task->group_leader;
2490 get_task_struct(leader);
2491 }
2492 rcu_read_unlock();
2493 put_task_struct(task);
2494 if (!leader)
2495 goto out_no_task;
2496 retval = 0;
2497
2498 switch (pos) {
2499 case 0:
2500 ino = inode->i_ino;
2501 if (filldir(dirent, ".", 1, pos, ino, DT_DIR) < 0)
2502 goto out;
2503 pos++;
2504 /* fall through */
2505 case 1:
2506 ino = parent_ino(dentry);
2507 if (filldir(dirent, "..", 2, pos, ino, DT_DIR) < 0)
2508 goto out;
2509 pos++;
2510 /* fall through */
2511 }
2512
2513 /* f_version caches the tgid value that the last readdir call couldn't
2514 * return. lseek aka telldir automagically resets f_version to 0.
2515 */
2516 tid = filp->f_version;
2517 filp->f_version = 0;
2518 for (task = first_tid(leader, tid, pos - 2);
2519 task;
2520 task = next_tid(task), pos++) {
2521 tid = task->pid;
2522 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
2523 /* returning this tgid failed, save it as the first
2524 * pid for the next readir call */
2525 filp->f_version = tid;
2526 put_task_struct(task);
2527 break;
2528 }
2529 }
2530 out:
2531 filp->f_pos = pos;
2532 put_task_struct(leader);
2533 out_no_task:
2534 return retval;
2535 }
2536
2537 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
2538 {
2539 struct inode *inode = dentry->d_inode;
2540 struct task_struct *p = get_proc_task(inode);
2541 generic_fillattr(inode, stat);
2542
2543 if (p) {
2544 rcu_read_lock();
2545 stat->nlink += get_nr_threads(p);
2546 rcu_read_unlock();
2547 put_task_struct(p);
2548 }
2549
2550 return 0;
2551 }
2552
2553 static const struct inode_operations proc_task_inode_operations = {
2554 .lookup = proc_task_lookup,
2555 .getattr = proc_task_getattr,
2556 .setattr = proc_setattr,
2557 };
2558
2559 static const struct file_operations proc_task_operations = {
2560 .read = generic_read_dir,
2561 .readdir = proc_task_readdir,
2562 };
CJK support for tty framebuffer vt