?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
header cleaning: don't include smp_lock.h when not used
[usb.git] / fs / proc / base.c
bloba3e77fb867eaadd17ebcc93449cb7297d32e3a54
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 char *modname;
282 const char *sym_name;
283 unsigned long wchan, size, offset;
284 char namebuf[KSYM_NAME_LEN+1];
285
286 wchan = get_wchan(task);
287
288 sym_name = kallsyms_lookup(wchan, &size, &offset, &modname, namebuf);
289 if (sym_name)
290 return sprintf(buffer, "%s", sym_name);
291 return sprintf(buffer, "%lu", wchan);
292 }
293 #endif /* CONFIG_KALLSYMS */
294
295 #ifdef CONFIG_SCHEDSTATS
296 /*
297 * Provides /proc/PID/schedstat
298 */
299 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
300 {
301 return sprintf(buffer, "%lu %lu %lu\n",
302 task->sched_info.cpu_time,
303 task->sched_info.run_delay,
304 task->sched_info.pcnt);
305 }
306 #endif
307
308 /* The badness from the OOM killer */
309 unsigned long badness(struct task_struct *p, unsigned long uptime);
310 static int proc_oom_score(struct task_struct *task, char *buffer)
311 {
312 unsigned long points;
313 struct timespec uptime;
314
315 do_posix_clock_monotonic_gettime(&uptime);
316 read_lock(&tasklist_lock);
317 points = badness(task, uptime.tv_sec);
318 read_unlock(&tasklist_lock);
319 return sprintf(buffer, "%lu\n", points);
320 }
321
322 /************************************************************************/
323 /* Here the fs part begins */
324 /************************************************************************/
325
326 /* permission checks */
327 static int proc_fd_access_allowed(struct inode *inode)
328 {
329 struct task_struct *task;
330 int allowed = 0;
331 /* Allow access to a task's file descriptors if it is us or we
332 * may use ptrace attach to the process and find out that
333 * information.
334 */
335 task = get_proc_task(inode);
336 if (task) {
337 allowed = ptrace_may_attach(task);
338 put_task_struct(task);
339 }
340 return allowed;
341 }
342
343 static int proc_setattr(struct dentry *dentry, struct iattr *attr)
344 {
345 int error;
346 struct inode *inode = dentry->d_inode;
347
348 if (attr->ia_valid & ATTR_MODE)
349 return -EPERM;
350
351 error = inode_change_ok(inode, attr);
352 if (!error) {
353 error = security_inode_setattr(dentry, attr);
354 if (!error)
355 error = inode_setattr(inode, attr);
356 }
357 return error;
358 }
359
360 static const struct inode_operations proc_def_inode_operations = {
361 .setattr = proc_setattr,
362 };
363
364 extern struct seq_operations mounts_op;
365 struct proc_mounts {
366 struct seq_file m;
367 int event;
368 };
369
370 static int mounts_open(struct inode *inode, struct file *file)
371 {
372 struct task_struct *task = get_proc_task(inode);
373 struct mnt_namespace *ns = NULL;
374 struct proc_mounts *p;
375 int ret = -EINVAL;
376
377 if (task) {
378 task_lock(task);
379 if (task->nsproxy) {
380 ns = task->nsproxy->mnt_ns;
381 if (ns)
382 get_mnt_ns(ns);
383 }
384 task_unlock(task);
385 put_task_struct(task);
386 }
387
388 if (ns) {
389 ret = -ENOMEM;
390 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
391 if (p) {
392 file->private_data = &p->m;
393 ret = seq_open(file, &mounts_op);
394 if (!ret) {
395 p->m.private = ns;
396 p->event = ns->event;
397 return 0;
398 }
399 kfree(p);
400 }
401 put_mnt_ns(ns);
402 }
403 return ret;
404 }
405
406 static int mounts_release(struct inode *inode, struct file *file)
407 {
408 struct seq_file *m = file->private_data;
409 struct mnt_namespace *ns = m->private;
410 put_mnt_ns(ns);
411 return seq_release(inode, file);
412 }
413
414 static unsigned mounts_poll(struct file *file, poll_table *wait)
415 {
416 struct proc_mounts *p = file->private_data;
417 struct mnt_namespace *ns = p->m.private;
418 unsigned res = 0;
419
420 poll_wait(file, &ns->poll, wait);
421
422 spin_lock(&vfsmount_lock);
423 if (p->event != ns->event) {
424 p->event = ns->event;
425 res = POLLERR;
426 }
427 spin_unlock(&vfsmount_lock);
428
429 return res;
430 }
431
432 static const struct file_operations proc_mounts_operations = {
433 .open = mounts_open,
434 .read = seq_read,
435 .llseek = seq_lseek,
436 .release = mounts_release,
437 .poll = mounts_poll,
438 };
439
440 extern struct seq_operations mountstats_op;
441 static int mountstats_open(struct inode *inode, struct file *file)
442 {
443 int ret = seq_open(file, &mountstats_op);
444
445 if (!ret) {
446 struct seq_file *m = file->private_data;
447 struct mnt_namespace *mnt_ns = NULL;
448 struct task_struct *task = get_proc_task(inode);
449
450 if (task) {
451 task_lock(task);
452 if (task->nsproxy)
453 mnt_ns = task->nsproxy->mnt_ns;
454 if (mnt_ns)
455 get_mnt_ns(mnt_ns);
456 task_unlock(task);
457 put_task_struct(task);
458 }
459
460 if (mnt_ns)
461 m->private = mnt_ns;
462 else {
463 seq_release(inode, file);
464 ret = -EINVAL;
465 }
466 }
467 return ret;
468 }
469
470 static const struct file_operations proc_mountstats_operations = {
471 .open = mountstats_open,
472 .read = seq_read,
473 .llseek = seq_lseek,
474 .release = mounts_release,
475 };
476
477 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
478
479 static ssize_t proc_info_read(struct file * file, char __user * buf,
480 size_t count, loff_t *ppos)
481 {
482 struct inode * inode = file->f_path.dentry->d_inode;
483 unsigned long page;
484 ssize_t length;
485 struct task_struct *task = get_proc_task(inode);
486
487 length = -ESRCH;
488 if (!task)
489 goto out_no_task;
490
491 if (count > PROC_BLOCK_SIZE)
492 count = PROC_BLOCK_SIZE;
493
494 length = -ENOMEM;
495 if (!(page = __get_free_page(GFP_KERNEL)))
496 goto out;
497
498 length = PROC_I(inode)->op.proc_read(task, (char*)page);
499
500 if (length >= 0)
501 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
502 free_page(page);
503 out:
504 put_task_struct(task);
505 out_no_task:
506 return length;
507 }
508
509 static const struct file_operations proc_info_file_operations = {
510 .read = proc_info_read,
511 };
512
513 static int mem_open(struct inode* inode, struct file* file)
514 {
515 file->private_data = (void*)((long)current->self_exec_id);
516 return 0;
517 }
518
519 static ssize_t mem_read(struct file * file, char __user * buf,
520 size_t count, loff_t *ppos)
521 {
522 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
523 char *page;
524 unsigned long src = *ppos;
525 int ret = -ESRCH;
526 struct mm_struct *mm;
527
528 if (!task)
529 goto out_no_task;
530
531 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
532 goto out;
533
534 ret = -ENOMEM;
535 page = (char *)__get_free_page(GFP_USER);
536 if (!page)
537 goto out;
538
539 ret = 0;
540
541 mm = get_task_mm(task);
542 if (!mm)
543 goto out_free;
544
545 ret = -EIO;
546
547 if (file->private_data != (void*)((long)current->self_exec_id))
548 goto out_put;
549
550 ret = 0;
551
552 while (count > 0) {
553 int this_len, retval;
554
555 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
556 retval = access_process_vm(task, src, page, this_len, 0);
557 if (!retval || !MAY_PTRACE(task) || !ptrace_may_attach(task)) {
558 if (!ret)
559 ret = -EIO;
560 break;
561 }
562
563 if (copy_to_user(buf, page, retval)) {
564 ret = -EFAULT;
565 break;
566 }
567
568 ret += retval;
569 src += retval;
570 buf += retval;
571 count -= retval;
572 }
573 *ppos = src;
574
575 out_put:
576 mmput(mm);
577 out_free:
578 free_page((unsigned long) page);
579 out:
580 put_task_struct(task);
581 out_no_task:
582 return ret;
583 }
584
585 #define mem_write NULL
586
587 #ifndef mem_write
588 /* This is a security hazard */
589 static ssize_t mem_write(struct file * file, const char __user *buf,
590 size_t count, loff_t *ppos)
591 {
592 int copied;
593 char *page;
594 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
595 unsigned long dst = *ppos;
596
597 copied = -ESRCH;
598 if (!task)
599 goto out_no_task;
600
601 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
602 goto out;
603
604 copied = -ENOMEM;
605 page = (char *)__get_free_page(GFP_USER);
606 if (!page)
607 goto out;
608
609 copied = 0;
610 while (count > 0) {
611 int this_len, retval;
612
613 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
614 if (copy_from_user(page, buf, this_len)) {
615 copied = -EFAULT;
616 break;
617 }
618 retval = access_process_vm(task, dst, page, this_len, 1);
619 if (!retval) {
620 if (!copied)
621 copied = -EIO;
622 break;
623 }
624 copied += retval;
625 buf += retval;
626 dst += retval;
627 count -= retval;
628 }
629 *ppos = dst;
630 free_page((unsigned long) page);
631 out:
632 put_task_struct(task);
633 out_no_task:
634 return copied;
635 }
636 #endif
637
638 static loff_t mem_lseek(struct file * file, loff_t offset, int orig)
639 {
640 switch (orig) {
641 case 0:
642 file->f_pos = offset;
643 break;
644 case 1:
645 file->f_pos += offset;
646 break;
647 default:
648 return -EINVAL;
649 }
650 force_successful_syscall_return();
651 return file->f_pos;
652 }
653
654 static const struct file_operations proc_mem_operations = {
655 .llseek = mem_lseek,
656 .read = mem_read,
657 .write = mem_write,
658 .open = mem_open,
659 };
660
661 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
662 size_t count, loff_t *ppos)
663 {
664 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
665 char buffer[PROC_NUMBUF];
666 size_t len;
667 int oom_adjust;
668 loff_t __ppos = *ppos;
669
670 if (!task)
671 return -ESRCH;
672 oom_adjust = task->oomkilladj;
673 put_task_struct(task);
674
675 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
676 if (__ppos >= len)
677 return 0;
678 if (count > len-__ppos)
679 count = len-__ppos;
680 if (copy_to_user(buf, buffer + __ppos, count))
681 return -EFAULT;
682 *ppos = __ppos + count;
683 return count;
684 }
685
686 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
687 size_t count, loff_t *ppos)
688 {
689 struct task_struct *task;
690 char buffer[PROC_NUMBUF], *end;
691 int oom_adjust;
692
693 memset(buffer, 0, sizeof(buffer));
694 if (count > sizeof(buffer) - 1)
695 count = sizeof(buffer) - 1;
696 if (copy_from_user(buffer, buf, count))
697 return -EFAULT;
698 oom_adjust = simple_strtol(buffer, &end, 0);
699 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
700 oom_adjust != OOM_DISABLE)
701 return -EINVAL;
702 if (*end == '\n')
703 end++;
704 task = get_proc_task(file->f_path.dentry->d_inode);
705 if (!task)
706 return -ESRCH;
707 if (oom_adjust < task->oomkilladj && !capable(CAP_SYS_RESOURCE)) {
708 put_task_struct(task);
709 return -EACCES;
710 }
711 task->oomkilladj = oom_adjust;
712 put_task_struct(task);
713 if (end - buffer == 0)
714 return -EIO;
715 return end - buffer;
716 }
717
718 static const struct file_operations proc_oom_adjust_operations = {
719 .read = oom_adjust_read,
720 .write = oom_adjust_write,
721 };
722
723 static ssize_t clear_refs_write(struct file *file, const char __user *buf,
724 size_t count, loff_t *ppos)
725 {
726 struct task_struct *task;
727 char buffer[PROC_NUMBUF], *end;
728 struct mm_struct *mm;
729
730 memset(buffer, 0, sizeof(buffer));
731 if (count > sizeof(buffer) - 1)
732 count = sizeof(buffer) - 1;
733 if (copy_from_user(buffer, buf, count))
734 return -EFAULT;
735 if (!simple_strtol(buffer, &end, 0))
736 return -EINVAL;
737 if (*end == '\n')
738 end++;
739 task = get_proc_task(file->f_path.dentry->d_inode);
740 if (!task)
741 return -ESRCH;
742 mm = get_task_mm(task);
743 if (mm) {
744 clear_refs_smap(mm);
745 mmput(mm);
746 }
747 put_task_struct(task);
748 if (end - buffer == 0)
749 return -EIO;
750 return end - buffer;
751 }
752
753 static struct file_operations proc_clear_refs_operations = {
754 .write = clear_refs_write,
755 };
756
757 #ifdef CONFIG_AUDITSYSCALL
758 #define TMPBUFLEN 21
759 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
760 size_t count, loff_t *ppos)
761 {
762 struct inode * inode = file->f_path.dentry->d_inode;
763 struct task_struct *task = get_proc_task(inode);
764 ssize_t length;
765 char tmpbuf[TMPBUFLEN];
766
767 if (!task)
768 return -ESRCH;
769 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
770 audit_get_loginuid(task->audit_context));
771 put_task_struct(task);
772 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
773 }
774
775 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
776 size_t count, loff_t *ppos)
777 {
778 struct inode * inode = file->f_path.dentry->d_inode;
779 char *page, *tmp;
780 ssize_t length;
781 uid_t loginuid;
782
783 if (!capable(CAP_AUDIT_CONTROL))
784 return -EPERM;
785
786 if (current != pid_task(proc_pid(inode), PIDTYPE_PID))
787 return -EPERM;
788
789 if (count >= PAGE_SIZE)
790 count = PAGE_SIZE - 1;
791
792 if (*ppos != 0) {
793 /* No partial writes. */
794 return -EINVAL;
795 }
796 page = (char*)__get_free_page(GFP_USER);
797 if (!page)
798 return -ENOMEM;
799 length = -EFAULT;
800 if (copy_from_user(page, buf, count))
801 goto out_free_page;
802
803 page[count] = '\0';
804 loginuid = simple_strtoul(page, &tmp, 10);
805 if (tmp == page) {
806 length = -EINVAL;
807 goto out_free_page;
808
809 }
810 length = audit_set_loginuid(current, loginuid);
811 if (likely(length == 0))
812 length = count;
813
814 out_free_page:
815 free_page((unsigned long) page);
816 return length;
817 }
818
819 static const struct file_operations proc_loginuid_operations = {
820 .read = proc_loginuid_read,
821 .write = proc_loginuid_write,
822 };
823 #endif
824
825 #ifdef CONFIG_SECCOMP
826 static ssize_t seccomp_read(struct file *file, char __user *buf,
827 size_t count, loff_t *ppos)
828 {
829 struct task_struct *tsk = get_proc_task(file->f_dentry->d_inode);
830 char __buf[20];
831 loff_t __ppos = *ppos;
832 size_t len;
833
834 if (!tsk)
835 return -ESRCH;
836 /* no need to print the trailing zero, so use only len */
837 len = sprintf(__buf, "%u\n", tsk->seccomp.mode);
838 put_task_struct(tsk);
839 if (__ppos >= len)
840 return 0;
841 if (count > len - __ppos)
842 count = len - __ppos;
843 if (copy_to_user(buf, __buf + __ppos, count))
844 return -EFAULT;
845 *ppos = __ppos + count;
846 return count;
847 }
848
849 static ssize_t seccomp_write(struct file *file, const char __user *buf,
850 size_t count, loff_t *ppos)
851 {
852 struct task_struct *tsk = get_proc_task(file->f_dentry->d_inode);
853 char __buf[20], *end;
854 unsigned int seccomp_mode;
855 ssize_t result;
856
857 result = -ESRCH;
858 if (!tsk)
859 goto out_no_task;
860
861 /* can set it only once to be even more secure */
862 result = -EPERM;
863 if (unlikely(tsk->seccomp.mode))
864 goto out;
865
866 result = -EFAULT;
867 memset(__buf, 0, sizeof(__buf));
868 count = min(count, sizeof(__buf) - 1);
869 if (copy_from_user(__buf, buf, count))
870 goto out;
871
872 seccomp_mode = simple_strtoul(__buf, &end, 0);
873 if (*end == '\n')
874 end++;
875 result = -EINVAL;
876 if (seccomp_mode && seccomp_mode <= NR_SECCOMP_MODES) {
877 tsk->seccomp.mode = seccomp_mode;
878 set_tsk_thread_flag(tsk, TIF_SECCOMP);
879 } else
880 goto out;
881 result = -EIO;
882 if (unlikely(!(end - __buf)))
883 goto out;
884 result = end - __buf;
885 out:
886 put_task_struct(tsk);
887 out_no_task:
888 return result;
889 }
890
891 static const struct file_operations proc_seccomp_operations = {
892 .read = seccomp_read,
893 .write = seccomp_write,
894 };
895 #endif /* CONFIG_SECCOMP */
896
897 #ifdef CONFIG_FAULT_INJECTION
898 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
899 size_t count, loff_t *ppos)
900 {
901 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
902 char buffer[PROC_NUMBUF];
903 size_t len;
904 int make_it_fail;
905 loff_t __ppos = *ppos;
906
907 if (!task)
908 return -ESRCH;
909 make_it_fail = task->make_it_fail;
910 put_task_struct(task);
911
912 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
913 if (__ppos >= len)
914 return 0;
915 if (count > len-__ppos)
916 count = len-__ppos;
917 if (copy_to_user(buf, buffer + __ppos, count))
918 return -EFAULT;
919 *ppos = __ppos + count;
920 return count;
921 }
922
923 static ssize_t proc_fault_inject_write(struct file * file,
924 const char __user * buf, size_t count, loff_t *ppos)
925 {
926 struct task_struct *task;
927 char buffer[PROC_NUMBUF], *end;
928 int make_it_fail;
929
930 if (!capable(CAP_SYS_RESOURCE))
931 return -EPERM;
932 memset(buffer, 0, sizeof(buffer));
933 if (count > sizeof(buffer) - 1)
934 count = sizeof(buffer) - 1;
935 if (copy_from_user(buffer, buf, count))
936 return -EFAULT;
937 make_it_fail = simple_strtol(buffer, &end, 0);
938 if (*end == '\n')
939 end++;
940 task = get_proc_task(file->f_dentry->d_inode);
941 if (!task)
942 return -ESRCH;
943 task->make_it_fail = make_it_fail;
944 put_task_struct(task);
945 if (end - buffer == 0)
946 return -EIO;
947 return end - buffer;
948 }
949
950 static const struct file_operations proc_fault_inject_operations = {
951 .read = proc_fault_inject_read,
952 .write = proc_fault_inject_write,
953 };
954 #endif
955
956 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
957 {
958 struct inode *inode = dentry->d_inode;
959 int error = -EACCES;
960
961 /* We don't need a base pointer in the /proc filesystem */
962 path_release(nd);
963
964 /* Are we allowed to snoop on the tasks file descriptors? */
965 if (!proc_fd_access_allowed(inode))
966 goto out;
967
968 error = PROC_I(inode)->op.proc_get_link(inode, &nd->dentry, &nd->mnt);
969 nd->last_type = LAST_BIND;
970 out:
971 return ERR_PTR(error);
972 }
973
974 static int do_proc_readlink(struct dentry *dentry, struct vfsmount *mnt,
975 char __user *buffer, int buflen)
976 {
977 struct inode * inode;
978 char *tmp = (char*)__get_free_page(GFP_KERNEL), *path;
979 int len;
980
981 if (!tmp)
982 return -ENOMEM;
983
984 inode = dentry->d_inode;
985 path = d_path(dentry, mnt, tmp, PAGE_SIZE);
986 len = PTR_ERR(path);
987 if (IS_ERR(path))
988 goto out;
989 len = tmp + PAGE_SIZE - 1 - path;
990
991 if (len > buflen)
992 len = buflen;
993 if (copy_to_user(buffer, path, len))
994 len = -EFAULT;
995 out:
996 free_page((unsigned long)tmp);
997 return len;
998 }
999
1000 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1001 {
1002 int error = -EACCES;
1003 struct inode *inode = dentry->d_inode;
1004 struct dentry *de;
1005 struct vfsmount *mnt = NULL;
1006
1007 /* Are we allowed to snoop on the tasks file descriptors? */
1008 if (!proc_fd_access_allowed(inode))
1009 goto out;
1010
1011 error = PROC_I(inode)->op.proc_get_link(inode, &de, &mnt);
1012 if (error)
1013 goto out;
1014
1015 error = do_proc_readlink(de, mnt, buffer, buflen);
1016 dput(de);
1017 mntput(mnt);
1018 out:
1019 return error;
1020 }
1021
1022 static const struct inode_operations proc_pid_link_inode_operations = {
1023 .readlink = proc_pid_readlink,
1024 .follow_link = proc_pid_follow_link,
1025 .setattr = proc_setattr,
1026 };
1027
1028
1029 /* building an inode */
1030
1031 static int task_dumpable(struct task_struct *task)
1032 {
1033 int dumpable = 0;
1034 struct mm_struct *mm;
1035
1036 task_lock(task);
1037 mm = task->mm;
1038 if (mm)
1039 dumpable = mm->dumpable;
1040 task_unlock(task);
1041 if(dumpable == 1)
1042 return 1;
1043 return 0;
1044 }
1045
1046
1047 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1048 {
1049 struct inode * inode;
1050 struct proc_inode *ei;
1051
1052 /* We need a new inode */
1053
1054 inode = new_inode(sb);
1055 if (!inode)
1056 goto out;
1057
1058 /* Common stuff */
1059 ei = PROC_I(inode);
1060 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1061 inode->i_op = &proc_def_inode_operations;
1062
1063 /*
1064 * grab the reference to task.
1065 */
1066 ei->pid = get_task_pid(task, PIDTYPE_PID);
1067 if (!ei->pid)
1068 goto out_unlock;
1069
1070 inode->i_uid = 0;
1071 inode->i_gid = 0;
1072 if (task_dumpable(task)) {
1073 inode->i_uid = task->euid;
1074 inode->i_gid = task->egid;
1075 }
1076 security_task_to_inode(task, inode);
1077
1078 out:
1079 return inode;
1080
1081 out_unlock:
1082 iput(inode);
1083 return NULL;
1084 }
1085
1086 static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1087 {
1088 struct inode *inode = dentry->d_inode;
1089 struct task_struct *task;
1090 generic_fillattr(inode, stat);
1091
1092 rcu_read_lock();
1093 stat->uid = 0;
1094 stat->gid = 0;
1095 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1096 if (task) {
1097 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1098 task_dumpable(task)) {
1099 stat->uid = task->euid;
1100 stat->gid = task->egid;
1101 }
1102 }
1103 rcu_read_unlock();
1104 return 0;
1105 }
1106
1107 /* dentry stuff */
1108
1109 /*
1110 * Exceptional case: normally we are not allowed to unhash a busy
1111 * directory. In this case, however, we can do it - no aliasing problems
1112 * due to the way we treat inodes.
1113 *
1114 * Rewrite the inode's ownerships here because the owning task may have
1115 * performed a setuid(), etc.
1116 *
1117 * Before the /proc/pid/status file was created the only way to read
1118 * the effective uid of a /process was to stat /proc/pid. Reading
1119 * /proc/pid/status is slow enough that procps and other packages
1120 * kept stating /proc/pid. To keep the rules in /proc simple I have
1121 * made this apply to all per process world readable and executable
1122 * directories.
1123 */
1124 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1125 {
1126 struct inode *inode = dentry->d_inode;
1127 struct task_struct *task = get_proc_task(inode);
1128 if (task) {
1129 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1130 task_dumpable(task)) {
1131 inode->i_uid = task->euid;
1132 inode->i_gid = task->egid;
1133 } else {
1134 inode->i_uid = 0;
1135 inode->i_gid = 0;
1136 }
1137 inode->i_mode &= ~(S_ISUID | S_ISGID);
1138 security_task_to_inode(task, inode);
1139 put_task_struct(task);
1140 return 1;
1141 }
1142 d_drop(dentry);
1143 return 0;
1144 }
1145
1146 static int pid_delete_dentry(struct dentry * dentry)
1147 {
1148 /* Is the task we represent dead?
1149 * If so, then don't put the dentry on the lru list,
1150 * kill it immediately.
1151 */
1152 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1153 }
1154
1155 static struct dentry_operations pid_dentry_operations =
1156 {
1157 .d_revalidate = pid_revalidate,
1158 .d_delete = pid_delete_dentry,
1159 };
1160
1161 /* Lookups */
1162
1163 typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1164 struct task_struct *, const void *);
1165
1166 /*
1167 * Fill a directory entry.
1168 *
1169 * If possible create the dcache entry and derive our inode number and
1170 * file type from dcache entry.
1171 *
1172 * Since all of the proc inode numbers are dynamically generated, the inode
1173 * numbers do not exist until the inode is cache. This means creating the
1174 * the dcache entry in readdir is necessary to keep the inode numbers
1175 * reported by readdir in sync with the inode numbers reported
1176 * by stat.
1177 */
1178 static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1179 char *name, int len,
1180 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1181 {
1182 struct dentry *child, *dir = filp->f_path.dentry;
1183 struct inode *inode;
1184 struct qstr qname;
1185 ino_t ino = 0;
1186 unsigned type = DT_UNKNOWN;
1187
1188 qname.name = name;
1189 qname.len = len;
1190 qname.hash = full_name_hash(name, len);
1191
1192 child = d_lookup(dir, &qname);
1193 if (!child) {
1194 struct dentry *new;
1195 new = d_alloc(dir, &qname);
1196 if (new) {
1197 child = instantiate(dir->d_inode, new, task, ptr);
1198 if (child)
1199 dput(new);
1200 else
1201 child = new;
1202 }
1203 }
1204 if (!child || IS_ERR(child) || !child->d_inode)
1205 goto end_instantiate;
1206 inode = child->d_inode;
1207 if (inode) {
1208 ino = inode->i_ino;
1209 type = inode->i_mode >> 12;
1210 }
1211 dput(child);
1212 end_instantiate:
1213 if (!ino)
1214 ino = find_inode_number(dir, &qname);
1215 if (!ino)
1216 ino = 1;
1217 return filldir(dirent, name, len, filp->f_pos, ino, type);
1218 }
1219
1220 static unsigned name_to_int(struct dentry *dentry)
1221 {
1222 const char *name = dentry->d_name.name;
1223 int len = dentry->d_name.len;
1224 unsigned n = 0;
1225
1226 if (len > 1 && *name == '0')
1227 goto out;
1228 while (len-- > 0) {
1229 unsigned c = *name++ - '0';
1230 if (c > 9)
1231 goto out;
1232 if (n >= (~0U-9)/10)
1233 goto out;
1234 n *= 10;
1235 n += c;
1236 }
1237 return n;
1238 out:
1239 return ~0U;
1240 }
1241
1242 #define PROC_FDINFO_MAX 64
1243
1244 static int proc_fd_info(struct inode *inode, struct dentry **dentry,
1245 struct vfsmount **mnt, char *info)
1246 {
1247 struct task_struct *task = get_proc_task(inode);
1248 struct files_struct *files = NULL;
1249 struct file *file;
1250 int fd = proc_fd(inode);
1251
1252 if (task) {
1253 files = get_files_struct(task);
1254 put_task_struct(task);
1255 }
1256 if (files) {
1257 /*
1258 * We are not taking a ref to the file structure, so we must
1259 * hold ->file_lock.
1260 */
1261 spin_lock(&files->file_lock);
1262 file = fcheck_files(files, fd);
1263 if (file) {
1264 if (mnt)
1265 *mnt = mntget(file->f_path.mnt);
1266 if (dentry)
1267 *dentry = dget(file->f_path.dentry);
1268 if (info)
1269 snprintf(info, PROC_FDINFO_MAX,
1270 "pos:\t%lli\n"
1271 "flags:\t0%o\n",
1272 (long long) file->f_pos,
1273 file->f_flags);
1274 spin_unlock(&files->file_lock);
1275 put_files_struct(files);
1276 return 0;
1277 }
1278 spin_unlock(&files->file_lock);
1279 put_files_struct(files);
1280 }
1281 return -ENOENT;
1282 }
1283
1284 static int proc_fd_link(struct inode *inode, struct dentry **dentry,
1285 struct vfsmount **mnt)
1286 {
1287 return proc_fd_info(inode, dentry, mnt, NULL);
1288 }
1289
1290 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1291 {
1292 struct inode *inode = dentry->d_inode;
1293 struct task_struct *task = get_proc_task(inode);
1294 int fd = proc_fd(inode);
1295 struct files_struct *files;
1296
1297 if (task) {
1298 files = get_files_struct(task);
1299 if (files) {
1300 rcu_read_lock();
1301 if (fcheck_files(files, fd)) {
1302 rcu_read_unlock();
1303 put_files_struct(files);
1304 if (task_dumpable(task)) {
1305 inode->i_uid = task->euid;
1306 inode->i_gid = task->egid;
1307 } else {
1308 inode->i_uid = 0;
1309 inode->i_gid = 0;
1310 }
1311 inode->i_mode &= ~(S_ISUID | S_ISGID);
1312 security_task_to_inode(task, inode);
1313 put_task_struct(task);
1314 return 1;
1315 }
1316 rcu_read_unlock();
1317 put_files_struct(files);
1318 }
1319 put_task_struct(task);
1320 }
1321 d_drop(dentry);
1322 return 0;
1323 }
1324
1325 static struct dentry_operations tid_fd_dentry_operations =
1326 {
1327 .d_revalidate = tid_fd_revalidate,
1328 .d_delete = pid_delete_dentry,
1329 };
1330
1331 static struct dentry *proc_fd_instantiate(struct inode *dir,
1332 struct dentry *dentry, struct task_struct *task, const void *ptr)
1333 {
1334 unsigned fd = *(const unsigned *)ptr;
1335 struct file *file;
1336 struct files_struct *files;
1337 struct inode *inode;
1338 struct proc_inode *ei;
1339 struct dentry *error = ERR_PTR(-ENOENT);
1340
1341 inode = proc_pid_make_inode(dir->i_sb, task);
1342 if (!inode)
1343 goto out;
1344 ei = PROC_I(inode);
1345 ei->fd = fd;
1346 files = get_files_struct(task);
1347 if (!files)
1348 goto out_iput;
1349 inode->i_mode = S_IFLNK;
1350
1351 /*
1352 * We are not taking a ref to the file structure, so we must
1353 * hold ->file_lock.
1354 */
1355 spin_lock(&files->file_lock);
1356 file = fcheck_files(files, fd);
1357 if (!file)
1358 goto out_unlock;
1359 if (file->f_mode & 1)
1360 inode->i_mode |= S_IRUSR | S_IXUSR;
1361 if (file->f_mode & 2)
1362 inode->i_mode |= S_IWUSR | S_IXUSR;
1363 spin_unlock(&files->file_lock);
1364 put_files_struct(files);
1365
1366 inode->i_op = &proc_pid_link_inode_operations;
1367 inode->i_size = 64;
1368 ei->op.proc_get_link = proc_fd_link;
1369 dentry->d_op = &tid_fd_dentry_operations;
1370 d_add(dentry, inode);
1371 /* Close the race of the process dying before we return the dentry */
1372 if (tid_fd_revalidate(dentry, NULL))
1373 error = NULL;
1374
1375 out:
1376 return error;
1377 out_unlock:
1378 spin_unlock(&files->file_lock);
1379 put_files_struct(files);
1380 out_iput:
1381 iput(inode);
1382 goto out;
1383 }
1384
1385 static struct dentry *proc_lookupfd_common(struct inode *dir,
1386 struct dentry *dentry,
1387 instantiate_t instantiate)
1388 {
1389 struct task_struct *task = get_proc_task(dir);
1390 unsigned fd = name_to_int(dentry);
1391 struct dentry *result = ERR_PTR(-ENOENT);
1392
1393 if (!task)
1394 goto out_no_task;
1395 if (fd == ~0U)
1396 goto out;
1397
1398 result = instantiate(dir, dentry, task, &fd);
1399 out:
1400 put_task_struct(task);
1401 out_no_task:
1402 return result;
1403 }
1404
1405 static int proc_readfd_common(struct file * filp, void * dirent,
1406 filldir_t filldir, instantiate_t instantiate)
1407 {
1408 struct dentry *dentry = filp->f_path.dentry;
1409 struct inode *inode = dentry->d_inode;
1410 struct task_struct *p = get_proc_task(inode);
1411 unsigned int fd, tid, ino;
1412 int retval;
1413 struct files_struct * files;
1414 struct fdtable *fdt;
1415
1416 retval = -ENOENT;
1417 if (!p)
1418 goto out_no_task;
1419 retval = 0;
1420 tid = p->pid;
1421
1422 fd = filp->f_pos;
1423 switch (fd) {
1424 case 0:
1425 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1426 goto out;
1427 filp->f_pos++;
1428 case 1:
1429 ino = parent_ino(dentry);
1430 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1431 goto out;
1432 filp->f_pos++;
1433 default:
1434 files = get_files_struct(p);
1435 if (!files)
1436 goto out;
1437 rcu_read_lock();
1438 fdt = files_fdtable(files);
1439 for (fd = filp->f_pos-2;
1440 fd < fdt->max_fds;
1441 fd++, filp->f_pos++) {
1442 char name[PROC_NUMBUF];
1443 int len;
1444
1445 if (!fcheck_files(files, fd))
1446 continue;
1447 rcu_read_unlock();
1448
1449 len = snprintf(name, sizeof(name), "%d", fd);
1450 if (proc_fill_cache(filp, dirent, filldir,
1451 name, len, instantiate,
1452 p, &fd) < 0) {
1453 rcu_read_lock();
1454 break;
1455 }
1456 rcu_read_lock();
1457 }
1458 rcu_read_unlock();
1459 put_files_struct(files);
1460 }
1461 out:
1462 put_task_struct(p);
1463 out_no_task:
1464 return retval;
1465 }
1466
1467 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1468 struct nameidata *nd)
1469 {
1470 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1471 }
1472
1473 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1474 {
1475 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1476 }
1477
1478 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1479 size_t len, loff_t *ppos)
1480 {
1481 char tmp[PROC_FDINFO_MAX];
1482 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, NULL, tmp);
1483 if (!err)
1484 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1485 return err;
1486 }
1487
1488 static const struct file_operations proc_fdinfo_file_operations = {
1489 .open = nonseekable_open,
1490 .read = proc_fdinfo_read,
1491 };
1492
1493 static const struct file_operations proc_fd_operations = {
1494 .read = generic_read_dir,
1495 .readdir = proc_readfd,
1496 };
1497
1498 /*
1499 * /proc/pid/fd needs a special permission handler so that a process can still
1500 * access /proc/self/fd after it has executed a setuid().
1501 */
1502 static int proc_fd_permission(struct inode *inode, int mask,
1503 struct nameidata *nd)
1504 {
1505 int rv;
1506
1507 rv = generic_permission(inode, mask, NULL);
1508 if (rv == 0)
1509 return 0;
1510 if (task_pid(current) == proc_pid(inode))
1511 rv = 0;
1512 return rv;
1513 }
1514
1515 /*
1516 * proc directories can do almost nothing..
1517 */
1518 static const struct inode_operations proc_fd_inode_operations = {
1519 .lookup = proc_lookupfd,
1520 .permission = proc_fd_permission,
1521 .setattr = proc_setattr,
1522 };
1523
1524 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
1525 struct dentry *dentry, struct task_struct *task, const void *ptr)
1526 {
1527 unsigned fd = *(unsigned *)ptr;
1528 struct inode *inode;
1529 struct proc_inode *ei;
1530 struct dentry *error = ERR_PTR(-ENOENT);
1531
1532 inode = proc_pid_make_inode(dir->i_sb, task);
1533 if (!inode)
1534 goto out;
1535 ei = PROC_I(inode);
1536 ei->fd = fd;
1537 inode->i_mode = S_IFREG | S_IRUSR;
1538 inode->i_fop = &proc_fdinfo_file_operations;
1539 dentry->d_op = &tid_fd_dentry_operations;
1540 d_add(dentry, inode);
1541 /* Close the race of the process dying before we return the dentry */
1542 if (tid_fd_revalidate(dentry, NULL))
1543 error = NULL;
1544
1545 out:
1546 return error;
1547 }
1548
1549 static struct dentry *proc_lookupfdinfo(struct inode *dir,
1550 struct dentry *dentry,
1551 struct nameidata *nd)
1552 {
1553 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
1554 }
1555
1556 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
1557 {
1558 return proc_readfd_common(filp, dirent, filldir,
1559 proc_fdinfo_instantiate);
1560 }
1561
1562 static const struct file_operations proc_fdinfo_operations = {
1563 .read = generic_read_dir,
1564 .readdir = proc_readfdinfo,
1565 };
1566
1567 /*
1568 * proc directories can do almost nothing..
1569 */
1570 static const struct inode_operations proc_fdinfo_inode_operations = {
1571 .lookup = proc_lookupfdinfo,
1572 .setattr = proc_setattr,
1573 };
1574
1575
1576 static struct dentry *proc_pident_instantiate(struct inode *dir,
1577 struct dentry *dentry, struct task_struct *task, const void *ptr)
1578 {
1579 const struct pid_entry *p = ptr;
1580 struct inode *inode;
1581 struct proc_inode *ei;
1582 struct dentry *error = ERR_PTR(-EINVAL);
1583
1584 inode = proc_pid_make_inode(dir->i_sb, task);
1585 if (!inode)
1586 goto out;
1587
1588 ei = PROC_I(inode);
1589 inode->i_mode = p->mode;
1590 if (S_ISDIR(inode->i_mode))
1591 inode->i_nlink = 2; /* Use getattr to fix if necessary */
1592 if (p->iop)
1593 inode->i_op = p->iop;
1594 if (p->fop)
1595 inode->i_fop = p->fop;
1596 ei->op = p->op;
1597 dentry->d_op = &pid_dentry_operations;
1598 d_add(dentry, inode);
1599 /* Close the race of the process dying before we return the dentry */
1600 if (pid_revalidate(dentry, NULL))
1601 error = NULL;
1602 out:
1603 return error;
1604 }
1605
1606 static struct dentry *proc_pident_lookup(struct inode *dir,
1607 struct dentry *dentry,
1608 const struct pid_entry *ents,
1609 unsigned int nents)
1610 {
1611 struct inode *inode;
1612 struct dentry *error;
1613 struct task_struct *task = get_proc_task(dir);
1614 const struct pid_entry *p, *last;
1615
1616 error = ERR_PTR(-ENOENT);
1617 inode = NULL;
1618
1619 if (!task)
1620 goto out_no_task;
1621
1622 /*
1623 * Yes, it does not scale. And it should not. Don't add
1624 * new entries into /proc/<tgid>/ without very good reasons.
1625 */
1626 last = &ents[nents - 1];
1627 for (p = ents; p <= last; p++) {
1628 if (p->len != dentry->d_name.len)
1629 continue;
1630 if (!memcmp(dentry->d_name.name, p->name, p->len))
1631 break;
1632 }
1633 if (p > last)
1634 goto out;
1635
1636 error = proc_pident_instantiate(dir, dentry, task, p);
1637 out:
1638 put_task_struct(task);
1639 out_no_task:
1640 return error;
1641 }
1642
1643 static int proc_pident_fill_cache(struct file *filp, void *dirent,
1644 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
1645 {
1646 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
1647 proc_pident_instantiate, task, p);
1648 }
1649
1650 static int proc_pident_readdir(struct file *filp,
1651 void *dirent, filldir_t filldir,
1652 const struct pid_entry *ents, unsigned int nents)
1653 {
1654 int i;
1655 int pid;
1656 struct dentry *dentry = filp->f_path.dentry;
1657 struct inode *inode = dentry->d_inode;
1658 struct task_struct *task = get_proc_task(inode);
1659 const struct pid_entry *p, *last;
1660 ino_t ino;
1661 int ret;
1662
1663 ret = -ENOENT;
1664 if (!task)
1665 goto out_no_task;
1666
1667 ret = 0;
1668 pid = task->pid;
1669 i = filp->f_pos;
1670 switch (i) {
1671 case 0:
1672 ino = inode->i_ino;
1673 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
1674 goto out;
1675 i++;
1676 filp->f_pos++;
1677 /* fall through */
1678 case 1:
1679 ino = parent_ino(dentry);
1680 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
1681 goto out;
1682 i++;
1683 filp->f_pos++;
1684 /* fall through */
1685 default:
1686 i -= 2;
1687 if (i >= nents) {
1688 ret = 1;
1689 goto out;
1690 }
1691 p = ents + i;
1692 last = &ents[nents - 1];
1693 while (p <= last) {
1694 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
1695 goto out;
1696 filp->f_pos++;
1697 p++;
1698 }
1699 }
1700
1701 ret = 1;
1702 out:
1703 put_task_struct(task);
1704 out_no_task:
1705 return ret;
1706 }
1707
1708 #ifdef CONFIG_SECURITY
1709 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
1710 size_t count, loff_t *ppos)
1711 {
1712 struct inode * inode = file->f_path.dentry->d_inode;
1713 char *p = NULL;
1714 ssize_t length;
1715 struct task_struct *task = get_proc_task(inode);
1716
1717 if (!task)
1718 return -ESRCH;
1719
1720 length = security_getprocattr(task,
1721 (char*)file->f_path.dentry->d_name.name,
1722 &p);
1723 put_task_struct(task);
1724 if (length > 0)
1725 length = simple_read_from_buffer(buf, count, ppos, p, length);
1726 kfree(p);
1727 return length;
1728 }
1729
1730 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
1731 size_t count, loff_t *ppos)
1732 {
1733 struct inode * inode = file->f_path.dentry->d_inode;
1734 char *page;
1735 ssize_t length;
1736 struct task_struct *task = get_proc_task(inode);
1737
1738 length = -ESRCH;
1739 if (!task)
1740 goto out_no_task;
1741 if (count > PAGE_SIZE)
1742 count = PAGE_SIZE;
1743
1744 /* No partial writes. */
1745 length = -EINVAL;
1746 if (*ppos != 0)
1747 goto out;
1748
1749 length = -ENOMEM;
1750 page = (char*)__get_free_page(GFP_USER);
1751 if (!page)
1752 goto out;
1753
1754 length = -EFAULT;
1755 if (copy_from_user(page, buf, count))
1756 goto out_free;
1757
1758 length = security_setprocattr(task,
1759 (char*)file->f_path.dentry->d_name.name,
1760 (void*)page, count);
1761 out_free:
1762 free_page((unsigned long) page);
1763 out:
1764 put_task_struct(task);
1765 out_no_task:
1766 return length;
1767 }
1768
1769 static const struct file_operations proc_pid_attr_operations = {
1770 .read = proc_pid_attr_read,
1771 .write = proc_pid_attr_write,
1772 };
1773
1774 static const struct pid_entry attr_dir_stuff[] = {
1775 REG("current", S_IRUGO|S_IWUGO, pid_attr),
1776 REG("prev", S_IRUGO, pid_attr),
1777 REG("exec", S_IRUGO|S_IWUGO, pid_attr),
1778 REG("fscreate", S_IRUGO|S_IWUGO, pid_attr),
1779 REG("keycreate", S_IRUGO|S_IWUGO, pid_attr),
1780 REG("sockcreate", S_IRUGO|S_IWUGO, pid_attr),
1781 };
1782
1783 static int proc_attr_dir_readdir(struct file * filp,
1784 void * dirent, filldir_t filldir)
1785 {
1786 return proc_pident_readdir(filp,dirent,filldir,
1787 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
1788 }
1789
1790 static const struct file_operations proc_attr_dir_operations = {
1791 .read = generic_read_dir,
1792 .readdir = proc_attr_dir_readdir,
1793 };
1794
1795 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
1796 struct dentry *dentry, struct nameidata *nd)
1797 {
1798 return proc_pident_lookup(dir, dentry,
1799 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
1800 }
1801
1802 static const struct inode_operations proc_attr_dir_inode_operations = {
1803 .lookup = proc_attr_dir_lookup,
1804 .getattr = pid_getattr,
1805 .setattr = proc_setattr,
1806 };
1807
1808 #endif
1809
1810 /*
1811 * /proc/self:
1812 */
1813 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
1814 int buflen)
1815 {
1816 char tmp[PROC_NUMBUF];
1817 sprintf(tmp, "%d", current->tgid);
1818 return vfs_readlink(dentry,buffer,buflen,tmp);
1819 }
1820
1821 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
1822 {
1823 char tmp[PROC_NUMBUF];
1824 sprintf(tmp, "%d", current->tgid);
1825 return ERR_PTR(vfs_follow_link(nd,tmp));
1826 }
1827
1828 static const struct inode_operations proc_self_inode_operations = {
1829 .readlink = proc_self_readlink,
1830 .follow_link = proc_self_follow_link,
1831 };
1832
1833 /*
1834 * proc base
1835 *
1836 * These are the directory entries in the root directory of /proc
1837 * that properly belong to the /proc filesystem, as they describe
1838 * describe something that is process related.
1839 */
1840 static const struct pid_entry proc_base_stuff[] = {
1841 NOD("self", S_IFLNK|S_IRWXUGO,
1842 &proc_self_inode_operations, NULL, {}),
1843 };
1844
1845 /*
1846 * Exceptional case: normally we are not allowed to unhash a busy
1847 * directory. In this case, however, we can do it - no aliasing problems
1848 * due to the way we treat inodes.
1849 */
1850 static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
1851 {
1852 struct inode *inode = dentry->d_inode;
1853 struct task_struct *task = get_proc_task(inode);
1854 if (task) {
1855 put_task_struct(task);
1856 return 1;
1857 }
1858 d_drop(dentry);
1859 return 0;
1860 }
1861
1862 static struct dentry_operations proc_base_dentry_operations =
1863 {
1864 .d_revalidate = proc_base_revalidate,
1865 .d_delete = pid_delete_dentry,
1866 };
1867
1868 static struct dentry *proc_base_instantiate(struct inode *dir,
1869 struct dentry *dentry, struct task_struct *task, const void *ptr)
1870 {
1871 const struct pid_entry *p = ptr;
1872 struct inode *inode;
1873 struct proc_inode *ei;
1874 struct dentry *error = ERR_PTR(-EINVAL);
1875
1876 /* Allocate the inode */
1877 error = ERR_PTR(-ENOMEM);
1878 inode = new_inode(dir->i_sb);
1879 if (!inode)
1880 goto out;
1881
1882 /* Initialize the inode */
1883 ei = PROC_I(inode);
1884 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1885
1886 /*
1887 * grab the reference to the task.
1888 */
1889 ei->pid = get_task_pid(task, PIDTYPE_PID);
1890 if (!ei->pid)
1891 goto out_iput;
1892
1893 inode->i_uid = 0;
1894 inode->i_gid = 0;
1895 inode->i_mode = p->mode;
1896 if (S_ISDIR(inode->i_mode))
1897 inode->i_nlink = 2;
1898 if (S_ISLNK(inode->i_mode))
1899 inode->i_size = 64;
1900 if (p->iop)
1901 inode->i_op = p->iop;
1902 if (p->fop)
1903 inode->i_fop = p->fop;
1904 ei->op = p->op;
1905 dentry->d_op = &proc_base_dentry_operations;
1906 d_add(dentry, inode);
1907 error = NULL;
1908 out:
1909 return error;
1910 out_iput:
1911 iput(inode);
1912 goto out;
1913 }
1914
1915 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
1916 {
1917 struct dentry *error;
1918 struct task_struct *task = get_proc_task(dir);
1919 const struct pid_entry *p, *last;
1920
1921 error = ERR_PTR(-ENOENT);
1922
1923 if (!task)
1924 goto out_no_task;
1925
1926 /* Lookup the directory entry */
1927 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
1928 for (p = proc_base_stuff; p <= last; p++) {
1929 if (p->len != dentry->d_name.len)
1930 continue;
1931 if (!memcmp(dentry->d_name.name, p->name, p->len))
1932 break;
1933 }
1934 if (p > last)
1935 goto out;
1936
1937 error = proc_base_instantiate(dir, dentry, task, p);
1938
1939 out:
1940 put_task_struct(task);
1941 out_no_task:
1942 return error;
1943 }
1944
1945 static int proc_base_fill_cache(struct file *filp, void *dirent,
1946 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
1947 {
1948 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
1949 proc_base_instantiate, task, p);
1950 }
1951
1952 #ifdef CONFIG_TASK_IO_ACCOUNTING
1953 static int proc_pid_io_accounting(struct task_struct *task, char *buffer)
1954 {
1955 return sprintf(buffer,
1956 #ifdef CONFIG_TASK_XACCT
1957 "rchar: %llu\n"
1958 "wchar: %llu\n"
1959 "syscr: %llu\n"
1960 "syscw: %llu\n"
1961 #endif
1962 "read_bytes: %llu\n"
1963 "write_bytes: %llu\n"
1964 "cancelled_write_bytes: %llu\n",
1965 #ifdef CONFIG_TASK_XACCT
1966 (unsigned long long)task->rchar,
1967 (unsigned long long)task->wchar,
1968 (unsigned long long)task->syscr,
1969 (unsigned long long)task->syscw,
1970 #endif
1971 (unsigned long long)task->ioac.read_bytes,
1972 (unsigned long long)task->ioac.write_bytes,
1973 (unsigned long long)task->ioac.cancelled_write_bytes);
1974 }
1975 #endif
1976
1977 /*
1978 * Thread groups
1979 */
1980 static const struct file_operations proc_task_operations;
1981 static const struct inode_operations proc_task_inode_operations;
1982
1983 static const struct pid_entry tgid_base_stuff[] = {
1984 DIR("task", S_IRUGO|S_IXUGO, task),
1985 DIR("fd", S_IRUSR|S_IXUSR, fd),
1986 DIR("fdinfo", S_IRUSR|S_IXUSR, fdinfo),
1987 INF("environ", S_IRUSR, pid_environ),
1988 INF("auxv", S_IRUSR, pid_auxv),
1989 INF("status", S_IRUGO, pid_status),
1990 INF("cmdline", S_IRUGO, pid_cmdline),
1991 INF("stat", S_IRUGO, tgid_stat),
1992 INF("statm", S_IRUGO, pid_statm),
1993 REG("maps", S_IRUGO, maps),
1994 #ifdef CONFIG_NUMA
1995 REG("numa_maps", S_IRUGO, numa_maps),
1996 #endif
1997 REG("mem", S_IRUSR|S_IWUSR, mem),
1998 #ifdef CONFIG_SECCOMP
1999 REG("seccomp", S_IRUSR|S_IWUSR, seccomp),
2000 #endif
2001 LNK("cwd", cwd),
2002 LNK("root", root),
2003 LNK("exe", exe),
2004 REG("mounts", S_IRUGO, mounts),
2005 REG("mountstats", S_IRUSR, mountstats),
2006 #ifdef CONFIG_MMU
2007 REG("clear_refs", S_IWUSR, clear_refs),
2008 REG("smaps", S_IRUGO, smaps),
2009 #endif
2010 #ifdef CONFIG_SECURITY
2011 DIR("attr", S_IRUGO|S_IXUGO, attr_dir),
2012 #endif
2013 #ifdef CONFIG_KALLSYMS
2014 INF("wchan", S_IRUGO, pid_wchan),
2015 #endif
2016 #ifdef CONFIG_SCHEDSTATS
2017 INF("schedstat", S_IRUGO, pid_schedstat),
2018 #endif
2019 #ifdef CONFIG_CPUSETS
2020 REG("cpuset", S_IRUGO, cpuset),
2021 #endif
2022 INF("oom_score", S_IRUGO, oom_score),
2023 REG("oom_adj", S_IRUGO|S_IWUSR, oom_adjust),
2024 #ifdef CONFIG_AUDITSYSCALL
2025 REG("loginuid", S_IWUSR|S_IRUGO, loginuid),
2026 #endif
2027 #ifdef CONFIG_FAULT_INJECTION
2028 REG("make-it-fail", S_IRUGO|S_IWUSR, fault_inject),
2029 #endif
2030 #ifdef CONFIG_TASK_IO_ACCOUNTING
2031 INF("io", S_IRUGO, pid_io_accounting),
2032 #endif
2033 };
2034
2035 static int proc_tgid_base_readdir(struct file * filp,
2036 void * dirent, filldir_t filldir)
2037 {
2038 return proc_pident_readdir(filp,dirent,filldir,
2039 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2040 }
2041
2042 static const struct file_operations proc_tgid_base_operations = {
2043 .read = generic_read_dir,
2044 .readdir = proc_tgid_base_readdir,
2045 };
2046
2047 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2048 return proc_pident_lookup(dir, dentry,
2049 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2050 }
2051
2052 static const struct inode_operations proc_tgid_base_inode_operations = {
2053 .lookup = proc_tgid_base_lookup,
2054 .getattr = pid_getattr,
2055 .setattr = proc_setattr,
2056 };
2057
2058 /**
2059 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2060 *
2061 * @task: task that should be flushed.
2062 *
2063 * Looks in the dcache for
2064 * /proc/@pid
2065 * /proc/@tgid/task/@pid
2066 * if either directory is present flushes it and all of it'ts children
2067 * from the dcache.
2068 *
2069 * It is safe and reasonable to cache /proc entries for a task until
2070 * that task exits. After that they just clog up the dcache with
2071 * useless entries, possibly causing useful dcache entries to be
2072 * flushed instead. This routine is proved to flush those useless
2073 * dcache entries at process exit time.
2074 *
2075 * NOTE: This routine is just an optimization so it does not guarantee
2076 * that no dcache entries will exist at process exit time it
2077 * just makes it very unlikely that any will persist.
2078 */
2079 void proc_flush_task(struct task_struct *task)
2080 {
2081 struct dentry *dentry, *leader, *dir;
2082 char buf[PROC_NUMBUF];
2083 struct qstr name;
2084
2085 name.name = buf;
2086 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
2087 dentry = d_hash_and_lookup(proc_mnt->mnt_root, &name);
2088 if (dentry) {
2089 shrink_dcache_parent(dentry);
2090 d_drop(dentry);
2091 dput(dentry);
2092 }
2093
2094 if (thread_group_leader(task))
2095 goto out;
2096
2097 name.name = buf;
2098 name.len = snprintf(buf, sizeof(buf), "%d", task->tgid);
2099 leader = d_hash_and_lookup(proc_mnt->mnt_root, &name);
2100 if (!leader)
2101 goto out;
2102
2103 name.name = "task";
2104 name.len = strlen(name.name);
2105 dir = d_hash_and_lookup(leader, &name);
2106 if (!dir)
2107 goto out_put_leader;
2108
2109 name.name = buf;
2110 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
2111 dentry = d_hash_and_lookup(dir, &name);
2112 if (dentry) {
2113 shrink_dcache_parent(dentry);
2114 d_drop(dentry);
2115 dput(dentry);
2116 }
2117
2118 dput(dir);
2119 out_put_leader:
2120 dput(leader);
2121 out:
2122 return;
2123 }
2124
2125 static struct dentry *proc_pid_instantiate(struct inode *dir,
2126 struct dentry * dentry,
2127 struct task_struct *task, const void *ptr)
2128 {
2129 struct dentry *error = ERR_PTR(-ENOENT);
2130 struct inode *inode;
2131
2132 inode = proc_pid_make_inode(dir->i_sb, task);
2133 if (!inode)
2134 goto out;
2135
2136 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2137 inode->i_op = &proc_tgid_base_inode_operations;
2138 inode->i_fop = &proc_tgid_base_operations;
2139 inode->i_flags|=S_IMMUTABLE;
2140 inode->i_nlink = 5;
2141 #ifdef CONFIG_SECURITY
2142 inode->i_nlink += 1;
2143 #endif
2144
2145 dentry->d_op = &pid_dentry_operations;
2146
2147 d_add(dentry, inode);
2148 /* Close the race of the process dying before we return the dentry */
2149 if (pid_revalidate(dentry, NULL))
2150 error = NULL;
2151 out:
2152 return error;
2153 }
2154
2155 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2156 {
2157 struct dentry *result = ERR_PTR(-ENOENT);
2158 struct task_struct *task;
2159 unsigned tgid;
2160
2161 result = proc_base_lookup(dir, dentry);
2162 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2163 goto out;
2164
2165 tgid = name_to_int(dentry);
2166 if (tgid == ~0U)
2167 goto out;
2168
2169 rcu_read_lock();
2170 task = find_task_by_pid(tgid);
2171 if (task)
2172 get_task_struct(task);
2173 rcu_read_unlock();
2174 if (!task)
2175 goto out;
2176
2177 result = proc_pid_instantiate(dir, dentry, task, NULL);
2178 put_task_struct(task);
2179 out:
2180 return result;
2181 }
2182
2183 /*
2184 * Find the first task with tgid >= tgid
2185 *
2186 */
2187 static struct task_struct *next_tgid(unsigned int tgid)
2188 {
2189 struct task_struct *task;
2190 struct pid *pid;
2191
2192 rcu_read_lock();
2193 retry:
2194 task = NULL;
2195 pid = find_ge_pid(tgid);
2196 if (pid) {
2197 tgid = pid->nr + 1;
2198 task = pid_task(pid, PIDTYPE_PID);
2199 /* What we to know is if the pid we have find is the
2200 * pid of a thread_group_leader. Testing for task
2201 * being a thread_group_leader is the obvious thing
2202 * todo but there is a window when it fails, due to
2203 * the pid transfer logic in de_thread.
2204 *
2205 * So we perform the straight forward test of seeing
2206 * if the pid we have found is the pid of a thread
2207 * group leader, and don't worry if the task we have
2208 * found doesn't happen to be a thread group leader.
2209 * As we don't care in the case of readdir.
2210 */
2211 if (!task || !has_group_leader_pid(task))
2212 goto retry;
2213 get_task_struct(task);
2214 }
2215 rcu_read_unlock();
2216 return task;
2217 }
2218
2219 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
2220
2221 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2222 struct task_struct *task, int tgid)
2223 {
2224 char name[PROC_NUMBUF];
2225 int len = snprintf(name, sizeof(name), "%d", tgid);
2226 return proc_fill_cache(filp, dirent, filldir, name, len,
2227 proc_pid_instantiate, task, NULL);
2228 }
2229
2230 /* for the /proc/ directory itself, after non-process stuff has been done */
2231 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2232 {
2233 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2234 struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
2235 struct task_struct *task;
2236 int tgid;
2237
2238 if (!reaper)
2239 goto out_no_task;
2240
2241 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
2242 const struct pid_entry *p = &proc_base_stuff[nr];
2243 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
2244 goto out;
2245 }
2246
2247 tgid = filp->f_pos - TGID_OFFSET;
2248 for (task = next_tgid(tgid);
2249 task;
2250 put_task_struct(task), task = next_tgid(tgid + 1)) {
2251 tgid = task->pid;
2252 filp->f_pos = tgid + TGID_OFFSET;
2253 if (proc_pid_fill_cache(filp, dirent, filldir, task, tgid) < 0) {
2254 put_task_struct(task);
2255 goto out;
2256 }
2257 }
2258 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
2259 out:
2260 put_task_struct(reaper);
2261 out_no_task:
2262 return 0;
2263 }
2264
2265 /*
2266 * Tasks
2267 */
2268 static const struct pid_entry tid_base_stuff[] = {
2269 DIR("fd", S_IRUSR|S_IXUSR, fd),
2270 DIR("fdinfo", S_IRUSR|S_IXUSR, fdinfo),
2271 INF("environ", S_IRUSR, pid_environ),
2272 INF("auxv", S_IRUSR, pid_auxv),
2273 INF("status", S_IRUGO, pid_status),
2274 INF("cmdline", S_IRUGO, pid_cmdline),
2275 INF("stat", S_IRUGO, tid_stat),
2276 INF("statm", S_IRUGO, pid_statm),
2277 REG("maps", S_IRUGO, maps),
2278 #ifdef CONFIG_NUMA
2279 REG("numa_maps", S_IRUGO, numa_maps),
2280 #endif
2281 REG("mem", S_IRUSR|S_IWUSR, mem),
2282 #ifdef CONFIG_SECCOMP
2283 REG("seccomp", S_IRUSR|S_IWUSR, seccomp),
2284 #endif
2285 LNK("cwd", cwd),
2286 LNK("root", root),
2287 LNK("exe", exe),
2288 REG("mounts", S_IRUGO, mounts),
2289 #ifdef CONFIG_MMU
2290 REG("clear_refs", S_IWUSR, clear_refs),
2291 REG("smaps", S_IRUGO, smaps),
2292 #endif
2293 #ifdef CONFIG_SECURITY
2294 DIR("attr", S_IRUGO|S_IXUGO, attr_dir),
2295 #endif
2296 #ifdef CONFIG_KALLSYMS
2297 INF("wchan", S_IRUGO, pid_wchan),
2298 #endif
2299 #ifdef CONFIG_SCHEDSTATS
2300 INF("schedstat", S_IRUGO, pid_schedstat),
2301 #endif
2302 #ifdef CONFIG_CPUSETS
2303 REG("cpuset", S_IRUGO, cpuset),
2304 #endif
2305 INF("oom_score", S_IRUGO, oom_score),
2306 REG("oom_adj", S_IRUGO|S_IWUSR, oom_adjust),
2307 #ifdef CONFIG_AUDITSYSCALL
2308 REG("loginuid", S_IWUSR|S_IRUGO, loginuid),
2309 #endif
2310 #ifdef CONFIG_FAULT_INJECTION
2311 REG("make-it-fail", S_IRUGO|S_IWUSR, fault_inject),
2312 #endif
2313 };
2314
2315 static int proc_tid_base_readdir(struct file * filp,
2316 void * dirent, filldir_t filldir)
2317 {
2318 return proc_pident_readdir(filp,dirent,filldir,
2319 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
2320 }
2321
2322 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2323 return proc_pident_lookup(dir, dentry,
2324 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
2325 }
2326
2327 static const struct file_operations proc_tid_base_operations = {
2328 .read = generic_read_dir,
2329 .readdir = proc_tid_base_readdir,
2330 };
2331
2332 static const struct inode_operations proc_tid_base_inode_operations = {
2333 .lookup = proc_tid_base_lookup,
2334 .getattr = pid_getattr,
2335 .setattr = proc_setattr,
2336 };
2337
2338 static struct dentry *proc_task_instantiate(struct inode *dir,
2339 struct dentry *dentry, struct task_struct *task, const void *ptr)
2340 {
2341 struct dentry *error = ERR_PTR(-ENOENT);
2342 struct inode *inode;
2343 inode = proc_pid_make_inode(dir->i_sb, task);
2344
2345 if (!inode)
2346 goto out;
2347 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2348 inode->i_op = &proc_tid_base_inode_operations;
2349 inode->i_fop = &proc_tid_base_operations;
2350 inode->i_flags|=S_IMMUTABLE;
2351 inode->i_nlink = 4;
2352 #ifdef CONFIG_SECURITY
2353 inode->i_nlink += 1;
2354 #endif
2355
2356 dentry->d_op = &pid_dentry_operations;
2357
2358 d_add(dentry, inode);
2359 /* Close the race of the process dying before we return the dentry */
2360 if (pid_revalidate(dentry, NULL))
2361 error = NULL;
2362 out:
2363 return error;
2364 }
2365
2366 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2367 {
2368 struct dentry *result = ERR_PTR(-ENOENT);
2369 struct task_struct *task;
2370 struct task_struct *leader = get_proc_task(dir);
2371 unsigned tid;
2372
2373 if (!leader)
2374 goto out_no_task;
2375
2376 tid = name_to_int(dentry);
2377 if (tid == ~0U)
2378 goto out;
2379
2380 rcu_read_lock();
2381 task = find_task_by_pid(tid);
2382 if (task)
2383 get_task_struct(task);
2384 rcu_read_unlock();
2385 if (!task)
2386 goto out;
2387 if (leader->tgid != task->tgid)
2388 goto out_drop_task;
2389
2390 result = proc_task_instantiate(dir, dentry, task, NULL);
2391 out_drop_task:
2392 put_task_struct(task);
2393 out:
2394 put_task_struct(leader);
2395 out_no_task:
2396 return result;
2397 }
2398
2399 /*
2400 * Find the first tid of a thread group to return to user space.
2401 *
2402 * Usually this is just the thread group leader, but if the users
2403 * buffer was too small or there was a seek into the middle of the
2404 * directory we have more work todo.
2405 *
2406 * In the case of a short read we start with find_task_by_pid.
2407 *
2408 * In the case of a seek we start with the leader and walk nr
2409 * threads past it.
2410 */
2411 static struct task_struct *first_tid(struct task_struct *leader,
2412 int tid, int nr)
2413 {
2414 struct task_struct *pos;
2415
2416 rcu_read_lock();
2417 /* Attempt to start with the pid of a thread */
2418 if (tid && (nr > 0)) {
2419 pos = find_task_by_pid(tid);
2420 if (pos && (pos->group_leader == leader))
2421 goto found;
2422 }
2423
2424 /* If nr exceeds the number of threads there is nothing todo */
2425 pos = NULL;
2426 if (nr && nr >= get_nr_threads(leader))
2427 goto out;
2428
2429 /* If we haven't found our starting place yet start
2430 * with the leader and walk nr threads forward.
2431 */
2432 for (pos = leader; nr > 0; --nr) {
2433 pos = next_thread(pos);
2434 if (pos == leader) {
2435 pos = NULL;
2436 goto out;
2437 }
2438 }
2439 found:
2440 get_task_struct(pos);
2441 out:
2442 rcu_read_unlock();
2443 return pos;
2444 }
2445
2446 /*
2447 * Find the next thread in the thread list.
2448 * Return NULL if there is an error or no next thread.
2449 *
2450 * The reference to the input task_struct is released.
2451 */
2452 static struct task_struct *next_tid(struct task_struct *start)
2453 {
2454 struct task_struct *pos = NULL;
2455 rcu_read_lock();
2456 if (pid_alive(start)) {
2457 pos = next_thread(start);
2458 if (thread_group_leader(pos))
2459 pos = NULL;
2460 else
2461 get_task_struct(pos);
2462 }
2463 rcu_read_unlock();
2464 put_task_struct(start);
2465 return pos;
2466 }
2467
2468 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2469 struct task_struct *task, int tid)
2470 {
2471 char name[PROC_NUMBUF];
2472 int len = snprintf(name, sizeof(name), "%d", tid);
2473 return proc_fill_cache(filp, dirent, filldir, name, len,
2474 proc_task_instantiate, task, NULL);
2475 }
2476
2477 /* for the /proc/TGID/task/ directories */
2478 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
2479 {
2480 struct dentry *dentry = filp->f_path.dentry;
2481 struct inode *inode = dentry->d_inode;
2482 struct task_struct *leader = NULL;
2483 struct task_struct *task;
2484 int retval = -ENOENT;
2485 ino_t ino;
2486 int tid;
2487 unsigned long pos = filp->f_pos; /* avoiding "long long" filp->f_pos */
2488
2489 task = get_proc_task(inode);
2490 if (!task)
2491 goto out_no_task;
2492 rcu_read_lock();
2493 if (pid_alive(task)) {
2494 leader = task->group_leader;
2495 get_task_struct(leader);
2496 }
2497 rcu_read_unlock();
2498 put_task_struct(task);
2499 if (!leader)
2500 goto out_no_task;
2501 retval = 0;
2502
2503 switch (pos) {
2504 case 0:
2505 ino = inode->i_ino;
2506 if (filldir(dirent, ".", 1, pos, ino, DT_DIR) < 0)
2507 goto out;
2508 pos++;
2509 /* fall through */
2510 case 1:
2511 ino = parent_ino(dentry);
2512 if (filldir(dirent, "..", 2, pos, ino, DT_DIR) < 0)
2513 goto out;
2514 pos++;
2515 /* fall through */
2516 }
2517
2518 /* f_version caches the tgid value that the last readdir call couldn't
2519 * return. lseek aka telldir automagically resets f_version to 0.
2520 */
2521 tid = filp->f_version;
2522 filp->f_version = 0;
2523 for (task = first_tid(leader, tid, pos - 2);
2524 task;
2525 task = next_tid(task), pos++) {
2526 tid = task->pid;
2527 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
2528 /* returning this tgid failed, save it as the first
2529 * pid for the next readir call */
2530 filp->f_version = tid;
2531 put_task_struct(task);
2532 break;
2533 }
2534 }
2535 out:
2536 filp->f_pos = pos;
2537 put_task_struct(leader);
2538 out_no_task:
2539 return retval;
2540 }
2541
2542 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
2543 {
2544 struct inode *inode = dentry->d_inode;
2545 struct task_struct *p = get_proc_task(inode);
2546 generic_fillattr(inode, stat);
2547
2548 if (p) {
2549 rcu_read_lock();
2550 stat->nlink += get_nr_threads(p);
2551 rcu_read_unlock();
2552 put_task_struct(p);
2553 }
2554
2555 return 0;
2556 }
2557
2558 static const struct inode_operations proc_task_inode_operations = {
2559 .lookup = proc_task_lookup,
2560 .getattr = proc_task_getattr,
2561 .setattr = proc_setattr,
2562 };
2563
2564 static const struct file_operations proc_task_operations = {
2565 .read = generic_read_dir,
2566 .readdir = proc_task_readdir,
2567 };
USB Experimental Work