?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
sata_nv: don't rely on NV_INT_DEV indication with ADMA
[pv_ops_mirror.git] / fs / proc / base.c
blob77a57b5799c41cfc2fe84b0bdba00d66b8072eec
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/smp_lock.h>
65 #include <linux/rcupdate.h>
66 #include <linux/kallsyms.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 int len;
94 char *name;
95 mode_t mode;
96 struct inode_operations *iop;
97 struct file_operations *fop;
98 union proc_op op;
99 };
100
101 #define NOD(NAME, MODE, IOP, FOP, OP) { \
102 .len = sizeof(NAME) - 1, \
103 .name = (NAME), \
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 static struct fs_struct *get_fs_struct(struct task_struct *task)
127 {
128 struct fs_struct *fs;
129 task_lock(task);
130 fs = task->fs;
131 if(fs)
132 atomic_inc(&fs->count);
133 task_unlock(task);
134 return fs;
135 }
136
137 static int get_nr_threads(struct task_struct *tsk)
138 {
139 /* Must be called with the rcu_read_lock held */
140 unsigned long flags;
141 int count = 0;
142
143 if (lock_task_sighand(tsk, &flags)) {
144 count = atomic_read(&tsk->signal->count);
145 unlock_task_sighand(tsk, &flags);
146 }
147 return count;
148 }
149
150 static int proc_cwd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
151 {
152 struct task_struct *task = get_proc_task(inode);
153 struct fs_struct *fs = NULL;
154 int result = -ENOENT;
155
156 if (task) {
157 fs = get_fs_struct(task);
158 put_task_struct(task);
159 }
160 if (fs) {
161 read_lock(&fs->lock);
162 *mnt = mntget(fs->pwdmnt);
163 *dentry = dget(fs->pwd);
164 read_unlock(&fs->lock);
165 result = 0;
166 put_fs_struct(fs);
167 }
168 return result;
169 }
170
171 static int proc_root_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
172 {
173 struct task_struct *task = get_proc_task(inode);
174 struct fs_struct *fs = NULL;
175 int result = -ENOENT;
176
177 if (task) {
178 fs = get_fs_struct(task);
179 put_task_struct(task);
180 }
181 if (fs) {
182 read_lock(&fs->lock);
183 *mnt = mntget(fs->rootmnt);
184 *dentry = dget(fs->root);
185 read_unlock(&fs->lock);
186 result = 0;
187 put_fs_struct(fs);
188 }
189 return result;
190 }
191
192 #define MAY_PTRACE(task) \
193 (task == current || \
194 (task->parent == current && \
195 (task->ptrace & PT_PTRACED) && \
196 (task->state == TASK_STOPPED || task->state == TASK_TRACED) && \
197 security_ptrace(current,task) == 0))
198
199 static int proc_pid_environ(struct task_struct *task, char * buffer)
200 {
201 int res = 0;
202 struct mm_struct *mm = get_task_mm(task);
203 if (mm) {
204 unsigned int len = mm->env_end - mm->env_start;
205 if (len > PAGE_SIZE)
206 len = PAGE_SIZE;
207 res = access_process_vm(task, mm->env_start, buffer, len, 0);
208 if (!ptrace_may_attach(task))
209 res = -ESRCH;
210 mmput(mm);
211 }
212 return res;
213 }
214
215 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
216 {
217 int res = 0;
218 unsigned int len;
219 struct mm_struct *mm = get_task_mm(task);
220 if (!mm)
221 goto out;
222 if (!mm->arg_end)
223 goto out_mm; /* Shh! No looking before we're done */
224
225 len = mm->arg_end - mm->arg_start;
226
227 if (len > PAGE_SIZE)
228 len = PAGE_SIZE;
229
230 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
231
232 // If the nul at the end of args has been overwritten, then
233 // assume application is using setproctitle(3).
234 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
235 len = strnlen(buffer, res);
236 if (len < res) {
237 res = len;
238 } else {
239 len = mm->env_end - mm->env_start;
240 if (len > PAGE_SIZE - res)
241 len = PAGE_SIZE - res;
242 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
243 res = strnlen(buffer, res);
244 }
245 }
246 out_mm:
247 mmput(mm);
248 out:
249 return res;
250 }
251
252 static int proc_pid_auxv(struct task_struct *task, char *buffer)
253 {
254 int res = 0;
255 struct mm_struct *mm = get_task_mm(task);
256 if (mm) {
257 unsigned int nwords = 0;
258 do
259 nwords += 2;
260 while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
261 res = nwords * sizeof(mm->saved_auxv[0]);
262 if (res > PAGE_SIZE)
263 res = PAGE_SIZE;
264 memcpy(buffer, mm->saved_auxv, res);
265 mmput(mm);
266 }
267 return res;
268 }
269
270
271 #ifdef CONFIG_KALLSYMS
272 /*
273 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
274 * Returns the resolved symbol. If that fails, simply return the address.
275 */
276 static int proc_pid_wchan(struct task_struct *task, char *buffer)
277 {
278 char *modname;
279 const char *sym_name;
280 unsigned long wchan, size, offset;
281 char namebuf[KSYM_NAME_LEN+1];
282
283 wchan = get_wchan(task);
284
285 sym_name = kallsyms_lookup(wchan, &size, &offset, &modname, namebuf);
286 if (sym_name)
287 return sprintf(buffer, "%s", sym_name);
288 return sprintf(buffer, "%lu", wchan);
289 }
290 #endif /* CONFIG_KALLSYMS */
291
292 #ifdef CONFIG_SCHEDSTATS
293 /*
294 * Provides /proc/PID/schedstat
295 */
296 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
297 {
298 return sprintf(buffer, "%lu %lu %lu\n",
299 task->sched_info.cpu_time,
300 task->sched_info.run_delay,
301 task->sched_info.pcnt);
302 }
303 #endif
304
305 /* The badness from the OOM killer */
306 unsigned long badness(struct task_struct *p, unsigned long uptime);
307 static int proc_oom_score(struct task_struct *task, char *buffer)
308 {
309 unsigned long points;
310 struct timespec uptime;
311
312 do_posix_clock_monotonic_gettime(&uptime);
313 points = badness(task, uptime.tv_sec);
314 return sprintf(buffer, "%lu\n", points);
315 }
316
317 /************************************************************************/
318 /* Here the fs part begins */
319 /************************************************************************/
320
321 /* permission checks */
322 static int proc_fd_access_allowed(struct inode *inode)
323 {
324 struct task_struct *task;
325 int allowed = 0;
326 /* Allow access to a task's file descriptors if it is us or we
327 * may use ptrace attach to the process and find out that
328 * information.
329 */
330 task = get_proc_task(inode);
331 if (task) {
332 allowed = ptrace_may_attach(task);
333 put_task_struct(task);
334 }
335 return allowed;
336 }
337
338 static int proc_setattr(struct dentry *dentry, struct iattr *attr)
339 {
340 int error;
341 struct inode *inode = dentry->d_inode;
342
343 if (attr->ia_valid & ATTR_MODE)
344 return -EPERM;
345
346 error = inode_change_ok(inode, attr);
347 if (!error) {
348 error = security_inode_setattr(dentry, attr);
349 if (!error)
350 error = inode_setattr(inode, attr);
351 }
352 return error;
353 }
354
355 static 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 ns = task->nsproxy->mnt_ns;
375 if (ns)
376 get_mnt_ns(ns);
377 task_unlock(task);
378 put_task_struct(task);
379 }
380
381 if (ns) {
382 ret = -ENOMEM;
383 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
384 if (p) {
385 file->private_data = &p->m;
386 ret = seq_open(file, &mounts_op);
387 if (!ret) {
388 p->m.private = ns;
389 p->event = ns->event;
390 return 0;
391 }
392 kfree(p);
393 }
394 put_mnt_ns(ns);
395 }
396 return ret;
397 }
398
399 static int mounts_release(struct inode *inode, struct file *file)
400 {
401 struct seq_file *m = file->private_data;
402 struct mnt_namespace *ns = m->private;
403 put_mnt_ns(ns);
404 return seq_release(inode, file);
405 }
406
407 static unsigned mounts_poll(struct file *file, poll_table *wait)
408 {
409 struct proc_mounts *p = file->private_data;
410 struct mnt_namespace *ns = p->m.private;
411 unsigned res = 0;
412
413 poll_wait(file, &ns->poll, wait);
414
415 spin_lock(&vfsmount_lock);
416 if (p->event != ns->event) {
417 p->event = ns->event;
418 res = POLLERR;
419 }
420 spin_unlock(&vfsmount_lock);
421
422 return res;
423 }
424
425 static struct file_operations proc_mounts_operations = {
426 .open = mounts_open,
427 .read = seq_read,
428 .llseek = seq_lseek,
429 .release = mounts_release,
430 .poll = mounts_poll,
431 };
432
433 extern struct seq_operations mountstats_op;
434 static int mountstats_open(struct inode *inode, struct file *file)
435 {
436 int ret = seq_open(file, &mountstats_op);
437
438 if (!ret) {
439 struct seq_file *m = file->private_data;
440 struct mnt_namespace *mnt_ns = NULL;
441 struct task_struct *task = get_proc_task(inode);
442
443 if (task) {
444 task_lock(task);
445 if (task->nsproxy)
446 mnt_ns = task->nsproxy->mnt_ns;
447 if (mnt_ns)
448 get_mnt_ns(mnt_ns);
449 task_unlock(task);
450 put_task_struct(task);
451 }
452
453 if (mnt_ns)
454 m->private = mnt_ns;
455 else {
456 seq_release(inode, file);
457 ret = -EINVAL;
458 }
459 }
460 return ret;
461 }
462
463 static struct file_operations proc_mountstats_operations = {
464 .open = mountstats_open,
465 .read = seq_read,
466 .llseek = seq_lseek,
467 .release = mounts_release,
468 };
469
470 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
471
472 static ssize_t proc_info_read(struct file * file, char __user * buf,
473 size_t count, loff_t *ppos)
474 {
475 struct inode * inode = file->f_path.dentry->d_inode;
476 unsigned long page;
477 ssize_t length;
478 struct task_struct *task = get_proc_task(inode);
479
480 length = -ESRCH;
481 if (!task)
482 goto out_no_task;
483
484 if (count > PROC_BLOCK_SIZE)
485 count = PROC_BLOCK_SIZE;
486
487 length = -ENOMEM;
488 if (!(page = __get_free_page(GFP_KERNEL)))
489 goto out;
490
491 length = PROC_I(inode)->op.proc_read(task, (char*)page);
492
493 if (length >= 0)
494 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
495 free_page(page);
496 out:
497 put_task_struct(task);
498 out_no_task:
499 return length;
500 }
501
502 static struct file_operations proc_info_file_operations = {
503 .read = proc_info_read,
504 };
505
506 static int mem_open(struct inode* inode, struct file* file)
507 {
508 file->private_data = (void*)((long)current->self_exec_id);
509 return 0;
510 }
511
512 static ssize_t mem_read(struct file * file, char __user * buf,
513 size_t count, loff_t *ppos)
514 {
515 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
516 char *page;
517 unsigned long src = *ppos;
518 int ret = -ESRCH;
519 struct mm_struct *mm;
520
521 if (!task)
522 goto out_no_task;
523
524 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
525 goto out;
526
527 ret = -ENOMEM;
528 page = (char *)__get_free_page(GFP_USER);
529 if (!page)
530 goto out;
531
532 ret = 0;
533
534 mm = get_task_mm(task);
535 if (!mm)
536 goto out_free;
537
538 ret = -EIO;
539
540 if (file->private_data != (void*)((long)current->self_exec_id))
541 goto out_put;
542
543 ret = 0;
544
545 while (count > 0) {
546 int this_len, retval;
547
548 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
549 retval = access_process_vm(task, src, page, this_len, 0);
550 if (!retval || !MAY_PTRACE(task) || !ptrace_may_attach(task)) {
551 if (!ret)
552 ret = -EIO;
553 break;
554 }
555
556 if (copy_to_user(buf, page, retval)) {
557 ret = -EFAULT;
558 break;
559 }
560
561 ret += retval;
562 src += retval;
563 buf += retval;
564 count -= retval;
565 }
566 *ppos = src;
567
568 out_put:
569 mmput(mm);
570 out_free:
571 free_page((unsigned long) page);
572 out:
573 put_task_struct(task);
574 out_no_task:
575 return ret;
576 }
577
578 #define mem_write NULL
579
580 #ifndef mem_write
581 /* This is a security hazard */
582 static ssize_t mem_write(struct file * file, const char * buf,
583 size_t count, loff_t *ppos)
584 {
585 int copied;
586 char *page;
587 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
588 unsigned long dst = *ppos;
589
590 copied = -ESRCH;
591 if (!task)
592 goto out_no_task;
593
594 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
595 goto out;
596
597 copied = -ENOMEM;
598 page = (char *)__get_free_page(GFP_USER);
599 if (!page)
600 goto out;
601
602 copied = 0;
603 while (count > 0) {
604 int this_len, retval;
605
606 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
607 if (copy_from_user(page, buf, this_len)) {
608 copied = -EFAULT;
609 break;
610 }
611 retval = access_process_vm(task, dst, page, this_len, 1);
612 if (!retval) {
613 if (!copied)
614 copied = -EIO;
615 break;
616 }
617 copied += retval;
618 buf += retval;
619 dst += retval;
620 count -= retval;
621 }
622 *ppos = dst;
623 free_page((unsigned long) page);
624 out:
625 put_task_struct(task);
626 out_no_task:
627 return copied;
628 }
629 #endif
630
631 static loff_t mem_lseek(struct file * file, loff_t offset, int orig)
632 {
633 switch (orig) {
634 case 0:
635 file->f_pos = offset;
636 break;
637 case 1:
638 file->f_pos += offset;
639 break;
640 default:
641 return -EINVAL;
642 }
643 force_successful_syscall_return();
644 return file->f_pos;
645 }
646
647 static struct file_operations proc_mem_operations = {
648 .llseek = mem_lseek,
649 .read = mem_read,
650 .write = mem_write,
651 .open = mem_open,
652 };
653
654 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
655 size_t count, loff_t *ppos)
656 {
657 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
658 char buffer[PROC_NUMBUF];
659 size_t len;
660 int oom_adjust;
661 loff_t __ppos = *ppos;
662
663 if (!task)
664 return -ESRCH;
665 oom_adjust = task->oomkilladj;
666 put_task_struct(task);
667
668 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
669 if (__ppos >= len)
670 return 0;
671 if (count > len-__ppos)
672 count = len-__ppos;
673 if (copy_to_user(buf, buffer + __ppos, count))
674 return -EFAULT;
675 *ppos = __ppos + count;
676 return count;
677 }
678
679 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
680 size_t count, loff_t *ppos)
681 {
682 struct task_struct *task;
683 char buffer[PROC_NUMBUF], *end;
684 int oom_adjust;
685
686 memset(buffer, 0, sizeof(buffer));
687 if (count > sizeof(buffer) - 1)
688 count = sizeof(buffer) - 1;
689 if (copy_from_user(buffer, buf, count))
690 return -EFAULT;
691 oom_adjust = simple_strtol(buffer, &end, 0);
692 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
693 oom_adjust != OOM_DISABLE)
694 return -EINVAL;
695 if (*end == '\n')
696 end++;
697 task = get_proc_task(file->f_path.dentry->d_inode);
698 if (!task)
699 return -ESRCH;
700 if (oom_adjust < task->oomkilladj && !capable(CAP_SYS_RESOURCE)) {
701 put_task_struct(task);
702 return -EACCES;
703 }
704 task->oomkilladj = oom_adjust;
705 put_task_struct(task);
706 if (end - buffer == 0)
707 return -EIO;
708 return end - buffer;
709 }
710
711 static struct file_operations proc_oom_adjust_operations = {
712 .read = oom_adjust_read,
713 .write = oom_adjust_write,
714 };
715
716 #ifdef CONFIG_AUDITSYSCALL
717 #define TMPBUFLEN 21
718 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
719 size_t count, loff_t *ppos)
720 {
721 struct inode * inode = file->f_path.dentry->d_inode;
722 struct task_struct *task = get_proc_task(inode);
723 ssize_t length;
724 char tmpbuf[TMPBUFLEN];
725
726 if (!task)
727 return -ESRCH;
728 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
729 audit_get_loginuid(task->audit_context));
730 put_task_struct(task);
731 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
732 }
733
734 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
735 size_t count, loff_t *ppos)
736 {
737 struct inode * inode = file->f_path.dentry->d_inode;
738 char *page, *tmp;
739 ssize_t length;
740 uid_t loginuid;
741
742 if (!capable(CAP_AUDIT_CONTROL))
743 return -EPERM;
744
745 if (current != pid_task(proc_pid(inode), PIDTYPE_PID))
746 return -EPERM;
747
748 if (count >= PAGE_SIZE)
749 count = PAGE_SIZE - 1;
750
751 if (*ppos != 0) {
752 /* No partial writes. */
753 return -EINVAL;
754 }
755 page = (char*)__get_free_page(GFP_USER);
756 if (!page)
757 return -ENOMEM;
758 length = -EFAULT;
759 if (copy_from_user(page, buf, count))
760 goto out_free_page;
761
762 page[count] = '\0';
763 loginuid = simple_strtoul(page, &tmp, 10);
764 if (tmp == page) {
765 length = -EINVAL;
766 goto out_free_page;
767
768 }
769 length = audit_set_loginuid(current, loginuid);
770 if (likely(length == 0))
771 length = count;
772
773 out_free_page:
774 free_page((unsigned long) page);
775 return length;
776 }
777
778 static struct file_operations proc_loginuid_operations = {
779 .read = proc_loginuid_read,
780 .write = proc_loginuid_write,
781 };
782 #endif
783
784 #ifdef CONFIG_SECCOMP
785 static ssize_t seccomp_read(struct file *file, char __user *buf,
786 size_t count, loff_t *ppos)
787 {
788 struct task_struct *tsk = get_proc_task(file->f_dentry->d_inode);
789 char __buf[20];
790 loff_t __ppos = *ppos;
791 size_t len;
792
793 if (!tsk)
794 return -ESRCH;
795 /* no need to print the trailing zero, so use only len */
796 len = sprintf(__buf, "%u\n", tsk->seccomp.mode);
797 put_task_struct(tsk);
798 if (__ppos >= len)
799 return 0;
800 if (count > len - __ppos)
801 count = len - __ppos;
802 if (copy_to_user(buf, __buf + __ppos, count))
803 return -EFAULT;
804 *ppos = __ppos + count;
805 return count;
806 }
807
808 static ssize_t seccomp_write(struct file *file, const char __user *buf,
809 size_t count, loff_t *ppos)
810 {
811 struct task_struct *tsk = get_proc_task(file->f_dentry->d_inode);
812 char __buf[20], *end;
813 unsigned int seccomp_mode;
814 ssize_t result;
815
816 result = -ESRCH;
817 if (!tsk)
818 goto out_no_task;
819
820 /* can set it only once to be even more secure */
821 result = -EPERM;
822 if (unlikely(tsk->seccomp.mode))
823 goto out;
824
825 result = -EFAULT;
826 memset(__buf, 0, sizeof(__buf));
827 count = min(count, sizeof(__buf) - 1);
828 if (copy_from_user(__buf, buf, count))
829 goto out;
830
831 seccomp_mode = simple_strtoul(__buf, &end, 0);
832 if (*end == '\n')
833 end++;
834 result = -EINVAL;
835 if (seccomp_mode && seccomp_mode <= NR_SECCOMP_MODES) {
836 tsk->seccomp.mode = seccomp_mode;
837 set_tsk_thread_flag(tsk, TIF_SECCOMP);
838 } else
839 goto out;
840 result = -EIO;
841 if (unlikely(!(end - __buf)))
842 goto out;
843 result = end - __buf;
844 out:
845 put_task_struct(tsk);
846 out_no_task:
847 return result;
848 }
849
850 static struct file_operations proc_seccomp_operations = {
851 .read = seccomp_read,
852 .write = seccomp_write,
853 };
854 #endif /* CONFIG_SECCOMP */
855
856 #ifdef CONFIG_FAULT_INJECTION
857 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
858 size_t count, loff_t *ppos)
859 {
860 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
861 char buffer[PROC_NUMBUF];
862 size_t len;
863 int make_it_fail;
864 loff_t __ppos = *ppos;
865
866 if (!task)
867 return -ESRCH;
868 make_it_fail = task->make_it_fail;
869 put_task_struct(task);
870
871 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
872 if (__ppos >= len)
873 return 0;
874 if (count > len-__ppos)
875 count = len-__ppos;
876 if (copy_to_user(buf, buffer + __ppos, count))
877 return -EFAULT;
878 *ppos = __ppos + count;
879 return count;
880 }
881
882 static ssize_t proc_fault_inject_write(struct file * file,
883 const char __user * buf, size_t count, loff_t *ppos)
884 {
885 struct task_struct *task;
886 char buffer[PROC_NUMBUF], *end;
887 int make_it_fail;
888
889 if (!capable(CAP_SYS_RESOURCE))
890 return -EPERM;
891 memset(buffer, 0, sizeof(buffer));
892 if (count > sizeof(buffer) - 1)
893 count = sizeof(buffer) - 1;
894 if (copy_from_user(buffer, buf, count))
895 return -EFAULT;
896 make_it_fail = simple_strtol(buffer, &end, 0);
897 if (*end == '\n')
898 end++;
899 task = get_proc_task(file->f_dentry->d_inode);
900 if (!task)
901 return -ESRCH;
902 task->make_it_fail = make_it_fail;
903 put_task_struct(task);
904 if (end - buffer == 0)
905 return -EIO;
906 return end - buffer;
907 }
908
909 static struct file_operations proc_fault_inject_operations = {
910 .read = proc_fault_inject_read,
911 .write = proc_fault_inject_write,
912 };
913 #endif
914
915 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
916 {
917 struct inode *inode = dentry->d_inode;
918 int error = -EACCES;
919
920 /* We don't need a base pointer in the /proc filesystem */
921 path_release(nd);
922
923 /* Are we allowed to snoop on the tasks file descriptors? */
924 if (!proc_fd_access_allowed(inode))
925 goto out;
926
927 error = PROC_I(inode)->op.proc_get_link(inode, &nd->dentry, &nd->mnt);
928 nd->last_type = LAST_BIND;
929 out:
930 return ERR_PTR(error);
931 }
932
933 static int do_proc_readlink(struct dentry *dentry, struct vfsmount *mnt,
934 char __user *buffer, int buflen)
935 {
936 struct inode * inode;
937 char *tmp = (char*)__get_free_page(GFP_KERNEL), *path;
938 int len;
939
940 if (!tmp)
941 return -ENOMEM;
942
943 inode = dentry->d_inode;
944 path = d_path(dentry, mnt, tmp, PAGE_SIZE);
945 len = PTR_ERR(path);
946 if (IS_ERR(path))
947 goto out;
948 len = tmp + PAGE_SIZE - 1 - path;
949
950 if (len > buflen)
951 len = buflen;
952 if (copy_to_user(buffer, path, len))
953 len = -EFAULT;
954 out:
955 free_page((unsigned long)tmp);
956 return len;
957 }
958
959 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
960 {
961 int error = -EACCES;
962 struct inode *inode = dentry->d_inode;
963 struct dentry *de;
964 struct vfsmount *mnt = NULL;
965
966 /* Are we allowed to snoop on the tasks file descriptors? */
967 if (!proc_fd_access_allowed(inode))
968 goto out;
969
970 error = PROC_I(inode)->op.proc_get_link(inode, &de, &mnt);
971 if (error)
972 goto out;
973
974 error = do_proc_readlink(de, mnt, buffer, buflen);
975 dput(de);
976 mntput(mnt);
977 out:
978 return error;
979 }
980
981 static struct inode_operations proc_pid_link_inode_operations = {
982 .readlink = proc_pid_readlink,
983 .follow_link = proc_pid_follow_link,
984 .setattr = proc_setattr,
985 };
986
987
988 /* building an inode */
989
990 static int task_dumpable(struct task_struct *task)
991 {
992 int dumpable = 0;
993 struct mm_struct *mm;
994
995 task_lock(task);
996 mm = task->mm;
997 if (mm)
998 dumpable = mm->dumpable;
999 task_unlock(task);
1000 if(dumpable == 1)
1001 return 1;
1002 return 0;
1003 }
1004
1005
1006 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1007 {
1008 struct inode * inode;
1009 struct proc_inode *ei;
1010
1011 /* We need a new inode */
1012
1013 inode = new_inode(sb);
1014 if (!inode)
1015 goto out;
1016
1017 /* Common stuff */
1018 ei = PROC_I(inode);
1019 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1020 inode->i_op = &proc_def_inode_operations;
1021
1022 /*
1023 * grab the reference to task.
1024 */
1025 ei->pid = get_task_pid(task, PIDTYPE_PID);
1026 if (!ei->pid)
1027 goto out_unlock;
1028
1029 inode->i_uid = 0;
1030 inode->i_gid = 0;
1031 if (task_dumpable(task)) {
1032 inode->i_uid = task->euid;
1033 inode->i_gid = task->egid;
1034 }
1035 security_task_to_inode(task, inode);
1036
1037 out:
1038 return inode;
1039
1040 out_unlock:
1041 iput(inode);
1042 return NULL;
1043 }
1044
1045 static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1046 {
1047 struct inode *inode = dentry->d_inode;
1048 struct task_struct *task;
1049 generic_fillattr(inode, stat);
1050
1051 rcu_read_lock();
1052 stat->uid = 0;
1053 stat->gid = 0;
1054 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1055 if (task) {
1056 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1057 task_dumpable(task)) {
1058 stat->uid = task->euid;
1059 stat->gid = task->egid;
1060 }
1061 }
1062 rcu_read_unlock();
1063 return 0;
1064 }
1065
1066 /* dentry stuff */
1067
1068 /*
1069 * Exceptional case: normally we are not allowed to unhash a busy
1070 * directory. In this case, however, we can do it - no aliasing problems
1071 * due to the way we treat inodes.
1072 *
1073 * Rewrite the inode's ownerships here because the owning task may have
1074 * performed a setuid(), etc.
1075 *
1076 * Before the /proc/pid/status file was created the only way to read
1077 * the effective uid of a /process was to stat /proc/pid. Reading
1078 * /proc/pid/status is slow enough that procps and other packages
1079 * kept stating /proc/pid. To keep the rules in /proc simple I have
1080 * made this apply to all per process world readable and executable
1081 * directories.
1082 */
1083 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1084 {
1085 struct inode *inode = dentry->d_inode;
1086 struct task_struct *task = get_proc_task(inode);
1087 if (task) {
1088 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1089 task_dumpable(task)) {
1090 inode->i_uid = task->euid;
1091 inode->i_gid = task->egid;
1092 } else {
1093 inode->i_uid = 0;
1094 inode->i_gid = 0;
1095 }
1096 inode->i_mode &= ~(S_ISUID | S_ISGID);
1097 security_task_to_inode(task, inode);
1098 put_task_struct(task);
1099 return 1;
1100 }
1101 d_drop(dentry);
1102 return 0;
1103 }
1104
1105 static int pid_delete_dentry(struct dentry * dentry)
1106 {
1107 /* Is the task we represent dead?
1108 * If so, then don't put the dentry on the lru list,
1109 * kill it immediately.
1110 */
1111 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1112 }
1113
1114 static struct dentry_operations pid_dentry_operations =
1115 {
1116 .d_revalidate = pid_revalidate,
1117 .d_delete = pid_delete_dentry,
1118 };
1119
1120 /* Lookups */
1121
1122 typedef struct dentry *instantiate_t(struct inode *, struct dentry *, struct task_struct *, void *);
1123
1124 /*
1125 * Fill a directory entry.
1126 *
1127 * If possible create the dcache entry and derive our inode number and
1128 * file type from dcache entry.
1129 *
1130 * Since all of the proc inode numbers are dynamically generated, the inode
1131 * numbers do not exist until the inode is cache. This means creating the
1132 * the dcache entry in readdir is necessary to keep the inode numbers
1133 * reported by readdir in sync with the inode numbers reported
1134 * by stat.
1135 */
1136 static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1137 char *name, int len,
1138 instantiate_t instantiate, struct task_struct *task, void *ptr)
1139 {
1140 struct dentry *child, *dir = filp->f_path.dentry;
1141 struct inode *inode;
1142 struct qstr qname;
1143 ino_t ino = 0;
1144 unsigned type = DT_UNKNOWN;
1145
1146 qname.name = name;
1147 qname.len = len;
1148 qname.hash = full_name_hash(name, len);
1149
1150 child = d_lookup(dir, &qname);
1151 if (!child) {
1152 struct dentry *new;
1153 new = d_alloc(dir, &qname);
1154 if (new) {
1155 child = instantiate(dir->d_inode, new, task, ptr);
1156 if (child)
1157 dput(new);
1158 else
1159 child = new;
1160 }
1161 }
1162 if (!child || IS_ERR(child) || !child->d_inode)
1163 goto end_instantiate;
1164 inode = child->d_inode;
1165 if (inode) {
1166 ino = inode->i_ino;
1167 type = inode->i_mode >> 12;
1168 }
1169 dput(child);
1170 end_instantiate:
1171 if (!ino)
1172 ino = find_inode_number(dir, &qname);
1173 if (!ino)
1174 ino = 1;
1175 return filldir(dirent, name, len, filp->f_pos, ino, type);
1176 }
1177
1178 static unsigned name_to_int(struct dentry *dentry)
1179 {
1180 const char *name = dentry->d_name.name;
1181 int len = dentry->d_name.len;
1182 unsigned n = 0;
1183
1184 if (len > 1 && *name == '0')
1185 goto out;
1186 while (len-- > 0) {
1187 unsigned c = *name++ - '0';
1188 if (c > 9)
1189 goto out;
1190 if (n >= (~0U-9)/10)
1191 goto out;
1192 n *= 10;
1193 n += c;
1194 }
1195 return n;
1196 out:
1197 return ~0U;
1198 }
1199
1200 static int proc_fd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
1201 {
1202 struct task_struct *task = get_proc_task(inode);
1203 struct files_struct *files = NULL;
1204 struct file *file;
1205 int fd = proc_fd(inode);
1206
1207 if (task) {
1208 files = get_files_struct(task);
1209 put_task_struct(task);
1210 }
1211 if (files) {
1212 /*
1213 * We are not taking a ref to the file structure, so we must
1214 * hold ->file_lock.
1215 */
1216 spin_lock(&files->file_lock);
1217 file = fcheck_files(files, fd);
1218 if (file) {
1219 *mnt = mntget(file->f_path.mnt);
1220 *dentry = dget(file->f_path.dentry);
1221 spin_unlock(&files->file_lock);
1222 put_files_struct(files);
1223 return 0;
1224 }
1225 spin_unlock(&files->file_lock);
1226 put_files_struct(files);
1227 }
1228 return -ENOENT;
1229 }
1230
1231 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1232 {
1233 struct inode *inode = dentry->d_inode;
1234 struct task_struct *task = get_proc_task(inode);
1235 int fd = proc_fd(inode);
1236 struct files_struct *files;
1237
1238 if (task) {
1239 files = get_files_struct(task);
1240 if (files) {
1241 rcu_read_lock();
1242 if (fcheck_files(files, fd)) {
1243 rcu_read_unlock();
1244 put_files_struct(files);
1245 if (task_dumpable(task)) {
1246 inode->i_uid = task->euid;
1247 inode->i_gid = task->egid;
1248 } else {
1249 inode->i_uid = 0;
1250 inode->i_gid = 0;
1251 }
1252 inode->i_mode &= ~(S_ISUID | S_ISGID);
1253 security_task_to_inode(task, inode);
1254 put_task_struct(task);
1255 return 1;
1256 }
1257 rcu_read_unlock();
1258 put_files_struct(files);
1259 }
1260 put_task_struct(task);
1261 }
1262 d_drop(dentry);
1263 return 0;
1264 }
1265
1266 static struct dentry_operations tid_fd_dentry_operations =
1267 {
1268 .d_revalidate = tid_fd_revalidate,
1269 .d_delete = pid_delete_dentry,
1270 };
1271
1272 static struct dentry *proc_fd_instantiate(struct inode *dir,
1273 struct dentry *dentry, struct task_struct *task, void *ptr)
1274 {
1275 unsigned fd = *(unsigned *)ptr;
1276 struct file *file;
1277 struct files_struct *files;
1278 struct inode *inode;
1279 struct proc_inode *ei;
1280 struct dentry *error = ERR_PTR(-ENOENT);
1281
1282 inode = proc_pid_make_inode(dir->i_sb, task);
1283 if (!inode)
1284 goto out;
1285 ei = PROC_I(inode);
1286 ei->fd = fd;
1287 files = get_files_struct(task);
1288 if (!files)
1289 goto out_iput;
1290 inode->i_mode = S_IFLNK;
1291
1292 /*
1293 * We are not taking a ref to the file structure, so we must
1294 * hold ->file_lock.
1295 */
1296 spin_lock(&files->file_lock);
1297 file = fcheck_files(files, fd);
1298 if (!file)
1299 goto out_unlock;
1300 if (file->f_mode & 1)
1301 inode->i_mode |= S_IRUSR | S_IXUSR;
1302 if (file->f_mode & 2)
1303 inode->i_mode |= S_IWUSR | S_IXUSR;
1304 spin_unlock(&files->file_lock);
1305 put_files_struct(files);
1306
1307 inode->i_op = &proc_pid_link_inode_operations;
1308 inode->i_size = 64;
1309 ei->op.proc_get_link = proc_fd_link;
1310 dentry->d_op = &tid_fd_dentry_operations;
1311 d_add(dentry, inode);
1312 /* Close the race of the process dying before we return the dentry */
1313 if (tid_fd_revalidate(dentry, NULL))
1314 error = NULL;
1315
1316 out:
1317 return error;
1318 out_unlock:
1319 spin_unlock(&files->file_lock);
1320 put_files_struct(files);
1321 out_iput:
1322 iput(inode);
1323 goto out;
1324 }
1325
1326 static struct dentry *proc_lookupfd(struct inode * dir, struct dentry * dentry, struct nameidata *nd)
1327 {
1328 struct task_struct *task = get_proc_task(dir);
1329 unsigned fd = name_to_int(dentry);
1330 struct dentry *result = ERR_PTR(-ENOENT);
1331
1332 if (!task)
1333 goto out_no_task;
1334 if (fd == ~0U)
1335 goto out;
1336
1337 result = proc_fd_instantiate(dir, dentry, task, &fd);
1338 out:
1339 put_task_struct(task);
1340 out_no_task:
1341 return result;
1342 }
1343
1344 static int proc_fd_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1345 struct task_struct *task, int fd)
1346 {
1347 char name[PROC_NUMBUF];
1348 int len = snprintf(name, sizeof(name), "%d", fd);
1349 return proc_fill_cache(filp, dirent, filldir, name, len,
1350 proc_fd_instantiate, task, &fd);
1351 }
1352
1353 static int proc_readfd(struct file * filp, void * dirent, filldir_t filldir)
1354 {
1355 struct dentry *dentry = filp->f_path.dentry;
1356 struct inode *inode = dentry->d_inode;
1357 struct task_struct *p = get_proc_task(inode);
1358 unsigned int fd, tid, ino;
1359 int retval;
1360 struct files_struct * files;
1361 struct fdtable *fdt;
1362
1363 retval = -ENOENT;
1364 if (!p)
1365 goto out_no_task;
1366 retval = 0;
1367 tid = p->pid;
1368
1369 fd = filp->f_pos;
1370 switch (fd) {
1371 case 0:
1372 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1373 goto out;
1374 filp->f_pos++;
1375 case 1:
1376 ino = parent_ino(dentry);
1377 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1378 goto out;
1379 filp->f_pos++;
1380 default:
1381 files = get_files_struct(p);
1382 if (!files)
1383 goto out;
1384 rcu_read_lock();
1385 fdt = files_fdtable(files);
1386 for (fd = filp->f_pos-2;
1387 fd < fdt->max_fds;
1388 fd++, filp->f_pos++) {
1389
1390 if (!fcheck_files(files, fd))
1391 continue;
1392 rcu_read_unlock();
1393
1394 if (proc_fd_fill_cache(filp, dirent, filldir, p, fd) < 0) {
1395 rcu_read_lock();
1396 break;
1397 }
1398 rcu_read_lock();
1399 }
1400 rcu_read_unlock();
1401 put_files_struct(files);
1402 }
1403 out:
1404 put_task_struct(p);
1405 out_no_task:
1406 return retval;
1407 }
1408
1409 static struct file_operations proc_fd_operations = {
1410 .read = generic_read_dir,
1411 .readdir = proc_readfd,
1412 };
1413
1414 /*
1415 * proc directories can do almost nothing..
1416 */
1417 static struct inode_operations proc_fd_inode_operations = {
1418 .lookup = proc_lookupfd,
1419 .setattr = proc_setattr,
1420 };
1421
1422 static struct dentry *proc_pident_instantiate(struct inode *dir,
1423 struct dentry *dentry, struct task_struct *task, void *ptr)
1424 {
1425 struct pid_entry *p = ptr;
1426 struct inode *inode;
1427 struct proc_inode *ei;
1428 struct dentry *error = ERR_PTR(-EINVAL);
1429
1430 inode = proc_pid_make_inode(dir->i_sb, task);
1431 if (!inode)
1432 goto out;
1433
1434 ei = PROC_I(inode);
1435 inode->i_mode = p->mode;
1436 if (S_ISDIR(inode->i_mode))
1437 inode->i_nlink = 2; /* Use getattr to fix if necessary */
1438 if (p->iop)
1439 inode->i_op = p->iop;
1440 if (p->fop)
1441 inode->i_fop = p->fop;
1442 ei->op = p->op;
1443 dentry->d_op = &pid_dentry_operations;
1444 d_add(dentry, inode);
1445 /* Close the race of the process dying before we return the dentry */
1446 if (pid_revalidate(dentry, NULL))
1447 error = NULL;
1448 out:
1449 return error;
1450 }
1451
1452 static struct dentry *proc_pident_lookup(struct inode *dir,
1453 struct dentry *dentry,
1454 struct pid_entry *ents,
1455 unsigned int nents)
1456 {
1457 struct inode *inode;
1458 struct dentry *error;
1459 struct task_struct *task = get_proc_task(dir);
1460 struct pid_entry *p, *last;
1461
1462 error = ERR_PTR(-ENOENT);
1463 inode = NULL;
1464
1465 if (!task)
1466 goto out_no_task;
1467
1468 /*
1469 * Yes, it does not scale. And it should not. Don't add
1470 * new entries into /proc/<tgid>/ without very good reasons.
1471 */
1472 last = &ents[nents - 1];
1473 for (p = ents; p <= last; p++) {
1474 if (p->len != dentry->d_name.len)
1475 continue;
1476 if (!memcmp(dentry->d_name.name, p->name, p->len))
1477 break;
1478 }
1479 if (p > last)
1480 goto out;
1481
1482 error = proc_pident_instantiate(dir, dentry, task, p);
1483 out:
1484 put_task_struct(task);
1485 out_no_task:
1486 return error;
1487 }
1488
1489 static int proc_pident_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1490 struct task_struct *task, struct pid_entry *p)
1491 {
1492 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
1493 proc_pident_instantiate, task, p);
1494 }
1495
1496 static int proc_pident_readdir(struct file *filp,
1497 void *dirent, filldir_t filldir,
1498 struct pid_entry *ents, unsigned int nents)
1499 {
1500 int i;
1501 int pid;
1502 struct dentry *dentry = filp->f_path.dentry;
1503 struct inode *inode = dentry->d_inode;
1504 struct task_struct *task = get_proc_task(inode);
1505 struct pid_entry *p, *last;
1506 ino_t ino;
1507 int ret;
1508
1509 ret = -ENOENT;
1510 if (!task)
1511 goto out_no_task;
1512
1513 ret = 0;
1514 pid = task->pid;
1515 i = filp->f_pos;
1516 switch (i) {
1517 case 0:
1518 ino = inode->i_ino;
1519 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
1520 goto out;
1521 i++;
1522 filp->f_pos++;
1523 /* fall through */
1524 case 1:
1525 ino = parent_ino(dentry);
1526 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
1527 goto out;
1528 i++;
1529 filp->f_pos++;
1530 /* fall through */
1531 default:
1532 i -= 2;
1533 if (i >= nents) {
1534 ret = 1;
1535 goto out;
1536 }
1537 p = ents + i;
1538 last = &ents[nents - 1];
1539 while (p <= last) {
1540 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
1541 goto out;
1542 filp->f_pos++;
1543 p++;
1544 }
1545 }
1546
1547 ret = 1;
1548 out:
1549 put_task_struct(task);
1550 out_no_task:
1551 return ret;
1552 }
1553
1554 #ifdef CONFIG_SECURITY
1555 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
1556 size_t count, loff_t *ppos)
1557 {
1558 struct inode * inode = file->f_path.dentry->d_inode;
1559 unsigned long page;
1560 ssize_t length;
1561 struct task_struct *task = get_proc_task(inode);
1562
1563 length = -ESRCH;
1564 if (!task)
1565 goto out_no_task;
1566
1567 if (count > PAGE_SIZE)
1568 count = PAGE_SIZE;
1569 length = -ENOMEM;
1570 if (!(page = __get_free_page(GFP_KERNEL)))
1571 goto out;
1572
1573 length = security_getprocattr(task,
1574 (char*)file->f_path.dentry->d_name.name,
1575 (void*)page, count);
1576 if (length >= 0)
1577 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
1578 free_page(page);
1579 out:
1580 put_task_struct(task);
1581 out_no_task:
1582 return length;
1583 }
1584
1585 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
1586 size_t count, loff_t *ppos)
1587 {
1588 struct inode * inode = file->f_path.dentry->d_inode;
1589 char *page;
1590 ssize_t length;
1591 struct task_struct *task = get_proc_task(inode);
1592
1593 length = -ESRCH;
1594 if (!task)
1595 goto out_no_task;
1596 if (count > PAGE_SIZE)
1597 count = PAGE_SIZE;
1598
1599 /* No partial writes. */
1600 length = -EINVAL;
1601 if (*ppos != 0)
1602 goto out;
1603
1604 length = -ENOMEM;
1605 page = (char*)__get_free_page(GFP_USER);
1606 if (!page)
1607 goto out;
1608
1609 length = -EFAULT;
1610 if (copy_from_user(page, buf, count))
1611 goto out_free;
1612
1613 length = security_setprocattr(task,
1614 (char*)file->f_path.dentry->d_name.name,
1615 (void*)page, count);
1616 out_free:
1617 free_page((unsigned long) page);
1618 out:
1619 put_task_struct(task);
1620 out_no_task:
1621 return length;
1622 }
1623
1624 static struct file_operations proc_pid_attr_operations = {
1625 .read = proc_pid_attr_read,
1626 .write = proc_pid_attr_write,
1627 };
1628
1629 static struct pid_entry attr_dir_stuff[] = {
1630 REG("current", S_IRUGO|S_IWUGO, pid_attr),
1631 REG("prev", S_IRUGO, pid_attr),
1632 REG("exec", S_IRUGO|S_IWUGO, pid_attr),
1633 REG("fscreate", S_IRUGO|S_IWUGO, pid_attr),
1634 REG("keycreate", S_IRUGO|S_IWUGO, pid_attr),
1635 REG("sockcreate", S_IRUGO|S_IWUGO, pid_attr),
1636 };
1637
1638 static int proc_attr_dir_readdir(struct file * filp,
1639 void * dirent, filldir_t filldir)
1640 {
1641 return proc_pident_readdir(filp,dirent,filldir,
1642 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
1643 }
1644
1645 static struct file_operations proc_attr_dir_operations = {
1646 .read = generic_read_dir,
1647 .readdir = proc_attr_dir_readdir,
1648 };
1649
1650 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
1651 struct dentry *dentry, struct nameidata *nd)
1652 {
1653 return proc_pident_lookup(dir, dentry,
1654 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
1655 }
1656
1657 static struct inode_operations proc_attr_dir_inode_operations = {
1658 .lookup = proc_attr_dir_lookup,
1659 .getattr = pid_getattr,
1660 .setattr = proc_setattr,
1661 };
1662
1663 #endif
1664
1665 /*
1666 * /proc/self:
1667 */
1668 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
1669 int buflen)
1670 {
1671 char tmp[PROC_NUMBUF];
1672 sprintf(tmp, "%d", current->tgid);
1673 return vfs_readlink(dentry,buffer,buflen,tmp);
1674 }
1675
1676 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
1677 {
1678 char tmp[PROC_NUMBUF];
1679 sprintf(tmp, "%d", current->tgid);
1680 return ERR_PTR(vfs_follow_link(nd,tmp));
1681 }
1682
1683 static struct inode_operations proc_self_inode_operations = {
1684 .readlink = proc_self_readlink,
1685 .follow_link = proc_self_follow_link,
1686 };
1687
1688 /*
1689 * proc base
1690 *
1691 * These are the directory entries in the root directory of /proc
1692 * that properly belong to the /proc filesystem, as they describe
1693 * describe something that is process related.
1694 */
1695 static struct pid_entry proc_base_stuff[] = {
1696 NOD("self", S_IFLNK|S_IRWXUGO,
1697 &proc_self_inode_operations, NULL, {}),
1698 };
1699
1700 /*
1701 * Exceptional case: normally we are not allowed to unhash a busy
1702 * directory. In this case, however, we can do it - no aliasing problems
1703 * due to the way we treat inodes.
1704 */
1705 static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
1706 {
1707 struct inode *inode = dentry->d_inode;
1708 struct task_struct *task = get_proc_task(inode);
1709 if (task) {
1710 put_task_struct(task);
1711 return 1;
1712 }
1713 d_drop(dentry);
1714 return 0;
1715 }
1716
1717 static struct dentry_operations proc_base_dentry_operations =
1718 {
1719 .d_revalidate = proc_base_revalidate,
1720 .d_delete = pid_delete_dentry,
1721 };
1722
1723 static struct dentry *proc_base_instantiate(struct inode *dir,
1724 struct dentry *dentry, struct task_struct *task, void *ptr)
1725 {
1726 struct pid_entry *p = ptr;
1727 struct inode *inode;
1728 struct proc_inode *ei;
1729 struct dentry *error = ERR_PTR(-EINVAL);
1730
1731 /* Allocate the inode */
1732 error = ERR_PTR(-ENOMEM);
1733 inode = new_inode(dir->i_sb);
1734 if (!inode)
1735 goto out;
1736
1737 /* Initialize the inode */
1738 ei = PROC_I(inode);
1739 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1740
1741 /*
1742 * grab the reference to the task.
1743 */
1744 ei->pid = get_task_pid(task, PIDTYPE_PID);
1745 if (!ei->pid)
1746 goto out_iput;
1747
1748 inode->i_uid = 0;
1749 inode->i_gid = 0;
1750 inode->i_mode = p->mode;
1751 if (S_ISDIR(inode->i_mode))
1752 inode->i_nlink = 2;
1753 if (S_ISLNK(inode->i_mode))
1754 inode->i_size = 64;
1755 if (p->iop)
1756 inode->i_op = p->iop;
1757 if (p->fop)
1758 inode->i_fop = p->fop;
1759 ei->op = p->op;
1760 dentry->d_op = &proc_base_dentry_operations;
1761 d_add(dentry, inode);
1762 error = NULL;
1763 out:
1764 return error;
1765 out_iput:
1766 iput(inode);
1767 goto out;
1768 }
1769
1770 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
1771 {
1772 struct dentry *error;
1773 struct task_struct *task = get_proc_task(dir);
1774 struct pid_entry *p, *last;
1775
1776 error = ERR_PTR(-ENOENT);
1777
1778 if (!task)
1779 goto out_no_task;
1780
1781 /* Lookup the directory entry */
1782 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
1783 for (p = proc_base_stuff; p <= last; p++) {
1784 if (p->len != dentry->d_name.len)
1785 continue;
1786 if (!memcmp(dentry->d_name.name, p->name, p->len))
1787 break;
1788 }
1789 if (p > last)
1790 goto out;
1791
1792 error = proc_base_instantiate(dir, dentry, task, p);
1793
1794 out:
1795 put_task_struct(task);
1796 out_no_task:
1797 return error;
1798 }
1799
1800 static int proc_base_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1801 struct task_struct *task, struct pid_entry *p)
1802 {
1803 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
1804 proc_base_instantiate, task, p);
1805 }
1806
1807 #ifdef CONFIG_TASK_IO_ACCOUNTING
1808 static int proc_pid_io_accounting(struct task_struct *task, char *buffer)
1809 {
1810 return sprintf(buffer,
1811 "rchar: %llu\n"
1812 "wchar: %llu\n"
1813 "syscr: %llu\n"
1814 "syscw: %llu\n"
1815 "read_bytes: %llu\n"
1816 "write_bytes: %llu\n"
1817 "cancelled_write_bytes: %llu\n",
1818 (unsigned long long)task->rchar,
1819 (unsigned long long)task->wchar,
1820 (unsigned long long)task->syscr,
1821 (unsigned long long)task->syscw,
1822 (unsigned long long)task->ioac.read_bytes,
1823 (unsigned long long)task->ioac.write_bytes,
1824 (unsigned long long)task->ioac.cancelled_write_bytes);
1825 }
1826 #endif
1827
1828 /*
1829 * Thread groups
1830 */
1831 static struct file_operations proc_task_operations;
1832 static struct inode_operations proc_task_inode_operations;
1833
1834 static struct pid_entry tgid_base_stuff[] = {
1835 DIR("task", S_IRUGO|S_IXUGO, task),
1836 DIR("fd", S_IRUSR|S_IXUSR, fd),
1837 INF("environ", S_IRUSR, pid_environ),
1838 INF("auxv", S_IRUSR, pid_auxv),
1839 INF("status", S_IRUGO, pid_status),
1840 INF("cmdline", S_IRUGO, pid_cmdline),
1841 INF("stat", S_IRUGO, tgid_stat),
1842 INF("statm", S_IRUGO, pid_statm),
1843 REG("maps", S_IRUGO, maps),
1844 #ifdef CONFIG_NUMA
1845 REG("numa_maps", S_IRUGO, numa_maps),
1846 #endif
1847 REG("mem", S_IRUSR|S_IWUSR, mem),
1848 #ifdef CONFIG_SECCOMP
1849 REG("seccomp", S_IRUSR|S_IWUSR, seccomp),
1850 #endif
1851 LNK("cwd", cwd),
1852 LNK("root", root),
1853 LNK("exe", exe),
1854 REG("mounts", S_IRUGO, mounts),
1855 REG("mountstats", S_IRUSR, mountstats),
1856 #ifdef CONFIG_MMU
1857 REG("smaps", S_IRUGO, smaps),
1858 #endif
1859 #ifdef CONFIG_SECURITY
1860 DIR("attr", S_IRUGO|S_IXUGO, attr_dir),
1861 #endif
1862 #ifdef CONFIG_KALLSYMS
1863 INF("wchan", S_IRUGO, pid_wchan),
1864 #endif
1865 #ifdef CONFIG_SCHEDSTATS
1866 INF("schedstat", S_IRUGO, pid_schedstat),
1867 #endif
1868 #ifdef CONFIG_CPUSETS
1869 REG("cpuset", S_IRUGO, cpuset),
1870 #endif
1871 INF("oom_score", S_IRUGO, oom_score),
1872 REG("oom_adj", S_IRUGO|S_IWUSR, oom_adjust),
1873 #ifdef CONFIG_AUDITSYSCALL
1874 REG("loginuid", S_IWUSR|S_IRUGO, loginuid),
1875 #endif
1876 #ifdef CONFIG_FAULT_INJECTION
1877 REG("make-it-fail", S_IRUGO|S_IWUSR, fault_inject),
1878 #endif
1879 #ifdef CONFIG_TASK_IO_ACCOUNTING
1880 INF("io", S_IRUGO, pid_io_accounting),
1881 #endif
1882 };
1883
1884 static int proc_tgid_base_readdir(struct file * filp,
1885 void * dirent, filldir_t filldir)
1886 {
1887 return proc_pident_readdir(filp,dirent,filldir,
1888 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
1889 }
1890
1891 static struct file_operations proc_tgid_base_operations = {
1892 .read = generic_read_dir,
1893 .readdir = proc_tgid_base_readdir,
1894 };
1895
1896 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
1897 return proc_pident_lookup(dir, dentry,
1898 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
1899 }
1900
1901 static struct inode_operations proc_tgid_base_inode_operations = {
1902 .lookup = proc_tgid_base_lookup,
1903 .getattr = pid_getattr,
1904 .setattr = proc_setattr,
1905 };
1906
1907 /**
1908 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
1909 *
1910 * @task: task that should be flushed.
1911 *
1912 * Looks in the dcache for
1913 * /proc/@pid
1914 * /proc/@tgid/task/@pid
1915 * if either directory is present flushes it and all of it'ts children
1916 * from the dcache.
1917 *
1918 * It is safe and reasonable to cache /proc entries for a task until
1919 * that task exits. After that they just clog up the dcache with
1920 * useless entries, possibly causing useful dcache entries to be
1921 * flushed instead. This routine is proved to flush those useless
1922 * dcache entries at process exit time.
1923 *
1924 * NOTE: This routine is just an optimization so it does not guarantee
1925 * that no dcache entries will exist at process exit time it
1926 * just makes it very unlikely that any will persist.
1927 */
1928 void proc_flush_task(struct task_struct *task)
1929 {
1930 struct dentry *dentry, *leader, *dir;
1931 char buf[PROC_NUMBUF];
1932 struct qstr name;
1933
1934 name.name = buf;
1935 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
1936 dentry = d_hash_and_lookup(proc_mnt->mnt_root, &name);
1937 if (dentry) {
1938 shrink_dcache_parent(dentry);
1939 d_drop(dentry);
1940 dput(dentry);
1941 }
1942
1943 if (thread_group_leader(task))
1944 goto out;
1945
1946 name.name = buf;
1947 name.len = snprintf(buf, sizeof(buf), "%d", task->tgid);
1948 leader = d_hash_and_lookup(proc_mnt->mnt_root, &name);
1949 if (!leader)
1950 goto out;
1951
1952 name.name = "task";
1953 name.len = strlen(name.name);
1954 dir = d_hash_and_lookup(leader, &name);
1955 if (!dir)
1956 goto out_put_leader;
1957
1958 name.name = buf;
1959 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
1960 dentry = d_hash_and_lookup(dir, &name);
1961 if (dentry) {
1962 shrink_dcache_parent(dentry);
1963 d_drop(dentry);
1964 dput(dentry);
1965 }
1966
1967 dput(dir);
1968 out_put_leader:
1969 dput(leader);
1970 out:
1971 return;
1972 }
1973
1974 static struct dentry *proc_pid_instantiate(struct inode *dir,
1975 struct dentry * dentry,
1976 struct task_struct *task, void *ptr)
1977 {
1978 struct dentry *error = ERR_PTR(-ENOENT);
1979 struct inode *inode;
1980
1981 inode = proc_pid_make_inode(dir->i_sb, task);
1982 if (!inode)
1983 goto out;
1984
1985 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
1986 inode->i_op = &proc_tgid_base_inode_operations;
1987 inode->i_fop = &proc_tgid_base_operations;
1988 inode->i_flags|=S_IMMUTABLE;
1989 inode->i_nlink = 4;
1990 #ifdef CONFIG_SECURITY
1991 inode->i_nlink += 1;
1992 #endif
1993
1994 dentry->d_op = &pid_dentry_operations;
1995
1996 d_add(dentry, inode);
1997 /* Close the race of the process dying before we return the dentry */
1998 if (pid_revalidate(dentry, NULL))
1999 error = NULL;
2000 out:
2001 return error;
2002 }
2003
2004 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2005 {
2006 struct dentry *result = ERR_PTR(-ENOENT);
2007 struct task_struct *task;
2008 unsigned tgid;
2009
2010 result = proc_base_lookup(dir, dentry);
2011 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2012 goto out;
2013
2014 tgid = name_to_int(dentry);
2015 if (tgid == ~0U)
2016 goto out;
2017
2018 rcu_read_lock();
2019 task = find_task_by_pid(tgid);
2020 if (task)
2021 get_task_struct(task);
2022 rcu_read_unlock();
2023 if (!task)
2024 goto out;
2025
2026 result = proc_pid_instantiate(dir, dentry, task, NULL);
2027 put_task_struct(task);
2028 out:
2029 return result;
2030 }
2031
2032 /*
2033 * Find the first task with tgid >= tgid
2034 *
2035 */
2036 static struct task_struct *next_tgid(unsigned int tgid)
2037 {
2038 struct task_struct *task;
2039 struct pid *pid;
2040
2041 rcu_read_lock();
2042 retry:
2043 task = NULL;
2044 pid = find_ge_pid(tgid);
2045 if (pid) {
2046 tgid = pid->nr + 1;
2047 task = pid_task(pid, PIDTYPE_PID);
2048 /* What we to know is if the pid we have find is the
2049 * pid of a thread_group_leader. Testing for task
2050 * being a thread_group_leader is the obvious thing
2051 * todo but there is a window when it fails, due to
2052 * the pid transfer logic in de_thread.
2053 *
2054 * So we perform the straight forward test of seeing
2055 * if the pid we have found is the pid of a thread
2056 * group leader, and don't worry if the task we have
2057 * found doesn't happen to be a thread group leader.
2058 * As we don't care in the case of readdir.
2059 */
2060 if (!task || !has_group_leader_pid(task))
2061 goto retry;
2062 get_task_struct(task);
2063 }
2064 rcu_read_unlock();
2065 return task;
2066 }
2067
2068 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
2069
2070 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2071 struct task_struct *task, int tgid)
2072 {
2073 char name[PROC_NUMBUF];
2074 int len = snprintf(name, sizeof(name), "%d", tgid);
2075 return proc_fill_cache(filp, dirent, filldir, name, len,
2076 proc_pid_instantiate, task, NULL);
2077 }
2078
2079 /* for the /proc/ directory itself, after non-process stuff has been done */
2080 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2081 {
2082 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2083 struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
2084 struct task_struct *task;
2085 int tgid;
2086
2087 if (!reaper)
2088 goto out_no_task;
2089
2090 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
2091 struct pid_entry *p = &proc_base_stuff[nr];
2092 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
2093 goto out;
2094 }
2095
2096 tgid = filp->f_pos - TGID_OFFSET;
2097 for (task = next_tgid(tgid);
2098 task;
2099 put_task_struct(task), task = next_tgid(tgid + 1)) {
2100 tgid = task->pid;
2101 filp->f_pos = tgid + TGID_OFFSET;
2102 if (proc_pid_fill_cache(filp, dirent, filldir, task, tgid) < 0) {
2103 put_task_struct(task);
2104 goto out;
2105 }
2106 }
2107 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
2108 out:
2109 put_task_struct(reaper);
2110 out_no_task:
2111 return 0;
2112 }
2113
2114 /*
2115 * Tasks
2116 */
2117 static struct pid_entry tid_base_stuff[] = {
2118 DIR("fd", S_IRUSR|S_IXUSR, fd),
2119 INF("environ", S_IRUSR, pid_environ),
2120 INF("auxv", S_IRUSR, pid_auxv),
2121 INF("status", S_IRUGO, pid_status),
2122 INF("cmdline", S_IRUGO, pid_cmdline),
2123 INF("stat", S_IRUGO, tid_stat),
2124 INF("statm", S_IRUGO, pid_statm),
2125 REG("maps", S_IRUGO, maps),
2126 #ifdef CONFIG_NUMA
2127 REG("numa_maps", S_IRUGO, numa_maps),
2128 #endif
2129 REG("mem", S_IRUSR|S_IWUSR, mem),
2130 #ifdef CONFIG_SECCOMP
2131 REG("seccomp", S_IRUSR|S_IWUSR, seccomp),
2132 #endif
2133 LNK("cwd", cwd),
2134 LNK("root", root),
2135 LNK("exe", exe),
2136 REG("mounts", S_IRUGO, mounts),
2137 #ifdef CONFIG_MMU
2138 REG("smaps", S_IRUGO, smaps),
2139 #endif
2140 #ifdef CONFIG_SECURITY
2141 DIR("attr", S_IRUGO|S_IXUGO, attr_dir),
2142 #endif
2143 #ifdef CONFIG_KALLSYMS
2144 INF("wchan", S_IRUGO, pid_wchan),
2145 #endif
2146 #ifdef CONFIG_SCHEDSTATS
2147 INF("schedstat", S_IRUGO, pid_schedstat),
2148 #endif
2149 #ifdef CONFIG_CPUSETS
2150 REG("cpuset", S_IRUGO, cpuset),
2151 #endif
2152 INF("oom_score", S_IRUGO, oom_score),
2153 REG("oom_adj", S_IRUGO|S_IWUSR, oom_adjust),
2154 #ifdef CONFIG_AUDITSYSCALL
2155 REG("loginuid", S_IWUSR|S_IRUGO, loginuid),
2156 #endif
2157 #ifdef CONFIG_FAULT_INJECTION
2158 REG("make-it-fail", S_IRUGO|S_IWUSR, fault_inject),
2159 #endif
2160 };
2161
2162 static int proc_tid_base_readdir(struct file * filp,
2163 void * dirent, filldir_t filldir)
2164 {
2165 return proc_pident_readdir(filp,dirent,filldir,
2166 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
2167 }
2168
2169 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2170 return proc_pident_lookup(dir, dentry,
2171 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
2172 }
2173
2174 static struct file_operations proc_tid_base_operations = {
2175 .read = generic_read_dir,
2176 .readdir = proc_tid_base_readdir,
2177 };
2178
2179 static struct inode_operations proc_tid_base_inode_operations = {
2180 .lookup = proc_tid_base_lookup,
2181 .getattr = pid_getattr,
2182 .setattr = proc_setattr,
2183 };
2184
2185 static struct dentry *proc_task_instantiate(struct inode *dir,
2186 struct dentry *dentry, struct task_struct *task, void *ptr)
2187 {
2188 struct dentry *error = ERR_PTR(-ENOENT);
2189 struct inode *inode;
2190 inode = proc_pid_make_inode(dir->i_sb, task);
2191
2192 if (!inode)
2193 goto out;
2194 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2195 inode->i_op = &proc_tid_base_inode_operations;
2196 inode->i_fop = &proc_tid_base_operations;
2197 inode->i_flags|=S_IMMUTABLE;
2198 inode->i_nlink = 3;
2199 #ifdef CONFIG_SECURITY
2200 inode->i_nlink += 1;
2201 #endif
2202
2203 dentry->d_op = &pid_dentry_operations;
2204
2205 d_add(dentry, inode);
2206 /* Close the race of the process dying before we return the dentry */
2207 if (pid_revalidate(dentry, NULL))
2208 error = NULL;
2209 out:
2210 return error;
2211 }
2212
2213 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2214 {
2215 struct dentry *result = ERR_PTR(-ENOENT);
2216 struct task_struct *task;
2217 struct task_struct *leader = get_proc_task(dir);
2218 unsigned tid;
2219
2220 if (!leader)
2221 goto out_no_task;
2222
2223 tid = name_to_int(dentry);
2224 if (tid == ~0U)
2225 goto out;
2226
2227 rcu_read_lock();
2228 task = find_task_by_pid(tid);
2229 if (task)
2230 get_task_struct(task);
2231 rcu_read_unlock();
2232 if (!task)
2233 goto out;
2234 if (leader->tgid != task->tgid)
2235 goto out_drop_task;
2236
2237 result = proc_task_instantiate(dir, dentry, task, NULL);
2238 out_drop_task:
2239 put_task_struct(task);
2240 out:
2241 put_task_struct(leader);
2242 out_no_task:
2243 return result;
2244 }
2245
2246 /*
2247 * Find the first tid of a thread group to return to user space.
2248 *
2249 * Usually this is just the thread group leader, but if the users
2250 * buffer was too small or there was a seek into the middle of the
2251 * directory we have more work todo.
2252 *
2253 * In the case of a short read we start with find_task_by_pid.
2254 *
2255 * In the case of a seek we start with the leader and walk nr
2256 * threads past it.
2257 */
2258 static struct task_struct *first_tid(struct task_struct *leader,
2259 int tid, int nr)
2260 {
2261 struct task_struct *pos;
2262
2263 rcu_read_lock();
2264 /* Attempt to start with the pid of a thread */
2265 if (tid && (nr > 0)) {
2266 pos = find_task_by_pid(tid);
2267 if (pos && (pos->group_leader == leader))
2268 goto found;
2269 }
2270
2271 /* If nr exceeds the number of threads there is nothing todo */
2272 pos = NULL;
2273 if (nr && nr >= get_nr_threads(leader))
2274 goto out;
2275
2276 /* If we haven't found our starting place yet start
2277 * with the leader and walk nr threads forward.
2278 */
2279 for (pos = leader; nr > 0; --nr) {
2280 pos = next_thread(pos);
2281 if (pos == leader) {
2282 pos = NULL;
2283 goto out;
2284 }
2285 }
2286 found:
2287 get_task_struct(pos);
2288 out:
2289 rcu_read_unlock();
2290 return pos;
2291 }
2292
2293 /*
2294 * Find the next thread in the thread list.
2295 * Return NULL if there is an error or no next thread.
2296 *
2297 * The reference to the input task_struct is released.
2298 */
2299 static struct task_struct *next_tid(struct task_struct *start)
2300 {
2301 struct task_struct *pos = NULL;
2302 rcu_read_lock();
2303 if (pid_alive(start)) {
2304 pos = next_thread(start);
2305 if (thread_group_leader(pos))
2306 pos = NULL;
2307 else
2308 get_task_struct(pos);
2309 }
2310 rcu_read_unlock();
2311 put_task_struct(start);
2312 return pos;
2313 }
2314
2315 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2316 struct task_struct *task, int tid)
2317 {
2318 char name[PROC_NUMBUF];
2319 int len = snprintf(name, sizeof(name), "%d", tid);
2320 return proc_fill_cache(filp, dirent, filldir, name, len,
2321 proc_task_instantiate, task, NULL);
2322 }
2323
2324 /* for the /proc/TGID/task/ directories */
2325 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
2326 {
2327 struct dentry *dentry = filp->f_path.dentry;
2328 struct inode *inode = dentry->d_inode;
2329 struct task_struct *leader = get_proc_task(inode);
2330 struct task_struct *task;
2331 int retval = -ENOENT;
2332 ino_t ino;
2333 int tid;
2334 unsigned long pos = filp->f_pos; /* avoiding "long long" filp->f_pos */
2335
2336 if (!leader)
2337 goto out_no_task;
2338 retval = 0;
2339
2340 switch (pos) {
2341 case 0:
2342 ino = inode->i_ino;
2343 if (filldir(dirent, ".", 1, pos, ino, DT_DIR) < 0)
2344 goto out;
2345 pos++;
2346 /* fall through */
2347 case 1:
2348 ino = parent_ino(dentry);
2349 if (filldir(dirent, "..", 2, pos, ino, DT_DIR) < 0)
2350 goto out;
2351 pos++;
2352 /* fall through */
2353 }
2354
2355 /* f_version caches the tgid value that the last readdir call couldn't
2356 * return. lseek aka telldir automagically resets f_version to 0.
2357 */
2358 tid = filp->f_version;
2359 filp->f_version = 0;
2360 for (task = first_tid(leader, tid, pos - 2);
2361 task;
2362 task = next_tid(task), pos++) {
2363 tid = task->pid;
2364 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
2365 /* returning this tgid failed, save it as the first
2366 * pid for the next readir call */
2367 filp->f_version = tid;
2368 put_task_struct(task);
2369 break;
2370 }
2371 }
2372 out:
2373 filp->f_pos = pos;
2374 put_task_struct(leader);
2375 out_no_task:
2376 return retval;
2377 }
2378
2379 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
2380 {
2381 struct inode *inode = dentry->d_inode;
2382 struct task_struct *p = get_proc_task(inode);
2383 generic_fillattr(inode, stat);
2384
2385 if (p) {
2386 rcu_read_lock();
2387 stat->nlink += get_nr_threads(p);
2388 rcu_read_unlock();
2389 put_task_struct(p);
2390 }
2391
2392 return 0;
2393 }
2394
2395 static struct inode_operations proc_task_inode_operations = {
2396 .lookup = proc_task_lookup,
2397 .getattr = proc_task_getattr,
2398 .setattr = proc_setattr,
2399 };
2400
2401 static struct file_operations proc_task_operations = {
2402 .read = generic_read_dir,
2403 .readdir = proc_task_readdir,
2404 };
Mirror for ia64_pv_ops.git