]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blob - fs/proc/base.c
projects / mirror_ubuntu-zesty-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
[PATCH] namespaces: incorporate fs namespace into nsproxy
[mirror_ubuntu-zesty-kernel.git] / fs / proc / base.c
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/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 "internal.h"
76
77 /* NOTE:
78 * Implementing inode permission operations in /proc is almost
79 * certainly an error. Permission checks need to happen during
80 * each system call not at open time. The reason is that most of
81 * what we wish to check for permissions in /proc varies at runtime.
82 *
83 * The classic example of a problem is opening file descriptors
84 * in /proc for a task before it execs a suid executable.
85 */
86
87 /*
88 * For hysterical raisins we keep the same inumbers as in the old procfs.
89 * Feel free to change the macro below - just keep the range distinct from
90 * inumbers of the rest of procfs (currently those are in 0x0000--0xffff).
91 * As soon as we'll get a separate superblock we will be able to forget
92 * about magical ranges too.
93 */
94
95 #define fake_ino(pid,ino) (((pid)<<16)|(ino))
96
97 enum pid_directory_inos {
98 PROC_TGID_INO = 2,
99 PROC_TGID_TASK,
100 PROC_TGID_STATUS,
101 PROC_TGID_MEM,
102 #ifdef CONFIG_SECCOMP
103 PROC_TGID_SECCOMP,
104 #endif
105 PROC_TGID_CWD,
106 PROC_TGID_ROOT,
107 PROC_TGID_EXE,
108 PROC_TGID_FD,
109 PROC_TGID_ENVIRON,
110 PROC_TGID_AUXV,
111 PROC_TGID_CMDLINE,
112 PROC_TGID_STAT,
113 PROC_TGID_STATM,
114 PROC_TGID_MAPS,
115 PROC_TGID_NUMA_MAPS,
116 PROC_TGID_MOUNTS,
117 PROC_TGID_MOUNTSTATS,
118 PROC_TGID_WCHAN,
119 #ifdef CONFIG_MMU
120 PROC_TGID_SMAPS,
121 #endif
122 #ifdef CONFIG_SCHEDSTATS
123 PROC_TGID_SCHEDSTAT,
124 #endif
125 #ifdef CONFIG_CPUSETS
126 PROC_TGID_CPUSET,
127 #endif
128 #ifdef CONFIG_SECURITY
129 PROC_TGID_ATTR,
130 PROC_TGID_ATTR_CURRENT,
131 PROC_TGID_ATTR_PREV,
132 PROC_TGID_ATTR_EXEC,
133 PROC_TGID_ATTR_FSCREATE,
134 PROC_TGID_ATTR_KEYCREATE,
135 PROC_TGID_ATTR_SOCKCREATE,
136 #endif
137 #ifdef CONFIG_AUDITSYSCALL
138 PROC_TGID_LOGINUID,
139 #endif
140 PROC_TGID_OOM_SCORE,
141 PROC_TGID_OOM_ADJUST,
142 PROC_TID_INO,
143 PROC_TID_STATUS,
144 PROC_TID_MEM,
145 #ifdef CONFIG_SECCOMP
146 PROC_TID_SECCOMP,
147 #endif
148 PROC_TID_CWD,
149 PROC_TID_ROOT,
150 PROC_TID_EXE,
151 PROC_TID_FD,
152 PROC_TID_ENVIRON,
153 PROC_TID_AUXV,
154 PROC_TID_CMDLINE,
155 PROC_TID_STAT,
156 PROC_TID_STATM,
157 PROC_TID_MAPS,
158 PROC_TID_NUMA_MAPS,
159 PROC_TID_MOUNTS,
160 PROC_TID_MOUNTSTATS,
161 PROC_TID_WCHAN,
162 #ifdef CONFIG_MMU
163 PROC_TID_SMAPS,
164 #endif
165 #ifdef CONFIG_SCHEDSTATS
166 PROC_TID_SCHEDSTAT,
167 #endif
168 #ifdef CONFIG_CPUSETS
169 PROC_TID_CPUSET,
170 #endif
171 #ifdef CONFIG_SECURITY
172 PROC_TID_ATTR,
173 PROC_TID_ATTR_CURRENT,
174 PROC_TID_ATTR_PREV,
175 PROC_TID_ATTR_EXEC,
176 PROC_TID_ATTR_FSCREATE,
177 PROC_TID_ATTR_KEYCREATE,
178 PROC_TID_ATTR_SOCKCREATE,
179 #endif
180 #ifdef CONFIG_AUDITSYSCALL
181 PROC_TID_LOGINUID,
182 #endif
183 PROC_TID_OOM_SCORE,
184 PROC_TID_OOM_ADJUST,
185
186 /* Add new entries before this */
187 PROC_TID_FD_DIR = 0x8000, /* 0x8000-0xffff */
188 };
189
190 /* Worst case buffer size needed for holding an integer. */
191 #define PROC_NUMBUF 10
192
193 struct pid_entry {
194 int type;
195 int len;
196 char *name;
197 mode_t mode;
198 struct inode_operations *iop;
199 struct file_operations *fop;
200 union proc_op op;
201 };
202
203 #define NOD(TYPE, NAME, MODE, IOP, FOP, OP) { \
204 .type = (TYPE), \
205 .len = sizeof(NAME) - 1, \
206 .name = (NAME), \
207 .mode = MODE, \
208 .iop = IOP, \
209 .fop = FOP, \
210 .op = OP, \
211 }
212
213 #define DIR(TYPE, NAME, MODE, OTYPE) \
214 NOD(TYPE, NAME, (S_IFDIR|(MODE)), \
215 &proc_##OTYPE##_inode_operations, &proc_##OTYPE##_operations, \
216 {} )
217 #define LNK(TYPE, NAME, OTYPE) \
218 NOD(TYPE, NAME, (S_IFLNK|S_IRWXUGO), \
219 &proc_pid_link_inode_operations, NULL, \
220 { .proc_get_link = &proc_##OTYPE##_link } )
221 #define REG(TYPE, NAME, MODE, OTYPE) \
222 NOD(TYPE, NAME, (S_IFREG|(MODE)), NULL, \
223 &proc_##OTYPE##_operations, {})
224 #define INF(TYPE, NAME, MODE, OTYPE) \
225 NOD(TYPE, NAME, (S_IFREG|(MODE)), \
226 NULL, &proc_info_file_operations, \
227 { .proc_read = &proc_##OTYPE } )
228
229 static struct fs_struct *get_fs_struct(struct task_struct *task)
230 {
231 struct fs_struct *fs;
232 task_lock(task);
233 fs = task->fs;
234 if(fs)
235 atomic_inc(&fs->count);
236 task_unlock(task);
237 return fs;
238 }
239
240 static int get_nr_threads(struct task_struct *tsk)
241 {
242 /* Must be called with the rcu_read_lock held */
243 unsigned long flags;
244 int count = 0;
245
246 if (lock_task_sighand(tsk, &flags)) {
247 count = atomic_read(&tsk->signal->count);
248 unlock_task_sighand(tsk, &flags);
249 }
250 return count;
251 }
252
253 static int proc_cwd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
254 {
255 struct task_struct *task = get_proc_task(inode);
256 struct fs_struct *fs = NULL;
257 int result = -ENOENT;
258
259 if (task) {
260 fs = get_fs_struct(task);
261 put_task_struct(task);
262 }
263 if (fs) {
264 read_lock(&fs->lock);
265 *mnt = mntget(fs->pwdmnt);
266 *dentry = dget(fs->pwd);
267 read_unlock(&fs->lock);
268 result = 0;
269 put_fs_struct(fs);
270 }
271 return result;
272 }
273
274 static int proc_root_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
275 {
276 struct task_struct *task = get_proc_task(inode);
277 struct fs_struct *fs = NULL;
278 int result = -ENOENT;
279
280 if (task) {
281 fs = get_fs_struct(task);
282 put_task_struct(task);
283 }
284 if (fs) {
285 read_lock(&fs->lock);
286 *mnt = mntget(fs->rootmnt);
287 *dentry = dget(fs->root);
288 read_unlock(&fs->lock);
289 result = 0;
290 put_fs_struct(fs);
291 }
292 return result;
293 }
294
295 #define MAY_PTRACE(task) \
296 (task == current || \
297 (task->parent == current && \
298 (task->ptrace & PT_PTRACED) && \
299 (task->state == TASK_STOPPED || task->state == TASK_TRACED) && \
300 security_ptrace(current,task) == 0))
301
302 static int proc_pid_environ(struct task_struct *task, char * buffer)
303 {
304 int res = 0;
305 struct mm_struct *mm = get_task_mm(task);
306 if (mm) {
307 unsigned int len = mm->env_end - mm->env_start;
308 if (len > PAGE_SIZE)
309 len = PAGE_SIZE;
310 res = access_process_vm(task, mm->env_start, buffer, len, 0);
311 if (!ptrace_may_attach(task))
312 res = -ESRCH;
313 mmput(mm);
314 }
315 return res;
316 }
317
318 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
319 {
320 int res = 0;
321 unsigned int len;
322 struct mm_struct *mm = get_task_mm(task);
323 if (!mm)
324 goto out;
325 if (!mm->arg_end)
326 goto out_mm; /* Shh! No looking before we're done */
327
328 len = mm->arg_end - mm->arg_start;
329
330 if (len > PAGE_SIZE)
331 len = PAGE_SIZE;
332
333 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
334
335 // If the nul at the end of args has been overwritten, then
336 // assume application is using setproctitle(3).
337 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
338 len = strnlen(buffer, res);
339 if (len < res) {
340 res = len;
341 } else {
342 len = mm->env_end - mm->env_start;
343 if (len > PAGE_SIZE - res)
344 len = PAGE_SIZE - res;
345 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
346 res = strnlen(buffer, res);
347 }
348 }
349 out_mm:
350 mmput(mm);
351 out:
352 return res;
353 }
354
355 static int proc_pid_auxv(struct task_struct *task, char *buffer)
356 {
357 int res = 0;
358 struct mm_struct *mm = get_task_mm(task);
359 if (mm) {
360 unsigned int nwords = 0;
361 do
362 nwords += 2;
363 while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
364 res = nwords * sizeof(mm->saved_auxv[0]);
365 if (res > PAGE_SIZE)
366 res = PAGE_SIZE;
367 memcpy(buffer, mm->saved_auxv, res);
368 mmput(mm);
369 }
370 return res;
371 }
372
373
374 #ifdef CONFIG_KALLSYMS
375 /*
376 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
377 * Returns the resolved symbol. If that fails, simply return the address.
378 */
379 static int proc_pid_wchan(struct task_struct *task, char *buffer)
380 {
381 char *modname;
382 const char *sym_name;
383 unsigned long wchan, size, offset;
384 char namebuf[KSYM_NAME_LEN+1];
385
386 wchan = get_wchan(task);
387
388 sym_name = kallsyms_lookup(wchan, &size, &offset, &modname, namebuf);
389 if (sym_name)
390 return sprintf(buffer, "%s", sym_name);
391 return sprintf(buffer, "%lu", wchan);
392 }
393 #endif /* CONFIG_KALLSYMS */
394
395 #ifdef CONFIG_SCHEDSTATS
396 /*
397 * Provides /proc/PID/schedstat
398 */
399 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
400 {
401 return sprintf(buffer, "%lu %lu %lu\n",
402 task->sched_info.cpu_time,
403 task->sched_info.run_delay,
404 task->sched_info.pcnt);
405 }
406 #endif
407
408 /* The badness from the OOM killer */
409 unsigned long badness(struct task_struct *p, unsigned long uptime);
410 static int proc_oom_score(struct task_struct *task, char *buffer)
411 {
412 unsigned long points;
413 struct timespec uptime;
414
415 do_posix_clock_monotonic_gettime(&uptime);
416 points = badness(task, uptime.tv_sec);
417 return sprintf(buffer, "%lu\n", points);
418 }
419
420 /************************************************************************/
421 /* Here the fs part begins */
422 /************************************************************************/
423
424 /* permission checks */
425 static int proc_fd_access_allowed(struct inode *inode)
426 {
427 struct task_struct *task;
428 int allowed = 0;
429 /* Allow access to a task's file descriptors if it is us or we
430 * may use ptrace attach to the process and find out that
431 * information.
432 */
433 task = get_proc_task(inode);
434 if (task) {
435 allowed = ptrace_may_attach(task);
436 put_task_struct(task);
437 }
438 return allowed;
439 }
440
441 static int proc_setattr(struct dentry *dentry, struct iattr *attr)
442 {
443 int error;
444 struct inode *inode = dentry->d_inode;
445
446 if (attr->ia_valid & ATTR_MODE)
447 return -EPERM;
448
449 error = inode_change_ok(inode, attr);
450 if (!error) {
451 error = security_inode_setattr(dentry, attr);
452 if (!error)
453 error = inode_setattr(inode, attr);
454 }
455 return error;
456 }
457
458 static struct inode_operations proc_def_inode_operations = {
459 .setattr = proc_setattr,
460 };
461
462 extern struct seq_operations mounts_op;
463 struct proc_mounts {
464 struct seq_file m;
465 int event;
466 };
467
468 static int mounts_open(struct inode *inode, struct file *file)
469 {
470 struct task_struct *task = get_proc_task(inode);
471 struct namespace *namespace = NULL;
472 struct proc_mounts *p;
473 int ret = -EINVAL;
474
475 if (task) {
476 task_lock(task);
477 namespace = task->nsproxy->namespace;
478 if (namespace)
479 get_namespace(namespace);
480 task_unlock(task);
481 put_task_struct(task);
482 }
483
484 if (namespace) {
485 ret = -ENOMEM;
486 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
487 if (p) {
488 file->private_data = &p->m;
489 ret = seq_open(file, &mounts_op);
490 if (!ret) {
491 p->m.private = namespace;
492 p->event = namespace->event;
493 return 0;
494 }
495 kfree(p);
496 }
497 put_namespace(namespace);
498 }
499 return ret;
500 }
501
502 static int mounts_release(struct inode *inode, struct file *file)
503 {
504 struct seq_file *m = file->private_data;
505 struct namespace *namespace = m->private;
506 put_namespace(namespace);
507 return seq_release(inode, file);
508 }
509
510 static unsigned mounts_poll(struct file *file, poll_table *wait)
511 {
512 struct proc_mounts *p = file->private_data;
513 struct namespace *ns = p->m.private;
514 unsigned res = 0;
515
516 poll_wait(file, &ns->poll, wait);
517
518 spin_lock(&vfsmount_lock);
519 if (p->event != ns->event) {
520 p->event = ns->event;
521 res = POLLERR;
522 }
523 spin_unlock(&vfsmount_lock);
524
525 return res;
526 }
527
528 static struct file_operations proc_mounts_operations = {
529 .open = mounts_open,
530 .read = seq_read,
531 .llseek = seq_lseek,
532 .release = mounts_release,
533 .poll = mounts_poll,
534 };
535
536 extern struct seq_operations mountstats_op;
537 static int mountstats_open(struct inode *inode, struct file *file)
538 {
539 int ret = seq_open(file, &mountstats_op);
540
541 if (!ret) {
542 struct seq_file *m = file->private_data;
543 struct namespace *namespace = NULL;
544 struct task_struct *task = get_proc_task(inode);
545
546 if (task) {
547 task_lock(task);
548 namespace = task->nsproxy->namespace;
549 if (namespace)
550 get_namespace(namespace);
551 task_unlock(task);
552 put_task_struct(task);
553 }
554
555 if (namespace)
556 m->private = namespace;
557 else {
558 seq_release(inode, file);
559 ret = -EINVAL;
560 }
561 }
562 return ret;
563 }
564
565 static struct file_operations proc_mountstats_operations = {
566 .open = mountstats_open,
567 .read = seq_read,
568 .llseek = seq_lseek,
569 .release = mounts_release,
570 };
571
572 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
573
574 static ssize_t proc_info_read(struct file * file, char __user * buf,
575 size_t count, loff_t *ppos)
576 {
577 struct inode * inode = file->f_dentry->d_inode;
578 unsigned long page;
579 ssize_t length;
580 struct task_struct *task = get_proc_task(inode);
581
582 length = -ESRCH;
583 if (!task)
584 goto out_no_task;
585
586 if (count > PROC_BLOCK_SIZE)
587 count = PROC_BLOCK_SIZE;
588
589 length = -ENOMEM;
590 if (!(page = __get_free_page(GFP_KERNEL)))
591 goto out;
592
593 length = PROC_I(inode)->op.proc_read(task, (char*)page);
594
595 if (length >= 0)
596 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
597 free_page(page);
598 out:
599 put_task_struct(task);
600 out_no_task:
601 return length;
602 }
603
604 static struct file_operations proc_info_file_operations = {
605 .read = proc_info_read,
606 };
607
608 static int mem_open(struct inode* inode, struct file* file)
609 {
610 file->private_data = (void*)((long)current->self_exec_id);
611 return 0;
612 }
613
614 static ssize_t mem_read(struct file * file, char __user * buf,
615 size_t count, loff_t *ppos)
616 {
617 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
618 char *page;
619 unsigned long src = *ppos;
620 int ret = -ESRCH;
621 struct mm_struct *mm;
622
623 if (!task)
624 goto out_no_task;
625
626 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
627 goto out;
628
629 ret = -ENOMEM;
630 page = (char *)__get_free_page(GFP_USER);
631 if (!page)
632 goto out;
633
634 ret = 0;
635
636 mm = get_task_mm(task);
637 if (!mm)
638 goto out_free;
639
640 ret = -EIO;
641
642 if (file->private_data != (void*)((long)current->self_exec_id))
643 goto out_put;
644
645 ret = 0;
646
647 while (count > 0) {
648 int this_len, retval;
649
650 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
651 retval = access_process_vm(task, src, page, this_len, 0);
652 if (!retval || !MAY_PTRACE(task) || !ptrace_may_attach(task)) {
653 if (!ret)
654 ret = -EIO;
655 break;
656 }
657
658 if (copy_to_user(buf, page, retval)) {
659 ret = -EFAULT;
660 break;
661 }
662
663 ret += retval;
664 src += retval;
665 buf += retval;
666 count -= retval;
667 }
668 *ppos = src;
669
670 out_put:
671 mmput(mm);
672 out_free:
673 free_page((unsigned long) page);
674 out:
675 put_task_struct(task);
676 out_no_task:
677 return ret;
678 }
679
680 #define mem_write NULL
681
682 #ifndef mem_write
683 /* This is a security hazard */
684 static ssize_t mem_write(struct file * file, const char * buf,
685 size_t count, loff_t *ppos)
686 {
687 int copied;
688 char *page;
689 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
690 unsigned long dst = *ppos;
691
692 copied = -ESRCH;
693 if (!task)
694 goto out_no_task;
695
696 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
697 goto out;
698
699 copied = -ENOMEM;
700 page = (char *)__get_free_page(GFP_USER);
701 if (!page)
702 goto out;
703
704 copied = 0;
705 while (count > 0) {
706 int this_len, retval;
707
708 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
709 if (copy_from_user(page, buf, this_len)) {
710 copied = -EFAULT;
711 break;
712 }
713 retval = access_process_vm(task, dst, page, this_len, 1);
714 if (!retval) {
715 if (!copied)
716 copied = -EIO;
717 break;
718 }
719 copied += retval;
720 buf += retval;
721 dst += retval;
722 count -= retval;
723 }
724 *ppos = dst;
725 free_page((unsigned long) page);
726 out:
727 put_task_struct(task);
728 out_no_task:
729 return copied;
730 }
731 #endif
732
733 static loff_t mem_lseek(struct file * file, loff_t offset, int orig)
734 {
735 switch (orig) {
736 case 0:
737 file->f_pos = offset;
738 break;
739 case 1:
740 file->f_pos += offset;
741 break;
742 default:
743 return -EINVAL;
744 }
745 force_successful_syscall_return();
746 return file->f_pos;
747 }
748
749 static struct file_operations proc_mem_operations = {
750 .llseek = mem_lseek,
751 .read = mem_read,
752 .write = mem_write,
753 .open = mem_open,
754 };
755
756 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
757 size_t count, loff_t *ppos)
758 {
759 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
760 char buffer[PROC_NUMBUF];
761 size_t len;
762 int oom_adjust;
763 loff_t __ppos = *ppos;
764
765 if (!task)
766 return -ESRCH;
767 oom_adjust = task->oomkilladj;
768 put_task_struct(task);
769
770 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
771 if (__ppos >= len)
772 return 0;
773 if (count > len-__ppos)
774 count = len-__ppos;
775 if (copy_to_user(buf, buffer + __ppos, count))
776 return -EFAULT;
777 *ppos = __ppos + count;
778 return count;
779 }
780
781 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
782 size_t count, loff_t *ppos)
783 {
784 struct task_struct *task;
785 char buffer[PROC_NUMBUF], *end;
786 int oom_adjust;
787
788 if (!capable(CAP_SYS_RESOURCE))
789 return -EPERM;
790 memset(buffer, 0, sizeof(buffer));
791 if (count > sizeof(buffer) - 1)
792 count = sizeof(buffer) - 1;
793 if (copy_from_user(buffer, buf, count))
794 return -EFAULT;
795 oom_adjust = simple_strtol(buffer, &end, 0);
796 if ((oom_adjust < -16 || oom_adjust > 15) && oom_adjust != OOM_DISABLE)
797 return -EINVAL;
798 if (*end == '\n')
799 end++;
800 task = get_proc_task(file->f_dentry->d_inode);
801 if (!task)
802 return -ESRCH;
803 task->oomkilladj = oom_adjust;
804 put_task_struct(task);
805 if (end - buffer == 0)
806 return -EIO;
807 return end - buffer;
808 }
809
810 static struct file_operations proc_oom_adjust_operations = {
811 .read = oom_adjust_read,
812 .write = oom_adjust_write,
813 };
814
815 #ifdef CONFIG_AUDITSYSCALL
816 #define TMPBUFLEN 21
817 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
818 size_t count, loff_t *ppos)
819 {
820 struct inode * inode = file->f_dentry->d_inode;
821 struct task_struct *task = get_proc_task(inode);
822 ssize_t length;
823 char tmpbuf[TMPBUFLEN];
824
825 if (!task)
826 return -ESRCH;
827 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
828 audit_get_loginuid(task->audit_context));
829 put_task_struct(task);
830 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
831 }
832
833 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
834 size_t count, loff_t *ppos)
835 {
836 struct inode * inode = file->f_dentry->d_inode;
837 char *page, *tmp;
838 ssize_t length;
839 uid_t loginuid;
840
841 if (!capable(CAP_AUDIT_CONTROL))
842 return -EPERM;
843
844 if (current != pid_task(proc_pid(inode), PIDTYPE_PID))
845 return -EPERM;
846
847 if (count >= PAGE_SIZE)
848 count = PAGE_SIZE - 1;
849
850 if (*ppos != 0) {
851 /* No partial writes. */
852 return -EINVAL;
853 }
854 page = (char*)__get_free_page(GFP_USER);
855 if (!page)
856 return -ENOMEM;
857 length = -EFAULT;
858 if (copy_from_user(page, buf, count))
859 goto out_free_page;
860
861 page[count] = '\0';
862 loginuid = simple_strtoul(page, &tmp, 10);
863 if (tmp == page) {
864 length = -EINVAL;
865 goto out_free_page;
866
867 }
868 length = audit_set_loginuid(current, loginuid);
869 if (likely(length == 0))
870 length = count;
871
872 out_free_page:
873 free_page((unsigned long) page);
874 return length;
875 }
876
877 static struct file_operations proc_loginuid_operations = {
878 .read = proc_loginuid_read,
879 .write = proc_loginuid_write,
880 };
881 #endif
882
883 #ifdef CONFIG_SECCOMP
884 static ssize_t seccomp_read(struct file *file, char __user *buf,
885 size_t count, loff_t *ppos)
886 {
887 struct task_struct *tsk = get_proc_task(file->f_dentry->d_inode);
888 char __buf[20];
889 loff_t __ppos = *ppos;
890 size_t len;
891
892 if (!tsk)
893 return -ESRCH;
894 /* no need to print the trailing zero, so use only len */
895 len = sprintf(__buf, "%u\n", tsk->seccomp.mode);
896 put_task_struct(tsk);
897 if (__ppos >= len)
898 return 0;
899 if (count > len - __ppos)
900 count = len - __ppos;
901 if (copy_to_user(buf, __buf + __ppos, count))
902 return -EFAULT;
903 *ppos = __ppos + count;
904 return count;
905 }
906
907 static ssize_t seccomp_write(struct file *file, const char __user *buf,
908 size_t count, loff_t *ppos)
909 {
910 struct task_struct *tsk = get_proc_task(file->f_dentry->d_inode);
911 char __buf[20], *end;
912 unsigned int seccomp_mode;
913 ssize_t result;
914
915 result = -ESRCH;
916 if (!tsk)
917 goto out_no_task;
918
919 /* can set it only once to be even more secure */
920 result = -EPERM;
921 if (unlikely(tsk->seccomp.mode))
922 goto out;
923
924 result = -EFAULT;
925 memset(__buf, 0, sizeof(__buf));
926 count = min(count, sizeof(__buf) - 1);
927 if (copy_from_user(__buf, buf, count))
928 goto out;
929
930 seccomp_mode = simple_strtoul(__buf, &end, 0);
931 if (*end == '\n')
932 end++;
933 result = -EINVAL;
934 if (seccomp_mode && seccomp_mode <= NR_SECCOMP_MODES) {
935 tsk->seccomp.mode = seccomp_mode;
936 set_tsk_thread_flag(tsk, TIF_SECCOMP);
937 } else
938 goto out;
939 result = -EIO;
940 if (unlikely(!(end - __buf)))
941 goto out;
942 result = end - __buf;
943 out:
944 put_task_struct(tsk);
945 out_no_task:
946 return result;
947 }
948
949 static struct file_operations proc_seccomp_operations = {
950 .read = seccomp_read,
951 .write = seccomp_write,
952 };
953 #endif /* CONFIG_SECCOMP */
954
955 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
956 {
957 struct inode *inode = dentry->d_inode;
958 int error = -EACCES;
959
960 /* We don't need a base pointer in the /proc filesystem */
961 path_release(nd);
962
963 /* Are we allowed to snoop on the tasks file descriptors? */
964 if (!proc_fd_access_allowed(inode))
965 goto out;
966
967 error = PROC_I(inode)->op.proc_get_link(inode, &nd->dentry, &nd->mnt);
968 nd->last_type = LAST_BIND;
969 out:
970 return ERR_PTR(error);
971 }
972
973 static int do_proc_readlink(struct dentry *dentry, struct vfsmount *mnt,
974 char __user *buffer, int buflen)
975 {
976 struct inode * inode;
977 char *tmp = (char*)__get_free_page(GFP_KERNEL), *path;
978 int len;
979
980 if (!tmp)
981 return -ENOMEM;
982
983 inode = dentry->d_inode;
984 path = d_path(dentry, mnt, tmp, PAGE_SIZE);
985 len = PTR_ERR(path);
986 if (IS_ERR(path))
987 goto out;
988 len = tmp + PAGE_SIZE - 1 - path;
989
990 if (len > buflen)
991 len = buflen;
992 if (copy_to_user(buffer, path, len))
993 len = -EFAULT;
994 out:
995 free_page((unsigned long)tmp);
996 return len;
997 }
998
999 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1000 {
1001 int error = -EACCES;
1002 struct inode *inode = dentry->d_inode;
1003 struct dentry *de;
1004 struct vfsmount *mnt = NULL;
1005
1006 /* Are we allowed to snoop on the tasks file descriptors? */
1007 if (!proc_fd_access_allowed(inode))
1008 goto out;
1009
1010 error = PROC_I(inode)->op.proc_get_link(inode, &de, &mnt);
1011 if (error)
1012 goto out;
1013
1014 error = do_proc_readlink(de, mnt, buffer, buflen);
1015 dput(de);
1016 mntput(mnt);
1017 out:
1018 return error;
1019 }
1020
1021 static struct inode_operations proc_pid_link_inode_operations = {
1022 .readlink = proc_pid_readlink,
1023 .follow_link = proc_pid_follow_link,
1024 .setattr = proc_setattr,
1025 };
1026
1027
1028 /* building an inode */
1029
1030 static int task_dumpable(struct task_struct *task)
1031 {
1032 int dumpable = 0;
1033 struct mm_struct *mm;
1034
1035 task_lock(task);
1036 mm = task->mm;
1037 if (mm)
1038 dumpable = mm->dumpable;
1039 task_unlock(task);
1040 if(dumpable == 1)
1041 return 1;
1042 return 0;
1043 }
1044
1045
1046 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task, int ino)
1047 {
1048 struct inode * inode;
1049 struct proc_inode *ei;
1050
1051 /* We need a new inode */
1052
1053 inode = new_inode(sb);
1054 if (!inode)
1055 goto out;
1056
1057 /* Common stuff */
1058 ei = PROC_I(inode);
1059 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1060 inode->i_ino = fake_ino(task->pid, ino);
1061 inode->i_op = &proc_def_inode_operations;
1062
1063 /*
1064 * grab the reference to task.
1065 */
1066 ei->pid = get_pid(task->pids[PIDTYPE_PID].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 static unsigned name_to_int(struct dentry *dentry)
1164 {
1165 const char *name = dentry->d_name.name;
1166 int len = dentry->d_name.len;
1167 unsigned n = 0;
1168
1169 if (len > 1 && *name == '0')
1170 goto out;
1171 while (len-- > 0) {
1172 unsigned c = *name++ - '0';
1173 if (c > 9)
1174 goto out;
1175 if (n >= (~0U-9)/10)
1176 goto out;
1177 n *= 10;
1178 n += c;
1179 }
1180 return n;
1181 out:
1182 return ~0U;
1183 }
1184
1185 static int proc_fd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
1186 {
1187 struct task_struct *task = get_proc_task(inode);
1188 struct files_struct *files = NULL;
1189 struct file *file;
1190 int fd = proc_fd(inode);
1191
1192 if (task) {
1193 files = get_files_struct(task);
1194 put_task_struct(task);
1195 }
1196 if (files) {
1197 /*
1198 * We are not taking a ref to the file structure, so we must
1199 * hold ->file_lock.
1200 */
1201 spin_lock(&files->file_lock);
1202 file = fcheck_files(files, fd);
1203 if (file) {
1204 *mnt = mntget(file->f_vfsmnt);
1205 *dentry = dget(file->f_dentry);
1206 spin_unlock(&files->file_lock);
1207 put_files_struct(files);
1208 return 0;
1209 }
1210 spin_unlock(&files->file_lock);
1211 put_files_struct(files);
1212 }
1213 return -ENOENT;
1214 }
1215
1216 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1217 {
1218 struct inode *inode = dentry->d_inode;
1219 struct task_struct *task = get_proc_task(inode);
1220 int fd = proc_fd(inode);
1221 struct files_struct *files;
1222
1223 if (task) {
1224 files = get_files_struct(task);
1225 if (files) {
1226 rcu_read_lock();
1227 if (fcheck_files(files, fd)) {
1228 rcu_read_unlock();
1229 put_files_struct(files);
1230 if (task_dumpable(task)) {
1231 inode->i_uid = task->euid;
1232 inode->i_gid = task->egid;
1233 } else {
1234 inode->i_uid = 0;
1235 inode->i_gid = 0;
1236 }
1237 inode->i_mode &= ~(S_ISUID | S_ISGID);
1238 security_task_to_inode(task, inode);
1239 put_task_struct(task);
1240 return 1;
1241 }
1242 rcu_read_unlock();
1243 put_files_struct(files);
1244 }
1245 put_task_struct(task);
1246 }
1247 d_drop(dentry);
1248 return 0;
1249 }
1250
1251 static struct dentry_operations tid_fd_dentry_operations =
1252 {
1253 .d_revalidate = tid_fd_revalidate,
1254 .d_delete = pid_delete_dentry,
1255 };
1256
1257 /* SMP-safe */
1258 static struct dentry *proc_lookupfd(struct inode * dir, struct dentry * dentry, struct nameidata *nd)
1259 {
1260 struct task_struct *task = get_proc_task(dir);
1261 unsigned fd = name_to_int(dentry);
1262 struct dentry *result = ERR_PTR(-ENOENT);
1263 struct file * file;
1264 struct files_struct * files;
1265 struct inode *inode;
1266 struct proc_inode *ei;
1267
1268 if (!task)
1269 goto out_no_task;
1270 if (fd == ~0U)
1271 goto out;
1272
1273 inode = proc_pid_make_inode(dir->i_sb, task, PROC_TID_FD_DIR+fd);
1274 if (!inode)
1275 goto out;
1276 ei = PROC_I(inode);
1277 ei->fd = fd;
1278 files = get_files_struct(task);
1279 if (!files)
1280 goto out_unlock;
1281 inode->i_mode = S_IFLNK;
1282
1283 /*
1284 * We are not taking a ref to the file structure, so we must
1285 * hold ->file_lock.
1286 */
1287 spin_lock(&files->file_lock);
1288 file = fcheck_files(files, fd);
1289 if (!file)
1290 goto out_unlock2;
1291 if (file->f_mode & 1)
1292 inode->i_mode |= S_IRUSR | S_IXUSR;
1293 if (file->f_mode & 2)
1294 inode->i_mode |= S_IWUSR | S_IXUSR;
1295 spin_unlock(&files->file_lock);
1296 put_files_struct(files);
1297 inode->i_op = &proc_pid_link_inode_operations;
1298 inode->i_size = 64;
1299 ei->op.proc_get_link = proc_fd_link;
1300 dentry->d_op = &tid_fd_dentry_operations;
1301 d_add(dentry, inode);
1302 /* Close the race of the process dying before we return the dentry */
1303 if (tid_fd_revalidate(dentry, NULL))
1304 result = NULL;
1305 out:
1306 put_task_struct(task);
1307 out_no_task:
1308 return result;
1309
1310 out_unlock2:
1311 spin_unlock(&files->file_lock);
1312 put_files_struct(files);
1313 out_unlock:
1314 iput(inode);
1315 goto out;
1316 }
1317
1318 static int proc_readfd(struct file * filp, void * dirent, filldir_t filldir)
1319 {
1320 struct dentry *dentry = filp->f_dentry;
1321 struct inode *inode = dentry->d_inode;
1322 struct task_struct *p = get_proc_task(inode);
1323 unsigned int fd, tid, ino;
1324 int retval;
1325 char buf[PROC_NUMBUF];
1326 struct files_struct * files;
1327 struct fdtable *fdt;
1328
1329 retval = -ENOENT;
1330 if (!p)
1331 goto out_no_task;
1332 retval = 0;
1333 tid = p->pid;
1334
1335 fd = filp->f_pos;
1336 switch (fd) {
1337 case 0:
1338 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1339 goto out;
1340 filp->f_pos++;
1341 case 1:
1342 ino = parent_ino(dentry);
1343 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1344 goto out;
1345 filp->f_pos++;
1346 default:
1347 files = get_files_struct(p);
1348 if (!files)
1349 goto out;
1350 rcu_read_lock();
1351 fdt = files_fdtable(files);
1352 for (fd = filp->f_pos-2;
1353 fd < fdt->max_fds;
1354 fd++, filp->f_pos++) {
1355 unsigned int i,j;
1356
1357 if (!fcheck_files(files, fd))
1358 continue;
1359 rcu_read_unlock();
1360
1361 j = PROC_NUMBUF;
1362 i = fd;
1363 do {
1364 j--;
1365 buf[j] = '0' + (i % 10);
1366 i /= 10;
1367 } while (i);
1368
1369 ino = fake_ino(tid, PROC_TID_FD_DIR + fd);
1370 if (filldir(dirent, buf+j, PROC_NUMBUF-j, fd+2, ino, DT_LNK) < 0) {
1371 rcu_read_lock();
1372 break;
1373 }
1374 rcu_read_lock();
1375 }
1376 rcu_read_unlock();
1377 put_files_struct(files);
1378 }
1379 out:
1380 put_task_struct(p);
1381 out_no_task:
1382 return retval;
1383 }
1384
1385 static struct file_operations proc_fd_operations = {
1386 .read = generic_read_dir,
1387 .readdir = proc_readfd,
1388 };
1389
1390 /*
1391 * proc directories can do almost nothing..
1392 */
1393 static struct inode_operations proc_fd_inode_operations = {
1394 .lookup = proc_lookupfd,
1395 .setattr = proc_setattr,
1396 };
1397
1398 /* SMP-safe */
1399 static struct dentry *proc_pident_lookup(struct inode *dir,
1400 struct dentry *dentry,
1401 struct pid_entry *ents)
1402 {
1403 struct inode *inode;
1404 struct dentry *error;
1405 struct task_struct *task = get_proc_task(dir);
1406 struct pid_entry *p;
1407 struct proc_inode *ei;
1408
1409 error = ERR_PTR(-ENOENT);
1410 inode = NULL;
1411
1412 if (!task)
1413 goto out_no_task;
1414
1415 /*
1416 * Yes, it does not scale. And it should not. Don't add
1417 * new entries into /proc/<tgid>/ without very good reasons.
1418 */
1419 for (p = ents; p->name; p++) {
1420 if (p->len != dentry->d_name.len)
1421 continue;
1422 if (!memcmp(dentry->d_name.name, p->name, p->len))
1423 break;
1424 }
1425 if (!p->name)
1426 goto out;
1427
1428 error = ERR_PTR(-EINVAL);
1429 inode = proc_pid_make_inode(dir->i_sb, task, p->type);
1430 if (!inode)
1431 goto out;
1432
1433 ei = PROC_I(inode);
1434 inode->i_mode = p->mode;
1435 if (S_ISDIR(inode->i_mode))
1436 inode->i_nlink = 2; /* Use getattr to fix if necessary */
1437 if (p->iop)
1438 inode->i_op = p->iop;
1439 if (p->fop)
1440 inode->i_fop = p->fop;
1441 ei->op = p->op;
1442 dentry->d_op = &pid_dentry_operations;
1443 d_add(dentry, inode);
1444 /* Close the race of the process dying before we return the dentry */
1445 if (pid_revalidate(dentry, NULL))
1446 error = NULL;
1447 out:
1448 put_task_struct(task);
1449 out_no_task:
1450 return error;
1451 }
1452
1453 static int proc_pident_readdir(struct file *filp,
1454 void *dirent, filldir_t filldir,
1455 struct pid_entry *ents, unsigned int nents)
1456 {
1457 int i;
1458 int pid;
1459 struct dentry *dentry = filp->f_dentry;
1460 struct inode *inode = dentry->d_inode;
1461 struct task_struct *task = get_proc_task(inode);
1462 struct pid_entry *p;
1463 ino_t ino;
1464 int ret;
1465
1466 ret = -ENOENT;
1467 if (!task)
1468 goto out;
1469
1470 ret = 0;
1471 pid = task->pid;
1472 put_task_struct(task);
1473 i = filp->f_pos;
1474 switch (i) {
1475 case 0:
1476 ino = inode->i_ino;
1477 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
1478 goto out;
1479 i++;
1480 filp->f_pos++;
1481 /* fall through */
1482 case 1:
1483 ino = parent_ino(dentry);
1484 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
1485 goto out;
1486 i++;
1487 filp->f_pos++;
1488 /* fall through */
1489 default:
1490 i -= 2;
1491 if (i >= nents) {
1492 ret = 1;
1493 goto out;
1494 }
1495 p = ents + i;
1496 while (p->name) {
1497 if (filldir(dirent, p->name, p->len, filp->f_pos,
1498 fake_ino(pid, p->type), p->mode >> 12) < 0)
1499 goto out;
1500 filp->f_pos++;
1501 p++;
1502 }
1503 }
1504
1505 ret = 1;
1506 out:
1507 return ret;
1508 }
1509
1510 #ifdef CONFIG_SECURITY
1511 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
1512 size_t count, loff_t *ppos)
1513 {
1514 struct inode * inode = file->f_dentry->d_inode;
1515 unsigned long page;
1516 ssize_t length;
1517 struct task_struct *task = get_proc_task(inode);
1518
1519 length = -ESRCH;
1520 if (!task)
1521 goto out_no_task;
1522
1523 if (count > PAGE_SIZE)
1524 count = PAGE_SIZE;
1525 length = -ENOMEM;
1526 if (!(page = __get_free_page(GFP_KERNEL)))
1527 goto out;
1528
1529 length = security_getprocattr(task,
1530 (char*)file->f_dentry->d_name.name,
1531 (void*)page, count);
1532 if (length >= 0)
1533 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
1534 free_page(page);
1535 out:
1536 put_task_struct(task);
1537 out_no_task:
1538 return length;
1539 }
1540
1541 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
1542 size_t count, loff_t *ppos)
1543 {
1544 struct inode * inode = file->f_dentry->d_inode;
1545 char *page;
1546 ssize_t length;
1547 struct task_struct *task = get_proc_task(inode);
1548
1549 length = -ESRCH;
1550 if (!task)
1551 goto out_no_task;
1552 if (count > PAGE_SIZE)
1553 count = PAGE_SIZE;
1554
1555 /* No partial writes. */
1556 length = -EINVAL;
1557 if (*ppos != 0)
1558 goto out;
1559
1560 length = -ENOMEM;
1561 page = (char*)__get_free_page(GFP_USER);
1562 if (!page)
1563 goto out;
1564
1565 length = -EFAULT;
1566 if (copy_from_user(page, buf, count))
1567 goto out_free;
1568
1569 length = security_setprocattr(task,
1570 (char*)file->f_dentry->d_name.name,
1571 (void*)page, count);
1572 out_free:
1573 free_page((unsigned long) page);
1574 out:
1575 put_task_struct(task);
1576 out_no_task:
1577 return length;
1578 }
1579
1580 static struct file_operations proc_pid_attr_operations = {
1581 .read = proc_pid_attr_read,
1582 .write = proc_pid_attr_write,
1583 };
1584
1585 static struct pid_entry tgid_attr_stuff[] = {
1586 REG(PROC_TGID_ATTR_CURRENT, "current", S_IRUGO|S_IWUGO, pid_attr),
1587 REG(PROC_TGID_ATTR_PREV, "prev", S_IRUGO, pid_attr),
1588 REG(PROC_TGID_ATTR_EXEC, "exec", S_IRUGO|S_IWUGO, pid_attr),
1589 REG(PROC_TGID_ATTR_FSCREATE, "fscreate", S_IRUGO|S_IWUGO, pid_attr),
1590 REG(PROC_TGID_ATTR_KEYCREATE, "keycreate", S_IRUGO|S_IWUGO, pid_attr),
1591 REG(PROC_TGID_ATTR_SOCKCREATE, "sockcreate", S_IRUGO|S_IWUGO, pid_attr),
1592 {}
1593 };
1594 static struct pid_entry tid_attr_stuff[] = {
1595 REG(PROC_TID_ATTR_CURRENT, "current", S_IRUGO|S_IWUGO, pid_attr),
1596 REG(PROC_TID_ATTR_PREV, "prev", S_IRUGO, pid_attr),
1597 REG(PROC_TID_ATTR_EXEC, "exec", S_IRUGO|S_IWUGO, pid_attr),
1598 REG(PROC_TID_ATTR_FSCREATE, "fscreate", S_IRUGO|S_IWUGO, pid_attr),
1599 REG(PROC_TID_ATTR_KEYCREATE, "keycreate", S_IRUGO|S_IWUGO, pid_attr),
1600 REG(PROC_TID_ATTR_SOCKCREATE, "sockcreate", S_IRUGO|S_IWUGO, pid_attr),
1601 {}
1602 };
1603
1604 static int proc_tgid_attr_readdir(struct file * filp,
1605 void * dirent, filldir_t filldir)
1606 {
1607 return proc_pident_readdir(filp,dirent,filldir,
1608 tgid_attr_stuff,ARRAY_SIZE(tgid_attr_stuff));
1609 }
1610
1611 static int proc_tid_attr_readdir(struct file * filp,
1612 void * dirent, filldir_t filldir)
1613 {
1614 return proc_pident_readdir(filp,dirent,filldir,
1615 tid_attr_stuff,ARRAY_SIZE(tid_attr_stuff));
1616 }
1617
1618 static struct file_operations proc_tgid_attr_operations = {
1619 .read = generic_read_dir,
1620 .readdir = proc_tgid_attr_readdir,
1621 };
1622
1623 static struct file_operations proc_tid_attr_operations = {
1624 .read = generic_read_dir,
1625 .readdir = proc_tid_attr_readdir,
1626 };
1627
1628 static struct dentry *proc_tgid_attr_lookup(struct inode *dir,
1629 struct dentry *dentry, struct nameidata *nd)
1630 {
1631 return proc_pident_lookup(dir, dentry, tgid_attr_stuff);
1632 }
1633
1634 static struct dentry *proc_tid_attr_lookup(struct inode *dir,
1635 struct dentry *dentry, struct nameidata *nd)
1636 {
1637 return proc_pident_lookup(dir, dentry, tid_attr_stuff);
1638 }
1639
1640 static struct inode_operations proc_tgid_attr_inode_operations = {
1641 .lookup = proc_tgid_attr_lookup,
1642 .getattr = pid_getattr,
1643 .setattr = proc_setattr,
1644 };
1645
1646 static struct inode_operations proc_tid_attr_inode_operations = {
1647 .lookup = proc_tid_attr_lookup,
1648 .getattr = pid_getattr,
1649 .setattr = proc_setattr,
1650 };
1651 #endif
1652
1653 /*
1654 * /proc/self:
1655 */
1656 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
1657 int buflen)
1658 {
1659 char tmp[PROC_NUMBUF];
1660 sprintf(tmp, "%d", current->tgid);
1661 return vfs_readlink(dentry,buffer,buflen,tmp);
1662 }
1663
1664 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
1665 {
1666 char tmp[PROC_NUMBUF];
1667 sprintf(tmp, "%d", current->tgid);
1668 return ERR_PTR(vfs_follow_link(nd,tmp));
1669 }
1670
1671 static struct inode_operations proc_self_inode_operations = {
1672 .readlink = proc_self_readlink,
1673 .follow_link = proc_self_follow_link,
1674 };
1675
1676 /*
1677 * Thread groups
1678 */
1679 static struct file_operations proc_task_operations;
1680 static struct inode_operations proc_task_inode_operations;
1681
1682 static struct pid_entry tgid_base_stuff[] = {
1683 DIR(PROC_TGID_TASK, "task", S_IRUGO|S_IXUGO, task),
1684 DIR(PROC_TGID_FD, "fd", S_IRUSR|S_IXUSR, fd),
1685 INF(PROC_TGID_ENVIRON, "environ", S_IRUSR, pid_environ),
1686 INF(PROC_TGID_AUXV, "auxv", S_IRUSR, pid_auxv),
1687 INF(PROC_TGID_STATUS, "status", S_IRUGO, pid_status),
1688 INF(PROC_TGID_CMDLINE, "cmdline", S_IRUGO, pid_cmdline),
1689 INF(PROC_TGID_STAT, "stat", S_IRUGO, tgid_stat),
1690 INF(PROC_TGID_STATM, "statm", S_IRUGO, pid_statm),
1691 REG(PROC_TGID_MAPS, "maps", S_IRUGO, maps),
1692 #ifdef CONFIG_NUMA
1693 REG(PROC_TGID_NUMA_MAPS, "numa_maps", S_IRUGO, numa_maps),
1694 #endif
1695 REG(PROC_TGID_MEM, "mem", S_IRUSR|S_IWUSR, mem),
1696 #ifdef CONFIG_SECCOMP
1697 REG(PROC_TGID_SECCOMP, "seccomp", S_IRUSR|S_IWUSR, seccomp),
1698 #endif
1699 LNK(PROC_TGID_CWD, "cwd", cwd),
1700 LNK(PROC_TGID_ROOT, "root", root),
1701 LNK(PROC_TGID_EXE, "exe", exe),
1702 REG(PROC_TGID_MOUNTS, "mounts", S_IRUGO, mounts),
1703 REG(PROC_TGID_MOUNTSTATS, "mountstats", S_IRUSR, mountstats),
1704 #ifdef CONFIG_MMU
1705 REG(PROC_TGID_SMAPS, "smaps", S_IRUGO, smaps),
1706 #endif
1707 #ifdef CONFIG_SECURITY
1708 DIR(PROC_TGID_ATTR, "attr", S_IRUGO|S_IXUGO, tgid_attr),
1709 #endif
1710 #ifdef CONFIG_KALLSYMS
1711 INF(PROC_TGID_WCHAN, "wchan", S_IRUGO, pid_wchan),
1712 #endif
1713 #ifdef CONFIG_SCHEDSTATS
1714 INF(PROC_TGID_SCHEDSTAT, "schedstat", S_IRUGO, pid_schedstat),
1715 #endif
1716 #ifdef CONFIG_CPUSETS
1717 REG(PROC_TGID_CPUSET, "cpuset", S_IRUGO, cpuset),
1718 #endif
1719 INF(PROC_TGID_OOM_SCORE, "oom_score", S_IRUGO, oom_score),
1720 REG(PROC_TGID_OOM_ADJUST, "oom_adj", S_IRUGO|S_IWUSR, oom_adjust),
1721 #ifdef CONFIG_AUDITSYSCALL
1722 REG(PROC_TGID_LOGINUID, "loginuid", S_IWUSR|S_IRUGO, loginuid),
1723 #endif
1724 {}
1725 };
1726
1727 static int proc_tgid_base_readdir(struct file * filp,
1728 void * dirent, filldir_t filldir)
1729 {
1730 return proc_pident_readdir(filp,dirent,filldir,
1731 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
1732 }
1733
1734 static struct file_operations proc_tgid_base_operations = {
1735 .read = generic_read_dir,
1736 .readdir = proc_tgid_base_readdir,
1737 };
1738
1739 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
1740 return proc_pident_lookup(dir, dentry, tgid_base_stuff);
1741 }
1742
1743 static struct inode_operations proc_tgid_base_inode_operations = {
1744 .lookup = proc_tgid_base_lookup,
1745 .getattr = pid_getattr,
1746 .setattr = proc_setattr,
1747 };
1748
1749 /**
1750 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
1751 *
1752 * @task: task that should be flushed.
1753 *
1754 * Looks in the dcache for
1755 * /proc/@pid
1756 * /proc/@tgid/task/@pid
1757 * if either directory is present flushes it and all of it'ts children
1758 * from the dcache.
1759 *
1760 * It is safe and reasonable to cache /proc entries for a task until
1761 * that task exits. After that they just clog up the dcache with
1762 * useless entries, possibly causing useful dcache entries to be
1763 * flushed instead. This routine is proved to flush those useless
1764 * dcache entries at process exit time.
1765 *
1766 * NOTE: This routine is just an optimization so it does not guarantee
1767 * that no dcache entries will exist at process exit time it
1768 * just makes it very unlikely that any will persist.
1769 */
1770 void proc_flush_task(struct task_struct *task)
1771 {
1772 struct dentry *dentry, *leader, *dir;
1773 char buf[PROC_NUMBUF];
1774 struct qstr name;
1775
1776 name.name = buf;
1777 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
1778 dentry = d_hash_and_lookup(proc_mnt->mnt_root, &name);
1779 if (dentry) {
1780 shrink_dcache_parent(dentry);
1781 d_drop(dentry);
1782 dput(dentry);
1783 }
1784
1785 if (thread_group_leader(task))
1786 goto out;
1787
1788 name.name = buf;
1789 name.len = snprintf(buf, sizeof(buf), "%d", task->tgid);
1790 leader = d_hash_and_lookup(proc_mnt->mnt_root, &name);
1791 if (!leader)
1792 goto out;
1793
1794 name.name = "task";
1795 name.len = strlen(name.name);
1796 dir = d_hash_and_lookup(leader, &name);
1797 if (!dir)
1798 goto out_put_leader;
1799
1800 name.name = buf;
1801 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
1802 dentry = d_hash_and_lookup(dir, &name);
1803 if (dentry) {
1804 shrink_dcache_parent(dentry);
1805 d_drop(dentry);
1806 dput(dentry);
1807 }
1808
1809 dput(dir);
1810 out_put_leader:
1811 dput(leader);
1812 out:
1813 return;
1814 }
1815
1816 /* SMP-safe */
1817 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
1818 {
1819 struct dentry *result = ERR_PTR(-ENOENT);
1820 struct task_struct *task;
1821 struct inode *inode;
1822 struct proc_inode *ei;
1823 unsigned tgid;
1824
1825 if (dentry->d_name.len == 4 && !memcmp(dentry->d_name.name,"self",4)) {
1826 inode = new_inode(dir->i_sb);
1827 if (!inode)
1828 return ERR_PTR(-ENOMEM);
1829 ei = PROC_I(inode);
1830 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1831 inode->i_ino = fake_ino(0, PROC_TGID_INO);
1832 ei->pde = NULL;
1833 inode->i_mode = S_IFLNK|S_IRWXUGO;
1834 inode->i_uid = inode->i_gid = 0;
1835 inode->i_size = 64;
1836 inode->i_op = &proc_self_inode_operations;
1837 d_add(dentry, inode);
1838 return NULL;
1839 }
1840 tgid = name_to_int(dentry);
1841 if (tgid == ~0U)
1842 goto out;
1843
1844 rcu_read_lock();
1845 task = find_task_by_pid(tgid);
1846 if (task)
1847 get_task_struct(task);
1848 rcu_read_unlock();
1849 if (!task)
1850 goto out;
1851
1852 inode = proc_pid_make_inode(dir->i_sb, task, PROC_TGID_INO);
1853 if (!inode)
1854 goto out_put_task;
1855
1856 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
1857 inode->i_op = &proc_tgid_base_inode_operations;
1858 inode->i_fop = &proc_tgid_base_operations;
1859 inode->i_flags|=S_IMMUTABLE;
1860 #ifdef CONFIG_SECURITY
1861 inode->i_nlink = 5;
1862 #else
1863 inode->i_nlink = 4;
1864 #endif
1865
1866 dentry->d_op = &pid_dentry_operations;
1867
1868 d_add(dentry, inode);
1869 /* Close the race of the process dying before we return the dentry */
1870 if (pid_revalidate(dentry, NULL))
1871 result = NULL;
1872
1873 out_put_task:
1874 put_task_struct(task);
1875 out:
1876 return result;
1877 }
1878
1879 /*
1880 * Find the first task with tgid >= tgid
1881 *
1882 */
1883 static struct task_struct *next_tgid(unsigned int tgid)
1884 {
1885 struct task_struct *task;
1886 struct pid *pid;
1887
1888 rcu_read_lock();
1889 retry:
1890 task = NULL;
1891 pid = find_ge_pid(tgid);
1892 if (pid) {
1893 tgid = pid->nr + 1;
1894 task = pid_task(pid, PIDTYPE_PID);
1895 /* What we to know is if the pid we have find is the
1896 * pid of a thread_group_leader. Testing for task
1897 * being a thread_group_leader is the obvious thing
1898 * todo but there is a window when it fails, due to
1899 * the pid transfer logic in de_thread.
1900 *
1901 * So we perform the straight forward test of seeing
1902 * if the pid we have found is the pid of a thread
1903 * group leader, and don't worry if the task we have
1904 * found doesn't happen to be a thread group leader.
1905 * As we don't care in the case of readdir.
1906 */
1907 if (!task || !has_group_leader_pid(task))
1908 goto retry;
1909 get_task_struct(task);
1910 }
1911 rcu_read_unlock();
1912 return task;
1913 }
1914
1915 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + (1 /* /proc/self */))
1916
1917 /* for the /proc/ directory itself, after non-process stuff has been done */
1918 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
1919 {
1920 char buf[PROC_NUMBUF];
1921 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
1922 struct task_struct *task;
1923 int tgid;
1924
1925 if (!nr) {
1926 ino_t ino = fake_ino(0,PROC_TGID_INO);
1927 if (filldir(dirent, "self", 4, filp->f_pos, ino, DT_LNK) < 0)
1928 return 0;
1929 filp->f_pos++;
1930 nr++;
1931 }
1932
1933 tgid = filp->f_pos - TGID_OFFSET;
1934 for (task = next_tgid(tgid);
1935 task;
1936 put_task_struct(task), task = next_tgid(tgid + 1)) {
1937 int len;
1938 ino_t ino;
1939 tgid = task->pid;
1940 filp->f_pos = tgid + TGID_OFFSET;
1941 len = snprintf(buf, sizeof(buf), "%d", tgid);
1942 ino = fake_ino(tgid, PROC_TGID_INO);
1943 if (filldir(dirent, buf, len, filp->f_pos, ino, DT_DIR) < 0) {
1944 put_task_struct(task);
1945 goto out;
1946 }
1947 }
1948 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
1949 out:
1950 return 0;
1951 }
1952
1953 /*
1954 * Tasks
1955 */
1956 static struct pid_entry tid_base_stuff[] = {
1957 DIR(PROC_TID_FD, "fd", S_IRUSR|S_IXUSR, fd),
1958 INF(PROC_TID_ENVIRON, "environ", S_IRUSR, pid_environ),
1959 INF(PROC_TID_AUXV, "auxv", S_IRUSR, pid_auxv),
1960 INF(PROC_TID_STATUS, "status", S_IRUGO, pid_status),
1961 INF(PROC_TID_CMDLINE, "cmdline", S_IRUGO, pid_cmdline),
1962 INF(PROC_TID_STAT, "stat", S_IRUGO, tid_stat),
1963 INF(PROC_TID_STATM, "statm", S_IRUGO, pid_statm),
1964 REG(PROC_TID_MAPS, "maps", S_IRUGO, maps),
1965 #ifdef CONFIG_NUMA
1966 REG(PROC_TID_NUMA_MAPS, "numa_maps", S_IRUGO, numa_maps),
1967 #endif
1968 REG(PROC_TID_MEM, "mem", S_IRUSR|S_IWUSR, mem),
1969 #ifdef CONFIG_SECCOMP
1970 REG(PROC_TID_SECCOMP, "seccomp", S_IRUSR|S_IWUSR, seccomp),
1971 #endif
1972 LNK(PROC_TID_CWD, "cwd", cwd),
1973 LNK(PROC_TID_ROOT, "root", root),
1974 LNK(PROC_TID_EXE, "exe", exe),
1975 REG(PROC_TID_MOUNTS, "mounts", S_IRUGO, mounts),
1976 #ifdef CONFIG_MMU
1977 REG(PROC_TID_SMAPS, "smaps", S_IRUGO, smaps),
1978 #endif
1979 #ifdef CONFIG_SECURITY
1980 DIR(PROC_TID_ATTR, "attr", S_IRUGO|S_IXUGO, tid_attr),
1981 #endif
1982 #ifdef CONFIG_KALLSYMS
1983 INF(PROC_TID_WCHAN, "wchan", S_IRUGO, pid_wchan),
1984 #endif
1985 #ifdef CONFIG_SCHEDSTATS
1986 INF(PROC_TID_SCHEDSTAT, "schedstat", S_IRUGO, pid_schedstat),
1987 #endif
1988 #ifdef CONFIG_CPUSETS
1989 REG(PROC_TID_CPUSET, "cpuset", S_IRUGO, cpuset),
1990 #endif
1991 INF(PROC_TID_OOM_SCORE, "oom_score", S_IRUGO, oom_score),
1992 REG(PROC_TID_OOM_ADJUST, "oom_adj", S_IRUGO|S_IWUSR, oom_adjust),
1993 #ifdef CONFIG_AUDITSYSCALL
1994 REG(PROC_TID_LOGINUID, "loginuid", S_IWUSR|S_IRUGO, loginuid),
1995 #endif
1996 {}
1997 };
1998
1999 static int proc_tid_base_readdir(struct file * filp,
2000 void * dirent, filldir_t filldir)
2001 {
2002 return proc_pident_readdir(filp,dirent,filldir,
2003 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
2004 }
2005
2006 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2007 return proc_pident_lookup(dir, dentry, tid_base_stuff);
2008 }
2009
2010 static struct file_operations proc_tid_base_operations = {
2011 .read = generic_read_dir,
2012 .readdir = proc_tid_base_readdir,
2013 };
2014
2015 static struct inode_operations proc_tid_base_inode_operations = {
2016 .lookup = proc_tid_base_lookup,
2017 .getattr = pid_getattr,
2018 .setattr = proc_setattr,
2019 };
2020
2021 /* SMP-safe */
2022 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2023 {
2024 struct dentry *result = ERR_PTR(-ENOENT);
2025 struct task_struct *task;
2026 struct task_struct *leader = get_proc_task(dir);
2027 struct inode *inode;
2028 unsigned tid;
2029
2030 if (!leader)
2031 goto out_no_task;
2032
2033 tid = name_to_int(dentry);
2034 if (tid == ~0U)
2035 goto out;
2036
2037 rcu_read_lock();
2038 task = find_task_by_pid(tid);
2039 if (task)
2040 get_task_struct(task);
2041 rcu_read_unlock();
2042 if (!task)
2043 goto out;
2044 if (leader->tgid != task->tgid)
2045 goto out_drop_task;
2046
2047 inode = proc_pid_make_inode(dir->i_sb, task, PROC_TID_INO);
2048
2049
2050 if (!inode)
2051 goto out_drop_task;
2052 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2053 inode->i_op = &proc_tid_base_inode_operations;
2054 inode->i_fop = &proc_tid_base_operations;
2055 inode->i_flags|=S_IMMUTABLE;
2056 #ifdef CONFIG_SECURITY
2057 inode->i_nlink = 4;
2058 #else
2059 inode->i_nlink = 3;
2060 #endif
2061
2062 dentry->d_op = &pid_dentry_operations;
2063
2064 d_add(dentry, inode);
2065 /* Close the race of the process dying before we return the dentry */
2066 if (pid_revalidate(dentry, NULL))
2067 result = NULL;
2068
2069 out_drop_task:
2070 put_task_struct(task);
2071 out:
2072 put_task_struct(leader);
2073 out_no_task:
2074 return result;
2075 }
2076
2077 /*
2078 * Find the first tid of a thread group to return to user space.
2079 *
2080 * Usually this is just the thread group leader, but if the users
2081 * buffer was too small or there was a seek into the middle of the
2082 * directory we have more work todo.
2083 *
2084 * In the case of a short read we start with find_task_by_pid.
2085 *
2086 * In the case of a seek we start with the leader and walk nr
2087 * threads past it.
2088 */
2089 static struct task_struct *first_tid(struct task_struct *leader,
2090 int tid, int nr)
2091 {
2092 struct task_struct *pos;
2093
2094 rcu_read_lock();
2095 /* Attempt to start with the pid of a thread */
2096 if (tid && (nr > 0)) {
2097 pos = find_task_by_pid(tid);
2098 if (pos && (pos->group_leader == leader))
2099 goto found;
2100 }
2101
2102 /* If nr exceeds the number of threads there is nothing todo */
2103 pos = NULL;
2104 if (nr && nr >= get_nr_threads(leader))
2105 goto out;
2106
2107 /* If we haven't found our starting place yet start
2108 * with the leader and walk nr threads forward.
2109 */
2110 for (pos = leader; nr > 0; --nr) {
2111 pos = next_thread(pos);
2112 if (pos == leader) {
2113 pos = NULL;
2114 goto out;
2115 }
2116 }
2117 found:
2118 get_task_struct(pos);
2119 out:
2120 rcu_read_unlock();
2121 return pos;
2122 }
2123
2124 /*
2125 * Find the next thread in the thread list.
2126 * Return NULL if there is an error or no next thread.
2127 *
2128 * The reference to the input task_struct is released.
2129 */
2130 static struct task_struct *next_tid(struct task_struct *start)
2131 {
2132 struct task_struct *pos = NULL;
2133 rcu_read_lock();
2134 if (pid_alive(start)) {
2135 pos = next_thread(start);
2136 if (thread_group_leader(pos))
2137 pos = NULL;
2138 else
2139 get_task_struct(pos);
2140 }
2141 rcu_read_unlock();
2142 put_task_struct(start);
2143 return pos;
2144 }
2145
2146 /* for the /proc/TGID/task/ directories */
2147 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
2148 {
2149 char buf[PROC_NUMBUF];
2150 struct dentry *dentry = filp->f_dentry;
2151 struct inode *inode = dentry->d_inode;
2152 struct task_struct *leader = get_proc_task(inode);
2153 struct task_struct *task;
2154 int retval = -ENOENT;
2155 ino_t ino;
2156 int tid;
2157 unsigned long pos = filp->f_pos; /* avoiding "long long" filp->f_pos */
2158
2159 if (!leader)
2160 goto out_no_task;
2161 retval = 0;
2162
2163 switch (pos) {
2164 case 0:
2165 ino = inode->i_ino;
2166 if (filldir(dirent, ".", 1, pos, ino, DT_DIR) < 0)
2167 goto out;
2168 pos++;
2169 /* fall through */
2170 case 1:
2171 ino = parent_ino(dentry);
2172 if (filldir(dirent, "..", 2, pos, ino, DT_DIR) < 0)
2173 goto out;
2174 pos++;
2175 /* fall through */
2176 }
2177
2178 /* f_version caches the tgid value that the last readdir call couldn't
2179 * return. lseek aka telldir automagically resets f_version to 0.
2180 */
2181 tid = filp->f_version;
2182 filp->f_version = 0;
2183 for (task = first_tid(leader, tid, pos - 2);
2184 task;
2185 task = next_tid(task), pos++) {
2186 int len;
2187 tid = task->pid;
2188 len = snprintf(buf, sizeof(buf), "%d", tid);
2189 ino = fake_ino(tid, PROC_TID_INO);
2190 if (filldir(dirent, buf, len, pos, ino, DT_DIR < 0)) {
2191 /* returning this tgid failed, save it as the first
2192 * pid for the next readir call */
2193 filp->f_version = tid;
2194 put_task_struct(task);
2195 break;
2196 }
2197 }
2198 out:
2199 filp->f_pos = pos;
2200 put_task_struct(leader);
2201 out_no_task:
2202 return retval;
2203 }
2204
2205 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
2206 {
2207 struct inode *inode = dentry->d_inode;
2208 struct task_struct *p = get_proc_task(inode);
2209 generic_fillattr(inode, stat);
2210
2211 if (p) {
2212 rcu_read_lock();
2213 stat->nlink += get_nr_threads(p);
2214 rcu_read_unlock();
2215 put_task_struct(p);
2216 }
2217
2218 return 0;
2219 }
2220
2221 static struct inode_operations proc_task_inode_operations = {
2222 .lookup = proc_task_lookup,
2223 .getattr = proc_task_getattr,
2224 .setattr = proc_setattr,
2225 };
2226
2227 static struct file_operations proc_task_operations = {
2228 .read = generic_read_dir,
2229 .readdir = proc_task_readdir,
2230 };
ubuntu-zesty-kernel mirror