· 9 years ago · Jul 17, 2017, 06:10 PM
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/task_io_accounting_ops.h>
57#include <linux/init.h>
58#include <linux/capability.h>
59#include <linux/file.h>
60#include <linux/fdtable.h>
61#include <linux/string.h>
62#include <linux/seq_file.h>
63#include <linux/namei.h>
64#include <linux/mnt_namespace.h>
65#include <linux/mm.h>
66#include <linux/swap.h>
67#include <linux/rcupdate.h>
68#include <linux/kallsyms.h>
69#include <linux/stacktrace.h>
70#include <linux/resource.h>
71#include <linux/module.h>
72#include <linux/mount.h>
73#include <linux/security.h>
74#include <linux/ptrace.h>
75#include <linux/tracehook.h>
76#include <linux/printk.h>
77#include <linux/cgroup.h>
78#include <linux/cpuset.h>
79#include <linux/audit.h>
80#include <linux/poll.h>
81#include <linux/nsproxy.h>
82#include <linux/oom.h>
83#include <linux/elf.h>
84#include <linux/pid_namespace.h>
85#include <linux/user_namespace.h>
86#include <linux/fs_struct.h>
87#include <linux/slab.h>
88#include <linux/flex_array.h>
89#include <linux/posix-timers.h>
90#ifdef CONFIG_HARDWALL
91#include <asm/hardwall.h>
92#endif
93#include <trace/events/oom.h>
94#include "internal.h"
95#include "fd.h"
96
97/* NOTE:
98 * Implementing inode permission operations in /proc is almost
99 * certainly an error. Permission checks need to happen during
100 * each system call not at open time. The reason is that most of
101 * what we wish to check for permissions in /proc varies at runtime.
102 *
103 * The classic example of a problem is opening file descriptors
104 * in /proc for a task before it execs a suid executable.
105 */
106
107struct pid_entry {
108 const char *name;
109 int len;
110 umode_t mode;
111 const struct inode_operations *iop;
112 const struct file_operations *fop;
113 union proc_op op;
114};
115
116#define NOD(NAME, MODE, IOP, FOP, OP) { \
117 .name = (NAME), \
118 .len = sizeof(NAME) - 1, \
119 .mode = MODE, \
120 .iop = IOP, \
121 .fop = FOP, \
122 .op = OP, \
123}
124
125#define DIR(NAME, MODE, iops, fops) \
126 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
127#define LNK(NAME, get_link) \
128 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
129 &proc_pid_link_inode_operations, NULL, \
130 { .proc_get_link = get_link } )
131#define REG(NAME, MODE, fops) \
132 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
133#define ONE(NAME, MODE, show) \
134 NOD(NAME, (S_IFREG|(MODE)), \
135 NULL, &proc_single_file_operations, \
136 { .proc_show = show } )
137
138/*
139 * Count the number of hardlinks for the pid_entry table, excluding the .
140 * and .. links.
141 */
142static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
143 unsigned int n)
144{
145 unsigned int i;
146 unsigned int count;
147
148 count = 0;
149 for (i = 0; i < n; ++i) {
150 if (S_ISDIR(entries[i].mode))
151 ++count;
152 }
153
154 return count;
155}
156
157static int get_task_root(struct task_struct *task, struct path *root)
158{
159 int result = -ENOENT;
160
161 task_lock(task);
162 if (task->fs) {
163 get_fs_root(task->fs, root);
164 result = 0;
165 }
166 task_unlock(task);
167 return result;
168}
169
170static int proc_cwd_link(struct dentry *dentry, struct path *path)
171{
172 struct task_struct *task = get_proc_task(dentry->d_inode);
173 int result = -ENOENT;
174
175 if (task) {
176 task_lock(task);
177 if (task->fs) {
178 get_fs_pwd(task->fs, path);
179 result = 0;
180 }
181 task_unlock(task);
182 put_task_struct(task);
183 }
184 return result;
185}
186
187static int proc_root_link(struct dentry *dentry, struct path *path)
188{
189 struct task_struct *task = get_proc_task(dentry->d_inode);
190 int result = -ENOENT;
191
192 if (task) {
193 result = get_task_root(task, path);
194 put_task_struct(task);
195 }
196 return result;
197}
198
199static int proc_pid_cmdline(struct seq_file *m, struct pid_namespace *ns,
200 struct pid *pid, struct task_struct *task)
201{
202 /*
203 * Rely on struct seq_operations::show() being called once
204 * per internal buffer allocation. See single_open(), traverse().
205 */
206 BUG_ON(m->size < PAGE_SIZE);
207 m->count += get_cmdline(task, m->buf, PAGE_SIZE);
208 return 0;
209}
210
211static int proc_pid_auxv(struct seq_file *m, struct pid_namespace *ns,
212 struct pid *pid, struct task_struct *task)
213{
214 struct mm_struct *mm = mm_access(task, PTRACE_MODE_READ);
215 if (mm && !IS_ERR(mm)) {
216 unsigned int nwords = 0;
217 do {
218 nwords += 2;
219 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
220 seq_write(m, mm->saved_auxv, nwords * sizeof(mm->saved_auxv[0]));
221 mmput(mm);
222 return 0;
223 } else
224 return PTR_ERR(mm);
225}
226
227
228#ifdef CONFIG_KALLSYMS
229/*
230 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
231 * Returns the resolved symbol. If that fails, simply return the address.
232 */
233static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns,
234 struct pid *pid, struct task_struct *task)
235{
236 unsigned long wchan;
237 char symname[KSYM_NAME_LEN];
238
239 wchan = get_wchan(task);
240
241 if (lookup_symbol_name(wchan, symname) < 0)
242 if (!ptrace_may_access(task, PTRACE_MODE_READ))
243 return 0;
244 else
245 return seq_printf(m, "%lu", wchan);
246 else
247 return seq_printf(m, "%s", symname);
248}
249#endif /* CONFIG_KALLSYMS */
250
251static int lock_trace(struct task_struct *task)
252{
253 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
254 if (err)
255 return err;
256 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
257 mutex_unlock(&task->signal->cred_guard_mutex);
258 return -EPERM;
259 }
260 return 0;
261}
262
263static void unlock_trace(struct task_struct *task)
264{
265 mutex_unlock(&task->signal->cred_guard_mutex);
266}
267
268#ifdef CONFIG_STACKTRACE
269
270#define MAX_STACK_TRACE_DEPTH 64
271
272static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
273 struct pid *pid, struct task_struct *task)
274{
275 struct stack_trace trace;
276 unsigned long *entries;
277 int err;
278 int i;
279
280 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
281 if (!entries)
282 return -ENOMEM;
283
284 trace.nr_entries = 0;
285 trace.max_entries = MAX_STACK_TRACE_DEPTH;
286 trace.entries = entries;
287 trace.skip = 0;
288
289 err = lock_trace(task);
290 if (!err) {
291 save_stack_trace_tsk(task, &trace);
292
293 for (i = 0; i < trace.nr_entries; i++) {
294 seq_printf(m, "[<%pK>] %pS\n",
295 (void *)entries[i], (void *)entries[i]);
296 }
297 unlock_trace(task);
298 }
299 kfree(entries);
300
301 return err;
302}
303#endif
304
305#ifdef CONFIG_SCHEDSTATS
306/*
307 * Provides /proc/PID/schedstat
308 */
309static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
310 struct pid *pid, struct task_struct *task)
311{
312 return seq_printf(m, "%llu %llu %lu\n",
313 (unsigned long long)task->se.sum_exec_runtime,
314 (unsigned long long)task->sched_info.run_delay,
315 task->sched_info.pcount);
316}
317#endif
318
319#ifdef CONFIG_LATENCYTOP
320static int lstats_show_proc(struct seq_file *m, void *v)
321{
322 int i;
323 struct inode *inode = m->private;
324 struct task_struct *task = get_proc_task(inode);
325
326 if (!task)
327 return -ESRCH;
328 seq_puts(m, "Latency Top version : v0.1\n");
329 for (i = 0; i < 32; i++) {
330 struct latency_record *lr = &task->latency_record[i];
331 if (lr->backtrace[0]) {
332 int q;
333 seq_printf(m, "%i %li %li",
334 lr->count, lr->time, lr->max);
335 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
336 unsigned long bt = lr->backtrace[q];
337 if (!bt)
338 break;
339 if (bt == ULONG_MAX)
340 break;
341 seq_printf(m, " %ps", (void *)bt);
342 }
343 seq_putc(m, '\n');
344 }
345
346 }
347 put_task_struct(task);
348 return 0;
349}
350
351static int lstats_open(struct inode *inode, struct file *file)
352{
353 return single_open(file, lstats_show_proc, inode);
354}
355
356static ssize_t lstats_write(struct file *file, const char __user *buf,
357 size_t count, loff_t *offs)
358{
359 struct task_struct *task = get_proc_task(file_inode(file));
360
361 if (!task)
362 return -ESRCH;
363 clear_all_latency_tracing(task);
364 put_task_struct(task);
365
366 return count;
367}
368
369static const struct file_operations proc_lstats_operations = {
370 .open = lstats_open,
371 .read = seq_read,
372 .write = lstats_write,
373 .llseek = seq_lseek,
374 .release = single_release,
375};
376
377#endif
378
379static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
380 struct pid *pid, struct task_struct *task)
381{
382 unsigned long totalpages = totalram_pages + total_swap_pages;
383 unsigned long points = 0;
384
385 read_lock(&tasklist_lock);
386 if (pid_alive(task))
387 points = oom_badness(task, NULL, NULL, totalpages) *
388 1000 / totalpages;
389 read_unlock(&tasklist_lock);
390 return seq_printf(m, "%lu\n", points);
391}
392
393struct limit_names {
394 const char *name;
395 const char *unit;
396};
397
398static const struct limit_names lnames[RLIM_NLIMITS] = {
399 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
400 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
401 [RLIMIT_DATA] = {"Max data size", "bytes"},
402 [RLIMIT_STACK] = {"Max stack size", "bytes"},
403 [RLIMIT_CORE] = {"Max core file size", "bytes"},
404 [RLIMIT_RSS] = {"Max resident set", "bytes"},
405 [RLIMIT_NPROC] = {"Max processes", "processes"},
406 [RLIMIT_NOFILE] = {"Max open files", "files"},
407 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
408 [RLIMIT_AS] = {"Max address space", "bytes"},
409 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
410 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
411 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
412 [RLIMIT_NICE] = {"Max nice priority", NULL},
413 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
414 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
415};
416
417/* Display limits for a process */
418static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
419 struct pid *pid, struct task_struct *task)
420{
421 unsigned int i;
422 unsigned long flags;
423
424 struct rlimit rlim[RLIM_NLIMITS];
425
426 if (!lock_task_sighand(task, &flags))
427 return 0;
428 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
429 unlock_task_sighand(task, &flags);
430
431 /*
432 * print the file header
433 */
434 seq_printf(m, "%-25s %-20s %-20s %-10s\n",
435 "Limit", "Soft Limit", "Hard Limit", "Units");
436
437 for (i = 0; i < RLIM_NLIMITS; i++) {
438 if (rlim[i].rlim_cur == RLIM_INFINITY)
439 seq_printf(m, "%-25s %-20s ",
440 lnames[i].name, "unlimited");
441 else
442 seq_printf(m, "%-25s %-20lu ",
443 lnames[i].name, rlim[i].rlim_cur);
444
445 if (rlim[i].rlim_max == RLIM_INFINITY)
446 seq_printf(m, "%-20s ", "unlimited");
447 else
448 seq_printf(m, "%-20lu ", rlim[i].rlim_max);
449
450 if (lnames[i].unit)
451 seq_printf(m, "%-10s\n", lnames[i].unit);
452 else
453 seq_putc(m, '\n');
454 }
455
456 return 0;
457}
458
459#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
460static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
461 struct pid *pid, struct task_struct *task)
462{
463 long nr;
464 unsigned long args[6], sp, pc;
465 int res = lock_trace(task);
466 if (res)
467 return res;
468
469 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
470 seq_puts(m, "running\n");
471 else if (nr < 0)
472 seq_printf(m, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
473 else
474 seq_printf(m,
475 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
476 nr,
477 args[0], args[1], args[2], args[3], args[4], args[5],
478 sp, pc);
479 unlock_trace(task);
480 return res;
481}
482#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
483
484/************************************************************************/
485/* Here the fs part begins */
486/************************************************************************/
487
488/* permission checks */
489static int proc_fd_access_allowed(struct inode *inode)
490{
491 struct task_struct *task;
492 int allowed = 0;
493 /* Allow access to a task's file descriptors if it is us or we
494 * may use ptrace attach to the process and find out that
495 * information.
496 */
497 task = get_proc_task(inode);
498 if (task) {
499 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
500 put_task_struct(task);
501 }
502 return allowed;
503}
504
505int proc_setattr(struct dentry *dentry, struct iattr *attr)
506{
507 int error;
508 struct inode *inode = dentry->d_inode;
509
510 if (attr->ia_valid & ATTR_MODE)
511 return -EPERM;
512
513 error = inode_change_ok(inode, attr);
514 if (error)
515 return error;
516
517 setattr_copy(inode, attr);
518 mark_inode_dirty(inode);
519 return 0;
520}
521
522/*
523 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
524 * or euid/egid (for hide_pid_min=2)?
525 */
526static bool has_pid_permissions(struct pid_namespace *pid,
527 struct task_struct *task,
528 int hide_pid_min)
529{
530 if (pid->hide_pid < hide_pid_min)
531 return true;
532 if (in_group_p(pid->pid_gid))
533 return true;
534 return ptrace_may_access(task, PTRACE_MODE_READ);
535}
536
537
538static int proc_pid_permission(struct inode *inode, int mask)
539{
540 struct pid_namespace *pid = inode->i_sb->s_fs_info;
541 struct task_struct *task;
542 bool has_perms;
543
544 task = get_proc_task(inode);
545 if (!task)
546 return -ESRCH;
547 has_perms = has_pid_permissions(pid, task, 1);
548 put_task_struct(task);
549
550 if (!has_perms) {
551 if (pid->hide_pid == 2) {
552 /*
553 * Let's make getdents(), stat(), and open()
554 * consistent with each other. If a process
555 * may not stat() a file, it shouldn't be seen
556 * in procfs at all.
557 */
558 return -ENOENT;
559 }
560
561 return -EPERM;
562 }
563 return generic_permission(inode, mask);
564}
565
566
567
568static const struct inode_operations proc_def_inode_operations = {
569 .setattr = proc_setattr,
570};
571
572static int proc_single_show(struct seq_file *m, void *v)
573{
574 struct inode *inode = m->private;
575 struct pid_namespace *ns;
576 struct pid *pid;
577 struct task_struct *task;
578 int ret;
579
580 ns = inode->i_sb->s_fs_info;
581 pid = proc_pid(inode);
582 task = get_pid_task(pid, PIDTYPE_PID);
583 if (!task)
584 return -ESRCH;
585
586 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
587
588 put_task_struct(task);
589 return ret;
590}
591
592static int proc_single_open(struct inode *inode, struct file *filp)
593{
594 return single_open(filp, proc_single_show, inode);
595}
596
597static const struct file_operations proc_single_file_operations = {
598 .open = proc_single_open,
599 .read = seq_read,
600 .llseek = seq_lseek,
601 .release = single_release,
602};
603
604
605struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
606{
607 struct task_struct *task = get_proc_task(inode);
608 struct mm_struct *mm = ERR_PTR(-ESRCH);
609
610 if (task) {
611 mm = mm_access(task, mode);
612 put_task_struct(task);
613
614 if (!IS_ERR_OR_NULL(mm)) {
615 /* ensure this mm_struct can't be freed */
616 atomic_inc(&mm->mm_count);
617 /* but do not pin its memory */
618 mmput(mm);
619 }
620 }
621
622 return mm;
623}
624
625static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
626{
627 struct mm_struct *mm = proc_mem_open(inode, mode);
628
629 if (IS_ERR(mm))
630 return PTR_ERR(mm);
631
632 file->private_data = mm;
633 return 0;
634}
635
636static int mem_open(struct inode *inode, struct file *file)
637{
638 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
639
640 /* OK to pass negative loff_t, we can catch out-of-range */
641 file->f_mode |= FMODE_UNSIGNED_OFFSET;
642
643 return ret;
644}
645
646static ssize_t mem_rw(struct file *file, char __user *buf,
647 size_t count, loff_t *ppos, int write)
648{
649 struct mm_struct *mm = file->private_data;
650 unsigned long addr = *ppos;
651 ssize_t copied;
652 char *page;
653
654 if (!mm)
655 return 0;
656
657 page = (char *)__get_free_page(GFP_TEMPORARY);
658 if (!page)
659 return -ENOMEM;
660
661 copied = 0;
662 if (!atomic_inc_not_zero(&mm->mm_users))
663 goto free;
664
665 while (count > 0) {
666 int this_len = min_t(int, count, PAGE_SIZE);
667
668 if (write && copy_from_user(page, buf, this_len)) {
669 copied = -EFAULT;
670 break;
671 }
672
673 this_len = access_remote_vm(mm, addr, page, this_len, write);
674 if (!this_len) {
675 if (!copied)
676 copied = -EIO;
677 break;
678 }
679
680 if (!write && copy_to_user(buf, page, this_len)) {
681 copied = -EFAULT;
682 break;
683 }
684
685 buf += this_len;
686 addr += this_len;
687 copied += this_len;
688 count -= this_len;
689 }
690 *ppos = addr;
691
692 mmput(mm);
693free:
694 free_page((unsigned long) page);
695 return copied;
696}
697
698static ssize_t mem_read(struct file *file, char __user *buf,
699 size_t count, loff_t *ppos)
700{
701 return mem_rw(file, buf, count, ppos, 0);
702}
703
704static ssize_t mem_write(struct file *file, const char __user *buf,
705 size_t count, loff_t *ppos)
706{
707 return mem_rw(file, (char __user*)buf, count, ppos, 1);
708}
709
710loff_t mem_lseek(struct file *file, loff_t offset, int orig)
711{
712 switch (orig) {
713 case 0:
714 file->f_pos = offset;
715 break;
716 case 1:
717 file->f_pos += offset;
718 break;
719 default:
720 return -EINVAL;
721 }
722 force_successful_syscall_return();
723 return file->f_pos;
724}
725
726static int mem_release(struct inode *inode, struct file *file)
727{
728 struct mm_struct *mm = file->private_data;
729 if (mm)
730 mmdrop(mm);
731 return 0;
732}
733
734static const struct file_operations proc_mem_operations = {
735 .llseek = mem_lseek,
736 .read = mem_read,
737 .write = mem_write,
738 .open = mem_open,
739 .release = mem_release,
740};
741
742static int environ_open(struct inode *inode, struct file *file)
743{
744 return __mem_open(inode, file, PTRACE_MODE_READ);
745}
746
747static ssize_t environ_read(struct file *file, char __user *buf,
748 size_t count, loff_t *ppos)
749{
750 char *page;
751 unsigned long src = *ppos;
752 int ret = 0;
753 struct mm_struct *mm = file->private_data;
754
755 if (!mm)
756 return 0;
757
758 page = (char *)__get_free_page(GFP_TEMPORARY);
759 if (!page)
760 return -ENOMEM;
761
762 ret = 0;
763 if (!atomic_inc_not_zero(&mm->mm_users))
764 goto free;
765 while (count > 0) {
766 size_t this_len, max_len;
767 int retval;
768
769 if (src >= (mm->env_end - mm->env_start))
770 break;
771
772 this_len = mm->env_end - (mm->env_start + src);
773
774 max_len = min_t(size_t, PAGE_SIZE, count);
775 this_len = min(max_len, this_len);
776
777 retval = access_remote_vm(mm, (mm->env_start + src),
778 page, this_len, 0);
779
780 if (retval <= 0) {
781 ret = retval;
782 break;
783 }
784
785 if (copy_to_user(buf, page, retval)) {
786 ret = -EFAULT;
787 break;
788 }
789
790 ret += retval;
791 src += retval;
792 buf += retval;
793 count -= retval;
794 }
795 *ppos = src;
796 mmput(mm);
797
798free:
799 free_page((unsigned long) page);
800 return ret;
801}
802
803static const struct file_operations proc_environ_operations = {
804 .open = environ_open,
805 .read = environ_read,
806 .llseek = generic_file_llseek,
807 .release = mem_release,
808};
809
810static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
811 loff_t *ppos)
812{
813 struct task_struct *task = get_proc_task(file_inode(file));
814 char buffer[PROC_NUMBUF];
815 int oom_adj = OOM_ADJUST_MIN;
816 size_t len;
817 unsigned long flags;
818 int mult = 1;
819
820 if (!task)
821 return -ESRCH;
822 if (lock_task_sighand(task, &flags)) {
823 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX) {
824 oom_adj = OOM_ADJUST_MAX;
825 } else {
826 if (task->signal->oom_score_adj < 0)
827 mult = -1;
828 oom_adj = roundup(mult * task->signal->oom_score_adj *
829 -OOM_DISABLE, OOM_SCORE_ADJ_MAX) /
830 OOM_SCORE_ADJ_MAX * mult;
831 }
832 unlock_task_sighand(task, &flags);
833 }
834 put_task_struct(task);
835 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
836 return simple_read_from_buffer(buf, count, ppos, buffer, len);
837}
838
839static ssize_t oom_adj_write(struct file *file, const char __user *buf,
840 size_t count, loff_t *ppos)
841{
842 struct task_struct *task;
843 char buffer[PROC_NUMBUF];
844 int oom_adj;
845 unsigned long flags;
846 int err;
847
848 memset(buffer, 0, sizeof(buffer));
849 if (count > sizeof(buffer) - 1)
850 count = sizeof(buffer) - 1;
851 if (copy_from_user(buffer, buf, count)) {
852 err = -EFAULT;
853 goto out;
854 }
855
856 err = kstrtoint(strstrip(buffer), 0, &oom_adj);
857 if (err)
858 goto out;
859 if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
860 oom_adj != OOM_DISABLE) {
861 err = -EINVAL;
862 goto out;
863 }
864
865 task = get_proc_task(file_inode(file));
866 if (!task) {
867 err = -ESRCH;
868 goto out;
869 }
870
871 task_lock(task);
872 if (!task->mm) {
873 err = -EINVAL;
874 goto err_task_lock;
875 }
876
877 if (!lock_task_sighand(task, &flags)) {
878 err = -ESRCH;
879 goto err_task_lock;
880 }
881
882 /*
883 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
884 * value is always attainable.
885 */
886 if (oom_adj == OOM_ADJUST_MAX)
887 oom_adj = OOM_SCORE_ADJ_MAX;
888 else
889 oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
890
891 if (oom_adj < task->signal->oom_score_adj &&
892 !capable(CAP_SYS_RESOURCE)) {
893 err = -EACCES;
894 goto err_sighand;
895 }
896
897 /*
898 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
899 * /proc/pid/oom_score_adj instead.
900 */
901 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
902 current->comm, task_pid_nr(current), task_pid_nr(task),
903 task_pid_nr(task));
904
905 task->signal->oom_score_adj = oom_adj;
906 trace_oom_score_adj_update(task);
907err_sighand:
908 unlock_task_sighand(task, &flags);
909err_task_lock:
910 task_unlock(task);
911 put_task_struct(task);
912out:
913 return err < 0 ? err : count;
914}
915
916static const struct file_operations proc_oom_adj_operations = {
917 .read = oom_adj_read,
918 .write = oom_adj_write,
919 .llseek = generic_file_llseek,
920};
921
922static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
923 size_t count, loff_t *ppos)
924{
925 struct task_struct *task = get_proc_task(file_inode(file));
926 char buffer[PROC_NUMBUF];
927 short oom_score_adj = OOM_SCORE_ADJ_MIN;
928 unsigned long flags;
929 size_t len;
930
931 if (!task)
932 return -ESRCH;
933 if (lock_task_sighand(task, &flags)) {
934 oom_score_adj = task->signal->oom_score_adj;
935 unlock_task_sighand(task, &flags);
936 }
937 put_task_struct(task);
938 len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
939 return simple_read_from_buffer(buf, count, ppos, buffer, len);
940}
941
942static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
943 size_t count, loff_t *ppos)
944{
945 struct task_struct *task;
946 char buffer[PROC_NUMBUF];
947 unsigned long flags;
948 int oom_score_adj;
949 int err;
950
951 memset(buffer, 0, sizeof(buffer));
952 if (count > sizeof(buffer) - 1)
953 count = sizeof(buffer) - 1;
954 if (copy_from_user(buffer, buf, count)) {
955 err = -EFAULT;
956 goto out;
957 }
958
959 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
960 if (err)
961 goto out;
962 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
963 oom_score_adj > OOM_SCORE_ADJ_MAX) {
964 err = -EINVAL;
965 goto out;
966 }
967
968 task = get_proc_task(file_inode(file));
969 if (!task) {
970 err = -ESRCH;
971 goto out;
972 }
973
974 task_lock(task);
975 if (!task->mm) {
976 err = -EINVAL;
977 goto err_task_lock;
978 }
979
980 if (!lock_task_sighand(task, &flags)) {
981 err = -ESRCH;
982 goto err_task_lock;
983 }
984
985 if ((short)oom_score_adj < task->signal->oom_score_adj_min &&
986 !capable(CAP_SYS_RESOURCE)) {
987 err = -EACCES;
988 goto err_sighand;
989 }
990
991 task->signal->oom_score_adj = (short)oom_score_adj;
992 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
993 task->signal->oom_score_adj_min = (short)oom_score_adj;
994 trace_oom_score_adj_update(task);
995
996err_sighand:
997 unlock_task_sighand(task, &flags);
998err_task_lock:
999 task_unlock(task);
1000 put_task_struct(task);
1001out:
1002 return err < 0 ? err : count;
1003}
1004
1005static const struct file_operations proc_oom_score_adj_operations = {
1006 .read = oom_score_adj_read,
1007 .write = oom_score_adj_write,
1008 .llseek = default_llseek,
1009};
1010
1011#ifdef CONFIG_AUDITSYSCALL
1012#define TMPBUFLEN 21
1013static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1014 size_t count, loff_t *ppos)
1015{
1016 struct inode * inode = file_inode(file);
1017 struct task_struct *task = get_proc_task(inode);
1018 ssize_t length;
1019 char tmpbuf[TMPBUFLEN];
1020
1021 if (!task)
1022 return -ESRCH;
1023 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1024 from_kuid(file->f_cred->user_ns,
1025 audit_get_loginuid(task)));
1026 put_task_struct(task);
1027 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1028}
1029
1030static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1031 size_t count, loff_t *ppos)
1032{
1033 struct inode * inode = file_inode(file);
1034 char *page, *tmp;
1035 ssize_t length;
1036 uid_t loginuid;
1037 kuid_t kloginuid;
1038
1039 rcu_read_lock();
1040 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1041 rcu_read_unlock();
1042 return -EPERM;
1043 }
1044 rcu_read_unlock();
1045
1046 if (count >= PAGE_SIZE)
1047 count = PAGE_SIZE - 1;
1048
1049 if (*ppos != 0) {
1050 /* No partial writes. */
1051 return -EINVAL;
1052 }
1053 page = (char*)__get_free_page(GFP_TEMPORARY);
1054 if (!page)
1055 return -ENOMEM;
1056 length = -EFAULT;
1057 if (copy_from_user(page, buf, count))
1058 goto out_free_page;
1059
1060 page[count] = '\0';
1061 loginuid = simple_strtoul(page, &tmp, 10);
1062 if (tmp == page) {
1063 length = -EINVAL;
1064 goto out_free_page;
1065
1066 }
1067
1068 /* is userspace tring to explicitly UNSET the loginuid? */
1069 if (loginuid == AUDIT_UID_UNSET) {
1070 kloginuid = INVALID_UID;
1071 } else {
1072 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1073 if (!uid_valid(kloginuid)) {
1074 length = -EINVAL;
1075 goto out_free_page;
1076 }
1077 }
1078
1079 length = audit_set_loginuid(kloginuid);
1080 if (likely(length == 0))
1081 length = count;
1082
1083out_free_page:
1084 free_page((unsigned long) page);
1085 return length;
1086}
1087
1088static const struct file_operations proc_loginuid_operations = {
1089 .read = proc_loginuid_read,
1090 .write = proc_loginuid_write,
1091 .llseek = generic_file_llseek,
1092};
1093
1094static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1095 size_t count, loff_t *ppos)
1096{
1097 struct inode * inode = file_inode(file);
1098 struct task_struct *task = get_proc_task(inode);
1099 ssize_t length;
1100 char tmpbuf[TMPBUFLEN];
1101
1102 if (!task)
1103 return -ESRCH;
1104 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1105 audit_get_sessionid(task));
1106 put_task_struct(task);
1107 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1108}
1109
1110static const struct file_operations proc_sessionid_operations = {
1111 .read = proc_sessionid_read,
1112 .llseek = generic_file_llseek,
1113};
1114#endif
1115
1116#ifdef CONFIG_FAULT_INJECTION
1117static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1118 size_t count, loff_t *ppos)
1119{
1120 struct task_struct *task = get_proc_task(file_inode(file));
1121 char buffer[PROC_NUMBUF];
1122 size_t len;
1123 int make_it_fail;
1124
1125 if (!task)
1126 return -ESRCH;
1127 make_it_fail = task->make_it_fail;
1128 put_task_struct(task);
1129
1130 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1131
1132 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1133}
1134
1135static ssize_t proc_fault_inject_write(struct file * file,
1136 const char __user * buf, size_t count, loff_t *ppos)
1137{
1138 struct task_struct *task;
1139 char buffer[PROC_NUMBUF], *end;
1140 int make_it_fail;
1141
1142 if (!capable(CAP_SYS_RESOURCE))
1143 return -EPERM;
1144 memset(buffer, 0, sizeof(buffer));
1145 if (count > sizeof(buffer) - 1)
1146 count = sizeof(buffer) - 1;
1147 if (copy_from_user(buffer, buf, count))
1148 return -EFAULT;
1149 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1150 if (*end)
1151 return -EINVAL;
1152 if (make_it_fail < 0 || make_it_fail > 1)
1153 return -EINVAL;
1154
1155 task = get_proc_task(file_inode(file));
1156 if (!task)
1157 return -ESRCH;
1158 task->make_it_fail = make_it_fail;
1159 put_task_struct(task);
1160
1161 return count;
1162}
1163
1164static const struct file_operations proc_fault_inject_operations = {
1165 .read = proc_fault_inject_read,
1166 .write = proc_fault_inject_write,
1167 .llseek = generic_file_llseek,
1168};
1169#endif
1170
1171
1172#ifdef CONFIG_SCHED_DEBUG
1173/*
1174 * Print out various scheduling related per-task fields:
1175 */
1176static int sched_show(struct seq_file *m, void *v)
1177{
1178 struct inode *inode = m->private;
1179 struct task_struct *p;
1180
1181 p = get_proc_task(inode);
1182 if (!p)
1183 return -ESRCH;
1184 proc_sched_show_task(p, m);
1185
1186 put_task_struct(p);
1187
1188 return 0;
1189}
1190
1191static ssize_t
1192sched_write(struct file *file, const char __user *buf,
1193 size_t count, loff_t *offset)
1194{
1195 struct inode *inode = file_inode(file);
1196 struct task_struct *p;
1197
1198 p = get_proc_task(inode);
1199 if (!p)
1200 return -ESRCH;
1201 proc_sched_set_task(p);
1202
1203 put_task_struct(p);
1204
1205 return count;
1206}
1207
1208static int sched_open(struct inode *inode, struct file *filp)
1209{
1210 return single_open(filp, sched_show, inode);
1211}
1212
1213static const struct file_operations proc_pid_sched_operations = {
1214 .open = sched_open,
1215 .read = seq_read,
1216 .write = sched_write,
1217 .llseek = seq_lseek,
1218 .release = single_release,
1219};
1220
1221#endif
1222
1223/*
1224 * Print out various scheduling related per-task fields:
1225 */
1226
1227#ifdef CONFIG_SMP
1228
1229static int sched_wake_up_idle_show(struct seq_file *m, void *v)
1230{
1231 struct inode *inode = m->private;
1232 struct task_struct *p;
1233
1234 p = get_proc_task(inode);
1235 if (!p)
1236 return -ESRCH;
1237
1238 seq_printf(m, "%d\n", sched_get_wake_up_idle(p));
1239
1240 put_task_struct(p);
1241
1242 return 0;
1243}
1244
1245static ssize_t
1246sched_wake_up_idle_write(struct file *file, const char __user *buf,
1247 size_t count, loff_t *offset)
1248{
1249 struct inode *inode = file_inode(file);
1250 struct task_struct *p;
1251 char buffer[PROC_NUMBUF];
1252 int wake_up_idle, err;
1253
1254 memset(buffer, 0, sizeof(buffer));
1255 if (count > sizeof(buffer) - 1)
1256 count = sizeof(buffer) - 1;
1257 if (copy_from_user(buffer, buf, count)) {
1258 err = -EFAULT;
1259 goto out;
1260 }
1261
1262 err = kstrtoint(strstrip(buffer), 0, &wake_up_idle);
1263 if (err)
1264 goto out;
1265
1266 p = get_proc_task(inode);
1267 if (!p)
1268 return -ESRCH;
1269
1270 err = sched_set_wake_up_idle(p, wake_up_idle);
1271
1272 put_task_struct(p);
1273
1274out:
1275 return err < 0 ? err : count;
1276}
1277
1278static int sched_wake_up_idle_open(struct inode *inode, struct file *filp)
1279{
1280 return single_open(filp, sched_wake_up_idle_show, inode);
1281}
1282
1283static const struct file_operations proc_pid_sched_wake_up_idle_operations = {
1284 .open = sched_wake_up_idle_open,
1285 .read = seq_read,
1286 .write = sched_wake_up_idle_write,
1287 .llseek = seq_lseek,
1288 .release = single_release,
1289};
1290
1291#endif /* CONFIG_SMP */
1292
1293#ifdef CONFIG_SCHED_HMP
1294
1295static int sched_init_task_load_show(struct seq_file *m, void *v)
1296{
1297 struct inode *inode = m->private;
1298 struct task_struct *p;
1299
1300 p = get_proc_task(inode);
1301 if (!p)
1302 return -ESRCH;
1303
1304 seq_printf(m, "%d\n", sched_get_init_task_load(p));
1305
1306 put_task_struct(p);
1307
1308 return 0;
1309}
1310
1311static ssize_t
1312sched_init_task_load_write(struct file *file, const char __user *buf,
1313 size_t count, loff_t *offset)
1314{
1315 struct inode *inode = file_inode(file);
1316 struct task_struct *p;
1317 char buffer[PROC_NUMBUF];
1318 int init_task_load, err;
1319
1320 memset(buffer, 0, sizeof(buffer));
1321 if (count > sizeof(buffer) - 1)
1322 count = sizeof(buffer) - 1;
1323 if (copy_from_user(buffer, buf, count)) {
1324 err = -EFAULT;
1325 goto out;
1326 }
1327
1328 err = kstrtoint(strstrip(buffer), 0, &init_task_load);
1329 if (err)
1330 goto out;
1331
1332 p = get_proc_task(inode);
1333 if (!p)
1334 return -ESRCH;
1335
1336 err = sched_set_init_task_load(p, init_task_load);
1337
1338 put_task_struct(p);
1339
1340out:
1341 return err < 0 ? err : count;
1342}
1343
1344static int sched_init_task_load_open(struct inode *inode, struct file *filp)
1345{
1346 return single_open(filp, sched_init_task_load_show, inode);
1347}
1348
1349static const struct file_operations proc_pid_sched_init_task_load_operations = {
1350 .open = sched_init_task_load_open,
1351 .read = seq_read,
1352 .write = sched_init_task_load_write,
1353 .llseek = seq_lseek,
1354 .release = single_release,
1355};
1356
1357#ifndef CONFIG_SCHED_QHMP
1358static int sched_group_id_show(struct seq_file *m, void *v)
1359{
1360 struct inode *inode = m->private;
1361 struct task_struct *p;
1362
1363 p = get_proc_task(inode);
1364 if (!p)
1365 return -ESRCH;
1366
1367 seq_printf(m, "%d\n", sched_get_group_id(p));
1368
1369 put_task_struct(p);
1370
1371 return 0;
1372}
1373
1374static ssize_t
1375sched_group_id_write(struct file *file, const char __user *buf,
1376 size_t count, loff_t *offset)
1377{
1378 struct inode *inode = file_inode(file);
1379 struct task_struct *p;
1380 char buffer[PROC_NUMBUF];
1381 int group_id, err;
1382
1383 memset(buffer, 0, sizeof(buffer));
1384 if (count > sizeof(buffer) - 1)
1385 count = sizeof(buffer) - 1;
1386 if (copy_from_user(buffer, buf, count)) {
1387 err = -EFAULT;
1388 goto out;
1389 }
1390
1391 err = kstrtoint(strstrip(buffer), 0, &group_id);
1392 if (err)
1393 goto out;
1394
1395 p = get_proc_task(inode);
1396 if (!p)
1397 return -ESRCH;
1398
1399 err = sched_set_group_id(p, group_id);
1400
1401 put_task_struct(p);
1402
1403out:
1404 return err < 0 ? err : count;
1405}
1406
1407static int sched_group_id_open(struct inode *inode, struct file *filp)
1408{
1409 return single_open(filp, sched_group_id_show, inode);
1410}
1411
1412static const struct file_operations proc_pid_sched_group_id_operations = {
1413 .open = sched_group_id_open,
1414 .read = seq_read,
1415 .write = sched_group_id_write,
1416 .llseek = seq_lseek,
1417 .release = single_release,
1418};
1419#endif /* !CONFIG_SCHED_QHMP */
1420#endif /* CONFIG_SCHED_HMP */
1421
1422#ifdef CONFIG_SCHED_AUTOGROUP
1423/*
1424 * Print out autogroup related information:
1425 */
1426static int sched_autogroup_show(struct seq_file *m, void *v)
1427{
1428 struct inode *inode = m->private;
1429 struct task_struct *p;
1430
1431 p = get_proc_task(inode);
1432 if (!p)
1433 return -ESRCH;
1434 proc_sched_autogroup_show_task(p, m);
1435
1436 put_task_struct(p);
1437
1438 return 0;
1439}
1440
1441static ssize_t
1442sched_autogroup_write(struct file *file, const char __user *buf,
1443 size_t count, loff_t *offset)
1444{
1445 struct inode *inode = file_inode(file);
1446 struct task_struct *p;
1447 char buffer[PROC_NUMBUF];
1448 int nice;
1449 int err;
1450
1451 memset(buffer, 0, sizeof(buffer));
1452 if (count > sizeof(buffer) - 1)
1453 count = sizeof(buffer) - 1;
1454 if (copy_from_user(buffer, buf, count))
1455 return -EFAULT;
1456
1457 err = kstrtoint(strstrip(buffer), 0, &nice);
1458 if (err < 0)
1459 return err;
1460
1461 p = get_proc_task(inode);
1462 if (!p)
1463 return -ESRCH;
1464
1465 err = proc_sched_autogroup_set_nice(p, nice);
1466 if (err)
1467 count = err;
1468
1469 put_task_struct(p);
1470
1471 return count;
1472}
1473
1474static int sched_autogroup_open(struct inode *inode, struct file *filp)
1475{
1476 int ret;
1477
1478 ret = single_open(filp, sched_autogroup_show, NULL);
1479 if (!ret) {
1480 struct seq_file *m = filp->private_data;
1481
1482 m->private = inode;
1483 }
1484 return ret;
1485}
1486
1487static const struct file_operations proc_pid_sched_autogroup_operations = {
1488 .open = sched_autogroup_open,
1489 .read = seq_read,
1490 .write = sched_autogroup_write,
1491 .llseek = seq_lseek,
1492 .release = single_release,
1493};
1494
1495#endif /* CONFIG_SCHED_AUTOGROUP */
1496
1497static ssize_t comm_write(struct file *file, const char __user *buf,
1498 size_t count, loff_t *offset)
1499{
1500 struct inode *inode = file_inode(file);
1501 struct task_struct *p;
1502 char buffer[TASK_COMM_LEN];
1503 const size_t maxlen = sizeof(buffer) - 1;
1504
1505 memset(buffer, 0, sizeof(buffer));
1506 if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
1507 return -EFAULT;
1508
1509 p = get_proc_task(inode);
1510 if (!p)
1511 return -ESRCH;
1512
1513 if (same_thread_group(current, p))
1514 set_task_comm(p, buffer);
1515 else
1516 count = -EINVAL;
1517
1518 put_task_struct(p);
1519
1520 return count;
1521}
1522
1523static int comm_show(struct seq_file *m, void *v)
1524{
1525 struct inode *inode = m->private;
1526 struct task_struct *p;
1527
1528 p = get_proc_task(inode);
1529 if (!p)
1530 return -ESRCH;
1531
1532 task_lock(p);
1533 seq_printf(m, "%s\n", p->comm);
1534 task_unlock(p);
1535
1536 put_task_struct(p);
1537
1538 return 0;
1539}
1540
1541static int comm_open(struct inode *inode, struct file *filp)
1542{
1543 return single_open(filp, comm_show, inode);
1544}
1545
1546static const struct file_operations proc_pid_set_comm_operations = {
1547 .open = comm_open,
1548 .read = seq_read,
1549 .write = comm_write,
1550 .llseek = seq_lseek,
1551 .release = single_release,
1552};
1553
1554static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1555{
1556 struct task_struct *task;
1557 struct mm_struct *mm;
1558 struct file *exe_file;
1559
1560 task = get_proc_task(dentry->d_inode);
1561 if (!task)
1562 return -ENOENT;
1563 mm = get_task_mm(task);
1564 put_task_struct(task);
1565 if (!mm)
1566 return -ENOENT;
1567 exe_file = get_mm_exe_file(mm);
1568 mmput(mm);
1569 if (exe_file) {
1570 *exe_path = exe_file->f_path;
1571 path_get(&exe_file->f_path);
1572 fput(exe_file);
1573 return 0;
1574 } else
1575 return -ENOENT;
1576}
1577
1578static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1579{
1580 struct inode *inode = dentry->d_inode;
1581 struct path path;
1582 int error = -EACCES;
1583
1584 /* Are we allowed to snoop on the tasks file descriptors? */
1585 if (!proc_fd_access_allowed(inode))
1586 goto out;
1587
1588 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1589 if (error)
1590 goto out;
1591
1592 nd_jump_link(nd, &path);
1593 return NULL;
1594out:
1595 return ERR_PTR(error);
1596}
1597
1598static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1599{
1600 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1601 char *pathname;
1602 int len;
1603
1604 if (!tmp)
1605 return -ENOMEM;
1606
1607 pathname = d_path(path, tmp, PAGE_SIZE);
1608 len = PTR_ERR(pathname);
1609 if (IS_ERR(pathname))
1610 goto out;
1611 len = tmp + PAGE_SIZE - 1 - pathname;
1612
1613 if (len > buflen)
1614 len = buflen;
1615 if (copy_to_user(buffer, pathname, len))
1616 len = -EFAULT;
1617 out:
1618 free_page((unsigned long)tmp);
1619 return len;
1620}
1621
1622static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1623{
1624 int error = -EACCES;
1625 struct inode *inode = dentry->d_inode;
1626 struct path path;
1627
1628 /* Are we allowed to snoop on the tasks file descriptors? */
1629 if (!proc_fd_access_allowed(inode))
1630 goto out;
1631
1632 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1633 if (error)
1634 goto out;
1635
1636 error = do_proc_readlink(&path, buffer, buflen);
1637 path_put(&path);
1638out:
1639 return error;
1640}
1641
1642const struct inode_operations proc_pid_link_inode_operations = {
1643 .readlink = proc_pid_readlink,
1644 .follow_link = proc_pid_follow_link,
1645 .setattr = proc_setattr,
1646};
1647
1648
1649/* building an inode */
1650
1651struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1652{
1653 struct inode * inode;
1654 struct proc_inode *ei;
1655 const struct cred *cred;
1656
1657 /* We need a new inode */
1658
1659 inode = new_inode(sb);
1660 if (!inode)
1661 goto out;
1662
1663 /* Common stuff */
1664 ei = PROC_I(inode);
1665 inode->i_ino = get_next_ino();
1666 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1667 inode->i_op = &proc_def_inode_operations;
1668
1669 /*
1670 * grab the reference to task.
1671 */
1672 ei->pid = get_task_pid(task, PIDTYPE_PID);
1673 if (!ei->pid)
1674 goto out_unlock;
1675
1676 if (task_dumpable(task)) {
1677 rcu_read_lock();
1678 cred = __task_cred(task);
1679 inode->i_uid = cred->euid;
1680 inode->i_gid = cred->egid;
1681 rcu_read_unlock();
1682 }
1683 security_task_to_inode(task, inode);
1684
1685out:
1686 return inode;
1687
1688out_unlock:
1689 iput(inode);
1690 return NULL;
1691}
1692
1693int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1694{
1695 struct inode *inode = dentry->d_inode;
1696 struct task_struct *task;
1697 const struct cred *cred;
1698 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
1699
1700 generic_fillattr(inode, stat);
1701
1702 rcu_read_lock();
1703 stat->uid = GLOBAL_ROOT_UID;
1704 stat->gid = GLOBAL_ROOT_GID;
1705 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1706 if (task) {
1707 if (!has_pid_permissions(pid, task, 2)) {
1708 rcu_read_unlock();
1709 /*
1710 * This doesn't prevent learning whether PID exists,
1711 * it only makes getattr() consistent with readdir().
1712 */
1713 return -ENOENT;
1714 }
1715 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1716 task_dumpable(task)) {
1717 cred = __task_cred(task);
1718 stat->uid = cred->euid;
1719 stat->gid = cred->egid;
1720 }
1721 }
1722 rcu_read_unlock();
1723 return 0;
1724}
1725
1726/* dentry stuff */
1727
1728/*
1729 * Exceptional case: normally we are not allowed to unhash a busy
1730 * directory. In this case, however, we can do it - no aliasing problems
1731 * due to the way we treat inodes.
1732 *
1733 * Rewrite the inode's ownerships here because the owning task may have
1734 * performed a setuid(), etc.
1735 *
1736 * Before the /proc/pid/status file was created the only way to read
1737 * the effective uid of a /process was to stat /proc/pid. Reading
1738 * /proc/pid/status is slow enough that procps and other packages
1739 * kept stating /proc/pid. To keep the rules in /proc simple I have
1740 * made this apply to all per process world readable and executable
1741 * directories.
1742 */
1743int pid_revalidate(struct dentry *dentry, unsigned int flags)
1744{
1745 struct inode *inode;
1746 struct task_struct *task;
1747 const struct cred *cred;
1748
1749 if (flags & LOOKUP_RCU)
1750 return -ECHILD;
1751
1752 inode = dentry->d_inode;
1753 task = get_proc_task(inode);
1754
1755 if (task) {
1756 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1757 task_dumpable(task)) {
1758 rcu_read_lock();
1759 cred = __task_cred(task);
1760 inode->i_uid = cred->euid;
1761 inode->i_gid = cred->egid;
1762 rcu_read_unlock();
1763 } else {
1764 inode->i_uid = GLOBAL_ROOT_UID;
1765 inode->i_gid = GLOBAL_ROOT_GID;
1766 }
1767 inode->i_mode &= ~(S_ISUID | S_ISGID);
1768 security_task_to_inode(task, inode);
1769 put_task_struct(task);
1770 return 1;
1771 }
1772 return 0;
1773}
1774
1775static inline bool proc_inode_is_dead(struct inode *inode)
1776{
1777 return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
1778}
1779
1780int pid_delete_dentry(const struct dentry *dentry)
1781{
1782 /* Is the task we represent dead?
1783 * If so, then don't put the dentry on the lru list,
1784 * kill it immediately.
1785 */
1786 return proc_inode_is_dead(dentry->d_inode);
1787}
1788
1789const struct dentry_operations pid_dentry_operations =
1790{
1791 .d_revalidate = pid_revalidate,
1792 .d_delete = pid_delete_dentry,
1793};
1794
1795/* Lookups */
1796
1797/*
1798 * Fill a directory entry.
1799 *
1800 * If possible create the dcache entry and derive our inode number and
1801 * file type from dcache entry.
1802 *
1803 * Since all of the proc inode numbers are dynamically generated, the inode
1804 * numbers do not exist until the inode is cache. This means creating the
1805 * the dcache entry in readdir is necessary to keep the inode numbers
1806 * reported by readdir in sync with the inode numbers reported
1807 * by stat.
1808 */
1809bool proc_fill_cache(struct file *file, struct dir_context *ctx,
1810 const char *name, int len,
1811 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1812{
1813 struct dentry *child, *dir = file->f_path.dentry;
1814 struct qstr qname = QSTR_INIT(name, len);
1815 struct inode *inode;
1816 unsigned type;
1817 ino_t ino;
1818
1819 child = d_hash_and_lookup(dir, &qname);
1820 if (!child) {
1821 child = d_alloc(dir, &qname);
1822 if (!child)
1823 goto end_instantiate;
1824 if (instantiate(dir->d_inode, child, task, ptr) < 0) {
1825 dput(child);
1826 goto end_instantiate;
1827 }
1828 }
1829 inode = child->d_inode;
1830 ino = inode->i_ino;
1831 type = inode->i_mode >> 12;
1832 dput(child);
1833 return dir_emit(ctx, name, len, ino, type);
1834
1835end_instantiate:
1836 return dir_emit(ctx, name, len, 1, DT_UNKNOWN);
1837}
1838
1839#ifdef CONFIG_CHECKPOINT_RESTORE
1840
1841/*
1842 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1843 * which represent vma start and end addresses.
1844 */
1845static int dname_to_vma_addr(struct dentry *dentry,
1846 unsigned long *start, unsigned long *end)
1847{
1848 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
1849 return -EINVAL;
1850
1851 return 0;
1852}
1853
1854static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
1855{
1856 unsigned long vm_start, vm_end;
1857 bool exact_vma_exists = false;
1858 struct mm_struct *mm = NULL;
1859 struct task_struct *task;
1860 const struct cred *cred;
1861 struct inode *inode;
1862 int status = 0;
1863
1864 if (flags & LOOKUP_RCU)
1865 return -ECHILD;
1866
1867 if (!capable(CAP_SYS_ADMIN)) {
1868 status = -EPERM;
1869 goto out_notask;
1870 }
1871
1872 inode = dentry->d_inode;
1873 task = get_proc_task(inode);
1874 if (!task)
1875 goto out_notask;
1876
1877 mm = mm_access(task, PTRACE_MODE_READ);
1878 if (IS_ERR_OR_NULL(mm))
1879 goto out;
1880
1881 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1882 down_read(&mm->mmap_sem);
1883 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1884 up_read(&mm->mmap_sem);
1885 }
1886
1887 mmput(mm);
1888
1889 if (exact_vma_exists) {
1890 if (task_dumpable(task)) {
1891 rcu_read_lock();
1892 cred = __task_cred(task);
1893 inode->i_uid = cred->euid;
1894 inode->i_gid = cred->egid;
1895 rcu_read_unlock();
1896 } else {
1897 inode->i_uid = GLOBAL_ROOT_UID;
1898 inode->i_gid = GLOBAL_ROOT_GID;
1899 }
1900 security_task_to_inode(task, inode);
1901 status = 1;
1902 }
1903
1904out:
1905 put_task_struct(task);
1906
1907out_notask:
1908 return status;
1909}
1910
1911static const struct dentry_operations tid_map_files_dentry_operations = {
1912 .d_revalidate = map_files_d_revalidate,
1913 .d_delete = pid_delete_dentry,
1914};
1915
1916static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
1917{
1918 unsigned long vm_start, vm_end;
1919 struct vm_area_struct *vma;
1920 struct task_struct *task;
1921 struct mm_struct *mm;
1922 int rc;
1923
1924 rc = -ENOENT;
1925 task = get_proc_task(dentry->d_inode);
1926 if (!task)
1927 goto out;
1928
1929 mm = get_task_mm(task);
1930 put_task_struct(task);
1931 if (!mm)
1932 goto out;
1933
1934 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
1935 if (rc)
1936 goto out_mmput;
1937
1938 rc = -ENOENT;
1939 down_read(&mm->mmap_sem);
1940 vma = find_exact_vma(mm, vm_start, vm_end);
1941 if (vma && vma->vm_file) {
1942 *path = vma->vm_file->f_path;
1943 path_get(path);
1944 rc = 0;
1945 }
1946 up_read(&mm->mmap_sem);
1947
1948out_mmput:
1949 mmput(mm);
1950out:
1951 return rc;
1952}
1953
1954struct map_files_info {
1955 fmode_t mode;
1956 unsigned long len;
1957 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
1958};
1959
1960static int
1961proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
1962 struct task_struct *task, const void *ptr)
1963{
1964 fmode_t mode = (fmode_t)(unsigned long)ptr;
1965 struct proc_inode *ei;
1966 struct inode *inode;
1967
1968 inode = proc_pid_make_inode(dir->i_sb, task);
1969 if (!inode)
1970 return -ENOENT;
1971
1972 ei = PROC_I(inode);
1973 ei->op.proc_get_link = proc_map_files_get_link;
1974
1975 inode->i_op = &proc_pid_link_inode_operations;
1976 inode->i_size = 64;
1977 inode->i_mode = S_IFLNK;
1978
1979 if (mode & FMODE_READ)
1980 inode->i_mode |= S_IRUSR;
1981 if (mode & FMODE_WRITE)
1982 inode->i_mode |= S_IWUSR;
1983
1984 d_set_d_op(dentry, &tid_map_files_dentry_operations);
1985 d_add(dentry, inode);
1986
1987 return 0;
1988}
1989
1990static struct dentry *proc_map_files_lookup(struct inode *dir,
1991 struct dentry *dentry, unsigned int flags)
1992{
1993 unsigned long vm_start, vm_end;
1994 struct vm_area_struct *vma;
1995 struct task_struct *task;
1996 int result;
1997 struct mm_struct *mm;
1998
1999 result = -EPERM;
2000 if (!capable(CAP_SYS_ADMIN))
2001 goto out;
2002
2003 result = -ENOENT;
2004 task = get_proc_task(dir);
2005 if (!task)
2006 goto out;
2007
2008 result = -EACCES;
2009 if (!ptrace_may_access(task, PTRACE_MODE_READ))
2010 goto out_put_task;
2011
2012 result = -ENOENT;
2013 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2014 goto out_put_task;
2015
2016 mm = get_task_mm(task);
2017 if (!mm)
2018 goto out_put_task;
2019
2020 down_read(&mm->mmap_sem);
2021 vma = find_exact_vma(mm, vm_start, vm_end);
2022 if (!vma)
2023 goto out_no_vma;
2024
2025 if (vma->vm_file)
2026 result = proc_map_files_instantiate(dir, dentry, task,
2027 (void *)(unsigned long)vma->vm_file->f_mode);
2028
2029out_no_vma:
2030 up_read(&mm->mmap_sem);
2031 mmput(mm);
2032out_put_task:
2033 put_task_struct(task);
2034out:
2035 return ERR_PTR(result);
2036}
2037
2038static const struct inode_operations proc_map_files_inode_operations = {
2039 .lookup = proc_map_files_lookup,
2040 .permission = proc_fd_permission,
2041 .setattr = proc_setattr,
2042};
2043
2044static int
2045proc_map_files_readdir(struct file *file, struct dir_context *ctx)
2046{
2047 struct vm_area_struct *vma;
2048 struct task_struct *task;
2049 struct mm_struct *mm;
2050 unsigned long nr_files, pos, i;
2051 struct flex_array *fa = NULL;
2052 struct map_files_info info;
2053 struct map_files_info *p;
2054 int ret;
2055
2056 ret = -EPERM;
2057 if (!capable(CAP_SYS_ADMIN))
2058 goto out;
2059
2060 ret = -ENOENT;
2061 task = get_proc_task(file_inode(file));
2062 if (!task)
2063 goto out;
2064
2065 ret = -EACCES;
2066 if (!ptrace_may_access(task, PTRACE_MODE_READ))
2067 goto out_put_task;
2068
2069 ret = 0;
2070 if (!dir_emit_dots(file, ctx))
2071 goto out_put_task;
2072
2073 mm = get_task_mm(task);
2074 if (!mm)
2075 goto out_put_task;
2076 down_read(&mm->mmap_sem);
2077
2078 nr_files = 0;
2079
2080 /*
2081 * We need two passes here:
2082 *
2083 * 1) Collect vmas of mapped files with mmap_sem taken
2084 * 2) Release mmap_sem and instantiate entries
2085 *
2086 * otherwise we get lockdep complained, since filldir()
2087 * routine might require mmap_sem taken in might_fault().
2088 */
2089
2090 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2091 if (vma->vm_file && ++pos > ctx->pos)
2092 nr_files++;
2093 }
2094
2095 if (nr_files) {
2096 fa = flex_array_alloc(sizeof(info), nr_files,
2097 GFP_KERNEL);
2098 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2099 GFP_KERNEL)) {
2100 ret = -ENOMEM;
2101 if (fa)
2102 flex_array_free(fa);
2103 up_read(&mm->mmap_sem);
2104 mmput(mm);
2105 goto out_put_task;
2106 }
2107 for (i = 0, vma = mm->mmap, pos = 2; vma;
2108 vma = vma->vm_next) {
2109 if (!vma->vm_file)
2110 continue;
2111 if (++pos <= ctx->pos)
2112 continue;
2113
2114 info.mode = vma->vm_file->f_mode;
2115 info.len = snprintf(info.name,
2116 sizeof(info.name), "%lx-%lx",
2117 vma->vm_start, vma->vm_end);
2118 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2119 BUG();
2120 }
2121 }
2122 up_read(&mm->mmap_sem);
2123
2124 for (i = 0; i < nr_files; i++) {
2125 p = flex_array_get(fa, i);
2126 if (!proc_fill_cache(file, ctx,
2127 p->name, p->len,
2128 proc_map_files_instantiate,
2129 task,
2130 (void *)(unsigned long)p->mode))
2131 break;
2132 ctx->pos++;
2133 }
2134 if (fa)
2135 flex_array_free(fa);
2136 mmput(mm);
2137
2138out_put_task:
2139 put_task_struct(task);
2140out:
2141 return ret;
2142}
2143
2144static const struct file_operations proc_map_files_operations = {
2145 .read = generic_read_dir,
2146 .iterate = proc_map_files_readdir,
2147 .llseek = default_llseek,
2148};
2149
2150struct timers_private {
2151 struct pid *pid;
2152 struct task_struct *task;
2153 struct sighand_struct *sighand;
2154 struct pid_namespace *ns;
2155 unsigned long flags;
2156};
2157
2158static void *timers_start(struct seq_file *m, loff_t *pos)
2159{
2160 struct timers_private *tp = m->private;
2161
2162 tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2163 if (!tp->task)
2164 return ERR_PTR(-ESRCH);
2165
2166 tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2167 if (!tp->sighand)
2168 return ERR_PTR(-ESRCH);
2169
2170 return seq_list_start(&tp->task->signal->posix_timers, *pos);
2171}
2172
2173static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2174{
2175 struct timers_private *tp = m->private;
2176 return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2177}
2178
2179static void timers_stop(struct seq_file *m, void *v)
2180{
2181 struct timers_private *tp = m->private;
2182
2183 if (tp->sighand) {
2184 unlock_task_sighand(tp->task, &tp->flags);
2185 tp->sighand = NULL;
2186 }
2187
2188 if (tp->task) {
2189 put_task_struct(tp->task);
2190 tp->task = NULL;
2191 }
2192}
2193
2194static int show_timer(struct seq_file *m, void *v)
2195{
2196 struct k_itimer *timer;
2197 struct timers_private *tp = m->private;
2198 int notify;
2199 static const char * const nstr[] = {
2200 [SIGEV_SIGNAL] = "signal",
2201 [SIGEV_NONE] = "none",
2202 [SIGEV_THREAD] = "thread",
2203 };
2204
2205 timer = list_entry((struct list_head *)v, struct k_itimer, list);
2206 notify = timer->it_sigev_notify;
2207
2208 seq_printf(m, "ID: %d\n", timer->it_id);
2209 seq_printf(m, "signal: %d/%p\n", timer->sigq->info.si_signo,
2210 timer->sigq->info.si_value.sival_ptr);
2211 seq_printf(m, "notify: %s/%s.%d\n",
2212 nstr[notify & ~SIGEV_THREAD_ID],
2213 (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2214 pid_nr_ns(timer->it_pid, tp->ns));
2215 seq_printf(m, "ClockID: %d\n", timer->it_clock);
2216
2217 return 0;
2218}
2219
2220static const struct seq_operations proc_timers_seq_ops = {
2221 .start = timers_start,
2222 .next = timers_next,
2223 .stop = timers_stop,
2224 .show = show_timer,
2225};
2226
2227static int proc_timers_open(struct inode *inode, struct file *file)
2228{
2229 struct timers_private *tp;
2230
2231 tp = __seq_open_private(file, &proc_timers_seq_ops,
2232 sizeof(struct timers_private));
2233 if (!tp)
2234 return -ENOMEM;
2235
2236 tp->pid = proc_pid(inode);
2237 tp->ns = inode->i_sb->s_fs_info;
2238 return 0;
2239}
2240
2241static const struct file_operations proc_timers_operations = {
2242 .open = proc_timers_open,
2243 .read = seq_read,
2244 .llseek = seq_lseek,
2245 .release = seq_release_private,
2246};
2247#endif /* CONFIG_CHECKPOINT_RESTORE */
2248
2249static int proc_pident_instantiate(struct inode *dir,
2250 struct dentry *dentry, struct task_struct *task, const void *ptr)
2251{
2252 const struct pid_entry *p = ptr;
2253 struct inode *inode;
2254 struct proc_inode *ei;
2255
2256 inode = proc_pid_make_inode(dir->i_sb, task);
2257 if (!inode)
2258 goto out;
2259
2260 ei = PROC_I(inode);
2261 inode->i_mode = p->mode;
2262 if (S_ISDIR(inode->i_mode))
2263 set_nlink(inode, 2); /* Use getattr to fix if necessary */
2264 if (p->iop)
2265 inode->i_op = p->iop;
2266 if (p->fop)
2267 inode->i_fop = p->fop;
2268 ei->op = p->op;
2269 d_set_d_op(dentry, &pid_dentry_operations);
2270 d_add(dentry, inode);
2271 /* Close the race of the process dying before we return the dentry */
2272 if (pid_revalidate(dentry, 0))
2273 return 0;
2274out:
2275 return -ENOENT;
2276}
2277
2278static struct dentry *proc_pident_lookup(struct inode *dir,
2279 struct dentry *dentry,
2280 const struct pid_entry *ents,
2281 unsigned int nents)
2282{
2283 int error;
2284 struct task_struct *task = get_proc_task(dir);
2285 const struct pid_entry *p, *last;
2286
2287 error = -ENOENT;
2288
2289 if (!task)
2290 goto out_no_task;
2291
2292 /*
2293 * Yes, it does not scale. And it should not. Don't add
2294 * new entries into /proc/<tgid>/ without very good reasons.
2295 */
2296 last = &ents[nents - 1];
2297 for (p = ents; p <= last; p++) {
2298 if (p->len != dentry->d_name.len)
2299 continue;
2300 if (!memcmp(dentry->d_name.name, p->name, p->len))
2301 break;
2302 }
2303 if (p > last)
2304 goto out;
2305
2306 error = proc_pident_instantiate(dir, dentry, task, p);
2307out:
2308 put_task_struct(task);
2309out_no_task:
2310 return ERR_PTR(error);
2311}
2312
2313static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
2314 const struct pid_entry *ents, unsigned int nents)
2315{
2316 struct task_struct *task = get_proc_task(file_inode(file));
2317 const struct pid_entry *p;
2318
2319 if (!task)
2320 return -ENOENT;
2321
2322 if (!dir_emit_dots(file, ctx))
2323 goto out;
2324
2325 if (ctx->pos >= nents + 2)
2326 goto out;
2327
2328 for (p = ents + (ctx->pos - 2); p <= ents + nents - 1; p++) {
2329 if (!proc_fill_cache(file, ctx, p->name, p->len,
2330 proc_pident_instantiate, task, p))
2331 break;
2332 ctx->pos++;
2333 }
2334out:
2335 put_task_struct(task);
2336 return 0;
2337}
2338
2339#ifdef CONFIG_SECURITY
2340static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2341 size_t count, loff_t *ppos)
2342{
2343 struct inode * inode = file_inode(file);
2344 char *p = NULL;
2345 ssize_t length;
2346 struct task_struct *task = get_proc_task(inode);
2347
2348 if (!task)
2349 return -ESRCH;
2350
2351 length = security_getprocattr(task,
2352 (char*)file->f_path.dentry->d_name.name,
2353 &p);
2354 put_task_struct(task);
2355 if (length > 0)
2356 length = simple_read_from_buffer(buf, count, ppos, p, length);
2357 kfree(p);
2358 return length;
2359}
2360
2361static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2362 size_t count, loff_t *ppos)
2363{
2364 struct inode * inode = file_inode(file);
2365 char *page;
2366 ssize_t length;
2367 struct task_struct *task = get_proc_task(inode);
2368
2369 length = -ESRCH;
2370 if (!task)
2371 goto out_no_task;
2372 if (count > PAGE_SIZE)
2373 count = PAGE_SIZE;
2374
2375 /* No partial writes. */
2376 length = -EINVAL;
2377 if (*ppos != 0)
2378 goto out;
2379
2380 length = -ENOMEM;
2381 page = (char*)__get_free_page(GFP_TEMPORARY);
2382 if (!page)
2383 goto out;
2384
2385 length = -EFAULT;
2386 if (copy_from_user(page, buf, count))
2387 goto out_free;
2388
2389 /* Guard against adverse ptrace interaction */
2390 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2391 if (length < 0)
2392 goto out_free;
2393
2394 length = security_setprocattr(task,
2395 (char*)file->f_path.dentry->d_name.name,
2396 (void*)page, count);
2397 mutex_unlock(&task->signal->cred_guard_mutex);
2398out_free:
2399 free_page((unsigned long) page);
2400out:
2401 put_task_struct(task);
2402out_no_task:
2403 return length;
2404}
2405
2406static const struct file_operations proc_pid_attr_operations = {
2407 .read = proc_pid_attr_read,
2408 .write = proc_pid_attr_write,
2409 .llseek = generic_file_llseek,
2410};
2411
2412static const struct pid_entry attr_dir_stuff[] = {
2413 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2414 REG("prev", S_IRUGO, proc_pid_attr_operations),
2415 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2416 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2417 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2418 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2419};
2420
2421static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
2422{
2423 return proc_pident_readdir(file, ctx,
2424 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2425}
2426
2427static const struct file_operations proc_attr_dir_operations = {
2428 .read = generic_read_dir,
2429 .iterate = proc_attr_dir_readdir,
2430 .llseek = default_llseek,
2431};
2432
2433static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2434 struct dentry *dentry, unsigned int flags)
2435{
2436 return proc_pident_lookup(dir, dentry,
2437 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2438}
2439
2440static const struct inode_operations proc_attr_dir_inode_operations = {
2441 .lookup = proc_attr_dir_lookup,
2442 .getattr = pid_getattr,
2443 .setattr = proc_setattr,
2444};
2445
2446#endif
2447
2448#ifdef CONFIG_ELF_CORE
2449static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2450 size_t count, loff_t *ppos)
2451{
2452 struct task_struct *task = get_proc_task(file_inode(file));
2453 struct mm_struct *mm;
2454 char buffer[PROC_NUMBUF];
2455 size_t len;
2456 int ret;
2457
2458 if (!task)
2459 return -ESRCH;
2460
2461 ret = 0;
2462 mm = get_task_mm(task);
2463 if (mm) {
2464 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2465 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2466 MMF_DUMP_FILTER_SHIFT));
2467 mmput(mm);
2468 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2469 }
2470
2471 put_task_struct(task);
2472
2473 return ret;
2474}
2475
2476static ssize_t proc_coredump_filter_write(struct file *file,
2477 const char __user *buf,
2478 size_t count,
2479 loff_t *ppos)
2480{
2481 struct task_struct *task;
2482 struct mm_struct *mm;
2483 char buffer[PROC_NUMBUF], *end;
2484 unsigned int val;
2485 int ret;
2486 int i;
2487 unsigned long mask;
2488
2489 ret = -EFAULT;
2490 memset(buffer, 0, sizeof(buffer));
2491 if (count > sizeof(buffer) - 1)
2492 count = sizeof(buffer) - 1;
2493 if (copy_from_user(buffer, buf, count))
2494 goto out_no_task;
2495
2496 ret = -EINVAL;
2497 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2498 if (*end == '\n')
2499 end++;
2500 if (end - buffer == 0)
2501 goto out_no_task;
2502
2503 ret = -ESRCH;
2504 task = get_proc_task(file_inode(file));
2505 if (!task)
2506 goto out_no_task;
2507
2508 ret = end - buffer;
2509 mm = get_task_mm(task);
2510 if (!mm)
2511 goto out_no_mm;
2512
2513 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2514 if (val & mask)
2515 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2516 else
2517 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2518 }
2519
2520 mmput(mm);
2521 out_no_mm:
2522 put_task_struct(task);
2523 out_no_task:
2524 return ret;
2525}
2526
2527static const struct file_operations proc_coredump_filter_operations = {
2528 .read = proc_coredump_filter_read,
2529 .write = proc_coredump_filter_write,
2530 .llseek = generic_file_llseek,
2531};
2532#endif
2533
2534#ifdef CONFIG_TASK_IO_ACCOUNTING
2535static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
2536{
2537 struct task_io_accounting acct = task->ioac;
2538 unsigned long flags;
2539 int result;
2540
2541 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2542 if (result)
2543 return result;
2544
2545 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2546 result = -EACCES;
2547 goto out_unlock;
2548 }
2549
2550 if (whole && lock_task_sighand(task, &flags)) {
2551 struct task_struct *t = task;
2552
2553 task_io_accounting_add(&acct, &task->signal->ioac);
2554 while_each_thread(task, t)
2555 task_io_accounting_add(&acct, &t->ioac);
2556
2557 unlock_task_sighand(task, &flags);
2558 }
2559 result = seq_printf(m,
2560 "rchar: %llu\n"
2561 "wchar: %llu\n"
2562 "syscr: %llu\n"
2563 "syscw: %llu\n"
2564 "read_bytes: %llu\n"
2565 "write_bytes: %llu\n"
2566 "cancelled_write_bytes: %llu\n",
2567 (unsigned long long)acct.rchar,
2568 (unsigned long long)acct.wchar,
2569 (unsigned long long)acct.syscr,
2570 (unsigned long long)acct.syscw,
2571 (unsigned long long)acct.read_bytes,
2572 (unsigned long long)acct.write_bytes,
2573 (unsigned long long)acct.cancelled_write_bytes);
2574out_unlock:
2575 mutex_unlock(&task->signal->cred_guard_mutex);
2576 return result;
2577}
2578
2579static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2580 struct pid *pid, struct task_struct *task)
2581{
2582 return do_io_accounting(task, m, 0);
2583}
2584
2585static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2586 struct pid *pid, struct task_struct *task)
2587{
2588 return do_io_accounting(task, m, 1);
2589}
2590#endif /* CONFIG_TASK_IO_ACCOUNTING */
2591
2592#ifdef CONFIG_USER_NS
2593static int proc_id_map_open(struct inode *inode, struct file *file,
2594 const struct seq_operations *seq_ops)
2595{
2596 struct user_namespace *ns = NULL;
2597 struct task_struct *task;
2598 struct seq_file *seq;
2599 int ret = -EINVAL;
2600
2601 task = get_proc_task(inode);
2602 if (task) {
2603 rcu_read_lock();
2604 ns = get_user_ns(task_cred_xxx(task, user_ns));
2605 rcu_read_unlock();
2606 put_task_struct(task);
2607 }
2608 if (!ns)
2609 goto err;
2610
2611 ret = seq_open(file, seq_ops);
2612 if (ret)
2613 goto err_put_ns;
2614
2615 seq = file->private_data;
2616 seq->private = ns;
2617
2618 return 0;
2619err_put_ns:
2620 put_user_ns(ns);
2621err:
2622 return ret;
2623}
2624
2625static int proc_id_map_release(struct inode *inode, struct file *file)
2626{
2627 struct seq_file *seq = file->private_data;
2628 struct user_namespace *ns = seq->private;
2629 put_user_ns(ns);
2630 return seq_release(inode, file);
2631}
2632
2633static int proc_uid_map_open(struct inode *inode, struct file *file)
2634{
2635 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2636}
2637
2638static int proc_gid_map_open(struct inode *inode, struct file *file)
2639{
2640 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2641}
2642
2643static int proc_projid_map_open(struct inode *inode, struct file *file)
2644{
2645 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2646}
2647
2648static const struct file_operations proc_uid_map_operations = {
2649 .open = proc_uid_map_open,
2650 .write = proc_uid_map_write,
2651 .read = seq_read,
2652 .llseek = seq_lseek,
2653 .release = proc_id_map_release,
2654};
2655
2656static const struct file_operations proc_gid_map_operations = {
2657 .open = proc_gid_map_open,
2658 .write = proc_gid_map_write,
2659 .read = seq_read,
2660 .llseek = seq_lseek,
2661 .release = proc_id_map_release,
2662};
2663
2664static const struct file_operations proc_projid_map_operations = {
2665 .open = proc_projid_map_open,
2666 .write = proc_projid_map_write,
2667 .read = seq_read,
2668 .llseek = seq_lseek,
2669 .release = proc_id_map_release,
2670};
2671
2672static int proc_setgroups_open(struct inode *inode, struct file *file)
2673{
2674 struct user_namespace *ns = NULL;
2675 struct task_struct *task;
2676 int ret;
2677
2678 ret = -ESRCH;
2679 task = get_proc_task(inode);
2680 if (task) {
2681 rcu_read_lock();
2682 ns = get_user_ns(task_cred_xxx(task, user_ns));
2683 rcu_read_unlock();
2684 put_task_struct(task);
2685 }
2686 if (!ns)
2687 goto err;
2688
2689 if (file->f_mode & FMODE_WRITE) {
2690 ret = -EACCES;
2691 if (!ns_capable(ns, CAP_SYS_ADMIN))
2692 goto err_put_ns;
2693 }
2694
2695 ret = single_open(file, &proc_setgroups_show, ns);
2696 if (ret)
2697 goto err_put_ns;
2698
2699 return 0;
2700err_put_ns:
2701 put_user_ns(ns);
2702err:
2703 return ret;
2704}
2705
2706static int proc_setgroups_release(struct inode *inode, struct file *file)
2707{
2708 struct seq_file *seq = file->private_data;
2709 struct user_namespace *ns = seq->private;
2710 int ret = single_release(inode, file);
2711 put_user_ns(ns);
2712 return ret;
2713}
2714
2715static const struct file_operations proc_setgroups_operations = {
2716 .open = proc_setgroups_open,
2717 .write = proc_setgroups_write,
2718 .read = seq_read,
2719 .llseek = seq_lseek,
2720 .release = proc_setgroups_release,
2721};
2722#endif /* CONFIG_USER_NS */
2723
2724static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2725 struct pid *pid, struct task_struct *task)
2726{
2727 int err = lock_trace(task);
2728 if (!err) {
2729 seq_printf(m, "%08x\n", task->personality);
2730 unlock_trace(task);
2731 }
2732 return err;
2733}
2734
2735/*
2736 * Thread groups
2737 */
2738static const struct file_operations proc_task_operations;
2739static const struct inode_operations proc_task_inode_operations;
2740
2741static const struct pid_entry tgid_base_stuff[] = {
2742 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2743 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2744#ifdef CONFIG_CHECKPOINT_RESTORE
2745 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
2746#endif
2747 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2748 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2749#ifdef CONFIG_NET
2750 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2751#endif
2752 REG("environ", S_IRUSR, proc_environ_operations),
2753 ONE("auxv", S_IRUSR, proc_pid_auxv),
2754 ONE("status", S_IRUGO, proc_pid_status),
2755 ONE("personality", S_IRUSR, proc_pid_personality),
2756 ONE("limits", S_IRUGO, proc_pid_limits),
2757#ifdef CONFIG_SMP
2758 REG("sched_wake_up_idle", S_IRUGO|S_IWUSR, proc_pid_sched_wake_up_idle_operations),
2759#endif
2760#ifdef CONFIG_SCHED_HMP
2761 REG("sched_init_task_load", S_IRUGO|S_IWUSR, proc_pid_sched_init_task_load_operations),
2762#ifndef CONFIG_SCHED_QHMP
2763 REG("sched_group_id", S_IRUGO|S_IWUSR, proc_pid_sched_group_id_operations),
2764#endif
2765#endif
2766#ifdef CONFIG_SCHED_DEBUG
2767 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2768#endif
2769#ifdef CONFIG_SCHED_AUTOGROUP
2770 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2771#endif
2772 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2773#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2774 ONE("syscall", S_IRUSR, proc_pid_syscall),
2775#endif
2776 ONE("cmdline", S_IRUGO, proc_pid_cmdline),
2777 ONE("stat", S_IRUGO, proc_tgid_stat),
2778 ONE("statm", S_IRUGO, proc_pid_statm),
2779 REG("maps", S_IRUGO, proc_pid_maps_operations),
2780#ifdef CONFIG_NUMA
2781 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
2782#endif
2783 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2784 LNK("cwd", proc_cwd_link),
2785 LNK("root", proc_root_link),
2786 LNK("exe", proc_exe_link),
2787 REG("mounts", S_IRUGO, proc_mounts_operations),
2788 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2789 REG("mountstats", S_IRUSR, proc_mountstats_operations),
2790#ifdef CONFIG_PROCESS_RECLAIM
2791 REG("reclaim", S_IWUSR, proc_reclaim_operations),
2792#endif
2793#ifdef CONFIG_PROC_PAGE_MONITOR
2794 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2795 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
2796 REG("pagemap", S_IRUSR, proc_pagemap_operations),
2797#endif
2798#ifdef CONFIG_SECURITY
2799 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2800#endif
2801#ifdef CONFIG_KALLSYMS
2802 ONE("wchan", S_IRUGO, proc_pid_wchan),
2803#endif
2804#ifdef CONFIG_STACKTRACE
2805 ONE("stack", S_IRUSR, proc_pid_stack),
2806#endif
2807#ifdef CONFIG_SCHEDSTATS
2808 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
2809#endif
2810#ifdef CONFIG_LATENCYTOP
2811 REG("latency", S_IRUGO, proc_lstats_operations),
2812#endif
2813#ifdef CONFIG_PROC_PID_CPUSET
2814 ONE("cpuset", S_IRUGO, proc_cpuset_show),
2815#endif
2816#ifdef CONFIG_CGROUPS
2817 ONE("cgroup", S_IRUGO, proc_cgroup_show),
2818#endif
2819 ONE("oom_score", S_IRUGO, proc_oom_score),
2820 REG("oom_adj", S_IRUSR, proc_oom_adj_operations),
2821 REG("oom_score_adj", S_IRUSR, proc_oom_score_adj_operations),
2822#ifdef CONFIG_AUDITSYSCALL
2823 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2824 REG("sessionid", S_IRUGO, proc_sessionid_operations),
2825#endif
2826#ifdef CONFIG_FAULT_INJECTION
2827 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2828#endif
2829#ifdef CONFIG_ELF_CORE
2830 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2831#endif
2832#ifdef CONFIG_TASK_IO_ACCOUNTING
2833 ONE("io", S_IRUSR, proc_tgid_io_accounting),
2834#endif
2835#ifdef CONFIG_HARDWALL
2836 ONE("hardwall", S_IRUGO, proc_pid_hardwall),
2837#endif
2838#ifdef CONFIG_USER_NS
2839 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
2840 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
2841 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
2842 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
2843#endif
2844#ifdef CONFIG_CHECKPOINT_RESTORE
2845 REG("timers", S_IRUGO, proc_timers_operations),
2846#endif
2847};
2848
2849static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
2850{
2851 return proc_pident_readdir(file, ctx,
2852 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2853}
2854
2855static const struct file_operations proc_tgid_base_operations = {
2856 .read = generic_read_dir,
2857 .iterate = proc_tgid_base_readdir,
2858 .llseek = default_llseek,
2859};
2860
2861static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2862{
2863 return proc_pident_lookup(dir, dentry,
2864 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2865}
2866
2867static const struct inode_operations proc_tgid_base_inode_operations = {
2868 .lookup = proc_tgid_base_lookup,
2869 .getattr = pid_getattr,
2870 .setattr = proc_setattr,
2871 .permission = proc_pid_permission,
2872};
2873
2874static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2875{
2876 struct dentry *dentry, *leader, *dir;
2877 char buf[PROC_NUMBUF];
2878 struct qstr name;
2879
2880 name.name = buf;
2881 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2882 /* no ->d_hash() rejects on procfs */
2883 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2884 if (dentry) {
2885 d_invalidate(dentry);
2886 dput(dentry);
2887 }
2888
2889 name.name = buf;
2890 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2891 leader = d_hash_and_lookup(mnt->mnt_root, &name);
2892 if (!leader)
2893 goto out;
2894
2895 name.name = "task";
2896 name.len = strlen(name.name);
2897 dir = d_hash_and_lookup(leader, &name);
2898 if (!dir)
2899 goto out_put_leader;
2900
2901 name.name = buf;
2902 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2903 dentry = d_hash_and_lookup(dir, &name);
2904 if (dentry) {
2905 d_invalidate(dentry);
2906 dput(dentry);
2907 }
2908
2909 dput(dir);
2910out_put_leader:
2911 dput(leader);
2912out:
2913 return;
2914}
2915
2916/**
2917 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2918 * @task: task that should be flushed.
2919 *
2920 * When flushing dentries from proc, one needs to flush them from global
2921 * proc (proc_mnt) and from all the namespaces' procs this task was seen
2922 * in. This call is supposed to do all of this job.
2923 *
2924 * Looks in the dcache for
2925 * /proc/@pid
2926 * /proc/@tgid/task/@pid
2927 * if either directory is present flushes it and all of it'ts children
2928 * from the dcache.
2929 *
2930 * It is safe and reasonable to cache /proc entries for a task until
2931 * that task exits. After that they just clog up the dcache with
2932 * useless entries, possibly causing useful dcache entries to be
2933 * flushed instead. This routine is proved to flush those useless
2934 * dcache entries at process exit time.
2935 *
2936 * NOTE: This routine is just an optimization so it does not guarantee
2937 * that no dcache entries will exist at process exit time it
2938 * just makes it very unlikely that any will persist.
2939 */
2940
2941void proc_flush_task(struct task_struct *task)
2942{
2943 int i;
2944 struct pid *pid, *tgid;
2945 struct upid *upid;
2946
2947 pid = task_pid(task);
2948 tgid = task_tgid(task);
2949
2950 for (i = 0; i <= pid->level; i++) {
2951 upid = &pid->numbers[i];
2952 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2953 tgid->numbers[i].nr);
2954 }
2955}
2956
2957static int proc_pid_instantiate(struct inode *dir,
2958 struct dentry * dentry,
2959 struct task_struct *task, const void *ptr)
2960{
2961 struct inode *inode;
2962
2963 inode = proc_pid_make_inode(dir->i_sb, task);
2964 if (!inode)
2965 goto out;
2966
2967 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2968 inode->i_op = &proc_tgid_base_inode_operations;
2969 inode->i_fop = &proc_tgid_base_operations;
2970 inode->i_flags|=S_IMMUTABLE;
2971
2972 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
2973 ARRAY_SIZE(tgid_base_stuff)));
2974
2975 d_set_d_op(dentry, &pid_dentry_operations);
2976
2977 d_add(dentry, inode);
2978 /* Close the race of the process dying before we return the dentry */
2979 if (pid_revalidate(dentry, 0))
2980 return 0;
2981out:
2982 return -ENOENT;
2983}
2984
2985struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
2986{
2987 int result = -ENOENT;
2988 struct task_struct *task;
2989 unsigned tgid;
2990 struct pid_namespace *ns;
2991
2992 tgid = name_to_int(&dentry->d_name);
2993 if (tgid == ~0U)
2994 goto out;
2995
2996 ns = dentry->d_sb->s_fs_info;
2997 rcu_read_lock();
2998 task = find_task_by_pid_ns(tgid, ns);
2999 if (task)
3000 get_task_struct(task);
3001 rcu_read_unlock();
3002 if (!task)
3003 goto out;
3004
3005 result = proc_pid_instantiate(dir, dentry, task, NULL);
3006 put_task_struct(task);
3007out:
3008 return ERR_PTR(result);
3009}
3010
3011/*
3012 * Find the first task with tgid >= tgid
3013 *
3014 */
3015struct tgid_iter {
3016 unsigned int tgid;
3017 struct task_struct *task;
3018};
3019static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3020{
3021 struct pid *pid;
3022
3023 if (iter.task)
3024 put_task_struct(iter.task);
3025 rcu_read_lock();
3026retry:
3027 iter.task = NULL;
3028 pid = find_ge_pid(iter.tgid, ns);
3029 if (pid) {
3030 iter.tgid = pid_nr_ns(pid, ns);
3031 iter.task = pid_task(pid, PIDTYPE_PID);
3032 /* What we to know is if the pid we have find is the
3033 * pid of a thread_group_leader. Testing for task
3034 * being a thread_group_leader is the obvious thing
3035 * todo but there is a window when it fails, due to
3036 * the pid transfer logic in de_thread.
3037 *
3038 * So we perform the straight forward test of seeing
3039 * if the pid we have found is the pid of a thread
3040 * group leader, and don't worry if the task we have
3041 * found doesn't happen to be a thread group leader.
3042 * As we don't care in the case of readdir.
3043 */
3044 if (!iter.task || !has_group_leader_pid(iter.task)) {
3045 iter.tgid += 1;
3046 goto retry;
3047 }
3048 get_task_struct(iter.task);
3049 }
3050 rcu_read_unlock();
3051 return iter;
3052}
3053
3054#define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
3055
3056/* for the /proc/ directory itself, after non-process stuff has been done */
3057int proc_pid_readdir(struct file *file, struct dir_context *ctx)
3058{
3059 struct tgid_iter iter;
3060 struct pid_namespace *ns = file->f_dentry->d_sb->s_fs_info;
3061 loff_t pos = ctx->pos;
3062
3063 if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
3064 return 0;
3065
3066 if (pos == TGID_OFFSET - 2) {
3067 struct inode *inode = ns->proc_self->d_inode;
3068 if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
3069 return 0;
3070 ctx->pos = pos = pos + 1;
3071 }
3072 if (pos == TGID_OFFSET - 1) {
3073 struct inode *inode = ns->proc_thread_self->d_inode;
3074 if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3075 return 0;
3076 ctx->pos = pos = pos + 1;
3077 }
3078 iter.tgid = pos - TGID_OFFSET;
3079 iter.task = NULL;
3080 for (iter = next_tgid(ns, iter);
3081 iter.task;
3082 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3083 char name[PROC_NUMBUF];
3084 int len;
3085 if (!has_pid_permissions(ns, iter.task, 2))
3086 continue;
3087
3088 len = snprintf(name, sizeof(name), "%d", iter.tgid);
3089 ctx->pos = iter.tgid + TGID_OFFSET;
3090 if (!proc_fill_cache(file, ctx, name, len,
3091 proc_pid_instantiate, iter.task, NULL)) {
3092 put_task_struct(iter.task);
3093 return 0;
3094 }
3095 }
3096 ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
3097 return 0;
3098}
3099
3100/*
3101 * Tasks
3102 */
3103static const struct pid_entry tid_base_stuff[] = {
3104 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3105 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3106 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3107#ifdef CONFIG_NET
3108 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3109#endif
3110 REG("environ", S_IRUSR, proc_environ_operations),
3111 ONE("auxv", S_IRUSR, proc_pid_auxv),
3112 ONE("status", S_IRUGO, proc_pid_status),
3113 ONE("personality", S_IRUSR, proc_pid_personality),
3114 ONE("limits", S_IRUGO, proc_pid_limits),
3115#ifdef CONFIG_SCHED_DEBUG
3116 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3117#endif
3118 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3119#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3120 ONE("syscall", S_IRUSR, proc_pid_syscall),
3121#endif
3122 ONE("cmdline", S_IRUGO, proc_pid_cmdline),
3123 ONE("stat", S_IRUGO, proc_tid_stat),
3124 ONE("statm", S_IRUGO, proc_pid_statm),
3125 REG("maps", S_IRUGO, proc_tid_maps_operations),
3126#ifdef CONFIG_CHECKPOINT_RESTORE
3127 REG("children", S_IRUGO, proc_tid_children_operations),
3128#endif
3129#ifdef CONFIG_NUMA
3130 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
3131#endif
3132 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3133 LNK("cwd", proc_cwd_link),
3134 LNK("root", proc_root_link),
3135 LNK("exe", proc_exe_link),
3136 REG("mounts", S_IRUGO, proc_mounts_operations),
3137 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3138#ifdef CONFIG_PROC_PAGE_MONITOR
3139 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3140 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
3141 REG("pagemap", S_IRUSR, proc_pagemap_operations),
3142#endif
3143#ifdef CONFIG_SECURITY
3144 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3145#endif
3146#ifdef CONFIG_KALLSYMS
3147 ONE("wchan", S_IRUGO, proc_pid_wchan),
3148#endif
3149#ifdef CONFIG_STACKTRACE
3150 ONE("stack", S_IRUSR, proc_pid_stack),
3151#endif
3152#ifdef CONFIG_SCHEDSTATS
3153 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
3154#endif
3155#ifdef CONFIG_LATENCYTOP
3156 REG("latency", S_IRUGO, proc_lstats_operations),
3157#endif
3158#ifdef CONFIG_PROC_PID_CPUSET
3159 ONE("cpuset", S_IRUGO, proc_cpuset_show),
3160#endif
3161#ifdef CONFIG_CGROUPS
3162 ONE("cgroup", S_IRUGO, proc_cgroup_show),
3163#endif
3164 ONE("oom_score", S_IRUGO, proc_oom_score),
3165 REG("oom_adj", S_IRUSR, proc_oom_adj_operations),
3166 REG("oom_score_adj", S_IRUSR, proc_oom_score_adj_operations),
3167#ifdef CONFIG_AUDITSYSCALL
3168 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3169 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3170#endif
3171#ifdef CONFIG_FAULT_INJECTION
3172 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3173#endif
3174#ifdef CONFIG_TASK_IO_ACCOUNTING
3175 ONE("io", S_IRUSR, proc_tid_io_accounting),
3176#endif
3177#ifdef CONFIG_HARDWALL
3178 ONE("hardwall", S_IRUGO, proc_pid_hardwall),
3179#endif
3180#ifdef CONFIG_USER_NS
3181 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3182 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3183 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3184 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
3185#endif
3186};
3187
3188static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
3189{
3190 return proc_pident_readdir(file, ctx,
3191 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3192}
3193
3194static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3195{
3196 return proc_pident_lookup(dir, dentry,
3197 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3198}
3199
3200static const struct file_operations proc_tid_base_operations = {
3201 .read = generic_read_dir,
3202 .iterate = proc_tid_base_readdir,
3203 .llseek = default_llseek,
3204};
3205
3206static const struct inode_operations proc_tid_base_inode_operations = {
3207 .lookup = proc_tid_base_lookup,
3208 .getattr = pid_getattr,
3209 .setattr = proc_setattr,
3210};
3211
3212static int proc_task_instantiate(struct inode *dir,
3213 struct dentry *dentry, struct task_struct *task, const void *ptr)
3214{
3215 struct inode *inode;
3216 inode = proc_pid_make_inode(dir->i_sb, task);
3217
3218 if (!inode)
3219 goto out;
3220 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3221 inode->i_op = &proc_tid_base_inode_operations;
3222 inode->i_fop = &proc_tid_base_operations;
3223 inode->i_flags|=S_IMMUTABLE;
3224
3225 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3226 ARRAY_SIZE(tid_base_stuff)));
3227
3228 d_set_d_op(dentry, &pid_dentry_operations);
3229
3230 d_add(dentry, inode);
3231 /* Close the race of the process dying before we return the dentry */
3232 if (pid_revalidate(dentry, 0))
3233 return 0;
3234out:
3235 return -ENOENT;
3236}
3237
3238static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3239{
3240 int result = -ENOENT;
3241 struct task_struct *task;
3242 struct task_struct *leader = get_proc_task(dir);
3243 unsigned tid;
3244 struct pid_namespace *ns;
3245
3246 if (!leader)
3247 goto out_no_task;
3248
3249 tid = name_to_int(&dentry->d_name);
3250 if (tid == ~0U)
3251 goto out;
3252
3253 ns = dentry->d_sb->s_fs_info;
3254 rcu_read_lock();
3255 task = find_task_by_pid_ns(tid, ns);
3256 if (task)
3257 get_task_struct(task);
3258 rcu_read_unlock();
3259 if (!task)
3260 goto out;
3261 if (!same_thread_group(leader, task))
3262 goto out_drop_task;
3263
3264 result = proc_task_instantiate(dir, dentry, task, NULL);
3265out_drop_task:
3266 put_task_struct(task);
3267out:
3268 put_task_struct(leader);
3269out_no_task:
3270 return ERR_PTR(result);
3271}
3272
3273/*
3274 * Find the first tid of a thread group to return to user space.
3275 *
3276 * Usually this is just the thread group leader, but if the users
3277 * buffer was too small or there was a seek into the middle of the
3278 * directory we have more work todo.
3279 *
3280 * In the case of a short read we start with find_task_by_pid.
3281 *
3282 * In the case of a seek we start with the leader and walk nr
3283 * threads past it.
3284 */
3285static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3286 struct pid_namespace *ns)
3287{
3288 struct task_struct *pos, *task;
3289 unsigned long nr = f_pos;
3290
3291 if (nr != f_pos) /* 32bit overflow? */
3292 return NULL;
3293
3294 rcu_read_lock();
3295 task = pid_task(pid, PIDTYPE_PID);
3296 if (!task)
3297 goto fail;
3298
3299 /* Attempt to start with the tid of a thread */
3300 if (tid && nr) {
3301 pos = find_task_by_pid_ns(tid, ns);
3302 if (pos && same_thread_group(pos, task))
3303 goto found;
3304 }
3305
3306 /* If nr exceeds the number of threads there is nothing todo */
3307 if (nr >= get_nr_threads(task))
3308 goto fail;
3309
3310 /* If we haven't found our starting place yet start
3311 * with the leader and walk nr threads forward.
3312 */
3313 pos = task = task->group_leader;
3314 do {
3315 if (!nr--)
3316 goto found;
3317 } while_each_thread(task, pos);
3318fail:
3319 pos = NULL;
3320 goto out;
3321found:
3322 get_task_struct(pos);
3323out:
3324 rcu_read_unlock();
3325 return pos;
3326}
3327
3328/*
3329 * Find the next thread in the thread list.
3330 * Return NULL if there is an error or no next thread.
3331 *
3332 * The reference to the input task_struct is released.
3333 */
3334static struct task_struct *next_tid(struct task_struct *start)
3335{
3336 struct task_struct *pos = NULL;
3337 rcu_read_lock();
3338 if (pid_alive(start)) {
3339 pos = next_thread(start);
3340 if (thread_group_leader(pos))
3341 pos = NULL;
3342 else
3343 get_task_struct(pos);
3344 }
3345 rcu_read_unlock();
3346 put_task_struct(start);
3347 return pos;
3348}
3349
3350/* for the /proc/TGID/task/ directories */
3351static int proc_task_readdir(struct file *file, struct dir_context *ctx)
3352{
3353 struct inode *inode = file_inode(file);
3354 struct task_struct *task;
3355 struct pid_namespace *ns;
3356 int tid;
3357
3358 if (proc_inode_is_dead(inode))
3359 return -ENOENT;
3360
3361 if (!dir_emit_dots(file, ctx))
3362 return 0;
3363
3364 /* f_version caches the tgid value that the last readdir call couldn't
3365 * return. lseek aka telldir automagically resets f_version to 0.
3366 */
3367 ns = file->f_dentry->d_sb->s_fs_info;
3368 tid = (int)file->f_version;
3369 file->f_version = 0;
3370 for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
3371 task;
3372 task = next_tid(task), ctx->pos++) {
3373 char name[PROC_NUMBUF];
3374 int len;
3375 tid = task_pid_nr_ns(task, ns);
3376 len = snprintf(name, sizeof(name), "%d", tid);
3377 if (!proc_fill_cache(file, ctx, name, len,
3378 proc_task_instantiate, task, NULL)) {
3379 /* returning this tgid failed, save it as the first
3380 * pid for the next readir call */
3381 file->f_version = (u64)tid;
3382 put_task_struct(task);
3383 break;
3384 }
3385 }
3386
3387 return 0;
3388}
3389
3390static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3391{
3392 struct inode *inode = dentry->d_inode;
3393 struct task_struct *p = get_proc_task(inode);
3394 generic_fillattr(inode, stat);
3395
3396 if (p) {
3397 stat->nlink += get_nr_threads(p);
3398 put_task_struct(p);
3399 }
3400
3401 return 0;
3402}
3403
3404static const struct inode_operations proc_task_inode_operations = {
3405 .lookup = proc_task_lookup,
3406 .getattr = proc_task_getattr,
3407 .setattr = proc_setattr,
3408 .permission = proc_pid_permission,
3409};
3410
3411static const struct file_operations proc_task_operations = {
3412 .read = generic_read_dir,
3413 .iterate = proc_task_readdir,
3414 .llseek = default_llseek,
3415};