1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * linux/fs/proc/base.c
4 *
5 * Copyright (C) 1991, 1992 Linus Torvalds
6 *
7 * proc base directory handling functions
8 *
9 * 1999, Al Viro. Rewritten. Now it covers the whole per-process part.
10 * Instead of using magical inumbers to determine the kind of object
11 * we allocate and fill in-core inodes upon lookup. They don't even
12 * go into icache. We cache the reference to task_struct upon lookup too.
13 * Eventually it should become a filesystem in its own. We don't use the
14 * rest of procfs anymore.
15 *
16 *
17 * Changelog:
18 * 17-Jan-2005
19 * Allan Bezerra
20 * Bruna Moreira <bruna.moreira@indt.org.br>
21 * Edjard Mota <edjard.mota@indt.org.br>
22 * Ilias Biris <ilias.biris@indt.org.br>
23 * Mauricio Lin <mauricio.lin@indt.org.br>
24 *
25 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
26 *
27 * A new process specific entry (smaps) included in /proc. It shows the
28 * size of rss for each memory area. The maps entry lacks information
29 * about physical memory size (rss) for each mapped file, i.e.,
30 * rss information for executables and library files.
31 * This additional information is useful for any tools that need to know
32 * about physical memory consumption for a process specific library.
33 *
34 * Changelog:
35 * 21-Feb-2005
36 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
37 * Pud inclusion in the page table walking.
38 *
39 * ChangeLog:
40 * 10-Mar-2005
41 * 10LE Instituto Nokia de Tecnologia - INdT:
42 * A better way to walks through the page table as suggested by Hugh Dickins.
43 *
44 * Simo Piiroinen <simo.piiroinen@nokia.com>:
45 * Smaps information related to shared, private, clean and dirty pages.
46 *
47 * Paul Mundt <paul.mundt@nokia.com>:
48 * Overall revision about smaps.
49 */
50
51 #include <linux/uaccess.h>
52
53 #include <linux/errno.h>
54 #include <linux/time.h>
55 #include <linux/proc_fs.h>
56 #include <linux/stat.h>
57 #include <linux/task_io_accounting_ops.h>
58 #include <linux/init.h>
59 #include <linux/capability.h>
60 #include <linux/file.h>
61 #include <linux/fdtable.h>
62 #include <linux/generic-radix-tree.h>
63 #include <linux/string.h>
64 #include <linux/seq_file.h>
65 #include <linux/namei.h>
66 #include <linux/mnt_namespace.h>
67 #include <linux/mm.h>
68 #include <linux/swap.h>
69 #include <linux/rcupdate.h>
70 #include <linux/kallsyms.h>
71 #include <linux/stacktrace.h>
72 #include <linux/resource.h>
73 #include <linux/module.h>
74 #include <linux/mount.h>
75 #include <linux/security.h>
76 #include <linux/ptrace.h>
77 #include <linux/printk.h>
78 #include <linux/cache.h>
79 #include <linux/cgroup.h>
80 #include <linux/cpuset.h>
81 #include <linux/audit.h>
82 #include <linux/poll.h>
83 #include <linux/nsproxy.h>
84 #include <linux/oom.h>
85 #include <linux/elf.h>
86 #include <linux/pid_namespace.h>
87 #include <linux/user_namespace.h>
88 #include <linux/fs_parser.h>
89 #include <linux/fs_struct.h>
90 #include <linux/slab.h>
91 #include <linux/sched/autogroup.h>
92 #include <linux/sched/mm.h>
93 #include <linux/sched/coredump.h>
94 #include <linux/sched/debug.h>
95 #include <linux/sched/stat.h>
96 #include <linux/posix-timers.h>
97 #include <linux/time_namespace.h>
98 #include <linux/resctrl.h>
99 #include <linux/cn_proc.h>
100 #include <linux/ksm.h>
101 #include <uapi/linux/lsm.h>
102 #include <trace/events/oom.h>
103 #include "internal.h"
104 #include "fd.h"
105
106 #include "../../lib/kstrtox.h"
107
108 /* NOTE:
109 * Implementing inode permission operations in /proc is almost
110 * certainly an error. Permission checks need to happen during
111 * each system call not at open time. The reason is that most of
112 * what we wish to check for permissions in /proc varies at runtime.
113 *
114 * The classic example of a problem is opening file descriptors
115 * in /proc for a task before it execs a suid executable.
116 */
117
118 static u8 nlink_tid __ro_after_init;
119 static u8 nlink_tgid __ro_after_init;
120
121 enum proc_mem_force {
122 PROC_MEM_FORCE_ALWAYS,
123 PROC_MEM_FORCE_PTRACE,
124 PROC_MEM_FORCE_NEVER
125 };
126
127 static enum proc_mem_force proc_mem_force_override __ro_after_init =
128 IS_ENABLED(CONFIG_PROC_MEM_NO_FORCE) ? PROC_MEM_FORCE_NEVER :
129 IS_ENABLED(CONFIG_PROC_MEM_FORCE_PTRACE) ? PROC_MEM_FORCE_PTRACE :
130 PROC_MEM_FORCE_ALWAYS;
131
132 static const struct constant_table proc_mem_force_table[] __initconst = {
133 { "always", PROC_MEM_FORCE_ALWAYS },
134 { "ptrace", PROC_MEM_FORCE_PTRACE },
135 { "never", PROC_MEM_FORCE_NEVER },
136 { }
137 };
138
139 static int __init early_proc_mem_force_override(char *buf)
140 {
141 if (!buf)
142 return -EINVAL;
143
144 /*
145 * lookup_constant() defaults to proc_mem_force_override to preseve
146 * the initial Kconfig choice in case an invalid param gets passed.
147 */
148 proc_mem_force_override = lookup_constant(proc_mem_force_table,
149 buf, proc_mem_force_override);
150
151 return 0;
152 }
153 early_param("proc_mem.force_override", early_proc_mem_force_override);
154
155 struct pid_entry {
156 const char *name;
157 unsigned int len;
158 umode_t mode;
159 const struct inode_operations *iop;
160 const struct file_operations *fop;
161 union proc_op op;
162 };
163
164 #define NOD(NAME, MODE, IOP, FOP, OP) { \
165 .name = (NAME), \
166 .len = sizeof(NAME) - 1, \
167 .mode = MODE, \
168 .iop = IOP, \
169 .fop = FOP, \
170 .op = OP, \
171 }
172
173 #define DIR(NAME, MODE, iops, fops) \
174 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
175 #define LNK(NAME, get_link) \
176 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
177 &proc_pid_link_inode_operations, NULL, \
178 { .proc_get_link = get_link } )
179 #define REG(NAME, MODE, fops) \
180 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
181 #define ONE(NAME, MODE, show) \
182 NOD(NAME, (S_IFREG|(MODE)), \
183 NULL, &proc_single_file_operations, \
184 { .proc_show = show } )
185 #define ATTR(LSMID, NAME, MODE) \
186 NOD(NAME, (S_IFREG|(MODE)), \
187 NULL, &proc_pid_attr_operations, \
188 { .lsmid = LSMID })
189
190 /*
191 * Count the number of hardlinks for the pid_entry table, excluding the .
192 * and .. links.
193 */
194 static unsigned int __init pid_entry_nlink(const struct pid_entry *entries,
195 unsigned int n)
196 {
197 unsigned int i;
198 unsigned int count;
199
200 count = 2;
201 for (i = 0; i < n; ++i) {
202 if (S_ISDIR(entries[i].mode))
203 ++count;
204 }
205
206 return count;
207 }
208
209 static int get_task_root(struct task_struct *task, struct path *root)
210 {
211 int result = -ENOENT;
212
213 task_lock(task);
214 if (task->fs) {
215 get_fs_root(task->fs, root);
216 result = 0;
217 }
218 task_unlock(task);
219 return result;
220 }
221
222 static int proc_cwd_link(struct dentry *dentry, struct path *path)
223 {
224 struct task_struct *task = get_proc_task(d_inode(dentry));
225 int result = -ENOENT;
226
227 if (task) {
228 task_lock(task);
229 if (task->fs) {
230 get_fs_pwd(task->fs, path);
231 result = 0;
232 }
233 task_unlock(task);
234 put_task_struct(task);
235 }
236 return result;
237 }
238
239 static int proc_root_link(struct dentry *dentry, struct path *path)
240 {
241 struct task_struct *task = get_proc_task(d_inode(dentry));
242 int result = -ENOENT;
243
244 if (task) {
245 result = get_task_root(task, path);
246 put_task_struct(task);
247 }
248 return result;
249 }
250
251 /*
252 * If the user used setproctitle(), we just get the string from
253 * user space at arg_start, and limit it to a maximum of one page.
254 */
255 static ssize_t get_mm_proctitle(struct mm_struct *mm, char __user *buf,
256 size_t count, unsigned long pos,
257 unsigned long arg_start)
258 {
259 char *page;
260 int ret, got;
261
262 if (pos >= PAGE_SIZE)
263 return 0;
264
265 page = (char *)__get_free_page(GFP_KERNEL);
266 if (!page)
267 return -ENOMEM;
268
269 ret = 0;
270 got = access_remote_vm(mm, arg_start, page, PAGE_SIZE, FOLL_ANON);
271 if (got > 0) {
272 int len = strnlen(page, got);
273
274 /* Include the NUL character if it was found */
275 if (len < got)
276 len++;
277
278 if (len > pos) {
279 len -= pos;
280 if (len > count)
281 len = count;
282 len -= copy_to_user(buf, page+pos, len);
283 if (!len)
284 len = -EFAULT;
285 ret = len;
286 }
287 }
288 free_page((unsigned long)page);
289 return ret;
290 }
291
292 static ssize_t get_mm_cmdline(struct mm_struct *mm, char __user *buf,
293 size_t count, loff_t *ppos)
294 {
295 unsigned long arg_start, arg_end, env_start, env_end;
296 unsigned long pos, len;
297 char *page, c;
298
299 /* Check if process spawned far enough to have cmdline. */
300 if (!mm->env_end)
301 return 0;
302
303 spin_lock(&mm->arg_lock);
304 arg_start = mm->arg_start;
305 arg_end = mm->arg_end;
306 env_start = mm->env_start;
307 env_end = mm->env_end;
308 spin_unlock(&mm->arg_lock);
309
310 if (arg_start >= arg_end)
311 return 0;
312
313 /*
314 * We allow setproctitle() to overwrite the argument
315 * strings, and overflow past the original end. But
316 * only when it overflows into the environment area.
317 */
318 if (env_start != arg_end || env_end < env_start)
319 env_start = env_end = arg_end;
320 len = env_end - arg_start;
321
322 /* We're not going to care if "*ppos" has high bits set */
323 pos = *ppos;
324 if (pos >= len)
325 return 0;
326 if (count > len - pos)
327 count = len - pos;
328 if (!count)
329 return 0;
330
331 /*
332 * Magical special case: if the argv[] end byte is not
333 * zero, the user has overwritten it with setproctitle(3).
334 *
335 * Possible future enhancement: do this only once when
336 * pos is 0, and set a flag in the 'struct file'.
337 */
338 if (access_remote_vm(mm, arg_end-1, &c, 1, FOLL_ANON) == 1 && c)
339 return get_mm_proctitle(mm, buf, count, pos, arg_start);
340
341 /*
342 * For the non-setproctitle() case we limit things strictly
343 * to the [arg_start, arg_end[ range.
344 */
345 pos += arg_start;
346 if (pos < arg_start || pos >= arg_end)
347 return 0;
348 if (count > arg_end - pos)
349 count = arg_end - pos;
350
351 page = (char *)__get_free_page(GFP_KERNEL);
352 if (!page)
353 return -ENOMEM;
354
355 len = 0;
356 while (count) {
357 int got;
358 size_t size = min_t(size_t, PAGE_SIZE, count);
359
360 got = access_remote_vm(mm, pos, page, size, FOLL_ANON);
361 if (got <= 0)
362 break;
363 got -= copy_to_user(buf, page, got);
364 if (unlikely(!got)) {
365 if (!len)
366 len = -EFAULT;
367 break;
368 }
369 pos += got;
370 buf += got;
371 len += got;
372 count -= got;
373 }
374
375 free_page((unsigned long)page);
376 return len;
377 }
378
379 static ssize_t get_task_cmdline(struct task_struct *tsk, char __user *buf,
380 size_t count, loff_t *pos)
381 {
382 struct mm_struct *mm;
383 ssize_t ret;
384
385 mm = get_task_mm(tsk);
386 if (!mm)
387 return 0;
388
389 ret = get_mm_cmdline(mm, buf, count, pos);
390 mmput(mm);
391 return ret;
392 }
393
394 static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,
395 size_t count, loff_t *pos)
396 {
397 struct task_struct *tsk;
398 ssize_t ret;
399
400 BUG_ON(*pos < 0);
401
402 tsk = get_proc_task(file_inode(file));
403 if (!tsk)
404 return -ESRCH;
405 ret = get_task_cmdline(tsk, buf, count, pos);
406 put_task_struct(tsk);
407 if (ret > 0)
408 *pos += ret;
409 return ret;
410 }
411
412 static const struct file_operations proc_pid_cmdline_ops = {
413 .read = proc_pid_cmdline_read,
414 .llseek = generic_file_llseek,
415 };
416
417 #ifdef CONFIG_KALLSYMS
418 /*
419 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
420 * Returns the resolved symbol. If that fails, simply return the address.
421 */
422 static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns,
423 struct pid *pid, struct task_struct *task)
424 {
425 unsigned long wchan;
426 char symname[KSYM_NAME_LEN];
427
428 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
429 goto print0;
430
431 wchan = get_wchan(task);
432 if (wchan && !lookup_symbol_name(wchan, symname)) {
433 seq_puts(m, symname);
434 return 0;
435 }
436
437 print0:
438 seq_putc(m, '');
439 return 0;
440 }
441 #endif /* CONFIG_KALLSYMS */
442
443 static int lock_trace(struct task_struct *task)
444 {
445 int err = down_read_killable(&task->signal->exec_update_lock);
446 if (err)
447 return err;
448 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) {
449 up_read(&task->signal->exec_update_lock);
450 return -EPERM;
451 }
452 return 0;
453 }
454
455 static void unlock_trace(struct task_struct *task)
456 {
457 up_read(&task->signal->exec_update_lock);
458 }
459
460 #ifdef CONFIG_STACKTRACE
461
462 #define MAX_STACK_TRACE_DEPTH 64
463
464 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
465 struct pid *pid, struct task_struct *task)
466 {
467 unsigned long *entries;
468 int err;
469
470 /*
471 * The ability to racily run the kernel stack unwinder on a running task
472 * and then observe the unwinder output is scary; while it is useful for
473 * debugging kernel issues, it can also allow an attacker to leak kernel
474 * stack contents.
475 * Doing this in a manner that is at least safe from races would require
476 * some work to ensure that the remote task can not be scheduled; and
477 * even then, this would still expose the unwinder as local attack
478 * surface.
479 * Therefore, this interface is restricted to root.
480 */
481 if (!file_ns_capable(m->file, &init_user_ns, CAP_SYS_ADMIN))
482 return -EACCES;
483
484 entries = kmalloc_array(MAX_STACK_TRACE_DEPTH, sizeof(*entries),
485 GFP_KERNEL);
486 if (!entries)
487 return -ENOMEM;
488
489 err = lock_trace(task);
490 if (!err) {
491 unsigned int i, nr_entries;
492
493 nr_entries = stack_trace_save_tsk(task, entries,
494 MAX_STACK_TRACE_DEPTH, 0);
495
496 for (i = 0; i < nr_entries; i++) {
497 seq_printf(m, "[<0>] %pB\n", (void *)entries[i]);
498 }
499
500 unlock_trace(task);
501 }
502 kfree(entries);
503
504 return err;
505 }
506 #endif
507
508 #ifdef CONFIG_SCHED_INFO
509 /*
510 * Provides /proc/PID/schedstat
511 */
512 static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
513 struct pid *pid, struct task_struct *task)
514 {
515 if (unlikely(!sched_info_on()))
516 seq_puts(m, "0 0 0\n");
517 else
518 seq_printf(m, "%llu %llu %lu\n",
519 (unsigned long long)task->se.sum_exec_runtime,
520 (unsigned long long)task->sched_info.run_delay,
521 task->sched_info.pcount);
522
523 return 0;
524 }
525 #endif
526
527 #ifdef CONFIG_LATENCYTOP
528 static int lstats_show_proc(struct seq_file *m, void *v)
529 {
530 int i;
531 struct inode *inode = m->private;
532 struct task_struct *task = get_proc_task(inode);
533
534 if (!task)
535 return -ESRCH;
536 seq_puts(m, "Latency Top version : v0.1\n");
537 for (i = 0; i < LT_SAVECOUNT; i++) {
538 struct latency_record *lr = &task->latency_record[i];
539 if (lr->backtrace[0]) {
540 int q;
541 seq_printf(m, "%i %li %li",
542 lr->count, lr->time, lr->max);
543 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
544 unsigned long bt = lr->backtrace[q];
545
546 if (!bt)
547 break;
548 seq_printf(m, " %ps", (void *)bt);
549 }
550 seq_putc(m, '\n');
551 }
552
553 }
554 put_task_struct(task);
555 return 0;
556 }
557
558 static int lstats_open(struct inode *inode, struct file *file)
559 {
560 return single_open(file, lstats_show_proc, inode);
561 }
562
563 static ssize_t lstats_write(struct file *file, const char __user *buf,
564 size_t count, loff_t *offs)
565 {
566 struct task_struct *task = get_proc_task(file_inode(file));
567
568 if (!task)
569 return -ESRCH;
570 clear_tsk_latency_tracing(task);
571 put_task_struct(task);
572
573 return count;
574 }
575
576 static const struct file_operations proc_lstats_operations = {
577 .open = lstats_open,
578 .read = seq_read,
579 .write = lstats_write,
580 .llseek = seq_lseek,
581 .release = single_release,
582 };
583
584 #endif
585
586 static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
587 struct pid *pid, struct task_struct *task)
588 {
589 unsigned long totalpages = totalram_pages() + total_swap_pages;
590 unsigned long points = 0;
591 long badness;
592
593 badness = oom_badness(task, totalpages);
594 /*
595 * Special case OOM_SCORE_ADJ_MIN for all others scale the
596 * badness value into [0, 2000] range which we have been
597 * exporting for a long time so userspace might depend on it.
598 */
599 if (badness != LONG_MIN)
600 points = (1000 + badness * 1000 / (long)totalpages) * 2 / 3;
601
602 seq_printf(m, "%lu\n", points);
603
604 return 0;
605 }
606
607 struct limit_names {
608 const char *name;
609 const char *unit;
610 };
611
612 static const struct limit_names lnames[RLIM_NLIMITS] = {
613 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
614 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
615 [RLIMIT_DATA] = {"Max data size", "bytes"},
616 [RLIMIT_STACK] = {"Max stack size", "bytes"},
617 [RLIMIT_CORE] = {"Max core file size", "bytes"},
618 [RLIMIT_RSS] = {"Max resident set", "bytes"},
619 [RLIMIT_NPROC] = {"Max processes", "processes"},
620 [RLIMIT_NOFILE] = {"Max open files", "files"},
621 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
622 [RLIMIT_AS] = {"Max address space", "bytes"},
623 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
624 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
625 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
626 [RLIMIT_NICE] = {"Max nice priority", NULL},
627 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
628 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
629 };
630
631 /* Display limits for a process */
632 static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
633 struct pid *pid, struct task_struct *task)
634 {
635 unsigned int i;
636 unsigned long flags;
637
638 struct rlimit rlim[RLIM_NLIMITS];
639
640 if (!lock_task_sighand(task, &flags))
641 return 0;
642 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
643 unlock_task_sighand(task, &flags);
644
645 /*
646 * print the file header
647 */
648 seq_puts(m, "Limit "
649 "Soft Limit "
650 "Hard Limit "
651 "Units \n");
652
653 for (i = 0; i < RLIM_NLIMITS; i++) {
654 if (rlim[i].rlim_cur == RLIM_INFINITY)
655 seq_printf(m, "%-25s %-20s ",
656 lnames[i].name, "unlimited");
657 else
658 seq_printf(m, "%-25s %-20lu ",
659 lnames[i].name, rlim[i].rlim_cur);
660
661 if (rlim[i].rlim_max == RLIM_INFINITY)
662 seq_printf(m, "%-20s ", "unlimited");
663 else
664 seq_printf(m, "%-20lu ", rlim[i].rlim_max);
665
666 if (lnames[i].unit)
667 seq_printf(m, "%-10s\n", lnames[i].unit);
668 else
669 seq_putc(m, '\n');
670 }
671
672 return 0;
673 }
674
675 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
676 static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
677 struct pid *pid, struct task_struct *task)
678 {
679 struct syscall_info info;
680 u64 *args = &info.data.args[0];
681 int res;
682
683 res = lock_trace(task);
684 if (res)
685 return res;
686
687 if (task_current_syscall(task, &info))
688 seq_puts(m, "running\n");
689 else if (info.data.nr < 0)
690 seq_printf(m, "%d 0x%llx 0x%llx\n",
691 info.data.nr, info.sp, info.data.instruction_pointer);
692 else
693 seq_printf(m,
694 "%d 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx\n",
695 info.data.nr,
696 args[0], args[1], args[2], args[3], args[4], args[5],
697 info.sp, info.data.instruction_pointer);
698 unlock_trace(task);
699
700 return 0;
701 }
702 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
703
704 /************************************************************************/
705 /* Here the fs part begins */
706 /************************************************************************/
707
708 /* permission checks */
709 static bool proc_fd_access_allowed(struct inode *inode)
710 {
711 struct task_struct *task;
712 bool allowed = false;
713 /* Allow access to a task's file descriptors if it is us or we
714 * may use ptrace attach to the process and find out that
715 * information.
716 */
717 task = get_proc_task(inode);
718 if (task) {
719 allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
720 put_task_struct(task);
721 }
722 return allowed;
723 }
724
725 int proc_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
726 struct iattr *attr)
727 {
728 int error;
729 struct inode *inode = d_inode(dentry);
730
731 if (attr->ia_valid & ATTR_MODE)
732 return -EPERM;
733
734 error = setattr_prepare(&nop_mnt_idmap, dentry, attr);
735 if (error)
736 return error;
737
738 setattr_copy(&nop_mnt_idmap, inode, attr);
739 return 0;
740 }
741
742 /*
743 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
744 * or euid/egid (for hide_pid_min=2)?
745 */
746 static bool has_pid_permissions(struct proc_fs_info *fs_info,
747 struct task_struct *task,
748 enum proc_hidepid hide_pid_min)
749 {
750 /*
751 * If 'hidpid' mount option is set force a ptrace check,
752 * we indicate that we are using a filesystem syscall
753 * by passing PTRACE_MODE_READ_FSCREDS
754 */
755 if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE)
756 return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
757
758 if (fs_info->hide_pid < hide_pid_min)
759 return true;
760 if (in_group_p(fs_info->pid_gid))
761 return true;
762 return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
763 }
764
765
766 static int proc_pid_permission(struct mnt_idmap *idmap,
767 struct inode *inode, int mask)
768 {
769 struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
770 struct task_struct *task;
771 bool has_perms;
772
773 task = get_proc_task(inode);
774 if (!task)
775 return -ESRCH;
776 has_perms = has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS);
777 put_task_struct(task);
778
779 if (!has_perms) {
780 if (fs_info->hide_pid == HIDEPID_INVISIBLE) {
781 /*
782 * Let's make getdents(), stat(), and open()
783 * consistent with each other. If a process
784 * may not stat() a file, it shouldn't be seen
785 * in procfs at all.
786 */
787 return -ENOENT;
788 }
789
790 return -EPERM;
791 }
792 return generic_permission(&nop_mnt_idmap, inode, mask);
793 }
794
795
796
797 static const struct inode_operations proc_def_inode_operations = {
798 .setattr = proc_setattr,
799 };
800
801 static int proc_single_show(struct seq_file *m, void *v)
802 {
803 struct inode *inode = m->private;
804 struct pid_namespace *ns = proc_pid_ns(inode->i_sb);
805 struct pid *pid = proc_pid(inode);
806 struct task_struct *task;
807 int ret;
808
809 task = get_pid_task(pid, PIDTYPE_PID);
810 if (!task)
811 return -ESRCH;
812
813 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
814
815 put_task_struct(task);
816 return ret;
817 }
818
819 static int proc_single_open(struct inode *inode, struct file *filp)
820 {
821 return single_open(filp, proc_single_show, inode);
822 }
823
824 static const struct file_operations proc_single_file_operations = {
825 .open = proc_single_open,
826 .read = seq_read,
827 .llseek = seq_lseek,
828 .release = single_release,
829 };
830
831
832 struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
833 {
834 struct task_struct *task = get_proc_task(inode);
835 struct mm_struct *mm = ERR_PTR(-ESRCH);
836
837 if (task) {
838 mm = mm_access(task, mode | PTRACE_MODE_FSCREDS);
839 put_task_struct(task);
840
841 if (!IS_ERR_OR_NULL(mm)) {
842 /* ensure this mm_struct can't be freed */
843 mmgrab(mm);
844 /* but do not pin its memory */
845 mmput(mm);
846 }
847 }
848
849 return mm;
850 }
851
852 static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
853 {
854 struct mm_struct *mm = proc_mem_open(inode, mode);
855
856 if (IS_ERR(mm))
857 return PTR_ERR(mm);
858
859 file->private_data = mm;
860 return 0;
861 }
862
863 static int mem_open(struct inode *inode, struct file *file)
864 {
865 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
866
867 /* OK to pass negative loff_t, we can catch out-of-range */
868 file->f_mode |= FMODE_UNSIGNED_OFFSET;
869
870 return ret;
871 }
872
873 static bool proc_mem_foll_force(struct file *file, struct mm_struct *mm)
874 {
875 struct task_struct *task;
876 bool ptrace_active = false;
877
878 switch (proc_mem_force_override) {
879 case PROC_MEM_FORCE_NEVER:
880 return false;
881 case PROC_MEM_FORCE_PTRACE:
882 task = get_proc_task(file_inode(file));
883 if (task) {
884 ptrace_active = READ_ONCE(task->ptrace) &&
885 READ_ONCE(task->mm) == mm &&
886 READ_ONCE(task->parent) == current;
887 put_task_struct(task);
888 }
889 return ptrace_active;
890 default:
891 return true;
892 }
893 }
894
895 static ssize_t mem_rw(struct file *file, char __user *buf,
896 size_t count, loff_t *ppos, int write)
897 {
898 struct mm_struct *mm = file->private_data;
899 unsigned long addr = *ppos;
900 ssize_t copied;
901 char *page;
902 unsigned int flags;
903
904 if (!mm)
905 return 0;
906
907 page = (char *)__get_free_page(GFP_KERNEL);
908 if (!page)
909 return -ENOMEM;
910
911 copied = 0;
912 if (!mmget_not_zero(mm))
913 goto free;
914
915 flags = write ? FOLL_WRITE : 0;
916 if (proc_mem_foll_force(file, mm))
917 flags |= FOLL_FORCE;
918
919 while (count > 0) {
920 size_t this_len = min_t(size_t, count, PAGE_SIZE);
921
922 if (write && copy_from_user(page, buf, this_len)) {
923 copied = -EFAULT;
924 break;
925 }
926
927 this_len = access_remote_vm(mm, addr, page, this_len, flags);
928 if (!this_len) {
929 if (!copied)
930 copied = -EIO;
931 break;
932 }
933
934 if (!write && copy_to_user(buf, page, this_len)) {
935 copied = -EFAULT;
936 break;
937 }
938
939 buf += this_len;
940 addr += this_len;
941 copied += this_len;
942 count -= this_len;
943 }
944 *ppos = addr;
945
946 mmput(mm);
947 free:
948 free_page((unsigned long) page);
949 return copied;
950 }
951
952 static ssize_t mem_read(struct file *file, char __user *buf,
953 size_t count, loff_t *ppos)
954 {
955 return mem_rw(file, buf, count, ppos, 0);
956 }
957
958 static ssize_t mem_write(struct file *file, const char __user *buf,
959 size_t count, loff_t *ppos)
960 {
961 return mem_rw(file, (char __user*)buf, count, ppos, 1);
962 }
963
964 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
965 {
966 switch (orig) {
967 case 0:
968 file->f_pos = offset;
969 break;
970 case 1:
971 file->f_pos += offset;
972 break;
973 default:
974 return -EINVAL;
975 }
976 force_successful_syscall_return();
977 return file->f_pos;
978 }
979
980 static int mem_release(struct inode *inode, struct file *file)
981 {
982 struct mm_struct *mm = file->private_data;
983 if (mm)
984 mmdrop(mm);
985 return 0;
986 }
987
988 static const struct file_operations proc_mem_operations = {
989 .llseek = mem_lseek,
990 .read = mem_read,
991 .write = mem_write,
992 .open = mem_open,
993 .release = mem_release,
994 };
995
996 static int environ_open(struct inode *inode, struct file *file)
997 {
998 return __mem_open(inode, file, PTRACE_MODE_READ);
999 }
1000
1001 static ssize_t environ_read(struct file *file, char __user *buf,
1002 size_t count, loff_t *ppos)
1003 {
1004 char *page;
1005 unsigned long src = *ppos;
1006 int ret = 0;
1007 struct mm_struct *mm = file->private_data;
1008 unsigned long env_start, env_end;
1009
1010 /* Ensure the process spawned far enough to have an environment. */
1011 if (!mm || !mm->env_end)
1012 return 0;
1013
1014 page = (char *)__get_free_page(GFP_KERNEL);
1015 if (!page)
1016 return -ENOMEM;
1017
1018 ret = 0;
1019 if (!mmget_not_zero(mm))
1020 goto free;
1021
1022 spin_lock(&mm->arg_lock);
1023 env_start = mm->env_start;
1024 env_end = mm->env_end;
1025 spin_unlock(&mm->arg_lock);
1026
1027 while (count > 0) {
1028 size_t this_len, max_len;
1029 int retval;
1030
1031 if (src >= (env_end - env_start))
1032 break;
1033
1034 this_len = env_end - (env_start + src);
1035
1036 max_len = min_t(size_t, PAGE_SIZE, count);
1037 this_len = min(max_len, this_len);
1038
1039 retval = access_remote_vm(mm, (env_start + src), page, this_len, FOLL_ANON);
1040
1041 if (retval <= 0) {
1042 ret = retval;
1043 break;
1044 }
1045
1046 if (copy_to_user(buf, page, retval)) {
1047 ret = -EFAULT;
1048 break;
1049 }
1050
1051 ret += retval;
1052 src += retval;
1053 buf += retval;
1054 count -= retval;
1055 }
1056 *ppos = src;
1057 mmput(mm);
1058
1059 free:
1060 free_page((unsigned long) page);
1061 return ret;
1062 }
1063
1064 static const struct file_operations proc_environ_operations = {
1065 .open = environ_open,
1066 .read = environ_read,
1067 .llseek = generic_file_llseek,
1068 .release = mem_release,
1069 };
1070
1071 static int auxv_open(struct inode *inode, struct file *file)
1072 {
1073 return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS);
1074 }
1075
1076 static ssize_t auxv_read(struct file *file, char __user *buf,
1077 size_t count, loff_t *ppos)
1078 {
1079 struct mm_struct *mm = file->private_data;
1080 unsigned int nwords = 0;
1081
1082 if (!mm)
1083 return 0;
1084 do {
1085 nwords += 2;
1086 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
1087 return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv,
1088 nwords * sizeof(mm->saved_auxv[0]));
1089 }
1090
1091 static const struct file_operations proc_auxv_operations = {
1092 .open = auxv_open,
1093 .read = auxv_read,
1094 .llseek = generic_file_llseek,
1095 .release = mem_release,
1096 };
1097
1098 static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
1099 loff_t *ppos)
1100 {
1101 struct task_struct *task = get_proc_task(file_inode(file));
1102 char buffer[PROC_NUMBUF];
1103 int oom_adj = OOM_ADJUST_MIN;
1104 size_t len;
1105
1106 if (!task)
1107 return -ESRCH;
1108 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
1109 oom_adj = OOM_ADJUST_MAX;
1110 else
1111 oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
1112 OOM_SCORE_ADJ_MAX;
1113 put_task_struct(task);
1114 if (oom_adj > OOM_ADJUST_MAX)
1115 oom_adj = OOM_ADJUST_MAX;
1116 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
1117 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1118 }
1119
1120 static int __set_oom_adj(struct file *file, int oom_adj, bool legacy)
1121 {
1122 struct mm_struct *mm = NULL;
1123 struct task_struct *task;
1124 int err = 0;
1125
1126 task = get_proc_task(file_inode(file));
1127 if (!task)
1128 return -ESRCH;
1129
1130 mutex_lock(&oom_adj_mutex);
1131 if (legacy) {
1132 if (oom_adj < task->signal->oom_score_adj &&
1133 !capable(CAP_SYS_RESOURCE)) {
1134 err = -EACCES;
1135 goto err_unlock;
1136 }
1137 /*
1138 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1139 * /proc/pid/oom_score_adj instead.
1140 */
1141 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1142 current->comm, task_pid_nr(current), task_pid_nr(task),
1143 task_pid_nr(task));
1144 } else {
1145 if ((short)oom_adj < task->signal->oom_score_adj_min &&
1146 !capable(CAP_SYS_RESOURCE)) {
1147 err = -EACCES;
1148 goto err_unlock;
1149 }
1150 }
1151
1152 /*
1153 * Make sure we will check other processes sharing the mm if this is
1154 * not vfrok which wants its own oom_score_adj.
1155 * pin the mm so it doesn't go away and get reused after task_unlock
1156 */
1157 if (!task->vfork_done) {
1158 struct task_struct *p = find_lock_task_mm(task);
1159
1160 if (p) {
1161 if (test_bit(MMF_MULTIPROCESS, &p->mm->flags)) {
1162 mm = p->mm;
1163 mmgrab(mm);
1164 }
1165 task_unlock(p);
1166 }
1167 }
1168
1169 task->signal->oom_score_adj = oom_adj;
1170 if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1171 task->signal->oom_score_adj_min = (short)oom_adj;
1172 trace_oom_score_adj_update(task);
1173
1174 if (mm) {
1175 struct task_struct *p;
1176
1177 rcu_read_lock();
1178 for_each_process(p) {
1179 if (same_thread_group(task, p))
1180 continue;
1181
1182 /* do not touch kernel threads or the global init */
1183 if (p->flags & PF_KTHREAD || is_global_init(p))
1184 continue;
1185
1186 task_lock(p);
1187 if (!p->vfork_done && process_shares_mm(p, mm)) {
1188 p->signal->oom_score_adj = oom_adj;
1189 if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1190 p->signal->oom_score_adj_min = (short)oom_adj;
1191 }
1192 task_unlock(p);
1193 }
1194 rcu_read_unlock();
1195 mmdrop(mm);
1196 }
1197 err_unlock:
1198 mutex_unlock(&oom_adj_mutex);
1199 put_task_struct(task);
1200 return err;
1201 }
1202
1203 /*
1204 * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1205 * kernels. The effective policy is defined by oom_score_adj, which has a
1206 * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1207 * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1208 * Processes that become oom disabled via oom_adj will still be oom disabled
1209 * with this implementation.
1210 *
1211 * oom_adj cannot be removed since existing userspace binaries use it.
1212 */
1213 static ssize_t oom_adj_write(struct file *file, const char __user *buf,
1214 size_t count, loff_t *ppos)
1215 {
1216 char buffer[PROC_NUMBUF] = {};
1217 int oom_adj;
1218 int err;
1219
1220 if (count > sizeof(buffer) - 1)
1221 count = sizeof(buffer) - 1;
1222 if (copy_from_user(buffer, buf, count)) {
1223 err = -EFAULT;
1224 goto out;
1225 }
1226
1227 err = kstrtoint(strstrip(buffer), 0, &oom_adj);
1228 if (err)
1229 goto out;
1230 if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
1231 oom_adj != OOM_DISABLE) {
1232 err = -EINVAL;
1233 goto out;
1234 }
1235
1236 /*
1237 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1238 * value is always attainable.
1239 */
1240 if (oom_adj == OOM_ADJUST_MAX)
1241 oom_adj = OOM_SCORE_ADJ_MAX;
1242 else
1243 oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
1244
1245 err = __set_oom_adj(file, oom_adj, true);
1246 out:
1247 return err < 0 ? err : count;
1248 }
1249
1250 static const struct file_operations proc_oom_adj_operations = {
1251 .read = oom_adj_read,
1252 .write = oom_adj_write,
1253 .llseek = generic_file_llseek,
1254 };
1255
1256 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1257 size_t count, loff_t *ppos)
1258 {
1259 struct task_struct *task = get_proc_task(file_inode(file));
1260 char buffer[PROC_NUMBUF];
1261 short oom_score_adj = OOM_SCORE_ADJ_MIN;
1262 size_t len;
1263
1264 if (!task)
1265 return -ESRCH;
1266 oom_score_adj = task->signal->oom_score_adj;
1267 put_task_struct(task);
1268 len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
1269 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1270 }
1271
1272 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1273 size_t count, loff_t *ppos)
1274 {
1275 char buffer[PROC_NUMBUF] = {};
1276 int oom_score_adj;
1277 int err;
1278
1279 if (count > sizeof(buffer) - 1)
1280 count = sizeof(buffer) - 1;
1281 if (copy_from_user(buffer, buf, count)) {
1282 err = -EFAULT;
1283 goto out;
1284 }
1285
1286 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1287 if (err)
1288 goto out;
1289 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1290 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1291 err = -EINVAL;
1292 goto out;
1293 }
1294
1295 err = __set_oom_adj(file, oom_score_adj, false);
1296 out:
1297 return err < 0 ? err : count;
1298 }
1299
1300 static const struct file_operations proc_oom_score_adj_operations = {
1301 .read = oom_score_adj_read,
1302 .write = oom_score_adj_write,
1303 .llseek = default_llseek,
1304 };
1305
1306 #ifdef CONFIG_AUDIT
1307 #define TMPBUFLEN 11
1308 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1309 size_t count, loff_t *ppos)
1310 {
1311 struct inode * inode = file_inode(file);
1312 struct task_struct *task = get_proc_task(inode);
1313 ssize_t length;
1314 char tmpbuf[TMPBUFLEN];
1315
1316 if (!task)
1317 return -ESRCH;
1318 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1319 from_kuid(file->f_cred->user_ns,
1320 audit_get_loginuid(task)));
1321 put_task_struct(task);
1322 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1323 }
1324
1325 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1326 size_t count, loff_t *ppos)
1327 {
1328 struct inode * inode = file_inode(file);
1329 uid_t loginuid;
1330 kuid_t kloginuid;
1331 int rv;
1332
1333 /* Don't let kthreads write their own loginuid */
1334 if (current->flags & PF_KTHREAD)
1335 return -EPERM;
1336
1337 rcu_read_lock();
1338 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1339 rcu_read_unlock();
1340 return -EPERM;
1341 }
1342 rcu_read_unlock();
1343
1344 if (*ppos != 0) {
1345 /* No partial writes. */
1346 return -EINVAL;
1347 }
1348
1349 rv = kstrtou32_from_user(buf, count, 10, &loginuid);
1350 if (rv < 0)
1351 return rv;
1352
1353 /* is userspace tring to explicitly UNSET the loginuid? */
1354 if (loginuid == AUDIT_UID_UNSET) {
1355 kloginuid = INVALID_UID;
1356 } else {
1357 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1358 if (!uid_valid(kloginuid))
1359 return -EINVAL;
1360 }
1361
1362 rv = audit_set_loginuid(kloginuid);
1363 if (rv < 0)
1364 return rv;
1365 return count;
1366 }
1367
1368 static const struct file_operations proc_loginuid_operations = {
1369 .read = proc_loginuid_read,
1370 .write = proc_loginuid_write,
1371 .llseek = generic_file_llseek,
1372 };
1373
1374 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1375 size_t count, loff_t *ppos)
1376 {
1377 struct inode * inode = file_inode(file);
1378 struct task_struct *task = get_proc_task(inode);
1379 ssize_t length;
1380 char tmpbuf[TMPBUFLEN];
1381
1382 if (!task)
1383 return -ESRCH;
1384 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1385 audit_get_sessionid(task));
1386 put_task_struct(task);
1387 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1388 }
1389
1390 static const struct file_operations proc_sessionid_operations = {
1391 .read = proc_sessionid_read,
1392 .llseek = generic_file_llseek,
1393 };
1394 #endif
1395
1396 #ifdef CONFIG_FAULT_INJECTION
1397 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1398 size_t count, loff_t *ppos)
1399 {
1400 struct task_struct *task = get_proc_task(file_inode(file));
1401 char buffer[PROC_NUMBUF];
1402 size_t len;
1403 int make_it_fail;
1404
1405 if (!task)
1406 return -ESRCH;
1407 make_it_fail = task->make_it_fail;
1408 put_task_struct(task);
1409
1410 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1411
1412 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1413 }
1414
1415 static ssize_t proc_fault_inject_write(struct file * file,
1416 const char __user * buf, size_t count, loff_t *ppos)
1417 {
1418 struct task_struct *task;
1419 char buffer[PROC_NUMBUF] = {};
1420 int make_it_fail;
1421 int rv;
1422
1423 if (!capable(CAP_SYS_RESOURCE))
1424 return -EPERM;
1425
1426 if (count > sizeof(buffer) - 1)
1427 count = sizeof(buffer) - 1;
1428 if (copy_from_user(buffer, buf, count))
1429 return -EFAULT;
1430 rv = kstrtoint(strstrip(buffer), 0, &make_it_fail);
1431 if (rv < 0)
1432 return rv;
1433 if (make_it_fail < 0 || make_it_fail > 1)
1434 return -EINVAL;
1435
1436 task = get_proc_task(file_inode(file));
1437 if (!task)
1438 return -ESRCH;
1439 task->make_it_fail = make_it_fail;
1440 put_task_struct(task);
1441
1442 return count;
1443 }
1444
1445 static const struct file_operations proc_fault_inject_operations = {
1446 .read = proc_fault_inject_read,
1447 .write = proc_fault_inject_write,
1448 .llseek = generic_file_llseek,
1449 };
1450
1451 static ssize_t proc_fail_nth_write(struct file *file, const char __user *buf,
1452 size_t count, loff_t *ppos)
1453 {
1454 struct task_struct *task;
1455 int err;
1456 unsigned int n;
1457
1458 err = kstrtouint_from_user(buf, count, 0, &n);
1459 if (err)
1460 return err;
1461
1462 task = get_proc_task(file_inode(file));
1463 if (!task)
1464 return -ESRCH;
1465 task->fail_nth = n;
1466 put_task_struct(task);
1467
1468 return count;
1469 }
1470
1471 static ssize_t proc_fail_nth_read(struct file *file, char __user *buf,
1472 size_t count, loff_t *ppos)
1473 {
1474 struct task_struct *task;
1475 char numbuf[PROC_NUMBUF];
1476 ssize_t len;
1477
1478 task = get_proc_task(file_inode(file));
1479 if (!task)
1480 return -ESRCH;
1481 len = snprintf(numbuf, sizeof(numbuf), "%u\n", task->fail_nth);
1482 put_task_struct(task);
1483 return simple_read_from_buffer(buf, count, ppos, numbuf, len);
1484 }
1485
1486 static const struct file_operations proc_fail_nth_operations = {
1487 .read = proc_fail_nth_read,
1488 .write = proc_fail_nth_write,
1489 };
1490 #endif
1491
1492
1493 #ifdef CONFIG_SCHED_DEBUG
1494 /*
1495 * Print out various scheduling related per-task fields:
1496 */
1497 static int sched_show(struct seq_file *m, void *v)
1498 {
1499 struct inode *inode = m->private;
1500 struct pid_namespace *ns = proc_pid_ns(inode->i_sb);
1501 struct task_struct *p;
1502
1503 p = get_proc_task(inode);
1504 if (!p)
1505 return -ESRCH;
1506 proc_sched_show_task(p, ns, m);
1507
1508 put_task_struct(p);
1509
1510 return 0;
1511 }
1512
1513 static ssize_t
1514 sched_write(struct file *file, const char __user *buf,
1515 size_t count, loff_t *offset)
1516 {
1517 struct inode *inode = file_inode(file);
1518 struct task_struct *p;
1519
1520 p = get_proc_task(inode);
1521 if (!p)
1522 return -ESRCH;
1523 proc_sched_set_task(p);
1524
1525 put_task_struct(p);
1526
1527 return count;
1528 }
1529
1530 static int sched_open(struct inode *inode, struct file *filp)
1531 {
1532 return single_open(filp, sched_show, inode);
1533 }
1534
1535 static const struct file_operations proc_pid_sched_operations = {
1536 .open = sched_open,
1537 .read = seq_read,
1538 .write = sched_write,
1539 .llseek = seq_lseek,
1540 .release = single_release,
1541 };
1542
1543 #endif
1544
1545 #ifdef CONFIG_SCHED_AUTOGROUP
1546 /*
1547 * Print out autogroup related information:
1548 */
1549 static int sched_autogroup_show(struct seq_file *m, void *v)
1550 {
1551 struct inode *inode = m->private;
1552 struct task_struct *p;
1553
1554 p = get_proc_task(inode);
1555 if (!p)
1556 return -ESRCH;
1557 proc_sched_autogroup_show_task(p, m);
1558
1559 put_task_struct(p);
1560
1561 return 0;
1562 }
1563
1564 static ssize_t
1565 sched_autogroup_write(struct file *file, const char __user *buf,
1566 size_t count, loff_t *offset)
1567 {
1568 struct inode *inode = file_inode(file);
1569 struct task_struct *p;
1570 char buffer[PROC_NUMBUF] = {};
1571 int nice;
1572 int err;
1573
1574 if (count > sizeof(buffer) - 1)
1575 count = sizeof(buffer) - 1;
1576 if (copy_from_user(buffer, buf, count))
1577 return -EFAULT;
1578
1579 err = kstrtoint(strstrip(buffer), 0, &nice);
1580 if (err < 0)
1581 return err;
1582
1583 p = get_proc_task(inode);
1584 if (!p)
1585 return -ESRCH;
1586
1587 err = proc_sched_autogroup_set_nice(p, nice);
1588 if (err)
1589 count = err;
1590
1591 put_task_struct(p);
1592
1593 return count;
1594 }
1595
1596 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1597 {
1598 int ret;
1599
1600 ret = single_open(filp, sched_autogroup_show, NULL);
1601 if (!ret) {
1602 struct seq_file *m = filp->private_data;
1603
1604 m->private = inode;
1605 }
1606 return ret;
1607 }
1608
1609 static const struct file_operations proc_pid_sched_autogroup_operations = {
1610 .open = sched_autogroup_open,
1611 .read = seq_read,
1612 .write = sched_autogroup_write,
1613 .llseek = seq_lseek,
1614 .release = single_release,
1615 };
1616
1617 #endif /* CONFIG_SCHED_AUTOGROUP */
1618
1619 #ifdef CONFIG_TIME_NS
1620 static int timens_offsets_show(struct seq_file *m, void *v)
1621 {
1622 struct task_struct *p;
1623
1624 p = get_proc_task(file_inode(m->file));
1625 if (!p)
1626 return -ESRCH;
1627 proc_timens_show_offsets(p, m);
1628
1629 put_task_struct(p);
1630
1631 return 0;
1632 }
1633
1634 static ssize_t timens_offsets_write(struct file *file, const char __user *buf,
1635 size_t count, loff_t *ppos)
1636 {
1637 struct inode *inode = file_inode(file);
1638 struct proc_timens_offset offsets[2];
1639 char *kbuf = NULL, *pos, *next_line;
1640 struct task_struct *p;
1641 int ret, noffsets;
1642
1643 /* Only allow < page size writes at the beginning of the file */
1644 if ((*ppos != 0) || (count >= PAGE_SIZE))
1645 return -EINVAL;
1646
1647 /* Slurp in the user data */
1648 kbuf = memdup_user_nul(buf, count);
1649 if (IS_ERR(kbuf))
1650 return PTR_ERR(kbuf);
1651
1652 /* Parse the user data */
1653 ret = -EINVAL;
1654 noffsets = 0;
1655 for (pos = kbuf; pos; pos = next_line) {
1656 struct proc_timens_offset *off = &offsets[noffsets];
1657 char clock[10];
1658 int err;
1659
1660 /* Find the end of line and ensure we don't look past it */
1661 next_line = strchr(pos, '\n');
1662 if (next_line) {
1663 *next_line = '\0';
1664 next_line++;
1665 if (*next_line == '\0')
1666 next_line = NULL;
1667 }
1668
1669 err = sscanf(pos, "%9s %lld %lu", clock,
1670 &off->val.tv_sec, &off->val.tv_nsec);
1671 if (err != 3 || off->val.tv_nsec >= NSEC_PER_SEC)
1672 goto out;
1673
1674 clock[sizeof(clock) - 1] = 0;
1675 if (strcmp(clock, "monotonic") == 0 ||
1676 strcmp(clock, __stringify(CLOCK_MONOTONIC)) == 0)
1677 off->clockid = CLOCK_MONOTONIC;
1678 else if (strcmp(clock, "boottime") == 0 ||
1679 strcmp(clock, __stringify(CLOCK_BOOTTIME)) == 0)
1680 off->clockid = CLOCK_BOOTTIME;
1681 else
1682 goto out;
1683
1684 noffsets++;
1685 if (noffsets == ARRAY_SIZE(offsets)) {
1686 if (next_line)
1687 count = next_line - kbuf;
1688 break;
1689 }
1690 }
1691
1692 ret = -ESRCH;
1693 p = get_proc_task(inode);
1694 if (!p)
1695 goto out;
1696 ret = proc_timens_set_offset(file, p, offsets, noffsets);
1697 put_task_struct(p);
1698 if (ret)
1699 goto out;
1700
1701 ret = count;
1702 out:
1703 kfree(kbuf);
1704 return ret;
1705 }
1706
1707 static int timens_offsets_open(struct inode *inode, struct file *filp)
1708 {
1709 return single_open(filp, timens_offsets_show, inode);
1710 }
1711
1712 static const struct file_operations proc_timens_offsets_operations = {
1713 .open = timens_offsets_open,
1714 .read = seq_read,
1715 .write = timens_offsets_write,
1716 .llseek = seq_lseek,
1717 .release = single_release,
1718 };
1719 #endif /* CONFIG_TIME_NS */
1720
1721 static ssize_t comm_write(struct file *file, const char __user *buf,
1722 size_t count, loff_t *offset)
1723 {
1724 struct inode *inode = file_inode(file);
1725 struct task_struct *p;
1726 char buffer[TASK_COMM_LEN] = {};
1727 const size_t maxlen = sizeof(buffer) - 1;
1728
1729 if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
1730 return -EFAULT;
1731
1732 p = get_proc_task(inode);
1733 if (!p)
1734 return -ESRCH;
1735
1736 if (same_thread_group(current, p)) {
1737 set_task_comm(p, buffer);
1738 proc_comm_connector(p);
1739 }
1740 else
1741 count = -EINVAL;
1742
1743 put_task_struct(p);
1744
1745 return count;
1746 }
1747
1748 static int comm_show(struct seq_file *m, void *v)
1749 {
1750 struct inode *inode = m->private;
1751 struct task_struct *p;
1752
1753 p = get_proc_task(inode);
1754 if (!p)
1755 return -ESRCH;
1756
1757 proc_task_name(m, p, false);
1758 seq_putc(m, '\n');
1759
1760 put_task_struct(p);
1761
1762 return 0;
1763 }
1764
1765 static int comm_open(struct inode *inode, struct file *filp)
1766 {
1767 return single_open(filp, comm_show, inode);
1768 }
1769
1770 static const struct file_operations proc_pid_set_comm_operations = {
1771 .open = comm_open,
1772 .read = seq_read,
1773 .write = comm_write,
1774 .llseek = seq_lseek,
1775 .release = single_release,
1776 };
1777
1778 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1779 {
1780 struct task_struct *task;
1781 struct file *exe_file;
1782
1783 task = get_proc_task(d_inode(dentry));
1784 if (!task)
1785 return -ENOENT;
1786 exe_file = get_task_exe_file(task);
1787 put_task_struct(task);
1788 if (exe_file) {
1789 *exe_path = exe_file->f_path;
1790 path_get(&exe_file->f_path);
1791 fput(exe_file);
1792 return 0;
1793 } else
1794 return -ENOENT;
1795 }
1796
1797 static const char *proc_pid_get_link(struct dentry *dentry,
1798 struct inode *inode,
1799 struct delayed_call *done)
1800 {
1801 struct path path;
1802 int error = -EACCES;
1803
1804 if (!dentry)
1805 return ERR_PTR(-ECHILD);
1806
1807 /* Are we allowed to snoop on the tasks file descriptors? */
1808 if (!proc_fd_access_allowed(inode))
1809 goto out;
1810
1811 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1812 if (error)
1813 goto out;
1814
1815 error = nd_jump_link(&path);
1816 out:
1817 return ERR_PTR(error);
1818 }
1819
1820 static int do_proc_readlink(const struct path *path, char __user *buffer, int buflen)
1821 {
1822 char *tmp = kmalloc(PATH_MAX, GFP_KERNEL);
1823 char *pathname;
1824 int len;
1825
1826 if (!tmp)
1827 return -ENOMEM;
1828
1829 pathname = d_path(path, tmp, PATH_MAX);
1830 len = PTR_ERR(pathname);
1831 if (IS_ERR(pathname))
1832 goto out;
1833 len = tmp + PATH_MAX - 1 - pathname;
1834
1835 if (len > buflen)
1836 len = buflen;
1837 if (copy_to_user(buffer, pathname, len))
1838 len = -EFAULT;
1839 out:
1840 kfree(tmp);
1841 return len;
1842 }
1843
1844 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1845 {
1846 int error = -EACCES;
1847 struct inode *inode = d_inode(dentry);
1848 struct path path;
1849
1850 /* Are we allowed to snoop on the tasks file descriptors? */
1851 if (!proc_fd_access_allowed(inode))
1852 goto out;
1853
1854 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1855 if (error)
1856 goto out;
1857
1858 error = do_proc_readlink(&path, buffer, buflen);
1859 path_put(&path);
1860 out:
1861 return error;
1862 }
1863
1864 const struct inode_operations proc_pid_link_inode_operations = {
1865 .readlink = proc_pid_readlink,
1866 .get_link = proc_pid_get_link,
1867 .setattr = proc_setattr,
1868 };
1869
1870
1871 /* building an inode */
1872
1873 void task_dump_owner(struct task_struct *task, umode_t mode,
1874 kuid_t *ruid, kgid_t *rgid)
1875 {
1876 /* Depending on the state of dumpable compute who should own a
1877 * proc file for a task.
1878 */
1879 const struct cred *cred;
1880 kuid_t uid;
1881 kgid_t gid;
1882
1883 if (unlikely(task->flags & PF_KTHREAD)) {
1884 *ruid = GLOBAL_ROOT_UID;
1885 *rgid = GLOBAL_ROOT_GID;
1886 return;
1887 }
1888
1889 /* Default to the tasks effective ownership */
1890 rcu_read_lock();
1891 cred = __task_cred(task);
1892 uid = cred->euid;
1893 gid = cred->egid;
1894 rcu_read_unlock();
1895
1896 /*
1897 * Before the /proc/pid/status file was created the only way to read
1898 * the effective uid of a /process was to stat /proc/pid. Reading
1899 * /proc/pid/status is slow enough that procps and other packages
1900 * kept stating /proc/pid. To keep the rules in /proc simple I have
1901 * made this apply to all per process world readable and executable
1902 * directories.
1903 */
1904 if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) {
1905 struct mm_struct *mm;
1906 task_lock(task);
1907 mm = task->mm;
1908 /* Make non-dumpable tasks owned by some root */
1909 if (mm) {
1910 if (get_dumpable(mm) != SUID_DUMP_USER) {
1911 struct user_namespace *user_ns = mm->user_ns;
1912
1913 uid = make_kuid(user_ns, 0);
1914 if (!uid_valid(uid))
1915 uid = GLOBAL_ROOT_UID;
1916
1917 gid = make_kgid(user_ns, 0);
1918 if (!gid_valid(gid))
1919 gid = GLOBAL_ROOT_GID;
1920 }
1921 } else {
1922 uid = GLOBAL_ROOT_UID;
1923 gid = GLOBAL_ROOT_GID;
1924 }
1925 task_unlock(task);
1926 }
1927 *ruid = uid;
1928 *rgid = gid;
1929 }
1930
1931 void proc_pid_evict_inode(struct proc_inode *ei)
1932 {
1933 struct pid *pid = ei->pid;
1934
1935 if (S_ISDIR(ei->vfs_inode.i_mode)) {
1936 spin_lock(&pid->lock);
1937 hlist_del_init_rcu(&ei->sibling_inodes);
1938 spin_unlock(&pid->lock);
1939 }
1940 }
1941
1942 struct inode *proc_pid_make_inode(struct super_block *sb,
1943 struct task_struct *task, umode_t mode)
1944 {
1945 struct inode * inode;
1946 struct proc_inode *ei;
1947 struct pid *pid;
1948
1949 /* We need a new inode */
1950
1951 inode = new_inode(sb);
1952 if (!inode)
1953 goto out;
1954
1955 /* Common stuff */
1956 ei = PROC_I(inode);
1957 inode->i_mode = mode;
1958 inode->i_ino = get_next_ino();
1959 simple_inode_init_ts(inode);
1960 inode->i_op = &proc_def_inode_operations;
1961
1962 /*
1963 * grab the reference to task.
1964 */
1965 pid = get_task_pid(task, PIDTYPE_PID);
1966 if (!pid)
1967 goto out_unlock;
1968
1969 /* Let the pid remember us for quick removal */
1970 ei->pid = pid;
1971
1972 task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
1973 security_task_to_inode(task, inode);
1974
1975 out:
1976 return inode;
1977
1978 out_unlock:
1979 iput(inode);
1980 return NULL;
1981 }
1982
1983 /*
1984 * Generating an inode and adding it into @pid->inodes, so that task will
1985 * invalidate inode's dentry before being released.
1986 *
1987 * This helper is used for creating dir-type entries under '/proc' and
1988 * '/proc/<tgid>/task'. Other entries(eg. fd, stat) under '/proc/<tgid>'
1989 * can be released by invalidating '/proc/<tgid>' dentry.
1990 * In theory, dentries under '/proc/<tgid>/task' can also be released by
1991 * invalidating '/proc/<tgid>' dentry, we reserve it to handle single
1992 * thread exiting situation: Any one of threads should invalidate its
1993 * '/proc/<tgid>/task/<pid>' dentry before released.
1994 */
1995 static struct inode *proc_pid_make_base_inode(struct super_block *sb,
1996 struct task_struct *task, umode_t mode)
1997 {
1998 struct inode *inode;
1999 struct proc_inode *ei;
2000 struct pid *pid;
2001
2002 inode = proc_pid_make_inode(sb, task, mode);
2003 if (!inode)
2004 return NULL;
2005
2006 /* Let proc_flush_pid find this directory inode */
2007 ei = PROC_I(inode);
2008 pid = ei->pid;
2009 spin_lock(&pid->lock);
2010 hlist_add_head_rcu(&ei->sibling_inodes, &pid->inodes);
2011 spin_unlock(&pid->lock);
2012
2013 return inode;
2014 }
2015
2016 int pid_getattr(struct mnt_idmap *idmap, const struct path *path,
2017 struct kstat *stat, u32 request_mask, unsigned int query_flags)
2018 {
2019 struct inode *inode = d_inode(path->dentry);
2020 struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
2021 struct task_struct *task;
2022
2023 generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat);
2024
2025 stat->uid = GLOBAL_ROOT_UID;
2026 stat->gid = GLOBAL_ROOT_GID;
2027 rcu_read_lock();
2028 task = pid_task(proc_pid(inode), PIDTYPE_PID);
2029 if (task) {
2030 if (!has_pid_permissions(fs_info, task, HIDEPID_INVISIBLE)) {
2031 rcu_read_unlock();
2032 /*
2033 * This doesn't prevent learning whether PID exists,
2034 * it only makes getattr() consistent with readdir().
2035 */
2036 return -ENOENT;
2037 }
2038 task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid);
2039 }
2040 rcu_read_unlock();
2041 return 0;
2042 }
2043
2044 /* dentry stuff */
2045
2046 /*
2047 * Set <pid>/... inode ownership (can change due to setuid(), etc.)
2048 */
2049 void pid_update_inode(struct task_struct *task, struct inode *inode)
2050 {
2051 task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid);
2052
2053 inode->i_mode &= ~(S_ISUID | S_ISGID);
2054 security_task_to_inode(task, inode);
2055 }
2056
2057 /*
2058 * Rewrite the inode's ownerships here because the owning task may have
2059 * performed a setuid(), etc.
2060 *
2061 */
2062 static int pid_revalidate(struct dentry *dentry, unsigned int flags)
2063 {
2064 struct inode *inode;
2065 struct task_struct *task;
2066 int ret = 0;
2067
2068 rcu_read_lock();
2069 inode = d_inode_rcu(dentry);
2070 if (!inode)
2071 goto out;
2072 task = pid_task(proc_pid(inode), PIDTYPE_PID);
2073
2074 if (task) {
2075 pid_update_inode(task, inode);
2076 ret = 1;
2077 }
2078 out:
2079 rcu_read_unlock();
2080 return ret;
2081 }
2082
2083 static inline bool proc_inode_is_dead(struct inode *inode)
2084 {
2085 return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
2086 }
2087
2088 int pid_delete_dentry(const struct dentry *dentry)
2089 {
2090 /* Is the task we represent dead?
2091 * If so, then don't put the dentry on the lru list,
2092 * kill it immediately.
2093 */
2094 return proc_inode_is_dead(d_inode(dentry));
2095 }
2096
2097 const struct dentry_operations pid_dentry_operations =
2098 {
2099 .d_revalidate = pid_revalidate,
2100 .d_delete = pid_delete_dentry,
2101 };
2102
2103 /* Lookups */
2104
2105 /*
2106 * Fill a directory entry.
2107 *
2108 * If possible create the dcache entry and derive our inode number and
2109 * file type from dcache entry.
2110 *
2111 * Since all of the proc inode numbers are dynamically generated, the inode
2112 * numbers do not exist until the inode is cache. This means creating
2113 * the dcache entry in readdir is necessary to keep the inode numbers
2114 * reported by readdir in sync with the inode numbers reported
2115 * by stat.
2116 */
2117 bool proc_fill_cache(struct file *file, struct dir_context *ctx,
2118 const char *name, unsigned int len,
2119 instantiate_t instantiate, struct task_struct *task, const void *ptr)
2120 {
2121 struct dentry *child, *dir = file->f_path.dentry;
2122 struct qstr qname = QSTR_INIT(name, len);
2123 struct inode *inode;
2124 unsigned type = DT_UNKNOWN;
2125 ino_t ino = 1;
2126
2127 child = d_hash_and_lookup(dir, &qname);
2128 if (!child) {
2129 DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq);
2130 child = d_alloc_parallel(dir, &qname, &wq);
2131 if (IS_ERR(child))
2132 goto end_instantiate;
2133 if (d_in_lookup(child)) {
2134 struct dentry *res;
2135 res = instantiate(child, task, ptr);
2136 d_lookup_done(child);
2137 if (unlikely(res)) {
2138 dput(child);
2139 child = res;
2140 if (IS_ERR(child))
2141 goto end_instantiate;
2142 }
2143 }
2144 }
2145 inode = d_inode(child);
2146 ino = inode->i_ino;
2147 type = inode->i_mode >> 12;
2148 dput(child);
2149 end_instantiate:
2150 return dir_emit(ctx, name, len, ino, type);
2151 }
2152
2153 /*
2154 * dname_to_vma_addr - maps a dentry name into two unsigned longs
2155 * which represent vma start and end addresses.
2156 */
2157 static int dname_to_vma_addr(struct dentry *dentry,
2158 unsigned long *start, unsigned long *end)
2159 {
2160 const char *str = dentry->d_name.name;
2161 unsigned long long sval, eval;
2162 unsigned int len;
2163
2164 if (str[0] == '' && str[1] != '-')
2165 return -EINVAL;
2166 len = _parse_integer(str, 16, &sval);
2167 if (len & KSTRTOX_OVERFLOW)
2168 return -EINVAL;
2169 if (sval != (unsigned long)sval)
2170 return -EINVAL;
2171 str += len;
2172
2173 if (*str != '-')
2174 return -EINVAL;
2175 str++;
2176
2177 if (str[0] == '' && str[1])
2178 return -EINVAL;
2179 len = _parse_integer(str, 16, &eval);
2180 if (len & KSTRTOX_OVERFLOW)
2181 return -EINVAL;
2182 if (eval != (unsigned long)eval)
2183 return -EINVAL;
2184 str += len;
2185
2186 if (*str != '\0')
2187 return -EINVAL;
2188
2189 *start = sval;
2190 *end = eval;
2191
2192 return 0;
2193 }
2194
2195 static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
2196 {
2197 unsigned long vm_start, vm_end;
2198 bool exact_vma_exists = false;
2199 struct mm_struct *mm = NULL;
2200 struct task_struct *task;
2201 struct inode *inode;
2202 int status = 0;
2203
2204 if (flags & LOOKUP_RCU)
2205 return -ECHILD;
2206
2207 inode = d_inode(dentry);
2208 task = get_proc_task(inode);
2209 if (!task)
2210 goto out_notask;
2211
2212 mm = mm_access(task, PTRACE_MODE_READ_FSCREDS);
2213 if (IS_ERR_OR_NULL(mm))
2214 goto out;
2215
2216 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
2217 status = mmap_read_lock_killable(mm);
2218 if (!status) {
2219 exact_vma_exists = !!find_exact_vma(mm, vm_start,
2220 vm_end);
2221 mmap_read_unlock(mm);
2222 }
2223 }
2224
2225 mmput(mm);
2226
2227 if (exact_vma_exists) {
2228 task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
2229
2230 security_task_to_inode(task, inode);
2231 status = 1;
2232 }
2233
2234 out:
2235 put_task_struct(task);
2236
2237 out_notask:
2238 return status;
2239 }
2240
2241 static const struct dentry_operations tid_map_files_dentry_operations = {
2242 .d_revalidate = map_files_d_revalidate,
2243 .d_delete = pid_delete_dentry,
2244 };
2245
2246 static int map_files_get_link(struct dentry *dentry, struct path *path)
2247 {
2248 unsigned long vm_start, vm_end;
2249 struct vm_area_struct *vma;
2250 struct task_struct *task;
2251 struct mm_struct *mm;
2252 int rc;
2253
2254 rc = -ENOENT;
2255 task = get_proc_task(d_inode(dentry));
2256 if (!task)
2257 goto out;
2258
2259 mm = get_task_mm(task);
2260 put_task_struct(task);
2261 if (!mm)
2262 goto out;
2263
2264 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2265 if (rc)
2266 goto out_mmput;
2267
2268 rc = mmap_read_lock_killable(mm);
2269 if (rc)
2270 goto out_mmput;
2271
2272 rc = -ENOENT;
2273 vma = find_exact_vma(mm, vm_start, vm_end);
2274 if (vma && vma->vm_file) {
2275 *path = *file_user_path(vma->vm_file);
2276 path_get(path);
2277 rc = 0;
2278 }
2279 mmap_read_unlock(mm);
2280
2281 out_mmput:
2282 mmput(mm);
2283 out:
2284 return rc;
2285 }
2286
2287 struct map_files_info {
2288 unsigned long start;
2289 unsigned long end;
2290 fmode_t mode;
2291 };
2292
2293 /*
2294 * Only allow CAP_SYS_ADMIN and CAP_CHECKPOINT_RESTORE to follow the links, due
2295 * to concerns about how the symlinks may be used to bypass permissions on
2296 * ancestor directories in the path to the file in question.
2297 */
2298 static const char *
2299 proc_map_files_get_link(struct dentry *dentry,
2300 struct inode *inode,
2301 struct delayed_call *done)
2302 {
2303 if (!checkpoint_restore_ns_capable(&init_user_ns))
2304 return ERR_PTR(-EPERM);
2305
2306 return proc_pid_get_link(dentry, inode, done);
2307 }
2308
2309 /*
2310 * Identical to proc_pid_link_inode_operations except for get_link()
2311 */
2312 static const struct inode_operations proc_map_files_link_inode_operations = {
2313 .readlink = proc_pid_readlink,
2314 .get_link = proc_map_files_get_link,
2315 .setattr = proc_setattr,
2316 };
2317
2318 static struct dentry *
2319 proc_map_files_instantiate(struct dentry *dentry,
2320 struct task_struct *task, const void *ptr)
2321 {
2322 fmode_t mode = (fmode_t)(unsigned long)ptr;
2323 struct proc_inode *ei;
2324 struct inode *inode;
2325
2326 inode = proc_pid_make_inode(dentry->d_sb, task, S_IFLNK |
2327 ((mode & FMODE_READ ) ? S_IRUSR : 0) |
2328 ((mode & FMODE_WRITE) ? S_IWUSR : 0));
2329 if (!inode)
2330 return ERR_PTR(-ENOENT);
2331
2332 ei = PROC_I(inode);
2333 ei->op.proc_get_link = map_files_get_link;
2334
2335 inode->i_op = &proc_map_files_link_inode_operations;
2336 inode->i_size = 64;
2337
2338 d_set_d_op(dentry, &tid_map_files_dentry_operations);
2339 return d_splice_alias(inode, dentry);
2340 }
2341
2342 static struct dentry *proc_map_files_lookup(struct inode *dir,
2343 struct dentry *dentry, unsigned int flags)
2344 {
2345 unsigned long vm_start, vm_end;
2346 struct vm_area_struct *vma;
2347 struct task_struct *task;
2348 struct dentry *result;
2349 struct mm_struct *mm;
2350
2351 result = ERR_PTR(-ENOENT);
2352 task = get_proc_task(dir);
2353 if (!task)
2354 goto out;
2355
2356 result = ERR_PTR(-EACCES);
2357 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2358 goto out_put_task;
2359
2360 result = ERR_PTR(-ENOENT);
2361 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2362 goto out_put_task;
2363
2364 mm = get_task_mm(task);
2365 if (!mm)
2366 goto out_put_task;
2367
2368 result = ERR_PTR(-EINTR);
2369 if (mmap_read_lock_killable(mm))
2370 goto out_put_mm;
2371
2372 result = ERR_PTR(-ENOENT);
2373 vma = find_exact_vma(mm, vm_start, vm_end);
2374 if (!vma)
2375 goto out_no_vma;
2376
2377 if (vma->vm_file)
2378 result = proc_map_files_instantiate(dentry, task,
2379 (void *)(unsigned long)vma->vm_file->f_mode);
2380
2381 out_no_vma:
2382 mmap_read_unlock(mm);
2383 out_put_mm:
2384 mmput(mm);
2385 out_put_task:
2386 put_task_struct(task);
2387 out:
2388 return result;
2389 }
2390
2391 static const struct inode_operations proc_map_files_inode_operations = {
2392 .lookup = proc_map_files_lookup,
2393 .permission = proc_fd_permission,
2394 .setattr = proc_setattr,
2395 };
2396
2397 static int
2398 proc_map_files_readdir(struct file *file, struct dir_context *ctx)
2399 {
2400 struct vm_area_struct *vma;
2401 struct task_struct *task;
2402 struct mm_struct *mm;
2403 unsigned long nr_files, pos, i;
2404 GENRADIX(struct map_files_info) fa;
2405 struct map_files_info *p;
2406 int ret;
2407 struct vma_iterator vmi;
2408
2409 genradix_init(&fa);
2410
2411 ret = -ENOENT;
2412 task = get_proc_task(file_inode(file));
2413 if (!task)
2414 goto out;
2415
2416 ret = -EACCES;
2417 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2418 goto out_put_task;
2419
2420 ret = 0;
2421 if (!dir_emit_dots(file, ctx))
2422 goto out_put_task;
2423
2424 mm = get_task_mm(task);
2425 if (!mm)
2426 goto out_put_task;
2427
2428 ret = mmap_read_lock_killable(mm);
2429 if (ret) {
2430 mmput(mm);
2431 goto out_put_task;
2432 }
2433
2434 nr_files = 0;
2435
2436 /*
2437 * We need two passes here:
2438 *
2439 * 1) Collect vmas of mapped files with mmap_lock taken
2440 * 2) Release mmap_lock and instantiate entries
2441 *
2442 * otherwise we get lockdep complained, since filldir()
2443 * routine might require mmap_lock taken in might_fault().
2444 */
2445
2446 pos = 2;
2447 vma_iter_init(&vmi, mm, 0);
2448 for_each_vma(vmi, vma) {
2449 if (!vma->vm_file)
2450 continue;
2451 if (++pos <= ctx->pos)
2452 continue;
2453
2454 p = genradix_ptr_alloc(&fa, nr_files++, GFP_KERNEL);
2455 if (!p) {
2456 ret = -ENOMEM;
2457 mmap_read_unlock(mm);
2458 mmput(mm);
2459 goto out_put_task;
2460 }
2461
2462 p->start = vma->vm_start;
2463 p->end = vma->vm_end;
2464 p->mode = vma->vm_file->f_mode;
2465 }
2466 mmap_read_unlock(mm);
2467 mmput(mm);
2468
2469 for (i = 0; i < nr_files; i++) {
2470 char buf[4 * sizeof(long) + 2]; /* max: %lx-%lx\0 */
2471 unsigned int len;
2472
2473 p = genradix_ptr(&fa, i);
2474 len = snprintf(buf, sizeof(buf), "%lx-%lx", p->start, p->end);
2475 if (!proc_fill_cache(file, ctx,
2476 buf, len,
2477 proc_map_files_instantiate,
2478 task,
2479 (void *)(unsigned long)p->mode))
2480 break;
2481 ctx->pos++;
2482 }
2483
2484 out_put_task:
2485 put_task_struct(task);
2486 out:
2487 genradix_free(&fa);
2488 return ret;
2489 }
2490
2491 static const struct file_operations proc_map_files_operations = {
2492 .read = generic_read_dir,
2493 .iterate_shared = proc_map_files_readdir,
2494 .llseek = generic_file_llseek,
2495 };
2496
2497 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
2498 struct timers_private {
2499 struct pid *pid;
2500 struct task_struct *task;
2501 struct sighand_struct *sighand;
2502 struct pid_namespace *ns;
2503 unsigned long flags;
2504 };
2505
2506 static void *timers_start(struct seq_file *m, loff_t *pos)
2507 {
2508 struct timers_private *tp = m->private;
2509
2510 tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2511 if (!tp->task)
2512 return ERR_PTR(-ESRCH);
2513
2514 tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2515 if (!tp->sighand)
2516 return ERR_PTR(-ESRCH);
2517
2518 return seq_list_start(&tp->task->signal->posix_timers, *pos);
2519 }
2520
2521 static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2522 {
2523 struct timers_private *tp = m->private;
2524 return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2525 }
2526
2527 static void timers_stop(struct seq_file *m, void *v)
2528 {
2529 struct timers_private *tp = m->private;
2530
2531 if (tp->sighand) {
2532 unlock_task_sighand(tp->task, &tp->flags);
2533 tp->sighand = NULL;
2534 }
2535
2536 if (tp->task) {
2537 put_task_struct(tp->task);
2538 tp->task = NULL;
2539 }
2540 }
2541
2542 static int show_timer(struct seq_file *m, void *v)
2543 {
2544 struct k_itimer *timer;
2545 struct timers_private *tp = m->private;
2546 int notify;
2547 static const char * const nstr[] = {
2548 [SIGEV_SIGNAL] = "signal",
2549 [SIGEV_NONE] = "none",
2550 [SIGEV_THREAD] = "thread",
2551 };
2552
2553 timer = list_entry((struct list_head *)v, struct k_itimer, list);
2554 notify = timer->it_sigev_notify;
2555
2556 seq_printf(m, "ID: %d\n", timer->it_id);
2557 seq_printf(m, "signal: %d/%px\n",
2558 timer->sigq->info.si_signo,
2559 timer->sigq->info.si_value.sival_ptr);
2560 seq_printf(m, "notify: %s/%s.%d\n",
2561 nstr[notify & ~SIGEV_THREAD_ID],
2562 (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2563 pid_nr_ns(timer->it_pid, tp->ns));
2564 seq_printf(m, "ClockID: %d\n", timer->it_clock);
2565
2566 return 0;
2567 }
2568
2569 static const struct seq_operations proc_timers_seq_ops = {
2570 .start = timers_start,
2571 .next = timers_next,
2572 .stop = timers_stop,
2573 .show = show_timer,
2574 };
2575
2576 static int proc_timers_open(struct inode *inode, struct file *file)
2577 {
2578 struct timers_private *tp;
2579
2580 tp = __seq_open_private(file, &proc_timers_seq_ops,
2581 sizeof(struct timers_private));
2582 if (!tp)
2583 return -ENOMEM;
2584
2585 tp->pid = proc_pid(inode);
2586 tp->ns = proc_pid_ns(inode->i_sb);
2587 return 0;
2588 }
2589
2590 static const struct file_operations proc_timers_operations = {
2591 .open = proc_timers_open,
2592 .read = seq_read,
2593 .llseek = seq_lseek,
2594 .release = seq_release_private,
2595 };
2596 #endif
2597
2598 static ssize_t timerslack_ns_write(struct file *file, const char __user *buf,
2599 size_t count, loff_t *offset)
2600 {
2601 struct inode *inode = file_inode(file);
2602 struct task_struct *p;
2603 u64 slack_ns;
2604 int err;
2605
2606 err = kstrtoull_from_user(buf, count, 10, &slack_ns);
2607 if (err < 0)
2608 return err;
2609
2610 p = get_proc_task(inode);
2611 if (!p)
2612 return -ESRCH;
2613
2614 if (p != current) {
2615 rcu_read_lock();
2616 if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
2617 rcu_read_unlock();
2618 count = -EPERM;
2619 goto out;
2620 }
2621 rcu_read_unlock();
2622
2623 err = security_task_setscheduler(p);
2624 if (err) {
2625 count = err;
2626 goto out;
2627 }
2628 }
2629
2630 task_lock(p);
2631 if (slack_ns == 0)
2632 p->timer_slack_ns = p->default_timer_slack_ns;
2633 else
2634 p->timer_slack_ns = slack_ns;
2635 task_unlock(p);
2636
2637 out:
2638 put_task_struct(p);
2639
2640 return count;
2641 }
2642
2643 static int timerslack_ns_show(struct seq_file *m, void *v)
2644 {
2645 struct inode *inode = m->private;
2646 struct task_struct *p;
2647 int err = 0;
2648
2649 p = get_proc_task(inode);
2650 if (!p)
2651 return -ESRCH;
2652
2653 if (p != current) {
2654 rcu_read_lock();
2655 if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
2656 rcu_read_unlock();
2657 err = -EPERM;
2658 goto out;
2659 }
2660 rcu_read_unlock();
2661
2662 err = security_task_getscheduler(p);
2663 if (err)
2664 goto out;
2665 }
2666
2667 task_lock(p);
2668 seq_printf(m, "%llu\n", p->timer_slack_ns);
2669 task_unlock(p);
2670
2671 out:
2672 put_task_struct(p);
2673
2674 return err;
2675 }
2676
2677 static int timerslack_ns_open(struct inode *inode, struct file *filp)
2678 {
2679 return single_open(filp, timerslack_ns_show, inode);
2680 }
2681
2682 static const struct file_operations proc_pid_set_timerslack_ns_operations = {
2683 .open = timerslack_ns_open,
2684 .read = seq_read,
2685 .write = timerslack_ns_write,
2686 .llseek = seq_lseek,
2687 .release = single_release,
2688 };
2689
2690 static struct dentry *proc_pident_instantiate(struct dentry *dentry,
2691 struct task_struct *task, const void *ptr)
2692 {
2693 const struct pid_entry *p = ptr;
2694 struct inode *inode;
2695 struct proc_inode *ei;
2696
2697 inode = proc_pid_make_inode(dentry->d_sb, task, p->mode);
2698 if (!inode)
2699 return ERR_PTR(-ENOENT);
2700
2701 ei = PROC_I(inode);
2702 if (S_ISDIR(inode->i_mode))
2703 set_nlink(inode, 2); /* Use getattr to fix if necessary */
2704 if (p->iop)
2705 inode->i_op = p->iop;
2706 if (p->fop)
2707 inode->i_fop = p->fop;
2708 ei->op = p->op;
2709 pid_update_inode(task, inode);
2710 d_set_d_op(dentry, &pid_dentry_operations);
2711 return d_splice_alias(inode, dentry);
2712 }
2713
2714 static struct dentry *proc_pident_lookup(struct inode *dir,
2715 struct dentry *dentry,
2716 const struct pid_entry *p,
2717 const struct pid_entry *end)
2718 {
2719 struct task_struct *task = get_proc_task(dir);
2720 struct dentry *res = ERR_PTR(-ENOENT);
2721
2722 if (!task)
2723 goto out_no_task;
2724
2725 /*
2726 * Yes, it does not scale. And it should not. Don't add
2727 * new entries into /proc/<tgid>/ without very good reasons.
2728 */
2729 for (; p < end; p++) {
2730 if (p->len != dentry->d_name.len)
2731 continue;
2732 if (!memcmp(dentry->d_name.name, p->name, p->len)) {
2733 res = proc_pident_instantiate(dentry, task, p);
2734 break;
2735 }
2736 }
2737 put_task_struct(task);
2738 out_no_task:
2739 return res;
2740 }
2741
2742 static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
2743 const struct pid_entry *ents, unsigned int nents)
2744 {
2745 struct task_struct *task = get_proc_task(file_inode(file));
2746 const struct pid_entry *p;
2747
2748 if (!task)
2749 return -ENOENT;
2750
2751 if (!dir_emit_dots(file, ctx))
2752 goto out;
2753
2754 if (ctx->pos >= nents + 2)
2755 goto out;
2756
2757 for (p = ents + (ctx->pos - 2); p < ents + nents; p++) {
2758 if (!proc_fill_cache(file, ctx, p->name, p->len,
2759 proc_pident_instantiate, task, p))
2760 break;
2761 ctx->pos++;
2762 }
2763 out:
2764 put_task_struct(task);
2765 return 0;
2766 }
2767
2768 #ifdef CONFIG_SECURITY
2769 static int proc_pid_attr_open(struct inode *inode, struct file *file)
2770 {
2771 file->private_data = NULL;
2772 __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS);
2773 return 0;
2774 }
2775
2776 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2777 size_t count, loff_t *ppos)
2778 {
2779 struct inode * inode = file_inode(file);
2780 char *p = NULL;
2781 ssize_t length;
2782 struct task_struct *task = get_proc_task(inode);
2783
2784 if (!task)
2785 return -ESRCH;
2786
2787 length = security_getprocattr(task, PROC_I(inode)->op.lsmid,
2788 file->f_path.dentry->d_name.name,
2789 &p);
2790 put_task_struct(task);
2791 if (length > 0)
2792 length = simple_read_from_buffer(buf, count, ppos, p, length);
2793 kfree(p);
2794 return length;
2795 }
2796
2797 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2798 size_t count, loff_t *ppos)
2799 {
2800 struct inode * inode = file_inode(file);
2801 struct task_struct *task;
2802 void *page;
2803 int rv;
2804
2805 /* A task may only write when it was the opener. */
2806 if (file->private_data != current->mm)
2807 return -EPERM;
2808
2809 rcu_read_lock();
2810 task = pid_task(proc_pid(inode), PIDTYPE_PID);
2811 if (!task) {
2812 rcu_read_unlock();
2813 return -ESRCH;
2814 }
2815 /* A task may only write its own attributes. */
2816 if (current != task) {
2817 rcu_read_unlock();
2818 return -EACCES;
2819 }
2820 /* Prevent changes to overridden credentials. */
2821 if (current_cred() != current_real_cred()) {
2822 rcu_read_unlock();
2823 return -EBUSY;
2824 }
2825 rcu_read_unlock();
2826
2827 if (count > PAGE_SIZE)
2828 count = PAGE_SIZE;
2829
2830 /* No partial writes. */
2831 if (*ppos != 0)
2832 return -EINVAL;
2833
2834 page = memdup_user(buf, count);
2835 if (IS_ERR(page)) {
2836 rv = PTR_ERR(page);
2837 goto out;
2838 }
2839
2840 /* Guard against adverse ptrace interaction */
2841 rv = mutex_lock_interruptible(¤t->signal->cred_guard_mutex);
2842 if (rv < 0)
2843 goto out_free;
2844
2845 rv = security_setprocattr(PROC_I(inode)->op.lsmid,
2846 file->f_path.dentry->d_name.name, page,
2847 count);
2848 mutex_unlock(¤t->signal->cred_guard_mutex);
2849 out_free:
2850 kfree(page);
2851 out:
2852 return rv;
2853 }
2854
2855 static const struct file_operations proc_pid_attr_operations = {
2856 .open = proc_pid_attr_open,
2857 .read = proc_pid_attr_read,
2858 .write = proc_pid_attr_write,
2859 .llseek = generic_file_llseek,
2860 .release = mem_release,
2861 };
2862
2863 #define LSM_DIR_OPS(LSM) \
2864 static int proc_##LSM##_attr_dir_iterate(struct file *filp, \
2865 struct dir_context *ctx) \
2866 { \
2867 return proc_pident_readdir(filp, ctx, \
2868 LSM##_attr_dir_stuff, \
2869 ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2870 } \
2871 \
2872 static const struct file_operations proc_##LSM##_attr_dir_ops = { \
2873 .read = generic_read_dir, \
2874 .iterate_shared = proc_##LSM##_attr_dir_iterate, \
2875 .llseek = default_llseek, \
2876 }; \
2877 \
2878 static struct dentry *proc_##LSM##_attr_dir_lookup(struct inode *dir, \
2879 struct dentry *dentry, unsigned int flags) \
2880 { \
2881 return proc_pident_lookup(dir, dentry, \
2882 LSM##_attr_dir_stuff, \
2883 LSM##_attr_dir_stuff + ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2884 } \
2885 \
2886 static const struct inode_operations proc_##LSM##_attr_dir_inode_ops = { \
2887 .lookup = proc_##LSM##_attr_dir_lookup, \
2888 .getattr = pid_getattr, \
2889 .setattr = proc_setattr, \
2890 }
2891
2892 #ifdef CONFIG_SECURITY_SMACK
2893 static const struct pid_entry smack_attr_dir_stuff[] = {
2894 ATTR(LSM_ID_SMACK, "current", 0666),
2895 };
2896 LSM_DIR_OPS(smack);
2897 #endif
2898
2899 #ifdef CONFIG_SECURITY_APPARMOR
2900 static const struct pid_entry apparmor_attr_dir_stuff[] = {
2901 ATTR(LSM_ID_APPARMOR, "current", 0666),
2902 ATTR(LSM_ID_APPARMOR, "prev", 0444),
2903 ATTR(LSM_ID_APPARMOR, "exec", 0666),
2904 };
2905 LSM_DIR_OPS(apparmor);
2906 #endif
2907
2908 static const struct pid_entry attr_dir_stuff[] = {
2909 ATTR(LSM_ID_UNDEF, "current", 0666),
2910 ATTR(LSM_ID_UNDEF, "prev", 0444),
2911 ATTR(LSM_ID_UNDEF, "exec", 0666),
2912 ATTR(LSM_ID_UNDEF, "fscreate", 0666),
2913 ATTR(LSM_ID_UNDEF, "keycreate", 0666),
2914 ATTR(LSM_ID_UNDEF, "sockcreate", 0666),
2915 #ifdef CONFIG_SECURITY_SMACK
2916 DIR("smack", 0555,
2917 proc_smack_attr_dir_inode_ops, proc_smack_attr_dir_ops),
2918 #endif
2919 #ifdef CONFIG_SECURITY_APPARMOR
2920 DIR("apparmor", 0555,
2921 proc_apparmor_attr_dir_inode_ops, proc_apparmor_attr_dir_ops),
2922 #endif
2923 };
2924
2925 static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
2926 {
2927 return proc_pident_readdir(file, ctx,
2928 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2929 }
2930
2931 static const struct file_operations proc_attr_dir_operations = {
2932 .read = generic_read_dir,
2933 .iterate_shared = proc_attr_dir_readdir,
2934 .llseek = generic_file_llseek,
2935 };
2936
2937 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2938 struct dentry *dentry, unsigned int flags)
2939 {
2940 return proc_pident_lookup(dir, dentry,
2941 attr_dir_stuff,
2942 attr_dir_stuff + ARRAY_SIZE(attr_dir_stuff));
2943 }
2944
2945 static const struct inode_operations proc_attr_dir_inode_operations = {
2946 .lookup = proc_attr_dir_lookup,
2947 .getattr = pid_getattr,
2948 .setattr = proc_setattr,
2949 };
2950
2951 #endif
2952
2953 #ifdef CONFIG_ELF_CORE
2954 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2955 size_t count, loff_t *ppos)
2956 {
2957 struct task_struct *task = get_proc_task(file_inode(file));
2958 struct mm_struct *mm;
2959 char buffer[PROC_NUMBUF];
2960 size_t len;
2961 int ret;
2962
2963 if (!task)
2964 return -ESRCH;
2965
2966 ret = 0;
2967 mm = get_task_mm(task);
2968 if (mm) {
2969 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2970 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2971 MMF_DUMP_FILTER_SHIFT));
2972 mmput(mm);
2973 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2974 }
2975
2976 put_task_struct(task);
2977
2978 return ret;
2979 }
2980
2981 static ssize_t proc_coredump_filter_write(struct file *file,
2982 const char __user *buf,
2983 size_t count,
2984 loff_t *ppos)
2985 {
2986 struct task_struct *task;
2987 struct mm_struct *mm;
2988 unsigned int val;
2989 int ret;
2990 int i;
2991 unsigned long mask;
2992
2993 ret = kstrtouint_from_user(buf, count, 0, &val);
2994 if (ret < 0)
2995 return ret;
2996
2997 ret = -ESRCH;
2998 task = get_proc_task(file_inode(file));
2999 if (!task)
3000 goto out_no_task;
3001
3002 mm = get_task_mm(task);
3003 if (!mm)
3004 goto out_no_mm;
3005 ret = 0;
3006
3007 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
3008 if (val & mask)
3009 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
3010 else
3011 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
3012 }
3013
3014 mmput(mm);
3015 out_no_mm:
3016 put_task_struct(task);
3017 out_no_task:
3018 if (ret < 0)
3019 return ret;
3020 return count;
3021 }
3022
3023 static const struct file_operations proc_coredump_filter_operations = {
3024 .read = proc_coredump_filter_read,
3025 .write = proc_coredump_filter_write,
3026 .llseek = generic_file_llseek,
3027 };
3028 #endif
3029
3030 #ifdef CONFIG_TASK_IO_ACCOUNTING
3031 static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
3032 {
3033 struct task_io_accounting acct;
3034 int result;
3035
3036 result = down_read_killable(&task->signal->exec_update_lock);
3037 if (result)
3038 return result;
3039
3040 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) {
3041 result = -EACCES;
3042 goto out_unlock;
3043 }
3044
3045 if (whole) {
3046 struct signal_struct *sig = task->signal;
3047 struct task_struct *t;
3048 unsigned int seq = 1;
3049 unsigned long flags;
3050
3051 rcu_read_lock();
3052 do {
3053 seq++; /* 2 on the 1st/lockless path, otherwise odd */
3054 flags = read_seqbegin_or_lock_irqsave(&sig->stats_lock, &seq);
3055
3056 acct = sig->ioac;
3057 __for_each_thread(sig, t)
3058 task_io_accounting_add(&acct, &t->ioac);
3059
3060 } while (need_seqretry(&sig->stats_lock, seq));
3061 done_seqretry_irqrestore(&sig->stats_lock, seq, flags);
3062 rcu_read_unlock();
3063 } else {
3064 acct = task->ioac;
3065 }
3066
3067 seq_printf(m,
3068 "rchar: %llu\n"
3069 "wchar: %llu\n"
3070 "syscr: %llu\n"
3071 "syscw: %llu\n"
3072 "read_bytes: %llu\n"
3073 "write_bytes: %llu\n"
3074 "cancelled_write_bytes: %llu\n",
3075 (unsigned long long)acct.rchar,
3076 (unsigned long long)acct.wchar,
3077 (unsigned long long)acct.syscr,
3078 (unsigned long long)acct.syscw,
3079 (unsigned long long)acct.read_bytes,
3080 (unsigned long long)acct.write_bytes,
3081 (unsigned long long)acct.cancelled_write_bytes);
3082 result = 0;
3083
3084 out_unlock:
3085 up_read(&task->signal->exec_update_lock);
3086 return result;
3087 }
3088
3089 static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
3090 struct pid *pid, struct task_struct *task)
3091 {
3092 return do_io_accounting(task, m, 0);
3093 }
3094
3095 static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
3096 struct pid *pid, struct task_struct *task)
3097 {
3098 return do_io_accounting(task, m, 1);
3099 }
3100 #endif /* CONFIG_TASK_IO_ACCOUNTING */
3101
3102 #ifdef CONFIG_USER_NS
3103 static int proc_id_map_open(struct inode *inode, struct file *file,
3104 const struct seq_operations *seq_ops)
3105 {
3106 struct user_namespace *ns = NULL;
3107 struct task_struct *task;
3108 struct seq_file *seq;
3109 int ret = -EINVAL;
3110
3111 task = get_proc_task(inode);
3112 if (task) {
3113 rcu_read_lock();
3114 ns = get_user_ns(task_cred_xxx(task, user_ns));
3115 rcu_read_unlock();
3116 put_task_struct(task);
3117 }
3118 if (!ns)
3119 goto err;
3120
3121 ret = seq_open(file, seq_ops);
3122 if (ret)
3123 goto err_put_ns;
3124
3125 seq = file->private_data;
3126 seq->private = ns;
3127
3128 return 0;
3129 err_put_ns:
3130 put_user_ns(ns);
3131 err:
3132 return ret;
3133 }
3134
3135 static int proc_id_map_release(struct inode *inode, struct file *file)
3136 {
3137 struct seq_file *seq = file->private_data;
3138 struct user_namespace *ns = seq->private;
3139 put_user_ns(ns);
3140 return seq_release(inode, file);
3141 }
3142
3143 static int proc_uid_map_open(struct inode *inode, struct file *file)
3144 {
3145 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
3146 }
3147
3148 static int proc_gid_map_open(struct inode *inode, struct file *file)
3149 {
3150 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
3151 }
3152
3153 static int proc_projid_map_open(struct inode *inode, struct file *file)
3154 {
3155 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
3156 }
3157
3158 static const struct file_operations proc_uid_map_operations = {
3159 .open = proc_uid_map_open,
3160 .write = proc_uid_map_write,
3161 .read = seq_read,
3162 .llseek = seq_lseek,
3163 .release = proc_id_map_release,
3164 };
3165
3166 static const struct file_operations proc_gid_map_operations = {
3167 .open = proc_gid_map_open,
3168 .write = proc_gid_map_write,
3169 .read = seq_read,
3170 .llseek = seq_lseek,
3171 .release = proc_id_map_release,
3172 };
3173
3174 static const struct file_operations proc_projid_map_operations = {
3175 .open = proc_projid_map_open,
3176 .write = proc_projid_map_write,
3177 .read = seq_read,
3178 .llseek = seq_lseek,
3179 .release = proc_id_map_release,
3180 };
3181
3182 static int proc_setgroups_open(struct inode *inode, struct file *file)
3183 {
3184 struct user_namespace *ns = NULL;
3185 struct task_struct *task;
3186 int ret;
3187
3188 ret = -ESRCH;
3189 task = get_proc_task(inode);
3190 if (task) {
3191 rcu_read_lock();
3192 ns = get_user_ns(task_cred_xxx(task, user_ns));
3193 rcu_read_unlock();
3194 put_task_struct(task);
3195 }
3196 if (!ns)
3197 goto err;
3198
3199 if (file->f_mode & FMODE_WRITE) {
3200 ret = -EACCES;
3201 if (!ns_capable(ns, CAP_SYS_ADMIN))
3202 goto err_put_ns;
3203 }
3204
3205 ret = single_open(file, &proc_setgroups_show, ns);
3206 if (ret)
3207 goto err_put_ns;
3208
3209 return 0;
3210 err_put_ns:
3211 put_user_ns(ns);
3212 err:
3213 return ret;
3214 }
3215
3216 static int proc_setgroups_release(struct inode *inode, struct file *file)
3217 {
3218 struct seq_file *seq = file->private_data;
3219 struct user_namespace *ns = seq->private;
3220 int ret = single_release(inode, file);
3221 put_user_ns(ns);
3222 return ret;
3223 }
3224
3225 static const struct file_operations proc_setgroups_operations = {
3226 .open = proc_setgroups_open,
3227 .write = proc_setgroups_write,
3228 .read = seq_read,
3229 .llseek = seq_lseek,
3230 .release = proc_setgroups_release,
3231 };
3232 #endif /* CONFIG_USER_NS */
3233
3234 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
3235 struct pid *pid, struct task_struct *task)
3236 {
3237 int err = lock_trace(task);
3238 if (!err) {
3239 seq_printf(m, "%08x\n", task->personality);
3240 unlock_trace(task);
3241 }
3242 return err;
3243 }
3244
3245 #ifdef CONFIG_LIVEPATCH
3246 static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns,
3247 struct pid *pid, struct task_struct *task)
3248 {
3249 seq_printf(m, "%d\n", task->patch_state);
3250 return 0;
3251 }
3252 #endif /* CONFIG_LIVEPATCH */
3253
3254 #ifdef CONFIG_KSM
3255 static int proc_pid_ksm_merging_pages(struct seq_file *m, struct pid_namespace *ns,
3256 struct pid *pid, struct task_struct *task)
3257 {
3258 struct mm_struct *mm;
3259
3260 mm = get_task_mm(task);
3261 if (mm) {
3262 seq_printf(m, "%lu\n", mm->ksm_merging_pages);
3263 mmput(mm);
3264 }
3265
3266 return 0;
3267 }
3268 static int proc_pid_ksm_stat(struct seq_file *m, struct pid_namespace *ns,
3269 struct pid *pid, struct task_struct *task)
3270 {
3271 struct mm_struct *mm;
3272
3273 mm = get_task_mm(task);
3274 if (mm) {
3275 seq_printf(m, "ksm_rmap_items %lu\n", mm->ksm_rmap_items);
3276 seq_printf(m, "ksm_zero_pages %ld\n", mm_ksm_zero_pages(mm));
3277 seq_printf(m, "ksm_merging_pages %lu\n", mm->ksm_merging_pages);
3278 seq_printf(m, "ksm_process_profit %ld\n", ksm_process_profit(mm));
3279 mmput(mm);
3280 }
3281
3282 return 0;
3283 }
3284 #endif /* CONFIG_KSM */
3285
3286 #ifdef CONFIG_STACKLEAK_METRICS
3287 static int proc_stack_depth(struct seq_file *m, struct pid_namespace *ns,
3288 struct pid *pid, struct task_struct *task)
3289 {
3290 unsigned long prev_depth = THREAD_SIZE -
3291 (task->prev_lowest_stack & (THREAD_SIZE - 1));
3292 unsigned long depth = THREAD_SIZE -
3293 (task->lowest_stack & (THREAD_SIZE - 1));
3294
3295 seq_printf(m, "previous stack depth: %lu\nstack depth: %lu\n",
3296 prev_depth, depth);
3297 return 0;
3298 }
3299 #endif /* CONFIG_STACKLEAK_METRICS */
3300
3301 /*
3302 * Thread groups
3303 */
3304 static const struct file_operations proc_task_operations;
3305 static const struct inode_operations proc_task_inode_operations;
3306
3307 static const struct pid_entry tgid_base_stuff[] = {
3308 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
3309 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3310 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
3311 DIR("fdinfo", S_IRUGO|S_IXUGO, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3312 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3313 #ifdef CONFIG_NET
3314 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3315 #endif
3316 REG("environ", S_IRUSR, proc_environ_operations),
3317 REG("auxv", S_IRUSR, proc_auxv_operations),
3318 ONE("status", S_IRUGO, proc_pid_status),
3319 ONE("personality", S_IRUSR, proc_pid_personality),
3320 ONE("limits", S_IRUGO, proc_pid_limits),
3321 #ifdef CONFIG_SCHED_DEBUG
3322 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3323 #endif
3324 #ifdef CONFIG_SCHED_AUTOGROUP
3325 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
3326 #endif
3327 #ifdef CONFIG_TIME_NS
3328 REG("timens_offsets", S_IRUGO|S_IWUSR, proc_timens_offsets_operations),
3329 #endif
3330 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3331 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3332 ONE("syscall", S_IRUSR, proc_pid_syscall),
3333 #endif
3334 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
3335 ONE("stat", S_IRUGO, proc_tgid_stat),
3336 ONE("statm", S_IRUGO, proc_pid_statm),
3337 REG("maps", S_IRUGO, proc_pid_maps_operations),
3338 #ifdef CONFIG_NUMA
3339 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
3340 #endif
3341 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3342 LNK("cwd", proc_cwd_link),
3343 LNK("root", proc_root_link),
3344 LNK("exe", proc_exe_link),
3345 REG("mounts", S_IRUGO, proc_mounts_operations),
3346 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3347 REG("mountstats", S_IRUSR, proc_mountstats_operations),
3348 #ifdef CONFIG_PROC_PAGE_MONITOR
3349 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3350 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
3351 REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
3352 REG("pagemap", S_IRUSR, proc_pagemap_operations),
3353 #endif
3354 #ifdef CONFIG_SECURITY
3355 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3356 #endif
3357 #ifdef CONFIG_KALLSYMS
3358 ONE("wchan", S_IRUGO, proc_pid_wchan),
3359 #endif
3360 #ifdef CONFIG_STACKTRACE
3361 ONE("stack", S_IRUSR, proc_pid_stack),
3362 #endif
3363 #ifdef CONFIG_SCHED_INFO
3364 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
3365 #endif
3366 #ifdef CONFIG_LATENCYTOP
3367 REG("latency", S_IRUGO, proc_lstats_operations),
3368 #endif
3369 #ifdef CONFIG_PROC_PID_CPUSET
3370 ONE("cpuset", S_IRUGO, proc_cpuset_show),
3371 #endif
3372 #ifdef CONFIG_CGROUPS
3373 ONE("cgroup", S_IRUGO, proc_cgroup_show),
3374 #endif
3375 #ifdef CONFIG_PROC_CPU_RESCTRL
3376 ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show),
3377 #endif
3378 ONE("oom_score", S_IRUGO, proc_oom_score),
3379 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3380 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3381 #ifdef CONFIG_AUDIT
3382 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3383 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3384 #endif
3385 #ifdef CONFIG_FAULT_INJECTION
3386 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3387 REG("fail-nth", 0644, proc_fail_nth_operations),
3388 #endif
3389 #ifdef CONFIG_ELF_CORE
3390 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3391 #endif
3392 #ifdef CONFIG_TASK_IO_ACCOUNTING
3393 ONE("io", S_IRUSR, proc_tgid_io_accounting),
3394 #endif
3395 #ifdef CONFIG_USER_NS
3396 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3397 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3398 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3399 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
3400 #endif
3401 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
3402 REG("timers", S_IRUGO, proc_timers_operations),
3403 #endif
3404 REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations),
3405 #ifdef CONFIG_LIVEPATCH
3406 ONE("patch_state", S_IRUSR, proc_pid_patch_state),
3407 #endif
3408 #ifdef CONFIG_STACKLEAK_METRICS
3409 ONE("stack_depth", S_IRUGO, proc_stack_depth),
3410 #endif
3411 #ifdef CONFIG_PROC_PID_ARCH_STATUS
3412 ONE("arch_status", S_IRUGO, proc_pid_arch_status),
3413 #endif
3414 #ifdef CONFIG_SECCOMP_CACHE_DEBUG
3415 ONE("seccomp_cache", S_IRUSR, proc_pid_seccomp_cache),
3416 #endif
3417 #ifdef CONFIG_KSM
3418 ONE("ksm_merging_pages", S_IRUSR, proc_pid_ksm_merging_pages),
3419 ONE("ksm_stat", S_IRUSR, proc_pid_ksm_stat),
3420 #endif
3421 };
3422
3423 static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
3424 {
3425 return proc_pident_readdir(file, ctx,
3426 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3427 }
3428
3429 static const struct file_operations proc_tgid_base_operations = {
3430 .read = generic_read_dir,
3431 .iterate_shared = proc_tgid_base_readdir,
3432 .llseek = generic_file_llseek,
3433 };
3434
3435 struct pid *tgid_pidfd_to_pid(const struct file *file)
3436 {
3437 if (file->f_op != &proc_tgid_base_operations)
3438 return ERR_PTR(-EBADF);
3439
3440 return proc_pid(file_inode(file));
3441 }
3442
3443 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3444 {
3445 return proc_pident_lookup(dir, dentry,
3446 tgid_base_stuff,
3447 tgid_base_stuff + ARRAY_SIZE(tgid_base_stuff));
3448 }
3449
3450 static const struct inode_operations proc_tgid_base_inode_operations = {
3451 .lookup = proc_tgid_base_lookup,
3452 .getattr = pid_getattr,
3453 .setattr = proc_setattr,
3454 .permission = proc_pid_permission,
3455 };
3456
3457 /**
3458 * proc_flush_pid - Remove dcache entries for @pid from the /proc dcache.
3459 * @pid: pid that should be flushed.
3460 *
3461 * This function walks a list of inodes (that belong to any proc
3462 * filesystem) that are attached to the pid and flushes them from
3463 * the dentry cache.
3464 *
3465 * It is safe and reasonable to cache /proc entries for a task until
3466 * that task exits. After that they just clog up the dcache with
3467 * useless entries, possibly causing useful dcache entries to be
3468 * flushed instead. This routine is provided to flush those useless
3469 * dcache entries when a process is reaped.
3470 *
3471 * NOTE: This routine is just an optimization so it does not guarantee
3472 * that no dcache entries will exist after a process is reaped
3473 * it just makes it very unlikely that any will persist.
3474 */
3475
3476 void proc_flush_pid(struct pid *pid)
3477 {
3478 proc_invalidate_siblings_dcache(&pid->inodes, &pid->lock);
3479 }
3480
3481 static struct dentry *proc_pid_instantiate(struct dentry * dentry,
3482 struct task_struct *task, const void *ptr)
3483 {
3484 struct inode *inode;
3485
3486 inode = proc_pid_make_base_inode(dentry->d_sb, task,
3487 S_IFDIR | S_IRUGO | S_IXUGO);
3488 if (!inode)
3489 return ERR_PTR(-ENOENT);
3490
3491 inode->i_op = &proc_tgid_base_inode_operations;
3492 inode->i_fop = &proc_tgid_base_operations;
3493 inode->i_flags|=S_IMMUTABLE;
3494
3495 set_nlink(inode, nlink_tgid);
3496 pid_update_inode(task, inode);
3497
3498 d_set_d_op(dentry, &pid_dentry_operations);
3499 return d_splice_alias(inode, dentry);
3500 }
3501
3502 struct dentry *proc_pid_lookup(struct dentry *dentry, unsigned int flags)
3503 {
3504 struct task_struct *task;
3505 unsigned tgid;
3506 struct proc_fs_info *fs_info;
3507 struct pid_namespace *ns;
3508 struct dentry *result = ERR_PTR(-ENOENT);
3509
3510 tgid = name_to_int(&dentry->d_name);
3511 if (tgid == ~0U)
3512 goto out;
3513
3514 fs_info = proc_sb_info(dentry->d_sb);
3515 ns = fs_info->pid_ns;
3516 rcu_read_lock();
3517 task = find_task_by_pid_ns(tgid, ns);
3518 if (task)
3519 get_task_struct(task);
3520 rcu_read_unlock();
3521 if (!task)
3522 goto out;
3523
3524 /* Limit procfs to only ptraceable tasks */
3525 if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE) {
3526 if (!has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS))
3527 goto out_put_task;
3528 }
3529
3530 result = proc_pid_instantiate(dentry, task, NULL);
3531 out_put_task:
3532 put_task_struct(task);
3533 out:
3534 return result;
3535 }
3536
3537 /*
3538 * Find the first task with tgid >= tgid
3539 *
3540 */
3541 struct tgid_iter {
3542 unsigned int tgid;
3543 struct task_struct *task;
3544 };
3545 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3546 {
3547 struct pid *pid;
3548
3549 if (iter.task)
3550 put_task_struct(iter.task);
3551 rcu_read_lock();
3552 retry:
3553 iter.task = NULL;
3554 pid = find_ge_pid(iter.tgid, ns);
3555 if (pid) {
3556 iter.tgid = pid_nr_ns(pid, ns);
3557 iter.task = pid_task(pid, PIDTYPE_TGID);
3558 if (!iter.task) {
3559 iter.tgid += 1;
3560 goto retry;
3561 }
3562 get_task_struct(iter.task);
3563 }
3564 rcu_read_unlock();
3565 return iter;
3566 }
3567
3568 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
3569
3570 /* for the /proc/ directory itself, after non-process stuff has been done */
3571 int proc_pid_readdir(struct file *file, struct dir_context *ctx)
3572 {
3573 struct tgid_iter iter;
3574 struct proc_fs_info *fs_info = proc_sb_info(file_inode(file)->i_sb);
3575 struct pid_namespace *ns = proc_pid_ns(file_inode(file)->i_sb);
3576 loff_t pos = ctx->pos;
3577
3578 if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
3579 return 0;
3580
3581 if (pos == TGID_OFFSET - 2) {
3582 struct inode *inode = d_inode(fs_info->proc_self);
3583 if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
3584 return 0;
3585 ctx->pos = pos = pos + 1;
3586 }
3587 if (pos == TGID_OFFSET - 1) {
3588 struct inode *inode = d_inode(fs_info->proc_thread_self);
3589 if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3590 return 0;
3591 ctx->pos = pos = pos + 1;
3592 }
3593 iter.tgid = pos - TGID_OFFSET;
3594 iter.task = NULL;
3595 for (iter = next_tgid(ns, iter);
3596 iter.task;
3597 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3598 char name[10 + 1];
3599 unsigned int len;
3600
3601 cond_resched();
3602 if (!has_pid_permissions(fs_info, iter.task, HIDEPID_INVISIBLE))
3603 continue;
3604
3605 len = snprintf(name, sizeof(name), "%u", iter.tgid);
3606 ctx->pos = iter.tgid + TGID_OFFSET;
3607 if (!proc_fill_cache(file, ctx, name, len,
3608 proc_pid_instantiate, iter.task, NULL)) {
3609 put_task_struct(iter.task);
3610 return 0;
3611 }
3612 }
3613 ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
3614 return 0;
3615 }
3616
3617 /*
3618 * proc_tid_comm_permission is a special permission function exclusively
3619 * used for the node /proc/<pid>/task/<tid>/comm.
3620 * It bypasses generic permission checks in the case where a task of the same
3621 * task group attempts to access the node.
3622 * The rationale behind this is that glibc and bionic access this node for
3623 * cross thread naming (pthread_set/getname_np(!self)). However, if
3624 * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3625 * which locks out the cross thread naming implementation.
3626 * This function makes sure that the node is always accessible for members of
3627 * same thread group.
3628 */
3629 static int proc_tid_comm_permission(struct mnt_idmap *idmap,
3630 struct inode *inode, int mask)
3631 {
3632 bool is_same_tgroup;
3633 struct task_struct *task;
3634
3635 task = get_proc_task(inode);
3636 if (!task)
3637 return -ESRCH;
3638 is_same_tgroup = same_thread_group(current, task);
3639 put_task_struct(task);
3640
3641 if (likely(is_same_tgroup && !(mask & MAY_EXEC))) {
3642 /* This file (/proc/<pid>/task/<tid>/comm) can always be
3643 * read or written by the members of the corresponding
3644 * thread group.
3645 */
3646 return 0;
3647 }
3648
3649 return generic_permission(&nop_mnt_idmap, inode, mask);
3650 }
3651
3652 static const struct inode_operations proc_tid_comm_inode_operations = {
3653 .setattr = proc_setattr,
3654 .permission = proc_tid_comm_permission,
3655 };
3656
3657 /*
3658 * Tasks
3659 */
3660 static const struct pid_entry tid_base_stuff[] = {
3661 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3662 DIR("fdinfo", S_IRUGO|S_IXUGO, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3663 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3664 #ifdef CONFIG_NET
3665 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3666 #endif
3667 REG("environ", S_IRUSR, proc_environ_operations),
3668 REG("auxv", S_IRUSR, proc_auxv_operations),
3669 ONE("status", S_IRUGO, proc_pid_status),
3670 ONE("personality", S_IRUSR, proc_pid_personality),
3671 ONE("limits", S_IRUGO, proc_pid_limits),
3672 #ifdef CONFIG_SCHED_DEBUG
3673 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3674 #endif
3675 NOD("comm", S_IFREG|S_IRUGO|S_IWUSR,
3676 &proc_tid_comm_inode_operations,
3677 &proc_pid_set_comm_operations, {}),
3678 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3679 ONE("syscall", S_IRUSR, proc_pid_syscall),
3680 #endif
3681 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
3682 ONE("stat", S_IRUGO, proc_tid_stat),
3683 ONE("statm", S_IRUGO, proc_pid_statm),
3684 REG("maps", S_IRUGO, proc_pid_maps_operations),
3685 #ifdef CONFIG_PROC_CHILDREN
3686 REG("children", S_IRUGO, proc_tid_children_operations),
3687 #endif
3688 #ifdef CONFIG_NUMA
3689 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
3690 #endif
3691 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3692 LNK("cwd", proc_cwd_link),
3693 LNK("root", proc_root_link),
3694 LNK("exe", proc_exe_link),
3695 REG("mounts", S_IRUGO, proc_mounts_operations),
3696 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3697 #ifdef CONFIG_PROC_PAGE_MONITOR
3698 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3699 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
3700 REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
3701 REG("pagemap", S_IRUSR, proc_pagemap_operations),
3702 #endif
3703 #ifdef CONFIG_SECURITY
3704 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3705 #endif
3706 #ifdef CONFIG_KALLSYMS
3707 ONE("wchan", S_IRUGO, proc_pid_wchan),
3708 #endif
3709 #ifdef CONFIG_STACKTRACE
3710 ONE("stack", S_IRUSR, proc_pid_stack),
3711 #endif
3712 #ifdef CONFIG_SCHED_INFO
3713 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
3714 #endif
3715 #ifdef CONFIG_LATENCYTOP
3716 REG("latency", S_IRUGO, proc_lstats_operations),
3717 #endif
3718 #ifdef CONFIG_PROC_PID_CPUSET
3719 ONE("cpuset", S_IRUGO, proc_cpuset_show),
3720 #endif
3721 #ifdef CONFIG_CGROUPS
3722 ONE("cgroup", S_IRUGO, proc_cgroup_show),
3723 #endif
3724 #ifdef CONFIG_PROC_CPU_RESCTRL
3725 ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show),
3726 #endif
3727 ONE("oom_score", S_IRUGO, proc_oom_score),
3728 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3729 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3730 #ifdef CONFIG_AUDIT
3731 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3732 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3733 #endif
3734 #ifdef CONFIG_FAULT_INJECTION
3735 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3736 REG("fail-nth", 0644, proc_fail_nth_operations),
3737 #endif
3738 #ifdef CONFIG_TASK_IO_ACCOUNTING
3739 ONE("io", S_IRUSR, proc_tid_io_accounting),
3740 #endif
3741 #ifdef CONFIG_USER_NS
3742 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3743 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3744 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3745 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
3746 #endif
3747 #ifdef CONFIG_LIVEPATCH
3748 ONE("patch_state", S_IRUSR, proc_pid_patch_state),
3749 #endif
3750 #ifdef CONFIG_PROC_PID_ARCH_STATUS
3751 ONE("arch_status", S_IRUGO, proc_pid_arch_status),
3752 #endif
3753 #ifdef CONFIG_SECCOMP_CACHE_DEBUG
3754 ONE("seccomp_cache", S_IRUSR, proc_pid_seccomp_cache),
3755 #endif
3756 #ifdef CONFIG_KSM
3757 ONE("ksm_merging_pages", S_IRUSR, proc_pid_ksm_merging_pages),
3758 ONE("ksm_stat", S_IRUSR, proc_pid_ksm_stat),
3759 #endif
3760 };
3761
3762 static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
3763 {
3764 return proc_pident_readdir(file, ctx,
3765 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3766 }
3767
3768 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3769 {
3770 return proc_pident_lookup(dir, dentry,
3771 tid_base_stuff,
3772 tid_base_stuff + ARRAY_SIZE(tid_base_stuff));
3773 }
3774
3775 static const struct file_operations proc_tid_base_operations = {
3776 .read = generic_read_dir,
3777 .iterate_shared = proc_tid_base_readdir,
3778 .llseek = generic_file_llseek,
3779 };
3780
3781 static const struct inode_operations proc_tid_base_inode_operations = {
3782 .lookup = proc_tid_base_lookup,
3783 .getattr = pid_getattr,
3784 .setattr = proc_setattr,
3785 };
3786
3787 static struct dentry *proc_task_instantiate(struct dentry *dentry,
3788 struct task_struct *task, const void *ptr)
3789 {
3790 struct inode *inode;
3791 inode = proc_pid_make_base_inode(dentry->d_sb, task,
3792 S_IFDIR | S_IRUGO | S_IXUGO);
3793 if (!inode)
3794 return ERR_PTR(-ENOENT);
3795
3796 inode->i_op = &proc_tid_base_inode_operations;
3797 inode->i_fop = &proc_tid_base_operations;
3798 inode->i_flags |= S_IMMUTABLE;
3799
3800 set_nlink(inode, nlink_tid);
3801 pid_update_inode(task, inode);
3802
3803 d_set_d_op(dentry, &pid_dentry_operations);
3804 return d_splice_alias(inode, dentry);
3805 }
3806
3807 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3808 {
3809 struct task_struct *task;
3810 struct task_struct *leader = get_proc_task(dir);
3811 unsigned tid;
3812 struct proc_fs_info *fs_info;
3813 struct pid_namespace *ns;
3814 struct dentry *result = ERR_PTR(-ENOENT);
3815
3816 if (!leader)
3817 goto out_no_task;
3818
3819 tid = name_to_int(&dentry->d_name);
3820 if (tid == ~0U)
3821 goto out;
3822
3823 fs_info = proc_sb_info(dentry->d_sb);
3824 ns = fs_info->pid_ns;
3825 rcu_read_lock();
3826 task = find_task_by_pid_ns(tid, ns);
3827 if (task)
3828 get_task_struct(task);
3829 rcu_read_unlock();
3830 if (!task)
3831 goto out;
3832 if (!same_thread_group(leader, task))
3833 goto out_drop_task;
3834
3835 result = proc_task_instantiate(dentry, task, NULL);
3836 out_drop_task:
3837 put_task_struct(task);
3838 out:
3839 put_task_struct(leader);
3840 out_no_task:
3841 return result;
3842 }
3843
3844 /*
3845 * Find the first tid of a thread group to return to user space.
3846 *
3847 * Usually this is just the thread group leader, but if the users
3848 * buffer was too small or there was a seek into the middle of the
3849 * directory we have more work todo.
3850 *
3851 * In the case of a short read we start with find_task_by_pid.
3852 *
3853 * In the case of a seek we start with the leader and walk nr
3854 * threads past it.
3855 */
3856 static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3857 struct pid_namespace *ns)
3858 {
3859 struct task_struct *pos, *task;
3860 unsigned long nr = f_pos;
3861
3862 if (nr != f_pos) /* 32bit overflow? */
3863 return NULL;
3864
3865 rcu_read_lock();
3866 task = pid_task(pid, PIDTYPE_PID);
3867 if (!task)
3868 goto fail;
3869
3870 /* Attempt to start with the tid of a thread */
3871 if (tid && nr) {
3872 pos = find_task_by_pid_ns(tid, ns);
3873 if (pos && same_thread_group(pos, task))
3874 goto found;
3875 }
3876
3877 /* If nr exceeds the number of threads there is nothing todo */
3878 if (nr >= get_nr_threads(task))
3879 goto fail;
3880
3881 /* If we haven't found our starting place yet start
3882 * with the leader and walk nr threads forward.
3883 */
3884 for_each_thread(task, pos) {
3885 if (!nr--)
3886 goto found;
3887 }
3888 fail:
3889 pos = NULL;
3890 goto out;
3891 found:
3892 get_task_struct(pos);
3893 out:
3894 rcu_read_unlock();
3895 return pos;
3896 }
3897
3898 /*
3899 * Find the next thread in the thread list.
3900 * Return NULL if there is an error or no next thread.
3901 *
3902 * The reference to the input task_struct is released.
3903 */
3904 static struct task_struct *next_tid(struct task_struct *start)
3905 {
3906 struct task_struct *pos = NULL;
3907 rcu_read_lock();
3908 if (pid_alive(start)) {
3909 pos = __next_thread(start);
3910 if (pos)
3911 get_task_struct(pos);
3912 }
3913 rcu_read_unlock();
3914 put_task_struct(start);
3915 return pos;
3916 }
3917
3918 /* for the /proc/TGID/task/ directories */
3919 static int proc_task_readdir(struct file *file, struct dir_context *ctx)
3920 {
3921 struct inode *inode = file_inode(file);
3922 struct task_struct *task;
3923 struct pid_namespace *ns;
3924 int tid;
3925
3926 if (proc_inode_is_dead(inode))
3927 return -ENOENT;
3928
3929 if (!dir_emit_dots(file, ctx))
3930 return 0;
3931
3932 /* f_version caches the tgid value that the last readdir call couldn't
3933 * return. lseek aka telldir automagically resets f_version to 0.
3934 */
3935 ns = proc_pid_ns(inode->i_sb);
3936 tid = (int)file->f_version;
3937 file->f_version = 0;
3938 for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
3939 task;
3940 task = next_tid(task), ctx->pos++) {
3941 char name[10 + 1];
3942 unsigned int len;
3943
3944 tid = task_pid_nr_ns(task, ns);
3945 if (!tid)
3946 continue; /* The task has just exited. */
3947 len = snprintf(name, sizeof(name), "%u", tid);
3948 if (!proc_fill_cache(file, ctx, name, len,
3949 proc_task_instantiate, task, NULL)) {
3950 /* returning this tgid failed, save it as the first
3951 * pid for the next readir call */
3952 file->f_version = (u64)tid;
3953 put_task_struct(task);
3954 break;
3955 }
3956 }
3957
3958 return 0;
3959 }
3960
3961 static int proc_task_getattr(struct mnt_idmap *idmap,
3962 const struct path *path, struct kstat *stat,
3963 u32 request_mask, unsigned int query_flags)
3964 {
3965 struct inode *inode = d_inode(path->dentry);
3966 struct task_struct *p = get_proc_task(inode);
3967 generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat);
3968
3969 if (p) {
3970 stat->nlink += get_nr_threads(p);
3971 put_task_struct(p);
3972 }
3973
3974 return 0;
3975 }
3976
3977 static const struct inode_operations proc_task_inode_operations = {
3978 .lookup = proc_task_lookup,
3979 .getattr = proc_task_getattr,
3980 .setattr = proc_setattr,
3981 .permission = proc_pid_permission,
3982 };
3983
3984 static const struct file_operations proc_task_operations = {
3985 .read = generic_read_dir,
3986 .iterate_shared = proc_task_readdir,
3987 .llseek = generic_file_llseek,
3988 };
3989
3990 void __init set_proc_pid_nlink(void)
3991 {
3992 nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3993 nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3994 }
3995
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.