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/pgsize_migration.h>
69 #include <linux/swap.h>
70 #include <linux/rcupdate.h>
71 #include <linux/kallsyms.h>
72 #include <linux/stacktrace.h>
73 #include <linux/resource.h>
74 #include <linux/module.h>
75 #include <linux/mount.h>
76 #include <linux/security.h>
77 #include <linux/ptrace.h>
78 #include <linux/printk.h>
79 #include <linux/cache.h>
80 #include <linux/cgroup.h>
81 #include <linux/cpuset.h>
82 #include <linux/audit.h>
83 #include <linux/poll.h>
84 #include <linux/nsproxy.h>
85 #include <linux/oom.h>
86 #include <linux/elf.h>
87 #include <linux/pid_namespace.h>
88 #include <linux/user_namespace.h>
89 #include <linux/fs_parser.h>
90 #include <linux/fs_struct.h>
91 #include <linux/slab.h>
92 #include <linux/sched/autogroup.h>
93 #include <linux/sched/mm.h>
94 #include <linux/sched/coredump.h>
95 #include <linux/sched/debug.h>
96 #include <linux/sched/stat.h>
97 #include <linux/posix-timers.h>
98 #include <linux/time_namespace.h>
99 #include <linux/resctrl.h>
100 #include <linux/cn_proc.h>
101 #include <linux/ksm.h>
102 #include <linux/cpufreq_times.h>
103 #include <uapi/linux/lsm.h>
104 #include <trace/events/oom.h>
105 #include <trace/hooks/sched.h>
106 #include "internal.h"
107 #include "fd.h"
108
109 #include "../../lib/kstrtox.h"
110
111 /* NOTE:
112 * Implementing inode permission operations in /proc is almost
113 * certainly an error. Permission checks need to happen during
114 * each system call not at open time. The reason is that most of
115 * what we wish to check for permissions in /proc varies at runtime.
116 *
117 * The classic example of a problem is opening file descriptors
118 * in /proc for a task before it execs a suid executable.
119 */
120
121 static u8 nlink_tid __ro_after_init;
122 static u8 nlink_tgid __ro_after_init;
123
124 enum proc_mem_force {
125 PROC_MEM_FORCE_ALWAYS,
126 PROC_MEM_FORCE_PTRACE,
127 PROC_MEM_FORCE_NEVER
128 };
129
130 static enum proc_mem_force proc_mem_force_override __ro_after_init =
131 IS_ENABLED(CONFIG_PROC_MEM_NO_FORCE) ? PROC_MEM_FORCE_NEVER :
132 IS_ENABLED(CONFIG_PROC_MEM_FORCE_PTRACE) ? PROC_MEM_FORCE_PTRACE :
133 PROC_MEM_FORCE_ALWAYS;
134
135 static const struct constant_table proc_mem_force_table[] __initconst = {
136 { "always", PROC_MEM_FORCE_ALWAYS },
137 { "ptrace", PROC_MEM_FORCE_PTRACE },
138 { "never", PROC_MEM_FORCE_NEVER },
139 { }
140 };
141
early_proc_mem_force_override(char * buf)142 static int __init early_proc_mem_force_override(char *buf)
143 {
144 if (!buf)
145 return -EINVAL;
146
147 /*
148 * lookup_constant() defaults to proc_mem_force_override to preseve
149 * the initial Kconfig choice in case an invalid param gets passed.
150 */
151 proc_mem_force_override = lookup_constant(proc_mem_force_table,
152 buf, proc_mem_force_override);
153
154 return 0;
155 }
156 early_param("proc_mem.force_override", early_proc_mem_force_override);
157
158 struct pid_entry {
159 const char *name;
160 unsigned int len;
161 umode_t mode;
162 const struct inode_operations *iop;
163 const struct file_operations *fop;
164 union proc_op op;
165 };
166
167 #define NOD(NAME, MODE, IOP, FOP, OP) { \
168 .name = (NAME), \
169 .len = sizeof(NAME) - 1, \
170 .mode = MODE, \
171 .iop = IOP, \
172 .fop = FOP, \
173 .op = OP, \
174 }
175
176 #define DIR(NAME, MODE, iops, fops) \
177 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
178 #define LNK(NAME, get_link) \
179 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
180 &proc_pid_link_inode_operations, NULL, \
181 { .proc_get_link = get_link } )
182 #define REG(NAME, MODE, fops) \
183 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
184 #define ONE(NAME, MODE, show) \
185 NOD(NAME, (S_IFREG|(MODE)), \
186 NULL, &proc_single_file_operations, \
187 { .proc_show = show } )
188 #define ATTR(LSMID, NAME, MODE) \
189 NOD(NAME, (S_IFREG|(MODE)), \
190 NULL, &proc_pid_attr_operations, \
191 { .lsmid = LSMID })
192
193 /*
194 * Count the number of hardlinks for the pid_entry table, excluding the .
195 * and .. links.
196 */
pid_entry_nlink(const struct pid_entry * entries,unsigned int n)197 static unsigned int __init pid_entry_nlink(const struct pid_entry *entries,
198 unsigned int n)
199 {
200 unsigned int i;
201 unsigned int count;
202
203 count = 2;
204 for (i = 0; i < n; ++i) {
205 if (S_ISDIR(entries[i].mode))
206 ++count;
207 }
208
209 return count;
210 }
211
get_task_root(struct task_struct * task,struct path * root)212 static int get_task_root(struct task_struct *task, struct path *root)
213 {
214 int result = -ENOENT;
215
216 task_lock(task);
217 if (task->fs) {
218 get_fs_root(task->fs, root);
219 result = 0;
220 }
221 task_unlock(task);
222 return result;
223 }
224
proc_cwd_link(struct dentry * dentry,struct path * path)225 static int proc_cwd_link(struct dentry *dentry, struct path *path)
226 {
227 struct task_struct *task = get_proc_task(d_inode(dentry));
228 int result = -ENOENT;
229
230 if (task) {
231 task_lock(task);
232 if (task->fs) {
233 get_fs_pwd(task->fs, path);
234 result = 0;
235 }
236 task_unlock(task);
237 put_task_struct(task);
238 }
239 return result;
240 }
241
proc_root_link(struct dentry * dentry,struct path * path)242 static int proc_root_link(struct dentry *dentry, struct path *path)
243 {
244 struct task_struct *task = get_proc_task(d_inode(dentry));
245 int result = -ENOENT;
246
247 if (task) {
248 result = get_task_root(task, path);
249 put_task_struct(task);
250 }
251 return result;
252 }
253
254 /*
255 * If the user used setproctitle(), we just get the string from
256 * user space at arg_start, and limit it to a maximum of one page.
257 */
get_mm_proctitle(struct mm_struct * mm,char __user * buf,size_t count,unsigned long pos,unsigned long arg_start)258 static ssize_t get_mm_proctitle(struct mm_struct *mm, char __user *buf,
259 size_t count, unsigned long pos,
260 unsigned long arg_start)
261 {
262 char *page;
263 int ret, got;
264
265 if (pos >= PAGE_SIZE)
266 return 0;
267
268 page = (char *)__get_free_page(GFP_KERNEL);
269 if (!page)
270 return -ENOMEM;
271
272 ret = 0;
273 got = access_remote_vm(mm, arg_start, page, PAGE_SIZE, FOLL_ANON);
274 if (got > 0) {
275 int len = strnlen(page, got);
276
277 /* Include the NUL character if it was found */
278 if (len < got)
279 len++;
280
281 if (len > pos) {
282 len -= pos;
283 if (len > count)
284 len = count;
285 len -= copy_to_user(buf, page+pos, len);
286 if (!len)
287 len = -EFAULT;
288 ret = len;
289 }
290 }
291 free_page((unsigned long)page);
292 return ret;
293 }
294
get_mm_cmdline(struct mm_struct * mm,char __user * buf,size_t count,loff_t * ppos)295 static ssize_t get_mm_cmdline(struct mm_struct *mm, char __user *buf,
296 size_t count, loff_t *ppos)
297 {
298 unsigned long arg_start, arg_end, env_start, env_end;
299 unsigned long pos, len;
300 char *page, c;
301
302 /* Check if process spawned far enough to have cmdline. */
303 if (!mm->env_end)
304 return 0;
305
306 spin_lock(&mm->arg_lock);
307 arg_start = mm->arg_start;
308 arg_end = mm->arg_end;
309 env_start = mm->env_start;
310 env_end = mm->env_end;
311 spin_unlock(&mm->arg_lock);
312
313 if (arg_start >= arg_end)
314 return 0;
315
316 /*
317 * We allow setproctitle() to overwrite the argument
318 * strings, and overflow past the original end. But
319 * only when it overflows into the environment area.
320 */
321 if (env_start != arg_end || env_end < env_start)
322 env_start = env_end = arg_end;
323 len = env_end - arg_start;
324
325 /* We're not going to care if "*ppos" has high bits set */
326 pos = *ppos;
327 if (pos >= len)
328 return 0;
329 if (count > len - pos)
330 count = len - pos;
331 if (!count)
332 return 0;
333
334 /*
335 * Magical special case: if the argv[] end byte is not
336 * zero, the user has overwritten it with setproctitle(3).
337 *
338 * Possible future enhancement: do this only once when
339 * pos is 0, and set a flag in the 'struct file'.
340 */
341 if (access_remote_vm(mm, arg_end-1, &c, 1, FOLL_ANON) == 1 && c)
342 return get_mm_proctitle(mm, buf, count, pos, arg_start);
343
344 /*
345 * For the non-setproctitle() case we limit things strictly
346 * to the [arg_start, arg_end[ range.
347 */
348 pos += arg_start;
349 if (pos < arg_start || pos >= arg_end)
350 return 0;
351 if (count > arg_end - pos)
352 count = arg_end - pos;
353
354 page = (char *)__get_free_page(GFP_KERNEL);
355 if (!page)
356 return -ENOMEM;
357
358 len = 0;
359 while (count) {
360 int got;
361 size_t size = min_t(size_t, PAGE_SIZE, count);
362
363 got = access_remote_vm(mm, pos, page, size, FOLL_ANON);
364 if (got <= 0)
365 break;
366 got -= copy_to_user(buf, page, got);
367 if (unlikely(!got)) {
368 if (!len)
369 len = -EFAULT;
370 break;
371 }
372 pos += got;
373 buf += got;
374 len += got;
375 count -= got;
376 }
377
378 free_page((unsigned long)page);
379 return len;
380 }
381
get_task_cmdline(struct task_struct * tsk,char __user * buf,size_t count,loff_t * pos)382 static ssize_t get_task_cmdline(struct task_struct *tsk, char __user *buf,
383 size_t count, loff_t *pos)
384 {
385 struct mm_struct *mm;
386 bool prio_inherited = false;
387 int saved_prio;
388 ssize_t ret;
389
390 mm = get_task_mm(tsk);
391 if (!mm)
392 return 0;
393
394 /*
395 * access_remote_vm() holds the hot mmap_sem lock which can cause the
396 * task for which we read cmdline etc for by some debug deamon to slow
397 * down and suffer a performance hit. Especially if the reader task has
398 * a low nice value.
399 */
400 trace_android_vh_prio_inheritance(tsk, &saved_prio, &prio_inherited);
401 ret = get_mm_cmdline(mm, buf, count, pos);
402 if (prio_inherited)
403 trace_android_vh_prio_restore(saved_prio);
404 mmput(mm);
405 return ret;
406 }
407
proc_pid_cmdline_read(struct file * file,char __user * buf,size_t count,loff_t * pos)408 static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,
409 size_t count, loff_t *pos)
410 {
411 struct task_struct *tsk;
412 ssize_t ret;
413
414 BUG_ON(*pos < 0);
415
416 tsk = get_proc_task(file_inode(file));
417 if (!tsk)
418 return -ESRCH;
419 ret = get_task_cmdline(tsk, buf, count, pos);
420 put_task_struct(tsk);
421 if (ret > 0)
422 *pos += ret;
423 return ret;
424 }
425
426 static const struct file_operations proc_pid_cmdline_ops = {
427 .read = proc_pid_cmdline_read,
428 .llseek = generic_file_llseek,
429 };
430
431 #ifdef CONFIG_KALLSYMS
432 /*
433 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
434 * Returns the resolved symbol to user space.
435 */
proc_pid_wchan(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)436 static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns,
437 struct pid *pid, struct task_struct *task)
438 {
439 unsigned long wchan;
440 char symname[KSYM_NAME_LEN];
441
442 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
443 goto print0;
444
445 wchan = get_wchan(task);
446 if (wchan && !lookup_symbol_name(wchan, symname)) {
447 seq_puts(m, symname);
448 return 0;
449 }
450
451 print0:
452 seq_putc(m, '0');
453 return 0;
454 }
455 #endif /* CONFIG_KALLSYMS */
456
lock_trace(struct task_struct * task)457 static int lock_trace(struct task_struct *task)
458 {
459 int err = down_read_killable(&task->signal->exec_update_lock);
460 if (err)
461 return err;
462 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) {
463 up_read(&task->signal->exec_update_lock);
464 return -EPERM;
465 }
466 return 0;
467 }
468
unlock_trace(struct task_struct * task)469 static void unlock_trace(struct task_struct *task)
470 {
471 up_read(&task->signal->exec_update_lock);
472 }
473
474 #ifdef CONFIG_STACKTRACE
475
476 #define MAX_STACK_TRACE_DEPTH 64
477
proc_pid_stack(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)478 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
479 struct pid *pid, struct task_struct *task)
480 {
481 unsigned long *entries;
482 int err;
483
484 /*
485 * The ability to racily run the kernel stack unwinder on a running task
486 * and then observe the unwinder output is scary; while it is useful for
487 * debugging kernel issues, it can also allow an attacker to leak kernel
488 * stack contents.
489 * Doing this in a manner that is at least safe from races would require
490 * some work to ensure that the remote task can not be scheduled; and
491 * even then, this would still expose the unwinder as local attack
492 * surface.
493 * Therefore, this interface is restricted to root.
494 */
495 if (!file_ns_capable(m->file, &init_user_ns, CAP_SYS_ADMIN))
496 return -EACCES;
497
498 entries = kmalloc_array(MAX_STACK_TRACE_DEPTH, sizeof(*entries),
499 GFP_KERNEL);
500 if (!entries)
501 return -ENOMEM;
502
503 err = lock_trace(task);
504 if (!err) {
505 unsigned int i, nr_entries;
506
507 nr_entries = stack_trace_save_tsk(task, entries,
508 MAX_STACK_TRACE_DEPTH, 0);
509
510 for (i = 0; i < nr_entries; i++) {
511 seq_printf(m, "[<0>] %pB\n", (void *)entries[i]);
512 }
513
514 unlock_trace(task);
515 }
516 kfree(entries);
517
518 return err;
519 }
520 #endif
521
522 #ifdef CONFIG_SCHED_INFO
523 /*
524 * Provides /proc/PID/schedstat
525 */
proc_pid_schedstat(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)526 static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
527 struct pid *pid, struct task_struct *task)
528 {
529 if (unlikely(!sched_info_on()))
530 seq_puts(m, "0 0 0\n");
531 else
532 seq_printf(m, "%llu %llu %lu\n",
533 (unsigned long long)task->se.sum_exec_runtime,
534 (unsigned long long)task->sched_info.run_delay,
535 task->sched_info.pcount);
536
537 return 0;
538 }
539 #endif
540
541 #ifdef CONFIG_LATENCYTOP
lstats_show_proc(struct seq_file * m,void * v)542 static int lstats_show_proc(struct seq_file *m, void *v)
543 {
544 int i;
545 struct inode *inode = m->private;
546 struct task_struct *task = get_proc_task(inode);
547
548 if (!task)
549 return -ESRCH;
550 seq_puts(m, "Latency Top version : v0.1\n");
551 for (i = 0; i < LT_SAVECOUNT; i++) {
552 struct latency_record *lr = &task->latency_record[i];
553 if (lr->backtrace[0]) {
554 int q;
555 seq_printf(m, "%i %li %li",
556 lr->count, lr->time, lr->max);
557 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
558 unsigned long bt = lr->backtrace[q];
559
560 if (!bt)
561 break;
562 seq_printf(m, " %ps", (void *)bt);
563 }
564 seq_putc(m, '\n');
565 }
566
567 }
568 put_task_struct(task);
569 return 0;
570 }
571
lstats_open(struct inode * inode,struct file * file)572 static int lstats_open(struct inode *inode, struct file *file)
573 {
574 return single_open(file, lstats_show_proc, inode);
575 }
576
lstats_write(struct file * file,const char __user * buf,size_t count,loff_t * offs)577 static ssize_t lstats_write(struct file *file, const char __user *buf,
578 size_t count, loff_t *offs)
579 {
580 struct task_struct *task = get_proc_task(file_inode(file));
581
582 if (!task)
583 return -ESRCH;
584 clear_tsk_latency_tracing(task);
585 put_task_struct(task);
586
587 return count;
588 }
589
590 static const struct file_operations proc_lstats_operations = {
591 .open = lstats_open,
592 .read = seq_read,
593 .write = lstats_write,
594 .llseek = seq_lseek,
595 .release = single_release,
596 };
597
598 #endif
599
proc_oom_score(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)600 static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
601 struct pid *pid, struct task_struct *task)
602 {
603 unsigned long totalpages = totalram_pages() + total_swap_pages;
604 unsigned long points = 0;
605 long badness;
606
607 badness = oom_badness(task, totalpages);
608 /*
609 * Special case OOM_SCORE_ADJ_MIN for all others scale the
610 * badness value into [0, 2000] range which we have been
611 * exporting for a long time so userspace might depend on it.
612 */
613 if (badness != LONG_MIN)
614 points = (1000 + badness * 1000 / (long)totalpages) * 2 / 3;
615
616 seq_printf(m, "%lu\n", points);
617
618 return 0;
619 }
620
621 struct limit_names {
622 const char *name;
623 const char *unit;
624 };
625
626 static const struct limit_names lnames[RLIM_NLIMITS] = {
627 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
628 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
629 [RLIMIT_DATA] = {"Max data size", "bytes"},
630 [RLIMIT_STACK] = {"Max stack size", "bytes"},
631 [RLIMIT_CORE] = {"Max core file size", "bytes"},
632 [RLIMIT_RSS] = {"Max resident set", "bytes"},
633 [RLIMIT_NPROC] = {"Max processes", "processes"},
634 [RLIMIT_NOFILE] = {"Max open files", "files"},
635 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
636 [RLIMIT_AS] = {"Max address space", "bytes"},
637 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
638 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
639 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
640 [RLIMIT_NICE] = {"Max nice priority", NULL},
641 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
642 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
643 };
644
645 /* Display limits for a process */
proc_pid_limits(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)646 static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
647 struct pid *pid, struct task_struct *task)
648 {
649 unsigned int i;
650 unsigned long flags;
651
652 struct rlimit rlim[RLIM_NLIMITS];
653
654 if (!lock_task_sighand(task, &flags))
655 return 0;
656 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
657 unlock_task_sighand(task, &flags);
658
659 /*
660 * print the file header
661 */
662 seq_puts(m, "Limit "
663 "Soft Limit "
664 "Hard Limit "
665 "Units \n");
666
667 for (i = 0; i < RLIM_NLIMITS; i++) {
668 if (rlim[i].rlim_cur == RLIM_INFINITY)
669 seq_printf(m, "%-25s %-20s ",
670 lnames[i].name, "unlimited");
671 else
672 seq_printf(m, "%-25s %-20lu ",
673 lnames[i].name, rlim[i].rlim_cur);
674
675 if (rlim[i].rlim_max == RLIM_INFINITY)
676 seq_printf(m, "%-20s ", "unlimited");
677 else
678 seq_printf(m, "%-20lu ", rlim[i].rlim_max);
679
680 if (lnames[i].unit)
681 seq_printf(m, "%-10s\n", lnames[i].unit);
682 else
683 seq_putc(m, '\n');
684 }
685
686 return 0;
687 }
688
689 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
proc_pid_syscall(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)690 static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
691 struct pid *pid, struct task_struct *task)
692 {
693 struct syscall_info info;
694 u64 *args = &info.data.args[0];
695 int res;
696
697 res = lock_trace(task);
698 if (res)
699 return res;
700
701 if (task_current_syscall(task, &info))
702 seq_puts(m, "running\n");
703 else if (info.data.nr < 0)
704 seq_printf(m, "%d 0x%llx 0x%llx\n",
705 info.data.nr, info.sp, info.data.instruction_pointer);
706 else
707 seq_printf(m,
708 "%d 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx\n",
709 info.data.nr,
710 args[0], args[1], args[2], args[3], args[4], args[5],
711 info.sp, info.data.instruction_pointer);
712 unlock_trace(task);
713
714 return 0;
715 }
716 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
717
718 /************************************************************************/
719 /* Here the fs part begins */
720 /************************************************************************/
721
722 /* permission checks */
proc_fd_access_allowed(struct inode * inode)723 static bool proc_fd_access_allowed(struct inode *inode)
724 {
725 struct task_struct *task;
726 bool allowed = false;
727 /* Allow access to a task's file descriptors if it is us or we
728 * may use ptrace attach to the process and find out that
729 * information.
730 */
731 task = get_proc_task(inode);
732 if (task) {
733 allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
734 put_task_struct(task);
735 }
736 return allowed;
737 }
738
proc_setattr(struct mnt_idmap * idmap,struct dentry * dentry,struct iattr * attr)739 int proc_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
740 struct iattr *attr)
741 {
742 int error;
743 struct inode *inode = d_inode(dentry);
744
745 if (attr->ia_valid & ATTR_MODE)
746 return -EPERM;
747
748 error = setattr_prepare(&nop_mnt_idmap, dentry, attr);
749 if (error)
750 return error;
751
752 setattr_copy(&nop_mnt_idmap, inode, attr);
753 return 0;
754 }
755
756 /*
757 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
758 * or euid/egid (for hide_pid_min=2)?
759 */
has_pid_permissions(struct proc_fs_info * fs_info,struct task_struct * task,enum proc_hidepid hide_pid_min)760 static bool has_pid_permissions(struct proc_fs_info *fs_info,
761 struct task_struct *task,
762 enum proc_hidepid hide_pid_min)
763 {
764 /*
765 * If 'hidpid' mount option is set force a ptrace check,
766 * we indicate that we are using a filesystem syscall
767 * by passing PTRACE_MODE_READ_FSCREDS
768 */
769 if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE)
770 return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
771
772 if (fs_info->hide_pid < hide_pid_min)
773 return true;
774 if (in_group_p(fs_info->pid_gid))
775 return true;
776 return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
777 }
778
779
proc_pid_permission(struct mnt_idmap * idmap,struct inode * inode,int mask)780 static int proc_pid_permission(struct mnt_idmap *idmap,
781 struct inode *inode, int mask)
782 {
783 struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
784 struct task_struct *task;
785 bool has_perms;
786
787 task = get_proc_task(inode);
788 if (!task)
789 return -ESRCH;
790 has_perms = has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS);
791 put_task_struct(task);
792
793 if (!has_perms) {
794 if (fs_info->hide_pid == HIDEPID_INVISIBLE) {
795 /*
796 * Let's make getdents(), stat(), and open()
797 * consistent with each other. If a process
798 * may not stat() a file, it shouldn't be seen
799 * in procfs at all.
800 */
801 return -ENOENT;
802 }
803
804 return -EPERM;
805 }
806 return generic_permission(&nop_mnt_idmap, inode, mask);
807 }
808
809
810
811 static const struct inode_operations proc_def_inode_operations = {
812 .setattr = proc_setattr,
813 };
814
proc_single_show(struct seq_file * m,void * v)815 static int proc_single_show(struct seq_file *m, void *v)
816 {
817 struct inode *inode = m->private;
818 struct pid_namespace *ns = proc_pid_ns(inode->i_sb);
819 struct pid *pid = proc_pid(inode);
820 struct task_struct *task;
821 int ret;
822
823 task = get_pid_task(pid, PIDTYPE_PID);
824 if (!task)
825 return -ESRCH;
826
827 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
828
829 put_task_struct(task);
830 return ret;
831 }
832
proc_single_open(struct inode * inode,struct file * filp)833 static int proc_single_open(struct inode *inode, struct file *filp)
834 {
835 return single_open(filp, proc_single_show, inode);
836 }
837
838 static const struct file_operations proc_single_file_operations = {
839 .open = proc_single_open,
840 .read = seq_read,
841 .llseek = seq_lseek,
842 .release = single_release,
843 };
844
845
proc_mem_open(struct inode * inode,unsigned int mode)846 struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
847 {
848 struct task_struct *task = get_proc_task(inode);
849 struct mm_struct *mm = ERR_PTR(-ESRCH);
850
851 if (task) {
852 mm = mm_access(task, mode | PTRACE_MODE_FSCREDS);
853 put_task_struct(task);
854
855 if (!IS_ERR_OR_NULL(mm)) {
856 /* ensure this mm_struct can't be freed */
857 mmgrab(mm);
858 /* but do not pin its memory */
859 mmput(mm);
860 }
861 }
862
863 return mm;
864 }
865
__mem_open(struct inode * inode,struct file * file,unsigned int mode)866 static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
867 {
868 struct mm_struct *mm = proc_mem_open(inode, mode);
869
870 if (IS_ERR(mm))
871 return PTR_ERR(mm);
872
873 file->private_data = mm;
874 return 0;
875 }
876
mem_open(struct inode * inode,struct file * file)877 static int mem_open(struct inode *inode, struct file *file)
878 {
879 if (WARN_ON_ONCE(!(file->f_op->fop_flags & FOP_UNSIGNED_OFFSET)))
880 return -EINVAL;
881 return __mem_open(inode, file, PTRACE_MODE_ATTACH);
882 }
883
proc_mem_foll_force(struct file * file,struct mm_struct * mm)884 static bool proc_mem_foll_force(struct file *file, struct mm_struct *mm)
885 {
886 struct task_struct *task;
887 bool ptrace_active = false;
888
889 switch (proc_mem_force_override) {
890 case PROC_MEM_FORCE_NEVER:
891 return false;
892 case PROC_MEM_FORCE_PTRACE:
893 task = get_proc_task(file_inode(file));
894 if (task) {
895 ptrace_active = READ_ONCE(task->ptrace) &&
896 READ_ONCE(task->mm) == mm &&
897 READ_ONCE(task->parent) == current;
898 put_task_struct(task);
899 }
900 return ptrace_active;
901 default:
902 return true;
903 }
904 }
905
mem_rw(struct file * file,char __user * buf,size_t count,loff_t * ppos,int write)906 static ssize_t mem_rw(struct file *file, char __user *buf,
907 size_t count, loff_t *ppos, int write)
908 {
909 struct mm_struct *mm = file->private_data;
910 unsigned long addr = *ppos;
911 ssize_t copied;
912 char *page;
913 unsigned int flags;
914
915 if (!mm)
916 return 0;
917
918 page = (char *)__get_free_page(GFP_KERNEL);
919 if (!page)
920 return -ENOMEM;
921
922 copied = 0;
923 if (!mmget_not_zero(mm))
924 goto free;
925
926 flags = write ? FOLL_WRITE : 0;
927 if (proc_mem_foll_force(file, mm))
928 flags |= FOLL_FORCE;
929
930 while (count > 0) {
931 size_t this_len = min_t(size_t, count, PAGE_SIZE);
932
933 if (write && copy_from_user(page, buf, this_len)) {
934 copied = -EFAULT;
935 break;
936 }
937
938 this_len = access_remote_vm(mm, addr, page, this_len, flags);
939 if (!this_len) {
940 if (!copied)
941 copied = -EIO;
942 break;
943 }
944
945 if (!write && copy_to_user(buf, page, this_len)) {
946 copied = -EFAULT;
947 break;
948 }
949
950 buf += this_len;
951 addr += this_len;
952 copied += this_len;
953 count -= this_len;
954 }
955 *ppos = addr;
956
957 mmput(mm);
958 free:
959 free_page((unsigned long) page);
960 return copied;
961 }
962
mem_read(struct file * file,char __user * buf,size_t count,loff_t * ppos)963 static ssize_t mem_read(struct file *file, char __user *buf,
964 size_t count, loff_t *ppos)
965 {
966 return mem_rw(file, buf, count, ppos, 0);
967 }
968
mem_write(struct file * file,const char __user * buf,size_t count,loff_t * ppos)969 static ssize_t mem_write(struct file *file, const char __user *buf,
970 size_t count, loff_t *ppos)
971 {
972 return mem_rw(file, (char __user*)buf, count, ppos, 1);
973 }
974
mem_lseek(struct file * file,loff_t offset,int orig)975 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
976 {
977 switch (orig) {
978 case 0:
979 file->f_pos = offset;
980 break;
981 case 1:
982 file->f_pos += offset;
983 break;
984 default:
985 return -EINVAL;
986 }
987 force_successful_syscall_return();
988 return file->f_pos;
989 }
990
mem_release(struct inode * inode,struct file * file)991 static int mem_release(struct inode *inode, struct file *file)
992 {
993 struct mm_struct *mm = file->private_data;
994 if (mm)
995 mmdrop(mm);
996 return 0;
997 }
998
999 static const struct file_operations proc_mem_operations = {
1000 .llseek = mem_lseek,
1001 .read = mem_read,
1002 .write = mem_write,
1003 .open = mem_open,
1004 .release = mem_release,
1005 .fop_flags = FOP_UNSIGNED_OFFSET,
1006 };
1007
environ_open(struct inode * inode,struct file * file)1008 static int environ_open(struct inode *inode, struct file *file)
1009 {
1010 return __mem_open(inode, file, PTRACE_MODE_READ);
1011 }
1012
environ_read(struct file * file,char __user * buf,size_t count,loff_t * ppos)1013 static ssize_t environ_read(struct file *file, char __user *buf,
1014 size_t count, loff_t *ppos)
1015 {
1016 char *page;
1017 unsigned long src = *ppos;
1018 int ret = 0;
1019 struct mm_struct *mm = file->private_data;
1020 unsigned long env_start, env_end;
1021
1022 /* Ensure the process spawned far enough to have an environment. */
1023 if (!mm || !mm->env_end)
1024 return 0;
1025
1026 page = (char *)__get_free_page(GFP_KERNEL);
1027 if (!page)
1028 return -ENOMEM;
1029
1030 ret = 0;
1031 if (!mmget_not_zero(mm))
1032 goto free;
1033
1034 spin_lock(&mm->arg_lock);
1035 env_start = mm->env_start;
1036 env_end = mm->env_end;
1037 spin_unlock(&mm->arg_lock);
1038
1039 while (count > 0) {
1040 size_t this_len, max_len;
1041 int retval;
1042
1043 if (src >= (env_end - env_start))
1044 break;
1045
1046 this_len = env_end - (env_start + src);
1047
1048 max_len = min_t(size_t, PAGE_SIZE, count);
1049 this_len = min(max_len, this_len);
1050
1051 retval = access_remote_vm(mm, (env_start + src), page, this_len, FOLL_ANON);
1052
1053 if (retval <= 0) {
1054 ret = retval;
1055 break;
1056 }
1057
1058 if (copy_to_user(buf, page, retval)) {
1059 ret = -EFAULT;
1060 break;
1061 }
1062
1063 ret += retval;
1064 src += retval;
1065 buf += retval;
1066 count -= retval;
1067 }
1068 *ppos = src;
1069 mmput(mm);
1070
1071 free:
1072 free_page((unsigned long) page);
1073 return ret;
1074 }
1075
1076 static const struct file_operations proc_environ_operations = {
1077 .open = environ_open,
1078 .read = environ_read,
1079 .llseek = generic_file_llseek,
1080 .release = mem_release,
1081 };
1082
auxv_open(struct inode * inode,struct file * file)1083 static int auxv_open(struct inode *inode, struct file *file)
1084 {
1085 return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS);
1086 }
1087
auxv_read(struct file * file,char __user * buf,size_t count,loff_t * ppos)1088 static ssize_t auxv_read(struct file *file, char __user *buf,
1089 size_t count, loff_t *ppos)
1090 {
1091 struct mm_struct *mm = file->private_data;
1092 unsigned int nwords = 0;
1093
1094 if (!mm)
1095 return 0;
1096 do {
1097 nwords += 2;
1098 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
1099 return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv,
1100 nwords * sizeof(mm->saved_auxv[0]));
1101 }
1102
1103 static const struct file_operations proc_auxv_operations = {
1104 .open = auxv_open,
1105 .read = auxv_read,
1106 .llseek = generic_file_llseek,
1107 .release = mem_release,
1108 };
1109
oom_adj_read(struct file * file,char __user * buf,size_t count,loff_t * ppos)1110 static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
1111 loff_t *ppos)
1112 {
1113 struct task_struct *task = get_proc_task(file_inode(file));
1114 char buffer[PROC_NUMBUF];
1115 int oom_adj = OOM_ADJUST_MIN;
1116 size_t len;
1117
1118 if (!task)
1119 return -ESRCH;
1120 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
1121 oom_adj = OOM_ADJUST_MAX;
1122 else
1123 oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
1124 OOM_SCORE_ADJ_MAX;
1125 put_task_struct(task);
1126 if (oom_adj > OOM_ADJUST_MAX)
1127 oom_adj = OOM_ADJUST_MAX;
1128 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
1129 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1130 }
1131
__set_oom_adj(struct file * file,int oom_adj,bool legacy)1132 static int __set_oom_adj(struct file *file, int oom_adj, bool legacy)
1133 {
1134 struct mm_struct *mm = NULL;
1135 struct task_struct *task;
1136 int err = 0;
1137
1138 task = get_proc_task(file_inode(file));
1139 if (!task)
1140 return -ESRCH;
1141
1142 mutex_lock(&oom_adj_mutex);
1143 if (legacy) {
1144 if (oom_adj < task->signal->oom_score_adj &&
1145 !capable(CAP_SYS_RESOURCE)) {
1146 err = -EACCES;
1147 goto err_unlock;
1148 }
1149 /*
1150 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1151 * /proc/pid/oom_score_adj instead.
1152 */
1153 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1154 current->comm, task_pid_nr(current), task_pid_nr(task),
1155 task_pid_nr(task));
1156 } else {
1157 if ((short)oom_adj < task->signal->oom_score_adj_min &&
1158 !capable(CAP_SYS_RESOURCE)) {
1159 err = -EACCES;
1160 goto err_unlock;
1161 }
1162 }
1163
1164 /*
1165 * Make sure we will check other processes sharing the mm if this is
1166 * not vfrok which wants its own oom_score_adj.
1167 * pin the mm so it doesn't go away and get reused after task_unlock
1168 */
1169 if (!task->vfork_done) {
1170 struct task_struct *p = find_lock_task_mm(task);
1171
1172 if (p) {
1173 if (test_bit(MMF_MULTIPROCESS, &p->mm->flags)) {
1174 mm = p->mm;
1175 mmgrab(mm);
1176 }
1177 task_unlock(p);
1178 }
1179 }
1180
1181 task->signal->oom_score_adj = oom_adj;
1182 if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1183 task->signal->oom_score_adj_min = (short)oom_adj;
1184 trace_oom_score_adj_update(task);
1185
1186 if (mm) {
1187 struct task_struct *p;
1188
1189 rcu_read_lock();
1190 for_each_process(p) {
1191 if (same_thread_group(task, p))
1192 continue;
1193
1194 /* do not touch kernel threads or the global init */
1195 if (p->flags & PF_KTHREAD || is_global_init(p))
1196 continue;
1197
1198 task_lock(p);
1199 if (!p->vfork_done && process_shares_mm(p, mm)) {
1200 p->signal->oom_score_adj = oom_adj;
1201 if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1202 p->signal->oom_score_adj_min = (short)oom_adj;
1203 }
1204 task_unlock(p);
1205 }
1206 rcu_read_unlock();
1207 mmdrop(mm);
1208 }
1209 err_unlock:
1210 mutex_unlock(&oom_adj_mutex);
1211 put_task_struct(task);
1212 return err;
1213 }
1214
1215 /*
1216 * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1217 * kernels. The effective policy is defined by oom_score_adj, which has a
1218 * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1219 * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1220 * Processes that become oom disabled via oom_adj will still be oom disabled
1221 * with this implementation.
1222 *
1223 * oom_adj cannot be removed since existing userspace binaries use it.
1224 */
oom_adj_write(struct file * file,const char __user * buf,size_t count,loff_t * ppos)1225 static ssize_t oom_adj_write(struct file *file, const char __user *buf,
1226 size_t count, loff_t *ppos)
1227 {
1228 char buffer[PROC_NUMBUF] = {};
1229 int oom_adj;
1230 int err;
1231
1232 if (count > sizeof(buffer) - 1)
1233 count = sizeof(buffer) - 1;
1234 if (copy_from_user(buffer, buf, count)) {
1235 err = -EFAULT;
1236 goto out;
1237 }
1238
1239 err = kstrtoint(strstrip(buffer), 0, &oom_adj);
1240 if (err)
1241 goto out;
1242 if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
1243 oom_adj != OOM_DISABLE) {
1244 err = -EINVAL;
1245 goto out;
1246 }
1247
1248 /*
1249 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1250 * value is always attainable.
1251 */
1252 if (oom_adj == OOM_ADJUST_MAX)
1253 oom_adj = OOM_SCORE_ADJ_MAX;
1254 else
1255 oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
1256
1257 err = __set_oom_adj(file, oom_adj, true);
1258 out:
1259 return err < 0 ? err : count;
1260 }
1261
1262 static const struct file_operations proc_oom_adj_operations = {
1263 .read = oom_adj_read,
1264 .write = oom_adj_write,
1265 .llseek = generic_file_llseek,
1266 };
1267
oom_score_adj_read(struct file * file,char __user * buf,size_t count,loff_t * ppos)1268 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1269 size_t count, loff_t *ppos)
1270 {
1271 struct task_struct *task = get_proc_task(file_inode(file));
1272 char buffer[PROC_NUMBUF];
1273 short oom_score_adj = OOM_SCORE_ADJ_MIN;
1274 size_t len;
1275
1276 if (!task)
1277 return -ESRCH;
1278 oom_score_adj = task->signal->oom_score_adj;
1279 put_task_struct(task);
1280 len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
1281 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1282 }
1283
oom_score_adj_write(struct file * file,const char __user * buf,size_t count,loff_t * ppos)1284 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1285 size_t count, loff_t *ppos)
1286 {
1287 char buffer[PROC_NUMBUF] = {};
1288 int oom_score_adj;
1289 int err;
1290
1291 if (count > sizeof(buffer) - 1)
1292 count = sizeof(buffer) - 1;
1293 if (copy_from_user(buffer, buf, count)) {
1294 err = -EFAULT;
1295 goto out;
1296 }
1297
1298 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1299 if (err)
1300 goto out;
1301 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1302 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1303 err = -EINVAL;
1304 goto out;
1305 }
1306
1307 err = __set_oom_adj(file, oom_score_adj, false);
1308 out:
1309 return err < 0 ? err : count;
1310 }
1311
1312 static const struct file_operations proc_oom_score_adj_operations = {
1313 .read = oom_score_adj_read,
1314 .write = oom_score_adj_write,
1315 .llseek = default_llseek,
1316 };
1317
1318 #ifdef CONFIG_AUDIT
1319 #define TMPBUFLEN 11
proc_loginuid_read(struct file * file,char __user * buf,size_t count,loff_t * ppos)1320 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1321 size_t count, loff_t *ppos)
1322 {
1323 struct inode * inode = file_inode(file);
1324 struct task_struct *task = get_proc_task(inode);
1325 ssize_t length;
1326 char tmpbuf[TMPBUFLEN];
1327
1328 if (!task)
1329 return -ESRCH;
1330 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1331 from_kuid(file->f_cred->user_ns,
1332 audit_get_loginuid(task)));
1333 put_task_struct(task);
1334 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1335 }
1336
proc_loginuid_write(struct file * file,const char __user * buf,size_t count,loff_t * ppos)1337 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1338 size_t count, loff_t *ppos)
1339 {
1340 struct inode * inode = file_inode(file);
1341 uid_t loginuid;
1342 kuid_t kloginuid;
1343 int rv;
1344
1345 /* Don't let kthreads write their own loginuid */
1346 if (current->flags & PF_KTHREAD)
1347 return -EPERM;
1348
1349 rcu_read_lock();
1350 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1351 rcu_read_unlock();
1352 return -EPERM;
1353 }
1354 rcu_read_unlock();
1355
1356 if (*ppos != 0) {
1357 /* No partial writes. */
1358 return -EINVAL;
1359 }
1360
1361 rv = kstrtou32_from_user(buf, count, 10, &loginuid);
1362 if (rv < 0)
1363 return rv;
1364
1365 /* is userspace tring to explicitly UNSET the loginuid? */
1366 if (loginuid == AUDIT_UID_UNSET) {
1367 kloginuid = INVALID_UID;
1368 } else {
1369 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1370 if (!uid_valid(kloginuid))
1371 return -EINVAL;
1372 }
1373
1374 rv = audit_set_loginuid(kloginuid);
1375 if (rv < 0)
1376 return rv;
1377 return count;
1378 }
1379
1380 static const struct file_operations proc_loginuid_operations = {
1381 .read = proc_loginuid_read,
1382 .write = proc_loginuid_write,
1383 .llseek = generic_file_llseek,
1384 };
1385
proc_sessionid_read(struct file * file,char __user * buf,size_t count,loff_t * ppos)1386 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1387 size_t count, loff_t *ppos)
1388 {
1389 struct inode * inode = file_inode(file);
1390 struct task_struct *task = get_proc_task(inode);
1391 ssize_t length;
1392 char tmpbuf[TMPBUFLEN];
1393
1394 if (!task)
1395 return -ESRCH;
1396 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1397 audit_get_sessionid(task));
1398 put_task_struct(task);
1399 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1400 }
1401
1402 static const struct file_operations proc_sessionid_operations = {
1403 .read = proc_sessionid_read,
1404 .llseek = generic_file_llseek,
1405 };
1406 #endif
1407
1408 #ifdef CONFIG_FAULT_INJECTION
proc_fault_inject_read(struct file * file,char __user * buf,size_t count,loff_t * ppos)1409 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1410 size_t count, loff_t *ppos)
1411 {
1412 struct task_struct *task = get_proc_task(file_inode(file));
1413 char buffer[PROC_NUMBUF];
1414 size_t len;
1415 int make_it_fail;
1416
1417 if (!task)
1418 return -ESRCH;
1419 make_it_fail = task->make_it_fail;
1420 put_task_struct(task);
1421
1422 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1423
1424 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1425 }
1426
proc_fault_inject_write(struct file * file,const char __user * buf,size_t count,loff_t * ppos)1427 static ssize_t proc_fault_inject_write(struct file * file,
1428 const char __user * buf, size_t count, loff_t *ppos)
1429 {
1430 struct task_struct *task;
1431 char buffer[PROC_NUMBUF] = {};
1432 int make_it_fail;
1433 int rv;
1434
1435 if (!capable(CAP_SYS_RESOURCE))
1436 return -EPERM;
1437
1438 if (count > sizeof(buffer) - 1)
1439 count = sizeof(buffer) - 1;
1440 if (copy_from_user(buffer, buf, count))
1441 return -EFAULT;
1442 rv = kstrtoint(strstrip(buffer), 0, &make_it_fail);
1443 if (rv < 0)
1444 return rv;
1445 if (make_it_fail < 0 || make_it_fail > 1)
1446 return -EINVAL;
1447
1448 task = get_proc_task(file_inode(file));
1449 if (!task)
1450 return -ESRCH;
1451 task->make_it_fail = make_it_fail;
1452 put_task_struct(task);
1453
1454 return count;
1455 }
1456
1457 static const struct file_operations proc_fault_inject_operations = {
1458 .read = proc_fault_inject_read,
1459 .write = proc_fault_inject_write,
1460 .llseek = generic_file_llseek,
1461 };
1462
proc_fail_nth_write(struct file * file,const char __user * buf,size_t count,loff_t * ppos)1463 static ssize_t proc_fail_nth_write(struct file *file, const char __user *buf,
1464 size_t count, loff_t *ppos)
1465 {
1466 struct task_struct *task;
1467 int err;
1468 unsigned int n;
1469
1470 err = kstrtouint_from_user(buf, count, 0, &n);
1471 if (err)
1472 return err;
1473
1474 task = get_proc_task(file_inode(file));
1475 if (!task)
1476 return -ESRCH;
1477 task->fail_nth = n;
1478 put_task_struct(task);
1479
1480 return count;
1481 }
1482
proc_fail_nth_read(struct file * file,char __user * buf,size_t count,loff_t * ppos)1483 static ssize_t proc_fail_nth_read(struct file *file, char __user *buf,
1484 size_t count, loff_t *ppos)
1485 {
1486 struct task_struct *task;
1487 char numbuf[PROC_NUMBUF];
1488 ssize_t len;
1489
1490 task = get_proc_task(file_inode(file));
1491 if (!task)
1492 return -ESRCH;
1493 len = snprintf(numbuf, sizeof(numbuf), "%u\n", task->fail_nth);
1494 put_task_struct(task);
1495 return simple_read_from_buffer(buf, count, ppos, numbuf, len);
1496 }
1497
1498 static const struct file_operations proc_fail_nth_operations = {
1499 .read = proc_fail_nth_read,
1500 .write = proc_fail_nth_write,
1501 };
1502 #endif
1503
1504
1505 #ifdef CONFIG_SCHED_DEBUG
1506 /*
1507 * Print out various scheduling related per-task fields:
1508 */
sched_show(struct seq_file * m,void * v)1509 static int sched_show(struct seq_file *m, void *v)
1510 {
1511 struct inode *inode = m->private;
1512 struct pid_namespace *ns = proc_pid_ns(inode->i_sb);
1513 struct task_struct *p;
1514
1515 p = get_proc_task(inode);
1516 if (!p)
1517 return -ESRCH;
1518 proc_sched_show_task(p, ns, m);
1519
1520 put_task_struct(p);
1521
1522 return 0;
1523 }
1524
1525 static ssize_t
sched_write(struct file * file,const char __user * buf,size_t count,loff_t * offset)1526 sched_write(struct file *file, const char __user *buf,
1527 size_t count, loff_t *offset)
1528 {
1529 struct inode *inode = file_inode(file);
1530 struct task_struct *p;
1531
1532 p = get_proc_task(inode);
1533 if (!p)
1534 return -ESRCH;
1535 proc_sched_set_task(p);
1536
1537 put_task_struct(p);
1538
1539 return count;
1540 }
1541
sched_open(struct inode * inode,struct file * filp)1542 static int sched_open(struct inode *inode, struct file *filp)
1543 {
1544 return single_open(filp, sched_show, inode);
1545 }
1546
1547 static const struct file_operations proc_pid_sched_operations = {
1548 .open = sched_open,
1549 .read = seq_read,
1550 .write = sched_write,
1551 .llseek = seq_lseek,
1552 .release = single_release,
1553 };
1554
1555 #endif
1556
1557 #ifdef CONFIG_SCHED_AUTOGROUP
1558 /*
1559 * Print out autogroup related information:
1560 */
sched_autogroup_show(struct seq_file * m,void * v)1561 static int sched_autogroup_show(struct seq_file *m, void *v)
1562 {
1563 struct inode *inode = m->private;
1564 struct task_struct *p;
1565
1566 p = get_proc_task(inode);
1567 if (!p)
1568 return -ESRCH;
1569 proc_sched_autogroup_show_task(p, m);
1570
1571 put_task_struct(p);
1572
1573 return 0;
1574 }
1575
1576 static ssize_t
sched_autogroup_write(struct file * file,const char __user * buf,size_t count,loff_t * offset)1577 sched_autogroup_write(struct file *file, const char __user *buf,
1578 size_t count, loff_t *offset)
1579 {
1580 struct inode *inode = file_inode(file);
1581 struct task_struct *p;
1582 char buffer[PROC_NUMBUF] = {};
1583 int nice;
1584 int err;
1585
1586 if (count > sizeof(buffer) - 1)
1587 count = sizeof(buffer) - 1;
1588 if (copy_from_user(buffer, buf, count))
1589 return -EFAULT;
1590
1591 err = kstrtoint(strstrip(buffer), 0, &nice);
1592 if (err < 0)
1593 return err;
1594
1595 p = get_proc_task(inode);
1596 if (!p)
1597 return -ESRCH;
1598
1599 err = proc_sched_autogroup_set_nice(p, nice);
1600 if (err)
1601 count = err;
1602
1603 put_task_struct(p);
1604
1605 return count;
1606 }
1607
sched_autogroup_open(struct inode * inode,struct file * filp)1608 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1609 {
1610 int ret;
1611
1612 ret = single_open(filp, sched_autogroup_show, NULL);
1613 if (!ret) {
1614 struct seq_file *m = filp->private_data;
1615
1616 m->private = inode;
1617 }
1618 return ret;
1619 }
1620
1621 static const struct file_operations proc_pid_sched_autogroup_operations = {
1622 .open = sched_autogroup_open,
1623 .read = seq_read,
1624 .write = sched_autogroup_write,
1625 .llseek = seq_lseek,
1626 .release = single_release,
1627 };
1628
1629 #endif /* CONFIG_SCHED_AUTOGROUP */
1630
1631 #ifdef CONFIG_TIME_NS
timens_offsets_show(struct seq_file * m,void * v)1632 static int timens_offsets_show(struct seq_file *m, void *v)
1633 {
1634 struct task_struct *p;
1635
1636 p = get_proc_task(file_inode(m->file));
1637 if (!p)
1638 return -ESRCH;
1639 proc_timens_show_offsets(p, m);
1640
1641 put_task_struct(p);
1642
1643 return 0;
1644 }
1645
timens_offsets_write(struct file * file,const char __user * buf,size_t count,loff_t * ppos)1646 static ssize_t timens_offsets_write(struct file *file, const char __user *buf,
1647 size_t count, loff_t *ppos)
1648 {
1649 struct inode *inode = file_inode(file);
1650 struct proc_timens_offset offsets[2];
1651 char *kbuf = NULL, *pos, *next_line;
1652 struct task_struct *p;
1653 int ret, noffsets;
1654
1655 /* Only allow < page size writes at the beginning of the file */
1656 if ((*ppos != 0) || (count >= PAGE_SIZE))
1657 return -EINVAL;
1658
1659 /* Slurp in the user data */
1660 kbuf = memdup_user_nul(buf, count);
1661 if (IS_ERR(kbuf))
1662 return PTR_ERR(kbuf);
1663
1664 /* Parse the user data */
1665 ret = -EINVAL;
1666 noffsets = 0;
1667 for (pos = kbuf; pos; pos = next_line) {
1668 struct proc_timens_offset *off = &offsets[noffsets];
1669 char clock[10];
1670 int err;
1671
1672 /* Find the end of line and ensure we don't look past it */
1673 next_line = strchr(pos, '\n');
1674 if (next_line) {
1675 *next_line = '\0';
1676 next_line++;
1677 if (*next_line == '\0')
1678 next_line = NULL;
1679 }
1680
1681 err = sscanf(pos, "%9s %lld %lu", clock,
1682 &off->val.tv_sec, &off->val.tv_nsec);
1683 if (err != 3 || off->val.tv_nsec >= NSEC_PER_SEC)
1684 goto out;
1685
1686 clock[sizeof(clock) - 1] = 0;
1687 if (strcmp(clock, "monotonic") == 0 ||
1688 strcmp(clock, __stringify(CLOCK_MONOTONIC)) == 0)
1689 off->clockid = CLOCK_MONOTONIC;
1690 else if (strcmp(clock, "boottime") == 0 ||
1691 strcmp(clock, __stringify(CLOCK_BOOTTIME)) == 0)
1692 off->clockid = CLOCK_BOOTTIME;
1693 else
1694 goto out;
1695
1696 noffsets++;
1697 if (noffsets == ARRAY_SIZE(offsets)) {
1698 if (next_line)
1699 count = next_line - kbuf;
1700 break;
1701 }
1702 }
1703
1704 ret = -ESRCH;
1705 p = get_proc_task(inode);
1706 if (!p)
1707 goto out;
1708 ret = proc_timens_set_offset(file, p, offsets, noffsets);
1709 put_task_struct(p);
1710 if (ret)
1711 goto out;
1712
1713 ret = count;
1714 out:
1715 kfree(kbuf);
1716 return ret;
1717 }
1718
timens_offsets_open(struct inode * inode,struct file * filp)1719 static int timens_offsets_open(struct inode *inode, struct file *filp)
1720 {
1721 return single_open(filp, timens_offsets_show, inode);
1722 }
1723
1724 static const struct file_operations proc_timens_offsets_operations = {
1725 .open = timens_offsets_open,
1726 .read = seq_read,
1727 .write = timens_offsets_write,
1728 .llseek = seq_lseek,
1729 .release = single_release,
1730 };
1731 #endif /* CONFIG_TIME_NS */
1732
comm_write(struct file * file,const char __user * buf,size_t count,loff_t * offset)1733 static ssize_t comm_write(struct file *file, const char __user *buf,
1734 size_t count, loff_t *offset)
1735 {
1736 struct inode *inode = file_inode(file);
1737 struct task_struct *p;
1738 char buffer[TASK_COMM_LEN] = {};
1739 const size_t maxlen = sizeof(buffer) - 1;
1740
1741 if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
1742 return -EFAULT;
1743
1744 p = get_proc_task(inode);
1745 if (!p)
1746 return -ESRCH;
1747
1748 if (same_thread_group(current, p)) {
1749 set_task_comm(p, buffer);
1750 proc_comm_connector(p);
1751 }
1752 else
1753 count = -EINVAL;
1754
1755 put_task_struct(p);
1756
1757 return count;
1758 }
1759
comm_show(struct seq_file * m,void * v)1760 static int comm_show(struct seq_file *m, void *v)
1761 {
1762 struct inode *inode = m->private;
1763 struct task_struct *p;
1764
1765 p = get_proc_task(inode);
1766 if (!p)
1767 return -ESRCH;
1768
1769 proc_task_name(m, p, false);
1770 seq_putc(m, '\n');
1771
1772 put_task_struct(p);
1773
1774 return 0;
1775 }
1776
comm_open(struct inode * inode,struct file * filp)1777 static int comm_open(struct inode *inode, struct file *filp)
1778 {
1779 return single_open(filp, comm_show, inode);
1780 }
1781
1782 static const struct file_operations proc_pid_set_comm_operations = {
1783 .open = comm_open,
1784 .read = seq_read,
1785 .write = comm_write,
1786 .llseek = seq_lseek,
1787 .release = single_release,
1788 };
1789
proc_exe_link(struct dentry * dentry,struct path * exe_path)1790 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1791 {
1792 struct task_struct *task;
1793 struct file *exe_file;
1794
1795 task = get_proc_task(d_inode(dentry));
1796 if (!task)
1797 return -ENOENT;
1798 exe_file = get_task_exe_file(task);
1799 put_task_struct(task);
1800 if (exe_file) {
1801 *exe_path = exe_file->f_path;
1802 path_get(&exe_file->f_path);
1803 fput(exe_file);
1804 return 0;
1805 } else
1806 return -ENOENT;
1807 }
1808
proc_pid_get_link(struct dentry * dentry,struct inode * inode,struct delayed_call * done)1809 static const char *proc_pid_get_link(struct dentry *dentry,
1810 struct inode *inode,
1811 struct delayed_call *done)
1812 {
1813 struct path path;
1814 int error = -EACCES;
1815
1816 if (!dentry)
1817 return ERR_PTR(-ECHILD);
1818
1819 /* Are we allowed to snoop on the tasks file descriptors? */
1820 if (!proc_fd_access_allowed(inode))
1821 goto out;
1822
1823 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1824 if (error)
1825 goto out;
1826
1827 error = nd_jump_link(&path);
1828 out:
1829 return ERR_PTR(error);
1830 }
1831
do_proc_readlink(const struct path * path,char __user * buffer,int buflen)1832 static int do_proc_readlink(const struct path *path, char __user *buffer, int buflen)
1833 {
1834 char *tmp = kmalloc(PATH_MAX, GFP_KERNEL);
1835 char *pathname;
1836 int len;
1837
1838 if (!tmp)
1839 return -ENOMEM;
1840
1841 pathname = d_path(path, tmp, PATH_MAX);
1842 len = PTR_ERR(pathname);
1843 if (IS_ERR(pathname))
1844 goto out;
1845 len = tmp + PATH_MAX - 1 - pathname;
1846
1847 if (len > buflen)
1848 len = buflen;
1849 if (copy_to_user(buffer, pathname, len))
1850 len = -EFAULT;
1851 out:
1852 kfree(tmp);
1853 return len;
1854 }
1855
proc_pid_readlink(struct dentry * dentry,char __user * buffer,int buflen)1856 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1857 {
1858 int error = -EACCES;
1859 struct inode *inode = d_inode(dentry);
1860 struct path path;
1861
1862 /* Are we allowed to snoop on the tasks file descriptors? */
1863 if (!proc_fd_access_allowed(inode))
1864 goto out;
1865
1866 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1867 if (error)
1868 goto out;
1869
1870 error = do_proc_readlink(&path, buffer, buflen);
1871 path_put(&path);
1872 out:
1873 return error;
1874 }
1875
1876 const struct inode_operations proc_pid_link_inode_operations = {
1877 .readlink = proc_pid_readlink,
1878 .get_link = proc_pid_get_link,
1879 .setattr = proc_setattr,
1880 };
1881
1882
1883 /* building an inode */
1884
task_dump_owner(struct task_struct * task,umode_t mode,kuid_t * ruid,kgid_t * rgid)1885 void task_dump_owner(struct task_struct *task, umode_t mode,
1886 kuid_t *ruid, kgid_t *rgid)
1887 {
1888 /* Depending on the state of dumpable compute who should own a
1889 * proc file for a task.
1890 */
1891 const struct cred *cred;
1892 kuid_t uid;
1893 kgid_t gid;
1894
1895 if (unlikely(task->flags & PF_KTHREAD)) {
1896 *ruid = GLOBAL_ROOT_UID;
1897 *rgid = GLOBAL_ROOT_GID;
1898 return;
1899 }
1900
1901 /* Default to the tasks effective ownership */
1902 rcu_read_lock();
1903 cred = __task_cred(task);
1904 uid = cred->euid;
1905 gid = cred->egid;
1906 rcu_read_unlock();
1907
1908 /*
1909 * Before the /proc/pid/status file was created the only way to read
1910 * the effective uid of a /process was to stat /proc/pid. Reading
1911 * /proc/pid/status is slow enough that procps and other packages
1912 * kept stating /proc/pid. To keep the rules in /proc simple I have
1913 * made this apply to all per process world readable and executable
1914 * directories.
1915 */
1916 if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) {
1917 struct mm_struct *mm;
1918 task_lock(task);
1919 mm = task->mm;
1920 /* Make non-dumpable tasks owned by some root */
1921 if (mm) {
1922 if (get_dumpable(mm) != SUID_DUMP_USER) {
1923 struct user_namespace *user_ns = mm->user_ns;
1924
1925 uid = make_kuid(user_ns, 0);
1926 if (!uid_valid(uid))
1927 uid = GLOBAL_ROOT_UID;
1928
1929 gid = make_kgid(user_ns, 0);
1930 if (!gid_valid(gid))
1931 gid = GLOBAL_ROOT_GID;
1932 }
1933 } else {
1934 uid = GLOBAL_ROOT_UID;
1935 gid = GLOBAL_ROOT_GID;
1936 }
1937 task_unlock(task);
1938 }
1939 *ruid = uid;
1940 *rgid = gid;
1941 }
1942
proc_pid_evict_inode(struct proc_inode * ei)1943 void proc_pid_evict_inode(struct proc_inode *ei)
1944 {
1945 struct pid *pid = ei->pid;
1946
1947 if (S_ISDIR(ei->vfs_inode.i_mode)) {
1948 spin_lock(&pid->lock);
1949 hlist_del_init_rcu(&ei->sibling_inodes);
1950 spin_unlock(&pid->lock);
1951 }
1952 }
1953
proc_pid_make_inode(struct super_block * sb,struct task_struct * task,umode_t mode)1954 struct inode *proc_pid_make_inode(struct super_block *sb,
1955 struct task_struct *task, umode_t mode)
1956 {
1957 struct inode * inode;
1958 struct proc_inode *ei;
1959 struct pid *pid;
1960
1961 /* We need a new inode */
1962
1963 inode = new_inode(sb);
1964 if (!inode)
1965 goto out;
1966
1967 /* Common stuff */
1968 ei = PROC_I(inode);
1969 inode->i_mode = mode;
1970 inode->i_ino = get_next_ino();
1971 simple_inode_init_ts(inode);
1972 inode->i_op = &proc_def_inode_operations;
1973
1974 /*
1975 * grab the reference to task.
1976 */
1977 pid = get_task_pid(task, PIDTYPE_PID);
1978 if (!pid)
1979 goto out_unlock;
1980
1981 /* Let the pid remember us for quick removal */
1982 ei->pid = pid;
1983
1984 task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
1985 security_task_to_inode(task, inode);
1986
1987 out:
1988 return inode;
1989
1990 out_unlock:
1991 iput(inode);
1992 return NULL;
1993 }
1994
1995 /*
1996 * Generating an inode and adding it into @pid->inodes, so that task will
1997 * invalidate inode's dentry before being released.
1998 *
1999 * This helper is used for creating dir-type entries under '/proc' and
2000 * '/proc/<tgid>/task'. Other entries(eg. fd, stat) under '/proc/<tgid>'
2001 * can be released by invalidating '/proc/<tgid>' dentry.
2002 * In theory, dentries under '/proc/<tgid>/task' can also be released by
2003 * invalidating '/proc/<tgid>' dentry, we reserve it to handle single
2004 * thread exiting situation: Any one of threads should invalidate its
2005 * '/proc/<tgid>/task/<pid>' dentry before released.
2006 */
proc_pid_make_base_inode(struct super_block * sb,struct task_struct * task,umode_t mode)2007 static struct inode *proc_pid_make_base_inode(struct super_block *sb,
2008 struct task_struct *task, umode_t mode)
2009 {
2010 struct inode *inode;
2011 struct proc_inode *ei;
2012 struct pid *pid;
2013
2014 inode = proc_pid_make_inode(sb, task, mode);
2015 if (!inode)
2016 return NULL;
2017
2018 /* Let proc_flush_pid find this directory inode */
2019 ei = PROC_I(inode);
2020 pid = ei->pid;
2021 spin_lock(&pid->lock);
2022 hlist_add_head_rcu(&ei->sibling_inodes, &pid->inodes);
2023 spin_unlock(&pid->lock);
2024
2025 return inode;
2026 }
2027
pid_getattr(struct mnt_idmap * idmap,const struct path * path,struct kstat * stat,u32 request_mask,unsigned int query_flags)2028 int pid_getattr(struct mnt_idmap *idmap, const struct path *path,
2029 struct kstat *stat, u32 request_mask, unsigned int query_flags)
2030 {
2031 struct inode *inode = d_inode(path->dentry);
2032 struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
2033 struct task_struct *task;
2034
2035 generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat);
2036
2037 stat->uid = GLOBAL_ROOT_UID;
2038 stat->gid = GLOBAL_ROOT_GID;
2039 rcu_read_lock();
2040 task = pid_task(proc_pid(inode), PIDTYPE_PID);
2041 if (task) {
2042 if (!has_pid_permissions(fs_info, task, HIDEPID_INVISIBLE)) {
2043 rcu_read_unlock();
2044 /*
2045 * This doesn't prevent learning whether PID exists,
2046 * it only makes getattr() consistent with readdir().
2047 */
2048 return -ENOENT;
2049 }
2050 task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid);
2051 }
2052 rcu_read_unlock();
2053 return 0;
2054 }
2055
2056 /* dentry stuff */
2057
2058 /*
2059 * Set <pid>/... inode ownership (can change due to setuid(), etc.)
2060 */
pid_update_inode(struct task_struct * task,struct inode * inode)2061 void pid_update_inode(struct task_struct *task, struct inode *inode)
2062 {
2063 task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid);
2064
2065 inode->i_mode &= ~(S_ISUID | S_ISGID);
2066 security_task_to_inode(task, inode);
2067 }
2068
2069 /*
2070 * Rewrite the inode's ownerships here because the owning task may have
2071 * performed a setuid(), etc.
2072 *
2073 */
pid_revalidate(struct dentry * dentry,unsigned int flags)2074 static int pid_revalidate(struct dentry *dentry, unsigned int flags)
2075 {
2076 struct inode *inode;
2077 struct task_struct *task;
2078 int ret = 0;
2079
2080 rcu_read_lock();
2081 inode = d_inode_rcu(dentry);
2082 if (!inode)
2083 goto out;
2084 task = pid_task(proc_pid(inode), PIDTYPE_PID);
2085
2086 if (task) {
2087 pid_update_inode(task, inode);
2088 ret = 1;
2089 }
2090 out:
2091 rcu_read_unlock();
2092 return ret;
2093 }
2094
proc_inode_is_dead(struct inode * inode)2095 static inline bool proc_inode_is_dead(struct inode *inode)
2096 {
2097 return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
2098 }
2099
pid_delete_dentry(const struct dentry * dentry)2100 int pid_delete_dentry(const struct dentry *dentry)
2101 {
2102 /* Is the task we represent dead?
2103 * If so, then don't put the dentry on the lru list,
2104 * kill it immediately.
2105 */
2106 return proc_inode_is_dead(d_inode(dentry));
2107 }
2108
2109 const struct dentry_operations pid_dentry_operations =
2110 {
2111 .d_revalidate = pid_revalidate,
2112 .d_delete = pid_delete_dentry,
2113 };
2114
2115 /* Lookups */
2116
2117 /*
2118 * Fill a directory entry.
2119 *
2120 * If possible create the dcache entry and derive our inode number and
2121 * file type from dcache entry.
2122 *
2123 * Since all of the proc inode numbers are dynamically generated, the inode
2124 * numbers do not exist until the inode is cache. This means creating
2125 * the dcache entry in readdir is necessary to keep the inode numbers
2126 * reported by readdir in sync with the inode numbers reported
2127 * by stat.
2128 */
proc_fill_cache(struct file * file,struct dir_context * ctx,const char * name,unsigned int len,instantiate_t instantiate,struct task_struct * task,const void * ptr)2129 bool proc_fill_cache(struct file *file, struct dir_context *ctx,
2130 const char *name, unsigned int len,
2131 instantiate_t instantiate, struct task_struct *task, const void *ptr)
2132 {
2133 struct dentry *child, *dir = file->f_path.dentry;
2134 struct qstr qname = QSTR_INIT(name, len);
2135 struct inode *inode;
2136 unsigned type = DT_UNKNOWN;
2137 ino_t ino = 1;
2138
2139 child = d_hash_and_lookup(dir, &qname);
2140 if (!child) {
2141 DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq);
2142 child = d_alloc_parallel(dir, &qname, &wq);
2143 if (IS_ERR(child))
2144 goto end_instantiate;
2145 if (d_in_lookup(child)) {
2146 struct dentry *res;
2147 res = instantiate(child, task, ptr);
2148 d_lookup_done(child);
2149 if (unlikely(res)) {
2150 dput(child);
2151 child = res;
2152 if (IS_ERR(child))
2153 goto end_instantiate;
2154 }
2155 }
2156 }
2157 inode = d_inode(child);
2158 ino = inode->i_ino;
2159 type = inode->i_mode >> 12;
2160 dput(child);
2161 end_instantiate:
2162 return dir_emit(ctx, name, len, ino, type);
2163 }
2164
2165 /*
2166 * dname_to_vma_addr - maps a dentry name into two unsigned longs
2167 * which represent vma start and end addresses.
2168 */
dname_to_vma_addr(struct dentry * dentry,unsigned long * start,unsigned long * end)2169 static int dname_to_vma_addr(struct dentry *dentry,
2170 unsigned long *start, unsigned long *end)
2171 {
2172 const char *str = dentry->d_name.name;
2173 unsigned long long sval, eval;
2174 unsigned int len;
2175
2176 if (str[0] == '0' && str[1] != '-')
2177 return -EINVAL;
2178 len = _parse_integer(str, 16, &sval);
2179 if (len & KSTRTOX_OVERFLOW)
2180 return -EINVAL;
2181 if (sval != (unsigned long)sval)
2182 return -EINVAL;
2183 str += len;
2184
2185 if (*str != '-')
2186 return -EINVAL;
2187 str++;
2188
2189 if (str[0] == '0' && str[1])
2190 return -EINVAL;
2191 len = _parse_integer(str, 16, &eval);
2192 if (len & KSTRTOX_OVERFLOW)
2193 return -EINVAL;
2194 if (eval != (unsigned long)eval)
2195 return -EINVAL;
2196 str += len;
2197
2198 if (*str != '\0')
2199 return -EINVAL;
2200
2201 *start = sval;
2202 *end = eval;
2203
2204 return 0;
2205 }
2206
map_files_d_revalidate(struct dentry * dentry,unsigned int flags)2207 static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
2208 {
2209 unsigned long vm_start, vm_end;
2210 bool exact_vma_exists = false;
2211 struct mm_struct *mm = NULL;
2212 struct task_struct *task;
2213 struct inode *inode;
2214 int status = 0;
2215
2216 if (flags & LOOKUP_RCU)
2217 return -ECHILD;
2218
2219 inode = d_inode(dentry);
2220 task = get_proc_task(inode);
2221 if (!task)
2222 goto out_notask;
2223
2224 mm = mm_access(task, PTRACE_MODE_READ_FSCREDS);
2225 if (IS_ERR_OR_NULL(mm))
2226 goto out;
2227
2228 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
2229 status = mmap_read_lock_killable(mm);
2230 if (!status) {
2231 exact_vma_exists = !!find_exact_vma(mm, vm_start,
2232 vm_end);
2233 mmap_read_unlock(mm);
2234 }
2235 }
2236
2237 mmput(mm);
2238
2239 if (exact_vma_exists) {
2240 task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
2241
2242 security_task_to_inode(task, inode);
2243 status = 1;
2244 }
2245
2246 out:
2247 put_task_struct(task);
2248
2249 out_notask:
2250 return status;
2251 }
2252
2253 static const struct dentry_operations tid_map_files_dentry_operations = {
2254 .d_revalidate = map_files_d_revalidate,
2255 .d_delete = pid_delete_dentry,
2256 };
2257
map_files_get_link(struct dentry * dentry,struct path * path)2258 static int map_files_get_link(struct dentry *dentry, struct path *path)
2259 {
2260 unsigned long vm_start, vm_end;
2261 struct vm_area_struct *vma;
2262 struct task_struct *task;
2263 struct mm_struct *mm;
2264 int rc;
2265
2266 rc = -ENOENT;
2267 task = get_proc_task(d_inode(dentry));
2268 if (!task)
2269 goto out;
2270
2271 mm = get_task_mm(task);
2272 put_task_struct(task);
2273 if (!mm)
2274 goto out;
2275
2276 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2277 if (rc)
2278 goto out_mmput;
2279
2280 rc = mmap_read_lock_killable(mm);
2281 if (rc)
2282 goto out_mmput;
2283
2284 rc = -ENOENT;
2285 vma = find_exact_vma(mm, vm_start, vm_end);
2286 if (vma && vma->vm_file) {
2287 *path = *file_user_path(vma->vm_file);
2288 path_get(path);
2289 rc = 0;
2290 }
2291 mmap_read_unlock(mm);
2292
2293 out_mmput:
2294 mmput(mm);
2295 out:
2296 return rc;
2297 }
2298
2299 struct map_files_info {
2300 unsigned long start;
2301 unsigned long end;
2302 fmode_t mode;
2303 };
2304
2305 /*
2306 * Only allow CAP_SYS_ADMIN and CAP_CHECKPOINT_RESTORE to follow the links, due
2307 * to concerns about how the symlinks may be used to bypass permissions on
2308 * ancestor directories in the path to the file in question.
2309 */
2310 static const char *
proc_map_files_get_link(struct dentry * dentry,struct inode * inode,struct delayed_call * done)2311 proc_map_files_get_link(struct dentry *dentry,
2312 struct inode *inode,
2313 struct delayed_call *done)
2314 {
2315 if (!checkpoint_restore_ns_capable(&init_user_ns))
2316 return ERR_PTR(-EPERM);
2317
2318 return proc_pid_get_link(dentry, inode, done);
2319 }
2320
2321 /*
2322 * Identical to proc_pid_link_inode_operations except for get_link()
2323 */
2324 static const struct inode_operations proc_map_files_link_inode_operations = {
2325 .readlink = proc_pid_readlink,
2326 .get_link = proc_map_files_get_link,
2327 .setattr = proc_setattr,
2328 };
2329
2330 static struct dentry *
proc_map_files_instantiate(struct dentry * dentry,struct task_struct * task,const void * ptr)2331 proc_map_files_instantiate(struct dentry *dentry,
2332 struct task_struct *task, const void *ptr)
2333 {
2334 fmode_t mode = (fmode_t)(unsigned long)ptr;
2335 struct proc_inode *ei;
2336 struct inode *inode;
2337
2338 inode = proc_pid_make_inode(dentry->d_sb, task, S_IFLNK |
2339 ((mode & FMODE_READ ) ? S_IRUSR : 0) |
2340 ((mode & FMODE_WRITE) ? S_IWUSR : 0));
2341 if (!inode)
2342 return ERR_PTR(-ENOENT);
2343
2344 ei = PROC_I(inode);
2345 ei->op.proc_get_link = map_files_get_link;
2346
2347 inode->i_op = &proc_map_files_link_inode_operations;
2348 inode->i_size = 64;
2349
2350 return proc_splice_unmountable(inode, dentry,
2351 &tid_map_files_dentry_operations);
2352 }
2353
proc_map_files_lookup(struct inode * dir,struct dentry * dentry,unsigned int flags)2354 static struct dentry *proc_map_files_lookup(struct inode *dir,
2355 struct dentry *dentry, unsigned int flags)
2356 {
2357 unsigned long vm_start, vm_end;
2358 struct vm_area_struct *vma;
2359 struct task_struct *task;
2360 struct dentry *result;
2361 struct mm_struct *mm;
2362
2363 result = ERR_PTR(-ENOENT);
2364 task = get_proc_task(dir);
2365 if (!task)
2366 goto out;
2367
2368 result = ERR_PTR(-EACCES);
2369 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2370 goto out_put_task;
2371
2372 result = ERR_PTR(-ENOENT);
2373 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2374 goto out_put_task;
2375
2376 mm = get_task_mm(task);
2377 if (!mm)
2378 goto out_put_task;
2379
2380 result = ERR_PTR(-EINTR);
2381 if (mmap_read_lock_killable(mm))
2382 goto out_put_mm;
2383
2384 result = ERR_PTR(-ENOENT);
2385 vma = find_exact_vma(mm, vm_start, vm_end);
2386 if (!vma)
2387 goto out_no_vma;
2388
2389 if (vma->vm_file)
2390 result = proc_map_files_instantiate(dentry, task,
2391 (void *)(unsigned long)vma->vm_file->f_mode);
2392
2393 out_no_vma:
2394 mmap_read_unlock(mm);
2395 out_put_mm:
2396 mmput(mm);
2397 out_put_task:
2398 put_task_struct(task);
2399 out:
2400 return result;
2401 }
2402
2403 static const struct inode_operations proc_map_files_inode_operations = {
2404 .lookup = proc_map_files_lookup,
2405 .permission = proc_fd_permission,
2406 .setattr = proc_setattr,
2407 };
2408
2409 static int
proc_map_files_readdir(struct file * file,struct dir_context * ctx)2410 proc_map_files_readdir(struct file *file, struct dir_context *ctx)
2411 {
2412 struct vm_area_struct *vma;
2413 struct task_struct *task;
2414 struct mm_struct *mm;
2415 unsigned long nr_files, pos, i;
2416 GENRADIX(struct map_files_info) fa;
2417 struct map_files_info *p;
2418 int ret;
2419 struct vma_iterator vmi;
2420
2421 genradix_init(&fa);
2422
2423 ret = -ENOENT;
2424 task = get_proc_task(file_inode(file));
2425 if (!task)
2426 goto out;
2427
2428 ret = -EACCES;
2429 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2430 goto out_put_task;
2431
2432 ret = 0;
2433 if (!dir_emit_dots(file, ctx))
2434 goto out_put_task;
2435
2436 mm = get_task_mm(task);
2437 if (!mm)
2438 goto out_put_task;
2439
2440 ret = mmap_read_lock_killable(mm);
2441 if (ret) {
2442 mmput(mm);
2443 goto out_put_task;
2444 }
2445
2446 nr_files = 0;
2447
2448 /*
2449 * We need two passes here:
2450 *
2451 * 1) Collect vmas of mapped files with mmap_lock taken
2452 * 2) Release mmap_lock and instantiate entries
2453 *
2454 * otherwise we get lockdep complained, since filldir()
2455 * routine might require mmap_lock taken in might_fault().
2456 */
2457
2458 pos = 2;
2459 vma_iter_init(&vmi, mm, 0);
2460 for_each_vma(vmi, vma) {
2461 if (!vma->vm_file)
2462 continue;
2463 if (++pos <= ctx->pos)
2464 continue;
2465
2466 p = genradix_ptr_alloc(&fa, nr_files++, GFP_KERNEL);
2467 if (!p) {
2468 ret = -ENOMEM;
2469 mmap_read_unlock(mm);
2470 mmput(mm);
2471 goto out_put_task;
2472 }
2473
2474 p->start = vma->vm_start;
2475 p->end = VMA_PAD_START(vma);
2476 p->mode = vma->vm_file->f_mode;
2477 }
2478 mmap_read_unlock(mm);
2479 mmput(mm);
2480
2481 for (i = 0; i < nr_files; i++) {
2482 char buf[4 * sizeof(long) + 2]; /* max: %lx-%lx\0 */
2483 unsigned int len;
2484
2485 p = genradix_ptr(&fa, i);
2486 len = snprintf(buf, sizeof(buf), "%lx-%lx", p->start, p->end);
2487 if (!proc_fill_cache(file, ctx,
2488 buf, len,
2489 proc_map_files_instantiate,
2490 task,
2491 (void *)(unsigned long)p->mode))
2492 break;
2493 ctx->pos++;
2494 }
2495
2496 out_put_task:
2497 put_task_struct(task);
2498 out:
2499 genradix_free(&fa);
2500 return ret;
2501 }
2502
2503 static const struct file_operations proc_map_files_operations = {
2504 .read = generic_read_dir,
2505 .iterate_shared = proc_map_files_readdir,
2506 .llseek = generic_file_llseek,
2507 };
2508
2509 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
2510 struct timers_private {
2511 struct pid *pid;
2512 struct task_struct *task;
2513 struct sighand_struct *sighand;
2514 struct pid_namespace *ns;
2515 unsigned long flags;
2516 };
2517
timers_start(struct seq_file * m,loff_t * pos)2518 static void *timers_start(struct seq_file *m, loff_t *pos)
2519 {
2520 struct timers_private *tp = m->private;
2521
2522 tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2523 if (!tp->task)
2524 return ERR_PTR(-ESRCH);
2525
2526 tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2527 if (!tp->sighand)
2528 return ERR_PTR(-ESRCH);
2529
2530 return seq_hlist_start(&tp->task->signal->posix_timers, *pos);
2531 }
2532
timers_next(struct seq_file * m,void * v,loff_t * pos)2533 static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2534 {
2535 struct timers_private *tp = m->private;
2536 return seq_hlist_next(v, &tp->task->signal->posix_timers, pos);
2537 }
2538
timers_stop(struct seq_file * m,void * v)2539 static void timers_stop(struct seq_file *m, void *v)
2540 {
2541 struct timers_private *tp = m->private;
2542
2543 if (tp->sighand) {
2544 unlock_task_sighand(tp->task, &tp->flags);
2545 tp->sighand = NULL;
2546 }
2547
2548 if (tp->task) {
2549 put_task_struct(tp->task);
2550 tp->task = NULL;
2551 }
2552 }
2553
show_timer(struct seq_file * m,void * v)2554 static int show_timer(struct seq_file *m, void *v)
2555 {
2556 struct k_itimer *timer;
2557 struct timers_private *tp = m->private;
2558 int notify;
2559 static const char * const nstr[] = {
2560 [SIGEV_SIGNAL] = "signal",
2561 [SIGEV_NONE] = "none",
2562 [SIGEV_THREAD] = "thread",
2563 };
2564
2565 timer = hlist_entry((struct hlist_node *)v, struct k_itimer, list);
2566 notify = timer->it_sigev_notify;
2567
2568 seq_printf(m, "ID: %d\n", timer->it_id);
2569 seq_printf(m, "signal: %d/%px\n",
2570 timer->sigq->info.si_signo,
2571 timer->sigq->info.si_value.sival_ptr);
2572 seq_printf(m, "notify: %s/%s.%d\n",
2573 nstr[notify & ~SIGEV_THREAD_ID],
2574 (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2575 pid_nr_ns(timer->it_pid, tp->ns));
2576 seq_printf(m, "ClockID: %d\n", timer->it_clock);
2577
2578 return 0;
2579 }
2580
2581 static const struct seq_operations proc_timers_seq_ops = {
2582 .start = timers_start,
2583 .next = timers_next,
2584 .stop = timers_stop,
2585 .show = show_timer,
2586 };
2587
proc_timers_open(struct inode * inode,struct file * file)2588 static int proc_timers_open(struct inode *inode, struct file *file)
2589 {
2590 struct timers_private *tp;
2591
2592 tp = __seq_open_private(file, &proc_timers_seq_ops,
2593 sizeof(struct timers_private));
2594 if (!tp)
2595 return -ENOMEM;
2596
2597 tp->pid = proc_pid(inode);
2598 tp->ns = proc_pid_ns(inode->i_sb);
2599 return 0;
2600 }
2601
2602 static const struct file_operations proc_timers_operations = {
2603 .open = proc_timers_open,
2604 .read = seq_read,
2605 .llseek = seq_lseek,
2606 .release = seq_release_private,
2607 };
2608 #endif
2609
timerslack_ns_write(struct file * file,const char __user * buf,size_t count,loff_t * offset)2610 static ssize_t timerslack_ns_write(struct file *file, const char __user *buf,
2611 size_t count, loff_t *offset)
2612 {
2613 struct inode *inode = file_inode(file);
2614 struct task_struct *p;
2615 u64 slack_ns;
2616 int err;
2617
2618 err = kstrtoull_from_user(buf, count, 10, &slack_ns);
2619 if (err < 0)
2620 return err;
2621
2622 p = get_proc_task(inode);
2623 if (!p)
2624 return -ESRCH;
2625
2626 if (p != current) {
2627 rcu_read_lock();
2628 if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
2629 rcu_read_unlock();
2630 count = -EPERM;
2631 goto out;
2632 }
2633 rcu_read_unlock();
2634
2635 err = security_task_setscheduler(p);
2636 if (err) {
2637 count = err;
2638 goto out;
2639 }
2640 }
2641
2642 task_lock(p);
2643 if (rt_or_dl_task_policy(p))
2644 slack_ns = 0;
2645 else if (slack_ns == 0)
2646 slack_ns = p->default_timer_slack_ns;
2647 p->timer_slack_ns = slack_ns;
2648 task_unlock(p);
2649
2650 out:
2651 put_task_struct(p);
2652
2653 return count;
2654 }
2655
timerslack_ns_show(struct seq_file * m,void * v)2656 static int timerslack_ns_show(struct seq_file *m, void *v)
2657 {
2658 struct inode *inode = m->private;
2659 struct task_struct *p;
2660 int err = 0;
2661
2662 p = get_proc_task(inode);
2663 if (!p)
2664 return -ESRCH;
2665
2666 if (p != current) {
2667 rcu_read_lock();
2668 if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
2669 rcu_read_unlock();
2670 err = -EPERM;
2671 goto out;
2672 }
2673 rcu_read_unlock();
2674
2675 err = security_task_getscheduler(p);
2676 if (err)
2677 goto out;
2678 }
2679
2680 task_lock(p);
2681 seq_printf(m, "%llu\n", p->timer_slack_ns);
2682 task_unlock(p);
2683
2684 out:
2685 put_task_struct(p);
2686
2687 return err;
2688 }
2689
timerslack_ns_open(struct inode * inode,struct file * filp)2690 static int timerslack_ns_open(struct inode *inode, struct file *filp)
2691 {
2692 return single_open(filp, timerslack_ns_show, inode);
2693 }
2694
2695 static const struct file_operations proc_pid_set_timerslack_ns_operations = {
2696 .open = timerslack_ns_open,
2697 .read = seq_read,
2698 .write = timerslack_ns_write,
2699 .llseek = seq_lseek,
2700 .release = single_release,
2701 };
2702
proc_pident_instantiate(struct dentry * dentry,struct task_struct * task,const void * ptr)2703 static struct dentry *proc_pident_instantiate(struct dentry *dentry,
2704 struct task_struct *task, const void *ptr)
2705 {
2706 const struct pid_entry *p = ptr;
2707 struct inode *inode;
2708 struct proc_inode *ei;
2709
2710 inode = proc_pid_make_inode(dentry->d_sb, task, p->mode);
2711 if (!inode)
2712 return ERR_PTR(-ENOENT);
2713
2714 ei = PROC_I(inode);
2715 if (S_ISDIR(inode->i_mode))
2716 set_nlink(inode, 2); /* Use getattr to fix if necessary */
2717 if (p->iop)
2718 inode->i_op = p->iop;
2719 if (p->fop)
2720 inode->i_fop = p->fop;
2721 ei->op = p->op;
2722 pid_update_inode(task, inode);
2723 d_set_d_op(dentry, &pid_dentry_operations);
2724 return d_splice_alias(inode, dentry);
2725 }
2726
proc_pident_lookup(struct inode * dir,struct dentry * dentry,const struct pid_entry * p,const struct pid_entry * end)2727 static struct dentry *proc_pident_lookup(struct inode *dir,
2728 struct dentry *dentry,
2729 const struct pid_entry *p,
2730 const struct pid_entry *end)
2731 {
2732 struct task_struct *task = get_proc_task(dir);
2733 struct dentry *res = ERR_PTR(-ENOENT);
2734
2735 if (!task)
2736 goto out_no_task;
2737
2738 /*
2739 * Yes, it does not scale. And it should not. Don't add
2740 * new entries into /proc/<tgid>/ without very good reasons.
2741 */
2742 for (; p < end; p++) {
2743 if (p->len != dentry->d_name.len)
2744 continue;
2745 if (!memcmp(dentry->d_name.name, p->name, p->len)) {
2746 res = proc_pident_instantiate(dentry, task, p);
2747 break;
2748 }
2749 }
2750 put_task_struct(task);
2751 out_no_task:
2752 return res;
2753 }
2754
proc_pident_readdir(struct file * file,struct dir_context * ctx,const struct pid_entry * ents,unsigned int nents)2755 static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
2756 const struct pid_entry *ents, unsigned int nents)
2757 {
2758 struct task_struct *task = get_proc_task(file_inode(file));
2759 const struct pid_entry *p;
2760
2761 if (!task)
2762 return -ENOENT;
2763
2764 if (!dir_emit_dots(file, ctx))
2765 goto out;
2766
2767 if (ctx->pos >= nents + 2)
2768 goto out;
2769
2770 for (p = ents + (ctx->pos - 2); p < ents + nents; p++) {
2771 if (!proc_fill_cache(file, ctx, p->name, p->len,
2772 proc_pident_instantiate, task, p))
2773 break;
2774 ctx->pos++;
2775 }
2776 out:
2777 put_task_struct(task);
2778 return 0;
2779 }
2780
2781 #ifdef CONFIG_SECURITY
proc_pid_attr_open(struct inode * inode,struct file * file)2782 static int proc_pid_attr_open(struct inode *inode, struct file *file)
2783 {
2784 file->private_data = NULL;
2785 __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS);
2786 return 0;
2787 }
2788
proc_pid_attr_read(struct file * file,char __user * buf,size_t count,loff_t * ppos)2789 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2790 size_t count, loff_t *ppos)
2791 {
2792 struct inode * inode = file_inode(file);
2793 char *p = NULL;
2794 ssize_t length;
2795 struct task_struct *task = get_proc_task(inode);
2796
2797 if (!task)
2798 return -ESRCH;
2799
2800 length = security_getprocattr(task, PROC_I(inode)->op.lsmid,
2801 file->f_path.dentry->d_name.name,
2802 &p);
2803 put_task_struct(task);
2804 if (length > 0)
2805 length = simple_read_from_buffer(buf, count, ppos, p, length);
2806 kfree(p);
2807 return length;
2808 }
2809
proc_pid_attr_write(struct file * file,const char __user * buf,size_t count,loff_t * ppos)2810 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2811 size_t count, loff_t *ppos)
2812 {
2813 struct inode * inode = file_inode(file);
2814 struct task_struct *task;
2815 void *page;
2816 int rv;
2817
2818 /* A task may only write when it was the opener. */
2819 if (file->private_data != current->mm)
2820 return -EPERM;
2821
2822 rcu_read_lock();
2823 task = pid_task(proc_pid(inode), PIDTYPE_PID);
2824 if (!task) {
2825 rcu_read_unlock();
2826 return -ESRCH;
2827 }
2828 /* A task may only write its own attributes. */
2829 if (current != task) {
2830 rcu_read_unlock();
2831 return -EACCES;
2832 }
2833 /* Prevent changes to overridden credentials. */
2834 if (current_cred() != current_real_cred()) {
2835 rcu_read_unlock();
2836 return -EBUSY;
2837 }
2838 rcu_read_unlock();
2839
2840 if (count > PAGE_SIZE)
2841 count = PAGE_SIZE;
2842
2843 /* No partial writes. */
2844 if (*ppos != 0)
2845 return -EINVAL;
2846
2847 page = memdup_user(buf, count);
2848 if (IS_ERR(page)) {
2849 rv = PTR_ERR(page);
2850 goto out;
2851 }
2852
2853 /* Guard against adverse ptrace interaction */
2854 rv = mutex_lock_interruptible(¤t->signal->cred_guard_mutex);
2855 if (rv < 0)
2856 goto out_free;
2857
2858 rv = security_setprocattr(PROC_I(inode)->op.lsmid,
2859 file->f_path.dentry->d_name.name, page,
2860 count);
2861 mutex_unlock(¤t->signal->cred_guard_mutex);
2862 out_free:
2863 kfree(page);
2864 out:
2865 return rv;
2866 }
2867
2868 static const struct file_operations proc_pid_attr_operations = {
2869 .open = proc_pid_attr_open,
2870 .read = proc_pid_attr_read,
2871 .write = proc_pid_attr_write,
2872 .llseek = generic_file_llseek,
2873 .release = mem_release,
2874 };
2875
2876 #define LSM_DIR_OPS(LSM) \
2877 static int proc_##LSM##_attr_dir_iterate(struct file *filp, \
2878 struct dir_context *ctx) \
2879 { \
2880 return proc_pident_readdir(filp, ctx, \
2881 LSM##_attr_dir_stuff, \
2882 ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2883 } \
2884 \
2885 static const struct file_operations proc_##LSM##_attr_dir_ops = { \
2886 .read = generic_read_dir, \
2887 .iterate_shared = proc_##LSM##_attr_dir_iterate, \
2888 .llseek = default_llseek, \
2889 }; \
2890 \
2891 static struct dentry *proc_##LSM##_attr_dir_lookup(struct inode *dir, \
2892 struct dentry *dentry, unsigned int flags) \
2893 { \
2894 return proc_pident_lookup(dir, dentry, \
2895 LSM##_attr_dir_stuff, \
2896 LSM##_attr_dir_stuff + ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2897 } \
2898 \
2899 static const struct inode_operations proc_##LSM##_attr_dir_inode_ops = { \
2900 .lookup = proc_##LSM##_attr_dir_lookup, \
2901 .getattr = pid_getattr, \
2902 .setattr = proc_setattr, \
2903 }
2904
2905 #ifdef CONFIG_SECURITY_SMACK
2906 static const struct pid_entry smack_attr_dir_stuff[] = {
2907 ATTR(LSM_ID_SMACK, "current", 0666),
2908 };
2909 LSM_DIR_OPS(smack);
2910 #endif
2911
2912 #ifdef CONFIG_SECURITY_APPARMOR
2913 static const struct pid_entry apparmor_attr_dir_stuff[] = {
2914 ATTR(LSM_ID_APPARMOR, "current", 0666),
2915 ATTR(LSM_ID_APPARMOR, "prev", 0444),
2916 ATTR(LSM_ID_APPARMOR, "exec", 0666),
2917 };
2918 LSM_DIR_OPS(apparmor);
2919 #endif
2920
2921 static const struct pid_entry attr_dir_stuff[] = {
2922 ATTR(LSM_ID_UNDEF, "current", 0666),
2923 ATTR(LSM_ID_UNDEF, "prev", 0444),
2924 ATTR(LSM_ID_UNDEF, "exec", 0666),
2925 ATTR(LSM_ID_UNDEF, "fscreate", 0666),
2926 ATTR(LSM_ID_UNDEF, "keycreate", 0666),
2927 ATTR(LSM_ID_UNDEF, "sockcreate", 0666),
2928 #ifdef CONFIG_SECURITY_SMACK
2929 DIR("smack", 0555,
2930 proc_smack_attr_dir_inode_ops, proc_smack_attr_dir_ops),
2931 #endif
2932 #ifdef CONFIG_SECURITY_APPARMOR
2933 DIR("apparmor", 0555,
2934 proc_apparmor_attr_dir_inode_ops, proc_apparmor_attr_dir_ops),
2935 #endif
2936 };
2937
proc_attr_dir_readdir(struct file * file,struct dir_context * ctx)2938 static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
2939 {
2940 return proc_pident_readdir(file, ctx,
2941 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2942 }
2943
2944 static const struct file_operations proc_attr_dir_operations = {
2945 .read = generic_read_dir,
2946 .iterate_shared = proc_attr_dir_readdir,
2947 .llseek = generic_file_llseek,
2948 };
2949
proc_attr_dir_lookup(struct inode * dir,struct dentry * dentry,unsigned int flags)2950 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2951 struct dentry *dentry, unsigned int flags)
2952 {
2953 return proc_pident_lookup(dir, dentry,
2954 attr_dir_stuff,
2955 attr_dir_stuff + ARRAY_SIZE(attr_dir_stuff));
2956 }
2957
2958 static const struct inode_operations proc_attr_dir_inode_operations = {
2959 .lookup = proc_attr_dir_lookup,
2960 .getattr = pid_getattr,
2961 .setattr = proc_setattr,
2962 };
2963
2964 #endif
2965
2966 #ifdef CONFIG_ELF_CORE
proc_coredump_filter_read(struct file * file,char __user * buf,size_t count,loff_t * ppos)2967 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2968 size_t count, loff_t *ppos)
2969 {
2970 struct task_struct *task = get_proc_task(file_inode(file));
2971 struct mm_struct *mm;
2972 char buffer[PROC_NUMBUF];
2973 size_t len;
2974 int ret;
2975
2976 if (!task)
2977 return -ESRCH;
2978
2979 ret = 0;
2980 mm = get_task_mm(task);
2981 if (mm) {
2982 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2983 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2984 MMF_DUMP_FILTER_SHIFT));
2985 mmput(mm);
2986 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2987 }
2988
2989 put_task_struct(task);
2990
2991 return ret;
2992 }
2993
proc_coredump_filter_write(struct file * file,const char __user * buf,size_t count,loff_t * ppos)2994 static ssize_t proc_coredump_filter_write(struct file *file,
2995 const char __user *buf,
2996 size_t count,
2997 loff_t *ppos)
2998 {
2999 struct task_struct *task;
3000 struct mm_struct *mm;
3001 unsigned int val;
3002 int ret;
3003 int i;
3004 unsigned long mask;
3005
3006 ret = kstrtouint_from_user(buf, count, 0, &val);
3007 if (ret < 0)
3008 return ret;
3009
3010 ret = -ESRCH;
3011 task = get_proc_task(file_inode(file));
3012 if (!task)
3013 goto out_no_task;
3014
3015 mm = get_task_mm(task);
3016 if (!mm)
3017 goto out_no_mm;
3018 ret = 0;
3019
3020 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
3021 if (val & mask)
3022 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
3023 else
3024 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
3025 }
3026
3027 mmput(mm);
3028 out_no_mm:
3029 put_task_struct(task);
3030 out_no_task:
3031 if (ret < 0)
3032 return ret;
3033 return count;
3034 }
3035
3036 static const struct file_operations proc_coredump_filter_operations = {
3037 .read = proc_coredump_filter_read,
3038 .write = proc_coredump_filter_write,
3039 .llseek = generic_file_llseek,
3040 };
3041 #endif
3042
3043 #ifdef CONFIG_TASK_IO_ACCOUNTING
do_io_accounting(struct task_struct * task,struct seq_file * m,int whole)3044 static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
3045 {
3046 struct task_io_accounting acct;
3047 int result;
3048
3049 result = down_read_killable(&task->signal->exec_update_lock);
3050 if (result)
3051 return result;
3052
3053 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) {
3054 result = -EACCES;
3055 goto out_unlock;
3056 }
3057
3058 if (whole) {
3059 struct signal_struct *sig = task->signal;
3060 struct task_struct *t;
3061 unsigned int seq = 1;
3062 unsigned long flags;
3063
3064 rcu_read_lock();
3065 do {
3066 seq++; /* 2 on the 1st/lockless path, otherwise odd */
3067 flags = read_seqbegin_or_lock_irqsave(&sig->stats_lock, &seq);
3068
3069 acct = sig->ioac;
3070 __for_each_thread(sig, t)
3071 task_io_accounting_add(&acct, &t->ioac);
3072
3073 } while (need_seqretry(&sig->stats_lock, seq));
3074 done_seqretry_irqrestore(&sig->stats_lock, seq, flags);
3075 rcu_read_unlock();
3076 } else {
3077 acct = task->ioac;
3078 }
3079
3080 seq_printf(m,
3081 "rchar: %llu\n"
3082 "wchar: %llu\n"
3083 "syscr: %llu\n"
3084 "syscw: %llu\n"
3085 "read_bytes: %llu\n"
3086 "write_bytes: %llu\n"
3087 "cancelled_write_bytes: %llu\n",
3088 (unsigned long long)acct.rchar,
3089 (unsigned long long)acct.wchar,
3090 (unsigned long long)acct.syscr,
3091 (unsigned long long)acct.syscw,
3092 (unsigned long long)acct.read_bytes,
3093 (unsigned long long)acct.write_bytes,
3094 (unsigned long long)acct.cancelled_write_bytes);
3095 result = 0;
3096
3097 out_unlock:
3098 up_read(&task->signal->exec_update_lock);
3099 return result;
3100 }
3101
proc_tid_io_accounting(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)3102 static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
3103 struct pid *pid, struct task_struct *task)
3104 {
3105 return do_io_accounting(task, m, 0);
3106 }
3107
proc_tgid_io_accounting(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)3108 static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
3109 struct pid *pid, struct task_struct *task)
3110 {
3111 return do_io_accounting(task, m, 1);
3112 }
3113 #endif /* CONFIG_TASK_IO_ACCOUNTING */
3114
3115 #ifdef CONFIG_USER_NS
proc_id_map_open(struct inode * inode,struct file * file,const struct seq_operations * seq_ops)3116 static int proc_id_map_open(struct inode *inode, struct file *file,
3117 const struct seq_operations *seq_ops)
3118 {
3119 struct user_namespace *ns = NULL;
3120 struct task_struct *task;
3121 struct seq_file *seq;
3122 int ret = -EINVAL;
3123
3124 task = get_proc_task(inode);
3125 if (task) {
3126 rcu_read_lock();
3127 ns = get_user_ns(task_cred_xxx(task, user_ns));
3128 rcu_read_unlock();
3129 put_task_struct(task);
3130 }
3131 if (!ns)
3132 goto err;
3133
3134 ret = seq_open(file, seq_ops);
3135 if (ret)
3136 goto err_put_ns;
3137
3138 seq = file->private_data;
3139 seq->private = ns;
3140
3141 return 0;
3142 err_put_ns:
3143 put_user_ns(ns);
3144 err:
3145 return ret;
3146 }
3147
proc_id_map_release(struct inode * inode,struct file * file)3148 static int proc_id_map_release(struct inode *inode, struct file *file)
3149 {
3150 struct seq_file *seq = file->private_data;
3151 struct user_namespace *ns = seq->private;
3152 put_user_ns(ns);
3153 return seq_release(inode, file);
3154 }
3155
proc_uid_map_open(struct inode * inode,struct file * file)3156 static int proc_uid_map_open(struct inode *inode, struct file *file)
3157 {
3158 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
3159 }
3160
proc_gid_map_open(struct inode * inode,struct file * file)3161 static int proc_gid_map_open(struct inode *inode, struct file *file)
3162 {
3163 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
3164 }
3165
proc_projid_map_open(struct inode * inode,struct file * file)3166 static int proc_projid_map_open(struct inode *inode, struct file *file)
3167 {
3168 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
3169 }
3170
3171 static const struct file_operations proc_uid_map_operations = {
3172 .open = proc_uid_map_open,
3173 .write = proc_uid_map_write,
3174 .read = seq_read,
3175 .llseek = seq_lseek,
3176 .release = proc_id_map_release,
3177 };
3178
3179 static const struct file_operations proc_gid_map_operations = {
3180 .open = proc_gid_map_open,
3181 .write = proc_gid_map_write,
3182 .read = seq_read,
3183 .llseek = seq_lseek,
3184 .release = proc_id_map_release,
3185 };
3186
3187 static const struct file_operations proc_projid_map_operations = {
3188 .open = proc_projid_map_open,
3189 .write = proc_projid_map_write,
3190 .read = seq_read,
3191 .llseek = seq_lseek,
3192 .release = proc_id_map_release,
3193 };
3194
proc_setgroups_open(struct inode * inode,struct file * file)3195 static int proc_setgroups_open(struct inode *inode, struct file *file)
3196 {
3197 struct user_namespace *ns = NULL;
3198 struct task_struct *task;
3199 int ret;
3200
3201 ret = -ESRCH;
3202 task = get_proc_task(inode);
3203 if (task) {
3204 rcu_read_lock();
3205 ns = get_user_ns(task_cred_xxx(task, user_ns));
3206 rcu_read_unlock();
3207 put_task_struct(task);
3208 }
3209 if (!ns)
3210 goto err;
3211
3212 if (file->f_mode & FMODE_WRITE) {
3213 ret = -EACCES;
3214 if (!ns_capable(ns, CAP_SYS_ADMIN))
3215 goto err_put_ns;
3216 }
3217
3218 ret = single_open(file, &proc_setgroups_show, ns);
3219 if (ret)
3220 goto err_put_ns;
3221
3222 return 0;
3223 err_put_ns:
3224 put_user_ns(ns);
3225 err:
3226 return ret;
3227 }
3228
proc_setgroups_release(struct inode * inode,struct file * file)3229 static int proc_setgroups_release(struct inode *inode, struct file *file)
3230 {
3231 struct seq_file *seq = file->private_data;
3232 struct user_namespace *ns = seq->private;
3233 int ret = single_release(inode, file);
3234 put_user_ns(ns);
3235 return ret;
3236 }
3237
3238 static const struct file_operations proc_setgroups_operations = {
3239 .open = proc_setgroups_open,
3240 .write = proc_setgroups_write,
3241 .read = seq_read,
3242 .llseek = seq_lseek,
3243 .release = proc_setgroups_release,
3244 };
3245 #endif /* CONFIG_USER_NS */
3246
proc_pid_personality(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)3247 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
3248 struct pid *pid, struct task_struct *task)
3249 {
3250 int err = lock_trace(task);
3251 if (!err) {
3252 seq_printf(m, "%08x\n", task->personality);
3253 unlock_trace(task);
3254 }
3255 return err;
3256 }
3257
3258 #ifdef CONFIG_LIVEPATCH
proc_pid_patch_state(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)3259 static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns,
3260 struct pid *pid, struct task_struct *task)
3261 {
3262 seq_printf(m, "%d\n", task->patch_state);
3263 return 0;
3264 }
3265 #endif /* CONFIG_LIVEPATCH */
3266
3267 #ifdef CONFIG_KSM
proc_pid_ksm_merging_pages(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)3268 static int proc_pid_ksm_merging_pages(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, "%lu\n", mm->ksm_merging_pages);
3276 mmput(mm);
3277 }
3278
3279 return 0;
3280 }
proc_pid_ksm_stat(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)3281 static int proc_pid_ksm_stat(struct seq_file *m, struct pid_namespace *ns,
3282 struct pid *pid, struct task_struct *task)
3283 {
3284 struct mm_struct *mm;
3285
3286 mm = get_task_mm(task);
3287 if (mm) {
3288 seq_printf(m, "ksm_rmap_items %lu\n", mm->ksm_rmap_items);
3289 seq_printf(m, "ksm_zero_pages %ld\n", mm_ksm_zero_pages(mm));
3290 seq_printf(m, "ksm_merging_pages %lu\n", mm->ksm_merging_pages);
3291 seq_printf(m, "ksm_process_profit %ld\n", ksm_process_profit(mm));
3292 mmput(mm);
3293 }
3294
3295 return 0;
3296 }
3297 #endif /* CONFIG_KSM */
3298
3299 #ifdef CONFIG_STACKLEAK_METRICS
proc_stack_depth(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)3300 static int proc_stack_depth(struct seq_file *m, struct pid_namespace *ns,
3301 struct pid *pid, struct task_struct *task)
3302 {
3303 unsigned long prev_depth = THREAD_SIZE -
3304 (task->prev_lowest_stack & (THREAD_SIZE - 1));
3305 unsigned long depth = THREAD_SIZE -
3306 (task->lowest_stack & (THREAD_SIZE - 1));
3307
3308 seq_printf(m, "previous stack depth: %lu\nstack depth: %lu\n",
3309 prev_depth, depth);
3310 return 0;
3311 }
3312 #endif /* CONFIG_STACKLEAK_METRICS */
3313
3314 /*
3315 * Thread groups
3316 */
3317 static const struct file_operations proc_task_operations;
3318 static const struct inode_operations proc_task_inode_operations;
3319
3320 static const struct pid_entry tgid_base_stuff[] = {
3321 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
3322 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3323 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
3324 DIR("fdinfo", S_IRUGO|S_IXUGO, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3325 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3326 #ifdef CONFIG_NET
3327 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3328 #endif
3329 REG("environ", S_IRUSR, proc_environ_operations),
3330 REG("auxv", S_IRUSR, proc_auxv_operations),
3331 ONE("status", S_IRUGO, proc_pid_status),
3332 ONE("personality", S_IRUSR, proc_pid_personality),
3333 ONE("limits", S_IRUGO, proc_pid_limits),
3334 #ifdef CONFIG_SCHED_DEBUG
3335 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3336 #endif
3337 #ifdef CONFIG_SCHED_AUTOGROUP
3338 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
3339 #endif
3340 #ifdef CONFIG_TIME_NS
3341 REG("timens_offsets", S_IRUGO|S_IWUSR, proc_timens_offsets_operations),
3342 #endif
3343 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3344 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3345 ONE("syscall", S_IRUSR, proc_pid_syscall),
3346 #endif
3347 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
3348 ONE("stat", S_IRUGO, proc_tgid_stat),
3349 ONE("statm", S_IRUGO, proc_pid_statm),
3350 REG("maps", S_IRUGO, proc_pid_maps_operations),
3351 #ifdef CONFIG_NUMA
3352 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
3353 #endif
3354 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3355 LNK("cwd", proc_cwd_link),
3356 LNK("root", proc_root_link),
3357 LNK("exe", proc_exe_link),
3358 REG("mounts", S_IRUGO, proc_mounts_operations),
3359 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3360 REG("mountstats", S_IRUSR, proc_mountstats_operations),
3361 #ifdef CONFIG_PROC_PAGE_MONITOR
3362 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3363 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
3364 REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
3365 REG("pagemap", S_IRUSR, proc_pagemap_operations),
3366 #endif
3367 #ifdef CONFIG_SECURITY
3368 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3369 #endif
3370 #ifdef CONFIG_KALLSYMS
3371 ONE("wchan", S_IRUGO, proc_pid_wchan),
3372 #endif
3373 #ifdef CONFIG_STACKTRACE
3374 ONE("stack", S_IRUSR, proc_pid_stack),
3375 #endif
3376 #ifdef CONFIG_SCHED_INFO
3377 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
3378 #endif
3379 #ifdef CONFIG_LATENCYTOP
3380 REG("latency", S_IRUGO, proc_lstats_operations),
3381 #endif
3382 #ifdef CONFIG_PROC_PID_CPUSET
3383 ONE("cpuset", S_IRUGO, proc_cpuset_show),
3384 #endif
3385 #ifdef CONFIG_CGROUPS
3386 ONE("cgroup", S_IRUGO, proc_cgroup_show),
3387 #endif
3388 #ifdef CONFIG_PROC_CPU_RESCTRL
3389 ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show),
3390 #endif
3391 ONE("oom_score", S_IRUGO, proc_oom_score),
3392 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3393 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3394 #ifdef CONFIG_AUDIT
3395 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3396 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3397 #endif
3398 #ifdef CONFIG_FAULT_INJECTION
3399 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3400 REG("fail-nth", 0644, proc_fail_nth_operations),
3401 #endif
3402 #ifdef CONFIG_ELF_CORE
3403 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3404 #endif
3405 #ifdef CONFIG_TASK_IO_ACCOUNTING
3406 ONE("io", S_IRUSR, proc_tgid_io_accounting),
3407 #endif
3408 #ifdef CONFIG_USER_NS
3409 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3410 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3411 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3412 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
3413 #endif
3414 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
3415 REG("timers", S_IRUGO, proc_timers_operations),
3416 #endif
3417 REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations),
3418 #ifdef CONFIG_LIVEPATCH
3419 ONE("patch_state", S_IRUSR, proc_pid_patch_state),
3420 #endif
3421 #ifdef CONFIG_CPU_FREQ_TIMES
3422 ONE("time_in_state", 0444, proc_time_in_state_show),
3423 #endif
3424 #ifdef CONFIG_STACKLEAK_METRICS
3425 ONE("stack_depth", S_IRUGO, proc_stack_depth),
3426 #endif
3427 #ifdef CONFIG_PROC_PID_ARCH_STATUS
3428 ONE("arch_status", S_IRUGO, proc_pid_arch_status),
3429 #endif
3430 #ifdef CONFIG_SECCOMP_CACHE_DEBUG
3431 ONE("seccomp_cache", S_IRUSR, proc_pid_seccomp_cache),
3432 #endif
3433 #ifdef CONFIG_KSM
3434 ONE("ksm_merging_pages", S_IRUSR, proc_pid_ksm_merging_pages),
3435 ONE("ksm_stat", S_IRUSR, proc_pid_ksm_stat),
3436 #endif
3437 };
3438
proc_tgid_base_readdir(struct file * file,struct dir_context * ctx)3439 static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
3440 {
3441 return proc_pident_readdir(file, ctx,
3442 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3443 }
3444
3445 static const struct file_operations proc_tgid_base_operations = {
3446 .read = generic_read_dir,
3447 .iterate_shared = proc_tgid_base_readdir,
3448 .llseek = generic_file_llseek,
3449 };
3450
tgid_pidfd_to_pid(const struct file * file)3451 struct pid *tgid_pidfd_to_pid(const struct file *file)
3452 {
3453 if (file->f_op != &proc_tgid_base_operations)
3454 return ERR_PTR(-EBADF);
3455
3456 return proc_pid(file_inode(file));
3457 }
3458
proc_tgid_base_lookup(struct inode * dir,struct dentry * dentry,unsigned int flags)3459 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3460 {
3461 return proc_pident_lookup(dir, dentry,
3462 tgid_base_stuff,
3463 tgid_base_stuff + ARRAY_SIZE(tgid_base_stuff));
3464 }
3465
3466 static const struct inode_operations proc_tgid_base_inode_operations = {
3467 .lookup = proc_tgid_base_lookup,
3468 .getattr = pid_getattr,
3469 .setattr = proc_setattr,
3470 .permission = proc_pid_permission,
3471 };
3472
3473 /**
3474 * proc_flush_pid - Remove dcache entries for @pid from the /proc dcache.
3475 * @pid: pid that should be flushed.
3476 *
3477 * This function walks a list of inodes (that belong to any proc
3478 * filesystem) that are attached to the pid and flushes them from
3479 * the dentry cache.
3480 *
3481 * It is safe and reasonable to cache /proc entries for a task until
3482 * that task exits. After that they just clog up the dcache with
3483 * useless entries, possibly causing useful dcache entries to be
3484 * flushed instead. This routine is provided to flush those useless
3485 * dcache entries when a process is reaped.
3486 *
3487 * NOTE: This routine is just an optimization so it does not guarantee
3488 * that no dcache entries will exist after a process is reaped
3489 * it just makes it very unlikely that any will persist.
3490 */
3491
proc_flush_pid(struct pid * pid)3492 void proc_flush_pid(struct pid *pid)
3493 {
3494 proc_invalidate_siblings_dcache(&pid->inodes, &pid->lock);
3495 }
3496
proc_pid_instantiate(struct dentry * dentry,struct task_struct * task,const void * ptr)3497 static struct dentry *proc_pid_instantiate(struct dentry * dentry,
3498 struct task_struct *task, const void *ptr)
3499 {
3500 struct inode *inode;
3501
3502 inode = proc_pid_make_base_inode(dentry->d_sb, task,
3503 S_IFDIR | S_IRUGO | S_IXUGO);
3504 if (!inode)
3505 return ERR_PTR(-ENOENT);
3506
3507 inode->i_op = &proc_tgid_base_inode_operations;
3508 inode->i_fop = &proc_tgid_base_operations;
3509 inode->i_flags|=S_IMMUTABLE;
3510
3511 set_nlink(inode, nlink_tgid);
3512 pid_update_inode(task, inode);
3513
3514 d_set_d_op(dentry, &pid_dentry_operations);
3515 return d_splice_alias(inode, dentry);
3516 }
3517
proc_pid_lookup(struct dentry * dentry,unsigned int flags)3518 struct dentry *proc_pid_lookup(struct dentry *dentry, unsigned int flags)
3519 {
3520 struct task_struct *task;
3521 unsigned tgid;
3522 struct proc_fs_info *fs_info;
3523 struct pid_namespace *ns;
3524 struct dentry *result = ERR_PTR(-ENOENT);
3525
3526 tgid = name_to_int(&dentry->d_name);
3527 if (tgid == ~0U)
3528 goto out;
3529
3530 fs_info = proc_sb_info(dentry->d_sb);
3531 ns = fs_info->pid_ns;
3532 rcu_read_lock();
3533 task = find_task_by_pid_ns(tgid, ns);
3534 if (task)
3535 get_task_struct(task);
3536 rcu_read_unlock();
3537 if (!task)
3538 goto out;
3539
3540 /* Limit procfs to only ptraceable tasks */
3541 if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE) {
3542 if (!has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS))
3543 goto out_put_task;
3544 }
3545
3546 result = proc_pid_instantiate(dentry, task, NULL);
3547 out_put_task:
3548 put_task_struct(task);
3549 out:
3550 return result;
3551 }
3552
3553 /*
3554 * Find the first task with tgid >= tgid
3555 *
3556 */
3557 struct tgid_iter {
3558 unsigned int tgid;
3559 struct task_struct *task;
3560 };
next_tgid(struct pid_namespace * ns,struct tgid_iter iter)3561 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3562 {
3563 struct pid *pid;
3564
3565 if (iter.task)
3566 put_task_struct(iter.task);
3567 rcu_read_lock();
3568 retry:
3569 iter.task = NULL;
3570 pid = find_ge_pid(iter.tgid, ns);
3571 if (pid) {
3572 iter.tgid = pid_nr_ns(pid, ns);
3573 iter.task = pid_task(pid, PIDTYPE_TGID);
3574 if (!iter.task) {
3575 iter.tgid += 1;
3576 goto retry;
3577 }
3578 get_task_struct(iter.task);
3579 }
3580 rcu_read_unlock();
3581 return iter;
3582 }
3583
3584 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
3585
3586 /* for the /proc/ directory itself, after non-process stuff has been done */
proc_pid_readdir(struct file * file,struct dir_context * ctx)3587 int proc_pid_readdir(struct file *file, struct dir_context *ctx)
3588 {
3589 struct tgid_iter iter;
3590 struct proc_fs_info *fs_info = proc_sb_info(file_inode(file)->i_sb);
3591 struct pid_namespace *ns = proc_pid_ns(file_inode(file)->i_sb);
3592 loff_t pos = ctx->pos;
3593
3594 if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
3595 return 0;
3596
3597 if (pos == TGID_OFFSET - 2) {
3598 struct inode *inode = d_inode(fs_info->proc_self);
3599 if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
3600 return 0;
3601 ctx->pos = pos = pos + 1;
3602 }
3603 if (pos == TGID_OFFSET - 1) {
3604 struct inode *inode = d_inode(fs_info->proc_thread_self);
3605 if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3606 return 0;
3607 ctx->pos = pos = pos + 1;
3608 }
3609 iter.tgid = pos - TGID_OFFSET;
3610 iter.task = NULL;
3611 for (iter = next_tgid(ns, iter);
3612 iter.task;
3613 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3614 char name[10 + 1];
3615 unsigned int len;
3616
3617 cond_resched();
3618 if (!has_pid_permissions(fs_info, iter.task, HIDEPID_INVISIBLE))
3619 continue;
3620
3621 len = snprintf(name, sizeof(name), "%u", iter.tgid);
3622 ctx->pos = iter.tgid + TGID_OFFSET;
3623 if (!proc_fill_cache(file, ctx, name, len,
3624 proc_pid_instantiate, iter.task, NULL)) {
3625 put_task_struct(iter.task);
3626 return 0;
3627 }
3628 }
3629 ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
3630 return 0;
3631 }
3632
3633 /*
3634 * proc_tid_comm_permission is a special permission function exclusively
3635 * used for the node /proc/<pid>/task/<tid>/comm.
3636 * It bypasses generic permission checks in the case where a task of the same
3637 * task group attempts to access the node.
3638 * The rationale behind this is that glibc and bionic access this node for
3639 * cross thread naming (pthread_set/getname_np(!self)). However, if
3640 * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3641 * which locks out the cross thread naming implementation.
3642 * This function makes sure that the node is always accessible for members of
3643 * same thread group.
3644 */
proc_tid_comm_permission(struct mnt_idmap * idmap,struct inode * inode,int mask)3645 static int proc_tid_comm_permission(struct mnt_idmap *idmap,
3646 struct inode *inode, int mask)
3647 {
3648 bool is_same_tgroup;
3649 struct task_struct *task;
3650
3651 task = get_proc_task(inode);
3652 if (!task)
3653 return -ESRCH;
3654 is_same_tgroup = same_thread_group(current, task);
3655 put_task_struct(task);
3656
3657 if (likely(is_same_tgroup && !(mask & MAY_EXEC))) {
3658 /* This file (/proc/<pid>/task/<tid>/comm) can always be
3659 * read or written by the members of the corresponding
3660 * thread group.
3661 */
3662 return 0;
3663 }
3664
3665 return generic_permission(&nop_mnt_idmap, inode, mask);
3666 }
3667
3668 static const struct inode_operations proc_tid_comm_inode_operations = {
3669 .setattr = proc_setattr,
3670 .permission = proc_tid_comm_permission,
3671 };
3672
3673 /*
3674 * Tasks
3675 */
3676 static const struct pid_entry tid_base_stuff[] = {
3677 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3678 DIR("fdinfo", S_IRUGO|S_IXUGO, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3679 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3680 #ifdef CONFIG_NET
3681 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3682 #endif
3683 REG("environ", S_IRUSR, proc_environ_operations),
3684 REG("auxv", S_IRUSR, proc_auxv_operations),
3685 ONE("status", S_IRUGO, proc_pid_status),
3686 ONE("personality", S_IRUSR, proc_pid_personality),
3687 ONE("limits", S_IRUGO, proc_pid_limits),
3688 #ifdef CONFIG_SCHED_DEBUG
3689 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3690 #endif
3691 NOD("comm", S_IFREG|S_IRUGO|S_IWUSR,
3692 &proc_tid_comm_inode_operations,
3693 &proc_pid_set_comm_operations, {}),
3694 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3695 ONE("syscall", S_IRUSR, proc_pid_syscall),
3696 #endif
3697 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
3698 ONE("stat", S_IRUGO, proc_tid_stat),
3699 ONE("statm", S_IRUGO, proc_pid_statm),
3700 REG("maps", S_IRUGO, proc_pid_maps_operations),
3701 #ifdef CONFIG_PROC_CHILDREN
3702 REG("children", S_IRUGO, proc_tid_children_operations),
3703 #endif
3704 #ifdef CONFIG_NUMA
3705 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
3706 #endif
3707 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3708 LNK("cwd", proc_cwd_link),
3709 LNK("root", proc_root_link),
3710 LNK("exe", proc_exe_link),
3711 REG("mounts", S_IRUGO, proc_mounts_operations),
3712 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3713 #ifdef CONFIG_PROC_PAGE_MONITOR
3714 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3715 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
3716 REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
3717 REG("pagemap", S_IRUSR, proc_pagemap_operations),
3718 #endif
3719 #ifdef CONFIG_SECURITY
3720 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3721 #endif
3722 #ifdef CONFIG_KALLSYMS
3723 ONE("wchan", S_IRUGO, proc_pid_wchan),
3724 #endif
3725 #ifdef CONFIG_STACKTRACE
3726 ONE("stack", S_IRUSR, proc_pid_stack),
3727 #endif
3728 #ifdef CONFIG_SCHED_INFO
3729 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
3730 #endif
3731 #ifdef CONFIG_LATENCYTOP
3732 REG("latency", S_IRUGO, proc_lstats_operations),
3733 #endif
3734 #ifdef CONFIG_PROC_PID_CPUSET
3735 ONE("cpuset", S_IRUGO, proc_cpuset_show),
3736 #endif
3737 #ifdef CONFIG_CGROUPS
3738 ONE("cgroup", S_IRUGO, proc_cgroup_show),
3739 #endif
3740 #ifdef CONFIG_PROC_CPU_RESCTRL
3741 ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show),
3742 #endif
3743 ONE("oom_score", S_IRUGO, proc_oom_score),
3744 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3745 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3746 #ifdef CONFIG_AUDIT
3747 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3748 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3749 #endif
3750 #ifdef CONFIG_FAULT_INJECTION
3751 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3752 REG("fail-nth", 0644, proc_fail_nth_operations),
3753 #endif
3754 #ifdef CONFIG_TASK_IO_ACCOUNTING
3755 ONE("io", S_IRUSR, proc_tid_io_accounting),
3756 #endif
3757 #ifdef CONFIG_USER_NS
3758 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3759 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3760 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3761 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
3762 #endif
3763 #ifdef CONFIG_LIVEPATCH
3764 ONE("patch_state", S_IRUSR, proc_pid_patch_state),
3765 #endif
3766 #ifdef CONFIG_PROC_PID_ARCH_STATUS
3767 ONE("arch_status", S_IRUGO, proc_pid_arch_status),
3768 #endif
3769 #ifdef CONFIG_SECCOMP_CACHE_DEBUG
3770 ONE("seccomp_cache", S_IRUSR, proc_pid_seccomp_cache),
3771 #endif
3772 #ifdef CONFIG_KSM
3773 ONE("ksm_merging_pages", S_IRUSR, proc_pid_ksm_merging_pages),
3774 ONE("ksm_stat", S_IRUSR, proc_pid_ksm_stat),
3775 #endif
3776 #ifdef CONFIG_CPU_FREQ_TIMES
3777 ONE("time_in_state", 0444, proc_time_in_state_show),
3778 #endif
3779 };
3780
proc_tid_base_readdir(struct file * file,struct dir_context * ctx)3781 static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
3782 {
3783 return proc_pident_readdir(file, ctx,
3784 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3785 }
3786
proc_tid_base_lookup(struct inode * dir,struct dentry * dentry,unsigned int flags)3787 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3788 {
3789 return proc_pident_lookup(dir, dentry,
3790 tid_base_stuff,
3791 tid_base_stuff + ARRAY_SIZE(tid_base_stuff));
3792 }
3793
3794 static const struct file_operations proc_tid_base_operations = {
3795 .read = generic_read_dir,
3796 .iterate_shared = proc_tid_base_readdir,
3797 .llseek = generic_file_llseek,
3798 };
3799
3800 static const struct inode_operations proc_tid_base_inode_operations = {
3801 .lookup = proc_tid_base_lookup,
3802 .getattr = pid_getattr,
3803 .setattr = proc_setattr,
3804 };
3805
proc_task_instantiate(struct dentry * dentry,struct task_struct * task,const void * ptr)3806 static struct dentry *proc_task_instantiate(struct dentry *dentry,
3807 struct task_struct *task, const void *ptr)
3808 {
3809 struct inode *inode;
3810 inode = proc_pid_make_base_inode(dentry->d_sb, task,
3811 S_IFDIR | S_IRUGO | S_IXUGO);
3812 if (!inode)
3813 return ERR_PTR(-ENOENT);
3814
3815 inode->i_op = &proc_tid_base_inode_operations;
3816 inode->i_fop = &proc_tid_base_operations;
3817 inode->i_flags |= S_IMMUTABLE;
3818
3819 set_nlink(inode, nlink_tid);
3820 pid_update_inode(task, inode);
3821
3822 d_set_d_op(dentry, &pid_dentry_operations);
3823 return d_splice_alias(inode, dentry);
3824 }
3825
proc_task_lookup(struct inode * dir,struct dentry * dentry,unsigned int flags)3826 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3827 {
3828 struct task_struct *task;
3829 struct task_struct *leader = get_proc_task(dir);
3830 unsigned tid;
3831 struct proc_fs_info *fs_info;
3832 struct pid_namespace *ns;
3833 struct dentry *result = ERR_PTR(-ENOENT);
3834
3835 if (!leader)
3836 goto out_no_task;
3837
3838 tid = name_to_int(&dentry->d_name);
3839 if (tid == ~0U)
3840 goto out;
3841
3842 fs_info = proc_sb_info(dentry->d_sb);
3843 ns = fs_info->pid_ns;
3844 rcu_read_lock();
3845 task = find_task_by_pid_ns(tid, ns);
3846 if (task)
3847 get_task_struct(task);
3848 rcu_read_unlock();
3849 if (!task)
3850 goto out;
3851 if (!same_thread_group(leader, task))
3852 goto out_drop_task;
3853
3854 result = proc_task_instantiate(dentry, task, NULL);
3855 out_drop_task:
3856 put_task_struct(task);
3857 out:
3858 put_task_struct(leader);
3859 out_no_task:
3860 return result;
3861 }
3862
3863 /*
3864 * Find the first tid of a thread group to return to user space.
3865 *
3866 * Usually this is just the thread group leader, but if the users
3867 * buffer was too small or there was a seek into the middle of the
3868 * directory we have more work todo.
3869 *
3870 * In the case of a short read we start with find_task_by_pid.
3871 *
3872 * In the case of a seek we start with the leader and walk nr
3873 * threads past it.
3874 */
first_tid(struct pid * pid,int tid,loff_t f_pos,struct pid_namespace * ns)3875 static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3876 struct pid_namespace *ns)
3877 {
3878 struct task_struct *pos, *task;
3879 unsigned long nr = f_pos;
3880
3881 if (nr != f_pos) /* 32bit overflow? */
3882 return NULL;
3883
3884 rcu_read_lock();
3885 task = pid_task(pid, PIDTYPE_PID);
3886 if (!task)
3887 goto fail;
3888
3889 /* Attempt to start with the tid of a thread */
3890 if (tid && nr) {
3891 pos = find_task_by_pid_ns(tid, ns);
3892 if (pos && same_thread_group(pos, task))
3893 goto found;
3894 }
3895
3896 /* If nr exceeds the number of threads there is nothing todo */
3897 if (nr >= get_nr_threads(task))
3898 goto fail;
3899
3900 /* If we haven't found our starting place yet start
3901 * with the leader and walk nr threads forward.
3902 */
3903 for_each_thread(task, pos) {
3904 if (!nr--)
3905 goto found;
3906 }
3907 fail:
3908 pos = NULL;
3909 goto out;
3910 found:
3911 get_task_struct(pos);
3912 out:
3913 rcu_read_unlock();
3914 return pos;
3915 }
3916
3917 /*
3918 * Find the next thread in the thread list.
3919 * Return NULL if there is an error or no next thread.
3920 *
3921 * The reference to the input task_struct is released.
3922 */
next_tid(struct task_struct * start)3923 static struct task_struct *next_tid(struct task_struct *start)
3924 {
3925 struct task_struct *pos = NULL;
3926 rcu_read_lock();
3927 if (pid_alive(start)) {
3928 pos = __next_thread(start);
3929 if (pos)
3930 get_task_struct(pos);
3931 }
3932 rcu_read_unlock();
3933 put_task_struct(start);
3934 return pos;
3935 }
3936
3937 /* for the /proc/TGID/task/ directories */
proc_task_readdir(struct file * file,struct dir_context * ctx)3938 static int proc_task_readdir(struct file *file, struct dir_context *ctx)
3939 {
3940 struct inode *inode = file_inode(file);
3941 struct task_struct *task;
3942 struct pid_namespace *ns;
3943 int tid;
3944
3945 if (proc_inode_is_dead(inode))
3946 return -ENOENT;
3947
3948 if (!dir_emit_dots(file, ctx))
3949 return 0;
3950
3951 /* We cache the tgid value that the last readdir call couldn't
3952 * return and lseek resets it to 0.
3953 */
3954 ns = proc_pid_ns(inode->i_sb);
3955 tid = (int)(intptr_t)file->private_data;
3956 file->private_data = NULL;
3957 for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
3958 task;
3959 task = next_tid(task), ctx->pos++) {
3960 char name[10 + 1];
3961 unsigned int len;
3962
3963 tid = task_pid_nr_ns(task, ns);
3964 if (!tid)
3965 continue; /* The task has just exited. */
3966 len = snprintf(name, sizeof(name), "%u", tid);
3967 if (!proc_fill_cache(file, ctx, name, len,
3968 proc_task_instantiate, task, NULL)) {
3969 /* returning this tgid failed, save it as the first
3970 * pid for the next readir call */
3971 file->private_data = (void *)(intptr_t)tid;
3972 put_task_struct(task);
3973 break;
3974 }
3975 }
3976
3977 return 0;
3978 }
3979
proc_task_getattr(struct mnt_idmap * idmap,const struct path * path,struct kstat * stat,u32 request_mask,unsigned int query_flags)3980 static int proc_task_getattr(struct mnt_idmap *idmap,
3981 const struct path *path, struct kstat *stat,
3982 u32 request_mask, unsigned int query_flags)
3983 {
3984 struct inode *inode = d_inode(path->dentry);
3985 struct task_struct *p = get_proc_task(inode);
3986 generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat);
3987
3988 if (p) {
3989 stat->nlink += get_nr_threads(p);
3990 put_task_struct(p);
3991 }
3992
3993 return 0;
3994 }
3995
3996 /*
3997 * proc_task_readdir() set @file->private_data to a positive integer
3998 * value, so casting that to u64 is safe. generic_llseek_cookie() will
3999 * set @cookie to 0, so casting to an int is safe. The WARN_ON_ONCE() is
4000 * here to catch any unexpected change in behavior either in
4001 * proc_task_readdir() or generic_llseek_cookie().
4002 */
proc_dir_llseek(struct file * file,loff_t offset,int whence)4003 static loff_t proc_dir_llseek(struct file *file, loff_t offset, int whence)
4004 {
4005 u64 cookie = (u64)(intptr_t)file->private_data;
4006 loff_t off;
4007
4008 off = generic_llseek_cookie(file, offset, whence, &cookie);
4009 WARN_ON_ONCE(cookie > INT_MAX);
4010 file->private_data = (void *)(intptr_t)cookie; /* serialized by f_pos_lock */
4011 return off;
4012 }
4013
4014 static const struct inode_operations proc_task_inode_operations = {
4015 .lookup = proc_task_lookup,
4016 .getattr = proc_task_getattr,
4017 .setattr = proc_setattr,
4018 .permission = proc_pid_permission,
4019 };
4020
4021 static const struct file_operations proc_task_operations = {
4022 .read = generic_read_dir,
4023 .iterate_shared = proc_task_readdir,
4024 .llseek = proc_dir_llseek,
4025 };
4026
set_proc_pid_nlink(void)4027 void __init set_proc_pid_nlink(void)
4028 {
4029 nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
4030 nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
4031 }
4032