]>
Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 LT |
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. | |
e070ad49 ML |
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. | |
1da177e4 LT |
49 | */ |
50 | ||
7c0f6ba6 | 51 | #include <linux/uaccess.h> |
1da177e4 | 52 | |
1da177e4 LT |
53 | #include <linux/errno.h> |
54 | #include <linux/time.h> | |
55 | #include <linux/proc_fs.h> | |
56 | #include <linux/stat.h> | |
5995477a | 57 | #include <linux/task_io_accounting_ops.h> |
1da177e4 | 58 | #include <linux/init.h> |
16f7e0fe | 59 | #include <linux/capability.h> |
1da177e4 | 60 | #include <linux/file.h> |
9f3acc31 | 61 | #include <linux/fdtable.h> |
94f8f3b0 | 62 | #include <linux/generic-radix-tree.h> |
1da177e4 LT |
63 | #include <linux/string.h> |
64 | #include <linux/seq_file.h> | |
65 | #include <linux/namei.h> | |
6b3286ed | 66 | #include <linux/mnt_namespace.h> |
1da177e4 | 67 | #include <linux/mm.h> |
a63d83f4 | 68 | #include <linux/swap.h> |
b835996f | 69 | #include <linux/rcupdate.h> |
1da177e4 | 70 | #include <linux/kallsyms.h> |
2ec220e2 | 71 | #include <linux/stacktrace.h> |
d85f50d5 | 72 | #include <linux/resource.h> |
5096add8 | 73 | #include <linux/module.h> |
1da177e4 LT |
74 | #include <linux/mount.h> |
75 | #include <linux/security.h> | |
76 | #include <linux/ptrace.h> | |
0d094efe | 77 | #include <linux/tracehook.h> |
87ebdc00 | 78 | #include <linux/printk.h> |
efb1a57d | 79 | #include <linux/cache.h> |
a424316c | 80 | #include <linux/cgroup.h> |
1da177e4 LT |
81 | #include <linux/cpuset.h> |
82 | #include <linux/audit.h> | |
5addc5dd | 83 | #include <linux/poll.h> |
1651e14e | 84 | #include <linux/nsproxy.h> |
8ac773b4 | 85 | #include <linux/oom.h> |
3cb4a0bb | 86 | #include <linux/elf.h> |
60347f67 | 87 | #include <linux/pid_namespace.h> |
22d917d8 | 88 | #include <linux/user_namespace.h> |
5ad4e53b | 89 | #include <linux/fs_struct.h> |
5a0e3ad6 | 90 | #include <linux/slab.h> |
4eb5aaa3 | 91 | #include <linux/sched/autogroup.h> |
6e84f315 | 92 | #include <linux/sched/mm.h> |
f7ccbae4 | 93 | #include <linux/sched/coredump.h> |
b17b0153 | 94 | #include <linux/sched/debug.h> |
3905f9ad | 95 | #include <linux/sched/stat.h> |
48f6a7a5 | 96 | #include <linux/posix-timers.h> |
04a8682a | 97 | #include <linux/time_namespace.h> |
e79f15a4 | 98 | #include <linux/resctrl.h> |
43d2b113 | 99 | #include <trace/events/oom.h> |
1da177e4 | 100 | #include "internal.h" |
faf60af1 | 101 | #include "fd.h" |
1da177e4 | 102 | |
ac7f1061 AD |
103 | #include "../../lib/kstrtox.h" |
104 | ||
0f2fe20f EB |
105 | /* NOTE: |
106 | * Implementing inode permission operations in /proc is almost | |
107 | * certainly an error. Permission checks need to happen during | |
108 | * each system call not at open time. The reason is that most of | |
109 | * what we wish to check for permissions in /proc varies at runtime. | |
110 | * | |
111 | * The classic example of a problem is opening file descriptors | |
112 | * in /proc for a task before it execs a suid executable. | |
113 | */ | |
114 | ||
efb1a57d AD |
115 | static u8 nlink_tid __ro_after_init; |
116 | static u8 nlink_tgid __ro_after_init; | |
1270dd8d | 117 | |
1da177e4 | 118 | struct pid_entry { |
cedbccab | 119 | const char *name; |
623f594e | 120 | unsigned int len; |
d161a13f | 121 | umode_t mode; |
c5ef1c42 | 122 | const struct inode_operations *iop; |
00977a59 | 123 | const struct file_operations *fop; |
20cdc894 | 124 | union proc_op op; |
1da177e4 LT |
125 | }; |
126 | ||
61a28784 | 127 | #define NOD(NAME, MODE, IOP, FOP, OP) { \ |
20cdc894 | 128 | .name = (NAME), \ |
c5141e6d | 129 | .len = sizeof(NAME) - 1, \ |
20cdc894 EB |
130 | .mode = MODE, \ |
131 | .iop = IOP, \ | |
132 | .fop = FOP, \ | |
133 | .op = OP, \ | |
134 | } | |
135 | ||
631f9c18 AD |
136 | #define DIR(NAME, MODE, iops, fops) \ |
137 | NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} ) | |
138 | #define LNK(NAME, get_link) \ | |
61a28784 | 139 | NOD(NAME, (S_IFLNK|S_IRWXUGO), \ |
20cdc894 | 140 | &proc_pid_link_inode_operations, NULL, \ |
631f9c18 AD |
141 | { .proc_get_link = get_link } ) |
142 | #define REG(NAME, MODE, fops) \ | |
143 | NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {}) | |
631f9c18 | 144 | #define ONE(NAME, MODE, show) \ |
6d9c939d | 145 | NOD(NAME, (S_IFREG|(MODE)), \ |
be614086 | 146 | NULL, &proc_single_file_operations, \ |
631f9c18 | 147 | { .proc_show = show } ) |
6d9c939d CS |
148 | #define ATTR(LSM, NAME, MODE) \ |
149 | NOD(NAME, (S_IFREG|(MODE)), \ | |
150 | NULL, &proc_pid_attr_operations, \ | |
151 | { .lsm = LSM }) | |
1da177e4 | 152 | |
aed54175 VN |
153 | /* |
154 | * Count the number of hardlinks for the pid_entry table, excluding the . | |
155 | * and .. links. | |
156 | */ | |
1270dd8d | 157 | static unsigned int __init pid_entry_nlink(const struct pid_entry *entries, |
aed54175 VN |
158 | unsigned int n) |
159 | { | |
160 | unsigned int i; | |
161 | unsigned int count; | |
162 | ||
1270dd8d | 163 | count = 2; |
aed54175 VN |
164 | for (i = 0; i < n; ++i) { |
165 | if (S_ISDIR(entries[i].mode)) | |
166 | ++count; | |
167 | } | |
168 | ||
169 | return count; | |
170 | } | |
171 | ||
f7ad3c6b | 172 | static int get_task_root(struct task_struct *task, struct path *root) |
1da177e4 | 173 | { |
7c2c7d99 HD |
174 | int result = -ENOENT; |
175 | ||
0494f6ec | 176 | task_lock(task); |
f7ad3c6b MS |
177 | if (task->fs) { |
178 | get_fs_root(task->fs, root); | |
7c2c7d99 HD |
179 | result = 0; |
180 | } | |
0494f6ec | 181 | task_unlock(task); |
7c2c7d99 | 182 | return result; |
0494f6ec MS |
183 | } |
184 | ||
7773fbc5 | 185 | static int proc_cwd_link(struct dentry *dentry, struct path *path) |
0494f6ec | 186 | { |
2b0143b5 | 187 | struct task_struct *task = get_proc_task(d_inode(dentry)); |
0494f6ec | 188 | int result = -ENOENT; |
99f89551 EB |
189 | |
190 | if (task) { | |
f7ad3c6b MS |
191 | task_lock(task); |
192 | if (task->fs) { | |
193 | get_fs_pwd(task->fs, path); | |
194 | result = 0; | |
195 | } | |
196 | task_unlock(task); | |
99f89551 EB |
197 | put_task_struct(task); |
198 | } | |
1da177e4 LT |
199 | return result; |
200 | } | |
201 | ||
7773fbc5 | 202 | static int proc_root_link(struct dentry *dentry, struct path *path) |
1da177e4 | 203 | { |
2b0143b5 | 204 | struct task_struct *task = get_proc_task(d_inode(dentry)); |
1da177e4 | 205 | int result = -ENOENT; |
99f89551 EB |
206 | |
207 | if (task) { | |
f7ad3c6b | 208 | result = get_task_root(task, path); |
99f89551 EB |
209 | put_task_struct(task); |
210 | } | |
1da177e4 LT |
211 | return result; |
212 | } | |
213 | ||
d26d0cd9 LT |
214 | /* |
215 | * If the user used setproctitle(), we just get the string from | |
216 | * user space at arg_start, and limit it to a maximum of one page. | |
217 | */ | |
218 | static ssize_t get_mm_proctitle(struct mm_struct *mm, char __user *buf, | |
219 | size_t count, unsigned long pos, | |
220 | unsigned long arg_start) | |
221 | { | |
222 | char *page; | |
223 | int ret, got; | |
224 | ||
225 | if (pos >= PAGE_SIZE) | |
226 | return 0; | |
227 | ||
228 | page = (char *)__get_free_page(GFP_KERNEL); | |
229 | if (!page) | |
230 | return -ENOMEM; | |
231 | ||
232 | ret = 0; | |
233 | got = access_remote_vm(mm, arg_start, page, PAGE_SIZE, FOLL_ANON); | |
234 | if (got > 0) { | |
235 | int len = strnlen(page, got); | |
236 | ||
237 | /* Include the NUL character if it was found */ | |
238 | if (len < got) | |
239 | len++; | |
240 | ||
241 | if (len > pos) { | |
242 | len -= pos; | |
243 | if (len > count) | |
244 | len = count; | |
245 | len -= copy_to_user(buf, page+pos, len); | |
246 | if (!len) | |
247 | len = -EFAULT; | |
248 | ret = len; | |
249 | } | |
250 | } | |
251 | free_page((unsigned long)page); | |
252 | return ret; | |
253 | } | |
254 | ||
e4b4e441 | 255 | static ssize_t get_mm_cmdline(struct mm_struct *mm, char __user *buf, |
5ab82718 | 256 | size_t count, loff_t *ppos) |
1da177e4 | 257 | { |
c2c0bb44 | 258 | unsigned long arg_start, arg_end, env_start, env_end; |
5ab82718 | 259 | unsigned long pos, len; |
d26d0cd9 | 260 | char *page, c; |
c2c0bb44 | 261 | |
c2c0bb44 | 262 | /* Check if process spawned far enough to have cmdline. */ |
e4b4e441 LT |
263 | if (!mm->env_end) |
264 | return 0; | |
c2c0bb44 | 265 | |
88aa7cc6 | 266 | spin_lock(&mm->arg_lock); |
c2c0bb44 AD |
267 | arg_start = mm->arg_start; |
268 | arg_end = mm->arg_end; | |
269 | env_start = mm->env_start; | |
270 | env_end = mm->env_end; | |
88aa7cc6 | 271 | spin_unlock(&mm->arg_lock); |
c2c0bb44 | 272 | |
5ab82718 LT |
273 | if (arg_start >= arg_end) |
274 | return 0; | |
6a6cbe75 | 275 | |
2ca66ff7 | 276 | /* |
d26d0cd9 LT |
277 | * We allow setproctitle() to overwrite the argument |
278 | * strings, and overflow past the original end. But | |
279 | * only when it overflows into the environment area. | |
2ca66ff7 | 280 | */ |
d26d0cd9 | 281 | if (env_start != arg_end || env_end < env_start) |
5ab82718 | 282 | env_start = env_end = arg_end; |
d26d0cd9 | 283 | len = env_end - arg_start; |
f5b65348 | 284 | |
5ab82718 | 285 | /* We're not going to care if "*ppos" has high bits set */ |
d26d0cd9 LT |
286 | pos = *ppos; |
287 | if (pos >= len) | |
5ab82718 | 288 | return 0; |
d26d0cd9 LT |
289 | if (count > len - pos) |
290 | count = len - pos; | |
291 | if (!count) | |
292 | return 0; | |
293 | ||
294 | /* | |
295 | * Magical special case: if the argv[] end byte is not | |
296 | * zero, the user has overwritten it with setproctitle(3). | |
297 | * | |
298 | * Possible future enhancement: do this only once when | |
299 | * pos is 0, and set a flag in the 'struct file'. | |
300 | */ | |
301 | if (access_remote_vm(mm, arg_end-1, &c, 1, FOLL_ANON) == 1 && c) | |
302 | return get_mm_proctitle(mm, buf, count, pos, arg_start); | |
a0a07b87 | 303 | |
d26d0cd9 LT |
304 | /* |
305 | * For the non-setproctitle() case we limit things strictly | |
306 | * to the [arg_start, arg_end[ range. | |
307 | */ | |
308 | pos += arg_start; | |
3d712546 | 309 | if (pos < arg_start || pos >= arg_end) |
5ab82718 | 310 | return 0; |
3d712546 LT |
311 | if (count > arg_end - pos) |
312 | count = arg_end - pos; | |
5ab82718 LT |
313 | |
314 | page = (char *)__get_free_page(GFP_KERNEL); | |
315 | if (!page) | |
316 | return -ENOMEM; | |
317 | ||
318 | len = 0; | |
319 | while (count) { | |
320 | int got; | |
321 | size_t size = min_t(size_t, PAGE_SIZE, count); | |
322 | ||
3d712546 LT |
323 | got = access_remote_vm(mm, pos, page, size, FOLL_ANON); |
324 | if (got <= 0) | |
5ab82718 | 325 | break; |
3d712546 | 326 | got -= copy_to_user(buf, page, got); |
5ab82718 LT |
327 | if (unlikely(!got)) { |
328 | if (!len) | |
329 | len = -EFAULT; | |
330 | break; | |
c2c0bb44 | 331 | } |
5ab82718 LT |
332 | pos += got; |
333 | buf += got; | |
334 | len += got; | |
335 | count -= got; | |
c2c0bb44 AD |
336 | } |
337 | ||
c2c0bb44 | 338 | free_page((unsigned long)page); |
5ab82718 | 339 | return len; |
1da177e4 LT |
340 | } |
341 | ||
e4b4e441 LT |
342 | static ssize_t get_task_cmdline(struct task_struct *tsk, char __user *buf, |
343 | size_t count, loff_t *pos) | |
344 | { | |
345 | struct mm_struct *mm; | |
346 | ssize_t ret; | |
347 | ||
348 | mm = get_task_mm(tsk); | |
349 | if (!mm) | |
350 | return 0; | |
351 | ||
352 | ret = get_mm_cmdline(mm, buf, count, pos); | |
c2c0bb44 | 353 | mmput(mm); |
e4b4e441 LT |
354 | return ret; |
355 | } | |
356 | ||
357 | static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf, | |
358 | size_t count, loff_t *pos) | |
359 | { | |
360 | struct task_struct *tsk; | |
361 | ssize_t ret; | |
362 | ||
363 | BUG_ON(*pos < 0); | |
364 | ||
365 | tsk = get_proc_task(file_inode(file)); | |
366 | if (!tsk) | |
367 | return -ESRCH; | |
368 | ret = get_task_cmdline(tsk, buf, count, pos); | |
369 | put_task_struct(tsk); | |
370 | if (ret > 0) | |
371 | *pos += ret; | |
372 | return ret; | |
1da177e4 LT |
373 | } |
374 | ||
c2c0bb44 AD |
375 | static const struct file_operations proc_pid_cmdline_ops = { |
376 | .read = proc_pid_cmdline_read, | |
377 | .llseek = generic_file_llseek, | |
378 | }; | |
379 | ||
1da177e4 LT |
380 | #ifdef CONFIG_KALLSYMS |
381 | /* | |
382 | * Provides a wchan file via kallsyms in a proper one-value-per-file format. | |
383 | * Returns the resolved symbol. If that fails, simply return the address. | |
384 | */ | |
edfcd606 AD |
385 | static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns, |
386 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 387 | { |
ffb45122 | 388 | unsigned long wchan; |
9281acea | 389 | char symname[KSYM_NAME_LEN]; |
1da177e4 | 390 | |
24b2ec21 AD |
391 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
392 | goto print0; | |
1da177e4 | 393 | |
24b2ec21 AD |
394 | wchan = get_wchan(task); |
395 | if (wchan && !lookup_symbol_name(wchan, symname)) { | |
21dae0ad | 396 | seq_puts(m, symname); |
24b2ec21 AD |
397 | return 0; |
398 | } | |
25ce3191 | 399 | |
24b2ec21 AD |
400 | print0: |
401 | seq_putc(m, '0'); | |
25ce3191 | 402 | return 0; |
1da177e4 LT |
403 | } |
404 | #endif /* CONFIG_KALLSYMS */ | |
405 | ||
a9712bc1 AV |
406 | static int lock_trace(struct task_struct *task) |
407 | { | |
2db9dbf7 | 408 | int err = mutex_lock_killable(&task->signal->exec_update_mutex); |
a9712bc1 AV |
409 | if (err) |
410 | return err; | |
caaee623 | 411 | if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) { |
2db9dbf7 | 412 | mutex_unlock(&task->signal->exec_update_mutex); |
a9712bc1 AV |
413 | return -EPERM; |
414 | } | |
415 | return 0; | |
416 | } | |
417 | ||
418 | static void unlock_trace(struct task_struct *task) | |
419 | { | |
2db9dbf7 | 420 | mutex_unlock(&task->signal->exec_update_mutex); |
a9712bc1 AV |
421 | } |
422 | ||
2ec220e2 KC |
423 | #ifdef CONFIG_STACKTRACE |
424 | ||
425 | #define MAX_STACK_TRACE_DEPTH 64 | |
426 | ||
427 | static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns, | |
428 | struct pid *pid, struct task_struct *task) | |
429 | { | |
2ec220e2 | 430 | unsigned long *entries; |
a9712bc1 | 431 | int err; |
2ec220e2 | 432 | |
f8a00cef JH |
433 | /* |
434 | * The ability to racily run the kernel stack unwinder on a running task | |
435 | * and then observe the unwinder output is scary; while it is useful for | |
436 | * debugging kernel issues, it can also allow an attacker to leak kernel | |
437 | * stack contents. | |
438 | * Doing this in a manner that is at least safe from races would require | |
439 | * some work to ensure that the remote task can not be scheduled; and | |
440 | * even then, this would still expose the unwinder as local attack | |
441 | * surface. | |
442 | * Therefore, this interface is restricted to root. | |
443 | */ | |
444 | if (!file_ns_capable(m->file, &init_user_ns, CAP_SYS_ADMIN)) | |
445 | return -EACCES; | |
446 | ||
6da2ec56 KC |
447 | entries = kmalloc_array(MAX_STACK_TRACE_DEPTH, sizeof(*entries), |
448 | GFP_KERNEL); | |
2ec220e2 KC |
449 | if (!entries) |
450 | return -ENOMEM; | |
451 | ||
a9712bc1 AV |
452 | err = lock_trace(task); |
453 | if (!err) { | |
e988e5ec | 454 | unsigned int i, nr_entries; |
5d008fb4 | 455 | |
e988e5ec TG |
456 | nr_entries = stack_trace_save_tsk(task, entries, |
457 | MAX_STACK_TRACE_DEPTH, 0); | |
a9712bc1 | 458 | |
e988e5ec | 459 | for (i = 0; i < nr_entries; i++) { |
8f5abe84 | 460 | seq_printf(m, "[<0>] %pB\n", (void *)entries[i]); |
a9712bc1 | 461 | } |
e988e5ec | 462 | |
a9712bc1 | 463 | unlock_trace(task); |
2ec220e2 KC |
464 | } |
465 | kfree(entries); | |
466 | ||
a9712bc1 | 467 | return err; |
2ec220e2 KC |
468 | } |
469 | #endif | |
470 | ||
5968cece | 471 | #ifdef CONFIG_SCHED_INFO |
1da177e4 LT |
472 | /* |
473 | * Provides /proc/PID/schedstat | |
474 | */ | |
f6e826ca AD |
475 | static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns, |
476 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 477 | { |
5968cece | 478 | if (unlikely(!sched_info_on())) |
08b55775 | 479 | seq_puts(m, "0 0 0\n"); |
5968cece NR |
480 | else |
481 | seq_printf(m, "%llu %llu %lu\n", | |
25ce3191 JP |
482 | (unsigned long long)task->se.sum_exec_runtime, |
483 | (unsigned long long)task->sched_info.run_delay, | |
484 | task->sched_info.pcount); | |
485 | ||
486 | return 0; | |
1da177e4 LT |
487 | } |
488 | #endif | |
489 | ||
9745512c AV |
490 | #ifdef CONFIG_LATENCYTOP |
491 | static int lstats_show_proc(struct seq_file *m, void *v) | |
492 | { | |
493 | int i; | |
13d77c37 HS |
494 | struct inode *inode = m->private; |
495 | struct task_struct *task = get_proc_task(inode); | |
9745512c | 496 | |
13d77c37 HS |
497 | if (!task) |
498 | return -ESRCH; | |
499 | seq_puts(m, "Latency Top version : v0.1\n"); | |
f6d2f584 | 500 | for (i = 0; i < LT_SAVECOUNT; i++) { |
34e49d4f JP |
501 | struct latency_record *lr = &task->latency_record[i]; |
502 | if (lr->backtrace[0]) { | |
9745512c | 503 | int q; |
34e49d4f JP |
504 | seq_printf(m, "%i %li %li", |
505 | lr->count, lr->time, lr->max); | |
9745512c | 506 | for (q = 0; q < LT_BACKTRACEDEPTH; q++) { |
34e49d4f | 507 | unsigned long bt = lr->backtrace[q]; |
accddc41 | 508 | |
34e49d4f | 509 | if (!bt) |
9745512c | 510 | break; |
34e49d4f | 511 | seq_printf(m, " %ps", (void *)bt); |
9745512c | 512 | } |
9d6de12f | 513 | seq_putc(m, '\n'); |
9745512c AV |
514 | } |
515 | ||
516 | } | |
13d77c37 | 517 | put_task_struct(task); |
9745512c AV |
518 | return 0; |
519 | } | |
520 | ||
521 | static int lstats_open(struct inode *inode, struct file *file) | |
522 | { | |
13d77c37 | 523 | return single_open(file, lstats_show_proc, inode); |
d6643d12 HS |
524 | } |
525 | ||
9745512c AV |
526 | static ssize_t lstats_write(struct file *file, const char __user *buf, |
527 | size_t count, loff_t *offs) | |
528 | { | |
496ad9aa | 529 | struct task_struct *task = get_proc_task(file_inode(file)); |
9745512c | 530 | |
13d77c37 HS |
531 | if (!task) |
532 | return -ESRCH; | |
e02c9b0d | 533 | clear_tsk_latency_tracing(task); |
13d77c37 | 534 | put_task_struct(task); |
9745512c AV |
535 | |
536 | return count; | |
537 | } | |
538 | ||
539 | static const struct file_operations proc_lstats_operations = { | |
540 | .open = lstats_open, | |
541 | .read = seq_read, | |
542 | .write = lstats_write, | |
543 | .llseek = seq_lseek, | |
13d77c37 | 544 | .release = single_release, |
9745512c AV |
545 | }; |
546 | ||
547 | #endif | |
548 | ||
6ba51e37 AD |
549 | static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns, |
550 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 551 | { |
ca79b0c2 | 552 | unsigned long totalpages = totalram_pages() + total_swap_pages; |
b95c35e7 | 553 | unsigned long points = 0; |
1da177e4 | 554 | |
ac311a14 | 555 | points = oom_badness(task, totalpages) * 1000 / totalpages; |
25ce3191 JP |
556 | seq_printf(m, "%lu\n", points); |
557 | ||
558 | return 0; | |
1da177e4 LT |
559 | } |
560 | ||
d85f50d5 | 561 | struct limit_names { |
cedbccab AD |
562 | const char *name; |
563 | const char *unit; | |
d85f50d5 NH |
564 | }; |
565 | ||
566 | static const struct limit_names lnames[RLIM_NLIMITS] = { | |
cff4edb5 | 567 | [RLIMIT_CPU] = {"Max cpu time", "seconds"}, |
d85f50d5 NH |
568 | [RLIMIT_FSIZE] = {"Max file size", "bytes"}, |
569 | [RLIMIT_DATA] = {"Max data size", "bytes"}, | |
570 | [RLIMIT_STACK] = {"Max stack size", "bytes"}, | |
571 | [RLIMIT_CORE] = {"Max core file size", "bytes"}, | |
572 | [RLIMIT_RSS] = {"Max resident set", "bytes"}, | |
573 | [RLIMIT_NPROC] = {"Max processes", "processes"}, | |
574 | [RLIMIT_NOFILE] = {"Max open files", "files"}, | |
575 | [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"}, | |
576 | [RLIMIT_AS] = {"Max address space", "bytes"}, | |
577 | [RLIMIT_LOCKS] = {"Max file locks", "locks"}, | |
578 | [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"}, | |
579 | [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"}, | |
580 | [RLIMIT_NICE] = {"Max nice priority", NULL}, | |
581 | [RLIMIT_RTPRIO] = {"Max realtime priority", NULL}, | |
8808117c | 582 | [RLIMIT_RTTIME] = {"Max realtime timeout", "us"}, |
d85f50d5 NH |
583 | }; |
584 | ||
585 | /* Display limits for a process */ | |
1c963eb1 AD |
586 | static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns, |
587 | struct pid *pid, struct task_struct *task) | |
d85f50d5 NH |
588 | { |
589 | unsigned int i; | |
d85f50d5 | 590 | unsigned long flags; |
d85f50d5 NH |
591 | |
592 | struct rlimit rlim[RLIM_NLIMITS]; | |
593 | ||
a6bebbc8 | 594 | if (!lock_task_sighand(task, &flags)) |
d85f50d5 | 595 | return 0; |
d85f50d5 NH |
596 | memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS); |
597 | unlock_task_sighand(task, &flags); | |
d85f50d5 NH |
598 | |
599 | /* | |
600 | * print the file header | |
601 | */ | |
afe922c2 AD |
602 | seq_puts(m, "Limit " |
603 | "Soft Limit " | |
604 | "Hard Limit " | |
605 | "Units \n"); | |
d85f50d5 NH |
606 | |
607 | for (i = 0; i < RLIM_NLIMITS; i++) { | |
608 | if (rlim[i].rlim_cur == RLIM_INFINITY) | |
1c963eb1 | 609 | seq_printf(m, "%-25s %-20s ", |
25ce3191 | 610 | lnames[i].name, "unlimited"); |
d85f50d5 | 611 | else |
1c963eb1 | 612 | seq_printf(m, "%-25s %-20lu ", |
25ce3191 | 613 | lnames[i].name, rlim[i].rlim_cur); |
d85f50d5 NH |
614 | |
615 | if (rlim[i].rlim_max == RLIM_INFINITY) | |
1c963eb1 | 616 | seq_printf(m, "%-20s ", "unlimited"); |
d85f50d5 | 617 | else |
1c963eb1 | 618 | seq_printf(m, "%-20lu ", rlim[i].rlim_max); |
d85f50d5 NH |
619 | |
620 | if (lnames[i].unit) | |
1c963eb1 | 621 | seq_printf(m, "%-10s\n", lnames[i].unit); |
d85f50d5 | 622 | else |
1c963eb1 | 623 | seq_putc(m, '\n'); |
d85f50d5 NH |
624 | } |
625 | ||
1c963eb1 | 626 | return 0; |
d85f50d5 NH |
627 | } |
628 | ||
ebcb6734 | 629 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 AD |
630 | static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns, |
631 | struct pid *pid, struct task_struct *task) | |
ebcb6734 | 632 | { |
631b7aba SRRH |
633 | struct syscall_info info; |
634 | u64 *args = &info.data.args[0]; | |
25ce3191 JP |
635 | int res; |
636 | ||
637 | res = lock_trace(task); | |
a9712bc1 AV |
638 | if (res) |
639 | return res; | |
ebcb6734 | 640 | |
631b7aba | 641 | if (task_current_syscall(task, &info)) |
09d93bd6 | 642 | seq_puts(m, "running\n"); |
631b7aba SRRH |
643 | else if (info.data.nr < 0) |
644 | seq_printf(m, "%d 0x%llx 0x%llx\n", | |
645 | info.data.nr, info.sp, info.data.instruction_pointer); | |
a9712bc1 | 646 | else |
09d93bd6 | 647 | seq_printf(m, |
631b7aba SRRH |
648 | "%d 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx\n", |
649 | info.data.nr, | |
ebcb6734 | 650 | args[0], args[1], args[2], args[3], args[4], args[5], |
631b7aba | 651 | info.sp, info.data.instruction_pointer); |
a9712bc1 | 652 | unlock_trace(task); |
25ce3191 JP |
653 | |
654 | return 0; | |
ebcb6734 RM |
655 | } |
656 | #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */ | |
657 | ||
1da177e4 LT |
658 | /************************************************************************/ |
659 | /* Here the fs part begins */ | |
660 | /************************************************************************/ | |
661 | ||
662 | /* permission checks */ | |
778c1144 | 663 | static int proc_fd_access_allowed(struct inode *inode) |
1da177e4 | 664 | { |
778c1144 EB |
665 | struct task_struct *task; |
666 | int allowed = 0; | |
df26c40e EB |
667 | /* Allow access to a task's file descriptors if it is us or we |
668 | * may use ptrace attach to the process and find out that | |
669 | * information. | |
778c1144 EB |
670 | */ |
671 | task = get_proc_task(inode); | |
df26c40e | 672 | if (task) { |
caaee623 | 673 | allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS); |
778c1144 | 674 | put_task_struct(task); |
df26c40e | 675 | } |
778c1144 | 676 | return allowed; |
1da177e4 LT |
677 | } |
678 | ||
6b4e306a | 679 | int proc_setattr(struct dentry *dentry, struct iattr *attr) |
6d76fa58 LT |
680 | { |
681 | int error; | |
2b0143b5 | 682 | struct inode *inode = d_inode(dentry); |
6d76fa58 LT |
683 | |
684 | if (attr->ia_valid & ATTR_MODE) | |
685 | return -EPERM; | |
686 | ||
31051c85 | 687 | error = setattr_prepare(dentry, attr); |
1025774c CH |
688 | if (error) |
689 | return error; | |
690 | ||
1025774c CH |
691 | setattr_copy(inode, attr); |
692 | mark_inode_dirty(inode); | |
693 | return 0; | |
6d76fa58 LT |
694 | } |
695 | ||
0499680a VK |
696 | /* |
697 | * May current process learn task's sched/cmdline info (for hide_pid_min=1) | |
698 | * or euid/egid (for hide_pid_min=2)? | |
699 | */ | |
700 | static bool has_pid_permissions(struct pid_namespace *pid, | |
701 | struct task_struct *task, | |
702 | int hide_pid_min) | |
703 | { | |
704 | if (pid->hide_pid < hide_pid_min) | |
705 | return true; | |
706 | if (in_group_p(pid->pid_gid)) | |
707 | return true; | |
caaee623 | 708 | return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS); |
0499680a VK |
709 | } |
710 | ||
711 | ||
712 | static int proc_pid_permission(struct inode *inode, int mask) | |
713 | { | |
76f668be | 714 | struct pid_namespace *pid = proc_pid_ns(inode); |
0499680a VK |
715 | struct task_struct *task; |
716 | bool has_perms; | |
717 | ||
718 | task = get_proc_task(inode); | |
a2ef990a XF |
719 | if (!task) |
720 | return -ESRCH; | |
796f571b | 721 | has_perms = has_pid_permissions(pid, task, HIDEPID_NO_ACCESS); |
0499680a VK |
722 | put_task_struct(task); |
723 | ||
724 | if (!has_perms) { | |
796f571b | 725 | if (pid->hide_pid == HIDEPID_INVISIBLE) { |
0499680a VK |
726 | /* |
727 | * Let's make getdents(), stat(), and open() | |
728 | * consistent with each other. If a process | |
729 | * may not stat() a file, it shouldn't be seen | |
730 | * in procfs at all. | |
731 | */ | |
732 | return -ENOENT; | |
733 | } | |
734 | ||
735 | return -EPERM; | |
736 | } | |
737 | return generic_permission(inode, mask); | |
738 | } | |
739 | ||
740 | ||
741 | ||
c5ef1c42 | 742 | static const struct inode_operations proc_def_inode_operations = { |
6d76fa58 LT |
743 | .setattr = proc_setattr, |
744 | }; | |
745 | ||
be614086 EB |
746 | static int proc_single_show(struct seq_file *m, void *v) |
747 | { | |
748 | struct inode *inode = m->private; | |
76f668be CH |
749 | struct pid_namespace *ns = proc_pid_ns(inode); |
750 | struct pid *pid = proc_pid(inode); | |
be614086 EB |
751 | struct task_struct *task; |
752 | int ret; | |
753 | ||
be614086 EB |
754 | task = get_pid_task(pid, PIDTYPE_PID); |
755 | if (!task) | |
756 | return -ESRCH; | |
757 | ||
758 | ret = PROC_I(inode)->op.proc_show(m, ns, pid, task); | |
759 | ||
760 | put_task_struct(task); | |
761 | return ret; | |
762 | } | |
763 | ||
764 | static int proc_single_open(struct inode *inode, struct file *filp) | |
765 | { | |
c6a34058 | 766 | return single_open(filp, proc_single_show, inode); |
be614086 EB |
767 | } |
768 | ||
769 | static const struct file_operations proc_single_file_operations = { | |
770 | .open = proc_single_open, | |
771 | .read = seq_read, | |
772 | .llseek = seq_lseek, | |
773 | .release = single_release, | |
774 | }; | |
775 | ||
5381e169 ON |
776 | |
777 | struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode) | |
1da177e4 | 778 | { |
5381e169 ON |
779 | struct task_struct *task = get_proc_task(inode); |
780 | struct mm_struct *mm = ERR_PTR(-ESRCH); | |
e268337d | 781 | |
5381e169 | 782 | if (task) { |
caaee623 | 783 | mm = mm_access(task, mode | PTRACE_MODE_FSCREDS); |
5381e169 | 784 | put_task_struct(task); |
e268337d | 785 | |
5381e169 ON |
786 | if (!IS_ERR_OR_NULL(mm)) { |
787 | /* ensure this mm_struct can't be freed */ | |
f1f10076 | 788 | mmgrab(mm); |
5381e169 ON |
789 | /* but do not pin its memory */ |
790 | mmput(mm); | |
791 | } | |
792 | } | |
793 | ||
794 | return mm; | |
795 | } | |
796 | ||
797 | static int __mem_open(struct inode *inode, struct file *file, unsigned int mode) | |
798 | { | |
799 | struct mm_struct *mm = proc_mem_open(inode, mode); | |
e268337d LT |
800 | |
801 | if (IS_ERR(mm)) | |
802 | return PTR_ERR(mm); | |
803 | ||
e268337d | 804 | file->private_data = mm; |
1da177e4 LT |
805 | return 0; |
806 | } | |
807 | ||
b409e578 CW |
808 | static int mem_open(struct inode *inode, struct file *file) |
809 | { | |
bc452b4b DH |
810 | int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH); |
811 | ||
812 | /* OK to pass negative loff_t, we can catch out-of-range */ | |
813 | file->f_mode |= FMODE_UNSIGNED_OFFSET; | |
814 | ||
815 | return ret; | |
b409e578 CW |
816 | } |
817 | ||
572d34b9 ON |
818 | static ssize_t mem_rw(struct file *file, char __user *buf, |
819 | size_t count, loff_t *ppos, int write) | |
1da177e4 | 820 | { |
e268337d | 821 | struct mm_struct *mm = file->private_data; |
572d34b9 ON |
822 | unsigned long addr = *ppos; |
823 | ssize_t copied; | |
1da177e4 | 824 | char *page; |
272ddc8b | 825 | unsigned int flags; |
1da177e4 | 826 | |
e268337d LT |
827 | if (!mm) |
828 | return 0; | |
99f89551 | 829 | |
0ee931c4 | 830 | page = (char *)__get_free_page(GFP_KERNEL); |
30cd8903 | 831 | if (!page) |
e268337d | 832 | return -ENOMEM; |
1da177e4 | 833 | |
f7ca54f4 | 834 | copied = 0; |
388f7934 | 835 | if (!mmget_not_zero(mm)) |
6d08f2c7 ON |
836 | goto free; |
837 | ||
f511c0b1 | 838 | flags = FOLL_FORCE | (write ? FOLL_WRITE : 0); |
6347e8d5 | 839 | |
1da177e4 | 840 | while (count > 0) { |
572d34b9 | 841 | int this_len = min_t(int, count, PAGE_SIZE); |
1da177e4 | 842 | |
572d34b9 | 843 | if (write && copy_from_user(page, buf, this_len)) { |
1da177e4 LT |
844 | copied = -EFAULT; |
845 | break; | |
846 | } | |
572d34b9 | 847 | |
6347e8d5 | 848 | this_len = access_remote_vm(mm, addr, page, this_len, flags); |
572d34b9 | 849 | if (!this_len) { |
1da177e4 LT |
850 | if (!copied) |
851 | copied = -EIO; | |
852 | break; | |
853 | } | |
572d34b9 ON |
854 | |
855 | if (!write && copy_to_user(buf, page, this_len)) { | |
856 | copied = -EFAULT; | |
857 | break; | |
858 | } | |
859 | ||
860 | buf += this_len; | |
861 | addr += this_len; | |
862 | copied += this_len; | |
863 | count -= this_len; | |
1da177e4 | 864 | } |
572d34b9 | 865 | *ppos = addr; |
30cd8903 | 866 | |
6d08f2c7 ON |
867 | mmput(mm); |
868 | free: | |
30cd8903 | 869 | free_page((unsigned long) page); |
1da177e4 LT |
870 | return copied; |
871 | } | |
1da177e4 | 872 | |
572d34b9 ON |
873 | static ssize_t mem_read(struct file *file, char __user *buf, |
874 | size_t count, loff_t *ppos) | |
875 | { | |
876 | return mem_rw(file, buf, count, ppos, 0); | |
877 | } | |
878 | ||
879 | static ssize_t mem_write(struct file *file, const char __user *buf, | |
880 | size_t count, loff_t *ppos) | |
881 | { | |
882 | return mem_rw(file, (char __user*)buf, count, ppos, 1); | |
883 | } | |
884 | ||
85863e47 | 885 | loff_t mem_lseek(struct file *file, loff_t offset, int orig) |
1da177e4 LT |
886 | { |
887 | switch (orig) { | |
888 | case 0: | |
889 | file->f_pos = offset; | |
890 | break; | |
891 | case 1: | |
892 | file->f_pos += offset; | |
893 | break; | |
894 | default: | |
895 | return -EINVAL; | |
896 | } | |
897 | force_successful_syscall_return(); | |
898 | return file->f_pos; | |
899 | } | |
900 | ||
e268337d LT |
901 | static int mem_release(struct inode *inode, struct file *file) |
902 | { | |
903 | struct mm_struct *mm = file->private_data; | |
71879d3c | 904 | if (mm) |
6d08f2c7 | 905 | mmdrop(mm); |
e268337d LT |
906 | return 0; |
907 | } | |
908 | ||
00977a59 | 909 | static const struct file_operations proc_mem_operations = { |
1da177e4 LT |
910 | .llseek = mem_lseek, |
911 | .read = mem_read, | |
912 | .write = mem_write, | |
913 | .open = mem_open, | |
e268337d | 914 | .release = mem_release, |
1da177e4 LT |
915 | }; |
916 | ||
b409e578 CW |
917 | static int environ_open(struct inode *inode, struct file *file) |
918 | { | |
919 | return __mem_open(inode, file, PTRACE_MODE_READ); | |
920 | } | |
921 | ||
315e28c8 JP |
922 | static ssize_t environ_read(struct file *file, char __user *buf, |
923 | size_t count, loff_t *ppos) | |
924 | { | |
315e28c8 JP |
925 | char *page; |
926 | unsigned long src = *ppos; | |
b409e578 CW |
927 | int ret = 0; |
928 | struct mm_struct *mm = file->private_data; | |
a3b609ef | 929 | unsigned long env_start, env_end; |
315e28c8 | 930 | |
8148a73c MK |
931 | /* Ensure the process spawned far enough to have an environment. */ |
932 | if (!mm || !mm->env_end) | |
b409e578 | 933 | return 0; |
315e28c8 | 934 | |
0ee931c4 | 935 | page = (char *)__get_free_page(GFP_KERNEL); |
315e28c8 | 936 | if (!page) |
b409e578 | 937 | return -ENOMEM; |
315e28c8 | 938 | |
d6f64b89 | 939 | ret = 0; |
388f7934 | 940 | if (!mmget_not_zero(mm)) |
b409e578 | 941 | goto free; |
a3b609ef | 942 | |
88aa7cc6 | 943 | spin_lock(&mm->arg_lock); |
a3b609ef MG |
944 | env_start = mm->env_start; |
945 | env_end = mm->env_end; | |
88aa7cc6 | 946 | spin_unlock(&mm->arg_lock); |
a3b609ef | 947 | |
315e28c8 | 948 | while (count > 0) { |
e8905ec2 DH |
949 | size_t this_len, max_len; |
950 | int retval; | |
315e28c8 | 951 | |
a3b609ef | 952 | if (src >= (env_end - env_start)) |
315e28c8 JP |
953 | break; |
954 | ||
a3b609ef | 955 | this_len = env_end - (env_start + src); |
e8905ec2 DH |
956 | |
957 | max_len = min_t(size_t, PAGE_SIZE, count); | |
958 | this_len = min(max_len, this_len); | |
315e28c8 | 959 | |
7f7ccc2c | 960 | retval = access_remote_vm(mm, (env_start + src), page, this_len, FOLL_ANON); |
315e28c8 JP |
961 | |
962 | if (retval <= 0) { | |
963 | ret = retval; | |
964 | break; | |
965 | } | |
966 | ||
967 | if (copy_to_user(buf, page, retval)) { | |
968 | ret = -EFAULT; | |
969 | break; | |
970 | } | |
971 | ||
972 | ret += retval; | |
973 | src += retval; | |
974 | buf += retval; | |
975 | count -= retval; | |
976 | } | |
977 | *ppos = src; | |
315e28c8 | 978 | mmput(mm); |
b409e578 CW |
979 | |
980 | free: | |
315e28c8 | 981 | free_page((unsigned long) page); |
315e28c8 JP |
982 | return ret; |
983 | } | |
984 | ||
985 | static const struct file_operations proc_environ_operations = { | |
b409e578 | 986 | .open = environ_open, |
315e28c8 | 987 | .read = environ_read, |
87df8424 | 988 | .llseek = generic_file_llseek, |
b409e578 | 989 | .release = mem_release, |
315e28c8 JP |
990 | }; |
991 | ||
c5317167 AV |
992 | static int auxv_open(struct inode *inode, struct file *file) |
993 | { | |
994 | return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS); | |
995 | } | |
996 | ||
997 | static ssize_t auxv_read(struct file *file, char __user *buf, | |
998 | size_t count, loff_t *ppos) | |
999 | { | |
1000 | struct mm_struct *mm = file->private_data; | |
1001 | unsigned int nwords = 0; | |
06b2849d LY |
1002 | |
1003 | if (!mm) | |
1004 | return 0; | |
c5317167 AV |
1005 | do { |
1006 | nwords += 2; | |
1007 | } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */ | |
1008 | return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv, | |
1009 | nwords * sizeof(mm->saved_auxv[0])); | |
1010 | } | |
1011 | ||
1012 | static const struct file_operations proc_auxv_operations = { | |
1013 | .open = auxv_open, | |
1014 | .read = auxv_read, | |
1015 | .llseek = generic_file_llseek, | |
1016 | .release = mem_release, | |
1017 | }; | |
1018 | ||
fa0cbbf1 DR |
1019 | static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count, |
1020 | loff_t *ppos) | |
1021 | { | |
496ad9aa | 1022 | struct task_struct *task = get_proc_task(file_inode(file)); |
fa0cbbf1 DR |
1023 | char buffer[PROC_NUMBUF]; |
1024 | int oom_adj = OOM_ADJUST_MIN; | |
1025 | size_t len; | |
fa0cbbf1 DR |
1026 | |
1027 | if (!task) | |
1028 | return -ESRCH; | |
f913da59 MH |
1029 | if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX) |
1030 | oom_adj = OOM_ADJUST_MAX; | |
1031 | else | |
1032 | oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) / | |
1033 | OOM_SCORE_ADJ_MAX; | |
fa0cbbf1 DR |
1034 | put_task_struct(task); |
1035 | len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj); | |
1036 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
1037 | } | |
1038 | ||
1d5f0acb MH |
1039 | static int __set_oom_adj(struct file *file, int oom_adj, bool legacy) |
1040 | { | |
1041 | static DEFINE_MUTEX(oom_adj_mutex); | |
44a70ade | 1042 | struct mm_struct *mm = NULL; |
1d5f0acb MH |
1043 | struct task_struct *task; |
1044 | int err = 0; | |
1045 | ||
1046 | task = get_proc_task(file_inode(file)); | |
1047 | if (!task) | |
1048 | return -ESRCH; | |
1049 | ||
1050 | mutex_lock(&oom_adj_mutex); | |
1051 | if (legacy) { | |
1052 | if (oom_adj < task->signal->oom_score_adj && | |
1053 | !capable(CAP_SYS_RESOURCE)) { | |
1054 | err = -EACCES; | |
1055 | goto err_unlock; | |
1056 | } | |
1057 | /* | |
1058 | * /proc/pid/oom_adj is provided for legacy purposes, ask users to use | |
1059 | * /proc/pid/oom_score_adj instead. | |
1060 | */ | |
1061 | pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n", | |
1062 | current->comm, task_pid_nr(current), task_pid_nr(task), | |
1063 | task_pid_nr(task)); | |
1064 | } else { | |
1065 | if ((short)oom_adj < task->signal->oom_score_adj_min && | |
1066 | !capable(CAP_SYS_RESOURCE)) { | |
1067 | err = -EACCES; | |
1068 | goto err_unlock; | |
1069 | } | |
1070 | } | |
1071 | ||
44a70ade MH |
1072 | /* |
1073 | * Make sure we will check other processes sharing the mm if this is | |
1074 | * not vfrok which wants its own oom_score_adj. | |
1075 | * pin the mm so it doesn't go away and get reused after task_unlock | |
1076 | */ | |
1077 | if (!task->vfork_done) { | |
1078 | struct task_struct *p = find_lock_task_mm(task); | |
1079 | ||
1080 | if (p) { | |
1081 | if (atomic_read(&p->mm->mm_users) > 1) { | |
1082 | mm = p->mm; | |
f1f10076 | 1083 | mmgrab(mm); |
44a70ade MH |
1084 | } |
1085 | task_unlock(p); | |
1086 | } | |
1087 | } | |
1088 | ||
1d5f0acb MH |
1089 | task->signal->oom_score_adj = oom_adj; |
1090 | if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE)) | |
1091 | task->signal->oom_score_adj_min = (short)oom_adj; | |
1092 | trace_oom_score_adj_update(task); | |
44a70ade MH |
1093 | |
1094 | if (mm) { | |
1095 | struct task_struct *p; | |
1096 | ||
1097 | rcu_read_lock(); | |
1098 | for_each_process(p) { | |
1099 | if (same_thread_group(task, p)) | |
1100 | continue; | |
1101 | ||
1102 | /* do not touch kernel threads or the global init */ | |
1103 | if (p->flags & PF_KTHREAD || is_global_init(p)) | |
1104 | continue; | |
1105 | ||
1106 | task_lock(p); | |
1107 | if (!p->vfork_done && process_shares_mm(p, mm)) { | |
44a70ade MH |
1108 | p->signal->oom_score_adj = oom_adj; |
1109 | if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE)) | |
1110 | p->signal->oom_score_adj_min = (short)oom_adj; | |
1111 | } | |
1112 | task_unlock(p); | |
1113 | } | |
1114 | rcu_read_unlock(); | |
1115 | mmdrop(mm); | |
1116 | } | |
1d5f0acb MH |
1117 | err_unlock: |
1118 | mutex_unlock(&oom_adj_mutex); | |
1119 | put_task_struct(task); | |
1120 | return err; | |
1121 | } | |
f913da59 | 1122 | |
b72bdfa7 DR |
1123 | /* |
1124 | * /proc/pid/oom_adj exists solely for backwards compatibility with previous | |
1125 | * kernels. The effective policy is defined by oom_score_adj, which has a | |
1126 | * different scale: oom_adj grew exponentially and oom_score_adj grows linearly. | |
1127 | * Values written to oom_adj are simply mapped linearly to oom_score_adj. | |
1128 | * Processes that become oom disabled via oom_adj will still be oom disabled | |
1129 | * with this implementation. | |
1130 | * | |
1131 | * oom_adj cannot be removed since existing userspace binaries use it. | |
1132 | */ | |
fa0cbbf1 DR |
1133 | static ssize_t oom_adj_write(struct file *file, const char __user *buf, |
1134 | size_t count, loff_t *ppos) | |
1135 | { | |
fa0cbbf1 DR |
1136 | char buffer[PROC_NUMBUF]; |
1137 | int oom_adj; | |
fa0cbbf1 DR |
1138 | int err; |
1139 | ||
1140 | memset(buffer, 0, sizeof(buffer)); | |
1141 | if (count > sizeof(buffer) - 1) | |
1142 | count = sizeof(buffer) - 1; | |
1143 | if (copy_from_user(buffer, buf, count)) { | |
1144 | err = -EFAULT; | |
1145 | goto out; | |
1146 | } | |
1147 | ||
1148 | err = kstrtoint(strstrip(buffer), 0, &oom_adj); | |
1149 | if (err) | |
1150 | goto out; | |
1151 | if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) && | |
1152 | oom_adj != OOM_DISABLE) { | |
1153 | err = -EINVAL; | |
1154 | goto out; | |
1155 | } | |
1156 | ||
fa0cbbf1 DR |
1157 | /* |
1158 | * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum | |
1159 | * value is always attainable. | |
1160 | */ | |
1161 | if (oom_adj == OOM_ADJUST_MAX) | |
1162 | oom_adj = OOM_SCORE_ADJ_MAX; | |
1163 | else | |
1164 | oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE; | |
1165 | ||
1d5f0acb | 1166 | err = __set_oom_adj(file, oom_adj, true); |
fa0cbbf1 DR |
1167 | out: |
1168 | return err < 0 ? err : count; | |
1169 | } | |
1170 | ||
1171 | static const struct file_operations proc_oom_adj_operations = { | |
1172 | .read = oom_adj_read, | |
1173 | .write = oom_adj_write, | |
1174 | .llseek = generic_file_llseek, | |
1175 | }; | |
1176 | ||
a63d83f4 DR |
1177 | static ssize_t oom_score_adj_read(struct file *file, char __user *buf, |
1178 | size_t count, loff_t *ppos) | |
1179 | { | |
496ad9aa | 1180 | struct task_struct *task = get_proc_task(file_inode(file)); |
a63d83f4 | 1181 | char buffer[PROC_NUMBUF]; |
a9c58b90 | 1182 | short oom_score_adj = OOM_SCORE_ADJ_MIN; |
a63d83f4 DR |
1183 | size_t len; |
1184 | ||
1185 | if (!task) | |
1186 | return -ESRCH; | |
f913da59 | 1187 | oom_score_adj = task->signal->oom_score_adj; |
a63d83f4 | 1188 | put_task_struct(task); |
a9c58b90 | 1189 | len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj); |
a63d83f4 DR |
1190 | return simple_read_from_buffer(buf, count, ppos, buffer, len); |
1191 | } | |
1192 | ||
1193 | static ssize_t oom_score_adj_write(struct file *file, const char __user *buf, | |
1194 | size_t count, loff_t *ppos) | |
1195 | { | |
a63d83f4 | 1196 | char buffer[PROC_NUMBUF]; |
0a8cb8e3 | 1197 | int oom_score_adj; |
a63d83f4 DR |
1198 | int err; |
1199 | ||
1200 | memset(buffer, 0, sizeof(buffer)); | |
1201 | if (count > sizeof(buffer) - 1) | |
1202 | count = sizeof(buffer) - 1; | |
723548bf DR |
1203 | if (copy_from_user(buffer, buf, count)) { |
1204 | err = -EFAULT; | |
1205 | goto out; | |
1206 | } | |
a63d83f4 | 1207 | |
0a8cb8e3 | 1208 | err = kstrtoint(strstrip(buffer), 0, &oom_score_adj); |
a63d83f4 | 1209 | if (err) |
723548bf | 1210 | goto out; |
a63d83f4 | 1211 | if (oom_score_adj < OOM_SCORE_ADJ_MIN || |
723548bf DR |
1212 | oom_score_adj > OOM_SCORE_ADJ_MAX) { |
1213 | err = -EINVAL; | |
1214 | goto out; | |
1215 | } | |
a63d83f4 | 1216 | |
1d5f0acb | 1217 | err = __set_oom_adj(file, oom_score_adj, false); |
723548bf DR |
1218 | out: |
1219 | return err < 0 ? err : count; | |
a63d83f4 DR |
1220 | } |
1221 | ||
1222 | static const struct file_operations proc_oom_score_adj_operations = { | |
1223 | .read = oom_score_adj_read, | |
1224 | .write = oom_score_adj_write, | |
6038f373 | 1225 | .llseek = default_llseek, |
a63d83f4 DR |
1226 | }; |
1227 | ||
4b7d248b | 1228 | #ifdef CONFIG_AUDIT |
b4eb4f7f | 1229 | #define TMPBUFLEN 11 |
1da177e4 LT |
1230 | static ssize_t proc_loginuid_read(struct file * file, char __user * buf, |
1231 | size_t count, loff_t *ppos) | |
1232 | { | |
496ad9aa | 1233 | struct inode * inode = file_inode(file); |
99f89551 | 1234 | struct task_struct *task = get_proc_task(inode); |
1da177e4 LT |
1235 | ssize_t length; |
1236 | char tmpbuf[TMPBUFLEN]; | |
1237 | ||
99f89551 EB |
1238 | if (!task) |
1239 | return -ESRCH; | |
1da177e4 | 1240 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", |
e1760bd5 EB |
1241 | from_kuid(file->f_cred->user_ns, |
1242 | audit_get_loginuid(task))); | |
99f89551 | 1243 | put_task_struct(task); |
1da177e4 LT |
1244 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); |
1245 | } | |
1246 | ||
1247 | static ssize_t proc_loginuid_write(struct file * file, const char __user * buf, | |
1248 | size_t count, loff_t *ppos) | |
1249 | { | |
496ad9aa | 1250 | struct inode * inode = file_inode(file); |
1da177e4 | 1251 | uid_t loginuid; |
e1760bd5 | 1252 | kuid_t kloginuid; |
774636e1 | 1253 | int rv; |
1da177e4 | 1254 | |
7dc52157 PM |
1255 | rcu_read_lock(); |
1256 | if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) { | |
1257 | rcu_read_unlock(); | |
1da177e4 | 1258 | return -EPERM; |
7dc52157 PM |
1259 | } |
1260 | rcu_read_unlock(); | |
1da177e4 | 1261 | |
1da177e4 LT |
1262 | if (*ppos != 0) { |
1263 | /* No partial writes. */ | |
1264 | return -EINVAL; | |
1265 | } | |
1da177e4 | 1266 | |
774636e1 AD |
1267 | rv = kstrtou32_from_user(buf, count, 10, &loginuid); |
1268 | if (rv < 0) | |
1269 | return rv; | |
81407c84 EP |
1270 | |
1271 | /* is userspace tring to explicitly UNSET the loginuid? */ | |
1272 | if (loginuid == AUDIT_UID_UNSET) { | |
1273 | kloginuid = INVALID_UID; | |
1274 | } else { | |
1275 | kloginuid = make_kuid(file->f_cred->user_ns, loginuid); | |
774636e1 AD |
1276 | if (!uid_valid(kloginuid)) |
1277 | return -EINVAL; | |
e1760bd5 EB |
1278 | } |
1279 | ||
774636e1 AD |
1280 | rv = audit_set_loginuid(kloginuid); |
1281 | if (rv < 0) | |
1282 | return rv; | |
1283 | return count; | |
1da177e4 LT |
1284 | } |
1285 | ||
00977a59 | 1286 | static const struct file_operations proc_loginuid_operations = { |
1da177e4 LT |
1287 | .read = proc_loginuid_read, |
1288 | .write = proc_loginuid_write, | |
87df8424 | 1289 | .llseek = generic_file_llseek, |
1da177e4 | 1290 | }; |
1e0bd755 EP |
1291 | |
1292 | static ssize_t proc_sessionid_read(struct file * file, char __user * buf, | |
1293 | size_t count, loff_t *ppos) | |
1294 | { | |
496ad9aa | 1295 | struct inode * inode = file_inode(file); |
1e0bd755 EP |
1296 | struct task_struct *task = get_proc_task(inode); |
1297 | ssize_t length; | |
1298 | char tmpbuf[TMPBUFLEN]; | |
1299 | ||
1300 | if (!task) | |
1301 | return -ESRCH; | |
1302 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", | |
1303 | audit_get_sessionid(task)); | |
1304 | put_task_struct(task); | |
1305 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); | |
1306 | } | |
1307 | ||
1308 | static const struct file_operations proc_sessionid_operations = { | |
1309 | .read = proc_sessionid_read, | |
87df8424 | 1310 | .llseek = generic_file_llseek, |
1e0bd755 | 1311 | }; |
1da177e4 LT |
1312 | #endif |
1313 | ||
f4f154fd AM |
1314 | #ifdef CONFIG_FAULT_INJECTION |
1315 | static ssize_t proc_fault_inject_read(struct file * file, char __user * buf, | |
1316 | size_t count, loff_t *ppos) | |
1317 | { | |
496ad9aa | 1318 | struct task_struct *task = get_proc_task(file_inode(file)); |
f4f154fd AM |
1319 | char buffer[PROC_NUMBUF]; |
1320 | size_t len; | |
1321 | int make_it_fail; | |
f4f154fd AM |
1322 | |
1323 | if (!task) | |
1324 | return -ESRCH; | |
1325 | make_it_fail = task->make_it_fail; | |
1326 | put_task_struct(task); | |
1327 | ||
1328 | len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail); | |
0c28f287 AM |
1329 | |
1330 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
f4f154fd AM |
1331 | } |
1332 | ||
1333 | static ssize_t proc_fault_inject_write(struct file * file, | |
1334 | const char __user * buf, size_t count, loff_t *ppos) | |
1335 | { | |
1336 | struct task_struct *task; | |
774636e1 | 1337 | char buffer[PROC_NUMBUF]; |
f4f154fd | 1338 | int make_it_fail; |
774636e1 | 1339 | int rv; |
f4f154fd AM |
1340 | |
1341 | if (!capable(CAP_SYS_RESOURCE)) | |
1342 | return -EPERM; | |
1343 | memset(buffer, 0, sizeof(buffer)); | |
1344 | if (count > sizeof(buffer) - 1) | |
1345 | count = sizeof(buffer) - 1; | |
1346 | if (copy_from_user(buffer, buf, count)) | |
1347 | return -EFAULT; | |
774636e1 AD |
1348 | rv = kstrtoint(strstrip(buffer), 0, &make_it_fail); |
1349 | if (rv < 0) | |
1350 | return rv; | |
16caed31 DJ |
1351 | if (make_it_fail < 0 || make_it_fail > 1) |
1352 | return -EINVAL; | |
1353 | ||
496ad9aa | 1354 | task = get_proc_task(file_inode(file)); |
f4f154fd AM |
1355 | if (!task) |
1356 | return -ESRCH; | |
1357 | task->make_it_fail = make_it_fail; | |
1358 | put_task_struct(task); | |
cba8aafe VL |
1359 | |
1360 | return count; | |
f4f154fd AM |
1361 | } |
1362 | ||
00977a59 | 1363 | static const struct file_operations proc_fault_inject_operations = { |
f4f154fd AM |
1364 | .read = proc_fault_inject_read, |
1365 | .write = proc_fault_inject_write, | |
87df8424 | 1366 | .llseek = generic_file_llseek, |
f4f154fd | 1367 | }; |
e41d5818 DV |
1368 | |
1369 | static ssize_t proc_fail_nth_write(struct file *file, const char __user *buf, | |
1370 | size_t count, loff_t *ppos) | |
1371 | { | |
1372 | struct task_struct *task; | |
9049f2f6 AM |
1373 | int err; |
1374 | unsigned int n; | |
e41d5818 | 1375 | |
1203c8e6 AM |
1376 | err = kstrtouint_from_user(buf, count, 0, &n); |
1377 | if (err) | |
1378 | return err; | |
1379 | ||
e41d5818 DV |
1380 | task = get_proc_task(file_inode(file)); |
1381 | if (!task) | |
1382 | return -ESRCH; | |
9f7118b2 | 1383 | task->fail_nth = n; |
e41d5818 | 1384 | put_task_struct(task); |
1203c8e6 | 1385 | |
e41d5818 DV |
1386 | return count; |
1387 | } | |
1388 | ||
1389 | static ssize_t proc_fail_nth_read(struct file *file, char __user *buf, | |
1390 | size_t count, loff_t *ppos) | |
1391 | { | |
1392 | struct task_struct *task; | |
bfc74093 AM |
1393 | char numbuf[PROC_NUMBUF]; |
1394 | ssize_t len; | |
e41d5818 DV |
1395 | |
1396 | task = get_proc_task(file_inode(file)); | |
1397 | if (!task) | |
1398 | return -ESRCH; | |
9f7118b2 | 1399 | len = snprintf(numbuf, sizeof(numbuf), "%u\n", task->fail_nth); |
1203c8e6 | 1400 | put_task_struct(task); |
a44937fe | 1401 | return simple_read_from_buffer(buf, count, ppos, numbuf, len); |
e41d5818 DV |
1402 | } |
1403 | ||
1404 | static const struct file_operations proc_fail_nth_operations = { | |
1405 | .read = proc_fail_nth_read, | |
1406 | .write = proc_fail_nth_write, | |
1407 | }; | |
f4f154fd AM |
1408 | #endif |
1409 | ||
9745512c | 1410 | |
43ae34cb IM |
1411 | #ifdef CONFIG_SCHED_DEBUG |
1412 | /* | |
1413 | * Print out various scheduling related per-task fields: | |
1414 | */ | |
1415 | static int sched_show(struct seq_file *m, void *v) | |
1416 | { | |
1417 | struct inode *inode = m->private; | |
76f668be | 1418 | struct pid_namespace *ns = proc_pid_ns(inode); |
43ae34cb IM |
1419 | struct task_struct *p; |
1420 | ||
43ae34cb IM |
1421 | p = get_proc_task(inode); |
1422 | if (!p) | |
1423 | return -ESRCH; | |
74dc3384 | 1424 | proc_sched_show_task(p, ns, m); |
43ae34cb IM |
1425 | |
1426 | put_task_struct(p); | |
1427 | ||
1428 | return 0; | |
1429 | } | |
1430 | ||
1431 | static ssize_t | |
1432 | sched_write(struct file *file, const char __user *buf, | |
1433 | size_t count, loff_t *offset) | |
1434 | { | |
496ad9aa | 1435 | struct inode *inode = file_inode(file); |
43ae34cb IM |
1436 | struct task_struct *p; |
1437 | ||
43ae34cb IM |
1438 | p = get_proc_task(inode); |
1439 | if (!p) | |
1440 | return -ESRCH; | |
1441 | proc_sched_set_task(p); | |
1442 | ||
1443 | put_task_struct(p); | |
1444 | ||
1445 | return count; | |
1446 | } | |
1447 | ||
1448 | static int sched_open(struct inode *inode, struct file *filp) | |
1449 | { | |
c6a34058 | 1450 | return single_open(filp, sched_show, inode); |
43ae34cb IM |
1451 | } |
1452 | ||
1453 | static const struct file_operations proc_pid_sched_operations = { | |
1454 | .open = sched_open, | |
1455 | .read = seq_read, | |
1456 | .write = sched_write, | |
1457 | .llseek = seq_lseek, | |
5ea473a1 | 1458 | .release = single_release, |
43ae34cb IM |
1459 | }; |
1460 | ||
1461 | #endif | |
1462 | ||
5091faa4 MG |
1463 | #ifdef CONFIG_SCHED_AUTOGROUP |
1464 | /* | |
1465 | * Print out autogroup related information: | |
1466 | */ | |
1467 | static int sched_autogroup_show(struct seq_file *m, void *v) | |
1468 | { | |
1469 | struct inode *inode = m->private; | |
1470 | struct task_struct *p; | |
1471 | ||
1472 | p = get_proc_task(inode); | |
1473 | if (!p) | |
1474 | return -ESRCH; | |
1475 | proc_sched_autogroup_show_task(p, m); | |
1476 | ||
1477 | put_task_struct(p); | |
1478 | ||
1479 | return 0; | |
1480 | } | |
1481 | ||
1482 | static ssize_t | |
1483 | sched_autogroup_write(struct file *file, const char __user *buf, | |
1484 | size_t count, loff_t *offset) | |
1485 | { | |
496ad9aa | 1486 | struct inode *inode = file_inode(file); |
5091faa4 MG |
1487 | struct task_struct *p; |
1488 | char buffer[PROC_NUMBUF]; | |
0a8cb8e3 | 1489 | int nice; |
5091faa4 MG |
1490 | int err; |
1491 | ||
1492 | memset(buffer, 0, sizeof(buffer)); | |
1493 | if (count > sizeof(buffer) - 1) | |
1494 | count = sizeof(buffer) - 1; | |
1495 | if (copy_from_user(buffer, buf, count)) | |
1496 | return -EFAULT; | |
1497 | ||
0a8cb8e3 AD |
1498 | err = kstrtoint(strstrip(buffer), 0, &nice); |
1499 | if (err < 0) | |
1500 | return err; | |
5091faa4 MG |
1501 | |
1502 | p = get_proc_task(inode); | |
1503 | if (!p) | |
1504 | return -ESRCH; | |
1505 | ||
2e5b5b3a | 1506 | err = proc_sched_autogroup_set_nice(p, nice); |
5091faa4 MG |
1507 | if (err) |
1508 | count = err; | |
1509 | ||
1510 | put_task_struct(p); | |
1511 | ||
1512 | return count; | |
1513 | } | |
1514 | ||
1515 | static int sched_autogroup_open(struct inode *inode, struct file *filp) | |
1516 | { | |
1517 | int ret; | |
1518 | ||
1519 | ret = single_open(filp, sched_autogroup_show, NULL); | |
1520 | if (!ret) { | |
1521 | struct seq_file *m = filp->private_data; | |
1522 | ||
1523 | m->private = inode; | |
1524 | } | |
1525 | return ret; | |
1526 | } | |
1527 | ||
1528 | static const struct file_operations proc_pid_sched_autogroup_operations = { | |
1529 | .open = sched_autogroup_open, | |
1530 | .read = seq_read, | |
1531 | .write = sched_autogroup_write, | |
1532 | .llseek = seq_lseek, | |
1533 | .release = single_release, | |
1534 | }; | |
1535 | ||
1536 | #endif /* CONFIG_SCHED_AUTOGROUP */ | |
1537 | ||
04a8682a AV |
1538 | #ifdef CONFIG_TIME_NS |
1539 | static int timens_offsets_show(struct seq_file *m, void *v) | |
1540 | { | |
1541 | struct task_struct *p; | |
1542 | ||
1543 | p = get_proc_task(file_inode(m->file)); | |
1544 | if (!p) | |
1545 | return -ESRCH; | |
1546 | proc_timens_show_offsets(p, m); | |
1547 | ||
1548 | put_task_struct(p); | |
1549 | ||
1550 | return 0; | |
1551 | } | |
1552 | ||
1553 | static ssize_t timens_offsets_write(struct file *file, const char __user *buf, | |
1554 | size_t count, loff_t *ppos) | |
1555 | { | |
1556 | struct inode *inode = file_inode(file); | |
1557 | struct proc_timens_offset offsets[2]; | |
1558 | char *kbuf = NULL, *pos, *next_line; | |
1559 | struct task_struct *p; | |
1560 | int ret, noffsets; | |
1561 | ||
1562 | /* Only allow < page size writes at the beginning of the file */ | |
1563 | if ((*ppos != 0) || (count >= PAGE_SIZE)) | |
1564 | return -EINVAL; | |
1565 | ||
1566 | /* Slurp in the user data */ | |
1567 | kbuf = memdup_user_nul(buf, count); | |
1568 | if (IS_ERR(kbuf)) | |
1569 | return PTR_ERR(kbuf); | |
1570 | ||
1571 | /* Parse the user data */ | |
1572 | ret = -EINVAL; | |
1573 | noffsets = 0; | |
1574 | for (pos = kbuf; pos; pos = next_line) { | |
1575 | struct proc_timens_offset *off = &offsets[noffsets]; | |
94d440d6 | 1576 | char clock[10]; |
04a8682a AV |
1577 | int err; |
1578 | ||
1579 | /* Find the end of line and ensure we don't look past it */ | |
1580 | next_line = strchr(pos, '\n'); | |
1581 | if (next_line) { | |
1582 | *next_line = '\0'; | |
1583 | next_line++; | |
1584 | if (*next_line == '\0') | |
1585 | next_line = NULL; | |
1586 | } | |
1587 | ||
94d440d6 | 1588 | err = sscanf(pos, "%9s %lld %lu", clock, |
04a8682a AV |
1589 | &off->val.tv_sec, &off->val.tv_nsec); |
1590 | if (err != 3 || off->val.tv_nsec >= NSEC_PER_SEC) | |
1591 | goto out; | |
94d440d6 AV |
1592 | |
1593 | clock[sizeof(clock) - 1] = 0; | |
1594 | if (strcmp(clock, "monotonic") == 0 || | |
1595 | strcmp(clock, __stringify(CLOCK_MONOTONIC)) == 0) | |
1596 | off->clockid = CLOCK_MONOTONIC; | |
1597 | else if (strcmp(clock, "boottime") == 0 || | |
1598 | strcmp(clock, __stringify(CLOCK_BOOTTIME)) == 0) | |
1599 | off->clockid = CLOCK_BOOTTIME; | |
1600 | else | |
1601 | goto out; | |
1602 | ||
04a8682a AV |
1603 | noffsets++; |
1604 | if (noffsets == ARRAY_SIZE(offsets)) { | |
1605 | if (next_line) | |
1606 | count = next_line - kbuf; | |
1607 | break; | |
1608 | } | |
1609 | } | |
1610 | ||
1611 | ret = -ESRCH; | |
1612 | p = get_proc_task(inode); | |
1613 | if (!p) | |
1614 | goto out; | |
1615 | ret = proc_timens_set_offset(file, p, offsets, noffsets); | |
1616 | put_task_struct(p); | |
1617 | if (ret) | |
1618 | goto out; | |
1619 | ||
1620 | ret = count; | |
1621 | out: | |
1622 | kfree(kbuf); | |
1623 | return ret; | |
1624 | } | |
1625 | ||
1626 | static int timens_offsets_open(struct inode *inode, struct file *filp) | |
1627 | { | |
1628 | return single_open(filp, timens_offsets_show, inode); | |
1629 | } | |
1630 | ||
1631 | static const struct file_operations proc_timens_offsets_operations = { | |
1632 | .open = timens_offsets_open, | |
1633 | .read = seq_read, | |
1634 | .write = timens_offsets_write, | |
1635 | .llseek = seq_lseek, | |
1636 | .release = single_release, | |
1637 | }; | |
1638 | #endif /* CONFIG_TIME_NS */ | |
1639 | ||
4614a696 JS |
1640 | static ssize_t comm_write(struct file *file, const char __user *buf, |
1641 | size_t count, loff_t *offset) | |
1642 | { | |
496ad9aa | 1643 | struct inode *inode = file_inode(file); |
4614a696 JS |
1644 | struct task_struct *p; |
1645 | char buffer[TASK_COMM_LEN]; | |
830e0fc9 | 1646 | const size_t maxlen = sizeof(buffer) - 1; |
4614a696 JS |
1647 | |
1648 | memset(buffer, 0, sizeof(buffer)); | |
830e0fc9 | 1649 | if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count)) |
4614a696 JS |
1650 | return -EFAULT; |
1651 | ||
1652 | p = get_proc_task(inode); | |
1653 | if (!p) | |
1654 | return -ESRCH; | |
1655 | ||
1656 | if (same_thread_group(current, p)) | |
1657 | set_task_comm(p, buffer); | |
1658 | else | |
1659 | count = -EINVAL; | |
1660 | ||
1661 | put_task_struct(p); | |
1662 | ||
1663 | return count; | |
1664 | } | |
1665 | ||
1666 | static int comm_show(struct seq_file *m, void *v) | |
1667 | { | |
1668 | struct inode *inode = m->private; | |
1669 | struct task_struct *p; | |
1670 | ||
1671 | p = get_proc_task(inode); | |
1672 | if (!p) | |
1673 | return -ESRCH; | |
1674 | ||
88b72b31 TH |
1675 | proc_task_name(m, p, false); |
1676 | seq_putc(m, '\n'); | |
4614a696 JS |
1677 | |
1678 | put_task_struct(p); | |
1679 | ||
1680 | return 0; | |
1681 | } | |
1682 | ||
1683 | static int comm_open(struct inode *inode, struct file *filp) | |
1684 | { | |
c6a34058 | 1685 | return single_open(filp, comm_show, inode); |
4614a696 JS |
1686 | } |
1687 | ||
1688 | static const struct file_operations proc_pid_set_comm_operations = { | |
1689 | .open = comm_open, | |
1690 | .read = seq_read, | |
1691 | .write = comm_write, | |
1692 | .llseek = seq_lseek, | |
1693 | .release = single_release, | |
1694 | }; | |
1695 | ||
7773fbc5 | 1696 | static int proc_exe_link(struct dentry *dentry, struct path *exe_path) |
925d1c40 MH |
1697 | { |
1698 | struct task_struct *task; | |
925d1c40 MH |
1699 | struct file *exe_file; |
1700 | ||
2b0143b5 | 1701 | task = get_proc_task(d_inode(dentry)); |
925d1c40 MH |
1702 | if (!task) |
1703 | return -ENOENT; | |
cd81a917 | 1704 | exe_file = get_task_exe_file(task); |
925d1c40 | 1705 | put_task_struct(task); |
925d1c40 MH |
1706 | if (exe_file) { |
1707 | *exe_path = exe_file->f_path; | |
1708 | path_get(&exe_file->f_path); | |
1709 | fput(exe_file); | |
1710 | return 0; | |
1711 | } else | |
1712 | return -ENOENT; | |
1713 | } | |
1714 | ||
6b255391 | 1715 | static const char *proc_pid_get_link(struct dentry *dentry, |
fceef393 AV |
1716 | struct inode *inode, |
1717 | struct delayed_call *done) | |
1da177e4 | 1718 | { |
408ef013 | 1719 | struct path path; |
1da177e4 LT |
1720 | int error = -EACCES; |
1721 | ||
6b255391 AV |
1722 | if (!dentry) |
1723 | return ERR_PTR(-ECHILD); | |
1724 | ||
778c1144 EB |
1725 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1726 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1727 | goto out; |
1da177e4 | 1728 | |
408ef013 CH |
1729 | error = PROC_I(inode)->op.proc_get_link(dentry, &path); |
1730 | if (error) | |
1731 | goto out; | |
1732 | ||
1bc82070 | 1733 | error = nd_jump_link(&path); |
1da177e4 | 1734 | out: |
008b150a | 1735 | return ERR_PTR(error); |
1da177e4 LT |
1736 | } |
1737 | ||
3dcd25f3 | 1738 | static int do_proc_readlink(struct path *path, char __user *buffer, int buflen) |
1da177e4 | 1739 | { |
0ee931c4 | 1740 | char *tmp = (char *)__get_free_page(GFP_KERNEL); |
3dcd25f3 | 1741 | char *pathname; |
1da177e4 LT |
1742 | int len; |
1743 | ||
1744 | if (!tmp) | |
1745 | return -ENOMEM; | |
0c28f287 | 1746 | |
7b2a69ba | 1747 | pathname = d_path(path, tmp, PAGE_SIZE); |
3dcd25f3 JB |
1748 | len = PTR_ERR(pathname); |
1749 | if (IS_ERR(pathname)) | |
1da177e4 | 1750 | goto out; |
3dcd25f3 | 1751 | len = tmp + PAGE_SIZE - 1 - pathname; |
1da177e4 LT |
1752 | |
1753 | if (len > buflen) | |
1754 | len = buflen; | |
3dcd25f3 | 1755 | if (copy_to_user(buffer, pathname, len)) |
1da177e4 LT |
1756 | len = -EFAULT; |
1757 | out: | |
1758 | free_page((unsigned long)tmp); | |
1759 | return len; | |
1760 | } | |
1761 | ||
1762 | static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen) | |
1763 | { | |
1764 | int error = -EACCES; | |
2b0143b5 | 1765 | struct inode *inode = d_inode(dentry); |
3dcd25f3 | 1766 | struct path path; |
1da177e4 | 1767 | |
778c1144 EB |
1768 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1769 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1770 | goto out; |
1da177e4 | 1771 | |
7773fbc5 | 1772 | error = PROC_I(inode)->op.proc_get_link(dentry, &path); |
1da177e4 LT |
1773 | if (error) |
1774 | goto out; | |
1775 | ||
3dcd25f3 JB |
1776 | error = do_proc_readlink(&path, buffer, buflen); |
1777 | path_put(&path); | |
1da177e4 | 1778 | out: |
1da177e4 LT |
1779 | return error; |
1780 | } | |
1781 | ||
faf60af1 | 1782 | const struct inode_operations proc_pid_link_inode_operations = { |
1da177e4 | 1783 | .readlink = proc_pid_readlink, |
6b255391 | 1784 | .get_link = proc_pid_get_link, |
6d76fa58 | 1785 | .setattr = proc_setattr, |
1da177e4 LT |
1786 | }; |
1787 | ||
28a6d671 EB |
1788 | |
1789 | /* building an inode */ | |
1790 | ||
c6eb50d2 | 1791 | void task_dump_owner(struct task_struct *task, umode_t mode, |
68eb94f1 EB |
1792 | kuid_t *ruid, kgid_t *rgid) |
1793 | { | |
1794 | /* Depending on the state of dumpable compute who should own a | |
1795 | * proc file for a task. | |
1796 | */ | |
1797 | const struct cred *cred; | |
1798 | kuid_t uid; | |
1799 | kgid_t gid; | |
1800 | ||
2e0ad552 AD |
1801 | if (unlikely(task->flags & PF_KTHREAD)) { |
1802 | *ruid = GLOBAL_ROOT_UID; | |
1803 | *rgid = GLOBAL_ROOT_GID; | |
1804 | return; | |
1805 | } | |
1806 | ||
68eb94f1 EB |
1807 | /* Default to the tasks effective ownership */ |
1808 | rcu_read_lock(); | |
1809 | cred = __task_cred(task); | |
1810 | uid = cred->euid; | |
1811 | gid = cred->egid; | |
1812 | rcu_read_unlock(); | |
1813 | ||
1814 | /* | |
1815 | * Before the /proc/pid/status file was created the only way to read | |
1816 | * the effective uid of a /process was to stat /proc/pid. Reading | |
1817 | * /proc/pid/status is slow enough that procps and other packages | |
1818 | * kept stating /proc/pid. To keep the rules in /proc simple I have | |
1819 | * made this apply to all per process world readable and executable | |
1820 | * directories. | |
1821 | */ | |
1822 | if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) { | |
1823 | struct mm_struct *mm; | |
1824 | task_lock(task); | |
1825 | mm = task->mm; | |
1826 | /* Make non-dumpable tasks owned by some root */ | |
1827 | if (mm) { | |
1828 | if (get_dumpable(mm) != SUID_DUMP_USER) { | |
1829 | struct user_namespace *user_ns = mm->user_ns; | |
1830 | ||
1831 | uid = make_kuid(user_ns, 0); | |
1832 | if (!uid_valid(uid)) | |
1833 | uid = GLOBAL_ROOT_UID; | |
1834 | ||
1835 | gid = make_kgid(user_ns, 0); | |
1836 | if (!gid_valid(gid)) | |
1837 | gid = GLOBAL_ROOT_GID; | |
1838 | } | |
1839 | } else { | |
1840 | uid = GLOBAL_ROOT_UID; | |
1841 | gid = GLOBAL_ROOT_GID; | |
1842 | } | |
1843 | task_unlock(task); | |
1844 | } | |
1845 | *ruid = uid; | |
1846 | *rgid = gid; | |
1847 | } | |
1848 | ||
7bc3e6e5 EB |
1849 | void proc_pid_evict_inode(struct proc_inode *ei) |
1850 | { | |
1851 | struct pid *pid = ei->pid; | |
1852 | ||
1853 | if (S_ISDIR(ei->vfs_inode.i_mode)) { | |
63f818f4 | 1854 | spin_lock(&pid->lock); |
7bc3e6e5 | 1855 | hlist_del_init_rcu(&ei->sibling_inodes); |
63f818f4 | 1856 | spin_unlock(&pid->lock); |
7bc3e6e5 EB |
1857 | } |
1858 | ||
1859 | put_pid(pid); | |
1860 | } | |
1861 | ||
db978da8 AG |
1862 | struct inode *proc_pid_make_inode(struct super_block * sb, |
1863 | struct task_struct *task, umode_t mode) | |
28a6d671 EB |
1864 | { |
1865 | struct inode * inode; | |
1866 | struct proc_inode *ei; | |
7bc3e6e5 | 1867 | struct pid *pid; |
1da177e4 | 1868 | |
28a6d671 | 1869 | /* We need a new inode */ |
1da177e4 | 1870 | |
28a6d671 EB |
1871 | inode = new_inode(sb); |
1872 | if (!inode) | |
1873 | goto out; | |
1874 | ||
1875 | /* Common stuff */ | |
1876 | ei = PROC_I(inode); | |
db978da8 | 1877 | inode->i_mode = mode; |
85fe4025 | 1878 | inode->i_ino = get_next_ino(); |
078cd827 | 1879 | inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode); |
28a6d671 EB |
1880 | inode->i_op = &proc_def_inode_operations; |
1881 | ||
1882 | /* | |
1883 | * grab the reference to task. | |
1884 | */ | |
7bc3e6e5 EB |
1885 | pid = get_task_pid(task, PIDTYPE_PID); |
1886 | if (!pid) | |
28a6d671 EB |
1887 | goto out_unlock; |
1888 | ||
7bc3e6e5 EB |
1889 | /* Let the pid remember us for quick removal */ |
1890 | ei->pid = pid; | |
1891 | if (S_ISDIR(mode)) { | |
63f818f4 | 1892 | spin_lock(&pid->lock); |
7bc3e6e5 | 1893 | hlist_add_head_rcu(&ei->sibling_inodes, &pid->inodes); |
63f818f4 | 1894 | spin_unlock(&pid->lock); |
7bc3e6e5 EB |
1895 | } |
1896 | ||
68eb94f1 | 1897 | task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid); |
28a6d671 EB |
1898 | security_task_to_inode(task, inode); |
1899 | ||
1da177e4 | 1900 | out: |
28a6d671 EB |
1901 | return inode; |
1902 | ||
1903 | out_unlock: | |
1904 | iput(inode); | |
1905 | return NULL; | |
1da177e4 LT |
1906 | } |
1907 | ||
a528d35e DH |
1908 | int pid_getattr(const struct path *path, struct kstat *stat, |
1909 | u32 request_mask, unsigned int query_flags) | |
1da177e4 | 1910 | { |
a528d35e | 1911 | struct inode *inode = d_inode(path->dentry); |
76f668be | 1912 | struct pid_namespace *pid = proc_pid_ns(inode); |
28a6d671 | 1913 | struct task_struct *task; |
c69e8d9c | 1914 | |
28a6d671 | 1915 | generic_fillattr(inode, stat); |
1da177e4 | 1916 | |
dcb0f222 EB |
1917 | stat->uid = GLOBAL_ROOT_UID; |
1918 | stat->gid = GLOBAL_ROOT_GID; | |
94116929 | 1919 | rcu_read_lock(); |
28a6d671 EB |
1920 | task = pid_task(proc_pid(inode), PIDTYPE_PID); |
1921 | if (task) { | |
796f571b | 1922 | if (!has_pid_permissions(pid, task, HIDEPID_INVISIBLE)) { |
0499680a VK |
1923 | rcu_read_unlock(); |
1924 | /* | |
1925 | * This doesn't prevent learning whether PID exists, | |
1926 | * it only makes getattr() consistent with readdir(). | |
1927 | */ | |
1928 | return -ENOENT; | |
1929 | } | |
68eb94f1 | 1930 | task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid); |
1da177e4 | 1931 | } |
28a6d671 | 1932 | rcu_read_unlock(); |
d6e71144 | 1933 | return 0; |
1da177e4 LT |
1934 | } |
1935 | ||
1da177e4 LT |
1936 | /* dentry stuff */ |
1937 | ||
1938 | /* | |
1bbc5513 AV |
1939 | * Set <pid>/... inode ownership (can change due to setuid(), etc.) |
1940 | */ | |
1941 | void pid_update_inode(struct task_struct *task, struct inode *inode) | |
1942 | { | |
1943 | task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid); | |
1944 | ||
1945 | inode->i_mode &= ~(S_ISUID | S_ISGID); | |
1946 | security_task_to_inode(task, inode); | |
1947 | } | |
1948 | ||
1949 | /* | |
1da177e4 LT |
1950 | * Rewrite the inode's ownerships here because the owning task may have |
1951 | * performed a setuid(), etc. | |
99f89551 | 1952 | * |
1da177e4 | 1953 | */ |
1bbc5513 | 1954 | static int pid_revalidate(struct dentry *dentry, unsigned int flags) |
1da177e4 | 1955 | { |
34286d66 NP |
1956 | struct inode *inode; |
1957 | struct task_struct *task; | |
c69e8d9c | 1958 | |
0b728e19 | 1959 | if (flags & LOOKUP_RCU) |
34286d66 NP |
1960 | return -ECHILD; |
1961 | ||
2b0143b5 | 1962 | inode = d_inode(dentry); |
34286d66 NP |
1963 | task = get_proc_task(inode); |
1964 | ||
99f89551 | 1965 | if (task) { |
1bbc5513 | 1966 | pid_update_inode(task, inode); |
99f89551 | 1967 | put_task_struct(task); |
1da177e4 LT |
1968 | return 1; |
1969 | } | |
1da177e4 LT |
1970 | return 0; |
1971 | } | |
1972 | ||
d855a4b7 ON |
1973 | static inline bool proc_inode_is_dead(struct inode *inode) |
1974 | { | |
1975 | return !proc_pid(inode)->tasks[PIDTYPE_PID].first; | |
1976 | } | |
1977 | ||
1dd704b6 DH |
1978 | int pid_delete_dentry(const struct dentry *dentry) |
1979 | { | |
1980 | /* Is the task we represent dead? | |
1981 | * If so, then don't put the dentry on the lru list, | |
1982 | * kill it immediately. | |
1983 | */ | |
2b0143b5 | 1984 | return proc_inode_is_dead(d_inode(dentry)); |
1dd704b6 DH |
1985 | } |
1986 | ||
6b4e306a | 1987 | const struct dentry_operations pid_dentry_operations = |
28a6d671 EB |
1988 | { |
1989 | .d_revalidate = pid_revalidate, | |
1990 | .d_delete = pid_delete_dentry, | |
1991 | }; | |
1992 | ||
1993 | /* Lookups */ | |
1994 | ||
1c0d04c9 EB |
1995 | /* |
1996 | * Fill a directory entry. | |
1997 | * | |
1998 | * If possible create the dcache entry and derive our inode number and | |
1999 | * file type from dcache entry. | |
2000 | * | |
2001 | * Since all of the proc inode numbers are dynamically generated, the inode | |
2002 | * numbers do not exist until the inode is cache. This means creating the | |
2003 | * the dcache entry in readdir is necessary to keep the inode numbers | |
2004 | * reported by readdir in sync with the inode numbers reported | |
2005 | * by stat. | |
2006 | */ | |
f0c3b509 | 2007 | bool proc_fill_cache(struct file *file, struct dir_context *ctx, |
a4ef3895 | 2008 | const char *name, unsigned int len, |
c5141e6d | 2009 | instantiate_t instantiate, struct task_struct *task, const void *ptr) |
61a28784 | 2010 | { |
f0c3b509 | 2011 | struct dentry *child, *dir = file->f_path.dentry; |
1df98b8b | 2012 | struct qstr qname = QSTR_INIT(name, len); |
61a28784 | 2013 | struct inode *inode; |
0168b9e3 AV |
2014 | unsigned type = DT_UNKNOWN; |
2015 | ino_t ino = 1; | |
61a28784 | 2016 | |
1df98b8b | 2017 | child = d_hash_and_lookup(dir, &qname); |
61a28784 | 2018 | if (!child) { |
3781764b AV |
2019 | DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq); |
2020 | child = d_alloc_parallel(dir, &qname, &wq); | |
2021 | if (IS_ERR(child)) | |
1df98b8b | 2022 | goto end_instantiate; |
3781764b | 2023 | if (d_in_lookup(child)) { |
0168b9e3 AV |
2024 | struct dentry *res; |
2025 | res = instantiate(child, task, ptr); | |
3781764b | 2026 | d_lookup_done(child); |
0168b9e3 AV |
2027 | if (unlikely(res)) { |
2028 | dput(child); | |
2029 | child = res; | |
d85b399b AV |
2030 | if (IS_ERR(child)) |
2031 | goto end_instantiate; | |
3781764b | 2032 | } |
61a28784 EB |
2033 | } |
2034 | } | |
2b0143b5 | 2035 | inode = d_inode(child); |
147ce699 AV |
2036 | ino = inode->i_ino; |
2037 | type = inode->i_mode >> 12; | |
61a28784 | 2038 | dput(child); |
d85b399b | 2039 | end_instantiate: |
f0c3b509 | 2040 | return dir_emit(ctx, name, len, ino, type); |
61a28784 EB |
2041 | } |
2042 | ||
640708a2 PE |
2043 | /* |
2044 | * dname_to_vma_addr - maps a dentry name into two unsigned longs | |
2045 | * which represent vma start and end addresses. | |
2046 | */ | |
2047 | static int dname_to_vma_addr(struct dentry *dentry, | |
2048 | unsigned long *start, unsigned long *end) | |
2049 | { | |
ac7f1061 AD |
2050 | const char *str = dentry->d_name.name; |
2051 | unsigned long long sval, eval; | |
2052 | unsigned int len; | |
2053 | ||
35318db5 AD |
2054 | if (str[0] == '0' && str[1] != '-') |
2055 | return -EINVAL; | |
ac7f1061 AD |
2056 | len = _parse_integer(str, 16, &sval); |
2057 | if (len & KSTRTOX_OVERFLOW) | |
2058 | return -EINVAL; | |
2059 | if (sval != (unsigned long)sval) | |
2060 | return -EINVAL; | |
2061 | str += len; | |
2062 | ||
2063 | if (*str != '-') | |
640708a2 | 2064 | return -EINVAL; |
ac7f1061 AD |
2065 | str++; |
2066 | ||
35318db5 AD |
2067 | if (str[0] == '0' && str[1]) |
2068 | return -EINVAL; | |
ac7f1061 AD |
2069 | len = _parse_integer(str, 16, &eval); |
2070 | if (len & KSTRTOX_OVERFLOW) | |
2071 | return -EINVAL; | |
2072 | if (eval != (unsigned long)eval) | |
2073 | return -EINVAL; | |
2074 | str += len; | |
2075 | ||
2076 | if (*str != '\0') | |
2077 | return -EINVAL; | |
2078 | ||
2079 | *start = sval; | |
2080 | *end = eval; | |
640708a2 PE |
2081 | |
2082 | return 0; | |
2083 | } | |
2084 | ||
0b728e19 | 2085 | static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags) |
640708a2 PE |
2086 | { |
2087 | unsigned long vm_start, vm_end; | |
2088 | bool exact_vma_exists = false; | |
2089 | struct mm_struct *mm = NULL; | |
2090 | struct task_struct *task; | |
640708a2 PE |
2091 | struct inode *inode; |
2092 | int status = 0; | |
2093 | ||
0b728e19 | 2094 | if (flags & LOOKUP_RCU) |
640708a2 PE |
2095 | return -ECHILD; |
2096 | ||
2b0143b5 | 2097 | inode = d_inode(dentry); |
640708a2 PE |
2098 | task = get_proc_task(inode); |
2099 | if (!task) | |
2100 | goto out_notask; | |
2101 | ||
caaee623 | 2102 | mm = mm_access(task, PTRACE_MODE_READ_FSCREDS); |
2344bec7 | 2103 | if (IS_ERR_OR_NULL(mm)) |
640708a2 PE |
2104 | goto out; |
2105 | ||
2106 | if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) { | |
cd9e2bb8 KK |
2107 | status = down_read_killable(&mm->mmap_sem); |
2108 | if (!status) { | |
2109 | exact_vma_exists = !!find_exact_vma(mm, vm_start, | |
2110 | vm_end); | |
2111 | up_read(&mm->mmap_sem); | |
2112 | } | |
640708a2 PE |
2113 | } |
2114 | ||
2115 | mmput(mm); | |
2116 | ||
2117 | if (exact_vma_exists) { | |
68eb94f1 EB |
2118 | task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid); |
2119 | ||
640708a2 PE |
2120 | security_task_to_inode(task, inode); |
2121 | status = 1; | |
2122 | } | |
2123 | ||
2124 | out: | |
2125 | put_task_struct(task); | |
2126 | ||
2127 | out_notask: | |
640708a2 PE |
2128 | return status; |
2129 | } | |
2130 | ||
2131 | static const struct dentry_operations tid_map_files_dentry_operations = { | |
2132 | .d_revalidate = map_files_d_revalidate, | |
2133 | .d_delete = pid_delete_dentry, | |
2134 | }; | |
2135 | ||
6b255391 | 2136 | static int map_files_get_link(struct dentry *dentry, struct path *path) |
640708a2 PE |
2137 | { |
2138 | unsigned long vm_start, vm_end; | |
2139 | struct vm_area_struct *vma; | |
2140 | struct task_struct *task; | |
2141 | struct mm_struct *mm; | |
2142 | int rc; | |
2143 | ||
2144 | rc = -ENOENT; | |
2b0143b5 | 2145 | task = get_proc_task(d_inode(dentry)); |
640708a2 PE |
2146 | if (!task) |
2147 | goto out; | |
2148 | ||
2149 | mm = get_task_mm(task); | |
2150 | put_task_struct(task); | |
2151 | if (!mm) | |
2152 | goto out; | |
2153 | ||
2154 | rc = dname_to_vma_addr(dentry, &vm_start, &vm_end); | |
2155 | if (rc) | |
2156 | goto out_mmput; | |
2157 | ||
cd9e2bb8 KK |
2158 | rc = down_read_killable(&mm->mmap_sem); |
2159 | if (rc) | |
2160 | goto out_mmput; | |
2161 | ||
70335abb | 2162 | rc = -ENOENT; |
640708a2 PE |
2163 | vma = find_exact_vma(mm, vm_start, vm_end); |
2164 | if (vma && vma->vm_file) { | |
2165 | *path = vma->vm_file->f_path; | |
2166 | path_get(path); | |
2167 | rc = 0; | |
2168 | } | |
2169 | up_read(&mm->mmap_sem); | |
2170 | ||
2171 | out_mmput: | |
2172 | mmput(mm); | |
2173 | out: | |
2174 | return rc; | |
2175 | } | |
2176 | ||
2177 | struct map_files_info { | |
20d28cde AD |
2178 | unsigned long start; |
2179 | unsigned long end; | |
7b540d06 | 2180 | fmode_t mode; |
640708a2 PE |
2181 | }; |
2182 | ||
bdb4d100 CO |
2183 | /* |
2184 | * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the | |
2185 | * symlinks may be used to bypass permissions on ancestor directories in the | |
2186 | * path to the file in question. | |
2187 | */ | |
2188 | static const char * | |
6b255391 | 2189 | proc_map_files_get_link(struct dentry *dentry, |
fceef393 AV |
2190 | struct inode *inode, |
2191 | struct delayed_call *done) | |
bdb4d100 CO |
2192 | { |
2193 | if (!capable(CAP_SYS_ADMIN)) | |
2194 | return ERR_PTR(-EPERM); | |
2195 | ||
fceef393 | 2196 | return proc_pid_get_link(dentry, inode, done); |
bdb4d100 CO |
2197 | } |
2198 | ||
2199 | /* | |
6b255391 | 2200 | * Identical to proc_pid_link_inode_operations except for get_link() |
bdb4d100 CO |
2201 | */ |
2202 | static const struct inode_operations proc_map_files_link_inode_operations = { | |
2203 | .readlink = proc_pid_readlink, | |
6b255391 | 2204 | .get_link = proc_map_files_get_link, |
bdb4d100 CO |
2205 | .setattr = proc_setattr, |
2206 | }; | |
2207 | ||
0168b9e3 AV |
2208 | static struct dentry * |
2209 | proc_map_files_instantiate(struct dentry *dentry, | |
640708a2 PE |
2210 | struct task_struct *task, const void *ptr) |
2211 | { | |
7b540d06 | 2212 | fmode_t mode = (fmode_t)(unsigned long)ptr; |
640708a2 PE |
2213 | struct proc_inode *ei; |
2214 | struct inode *inode; | |
2215 | ||
0168b9e3 | 2216 | inode = proc_pid_make_inode(dentry->d_sb, task, S_IFLNK | |
db978da8 AG |
2217 | ((mode & FMODE_READ ) ? S_IRUSR : 0) | |
2218 | ((mode & FMODE_WRITE) ? S_IWUSR : 0)); | |
640708a2 | 2219 | if (!inode) |
0168b9e3 | 2220 | return ERR_PTR(-ENOENT); |
640708a2 PE |
2221 | |
2222 | ei = PROC_I(inode); | |
6b255391 | 2223 | ei->op.proc_get_link = map_files_get_link; |
640708a2 | 2224 | |
bdb4d100 | 2225 | inode->i_op = &proc_map_files_link_inode_operations; |
640708a2 | 2226 | inode->i_size = 64; |
640708a2 PE |
2227 | |
2228 | d_set_d_op(dentry, &tid_map_files_dentry_operations); | |
0168b9e3 | 2229 | return d_splice_alias(inode, dentry); |
640708a2 PE |
2230 | } |
2231 | ||
2232 | static struct dentry *proc_map_files_lookup(struct inode *dir, | |
00cd8dd3 | 2233 | struct dentry *dentry, unsigned int flags) |
640708a2 PE |
2234 | { |
2235 | unsigned long vm_start, vm_end; | |
2236 | struct vm_area_struct *vma; | |
2237 | struct task_struct *task; | |
0168b9e3 | 2238 | struct dentry *result; |
640708a2 PE |
2239 | struct mm_struct *mm; |
2240 | ||
0168b9e3 | 2241 | result = ERR_PTR(-ENOENT); |
640708a2 PE |
2242 | task = get_proc_task(dir); |
2243 | if (!task) | |
2244 | goto out; | |
2245 | ||
0168b9e3 | 2246 | result = ERR_PTR(-EACCES); |
caaee623 | 2247 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
640708a2 PE |
2248 | goto out_put_task; |
2249 | ||
0168b9e3 | 2250 | result = ERR_PTR(-ENOENT); |
640708a2 | 2251 | if (dname_to_vma_addr(dentry, &vm_start, &vm_end)) |
eb94cd96 | 2252 | goto out_put_task; |
640708a2 PE |
2253 | |
2254 | mm = get_task_mm(task); | |
2255 | if (!mm) | |
eb94cd96 | 2256 | goto out_put_task; |
640708a2 | 2257 | |
cd9e2bb8 KK |
2258 | result = ERR_PTR(-EINTR); |
2259 | if (down_read_killable(&mm->mmap_sem)) | |
2260 | goto out_put_mm; | |
2261 | ||
2262 | result = ERR_PTR(-ENOENT); | |
640708a2 PE |
2263 | vma = find_exact_vma(mm, vm_start, vm_end); |
2264 | if (!vma) | |
2265 | goto out_no_vma; | |
2266 | ||
05f56484 | 2267 | if (vma->vm_file) |
0168b9e3 | 2268 | result = proc_map_files_instantiate(dentry, task, |
05f56484 | 2269 | (void *)(unsigned long)vma->vm_file->f_mode); |
640708a2 PE |
2270 | |
2271 | out_no_vma: | |
2272 | up_read(&mm->mmap_sem); | |
cd9e2bb8 | 2273 | out_put_mm: |
640708a2 | 2274 | mmput(mm); |
640708a2 PE |
2275 | out_put_task: |
2276 | put_task_struct(task); | |
2277 | out: | |
0168b9e3 | 2278 | return result; |
640708a2 PE |
2279 | } |
2280 | ||
2281 | static const struct inode_operations proc_map_files_inode_operations = { | |
2282 | .lookup = proc_map_files_lookup, | |
2283 | .permission = proc_fd_permission, | |
2284 | .setattr = proc_setattr, | |
2285 | }; | |
2286 | ||
2287 | static int | |
f0c3b509 | 2288 | proc_map_files_readdir(struct file *file, struct dir_context *ctx) |
640708a2 | 2289 | { |
640708a2 PE |
2290 | struct vm_area_struct *vma; |
2291 | struct task_struct *task; | |
2292 | struct mm_struct *mm; | |
f0c3b509 | 2293 | unsigned long nr_files, pos, i; |
94f8f3b0 | 2294 | GENRADIX(struct map_files_info) fa; |
f0c3b509 | 2295 | struct map_files_info *p; |
640708a2 PE |
2296 | int ret; |
2297 | ||
94f8f3b0 KO |
2298 | genradix_init(&fa); |
2299 | ||
640708a2 | 2300 | ret = -ENOENT; |
f0c3b509 | 2301 | task = get_proc_task(file_inode(file)); |
640708a2 PE |
2302 | if (!task) |
2303 | goto out; | |
2304 | ||
2305 | ret = -EACCES; | |
caaee623 | 2306 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
640708a2 PE |
2307 | goto out_put_task; |
2308 | ||
2309 | ret = 0; | |
f0c3b509 AV |
2310 | if (!dir_emit_dots(file, ctx)) |
2311 | goto out_put_task; | |
640708a2 | 2312 | |
f0c3b509 AV |
2313 | mm = get_task_mm(task); |
2314 | if (!mm) | |
2315 | goto out_put_task; | |
cd9e2bb8 KK |
2316 | |
2317 | ret = down_read_killable(&mm->mmap_sem); | |
2318 | if (ret) { | |
2319 | mmput(mm); | |
2320 | goto out_put_task; | |
2321 | } | |
640708a2 | 2322 | |
f0c3b509 | 2323 | nr_files = 0; |
640708a2 | 2324 | |
f0c3b509 AV |
2325 | /* |
2326 | * We need two passes here: | |
2327 | * | |
2328 | * 1) Collect vmas of mapped files with mmap_sem taken | |
2329 | * 2) Release mmap_sem and instantiate entries | |
2330 | * | |
2331 | * otherwise we get lockdep complained, since filldir() | |
2332 | * routine might require mmap_sem taken in might_fault(). | |
2333 | */ | |
640708a2 | 2334 | |
f0c3b509 | 2335 | for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) { |
94f8f3b0 KO |
2336 | if (!vma->vm_file) |
2337 | continue; | |
2338 | if (++pos <= ctx->pos) | |
2339 | continue; | |
f0c3b509 | 2340 | |
94f8f3b0 KO |
2341 | p = genradix_ptr_alloc(&fa, nr_files++, GFP_KERNEL); |
2342 | if (!p) { | |
f0c3b509 | 2343 | ret = -ENOMEM; |
f0c3b509 AV |
2344 | up_read(&mm->mmap_sem); |
2345 | mmput(mm); | |
2346 | goto out_put_task; | |
640708a2 | 2347 | } |
f0c3b509 | 2348 | |
94f8f3b0 KO |
2349 | p->start = vma->vm_start; |
2350 | p->end = vma->vm_end; | |
2351 | p->mode = vma->vm_file->f_mode; | |
640708a2 | 2352 | } |
f0c3b509 | 2353 | up_read(&mm->mmap_sem); |
fe079a5e | 2354 | mmput(mm); |
f0c3b509 AV |
2355 | |
2356 | for (i = 0; i < nr_files; i++) { | |
20d28cde AD |
2357 | char buf[4 * sizeof(long) + 2]; /* max: %lx-%lx\0 */ |
2358 | unsigned int len; | |
2359 | ||
94f8f3b0 | 2360 | p = genradix_ptr(&fa, i); |
20d28cde | 2361 | len = snprintf(buf, sizeof(buf), "%lx-%lx", p->start, p->end); |
f0c3b509 | 2362 | if (!proc_fill_cache(file, ctx, |
20d28cde | 2363 | buf, len, |
f0c3b509 AV |
2364 | proc_map_files_instantiate, |
2365 | task, | |
2366 | (void *)(unsigned long)p->mode)) | |
2367 | break; | |
2368 | ctx->pos++; | |
640708a2 PE |
2369 | } |
2370 | ||
640708a2 PE |
2371 | out_put_task: |
2372 | put_task_struct(task); | |
2373 | out: | |
94f8f3b0 | 2374 | genradix_free(&fa); |
640708a2 PE |
2375 | return ret; |
2376 | } | |
2377 | ||
2378 | static const struct file_operations proc_map_files_operations = { | |
2379 | .read = generic_read_dir, | |
f50752ea AV |
2380 | .iterate_shared = proc_map_files_readdir, |
2381 | .llseek = generic_file_llseek, | |
640708a2 PE |
2382 | }; |
2383 | ||
b18b6a9c | 2384 | #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS) |
48f6a7a5 PE |
2385 | struct timers_private { |
2386 | struct pid *pid; | |
2387 | struct task_struct *task; | |
2388 | struct sighand_struct *sighand; | |
57b8015e | 2389 | struct pid_namespace *ns; |
48f6a7a5 PE |
2390 | unsigned long flags; |
2391 | }; | |
2392 | ||
2393 | static void *timers_start(struct seq_file *m, loff_t *pos) | |
2394 | { | |
2395 | struct timers_private *tp = m->private; | |
2396 | ||
2397 | tp->task = get_pid_task(tp->pid, PIDTYPE_PID); | |
2398 | if (!tp->task) | |
2399 | return ERR_PTR(-ESRCH); | |
2400 | ||
2401 | tp->sighand = lock_task_sighand(tp->task, &tp->flags); | |
2402 | if (!tp->sighand) | |
2403 | return ERR_PTR(-ESRCH); | |
2404 | ||
2405 | return seq_list_start(&tp->task->signal->posix_timers, *pos); | |
2406 | } | |
2407 | ||
2408 | static void *timers_next(struct seq_file *m, void *v, loff_t *pos) | |
2409 | { | |
2410 | struct timers_private *tp = m->private; | |
2411 | return seq_list_next(v, &tp->task->signal->posix_timers, pos); | |
2412 | } | |
2413 | ||
2414 | static void timers_stop(struct seq_file *m, void *v) | |
2415 | { | |
2416 | struct timers_private *tp = m->private; | |
2417 | ||
2418 | if (tp->sighand) { | |
2419 | unlock_task_sighand(tp->task, &tp->flags); | |
2420 | tp->sighand = NULL; | |
2421 | } | |
2422 | ||
2423 | if (tp->task) { | |
2424 | put_task_struct(tp->task); | |
2425 | tp->task = NULL; | |
2426 | } | |
2427 | } | |
2428 | ||
2429 | static int show_timer(struct seq_file *m, void *v) | |
2430 | { | |
2431 | struct k_itimer *timer; | |
57b8015e PE |
2432 | struct timers_private *tp = m->private; |
2433 | int notify; | |
cedbccab | 2434 | static const char * const nstr[] = { |
57b8015e PE |
2435 | [SIGEV_SIGNAL] = "signal", |
2436 | [SIGEV_NONE] = "none", | |
2437 | [SIGEV_THREAD] = "thread", | |
2438 | }; | |
48f6a7a5 PE |
2439 | |
2440 | timer = list_entry((struct list_head *)v, struct k_itimer, list); | |
57b8015e PE |
2441 | notify = timer->it_sigev_notify; |
2442 | ||
48f6a7a5 | 2443 | seq_printf(m, "ID: %d\n", timer->it_id); |
ba3edf1f | 2444 | seq_printf(m, "signal: %d/%px\n", |
25ce3191 JP |
2445 | timer->sigq->info.si_signo, |
2446 | timer->sigq->info.si_value.sival_ptr); | |
57b8015e | 2447 | seq_printf(m, "notify: %s/%s.%d\n", |
25ce3191 JP |
2448 | nstr[notify & ~SIGEV_THREAD_ID], |
2449 | (notify & SIGEV_THREAD_ID) ? "tid" : "pid", | |
2450 | pid_nr_ns(timer->it_pid, tp->ns)); | |
15ef0298 | 2451 | seq_printf(m, "ClockID: %d\n", timer->it_clock); |
48f6a7a5 PE |
2452 | |
2453 | return 0; | |
2454 | } | |
2455 | ||
2456 | static const struct seq_operations proc_timers_seq_ops = { | |
2457 | .start = timers_start, | |
2458 | .next = timers_next, | |
2459 | .stop = timers_stop, | |
2460 | .show = show_timer, | |
2461 | }; | |
2462 | ||
2463 | static int proc_timers_open(struct inode *inode, struct file *file) | |
2464 | { | |
2465 | struct timers_private *tp; | |
2466 | ||
2467 | tp = __seq_open_private(file, &proc_timers_seq_ops, | |
2468 | sizeof(struct timers_private)); | |
2469 | if (!tp) | |
2470 | return -ENOMEM; | |
2471 | ||
2472 | tp->pid = proc_pid(inode); | |
76f668be | 2473 | tp->ns = proc_pid_ns(inode); |
48f6a7a5 PE |
2474 | return 0; |
2475 | } | |
2476 | ||
2477 | static const struct file_operations proc_timers_operations = { | |
2478 | .open = proc_timers_open, | |
2479 | .read = seq_read, | |
2480 | .llseek = seq_lseek, | |
2481 | .release = seq_release_private, | |
2482 | }; | |
b5946bea | 2483 | #endif |
640708a2 | 2484 | |
5de23d43 JS |
2485 | static ssize_t timerslack_ns_write(struct file *file, const char __user *buf, |
2486 | size_t count, loff_t *offset) | |
2487 | { | |
2488 | struct inode *inode = file_inode(file); | |
2489 | struct task_struct *p; | |
2490 | u64 slack_ns; | |
2491 | int err; | |
2492 | ||
2493 | err = kstrtoull_from_user(buf, count, 10, &slack_ns); | |
2494 | if (err < 0) | |
2495 | return err; | |
2496 | ||
2497 | p = get_proc_task(inode); | |
2498 | if (!p) | |
2499 | return -ESRCH; | |
2500 | ||
4b2bd5fe | 2501 | if (p != current) { |
8da0b4f6 BG |
2502 | rcu_read_lock(); |
2503 | if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) { | |
2504 | rcu_read_unlock(); | |
4b2bd5fe JS |
2505 | count = -EPERM; |
2506 | goto out; | |
2507 | } | |
8da0b4f6 | 2508 | rcu_read_unlock(); |
5de23d43 | 2509 | |
4b2bd5fe JS |
2510 | err = security_task_setscheduler(p); |
2511 | if (err) { | |
2512 | count = err; | |
2513 | goto out; | |
2514 | } | |
904763e1 JS |
2515 | } |
2516 | ||
7abbaf94 JS |
2517 | task_lock(p); |
2518 | if (slack_ns == 0) | |
2519 | p->timer_slack_ns = p->default_timer_slack_ns; | |
2520 | else | |
2521 | p->timer_slack_ns = slack_ns; | |
2522 | task_unlock(p); | |
2523 | ||
2524 | out: | |
5de23d43 JS |
2525 | put_task_struct(p); |
2526 | ||
2527 | return count; | |
2528 | } | |
2529 | ||
2530 | static int timerslack_ns_show(struct seq_file *m, void *v) | |
2531 | { | |
2532 | struct inode *inode = m->private; | |
2533 | struct task_struct *p; | |
7abbaf94 | 2534 | int err = 0; |
5de23d43 JS |
2535 | |
2536 | p = get_proc_task(inode); | |
2537 | if (!p) | |
2538 | return -ESRCH; | |
2539 | ||
4b2bd5fe | 2540 | if (p != current) { |
8da0b4f6 BG |
2541 | rcu_read_lock(); |
2542 | if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) { | |
2543 | rcu_read_unlock(); | |
4b2bd5fe JS |
2544 | err = -EPERM; |
2545 | goto out; | |
2546 | } | |
8da0b4f6 BG |
2547 | rcu_read_unlock(); |
2548 | ||
4b2bd5fe JS |
2549 | err = security_task_getscheduler(p); |
2550 | if (err) | |
2551 | goto out; | |
2552 | } | |
904763e1 | 2553 | |
7abbaf94 JS |
2554 | task_lock(p); |
2555 | seq_printf(m, "%llu\n", p->timer_slack_ns); | |
2556 | task_unlock(p); | |
2557 | ||
2558 | out: | |
5de23d43 JS |
2559 | put_task_struct(p); |
2560 | ||
2561 | return err; | |
2562 | } | |
2563 | ||
2564 | static int timerslack_ns_open(struct inode *inode, struct file *filp) | |
2565 | { | |
2566 | return single_open(filp, timerslack_ns_show, inode); | |
2567 | } | |
2568 | ||
2569 | static const struct file_operations proc_pid_set_timerslack_ns_operations = { | |
2570 | .open = timerslack_ns_open, | |
2571 | .read = seq_read, | |
2572 | .write = timerslack_ns_write, | |
2573 | .llseek = seq_lseek, | |
2574 | .release = single_release, | |
2575 | }; | |
2576 | ||
0168b9e3 AV |
2577 | static struct dentry *proc_pident_instantiate(struct dentry *dentry, |
2578 | struct task_struct *task, const void *ptr) | |
444ceed8 | 2579 | { |
c5141e6d | 2580 | const struct pid_entry *p = ptr; |
444ceed8 EB |
2581 | struct inode *inode; |
2582 | struct proc_inode *ei; | |
444ceed8 | 2583 | |
0168b9e3 | 2584 | inode = proc_pid_make_inode(dentry->d_sb, task, p->mode); |
444ceed8 | 2585 | if (!inode) |
0168b9e3 | 2586 | return ERR_PTR(-ENOENT); |
444ceed8 EB |
2587 | |
2588 | ei = PROC_I(inode); | |
444ceed8 | 2589 | if (S_ISDIR(inode->i_mode)) |
bfe86848 | 2590 | set_nlink(inode, 2); /* Use getattr to fix if necessary */ |
444ceed8 EB |
2591 | if (p->iop) |
2592 | inode->i_op = p->iop; | |
2593 | if (p->fop) | |
2594 | inode->i_fop = p->fop; | |
2595 | ei->op = p->op; | |
1bbc5513 | 2596 | pid_update_inode(task, inode); |
fb045adb | 2597 | d_set_d_op(dentry, &pid_dentry_operations); |
0168b9e3 | 2598 | return d_splice_alias(inode, dentry); |
444ceed8 EB |
2599 | } |
2600 | ||
1da177e4 LT |
2601 | static struct dentry *proc_pident_lookup(struct inode *dir, |
2602 | struct dentry *dentry, | |
d5a572a4 AD |
2603 | const struct pid_entry *p, |
2604 | const struct pid_entry *end) | |
1da177e4 | 2605 | { |
99f89551 | 2606 | struct task_struct *task = get_proc_task(dir); |
0168b9e3 | 2607 | struct dentry *res = ERR_PTR(-ENOENT); |
1da177e4 | 2608 | |
99f89551 EB |
2609 | if (!task) |
2610 | goto out_no_task; | |
1da177e4 | 2611 | |
20cdc894 EB |
2612 | /* |
2613 | * Yes, it does not scale. And it should not. Don't add | |
2614 | * new entries into /proc/<tgid>/ without very good reasons. | |
2615 | */ | |
d5a572a4 | 2616 | for (; p < end; p++) { |
1da177e4 LT |
2617 | if (p->len != dentry->d_name.len) |
2618 | continue; | |
26b95137 AD |
2619 | if (!memcmp(dentry->d_name.name, p->name, p->len)) { |
2620 | res = proc_pident_instantiate(dentry, task, p); | |
1da177e4 | 2621 | break; |
26b95137 | 2622 | } |
1da177e4 | 2623 | } |
99f89551 EB |
2624 | put_task_struct(task); |
2625 | out_no_task: | |
0168b9e3 | 2626 | return res; |
1da177e4 LT |
2627 | } |
2628 | ||
f0c3b509 | 2629 | static int proc_pident_readdir(struct file *file, struct dir_context *ctx, |
c5141e6d | 2630 | const struct pid_entry *ents, unsigned int nents) |
28a6d671 | 2631 | { |
f0c3b509 AV |
2632 | struct task_struct *task = get_proc_task(file_inode(file)); |
2633 | const struct pid_entry *p; | |
28a6d671 | 2634 | |
28a6d671 | 2635 | if (!task) |
f0c3b509 | 2636 | return -ENOENT; |
28a6d671 | 2637 | |
f0c3b509 AV |
2638 | if (!dir_emit_dots(file, ctx)) |
2639 | goto out; | |
2640 | ||
2641 | if (ctx->pos >= nents + 2) | |
2642 | goto out; | |
28a6d671 | 2643 | |
bac5f5d5 | 2644 | for (p = ents + (ctx->pos - 2); p < ents + nents; p++) { |
f0c3b509 AV |
2645 | if (!proc_fill_cache(file, ctx, p->name, p->len, |
2646 | proc_pident_instantiate, task, p)) | |
2647 | break; | |
2648 | ctx->pos++; | |
2649 | } | |
28a6d671 | 2650 | out: |
61a28784 | 2651 | put_task_struct(task); |
f0c3b509 | 2652 | return 0; |
1da177e4 LT |
2653 | } |
2654 | ||
28a6d671 EB |
2655 | #ifdef CONFIG_SECURITY |
2656 | static ssize_t proc_pid_attr_read(struct file * file, char __user * buf, | |
2657 | size_t count, loff_t *ppos) | |
2658 | { | |
496ad9aa | 2659 | struct inode * inode = file_inode(file); |
04ff9708 | 2660 | char *p = NULL; |
28a6d671 EB |
2661 | ssize_t length; |
2662 | struct task_struct *task = get_proc_task(inode); | |
2663 | ||
28a6d671 | 2664 | if (!task) |
04ff9708 | 2665 | return -ESRCH; |
28a6d671 | 2666 | |
6d9c939d | 2667 | length = security_getprocattr(task, PROC_I(inode)->op.lsm, |
2fddfeef | 2668 | (char*)file->f_path.dentry->d_name.name, |
04ff9708 | 2669 | &p); |
28a6d671 | 2670 | put_task_struct(task); |
04ff9708 AV |
2671 | if (length > 0) |
2672 | length = simple_read_from_buffer(buf, count, ppos, p, length); | |
2673 | kfree(p); | |
28a6d671 | 2674 | return length; |
1da177e4 LT |
2675 | } |
2676 | ||
28a6d671 EB |
2677 | static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf, |
2678 | size_t count, loff_t *ppos) | |
2679 | { | |
496ad9aa | 2680 | struct inode * inode = file_inode(file); |
41089b6d | 2681 | struct task_struct *task; |
bb646cdb | 2682 | void *page; |
41089b6d | 2683 | int rv; |
b21507e2 | 2684 | |
41089b6d AD |
2685 | rcu_read_lock(); |
2686 | task = pid_task(proc_pid(inode), PIDTYPE_PID); | |
2687 | if (!task) { | |
2688 | rcu_read_unlock(); | |
2689 | return -ESRCH; | |
2690 | } | |
b21507e2 | 2691 | /* A task may only write its own attributes. */ |
41089b6d AD |
2692 | if (current != task) { |
2693 | rcu_read_unlock(); | |
2694 | return -EACCES; | |
2695 | } | |
35a196be PM |
2696 | /* Prevent changes to overridden credentials. */ |
2697 | if (current_cred() != current_real_cred()) { | |
2698 | rcu_read_unlock(); | |
2699 | return -EBUSY; | |
2700 | } | |
41089b6d | 2701 | rcu_read_unlock(); |
b21507e2 | 2702 | |
28a6d671 EB |
2703 | if (count > PAGE_SIZE) |
2704 | count = PAGE_SIZE; | |
2705 | ||
2706 | /* No partial writes. */ | |
28a6d671 | 2707 | if (*ppos != 0) |
41089b6d | 2708 | return -EINVAL; |
28a6d671 | 2709 | |
bb646cdb AV |
2710 | page = memdup_user(buf, count); |
2711 | if (IS_ERR(page)) { | |
41089b6d | 2712 | rv = PTR_ERR(page); |
28a6d671 | 2713 | goto out; |
bb646cdb | 2714 | } |
28a6d671 | 2715 | |
107db7c7 | 2716 | /* Guard against adverse ptrace interaction */ |
41089b6d AD |
2717 | rv = mutex_lock_interruptible(¤t->signal->cred_guard_mutex); |
2718 | if (rv < 0) | |
107db7c7 DH |
2719 | goto out_free; |
2720 | ||
6d9c939d CS |
2721 | rv = security_setprocattr(PROC_I(inode)->op.lsm, |
2722 | file->f_path.dentry->d_name.name, page, | |
2723 | count); | |
b21507e2 | 2724 | mutex_unlock(¤t->signal->cred_guard_mutex); |
28a6d671 | 2725 | out_free: |
bb646cdb | 2726 | kfree(page); |
28a6d671 | 2727 | out: |
41089b6d | 2728 | return rv; |
28a6d671 EB |
2729 | } |
2730 | ||
00977a59 | 2731 | static const struct file_operations proc_pid_attr_operations = { |
28a6d671 EB |
2732 | .read = proc_pid_attr_read, |
2733 | .write = proc_pid_attr_write, | |
87df8424 | 2734 | .llseek = generic_file_llseek, |
28a6d671 EB |
2735 | }; |
2736 | ||
6d9c939d CS |
2737 | #define LSM_DIR_OPS(LSM) \ |
2738 | static int proc_##LSM##_attr_dir_iterate(struct file *filp, \ | |
2739 | struct dir_context *ctx) \ | |
2740 | { \ | |
2741 | return proc_pident_readdir(filp, ctx, \ | |
2742 | LSM##_attr_dir_stuff, \ | |
2743 | ARRAY_SIZE(LSM##_attr_dir_stuff)); \ | |
2744 | } \ | |
2745 | \ | |
2746 | static const struct file_operations proc_##LSM##_attr_dir_ops = { \ | |
2747 | .read = generic_read_dir, \ | |
2748 | .iterate = proc_##LSM##_attr_dir_iterate, \ | |
2749 | .llseek = default_llseek, \ | |
2750 | }; \ | |
2751 | \ | |
2752 | static struct dentry *proc_##LSM##_attr_dir_lookup(struct inode *dir, \ | |
2753 | struct dentry *dentry, unsigned int flags) \ | |
2754 | { \ | |
2755 | return proc_pident_lookup(dir, dentry, \ | |
2756 | LSM##_attr_dir_stuff, \ | |
d5a572a4 | 2757 | LSM##_attr_dir_stuff + ARRAY_SIZE(LSM##_attr_dir_stuff)); \ |
6d9c939d CS |
2758 | } \ |
2759 | \ | |
2760 | static const struct inode_operations proc_##LSM##_attr_dir_inode_ops = { \ | |
2761 | .lookup = proc_##LSM##_attr_dir_lookup, \ | |
2762 | .getattr = pid_getattr, \ | |
2763 | .setattr = proc_setattr, \ | |
2764 | } | |
2765 | ||
2766 | #ifdef CONFIG_SECURITY_SMACK | |
2767 | static const struct pid_entry smack_attr_dir_stuff[] = { | |
2768 | ATTR("smack", "current", 0666), | |
2769 | }; | |
2770 | LSM_DIR_OPS(smack); | |
2771 | #endif | |
2772 | ||
c5141e6d | 2773 | static const struct pid_entry attr_dir_stuff[] = { |
6d9c939d CS |
2774 | ATTR(NULL, "current", 0666), |
2775 | ATTR(NULL, "prev", 0444), | |
2776 | ATTR(NULL, "exec", 0666), | |
2777 | ATTR(NULL, "fscreate", 0666), | |
2778 | ATTR(NULL, "keycreate", 0666), | |
2779 | ATTR(NULL, "sockcreate", 0666), | |
2780 | #ifdef CONFIG_SECURITY_SMACK | |
2781 | DIR("smack", 0555, | |
2782 | proc_smack_attr_dir_inode_ops, proc_smack_attr_dir_ops), | |
2783 | #endif | |
28a6d671 EB |
2784 | }; |
2785 | ||
f0c3b509 | 2786 | static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx) |
28a6d671 | 2787 | { |
f0c3b509 AV |
2788 | return proc_pident_readdir(file, ctx, |
2789 | attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2790 | } |
2791 | ||
00977a59 | 2792 | static const struct file_operations proc_attr_dir_operations = { |
1da177e4 | 2793 | .read = generic_read_dir, |
f50752ea AV |
2794 | .iterate_shared = proc_attr_dir_readdir, |
2795 | .llseek = generic_file_llseek, | |
1da177e4 LT |
2796 | }; |
2797 | ||
72d9dcfc | 2798 | static struct dentry *proc_attr_dir_lookup(struct inode *dir, |
00cd8dd3 | 2799 | struct dentry *dentry, unsigned int flags) |
28a6d671 | 2800 | { |
7bcd6b0e | 2801 | return proc_pident_lookup(dir, dentry, |
d5a572a4 AD |
2802 | attr_dir_stuff, |
2803 | attr_dir_stuff + ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2804 | } |
2805 | ||
c5ef1c42 | 2806 | static const struct inode_operations proc_attr_dir_inode_operations = { |
72d9dcfc | 2807 | .lookup = proc_attr_dir_lookup, |
99f89551 | 2808 | .getattr = pid_getattr, |
6d76fa58 | 2809 | .setattr = proc_setattr, |
1da177e4 LT |
2810 | }; |
2811 | ||
28a6d671 EB |
2812 | #endif |
2813 | ||
698ba7b5 | 2814 | #ifdef CONFIG_ELF_CORE |
3cb4a0bb KH |
2815 | static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf, |
2816 | size_t count, loff_t *ppos) | |
2817 | { | |
496ad9aa | 2818 | struct task_struct *task = get_proc_task(file_inode(file)); |
3cb4a0bb KH |
2819 | struct mm_struct *mm; |
2820 | char buffer[PROC_NUMBUF]; | |
2821 | size_t len; | |
2822 | int ret; | |
2823 | ||
2824 | if (!task) | |
2825 | return -ESRCH; | |
2826 | ||
2827 | ret = 0; | |
2828 | mm = get_task_mm(task); | |
2829 | if (mm) { | |
2830 | len = snprintf(buffer, sizeof(buffer), "%08lx\n", | |
2831 | ((mm->flags & MMF_DUMP_FILTER_MASK) >> | |
2832 | MMF_DUMP_FILTER_SHIFT)); | |
2833 | mmput(mm); | |
2834 | ret = simple_read_from_buffer(buf, count, ppos, buffer, len); | |
2835 | } | |
2836 | ||
2837 | put_task_struct(task); | |
2838 | ||
2839 | return ret; | |
2840 | } | |
2841 | ||
2842 | static ssize_t proc_coredump_filter_write(struct file *file, | |
2843 | const char __user *buf, | |
2844 | size_t count, | |
2845 | loff_t *ppos) | |
2846 | { | |
2847 | struct task_struct *task; | |
2848 | struct mm_struct *mm; | |
3cb4a0bb KH |
2849 | unsigned int val; |
2850 | int ret; | |
2851 | int i; | |
2852 | unsigned long mask; | |
2853 | ||
774636e1 AD |
2854 | ret = kstrtouint_from_user(buf, count, 0, &val); |
2855 | if (ret < 0) | |
2856 | return ret; | |
3cb4a0bb KH |
2857 | |
2858 | ret = -ESRCH; | |
496ad9aa | 2859 | task = get_proc_task(file_inode(file)); |
3cb4a0bb KH |
2860 | if (!task) |
2861 | goto out_no_task; | |
2862 | ||
3cb4a0bb KH |
2863 | mm = get_task_mm(task); |
2864 | if (!mm) | |
2865 | goto out_no_mm; | |
41a0c249 | 2866 | ret = 0; |
3cb4a0bb KH |
2867 | |
2868 | for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) { | |
2869 | if (val & mask) | |
2870 | set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2871 | else | |
2872 | clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2873 | } | |
2874 | ||
2875 | mmput(mm); | |
2876 | out_no_mm: | |
2877 | put_task_struct(task); | |
2878 | out_no_task: | |
774636e1 AD |
2879 | if (ret < 0) |
2880 | return ret; | |
2881 | return count; | |
3cb4a0bb KH |
2882 | } |
2883 | ||
2884 | static const struct file_operations proc_coredump_filter_operations = { | |
2885 | .read = proc_coredump_filter_read, | |
2886 | .write = proc_coredump_filter_write, | |
87df8424 | 2887 | .llseek = generic_file_llseek, |
3cb4a0bb KH |
2888 | }; |
2889 | #endif | |
2890 | ||
aba76fdb | 2891 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 2892 | static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole) |
297c5d92 | 2893 | { |
940389b8 | 2894 | struct task_io_accounting acct = task->ioac; |
5995477a | 2895 | unsigned long flags; |
293eb1e7 | 2896 | int result; |
5995477a | 2897 | |
76518d37 | 2898 | result = mutex_lock_killable(&task->signal->exec_update_mutex); |
293eb1e7 VK |
2899 | if (result) |
2900 | return result; | |
2901 | ||
caaee623 | 2902 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) { |
293eb1e7 VK |
2903 | result = -EACCES; |
2904 | goto out_unlock; | |
2905 | } | |
1d1221f3 | 2906 | |
5995477a AR |
2907 | if (whole && lock_task_sighand(task, &flags)) { |
2908 | struct task_struct *t = task; | |
2909 | ||
2910 | task_io_accounting_add(&acct, &task->signal->ioac); | |
2911 | while_each_thread(task, t) | |
2912 | task_io_accounting_add(&acct, &t->ioac); | |
2913 | ||
2914 | unlock_task_sighand(task, &flags); | |
297c5d92 | 2915 | } |
25ce3191 JP |
2916 | seq_printf(m, |
2917 | "rchar: %llu\n" | |
2918 | "wchar: %llu\n" | |
2919 | "syscr: %llu\n" | |
2920 | "syscw: %llu\n" | |
2921 | "read_bytes: %llu\n" | |
2922 | "write_bytes: %llu\n" | |
2923 | "cancelled_write_bytes: %llu\n", | |
2924 | (unsigned long long)acct.rchar, | |
2925 | (unsigned long long)acct.wchar, | |
2926 | (unsigned long long)acct.syscr, | |
2927 | (unsigned long long)acct.syscw, | |
2928 | (unsigned long long)acct.read_bytes, | |
2929 | (unsigned long long)acct.write_bytes, | |
2930 | (unsigned long long)acct.cancelled_write_bytes); | |
2931 | result = 0; | |
2932 | ||
293eb1e7 | 2933 | out_unlock: |
76518d37 | 2934 | mutex_unlock(&task->signal->exec_update_mutex); |
293eb1e7 | 2935 | return result; |
297c5d92 AR |
2936 | } |
2937 | ||
19aadc98 AD |
2938 | static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns, |
2939 | struct pid *pid, struct task_struct *task) | |
297c5d92 | 2940 | { |
19aadc98 | 2941 | return do_io_accounting(task, m, 0); |
aba76fdb | 2942 | } |
297c5d92 | 2943 | |
19aadc98 AD |
2944 | static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns, |
2945 | struct pid *pid, struct task_struct *task) | |
297c5d92 | 2946 | { |
19aadc98 | 2947 | return do_io_accounting(task, m, 1); |
297c5d92 AR |
2948 | } |
2949 | #endif /* CONFIG_TASK_IO_ACCOUNTING */ | |
aba76fdb | 2950 | |
22d917d8 EB |
2951 | #ifdef CONFIG_USER_NS |
2952 | static int proc_id_map_open(struct inode *inode, struct file *file, | |
ccf94f1b | 2953 | const struct seq_operations *seq_ops) |
22d917d8 EB |
2954 | { |
2955 | struct user_namespace *ns = NULL; | |
2956 | struct task_struct *task; | |
2957 | struct seq_file *seq; | |
2958 | int ret = -EINVAL; | |
2959 | ||
2960 | task = get_proc_task(inode); | |
2961 | if (task) { | |
2962 | rcu_read_lock(); | |
2963 | ns = get_user_ns(task_cred_xxx(task, user_ns)); | |
2964 | rcu_read_unlock(); | |
2965 | put_task_struct(task); | |
2966 | } | |
2967 | if (!ns) | |
2968 | goto err; | |
2969 | ||
2970 | ret = seq_open(file, seq_ops); | |
2971 | if (ret) | |
2972 | goto err_put_ns; | |
2973 | ||
2974 | seq = file->private_data; | |
2975 | seq->private = ns; | |
2976 | ||
2977 | return 0; | |
2978 | err_put_ns: | |
2979 | put_user_ns(ns); | |
2980 | err: | |
2981 | return ret; | |
2982 | } | |
2983 | ||
2984 | static int proc_id_map_release(struct inode *inode, struct file *file) | |
2985 | { | |
2986 | struct seq_file *seq = file->private_data; | |
2987 | struct user_namespace *ns = seq->private; | |
2988 | put_user_ns(ns); | |
2989 | return seq_release(inode, file); | |
2990 | } | |
2991 | ||
2992 | static int proc_uid_map_open(struct inode *inode, struct file *file) | |
2993 | { | |
2994 | return proc_id_map_open(inode, file, &proc_uid_seq_operations); | |
2995 | } | |
2996 | ||
2997 | static int proc_gid_map_open(struct inode *inode, struct file *file) | |
2998 | { | |
2999 | return proc_id_map_open(inode, file, &proc_gid_seq_operations); | |
3000 | } | |
3001 | ||
f76d207a EB |
3002 | static int proc_projid_map_open(struct inode *inode, struct file *file) |
3003 | { | |
3004 | return proc_id_map_open(inode, file, &proc_projid_seq_operations); | |
3005 | } | |
3006 | ||
22d917d8 EB |
3007 | static const struct file_operations proc_uid_map_operations = { |
3008 | .open = proc_uid_map_open, | |
3009 | .write = proc_uid_map_write, | |
3010 | .read = seq_read, | |
3011 | .llseek = seq_lseek, | |
3012 | .release = proc_id_map_release, | |
3013 | }; | |
3014 | ||
3015 | static const struct file_operations proc_gid_map_operations = { | |
3016 | .open = proc_gid_map_open, | |
3017 | .write = proc_gid_map_write, | |
3018 | .read = seq_read, | |
3019 | .llseek = seq_lseek, | |
3020 | .release = proc_id_map_release, | |
3021 | }; | |
f76d207a EB |
3022 | |
3023 | static const struct file_operations proc_projid_map_operations = { | |
3024 | .open = proc_projid_map_open, | |
3025 | .write = proc_projid_map_write, | |
3026 | .read = seq_read, | |
3027 | .llseek = seq_lseek, | |
3028 | .release = proc_id_map_release, | |
3029 | }; | |
9cc46516 EB |
3030 | |
3031 | static int proc_setgroups_open(struct inode *inode, struct file *file) | |
3032 | { | |
3033 | struct user_namespace *ns = NULL; | |
3034 | struct task_struct *task; | |
3035 | int ret; | |
3036 | ||
3037 | ret = -ESRCH; | |
3038 | task = get_proc_task(inode); | |
3039 | if (task) { | |
3040 | rcu_read_lock(); | |
3041 | ns = get_user_ns(task_cred_xxx(task, user_ns)); | |
3042 | rcu_read_unlock(); | |
3043 | put_task_struct(task); | |
3044 | } | |
3045 | if (!ns) | |
3046 | goto err; | |
3047 | ||
3048 | if (file->f_mode & FMODE_WRITE) { | |
3049 | ret = -EACCES; | |
3050 | if (!ns_capable(ns, CAP_SYS_ADMIN)) | |
3051 | goto err_put_ns; | |
3052 | } | |
3053 | ||
3054 | ret = single_open(file, &proc_setgroups_show, ns); | |
3055 | if (ret) | |
3056 | goto err_put_ns; | |
3057 | ||
3058 | return 0; | |
3059 | err_put_ns: | |
3060 | put_user_ns(ns); | |
3061 | err: | |
3062 | return ret; | |
3063 | } | |
3064 | ||
3065 | static int proc_setgroups_release(struct inode *inode, struct file *file) | |
3066 | { | |
3067 | struct seq_file *seq = file->private_data; | |
3068 | struct user_namespace *ns = seq->private; | |
3069 | int ret = single_release(inode, file); | |
3070 | put_user_ns(ns); | |
3071 | return ret; | |
3072 | } | |
3073 | ||
3074 | static const struct file_operations proc_setgroups_operations = { | |
3075 | .open = proc_setgroups_open, | |
3076 | .write = proc_setgroups_write, | |
3077 | .read = seq_read, | |
3078 | .llseek = seq_lseek, | |
3079 | .release = proc_setgroups_release, | |
3080 | }; | |
22d917d8 EB |
3081 | #endif /* CONFIG_USER_NS */ |
3082 | ||
47830723 KC |
3083 | static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns, |
3084 | struct pid *pid, struct task_struct *task) | |
3085 | { | |
a9712bc1 AV |
3086 | int err = lock_trace(task); |
3087 | if (!err) { | |
3088 | seq_printf(m, "%08x\n", task->personality); | |
3089 | unlock_trace(task); | |
3090 | } | |
3091 | return err; | |
47830723 KC |
3092 | } |
3093 | ||
7c23b330 JP |
3094 | #ifdef CONFIG_LIVEPATCH |
3095 | static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns, | |
3096 | struct pid *pid, struct task_struct *task) | |
3097 | { | |
3098 | seq_printf(m, "%d\n", task->patch_state); | |
3099 | return 0; | |
3100 | } | |
3101 | #endif /* CONFIG_LIVEPATCH */ | |
3102 | ||
c8d12627 AP |
3103 | #ifdef CONFIG_STACKLEAK_METRICS |
3104 | static int proc_stack_depth(struct seq_file *m, struct pid_namespace *ns, | |
3105 | struct pid *pid, struct task_struct *task) | |
3106 | { | |
3107 | unsigned long prev_depth = THREAD_SIZE - | |
3108 | (task->prev_lowest_stack & (THREAD_SIZE - 1)); | |
3109 | unsigned long depth = THREAD_SIZE - | |
3110 | (task->lowest_stack & (THREAD_SIZE - 1)); | |
3111 | ||
3112 | seq_printf(m, "previous stack depth: %lu\nstack depth: %lu\n", | |
3113 | prev_depth, depth); | |
3114 | return 0; | |
3115 | } | |
3116 | #endif /* CONFIG_STACKLEAK_METRICS */ | |
3117 | ||
28a6d671 EB |
3118 | /* |
3119 | * Thread groups | |
3120 | */ | |
00977a59 | 3121 | static const struct file_operations proc_task_operations; |
c5ef1c42 | 3122 | static const struct inode_operations proc_task_inode_operations; |
20cdc894 | 3123 | |
c5141e6d | 3124 | static const struct pid_entry tgid_base_stuff[] = { |
631f9c18 AD |
3125 | DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations), |
3126 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), | |
640708a2 | 3127 | DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations), |
631f9c18 | 3128 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
6b4e306a | 3129 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
b2211a36 | 3130 | #ifdef CONFIG_NET |
631f9c18 | 3131 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), |
b2211a36 | 3132 | #endif |
631f9c18 | 3133 | REG("environ", S_IRUSR, proc_environ_operations), |
c5317167 | 3134 | REG("auxv", S_IRUSR, proc_auxv_operations), |
631f9c18 | 3135 | ONE("status", S_IRUGO, proc_pid_status), |
35a35046 | 3136 | ONE("personality", S_IRUSR, proc_pid_personality), |
1c963eb1 | 3137 | ONE("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 3138 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 3139 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
5091faa4 MG |
3140 | #endif |
3141 | #ifdef CONFIG_SCHED_AUTOGROUP | |
3142 | REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations), | |
04a8682a AV |
3143 | #endif |
3144 | #ifdef CONFIG_TIME_NS | |
3145 | REG("timens_offsets", S_IRUGO|S_IWUSR, proc_timens_offsets_operations), | |
ebcb6734 | 3146 | #endif |
4614a696 | 3147 | REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations), |
ebcb6734 | 3148 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 | 3149 | ONE("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 3150 | #endif |
c2c0bb44 | 3151 | REG("cmdline", S_IRUGO, proc_pid_cmdline_ops), |
631f9c18 AD |
3152 | ONE("stat", S_IRUGO, proc_tgid_stat), |
3153 | ONE("statm", S_IRUGO, proc_pid_statm), | |
b7643757 | 3154 | REG("maps", S_IRUGO, proc_pid_maps_operations), |
28a6d671 | 3155 | #ifdef CONFIG_NUMA |
b7643757 | 3156 | REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations), |
28a6d671 | 3157 | #endif |
631f9c18 AD |
3158 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
3159 | LNK("cwd", proc_cwd_link), | |
3160 | LNK("root", proc_root_link), | |
3161 | LNK("exe", proc_exe_link), | |
3162 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
3163 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
3164 | REG("mountstats", S_IRUSR, proc_mountstats_operations), | |
1e883281 | 3165 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 | 3166 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
b7643757 | 3167 | REG("smaps", S_IRUGO, proc_pid_smaps_operations), |
493b0e9d | 3168 | REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations), |
32ed74a4 | 3169 | REG("pagemap", S_IRUSR, proc_pagemap_operations), |
28a6d671 EB |
3170 | #endif |
3171 | #ifdef CONFIG_SECURITY | |
631f9c18 | 3172 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
3173 | #endif |
3174 | #ifdef CONFIG_KALLSYMS | |
edfcd606 | 3175 | ONE("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 3176 | #endif |
2ec220e2 | 3177 | #ifdef CONFIG_STACKTRACE |
35a35046 | 3178 | ONE("stack", S_IRUSR, proc_pid_stack), |
28a6d671 | 3179 | #endif |
5968cece | 3180 | #ifdef CONFIG_SCHED_INFO |
f6e826ca | 3181 | ONE("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 3182 | #endif |
9745512c | 3183 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 3184 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 3185 | #endif |
8793d854 | 3186 | #ifdef CONFIG_PROC_PID_CPUSET |
52de4779 | 3187 | ONE("cpuset", S_IRUGO, proc_cpuset_show), |
a424316c PM |
3188 | #endif |
3189 | #ifdef CONFIG_CGROUPS | |
006f4ac4 | 3190 | ONE("cgroup", S_IRUGO, proc_cgroup_show), |
e79f15a4 CY |
3191 | #endif |
3192 | #ifdef CONFIG_PROC_CPU_RESCTRL | |
3193 | ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show), | |
28a6d671 | 3194 | #endif |
6ba51e37 | 3195 | ONE("oom_score", S_IRUGO, proc_oom_score), |
fa0cbbf1 | 3196 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations), |
a63d83f4 | 3197 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
4b7d248b | 3198 | #ifdef CONFIG_AUDIT |
631f9c18 AD |
3199 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
3200 | REG("sessionid", S_IRUGO, proc_sessionid_operations), | |
28a6d671 | 3201 | #endif |
f4f154fd | 3202 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 3203 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
168c42bc | 3204 | REG("fail-nth", 0644, proc_fail_nth_operations), |
f4f154fd | 3205 | #endif |
698ba7b5 | 3206 | #ifdef CONFIG_ELF_CORE |
631f9c18 | 3207 | REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations), |
3cb4a0bb | 3208 | #endif |
aba76fdb | 3209 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 3210 | ONE("io", S_IRUSR, proc_tgid_io_accounting), |
aba76fdb | 3211 | #endif |
22d917d8 EB |
3212 | #ifdef CONFIG_USER_NS |
3213 | REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations), | |
3214 | REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations), | |
f76d207a | 3215 | REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations), |
9cc46516 | 3216 | REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations), |
22d917d8 | 3217 | #endif |
b18b6a9c | 3218 | #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS) |
48f6a7a5 PE |
3219 | REG("timers", S_IRUGO, proc_timers_operations), |
3220 | #endif | |
5de23d43 | 3221 | REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations), |
7c23b330 JP |
3222 | #ifdef CONFIG_LIVEPATCH |
3223 | ONE("patch_state", S_IRUSR, proc_pid_patch_state), | |
3224 | #endif | |
c8d12627 AP |
3225 | #ifdef CONFIG_STACKLEAK_METRICS |
3226 | ONE("stack_depth", S_IRUGO, proc_stack_depth), | |
3227 | #endif | |
68bc30bb AL |
3228 | #ifdef CONFIG_PROC_PID_ARCH_STATUS |
3229 | ONE("arch_status", S_IRUGO, proc_pid_arch_status), | |
3230 | #endif | |
28a6d671 | 3231 | }; |
1da177e4 | 3232 | |
f0c3b509 | 3233 | static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3234 | { |
f0c3b509 AV |
3235 | return proc_pident_readdir(file, ctx, |
3236 | tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
3237 | } |
3238 | ||
00977a59 | 3239 | static const struct file_operations proc_tgid_base_operations = { |
1da177e4 | 3240 | .read = generic_read_dir, |
f50752ea AV |
3241 | .iterate_shared = proc_tgid_base_readdir, |
3242 | .llseek = generic_file_llseek, | |
1da177e4 LT |
3243 | }; |
3244 | ||
3eb39f47 CB |
3245 | struct pid *tgid_pidfd_to_pid(const struct file *file) |
3246 | { | |
30d158b1 | 3247 | if (file->f_op != &proc_tgid_base_operations) |
3eb39f47 CB |
3248 | return ERR_PTR(-EBADF); |
3249 | ||
3250 | return proc_pid(file_inode(file)); | |
3251 | } | |
3252 | ||
00cd8dd3 AV |
3253 | static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
3254 | { | |
7bcd6b0e | 3255 | return proc_pident_lookup(dir, dentry, |
d5a572a4 AD |
3256 | tgid_base_stuff, |
3257 | tgid_base_stuff + ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
3258 | } |
3259 | ||
c5ef1c42 | 3260 | static const struct inode_operations proc_tgid_base_inode_operations = { |
28a6d671 | 3261 | .lookup = proc_tgid_base_lookup, |
99f89551 | 3262 | .getattr = pid_getattr, |
6d76fa58 | 3263 | .setattr = proc_setattr, |
0499680a | 3264 | .permission = proc_pid_permission, |
1da177e4 | 3265 | }; |
1da177e4 | 3266 | |
0895e91d | 3267 | /** |
7bc3e6e5 EB |
3268 | * proc_flush_pid - Remove dcache entries for @pid from the /proc dcache. |
3269 | * @pid: pid that should be flushed. | |
0895e91d | 3270 | * |
7bc3e6e5 EB |
3271 | * This function walks a list of inodes (that belong to any proc |
3272 | * filesystem) that are attached to the pid and flushes them from | |
3273 | * the dentry cache. | |
0895e91d RD |
3274 | * |
3275 | * It is safe and reasonable to cache /proc entries for a task until | |
3276 | * that task exits. After that they just clog up the dcache with | |
3277 | * useless entries, possibly causing useful dcache entries to be | |
7bc3e6e5 EB |
3278 | * flushed instead. This routine is provided to flush those useless |
3279 | * dcache entries when a process is reaped. | |
0895e91d RD |
3280 | * |
3281 | * NOTE: This routine is just an optimization so it does not guarantee | |
7bc3e6e5 EB |
3282 | * that no dcache entries will exist after a process is reaped |
3283 | * it just makes it very unlikely that any will persist. | |
60347f67 PE |
3284 | */ |
3285 | ||
7bc3e6e5 | 3286 | void proc_flush_pid(struct pid *pid) |
60347f67 | 3287 | { |
63f818f4 | 3288 | proc_invalidate_siblings_dcache(&pid->inodes, &pid->lock); |
60347f67 PE |
3289 | } |
3290 | ||
0168b9e3 | 3291 | static struct dentry *proc_pid_instantiate(struct dentry * dentry, |
c52a47ac | 3292 | struct task_struct *task, const void *ptr) |
444ceed8 | 3293 | { |
444ceed8 EB |
3294 | struct inode *inode; |
3295 | ||
0168b9e3 | 3296 | inode = proc_pid_make_inode(dentry->d_sb, task, S_IFDIR | S_IRUGO | S_IXUGO); |
444ceed8 | 3297 | if (!inode) |
0168b9e3 | 3298 | return ERR_PTR(-ENOENT); |
444ceed8 | 3299 | |
444ceed8 EB |
3300 | inode->i_op = &proc_tgid_base_inode_operations; |
3301 | inode->i_fop = &proc_tgid_base_operations; | |
3302 | inode->i_flags|=S_IMMUTABLE; | |
aed54175 | 3303 | |
1270dd8d | 3304 | set_nlink(inode, nlink_tgid); |
1bbc5513 | 3305 | pid_update_inode(task, inode); |
444ceed8 | 3306 | |
fb045adb | 3307 | d_set_d_op(dentry, &pid_dentry_operations); |
0168b9e3 | 3308 | return d_splice_alias(inode, dentry); |
444ceed8 EB |
3309 | } |
3310 | ||
867aaccf | 3311 | struct dentry *proc_pid_lookup(struct dentry *dentry, unsigned int flags) |
1da177e4 LT |
3312 | { |
3313 | struct task_struct *task; | |
1da177e4 | 3314 | unsigned tgid; |
b488893a | 3315 | struct pid_namespace *ns; |
0168b9e3 | 3316 | struct dentry *result = ERR_PTR(-ENOENT); |
1da177e4 | 3317 | |
dbcdb504 | 3318 | tgid = name_to_int(&dentry->d_name); |
1da177e4 LT |
3319 | if (tgid == ~0U) |
3320 | goto out; | |
3321 | ||
b488893a | 3322 | ns = dentry->d_sb->s_fs_info; |
de758734 | 3323 | rcu_read_lock(); |
b488893a | 3324 | task = find_task_by_pid_ns(tgid, ns); |
1da177e4 LT |
3325 | if (task) |
3326 | get_task_struct(task); | |
de758734 | 3327 | rcu_read_unlock(); |
1da177e4 LT |
3328 | if (!task) |
3329 | goto out; | |
3330 | ||
0168b9e3 | 3331 | result = proc_pid_instantiate(dentry, task, NULL); |
1da177e4 | 3332 | put_task_struct(task); |
1da177e4 | 3333 | out: |
0168b9e3 | 3334 | return result; |
1da177e4 LT |
3335 | } |
3336 | ||
1da177e4 | 3337 | /* |
0804ef4b | 3338 | * Find the first task with tgid >= tgid |
0bc58a91 | 3339 | * |
1da177e4 | 3340 | */ |
19fd4bb2 EB |
3341 | struct tgid_iter { |
3342 | unsigned int tgid; | |
0804ef4b | 3343 | struct task_struct *task; |
19fd4bb2 EB |
3344 | }; |
3345 | static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter) | |
3346 | { | |
0804ef4b | 3347 | struct pid *pid; |
1da177e4 | 3348 | |
19fd4bb2 EB |
3349 | if (iter.task) |
3350 | put_task_struct(iter.task); | |
454cc105 | 3351 | rcu_read_lock(); |
0804ef4b | 3352 | retry: |
19fd4bb2 EB |
3353 | iter.task = NULL; |
3354 | pid = find_ge_pid(iter.tgid, ns); | |
0804ef4b | 3355 | if (pid) { |
19fd4bb2 EB |
3356 | iter.tgid = pid_nr_ns(pid, ns); |
3357 | iter.task = pid_task(pid, PIDTYPE_PID); | |
0804ef4b EB |
3358 | /* What we to know is if the pid we have find is the |
3359 | * pid of a thread_group_leader. Testing for task | |
3360 | * being a thread_group_leader is the obvious thing | |
3361 | * todo but there is a window when it fails, due to | |
3362 | * the pid transfer logic in de_thread. | |
3363 | * | |
3364 | * So we perform the straight forward test of seeing | |
3365 | * if the pid we have found is the pid of a thread | |
3366 | * group leader, and don't worry if the task we have | |
3367 | * found doesn't happen to be a thread group leader. | |
3368 | * As we don't care in the case of readdir. | |
3369 | */ | |
19fd4bb2 EB |
3370 | if (!iter.task || !has_group_leader_pid(iter.task)) { |
3371 | iter.tgid += 1; | |
0804ef4b | 3372 | goto retry; |
19fd4bb2 EB |
3373 | } |
3374 | get_task_struct(iter.task); | |
0bc58a91 | 3375 | } |
454cc105 | 3376 | rcu_read_unlock(); |
19fd4bb2 | 3377 | return iter; |
1da177e4 LT |
3378 | } |
3379 | ||
0097875b | 3380 | #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2) |
0804ef4b | 3381 | |
1da177e4 | 3382 | /* for the /proc/ directory itself, after non-process stuff has been done */ |
f0c3b509 | 3383 | int proc_pid_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3384 | { |
19fd4bb2 | 3385 | struct tgid_iter iter; |
76f668be | 3386 | struct pid_namespace *ns = proc_pid_ns(file_inode(file)); |
f0c3b509 | 3387 | loff_t pos = ctx->pos; |
1da177e4 | 3388 | |
021ada7d | 3389 | if (pos >= PID_MAX_LIMIT + TGID_OFFSET) |
f0c3b509 | 3390 | return 0; |
1da177e4 | 3391 | |
0097875b | 3392 | if (pos == TGID_OFFSET - 2) { |
2b0143b5 | 3393 | struct inode *inode = d_inode(ns->proc_self); |
db963164 | 3394 | if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK)) |
f0c3b509 | 3395 | return 0; |
0097875b EB |
3396 | ctx->pos = pos = pos + 1; |
3397 | } | |
3398 | if (pos == TGID_OFFSET - 1) { | |
2b0143b5 | 3399 | struct inode *inode = d_inode(ns->proc_thread_self); |
0097875b EB |
3400 | if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK)) |
3401 | return 0; | |
3402 | ctx->pos = pos = pos + 1; | |
021ada7d | 3403 | } |
0097875b | 3404 | iter.tgid = pos - TGID_OFFSET; |
19fd4bb2 | 3405 | iter.task = NULL; |
19fd4bb2 EB |
3406 | for (iter = next_tgid(ns, iter); |
3407 | iter.task; | |
3408 | iter.tgid += 1, iter = next_tgid(ns, iter)) { | |
e3912ac3 | 3409 | char name[10 + 1]; |
a4ef3895 | 3410 | unsigned int len; |
3ba4bcee ED |
3411 | |
3412 | cond_resched(); | |
796f571b | 3413 | if (!has_pid_permissions(ns, iter.task, HIDEPID_INVISIBLE)) |
f0c3b509 | 3414 | continue; |
0499680a | 3415 | |
e3912ac3 | 3416 | len = snprintf(name, sizeof(name), "%u", iter.tgid); |
f0c3b509 AV |
3417 | ctx->pos = iter.tgid + TGID_OFFSET; |
3418 | if (!proc_fill_cache(file, ctx, name, len, | |
3419 | proc_pid_instantiate, iter.task, NULL)) { | |
19fd4bb2 | 3420 | put_task_struct(iter.task); |
f0c3b509 | 3421 | return 0; |
1da177e4 | 3422 | } |
0bc58a91 | 3423 | } |
f0c3b509 | 3424 | ctx->pos = PID_MAX_LIMIT + TGID_OFFSET; |
0bc58a91 EB |
3425 | return 0; |
3426 | } | |
1da177e4 | 3427 | |
1b3044e3 JD |
3428 | /* |
3429 | * proc_tid_comm_permission is a special permission function exclusively | |
3430 | * used for the node /proc/<pid>/task/<tid>/comm. | |
3431 | * It bypasses generic permission checks in the case where a task of the same | |
3432 | * task group attempts to access the node. | |
3433 | * The rationale behind this is that glibc and bionic access this node for | |
3434 | * cross thread naming (pthread_set/getname_np(!self)). However, if | |
3435 | * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0, | |
3436 | * which locks out the cross thread naming implementation. | |
3437 | * This function makes sure that the node is always accessible for members of | |
3438 | * same thread group. | |
3439 | */ | |
3440 | static int proc_tid_comm_permission(struct inode *inode, int mask) | |
3441 | { | |
3442 | bool is_same_tgroup; | |
3443 | struct task_struct *task; | |
3444 | ||
3445 | task = get_proc_task(inode); | |
3446 | if (!task) | |
3447 | return -ESRCH; | |
3448 | is_same_tgroup = same_thread_group(current, task); | |
3449 | put_task_struct(task); | |
3450 | ||
3451 | if (likely(is_same_tgroup && !(mask & MAY_EXEC))) { | |
3452 | /* This file (/proc/<pid>/task/<tid>/comm) can always be | |
3453 | * read or written by the members of the corresponding | |
3454 | * thread group. | |
3455 | */ | |
3456 | return 0; | |
3457 | } | |
3458 | ||
3459 | return generic_permission(inode, mask); | |
3460 | } | |
3461 | ||
3462 | static const struct inode_operations proc_tid_comm_inode_operations = { | |
3463 | .permission = proc_tid_comm_permission, | |
3464 | }; | |
3465 | ||
28a6d671 EB |
3466 | /* |
3467 | * Tasks | |
3468 | */ | |
c5141e6d | 3469 | static const struct pid_entry tid_base_stuff[] = { |
631f9c18 | 3470 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), |
3835541d | 3471 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
6b4e306a | 3472 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
6ba8ed79 EB |
3473 | #ifdef CONFIG_NET |
3474 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), | |
3475 | #endif | |
631f9c18 | 3476 | REG("environ", S_IRUSR, proc_environ_operations), |
c5317167 | 3477 | REG("auxv", S_IRUSR, proc_auxv_operations), |
631f9c18 | 3478 | ONE("status", S_IRUGO, proc_pid_status), |
35a35046 | 3479 | ONE("personality", S_IRUSR, proc_pid_personality), |
1c963eb1 | 3480 | ONE("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 3481 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 3482 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
ebcb6734 | 3483 | #endif |
1b3044e3 JD |
3484 | NOD("comm", S_IFREG|S_IRUGO|S_IWUSR, |
3485 | &proc_tid_comm_inode_operations, | |
3486 | &proc_pid_set_comm_operations, {}), | |
ebcb6734 | 3487 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 | 3488 | ONE("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 3489 | #endif |
c2c0bb44 | 3490 | REG("cmdline", S_IRUGO, proc_pid_cmdline_ops), |
631f9c18 AD |
3491 | ONE("stat", S_IRUGO, proc_tid_stat), |
3492 | ONE("statm", S_IRUGO, proc_pid_statm), | |
871305bb | 3493 | REG("maps", S_IRUGO, proc_pid_maps_operations), |
2e13ba54 | 3494 | #ifdef CONFIG_PROC_CHILDREN |
81841161 CG |
3495 | REG("children", S_IRUGO, proc_tid_children_operations), |
3496 | #endif | |
28a6d671 | 3497 | #ifdef CONFIG_NUMA |
871305bb | 3498 | REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations), |
28a6d671 | 3499 | #endif |
631f9c18 AD |
3500 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
3501 | LNK("cwd", proc_cwd_link), | |
3502 | LNK("root", proc_root_link), | |
3503 | LNK("exe", proc_exe_link), | |
3504 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
3505 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
1e883281 | 3506 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 | 3507 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
871305bb | 3508 | REG("smaps", S_IRUGO, proc_pid_smaps_operations), |
493b0e9d | 3509 | REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations), |
32ed74a4 | 3510 | REG("pagemap", S_IRUSR, proc_pagemap_operations), |
28a6d671 EB |
3511 | #endif |
3512 | #ifdef CONFIG_SECURITY | |
631f9c18 | 3513 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
3514 | #endif |
3515 | #ifdef CONFIG_KALLSYMS | |
edfcd606 | 3516 | ONE("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 3517 | #endif |
2ec220e2 | 3518 | #ifdef CONFIG_STACKTRACE |
35a35046 | 3519 | ONE("stack", S_IRUSR, proc_pid_stack), |
28a6d671 | 3520 | #endif |
5968cece | 3521 | #ifdef CONFIG_SCHED_INFO |
f6e826ca | 3522 | ONE("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 3523 | #endif |
9745512c | 3524 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 3525 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 3526 | #endif |
8793d854 | 3527 | #ifdef CONFIG_PROC_PID_CPUSET |
52de4779 | 3528 | ONE("cpuset", S_IRUGO, proc_cpuset_show), |
a424316c PM |
3529 | #endif |
3530 | #ifdef CONFIG_CGROUPS | |
006f4ac4 | 3531 | ONE("cgroup", S_IRUGO, proc_cgroup_show), |
e79f15a4 CY |
3532 | #endif |
3533 | #ifdef CONFIG_PROC_CPU_RESCTRL | |
3534 | ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show), | |
28a6d671 | 3535 | #endif |
6ba51e37 | 3536 | ONE("oom_score", S_IRUGO, proc_oom_score), |
fa0cbbf1 | 3537 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations), |
a63d83f4 | 3538 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
4b7d248b | 3539 | #ifdef CONFIG_AUDIT |
631f9c18 | 3540 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
26ec3c64 | 3541 | REG("sessionid", S_IRUGO, proc_sessionid_operations), |
28a6d671 | 3542 | #endif |
f4f154fd | 3543 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 3544 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
1203c8e6 | 3545 | REG("fail-nth", 0644, proc_fail_nth_operations), |
f4f154fd | 3546 | #endif |
297c5d92 | 3547 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 3548 | ONE("io", S_IRUSR, proc_tid_io_accounting), |
297c5d92 | 3549 | #endif |
22d917d8 EB |
3550 | #ifdef CONFIG_USER_NS |
3551 | REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations), | |
3552 | REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations), | |
f76d207a | 3553 | REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations), |
9cc46516 | 3554 | REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations), |
22d917d8 | 3555 | #endif |
7c23b330 JP |
3556 | #ifdef CONFIG_LIVEPATCH |
3557 | ONE("patch_state", S_IRUSR, proc_pid_patch_state), | |
3558 | #endif | |
68bc30bb AL |
3559 | #ifdef CONFIG_PROC_PID_ARCH_STATUS |
3560 | ONE("arch_status", S_IRUGO, proc_pid_arch_status), | |
3561 | #endif | |
28a6d671 EB |
3562 | }; |
3563 | ||
f0c3b509 | 3564 | static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx) |
28a6d671 | 3565 | { |
f0c3b509 AV |
3566 | return proc_pident_readdir(file, ctx, |
3567 | tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
3568 | } |
3569 | ||
00cd8dd3 AV |
3570 | static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
3571 | { | |
7bcd6b0e | 3572 | return proc_pident_lookup(dir, dentry, |
d5a572a4 AD |
3573 | tid_base_stuff, |
3574 | tid_base_stuff + ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
3575 | } |
3576 | ||
00977a59 | 3577 | static const struct file_operations proc_tid_base_operations = { |
28a6d671 | 3578 | .read = generic_read_dir, |
f50752ea AV |
3579 | .iterate_shared = proc_tid_base_readdir, |
3580 | .llseek = generic_file_llseek, | |
28a6d671 EB |
3581 | }; |
3582 | ||
c5ef1c42 | 3583 | static const struct inode_operations proc_tid_base_inode_operations = { |
28a6d671 EB |
3584 | .lookup = proc_tid_base_lookup, |
3585 | .getattr = pid_getattr, | |
3586 | .setattr = proc_setattr, | |
3587 | }; | |
3588 | ||
0168b9e3 AV |
3589 | static struct dentry *proc_task_instantiate(struct dentry *dentry, |
3590 | struct task_struct *task, const void *ptr) | |
444ceed8 | 3591 | { |
444ceed8 | 3592 | struct inode *inode; |
0168b9e3 | 3593 | inode = proc_pid_make_inode(dentry->d_sb, task, S_IFDIR | S_IRUGO | S_IXUGO); |
444ceed8 | 3594 | if (!inode) |
0168b9e3 | 3595 | return ERR_PTR(-ENOENT); |
1bbc5513 | 3596 | |
444ceed8 EB |
3597 | inode->i_op = &proc_tid_base_inode_operations; |
3598 | inode->i_fop = &proc_tid_base_operations; | |
1bbc5513 | 3599 | inode->i_flags |= S_IMMUTABLE; |
aed54175 | 3600 | |
1270dd8d | 3601 | set_nlink(inode, nlink_tid); |
1bbc5513 | 3602 | pid_update_inode(task, inode); |
444ceed8 | 3603 | |
fb045adb | 3604 | d_set_d_op(dentry, &pid_dentry_operations); |
0168b9e3 | 3605 | return d_splice_alias(inode, dentry); |
444ceed8 EB |
3606 | } |
3607 | ||
00cd8dd3 | 3608 | static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags) |
28a6d671 | 3609 | { |
28a6d671 EB |
3610 | struct task_struct *task; |
3611 | struct task_struct *leader = get_proc_task(dir); | |
28a6d671 | 3612 | unsigned tid; |
b488893a | 3613 | struct pid_namespace *ns; |
0168b9e3 | 3614 | struct dentry *result = ERR_PTR(-ENOENT); |
28a6d671 EB |
3615 | |
3616 | if (!leader) | |
3617 | goto out_no_task; | |
3618 | ||
dbcdb504 | 3619 | tid = name_to_int(&dentry->d_name); |
28a6d671 EB |
3620 | if (tid == ~0U) |
3621 | goto out; | |
3622 | ||
b488893a | 3623 | ns = dentry->d_sb->s_fs_info; |
28a6d671 | 3624 | rcu_read_lock(); |
b488893a | 3625 | task = find_task_by_pid_ns(tid, ns); |
28a6d671 EB |
3626 | if (task) |
3627 | get_task_struct(task); | |
3628 | rcu_read_unlock(); | |
3629 | if (!task) | |
3630 | goto out; | |
bac0abd6 | 3631 | if (!same_thread_group(leader, task)) |
28a6d671 EB |
3632 | goto out_drop_task; |
3633 | ||
0168b9e3 | 3634 | result = proc_task_instantiate(dentry, task, NULL); |
28a6d671 EB |
3635 | out_drop_task: |
3636 | put_task_struct(task); | |
3637 | out: | |
3638 | put_task_struct(leader); | |
3639 | out_no_task: | |
0168b9e3 | 3640 | return result; |
28a6d671 EB |
3641 | } |
3642 | ||
0bc58a91 EB |
3643 | /* |
3644 | * Find the first tid of a thread group to return to user space. | |
3645 | * | |
3646 | * Usually this is just the thread group leader, but if the users | |
3647 | * buffer was too small or there was a seek into the middle of the | |
3648 | * directory we have more work todo. | |
3649 | * | |
3650 | * In the case of a short read we start with find_task_by_pid. | |
3651 | * | |
3652 | * In the case of a seek we start with the leader and walk nr | |
3653 | * threads past it. | |
3654 | */ | |
9f6e963f ON |
3655 | static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos, |
3656 | struct pid_namespace *ns) | |
0bc58a91 | 3657 | { |
d855a4b7 | 3658 | struct task_struct *pos, *task; |
9f6e963f ON |
3659 | unsigned long nr = f_pos; |
3660 | ||
3661 | if (nr != f_pos) /* 32bit overflow? */ | |
3662 | return NULL; | |
1da177e4 | 3663 | |
cc288738 | 3664 | rcu_read_lock(); |
d855a4b7 ON |
3665 | task = pid_task(pid, PIDTYPE_PID); |
3666 | if (!task) | |
3667 | goto fail; | |
3668 | ||
3669 | /* Attempt to start with the tid of a thread */ | |
9f6e963f | 3670 | if (tid && nr) { |
b488893a | 3671 | pos = find_task_by_pid_ns(tid, ns); |
d855a4b7 | 3672 | if (pos && same_thread_group(pos, task)) |
a872ff0c | 3673 | goto found; |
0bc58a91 | 3674 | } |
1da177e4 | 3675 | |
0bc58a91 | 3676 | /* If nr exceeds the number of threads there is nothing todo */ |
9f6e963f | 3677 | if (nr >= get_nr_threads(task)) |
c986c14a | 3678 | goto fail; |
1da177e4 | 3679 | |
a872ff0c ON |
3680 | /* If we haven't found our starting place yet start |
3681 | * with the leader and walk nr threads forward. | |
0bc58a91 | 3682 | */ |
d855a4b7 | 3683 | pos = task = task->group_leader; |
c986c14a | 3684 | do { |
9f6e963f | 3685 | if (!nr--) |
c986c14a | 3686 | goto found; |
d855a4b7 | 3687 | } while_each_thread(task, pos); |
c986c14a ON |
3688 | fail: |
3689 | pos = NULL; | |
3690 | goto out; | |
a872ff0c ON |
3691 | found: |
3692 | get_task_struct(pos); | |
3693 | out: | |
cc288738 | 3694 | rcu_read_unlock(); |
0bc58a91 EB |
3695 | return pos; |
3696 | } | |
3697 | ||
3698 | /* | |
3699 | * Find the next thread in the thread list. | |
3700 | * Return NULL if there is an error or no next thread. | |
3701 | * | |
3702 | * The reference to the input task_struct is released. | |
3703 | */ | |
3704 | static struct task_struct *next_tid(struct task_struct *start) | |
3705 | { | |
c1df7fb8 | 3706 | struct task_struct *pos = NULL; |
cc288738 | 3707 | rcu_read_lock(); |
c1df7fb8 | 3708 | if (pid_alive(start)) { |
0bc58a91 | 3709 | pos = next_thread(start); |
c1df7fb8 ON |
3710 | if (thread_group_leader(pos)) |
3711 | pos = NULL; | |
3712 | else | |
3713 | get_task_struct(pos); | |
3714 | } | |
cc288738 | 3715 | rcu_read_unlock(); |
0bc58a91 EB |
3716 | put_task_struct(start); |
3717 | return pos; | |
1da177e4 LT |
3718 | } |
3719 | ||
3720 | /* for the /proc/TGID/task/ directories */ | |
f0c3b509 | 3721 | static int proc_task_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3722 | { |
d855a4b7 ON |
3723 | struct inode *inode = file_inode(file); |
3724 | struct task_struct *task; | |
b488893a | 3725 | struct pid_namespace *ns; |
f0c3b509 | 3726 | int tid; |
1da177e4 | 3727 | |
d855a4b7 | 3728 | if (proc_inode_is_dead(inode)) |
f0c3b509 | 3729 | return -ENOENT; |
1da177e4 | 3730 | |
f0c3b509 | 3731 | if (!dir_emit_dots(file, ctx)) |
d855a4b7 | 3732 | return 0; |
1da177e4 | 3733 | |
0bc58a91 EB |
3734 | /* f_version caches the tgid value that the last readdir call couldn't |
3735 | * return. lseek aka telldir automagically resets f_version to 0. | |
3736 | */ | |
76f668be | 3737 | ns = proc_pid_ns(inode); |
f0c3b509 AV |
3738 | tid = (int)file->f_version; |
3739 | file->f_version = 0; | |
d855a4b7 | 3740 | for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns); |
0bc58a91 | 3741 | task; |
f0c3b509 | 3742 | task = next_tid(task), ctx->pos++) { |
e3912ac3 | 3743 | char name[10 + 1]; |
a4ef3895 | 3744 | unsigned int len; |
b488893a | 3745 | tid = task_pid_nr_ns(task, ns); |
e3912ac3 | 3746 | len = snprintf(name, sizeof(name), "%u", tid); |
f0c3b509 AV |
3747 | if (!proc_fill_cache(file, ctx, name, len, |
3748 | proc_task_instantiate, task, NULL)) { | |
0bc58a91 EB |
3749 | /* returning this tgid failed, save it as the first |
3750 | * pid for the next readir call */ | |
f0c3b509 | 3751 | file->f_version = (u64)tid; |
0bc58a91 | 3752 | put_task_struct(task); |
1da177e4 | 3753 | break; |
0bc58a91 | 3754 | } |
1da177e4 | 3755 | } |
d855a4b7 | 3756 | |
f0c3b509 | 3757 | return 0; |
1da177e4 | 3758 | } |
6e66b52b | 3759 | |
a528d35e DH |
3760 | static int proc_task_getattr(const struct path *path, struct kstat *stat, |
3761 | u32 request_mask, unsigned int query_flags) | |
6e66b52b | 3762 | { |
a528d35e | 3763 | struct inode *inode = d_inode(path->dentry); |
99f89551 | 3764 | struct task_struct *p = get_proc_task(inode); |
6e66b52b EB |
3765 | generic_fillattr(inode, stat); |
3766 | ||
99f89551 | 3767 | if (p) { |
99f89551 | 3768 | stat->nlink += get_nr_threads(p); |
99f89551 | 3769 | put_task_struct(p); |
6e66b52b EB |
3770 | } |
3771 | ||
3772 | return 0; | |
3773 | } | |
28a6d671 | 3774 | |
c5ef1c42 | 3775 | static const struct inode_operations proc_task_inode_operations = { |
28a6d671 EB |
3776 | .lookup = proc_task_lookup, |
3777 | .getattr = proc_task_getattr, | |
3778 | .setattr = proc_setattr, | |
0499680a | 3779 | .permission = proc_pid_permission, |
28a6d671 EB |
3780 | }; |
3781 | ||
00977a59 | 3782 | static const struct file_operations proc_task_operations = { |
28a6d671 | 3783 | .read = generic_read_dir, |
f50752ea AV |
3784 | .iterate_shared = proc_task_readdir, |
3785 | .llseek = generic_file_llseek, | |
28a6d671 | 3786 | }; |
1270dd8d AD |
3787 | |
3788 | void __init set_proc_pid_nlink(void) | |
3789 | { | |
3790 | nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
3791 | nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
3792 | } |