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> |
2ec220e2 | 70 | #include <linux/stacktrace.h> |
d85f50d5 | 71 | #include <linux/resource.h> |
5096add8 | 72 | #include <linux/module.h> |
1da177e4 LT |
73 | #include <linux/mount.h> |
74 | #include <linux/security.h> | |
75 | #include <linux/ptrace.h> | |
0d094efe | 76 | #include <linux/tracehook.h> |
87ebdc00 | 77 | #include <linux/printk.h> |
efb1a57d | 78 | #include <linux/cache.h> |
a424316c | 79 | #include <linux/cgroup.h> |
1da177e4 LT |
80 | #include <linux/cpuset.h> |
81 | #include <linux/audit.h> | |
5addc5dd | 82 | #include <linux/poll.h> |
1651e14e | 83 | #include <linux/nsproxy.h> |
8ac773b4 | 84 | #include <linux/oom.h> |
3cb4a0bb | 85 | #include <linux/elf.h> |
60347f67 | 86 | #include <linux/pid_namespace.h> |
22d917d8 | 87 | #include <linux/user_namespace.h> |
5ad4e53b | 88 | #include <linux/fs_struct.h> |
5a0e3ad6 | 89 | #include <linux/slab.h> |
4eb5aaa3 | 90 | #include <linux/sched/autogroup.h> |
6e84f315 | 91 | #include <linux/sched/mm.h> |
f7ccbae4 | 92 | #include <linux/sched/coredump.h> |
b17b0153 | 93 | #include <linux/sched/debug.h> |
3905f9ad | 94 | #include <linux/sched/stat.h> |
48f6a7a5 | 95 | #include <linux/posix-timers.h> |
04a8682a | 96 | #include <linux/time_namespace.h> |
e79f15a4 | 97 | #include <linux/resctrl.h> |
43d2b113 | 98 | #include <trace/events/oom.h> |
1da177e4 | 99 | #include "internal.h" |
faf60af1 | 100 | #include "fd.h" |
1da177e4 | 101 | |
ac7f1061 AD |
102 | #include "../../lib/kstrtox.h" |
103 | ||
0f2fe20f EB |
104 | /* NOTE: |
105 | * Implementing inode permission operations in /proc is almost | |
106 | * certainly an error. Permission checks need to happen during | |
107 | * each system call not at open time. The reason is that most of | |
108 | * what we wish to check for permissions in /proc varies at runtime. | |
109 | * | |
110 | * The classic example of a problem is opening file descriptors | |
111 | * in /proc for a task before it execs a suid executable. | |
112 | */ | |
113 | ||
efb1a57d AD |
114 | static u8 nlink_tid __ro_after_init; |
115 | static u8 nlink_tgid __ro_after_init; | |
1270dd8d | 116 | |
1da177e4 | 117 | struct pid_entry { |
cedbccab | 118 | const char *name; |
623f594e | 119 | unsigned int len; |
d161a13f | 120 | umode_t mode; |
c5ef1c42 | 121 | const struct inode_operations *iop; |
00977a59 | 122 | const struct file_operations *fop; |
20cdc894 | 123 | union proc_op op; |
1da177e4 LT |
124 | }; |
125 | ||
61a28784 | 126 | #define NOD(NAME, MODE, IOP, FOP, OP) { \ |
20cdc894 | 127 | .name = (NAME), \ |
c5141e6d | 128 | .len = sizeof(NAME) - 1, \ |
20cdc894 EB |
129 | .mode = MODE, \ |
130 | .iop = IOP, \ | |
131 | .fop = FOP, \ | |
132 | .op = OP, \ | |
133 | } | |
134 | ||
631f9c18 AD |
135 | #define DIR(NAME, MODE, iops, fops) \ |
136 | NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} ) | |
137 | #define LNK(NAME, get_link) \ | |
61a28784 | 138 | NOD(NAME, (S_IFLNK|S_IRWXUGO), \ |
20cdc894 | 139 | &proc_pid_link_inode_operations, NULL, \ |
631f9c18 AD |
140 | { .proc_get_link = get_link } ) |
141 | #define REG(NAME, MODE, fops) \ | |
142 | NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {}) | |
631f9c18 | 143 | #define ONE(NAME, MODE, show) \ |
6d9c939d | 144 | NOD(NAME, (S_IFREG|(MODE)), \ |
be614086 | 145 | NULL, &proc_single_file_operations, \ |
631f9c18 | 146 | { .proc_show = show } ) |
6d9c939d CS |
147 | #define ATTR(LSM, NAME, MODE) \ |
148 | NOD(NAME, (S_IFREG|(MODE)), \ | |
149 | NULL, &proc_pid_attr_operations, \ | |
150 | { .lsm = LSM }) | |
1da177e4 | 151 | |
aed54175 VN |
152 | /* |
153 | * Count the number of hardlinks for the pid_entry table, excluding the . | |
154 | * and .. links. | |
155 | */ | |
1270dd8d | 156 | static unsigned int __init pid_entry_nlink(const struct pid_entry *entries, |
aed54175 VN |
157 | unsigned int n) |
158 | { | |
159 | unsigned int i; | |
160 | unsigned int count; | |
161 | ||
1270dd8d | 162 | count = 2; |
aed54175 VN |
163 | for (i = 0; i < n; ++i) { |
164 | if (S_ISDIR(entries[i].mode)) | |
165 | ++count; | |
166 | } | |
167 | ||
168 | return count; | |
169 | } | |
170 | ||
f7ad3c6b | 171 | static int get_task_root(struct task_struct *task, struct path *root) |
1da177e4 | 172 | { |
7c2c7d99 HD |
173 | int result = -ENOENT; |
174 | ||
0494f6ec | 175 | task_lock(task); |
f7ad3c6b MS |
176 | if (task->fs) { |
177 | get_fs_root(task->fs, root); | |
7c2c7d99 HD |
178 | result = 0; |
179 | } | |
0494f6ec | 180 | task_unlock(task); |
7c2c7d99 | 181 | return result; |
0494f6ec MS |
182 | } |
183 | ||
7773fbc5 | 184 | static int proc_cwd_link(struct dentry *dentry, struct path *path) |
0494f6ec | 185 | { |
2b0143b5 | 186 | struct task_struct *task = get_proc_task(d_inode(dentry)); |
0494f6ec | 187 | int result = -ENOENT; |
99f89551 EB |
188 | |
189 | if (task) { | |
f7ad3c6b MS |
190 | task_lock(task); |
191 | if (task->fs) { | |
192 | get_fs_pwd(task->fs, path); | |
193 | result = 0; | |
194 | } | |
195 | task_unlock(task); | |
99f89551 EB |
196 | put_task_struct(task); |
197 | } | |
1da177e4 LT |
198 | return result; |
199 | } | |
200 | ||
7773fbc5 | 201 | static int proc_root_link(struct dentry *dentry, struct path *path) |
1da177e4 | 202 | { |
2b0143b5 | 203 | struct task_struct *task = get_proc_task(d_inode(dentry)); |
1da177e4 | 204 | int result = -ENOENT; |
99f89551 EB |
205 | |
206 | if (task) { | |
f7ad3c6b | 207 | result = get_task_root(task, path); |
99f89551 EB |
208 | put_task_struct(task); |
209 | } | |
1da177e4 LT |
210 | return result; |
211 | } | |
212 | ||
d26d0cd9 LT |
213 | /* |
214 | * If the user used setproctitle(), we just get the string from | |
215 | * user space at arg_start, and limit it to a maximum of one page. | |
216 | */ | |
217 | static ssize_t get_mm_proctitle(struct mm_struct *mm, char __user *buf, | |
218 | size_t count, unsigned long pos, | |
219 | unsigned long arg_start) | |
220 | { | |
221 | char *page; | |
222 | int ret, got; | |
223 | ||
224 | if (pos >= PAGE_SIZE) | |
225 | return 0; | |
226 | ||
227 | page = (char *)__get_free_page(GFP_KERNEL); | |
228 | if (!page) | |
229 | return -ENOMEM; | |
230 | ||
231 | ret = 0; | |
232 | got = access_remote_vm(mm, arg_start, page, PAGE_SIZE, FOLL_ANON); | |
233 | if (got > 0) { | |
234 | int len = strnlen(page, got); | |
235 | ||
236 | /* Include the NUL character if it was found */ | |
237 | if (len < got) | |
238 | len++; | |
239 | ||
240 | if (len > pos) { | |
241 | len -= pos; | |
242 | if (len > count) | |
243 | len = count; | |
244 | len -= copy_to_user(buf, page+pos, len); | |
245 | if (!len) | |
246 | len = -EFAULT; | |
247 | ret = len; | |
248 | } | |
249 | } | |
250 | free_page((unsigned long)page); | |
251 | return ret; | |
252 | } | |
253 | ||
e4b4e441 | 254 | static ssize_t get_mm_cmdline(struct mm_struct *mm, char __user *buf, |
5ab82718 | 255 | size_t count, loff_t *ppos) |
1da177e4 | 256 | { |
c2c0bb44 | 257 | unsigned long arg_start, arg_end, env_start, env_end; |
5ab82718 | 258 | unsigned long pos, len; |
d26d0cd9 | 259 | char *page, c; |
c2c0bb44 | 260 | |
c2c0bb44 | 261 | /* Check if process spawned far enough to have cmdline. */ |
e4b4e441 LT |
262 | if (!mm->env_end) |
263 | return 0; | |
c2c0bb44 | 264 | |
88aa7cc6 | 265 | spin_lock(&mm->arg_lock); |
c2c0bb44 AD |
266 | arg_start = mm->arg_start; |
267 | arg_end = mm->arg_end; | |
268 | env_start = mm->env_start; | |
269 | env_end = mm->env_end; | |
88aa7cc6 | 270 | spin_unlock(&mm->arg_lock); |
c2c0bb44 | 271 | |
5ab82718 LT |
272 | if (arg_start >= arg_end) |
273 | return 0; | |
6a6cbe75 | 274 | |
2ca66ff7 | 275 | /* |
d26d0cd9 LT |
276 | * We allow setproctitle() to overwrite the argument |
277 | * strings, and overflow past the original end. But | |
278 | * only when it overflows into the environment area. | |
2ca66ff7 | 279 | */ |
d26d0cd9 | 280 | if (env_start != arg_end || env_end < env_start) |
5ab82718 | 281 | env_start = env_end = arg_end; |
d26d0cd9 | 282 | len = env_end - arg_start; |
f5b65348 | 283 | |
5ab82718 | 284 | /* We're not going to care if "*ppos" has high bits set */ |
d26d0cd9 LT |
285 | pos = *ppos; |
286 | if (pos >= len) | |
5ab82718 | 287 | return 0; |
d26d0cd9 LT |
288 | if (count > len - pos) |
289 | count = len - pos; | |
290 | if (!count) | |
291 | return 0; | |
292 | ||
293 | /* | |
294 | * Magical special case: if the argv[] end byte is not | |
295 | * zero, the user has overwritten it with setproctitle(3). | |
296 | * | |
297 | * Possible future enhancement: do this only once when | |
298 | * pos is 0, and set a flag in the 'struct file'. | |
299 | */ | |
300 | if (access_remote_vm(mm, arg_end-1, &c, 1, FOLL_ANON) == 1 && c) | |
301 | return get_mm_proctitle(mm, buf, count, pos, arg_start); | |
a0a07b87 | 302 | |
d26d0cd9 LT |
303 | /* |
304 | * For the non-setproctitle() case we limit things strictly | |
305 | * to the [arg_start, arg_end[ range. | |
306 | */ | |
307 | pos += arg_start; | |
3d712546 | 308 | if (pos < arg_start || pos >= arg_end) |
5ab82718 | 309 | return 0; |
3d712546 LT |
310 | if (count > arg_end - pos) |
311 | count = arg_end - pos; | |
5ab82718 LT |
312 | |
313 | page = (char *)__get_free_page(GFP_KERNEL); | |
314 | if (!page) | |
315 | return -ENOMEM; | |
316 | ||
317 | len = 0; | |
318 | while (count) { | |
319 | int got; | |
320 | size_t size = min_t(size_t, PAGE_SIZE, count); | |
321 | ||
3d712546 LT |
322 | got = access_remote_vm(mm, pos, page, size, FOLL_ANON); |
323 | if (got <= 0) | |
5ab82718 | 324 | break; |
3d712546 | 325 | got -= copy_to_user(buf, page, got); |
5ab82718 LT |
326 | if (unlikely(!got)) { |
327 | if (!len) | |
328 | len = -EFAULT; | |
329 | break; | |
c2c0bb44 | 330 | } |
5ab82718 LT |
331 | pos += got; |
332 | buf += got; | |
333 | len += got; | |
334 | count -= got; | |
c2c0bb44 AD |
335 | } |
336 | ||
c2c0bb44 | 337 | free_page((unsigned long)page); |
5ab82718 | 338 | return len; |
1da177e4 LT |
339 | } |
340 | ||
e4b4e441 LT |
341 | static ssize_t get_task_cmdline(struct task_struct *tsk, char __user *buf, |
342 | size_t count, loff_t *pos) | |
343 | { | |
344 | struct mm_struct *mm; | |
345 | ssize_t ret; | |
346 | ||
347 | mm = get_task_mm(tsk); | |
348 | if (!mm) | |
349 | return 0; | |
350 | ||
351 | ret = get_mm_cmdline(mm, buf, count, pos); | |
c2c0bb44 | 352 | mmput(mm); |
e4b4e441 LT |
353 | return ret; |
354 | } | |
355 | ||
356 | static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf, | |
357 | size_t count, loff_t *pos) | |
358 | { | |
359 | struct task_struct *tsk; | |
360 | ssize_t ret; | |
361 | ||
362 | BUG_ON(*pos < 0); | |
363 | ||
364 | tsk = get_proc_task(file_inode(file)); | |
365 | if (!tsk) | |
366 | return -ESRCH; | |
367 | ret = get_task_cmdline(tsk, buf, count, pos); | |
368 | put_task_struct(tsk); | |
369 | if (ret > 0) | |
370 | *pos += ret; | |
371 | return ret; | |
1da177e4 LT |
372 | } |
373 | ||
c2c0bb44 AD |
374 | static const struct file_operations proc_pid_cmdline_ops = { |
375 | .read = proc_pid_cmdline_read, | |
376 | .llseek = generic_file_llseek, | |
377 | }; | |
378 | ||
1da177e4 LT |
379 | #ifdef CONFIG_KALLSYMS |
380 | /* | |
381 | * Provides a wchan file via kallsyms in a proper one-value-per-file format. | |
382 | * Returns the resolved symbol. If that fails, simply return the address. | |
383 | */ | |
edfcd606 AD |
384 | static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns, |
385 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 386 | { |
ffb45122 | 387 | unsigned long wchan; |
1da177e4 | 388 | |
152c432b HD |
389 | if (ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
390 | wchan = get_wchan(task); | |
391 | else | |
392 | wchan = 0; | |
1da177e4 | 393 | |
152c432b HD |
394 | if (wchan) |
395 | seq_printf(m, "%ps", (void *) wchan); | |
396 | else | |
397 | seq_putc(m, '0'); | |
25ce3191 JP |
398 | |
399 | return 0; | |
1da177e4 LT |
400 | } |
401 | #endif /* CONFIG_KALLSYMS */ | |
402 | ||
a9712bc1 AV |
403 | static int lock_trace(struct task_struct *task) |
404 | { | |
f7cfd871 | 405 | int err = down_read_killable(&task->signal->exec_update_lock); |
a9712bc1 AV |
406 | if (err) |
407 | return err; | |
caaee623 | 408 | if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) { |
f7cfd871 | 409 | up_read(&task->signal->exec_update_lock); |
a9712bc1 AV |
410 | return -EPERM; |
411 | } | |
412 | return 0; | |
413 | } | |
414 | ||
415 | static void unlock_trace(struct task_struct *task) | |
416 | { | |
f7cfd871 | 417 | up_read(&task->signal->exec_update_lock); |
a9712bc1 AV |
418 | } |
419 | ||
2ec220e2 KC |
420 | #ifdef CONFIG_STACKTRACE |
421 | ||
422 | #define MAX_STACK_TRACE_DEPTH 64 | |
423 | ||
424 | static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns, | |
425 | struct pid *pid, struct task_struct *task) | |
426 | { | |
2ec220e2 | 427 | unsigned long *entries; |
a9712bc1 | 428 | int err; |
2ec220e2 | 429 | |
f8a00cef JH |
430 | /* |
431 | * The ability to racily run the kernel stack unwinder on a running task | |
432 | * and then observe the unwinder output is scary; while it is useful for | |
433 | * debugging kernel issues, it can also allow an attacker to leak kernel | |
434 | * stack contents. | |
435 | * Doing this in a manner that is at least safe from races would require | |
436 | * some work to ensure that the remote task can not be scheduled; and | |
437 | * even then, this would still expose the unwinder as local attack | |
438 | * surface. | |
439 | * Therefore, this interface is restricted to root. | |
440 | */ | |
441 | if (!file_ns_capable(m->file, &init_user_ns, CAP_SYS_ADMIN)) | |
442 | return -EACCES; | |
443 | ||
6da2ec56 KC |
444 | entries = kmalloc_array(MAX_STACK_TRACE_DEPTH, sizeof(*entries), |
445 | GFP_KERNEL); | |
2ec220e2 KC |
446 | if (!entries) |
447 | return -ENOMEM; | |
448 | ||
a9712bc1 AV |
449 | err = lock_trace(task); |
450 | if (!err) { | |
e988e5ec | 451 | unsigned int i, nr_entries; |
5d008fb4 | 452 | |
e988e5ec TG |
453 | nr_entries = stack_trace_save_tsk(task, entries, |
454 | MAX_STACK_TRACE_DEPTH, 0); | |
a9712bc1 | 455 | |
e988e5ec | 456 | for (i = 0; i < nr_entries; i++) { |
8f5abe84 | 457 | seq_printf(m, "[<0>] %pB\n", (void *)entries[i]); |
a9712bc1 | 458 | } |
e988e5ec | 459 | |
a9712bc1 | 460 | unlock_trace(task); |
2ec220e2 KC |
461 | } |
462 | kfree(entries); | |
463 | ||
a9712bc1 | 464 | return err; |
2ec220e2 KC |
465 | } |
466 | #endif | |
467 | ||
5968cece | 468 | #ifdef CONFIG_SCHED_INFO |
1da177e4 LT |
469 | /* |
470 | * Provides /proc/PID/schedstat | |
471 | */ | |
f6e826ca AD |
472 | static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns, |
473 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 474 | { |
5968cece | 475 | if (unlikely(!sched_info_on())) |
08b55775 | 476 | seq_puts(m, "0 0 0\n"); |
5968cece NR |
477 | else |
478 | seq_printf(m, "%llu %llu %lu\n", | |
25ce3191 JP |
479 | (unsigned long long)task->se.sum_exec_runtime, |
480 | (unsigned long long)task->sched_info.run_delay, | |
481 | task->sched_info.pcount); | |
482 | ||
483 | return 0; | |
1da177e4 LT |
484 | } |
485 | #endif | |
486 | ||
9745512c AV |
487 | #ifdef CONFIG_LATENCYTOP |
488 | static int lstats_show_proc(struct seq_file *m, void *v) | |
489 | { | |
490 | int i; | |
13d77c37 HS |
491 | struct inode *inode = m->private; |
492 | struct task_struct *task = get_proc_task(inode); | |
9745512c | 493 | |
13d77c37 HS |
494 | if (!task) |
495 | return -ESRCH; | |
496 | seq_puts(m, "Latency Top version : v0.1\n"); | |
f6d2f584 | 497 | for (i = 0; i < LT_SAVECOUNT; i++) { |
34e49d4f JP |
498 | struct latency_record *lr = &task->latency_record[i]; |
499 | if (lr->backtrace[0]) { | |
9745512c | 500 | int q; |
34e49d4f JP |
501 | seq_printf(m, "%i %li %li", |
502 | lr->count, lr->time, lr->max); | |
9745512c | 503 | for (q = 0; q < LT_BACKTRACEDEPTH; q++) { |
34e49d4f | 504 | unsigned long bt = lr->backtrace[q]; |
accddc41 | 505 | |
34e49d4f | 506 | if (!bt) |
9745512c | 507 | break; |
34e49d4f | 508 | seq_printf(m, " %ps", (void *)bt); |
9745512c | 509 | } |
9d6de12f | 510 | seq_putc(m, '\n'); |
9745512c AV |
511 | } |
512 | ||
513 | } | |
13d77c37 | 514 | put_task_struct(task); |
9745512c AV |
515 | return 0; |
516 | } | |
517 | ||
518 | static int lstats_open(struct inode *inode, struct file *file) | |
519 | { | |
13d77c37 | 520 | return single_open(file, lstats_show_proc, inode); |
d6643d12 HS |
521 | } |
522 | ||
9745512c AV |
523 | static ssize_t lstats_write(struct file *file, const char __user *buf, |
524 | size_t count, loff_t *offs) | |
525 | { | |
496ad9aa | 526 | struct task_struct *task = get_proc_task(file_inode(file)); |
9745512c | 527 | |
13d77c37 HS |
528 | if (!task) |
529 | return -ESRCH; | |
e02c9b0d | 530 | clear_tsk_latency_tracing(task); |
13d77c37 | 531 | put_task_struct(task); |
9745512c AV |
532 | |
533 | return count; | |
534 | } | |
535 | ||
536 | static const struct file_operations proc_lstats_operations = { | |
537 | .open = lstats_open, | |
538 | .read = seq_read, | |
539 | .write = lstats_write, | |
540 | .llseek = seq_lseek, | |
13d77c37 | 541 | .release = single_release, |
9745512c AV |
542 | }; |
543 | ||
544 | #endif | |
545 | ||
6ba51e37 AD |
546 | static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns, |
547 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 548 | { |
ca79b0c2 | 549 | unsigned long totalpages = totalram_pages() + total_swap_pages; |
b95c35e7 | 550 | unsigned long points = 0; |
9066e5cf YS |
551 | long badness; |
552 | ||
553 | badness = oom_badness(task, totalpages); | |
554 | /* | |
555 | * Special case OOM_SCORE_ADJ_MIN for all others scale the | |
556 | * badness value into [0, 2000] range which we have been | |
557 | * exporting for a long time so userspace might depend on it. | |
558 | */ | |
559 | if (badness != LONG_MIN) | |
560 | points = (1000 + badness * 1000 / (long)totalpages) * 2 / 3; | |
1da177e4 | 561 | |
25ce3191 JP |
562 | seq_printf(m, "%lu\n", points); |
563 | ||
564 | return 0; | |
1da177e4 LT |
565 | } |
566 | ||
d85f50d5 | 567 | struct limit_names { |
cedbccab AD |
568 | const char *name; |
569 | const char *unit; | |
d85f50d5 NH |
570 | }; |
571 | ||
572 | static const struct limit_names lnames[RLIM_NLIMITS] = { | |
cff4edb5 | 573 | [RLIMIT_CPU] = {"Max cpu time", "seconds"}, |
d85f50d5 NH |
574 | [RLIMIT_FSIZE] = {"Max file size", "bytes"}, |
575 | [RLIMIT_DATA] = {"Max data size", "bytes"}, | |
576 | [RLIMIT_STACK] = {"Max stack size", "bytes"}, | |
577 | [RLIMIT_CORE] = {"Max core file size", "bytes"}, | |
578 | [RLIMIT_RSS] = {"Max resident set", "bytes"}, | |
579 | [RLIMIT_NPROC] = {"Max processes", "processes"}, | |
580 | [RLIMIT_NOFILE] = {"Max open files", "files"}, | |
581 | [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"}, | |
582 | [RLIMIT_AS] = {"Max address space", "bytes"}, | |
583 | [RLIMIT_LOCKS] = {"Max file locks", "locks"}, | |
584 | [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"}, | |
585 | [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"}, | |
586 | [RLIMIT_NICE] = {"Max nice priority", NULL}, | |
587 | [RLIMIT_RTPRIO] = {"Max realtime priority", NULL}, | |
8808117c | 588 | [RLIMIT_RTTIME] = {"Max realtime timeout", "us"}, |
d85f50d5 NH |
589 | }; |
590 | ||
591 | /* Display limits for a process */ | |
1c963eb1 AD |
592 | static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns, |
593 | struct pid *pid, struct task_struct *task) | |
d85f50d5 NH |
594 | { |
595 | unsigned int i; | |
d85f50d5 | 596 | unsigned long flags; |
d85f50d5 NH |
597 | |
598 | struct rlimit rlim[RLIM_NLIMITS]; | |
599 | ||
a6bebbc8 | 600 | if (!lock_task_sighand(task, &flags)) |
d85f50d5 | 601 | return 0; |
d85f50d5 NH |
602 | memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS); |
603 | unlock_task_sighand(task, &flags); | |
d85f50d5 NH |
604 | |
605 | /* | |
606 | * print the file header | |
607 | */ | |
afe922c2 AD |
608 | seq_puts(m, "Limit " |
609 | "Soft Limit " | |
610 | "Hard Limit " | |
611 | "Units \n"); | |
d85f50d5 NH |
612 | |
613 | for (i = 0; i < RLIM_NLIMITS; i++) { | |
614 | if (rlim[i].rlim_cur == RLIM_INFINITY) | |
1c963eb1 | 615 | seq_printf(m, "%-25s %-20s ", |
25ce3191 | 616 | lnames[i].name, "unlimited"); |
d85f50d5 | 617 | else |
1c963eb1 | 618 | seq_printf(m, "%-25s %-20lu ", |
25ce3191 | 619 | lnames[i].name, rlim[i].rlim_cur); |
d85f50d5 NH |
620 | |
621 | if (rlim[i].rlim_max == RLIM_INFINITY) | |
1c963eb1 | 622 | seq_printf(m, "%-20s ", "unlimited"); |
d85f50d5 | 623 | else |
1c963eb1 | 624 | seq_printf(m, "%-20lu ", rlim[i].rlim_max); |
d85f50d5 NH |
625 | |
626 | if (lnames[i].unit) | |
1c963eb1 | 627 | seq_printf(m, "%-10s\n", lnames[i].unit); |
d85f50d5 | 628 | else |
1c963eb1 | 629 | seq_putc(m, '\n'); |
d85f50d5 NH |
630 | } |
631 | ||
1c963eb1 | 632 | return 0; |
d85f50d5 NH |
633 | } |
634 | ||
ebcb6734 | 635 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 AD |
636 | static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns, |
637 | struct pid *pid, struct task_struct *task) | |
ebcb6734 | 638 | { |
631b7aba SRRH |
639 | struct syscall_info info; |
640 | u64 *args = &info.data.args[0]; | |
25ce3191 JP |
641 | int res; |
642 | ||
643 | res = lock_trace(task); | |
a9712bc1 AV |
644 | if (res) |
645 | return res; | |
ebcb6734 | 646 | |
631b7aba | 647 | if (task_current_syscall(task, &info)) |
09d93bd6 | 648 | seq_puts(m, "running\n"); |
631b7aba SRRH |
649 | else if (info.data.nr < 0) |
650 | seq_printf(m, "%d 0x%llx 0x%llx\n", | |
651 | info.data.nr, info.sp, info.data.instruction_pointer); | |
a9712bc1 | 652 | else |
09d93bd6 | 653 | seq_printf(m, |
631b7aba SRRH |
654 | "%d 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx\n", |
655 | info.data.nr, | |
ebcb6734 | 656 | args[0], args[1], args[2], args[3], args[4], args[5], |
631b7aba | 657 | info.sp, info.data.instruction_pointer); |
a9712bc1 | 658 | unlock_trace(task); |
25ce3191 JP |
659 | |
660 | return 0; | |
ebcb6734 RM |
661 | } |
662 | #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */ | |
663 | ||
1da177e4 LT |
664 | /************************************************************************/ |
665 | /* Here the fs part begins */ | |
666 | /************************************************************************/ | |
667 | ||
668 | /* permission checks */ | |
778c1144 | 669 | static int proc_fd_access_allowed(struct inode *inode) |
1da177e4 | 670 | { |
778c1144 EB |
671 | struct task_struct *task; |
672 | int allowed = 0; | |
df26c40e EB |
673 | /* Allow access to a task's file descriptors if it is us or we |
674 | * may use ptrace attach to the process and find out that | |
675 | * information. | |
778c1144 EB |
676 | */ |
677 | task = get_proc_task(inode); | |
df26c40e | 678 | if (task) { |
caaee623 | 679 | allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS); |
778c1144 | 680 | put_task_struct(task); |
df26c40e | 681 | } |
778c1144 | 682 | return allowed; |
1da177e4 LT |
683 | } |
684 | ||
549c7297 CB |
685 | int proc_setattr(struct user_namespace *mnt_userns, struct dentry *dentry, |
686 | struct iattr *attr) | |
6d76fa58 LT |
687 | { |
688 | int error; | |
2b0143b5 | 689 | struct inode *inode = d_inode(dentry); |
6d76fa58 LT |
690 | |
691 | if (attr->ia_valid & ATTR_MODE) | |
692 | return -EPERM; | |
693 | ||
2f221d6f | 694 | error = setattr_prepare(&init_user_ns, dentry, attr); |
1025774c CH |
695 | if (error) |
696 | return error; | |
697 | ||
2f221d6f | 698 | setattr_copy(&init_user_ns, inode, attr); |
1025774c CH |
699 | mark_inode_dirty(inode); |
700 | return 0; | |
6d76fa58 LT |
701 | } |
702 | ||
0499680a VK |
703 | /* |
704 | * May current process learn task's sched/cmdline info (for hide_pid_min=1) | |
705 | * or euid/egid (for hide_pid_min=2)? | |
706 | */ | |
fa10fed3 | 707 | static bool has_pid_permissions(struct proc_fs_info *fs_info, |
0499680a | 708 | struct task_struct *task, |
e61bb8b3 | 709 | enum proc_hidepid hide_pid_min) |
0499680a | 710 | { |
24a71ce5 AG |
711 | /* |
712 | * If 'hidpid' mount option is set force a ptrace check, | |
713 | * we indicate that we are using a filesystem syscall | |
714 | * by passing PTRACE_MODE_READ_FSCREDS | |
715 | */ | |
716 | if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE) | |
717 | return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS); | |
718 | ||
fa10fed3 | 719 | if (fs_info->hide_pid < hide_pid_min) |
0499680a | 720 | return true; |
fa10fed3 | 721 | if (in_group_p(fs_info->pid_gid)) |
0499680a | 722 | return true; |
caaee623 | 723 | return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS); |
0499680a VK |
724 | } |
725 | ||
726 | ||
549c7297 CB |
727 | static int proc_pid_permission(struct user_namespace *mnt_userns, |
728 | struct inode *inode, int mask) | |
0499680a | 729 | { |
fa10fed3 | 730 | struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb); |
0499680a VK |
731 | struct task_struct *task; |
732 | bool has_perms; | |
733 | ||
734 | task = get_proc_task(inode); | |
a2ef990a XF |
735 | if (!task) |
736 | return -ESRCH; | |
fa10fed3 | 737 | has_perms = has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS); |
0499680a VK |
738 | put_task_struct(task); |
739 | ||
740 | if (!has_perms) { | |
fa10fed3 | 741 | if (fs_info->hide_pid == HIDEPID_INVISIBLE) { |
0499680a VK |
742 | /* |
743 | * Let's make getdents(), stat(), and open() | |
744 | * consistent with each other. If a process | |
745 | * may not stat() a file, it shouldn't be seen | |
746 | * in procfs at all. | |
747 | */ | |
748 | return -ENOENT; | |
749 | } | |
750 | ||
751 | return -EPERM; | |
752 | } | |
47291baa | 753 | return generic_permission(&init_user_ns, inode, mask); |
0499680a VK |
754 | } |
755 | ||
756 | ||
757 | ||
c5ef1c42 | 758 | static const struct inode_operations proc_def_inode_operations = { |
6d76fa58 LT |
759 | .setattr = proc_setattr, |
760 | }; | |
761 | ||
be614086 EB |
762 | static int proc_single_show(struct seq_file *m, void *v) |
763 | { | |
764 | struct inode *inode = m->private; | |
9d78edea | 765 | struct pid_namespace *ns = proc_pid_ns(inode->i_sb); |
76f668be | 766 | struct pid *pid = proc_pid(inode); |
be614086 EB |
767 | struct task_struct *task; |
768 | int ret; | |
769 | ||
be614086 EB |
770 | task = get_pid_task(pid, PIDTYPE_PID); |
771 | if (!task) | |
772 | return -ESRCH; | |
773 | ||
774 | ret = PROC_I(inode)->op.proc_show(m, ns, pid, task); | |
775 | ||
776 | put_task_struct(task); | |
777 | return ret; | |
778 | } | |
779 | ||
780 | static int proc_single_open(struct inode *inode, struct file *filp) | |
781 | { | |
c6a34058 | 782 | return single_open(filp, proc_single_show, inode); |
be614086 EB |
783 | } |
784 | ||
785 | static const struct file_operations proc_single_file_operations = { | |
786 | .open = proc_single_open, | |
787 | .read = seq_read, | |
788 | .llseek = seq_lseek, | |
789 | .release = single_release, | |
790 | }; | |
791 | ||
5381e169 ON |
792 | |
793 | struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode) | |
1da177e4 | 794 | { |
5381e169 ON |
795 | struct task_struct *task = get_proc_task(inode); |
796 | struct mm_struct *mm = ERR_PTR(-ESRCH); | |
e268337d | 797 | |
5381e169 | 798 | if (task) { |
caaee623 | 799 | mm = mm_access(task, mode | PTRACE_MODE_FSCREDS); |
5381e169 | 800 | put_task_struct(task); |
e268337d | 801 | |
5381e169 ON |
802 | if (!IS_ERR_OR_NULL(mm)) { |
803 | /* ensure this mm_struct can't be freed */ | |
f1f10076 | 804 | mmgrab(mm); |
5381e169 ON |
805 | /* but do not pin its memory */ |
806 | mmput(mm); | |
807 | } | |
808 | } | |
809 | ||
810 | return mm; | |
811 | } | |
812 | ||
813 | static int __mem_open(struct inode *inode, struct file *file, unsigned int mode) | |
814 | { | |
815 | struct mm_struct *mm = proc_mem_open(inode, mode); | |
e268337d LT |
816 | |
817 | if (IS_ERR(mm)) | |
818 | return PTR_ERR(mm); | |
819 | ||
e268337d | 820 | file->private_data = mm; |
1da177e4 LT |
821 | return 0; |
822 | } | |
823 | ||
b409e578 CW |
824 | static int mem_open(struct inode *inode, struct file *file) |
825 | { | |
bc452b4b DH |
826 | int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH); |
827 | ||
828 | /* OK to pass negative loff_t, we can catch out-of-range */ | |
829 | file->f_mode |= FMODE_UNSIGNED_OFFSET; | |
830 | ||
831 | return ret; | |
b409e578 CW |
832 | } |
833 | ||
572d34b9 ON |
834 | static ssize_t mem_rw(struct file *file, char __user *buf, |
835 | size_t count, loff_t *ppos, int write) | |
1da177e4 | 836 | { |
e268337d | 837 | struct mm_struct *mm = file->private_data; |
572d34b9 ON |
838 | unsigned long addr = *ppos; |
839 | ssize_t copied; | |
1da177e4 | 840 | char *page; |
272ddc8b | 841 | unsigned int flags; |
1da177e4 | 842 | |
e268337d LT |
843 | if (!mm) |
844 | return 0; | |
99f89551 | 845 | |
0ee931c4 | 846 | page = (char *)__get_free_page(GFP_KERNEL); |
30cd8903 | 847 | if (!page) |
e268337d | 848 | return -ENOMEM; |
1da177e4 | 849 | |
f7ca54f4 | 850 | copied = 0; |
388f7934 | 851 | if (!mmget_not_zero(mm)) |
6d08f2c7 ON |
852 | goto free; |
853 | ||
f511c0b1 | 854 | flags = FOLL_FORCE | (write ? FOLL_WRITE : 0); |
6347e8d5 | 855 | |
1da177e4 | 856 | while (count > 0) { |
572d34b9 | 857 | int this_len = min_t(int, count, PAGE_SIZE); |
1da177e4 | 858 | |
572d34b9 | 859 | if (write && copy_from_user(page, buf, this_len)) { |
1da177e4 LT |
860 | copied = -EFAULT; |
861 | break; | |
862 | } | |
572d34b9 | 863 | |
6347e8d5 | 864 | this_len = access_remote_vm(mm, addr, page, this_len, flags); |
572d34b9 | 865 | if (!this_len) { |
1da177e4 LT |
866 | if (!copied) |
867 | copied = -EIO; | |
868 | break; | |
869 | } | |
572d34b9 ON |
870 | |
871 | if (!write && copy_to_user(buf, page, this_len)) { | |
872 | copied = -EFAULT; | |
873 | break; | |
874 | } | |
875 | ||
876 | buf += this_len; | |
877 | addr += this_len; | |
878 | copied += this_len; | |
879 | count -= this_len; | |
1da177e4 | 880 | } |
572d34b9 | 881 | *ppos = addr; |
30cd8903 | 882 | |
6d08f2c7 ON |
883 | mmput(mm); |
884 | free: | |
30cd8903 | 885 | free_page((unsigned long) page); |
1da177e4 LT |
886 | return copied; |
887 | } | |
1da177e4 | 888 | |
572d34b9 ON |
889 | static ssize_t mem_read(struct file *file, char __user *buf, |
890 | size_t count, loff_t *ppos) | |
891 | { | |
892 | return mem_rw(file, buf, count, ppos, 0); | |
893 | } | |
894 | ||
895 | static ssize_t mem_write(struct file *file, const char __user *buf, | |
896 | size_t count, loff_t *ppos) | |
897 | { | |
898 | return mem_rw(file, (char __user*)buf, count, ppos, 1); | |
899 | } | |
900 | ||
85863e47 | 901 | loff_t mem_lseek(struct file *file, loff_t offset, int orig) |
1da177e4 LT |
902 | { |
903 | switch (orig) { | |
904 | case 0: | |
905 | file->f_pos = offset; | |
906 | break; | |
907 | case 1: | |
908 | file->f_pos += offset; | |
909 | break; | |
910 | default: | |
911 | return -EINVAL; | |
912 | } | |
913 | force_successful_syscall_return(); | |
914 | return file->f_pos; | |
915 | } | |
916 | ||
e268337d LT |
917 | static int mem_release(struct inode *inode, struct file *file) |
918 | { | |
919 | struct mm_struct *mm = file->private_data; | |
71879d3c | 920 | if (mm) |
6d08f2c7 | 921 | mmdrop(mm); |
e268337d LT |
922 | return 0; |
923 | } | |
924 | ||
00977a59 | 925 | static const struct file_operations proc_mem_operations = { |
1da177e4 LT |
926 | .llseek = mem_lseek, |
927 | .read = mem_read, | |
928 | .write = mem_write, | |
929 | .open = mem_open, | |
e268337d | 930 | .release = mem_release, |
1da177e4 LT |
931 | }; |
932 | ||
b409e578 CW |
933 | static int environ_open(struct inode *inode, struct file *file) |
934 | { | |
935 | return __mem_open(inode, file, PTRACE_MODE_READ); | |
936 | } | |
937 | ||
315e28c8 JP |
938 | static ssize_t environ_read(struct file *file, char __user *buf, |
939 | size_t count, loff_t *ppos) | |
940 | { | |
315e28c8 JP |
941 | char *page; |
942 | unsigned long src = *ppos; | |
b409e578 CW |
943 | int ret = 0; |
944 | struct mm_struct *mm = file->private_data; | |
a3b609ef | 945 | unsigned long env_start, env_end; |
315e28c8 | 946 | |
8148a73c MK |
947 | /* Ensure the process spawned far enough to have an environment. */ |
948 | if (!mm || !mm->env_end) | |
b409e578 | 949 | return 0; |
315e28c8 | 950 | |
0ee931c4 | 951 | page = (char *)__get_free_page(GFP_KERNEL); |
315e28c8 | 952 | if (!page) |
b409e578 | 953 | return -ENOMEM; |
315e28c8 | 954 | |
d6f64b89 | 955 | ret = 0; |
388f7934 | 956 | if (!mmget_not_zero(mm)) |
b409e578 | 957 | goto free; |
a3b609ef | 958 | |
88aa7cc6 | 959 | spin_lock(&mm->arg_lock); |
a3b609ef MG |
960 | env_start = mm->env_start; |
961 | env_end = mm->env_end; | |
88aa7cc6 | 962 | spin_unlock(&mm->arg_lock); |
a3b609ef | 963 | |
315e28c8 | 964 | while (count > 0) { |
e8905ec2 DH |
965 | size_t this_len, max_len; |
966 | int retval; | |
315e28c8 | 967 | |
a3b609ef | 968 | if (src >= (env_end - env_start)) |
315e28c8 JP |
969 | break; |
970 | ||
a3b609ef | 971 | this_len = env_end - (env_start + src); |
e8905ec2 DH |
972 | |
973 | max_len = min_t(size_t, PAGE_SIZE, count); | |
974 | this_len = min(max_len, this_len); | |
315e28c8 | 975 | |
7f7ccc2c | 976 | retval = access_remote_vm(mm, (env_start + src), page, this_len, FOLL_ANON); |
315e28c8 JP |
977 | |
978 | if (retval <= 0) { | |
979 | ret = retval; | |
980 | break; | |
981 | } | |
982 | ||
983 | if (copy_to_user(buf, page, retval)) { | |
984 | ret = -EFAULT; | |
985 | break; | |
986 | } | |
987 | ||
988 | ret += retval; | |
989 | src += retval; | |
990 | buf += retval; | |
991 | count -= retval; | |
992 | } | |
993 | *ppos = src; | |
315e28c8 | 994 | mmput(mm); |
b409e578 CW |
995 | |
996 | free: | |
315e28c8 | 997 | free_page((unsigned long) page); |
315e28c8 JP |
998 | return ret; |
999 | } | |
1000 | ||
1001 | static const struct file_operations proc_environ_operations = { | |
b409e578 | 1002 | .open = environ_open, |
315e28c8 | 1003 | .read = environ_read, |
87df8424 | 1004 | .llseek = generic_file_llseek, |
b409e578 | 1005 | .release = mem_release, |
315e28c8 JP |
1006 | }; |
1007 | ||
c5317167 AV |
1008 | static int auxv_open(struct inode *inode, struct file *file) |
1009 | { | |
1010 | return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS); | |
1011 | } | |
1012 | ||
1013 | static ssize_t auxv_read(struct file *file, char __user *buf, | |
1014 | size_t count, loff_t *ppos) | |
1015 | { | |
1016 | struct mm_struct *mm = file->private_data; | |
1017 | unsigned int nwords = 0; | |
06b2849d LY |
1018 | |
1019 | if (!mm) | |
1020 | return 0; | |
c5317167 AV |
1021 | do { |
1022 | nwords += 2; | |
1023 | } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */ | |
1024 | return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv, | |
1025 | nwords * sizeof(mm->saved_auxv[0])); | |
1026 | } | |
1027 | ||
1028 | static const struct file_operations proc_auxv_operations = { | |
1029 | .open = auxv_open, | |
1030 | .read = auxv_read, | |
1031 | .llseek = generic_file_llseek, | |
1032 | .release = mem_release, | |
1033 | }; | |
1034 | ||
fa0cbbf1 DR |
1035 | static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count, |
1036 | loff_t *ppos) | |
1037 | { | |
496ad9aa | 1038 | struct task_struct *task = get_proc_task(file_inode(file)); |
fa0cbbf1 DR |
1039 | char buffer[PROC_NUMBUF]; |
1040 | int oom_adj = OOM_ADJUST_MIN; | |
1041 | size_t len; | |
fa0cbbf1 DR |
1042 | |
1043 | if (!task) | |
1044 | return -ESRCH; | |
f913da59 MH |
1045 | if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX) |
1046 | oom_adj = OOM_ADJUST_MAX; | |
1047 | else | |
1048 | oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) / | |
1049 | OOM_SCORE_ADJ_MAX; | |
fa0cbbf1 | 1050 | put_task_struct(task); |
66606567 CH |
1051 | if (oom_adj > OOM_ADJUST_MAX) |
1052 | oom_adj = OOM_ADJUST_MAX; | |
fa0cbbf1 DR |
1053 | len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj); |
1054 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
1055 | } | |
1056 | ||
1d5f0acb MH |
1057 | static int __set_oom_adj(struct file *file, int oom_adj, bool legacy) |
1058 | { | |
44a70ade | 1059 | struct mm_struct *mm = NULL; |
1d5f0acb MH |
1060 | struct task_struct *task; |
1061 | int err = 0; | |
1062 | ||
1063 | task = get_proc_task(file_inode(file)); | |
1064 | if (!task) | |
1065 | return -ESRCH; | |
1066 | ||
1067 | mutex_lock(&oom_adj_mutex); | |
1068 | if (legacy) { | |
1069 | if (oom_adj < task->signal->oom_score_adj && | |
1070 | !capable(CAP_SYS_RESOURCE)) { | |
1071 | err = -EACCES; | |
1072 | goto err_unlock; | |
1073 | } | |
1074 | /* | |
1075 | * /proc/pid/oom_adj is provided for legacy purposes, ask users to use | |
1076 | * /proc/pid/oom_score_adj instead. | |
1077 | */ | |
1078 | pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n", | |
1079 | current->comm, task_pid_nr(current), task_pid_nr(task), | |
1080 | task_pid_nr(task)); | |
1081 | } else { | |
1082 | if ((short)oom_adj < task->signal->oom_score_adj_min && | |
1083 | !capable(CAP_SYS_RESOURCE)) { | |
1084 | err = -EACCES; | |
1085 | goto err_unlock; | |
1086 | } | |
1087 | } | |
1088 | ||
44a70ade MH |
1089 | /* |
1090 | * Make sure we will check other processes sharing the mm if this is | |
1091 | * not vfrok which wants its own oom_score_adj. | |
1092 | * pin the mm so it doesn't go away and get reused after task_unlock | |
1093 | */ | |
1094 | if (!task->vfork_done) { | |
1095 | struct task_struct *p = find_lock_task_mm(task); | |
1096 | ||
1097 | if (p) { | |
67197a4f | 1098 | if (test_bit(MMF_MULTIPROCESS, &p->mm->flags)) { |
44a70ade | 1099 | mm = p->mm; |
f1f10076 | 1100 | mmgrab(mm); |
44a70ade MH |
1101 | } |
1102 | task_unlock(p); | |
1103 | } | |
1104 | } | |
1105 | ||
1d5f0acb MH |
1106 | task->signal->oom_score_adj = oom_adj; |
1107 | if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE)) | |
1108 | task->signal->oom_score_adj_min = (short)oom_adj; | |
1109 | trace_oom_score_adj_update(task); | |
44a70ade MH |
1110 | |
1111 | if (mm) { | |
1112 | struct task_struct *p; | |
1113 | ||
1114 | rcu_read_lock(); | |
1115 | for_each_process(p) { | |
1116 | if (same_thread_group(task, p)) | |
1117 | continue; | |
1118 | ||
1119 | /* do not touch kernel threads or the global init */ | |
1120 | if (p->flags & PF_KTHREAD || is_global_init(p)) | |
1121 | continue; | |
1122 | ||
1123 | task_lock(p); | |
1124 | if (!p->vfork_done && process_shares_mm(p, mm)) { | |
44a70ade MH |
1125 | p->signal->oom_score_adj = oom_adj; |
1126 | if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE)) | |
1127 | p->signal->oom_score_adj_min = (short)oom_adj; | |
1128 | } | |
1129 | task_unlock(p); | |
1130 | } | |
1131 | rcu_read_unlock(); | |
1132 | mmdrop(mm); | |
1133 | } | |
1d5f0acb MH |
1134 | err_unlock: |
1135 | mutex_unlock(&oom_adj_mutex); | |
1136 | put_task_struct(task); | |
1137 | return err; | |
1138 | } | |
f913da59 | 1139 | |
b72bdfa7 DR |
1140 | /* |
1141 | * /proc/pid/oom_adj exists solely for backwards compatibility with previous | |
1142 | * kernels. The effective policy is defined by oom_score_adj, which has a | |
1143 | * different scale: oom_adj grew exponentially and oom_score_adj grows linearly. | |
1144 | * Values written to oom_adj are simply mapped linearly to oom_score_adj. | |
1145 | * Processes that become oom disabled via oom_adj will still be oom disabled | |
1146 | * with this implementation. | |
1147 | * | |
1148 | * oom_adj cannot be removed since existing userspace binaries use it. | |
1149 | */ | |
fa0cbbf1 DR |
1150 | static ssize_t oom_adj_write(struct file *file, const char __user *buf, |
1151 | size_t count, loff_t *ppos) | |
1152 | { | |
fa0cbbf1 DR |
1153 | char buffer[PROC_NUMBUF]; |
1154 | int oom_adj; | |
fa0cbbf1 DR |
1155 | int err; |
1156 | ||
1157 | memset(buffer, 0, sizeof(buffer)); | |
1158 | if (count > sizeof(buffer) - 1) | |
1159 | count = sizeof(buffer) - 1; | |
1160 | if (copy_from_user(buffer, buf, count)) { | |
1161 | err = -EFAULT; | |
1162 | goto out; | |
1163 | } | |
1164 | ||
1165 | err = kstrtoint(strstrip(buffer), 0, &oom_adj); | |
1166 | if (err) | |
1167 | goto out; | |
1168 | if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) && | |
1169 | oom_adj != OOM_DISABLE) { | |
1170 | err = -EINVAL; | |
1171 | goto out; | |
1172 | } | |
1173 | ||
fa0cbbf1 DR |
1174 | /* |
1175 | * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum | |
1176 | * value is always attainable. | |
1177 | */ | |
1178 | if (oom_adj == OOM_ADJUST_MAX) | |
1179 | oom_adj = OOM_SCORE_ADJ_MAX; | |
1180 | else | |
1181 | oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE; | |
1182 | ||
1d5f0acb | 1183 | err = __set_oom_adj(file, oom_adj, true); |
fa0cbbf1 DR |
1184 | out: |
1185 | return err < 0 ? err : count; | |
1186 | } | |
1187 | ||
1188 | static const struct file_operations proc_oom_adj_operations = { | |
1189 | .read = oom_adj_read, | |
1190 | .write = oom_adj_write, | |
1191 | .llseek = generic_file_llseek, | |
1192 | }; | |
1193 | ||
a63d83f4 DR |
1194 | static ssize_t oom_score_adj_read(struct file *file, char __user *buf, |
1195 | size_t count, loff_t *ppos) | |
1196 | { | |
496ad9aa | 1197 | struct task_struct *task = get_proc_task(file_inode(file)); |
a63d83f4 | 1198 | char buffer[PROC_NUMBUF]; |
a9c58b90 | 1199 | short oom_score_adj = OOM_SCORE_ADJ_MIN; |
a63d83f4 DR |
1200 | size_t len; |
1201 | ||
1202 | if (!task) | |
1203 | return -ESRCH; | |
f913da59 | 1204 | oom_score_adj = task->signal->oom_score_adj; |
a63d83f4 | 1205 | put_task_struct(task); |
a9c58b90 | 1206 | len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj); |
a63d83f4 DR |
1207 | return simple_read_from_buffer(buf, count, ppos, buffer, len); |
1208 | } | |
1209 | ||
1210 | static ssize_t oom_score_adj_write(struct file *file, const char __user *buf, | |
1211 | size_t count, loff_t *ppos) | |
1212 | { | |
a63d83f4 | 1213 | char buffer[PROC_NUMBUF]; |
0a8cb8e3 | 1214 | int oom_score_adj; |
a63d83f4 DR |
1215 | int err; |
1216 | ||
1217 | memset(buffer, 0, sizeof(buffer)); | |
1218 | if (count > sizeof(buffer) - 1) | |
1219 | count = sizeof(buffer) - 1; | |
723548bf DR |
1220 | if (copy_from_user(buffer, buf, count)) { |
1221 | err = -EFAULT; | |
1222 | goto out; | |
1223 | } | |
a63d83f4 | 1224 | |
0a8cb8e3 | 1225 | err = kstrtoint(strstrip(buffer), 0, &oom_score_adj); |
a63d83f4 | 1226 | if (err) |
723548bf | 1227 | goto out; |
a63d83f4 | 1228 | if (oom_score_adj < OOM_SCORE_ADJ_MIN || |
723548bf DR |
1229 | oom_score_adj > OOM_SCORE_ADJ_MAX) { |
1230 | err = -EINVAL; | |
1231 | goto out; | |
1232 | } | |
a63d83f4 | 1233 | |
1d5f0acb | 1234 | err = __set_oom_adj(file, oom_score_adj, false); |
723548bf DR |
1235 | out: |
1236 | return err < 0 ? err : count; | |
a63d83f4 DR |
1237 | } |
1238 | ||
1239 | static const struct file_operations proc_oom_score_adj_operations = { | |
1240 | .read = oom_score_adj_read, | |
1241 | .write = oom_score_adj_write, | |
6038f373 | 1242 | .llseek = default_llseek, |
a63d83f4 DR |
1243 | }; |
1244 | ||
4b7d248b | 1245 | #ifdef CONFIG_AUDIT |
b4eb4f7f | 1246 | #define TMPBUFLEN 11 |
1da177e4 LT |
1247 | static ssize_t proc_loginuid_read(struct file * file, char __user * buf, |
1248 | size_t count, loff_t *ppos) | |
1249 | { | |
496ad9aa | 1250 | struct inode * inode = file_inode(file); |
99f89551 | 1251 | struct task_struct *task = get_proc_task(inode); |
1da177e4 LT |
1252 | ssize_t length; |
1253 | char tmpbuf[TMPBUFLEN]; | |
1254 | ||
99f89551 EB |
1255 | if (!task) |
1256 | return -ESRCH; | |
1da177e4 | 1257 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", |
e1760bd5 EB |
1258 | from_kuid(file->f_cred->user_ns, |
1259 | audit_get_loginuid(task))); | |
99f89551 | 1260 | put_task_struct(task); |
1da177e4 LT |
1261 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); |
1262 | } | |
1263 | ||
1264 | static ssize_t proc_loginuid_write(struct file * file, const char __user * buf, | |
1265 | size_t count, loff_t *ppos) | |
1266 | { | |
496ad9aa | 1267 | struct inode * inode = file_inode(file); |
1da177e4 | 1268 | uid_t loginuid; |
e1760bd5 | 1269 | kuid_t kloginuid; |
774636e1 | 1270 | int rv; |
1da177e4 | 1271 | |
4ea33a97 JA |
1272 | /* Don't let kthreads write their own loginuid */ |
1273 | if (current->flags & PF_KTHREAD) | |
1274 | return -EPERM; | |
1275 | ||
7dc52157 PM |
1276 | rcu_read_lock(); |
1277 | if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) { | |
1278 | rcu_read_unlock(); | |
1da177e4 | 1279 | return -EPERM; |
7dc52157 PM |
1280 | } |
1281 | rcu_read_unlock(); | |
1da177e4 | 1282 | |
1da177e4 LT |
1283 | if (*ppos != 0) { |
1284 | /* No partial writes. */ | |
1285 | return -EINVAL; | |
1286 | } | |
1da177e4 | 1287 | |
774636e1 AD |
1288 | rv = kstrtou32_from_user(buf, count, 10, &loginuid); |
1289 | if (rv < 0) | |
1290 | return rv; | |
81407c84 EP |
1291 | |
1292 | /* is userspace tring to explicitly UNSET the loginuid? */ | |
1293 | if (loginuid == AUDIT_UID_UNSET) { | |
1294 | kloginuid = INVALID_UID; | |
1295 | } else { | |
1296 | kloginuid = make_kuid(file->f_cred->user_ns, loginuid); | |
774636e1 AD |
1297 | if (!uid_valid(kloginuid)) |
1298 | return -EINVAL; | |
e1760bd5 EB |
1299 | } |
1300 | ||
774636e1 AD |
1301 | rv = audit_set_loginuid(kloginuid); |
1302 | if (rv < 0) | |
1303 | return rv; | |
1304 | return count; | |
1da177e4 LT |
1305 | } |
1306 | ||
00977a59 | 1307 | static const struct file_operations proc_loginuid_operations = { |
1da177e4 LT |
1308 | .read = proc_loginuid_read, |
1309 | .write = proc_loginuid_write, | |
87df8424 | 1310 | .llseek = generic_file_llseek, |
1da177e4 | 1311 | }; |
1e0bd755 EP |
1312 | |
1313 | static ssize_t proc_sessionid_read(struct file * file, char __user * buf, | |
1314 | size_t count, loff_t *ppos) | |
1315 | { | |
496ad9aa | 1316 | struct inode * inode = file_inode(file); |
1e0bd755 EP |
1317 | struct task_struct *task = get_proc_task(inode); |
1318 | ssize_t length; | |
1319 | char tmpbuf[TMPBUFLEN]; | |
1320 | ||
1321 | if (!task) | |
1322 | return -ESRCH; | |
1323 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", | |
1324 | audit_get_sessionid(task)); | |
1325 | put_task_struct(task); | |
1326 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); | |
1327 | } | |
1328 | ||
1329 | static const struct file_operations proc_sessionid_operations = { | |
1330 | .read = proc_sessionid_read, | |
87df8424 | 1331 | .llseek = generic_file_llseek, |
1e0bd755 | 1332 | }; |
1da177e4 LT |
1333 | #endif |
1334 | ||
f4f154fd AM |
1335 | #ifdef CONFIG_FAULT_INJECTION |
1336 | static ssize_t proc_fault_inject_read(struct file * file, char __user * buf, | |
1337 | size_t count, loff_t *ppos) | |
1338 | { | |
496ad9aa | 1339 | struct task_struct *task = get_proc_task(file_inode(file)); |
f4f154fd AM |
1340 | char buffer[PROC_NUMBUF]; |
1341 | size_t len; | |
1342 | int make_it_fail; | |
f4f154fd AM |
1343 | |
1344 | if (!task) | |
1345 | return -ESRCH; | |
1346 | make_it_fail = task->make_it_fail; | |
1347 | put_task_struct(task); | |
1348 | ||
1349 | len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail); | |
0c28f287 AM |
1350 | |
1351 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
f4f154fd AM |
1352 | } |
1353 | ||
1354 | static ssize_t proc_fault_inject_write(struct file * file, | |
1355 | const char __user * buf, size_t count, loff_t *ppos) | |
1356 | { | |
1357 | struct task_struct *task; | |
774636e1 | 1358 | char buffer[PROC_NUMBUF]; |
f4f154fd | 1359 | int make_it_fail; |
774636e1 | 1360 | int rv; |
f4f154fd AM |
1361 | |
1362 | if (!capable(CAP_SYS_RESOURCE)) | |
1363 | return -EPERM; | |
1364 | memset(buffer, 0, sizeof(buffer)); | |
1365 | if (count > sizeof(buffer) - 1) | |
1366 | count = sizeof(buffer) - 1; | |
1367 | if (copy_from_user(buffer, buf, count)) | |
1368 | return -EFAULT; | |
774636e1 AD |
1369 | rv = kstrtoint(strstrip(buffer), 0, &make_it_fail); |
1370 | if (rv < 0) | |
1371 | return rv; | |
16caed31 DJ |
1372 | if (make_it_fail < 0 || make_it_fail > 1) |
1373 | return -EINVAL; | |
1374 | ||
496ad9aa | 1375 | task = get_proc_task(file_inode(file)); |
f4f154fd AM |
1376 | if (!task) |
1377 | return -ESRCH; | |
1378 | task->make_it_fail = make_it_fail; | |
1379 | put_task_struct(task); | |
cba8aafe VL |
1380 | |
1381 | return count; | |
f4f154fd AM |
1382 | } |
1383 | ||
00977a59 | 1384 | static const struct file_operations proc_fault_inject_operations = { |
f4f154fd AM |
1385 | .read = proc_fault_inject_read, |
1386 | .write = proc_fault_inject_write, | |
87df8424 | 1387 | .llseek = generic_file_llseek, |
f4f154fd | 1388 | }; |
e41d5818 DV |
1389 | |
1390 | static ssize_t proc_fail_nth_write(struct file *file, const char __user *buf, | |
1391 | size_t count, loff_t *ppos) | |
1392 | { | |
1393 | struct task_struct *task; | |
9049f2f6 AM |
1394 | int err; |
1395 | unsigned int n; | |
e41d5818 | 1396 | |
1203c8e6 AM |
1397 | err = kstrtouint_from_user(buf, count, 0, &n); |
1398 | if (err) | |
1399 | return err; | |
1400 | ||
e41d5818 DV |
1401 | task = get_proc_task(file_inode(file)); |
1402 | if (!task) | |
1403 | return -ESRCH; | |
9f7118b2 | 1404 | task->fail_nth = n; |
e41d5818 | 1405 | put_task_struct(task); |
1203c8e6 | 1406 | |
e41d5818 DV |
1407 | return count; |
1408 | } | |
1409 | ||
1410 | static ssize_t proc_fail_nth_read(struct file *file, char __user *buf, | |
1411 | size_t count, loff_t *ppos) | |
1412 | { | |
1413 | struct task_struct *task; | |
bfc74093 AM |
1414 | char numbuf[PROC_NUMBUF]; |
1415 | ssize_t len; | |
e41d5818 DV |
1416 | |
1417 | task = get_proc_task(file_inode(file)); | |
1418 | if (!task) | |
1419 | return -ESRCH; | |
9f7118b2 | 1420 | len = snprintf(numbuf, sizeof(numbuf), "%u\n", task->fail_nth); |
1203c8e6 | 1421 | put_task_struct(task); |
a44937fe | 1422 | return simple_read_from_buffer(buf, count, ppos, numbuf, len); |
e41d5818 DV |
1423 | } |
1424 | ||
1425 | static const struct file_operations proc_fail_nth_operations = { | |
1426 | .read = proc_fail_nth_read, | |
1427 | .write = proc_fail_nth_write, | |
1428 | }; | |
f4f154fd AM |
1429 | #endif |
1430 | ||
9745512c | 1431 | |
43ae34cb IM |
1432 | #ifdef CONFIG_SCHED_DEBUG |
1433 | /* | |
1434 | * Print out various scheduling related per-task fields: | |
1435 | */ | |
1436 | static int sched_show(struct seq_file *m, void *v) | |
1437 | { | |
1438 | struct inode *inode = m->private; | |
9d78edea | 1439 | struct pid_namespace *ns = proc_pid_ns(inode->i_sb); |
43ae34cb IM |
1440 | struct task_struct *p; |
1441 | ||
43ae34cb IM |
1442 | p = get_proc_task(inode); |
1443 | if (!p) | |
1444 | return -ESRCH; | |
74dc3384 | 1445 | proc_sched_show_task(p, ns, m); |
43ae34cb IM |
1446 | |
1447 | put_task_struct(p); | |
1448 | ||
1449 | return 0; | |
1450 | } | |
1451 | ||
1452 | static ssize_t | |
1453 | sched_write(struct file *file, const char __user *buf, | |
1454 | size_t count, loff_t *offset) | |
1455 | { | |
496ad9aa | 1456 | struct inode *inode = file_inode(file); |
43ae34cb IM |
1457 | struct task_struct *p; |
1458 | ||
43ae34cb IM |
1459 | p = get_proc_task(inode); |
1460 | if (!p) | |
1461 | return -ESRCH; | |
1462 | proc_sched_set_task(p); | |
1463 | ||
1464 | put_task_struct(p); | |
1465 | ||
1466 | return count; | |
1467 | } | |
1468 | ||
1469 | static int sched_open(struct inode *inode, struct file *filp) | |
1470 | { | |
c6a34058 | 1471 | return single_open(filp, sched_show, inode); |
43ae34cb IM |
1472 | } |
1473 | ||
1474 | static const struct file_operations proc_pid_sched_operations = { | |
1475 | .open = sched_open, | |
1476 | .read = seq_read, | |
1477 | .write = sched_write, | |
1478 | .llseek = seq_lseek, | |
5ea473a1 | 1479 | .release = single_release, |
43ae34cb IM |
1480 | }; |
1481 | ||
1482 | #endif | |
1483 | ||
5091faa4 MG |
1484 | #ifdef CONFIG_SCHED_AUTOGROUP |
1485 | /* | |
1486 | * Print out autogroup related information: | |
1487 | */ | |
1488 | static int sched_autogroup_show(struct seq_file *m, void *v) | |
1489 | { | |
1490 | struct inode *inode = m->private; | |
1491 | struct task_struct *p; | |
1492 | ||
1493 | p = get_proc_task(inode); | |
1494 | if (!p) | |
1495 | return -ESRCH; | |
1496 | proc_sched_autogroup_show_task(p, m); | |
1497 | ||
1498 | put_task_struct(p); | |
1499 | ||
1500 | return 0; | |
1501 | } | |
1502 | ||
1503 | static ssize_t | |
1504 | sched_autogroup_write(struct file *file, const char __user *buf, | |
1505 | size_t count, loff_t *offset) | |
1506 | { | |
496ad9aa | 1507 | struct inode *inode = file_inode(file); |
5091faa4 MG |
1508 | struct task_struct *p; |
1509 | char buffer[PROC_NUMBUF]; | |
0a8cb8e3 | 1510 | int nice; |
5091faa4 MG |
1511 | int err; |
1512 | ||
1513 | memset(buffer, 0, sizeof(buffer)); | |
1514 | if (count > sizeof(buffer) - 1) | |
1515 | count = sizeof(buffer) - 1; | |
1516 | if (copy_from_user(buffer, buf, count)) | |
1517 | return -EFAULT; | |
1518 | ||
0a8cb8e3 AD |
1519 | err = kstrtoint(strstrip(buffer), 0, &nice); |
1520 | if (err < 0) | |
1521 | return err; | |
5091faa4 MG |
1522 | |
1523 | p = get_proc_task(inode); | |
1524 | if (!p) | |
1525 | return -ESRCH; | |
1526 | ||
2e5b5b3a | 1527 | err = proc_sched_autogroup_set_nice(p, nice); |
5091faa4 MG |
1528 | if (err) |
1529 | count = err; | |
1530 | ||
1531 | put_task_struct(p); | |
1532 | ||
1533 | return count; | |
1534 | } | |
1535 | ||
1536 | static int sched_autogroup_open(struct inode *inode, struct file *filp) | |
1537 | { | |
1538 | int ret; | |
1539 | ||
1540 | ret = single_open(filp, sched_autogroup_show, NULL); | |
1541 | if (!ret) { | |
1542 | struct seq_file *m = filp->private_data; | |
1543 | ||
1544 | m->private = inode; | |
1545 | } | |
1546 | return ret; | |
1547 | } | |
1548 | ||
1549 | static const struct file_operations proc_pid_sched_autogroup_operations = { | |
1550 | .open = sched_autogroup_open, | |
1551 | .read = seq_read, | |
1552 | .write = sched_autogroup_write, | |
1553 | .llseek = seq_lseek, | |
1554 | .release = single_release, | |
1555 | }; | |
1556 | ||
1557 | #endif /* CONFIG_SCHED_AUTOGROUP */ | |
1558 | ||
04a8682a AV |
1559 | #ifdef CONFIG_TIME_NS |
1560 | static int timens_offsets_show(struct seq_file *m, void *v) | |
1561 | { | |
1562 | struct task_struct *p; | |
1563 | ||
1564 | p = get_proc_task(file_inode(m->file)); | |
1565 | if (!p) | |
1566 | return -ESRCH; | |
1567 | proc_timens_show_offsets(p, m); | |
1568 | ||
1569 | put_task_struct(p); | |
1570 | ||
1571 | return 0; | |
1572 | } | |
1573 | ||
1574 | static ssize_t timens_offsets_write(struct file *file, const char __user *buf, | |
1575 | size_t count, loff_t *ppos) | |
1576 | { | |
1577 | struct inode *inode = file_inode(file); | |
1578 | struct proc_timens_offset offsets[2]; | |
1579 | char *kbuf = NULL, *pos, *next_line; | |
1580 | struct task_struct *p; | |
1581 | int ret, noffsets; | |
1582 | ||
1583 | /* Only allow < page size writes at the beginning of the file */ | |
1584 | if ((*ppos != 0) || (count >= PAGE_SIZE)) | |
1585 | return -EINVAL; | |
1586 | ||
1587 | /* Slurp in the user data */ | |
1588 | kbuf = memdup_user_nul(buf, count); | |
1589 | if (IS_ERR(kbuf)) | |
1590 | return PTR_ERR(kbuf); | |
1591 | ||
1592 | /* Parse the user data */ | |
1593 | ret = -EINVAL; | |
1594 | noffsets = 0; | |
1595 | for (pos = kbuf; pos; pos = next_line) { | |
1596 | struct proc_timens_offset *off = &offsets[noffsets]; | |
94d440d6 | 1597 | char clock[10]; |
04a8682a AV |
1598 | int err; |
1599 | ||
1600 | /* Find the end of line and ensure we don't look past it */ | |
1601 | next_line = strchr(pos, '\n'); | |
1602 | if (next_line) { | |
1603 | *next_line = '\0'; | |
1604 | next_line++; | |
1605 | if (*next_line == '\0') | |
1606 | next_line = NULL; | |
1607 | } | |
1608 | ||
94d440d6 | 1609 | err = sscanf(pos, "%9s %lld %lu", clock, |
04a8682a AV |
1610 | &off->val.tv_sec, &off->val.tv_nsec); |
1611 | if (err != 3 || off->val.tv_nsec >= NSEC_PER_SEC) | |
1612 | goto out; | |
94d440d6 AV |
1613 | |
1614 | clock[sizeof(clock) - 1] = 0; | |
1615 | if (strcmp(clock, "monotonic") == 0 || | |
1616 | strcmp(clock, __stringify(CLOCK_MONOTONIC)) == 0) | |
1617 | off->clockid = CLOCK_MONOTONIC; | |
1618 | else if (strcmp(clock, "boottime") == 0 || | |
1619 | strcmp(clock, __stringify(CLOCK_BOOTTIME)) == 0) | |
1620 | off->clockid = CLOCK_BOOTTIME; | |
1621 | else | |
1622 | goto out; | |
1623 | ||
04a8682a AV |
1624 | noffsets++; |
1625 | if (noffsets == ARRAY_SIZE(offsets)) { | |
1626 | if (next_line) | |
1627 | count = next_line - kbuf; | |
1628 | break; | |
1629 | } | |
1630 | } | |
1631 | ||
1632 | ret = -ESRCH; | |
1633 | p = get_proc_task(inode); | |
1634 | if (!p) | |
1635 | goto out; | |
1636 | ret = proc_timens_set_offset(file, p, offsets, noffsets); | |
1637 | put_task_struct(p); | |
1638 | if (ret) | |
1639 | goto out; | |
1640 | ||
1641 | ret = count; | |
1642 | out: | |
1643 | kfree(kbuf); | |
1644 | return ret; | |
1645 | } | |
1646 | ||
1647 | static int timens_offsets_open(struct inode *inode, struct file *filp) | |
1648 | { | |
1649 | return single_open(filp, timens_offsets_show, inode); | |
1650 | } | |
1651 | ||
1652 | static const struct file_operations proc_timens_offsets_operations = { | |
1653 | .open = timens_offsets_open, | |
1654 | .read = seq_read, | |
1655 | .write = timens_offsets_write, | |
1656 | .llseek = seq_lseek, | |
1657 | .release = single_release, | |
1658 | }; | |
1659 | #endif /* CONFIG_TIME_NS */ | |
1660 | ||
4614a696 | 1661 | static ssize_t comm_write(struct file *file, const char __user *buf, |
1662 | size_t count, loff_t *offset) | |
1663 | { | |
496ad9aa | 1664 | struct inode *inode = file_inode(file); |
4614a696 | 1665 | struct task_struct *p; |
1666 | char buffer[TASK_COMM_LEN]; | |
830e0fc9 | 1667 | const size_t maxlen = sizeof(buffer) - 1; |
4614a696 | 1668 | |
1669 | memset(buffer, 0, sizeof(buffer)); | |
830e0fc9 | 1670 | if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count)) |
4614a696 | 1671 | return -EFAULT; |
1672 | ||
1673 | p = get_proc_task(inode); | |
1674 | if (!p) | |
1675 | return -ESRCH; | |
1676 | ||
1677 | if (same_thread_group(current, p)) | |
1678 | set_task_comm(p, buffer); | |
1679 | else | |
1680 | count = -EINVAL; | |
1681 | ||
1682 | put_task_struct(p); | |
1683 | ||
1684 | return count; | |
1685 | } | |
1686 | ||
1687 | static int comm_show(struct seq_file *m, void *v) | |
1688 | { | |
1689 | struct inode *inode = m->private; | |
1690 | struct task_struct *p; | |
1691 | ||
1692 | p = get_proc_task(inode); | |
1693 | if (!p) | |
1694 | return -ESRCH; | |
1695 | ||
88b72b31 TH |
1696 | proc_task_name(m, p, false); |
1697 | seq_putc(m, '\n'); | |
4614a696 | 1698 | |
1699 | put_task_struct(p); | |
1700 | ||
1701 | return 0; | |
1702 | } | |
1703 | ||
1704 | static int comm_open(struct inode *inode, struct file *filp) | |
1705 | { | |
c6a34058 | 1706 | return single_open(filp, comm_show, inode); |
4614a696 | 1707 | } |
1708 | ||
1709 | static const struct file_operations proc_pid_set_comm_operations = { | |
1710 | .open = comm_open, | |
1711 | .read = seq_read, | |
1712 | .write = comm_write, | |
1713 | .llseek = seq_lseek, | |
1714 | .release = single_release, | |
1715 | }; | |
1716 | ||
7773fbc5 | 1717 | static int proc_exe_link(struct dentry *dentry, struct path *exe_path) |
925d1c40 MH |
1718 | { |
1719 | struct task_struct *task; | |
925d1c40 MH |
1720 | struct file *exe_file; |
1721 | ||
2b0143b5 | 1722 | task = get_proc_task(d_inode(dentry)); |
925d1c40 MH |
1723 | if (!task) |
1724 | return -ENOENT; | |
cd81a917 | 1725 | exe_file = get_task_exe_file(task); |
925d1c40 | 1726 | put_task_struct(task); |
925d1c40 MH |
1727 | if (exe_file) { |
1728 | *exe_path = exe_file->f_path; | |
1729 | path_get(&exe_file->f_path); | |
1730 | fput(exe_file); | |
1731 | return 0; | |
1732 | } else | |
1733 | return -ENOENT; | |
1734 | } | |
1735 | ||
6b255391 | 1736 | static const char *proc_pid_get_link(struct dentry *dentry, |
fceef393 AV |
1737 | struct inode *inode, |
1738 | struct delayed_call *done) | |
1da177e4 | 1739 | { |
408ef013 | 1740 | struct path path; |
1da177e4 LT |
1741 | int error = -EACCES; |
1742 | ||
6b255391 AV |
1743 | if (!dentry) |
1744 | return ERR_PTR(-ECHILD); | |
1745 | ||
778c1144 EB |
1746 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1747 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1748 | goto out; |
1da177e4 | 1749 | |
408ef013 CH |
1750 | error = PROC_I(inode)->op.proc_get_link(dentry, &path); |
1751 | if (error) | |
1752 | goto out; | |
1753 | ||
1bc82070 | 1754 | error = nd_jump_link(&path); |
1da177e4 | 1755 | out: |
008b150a | 1756 | return ERR_PTR(error); |
1da177e4 LT |
1757 | } |
1758 | ||
3dcd25f3 | 1759 | static int do_proc_readlink(struct path *path, char __user *buffer, int buflen) |
1da177e4 | 1760 | { |
0ee931c4 | 1761 | char *tmp = (char *)__get_free_page(GFP_KERNEL); |
3dcd25f3 | 1762 | char *pathname; |
1da177e4 LT |
1763 | int len; |
1764 | ||
1765 | if (!tmp) | |
1766 | return -ENOMEM; | |
0c28f287 | 1767 | |
7b2a69ba | 1768 | pathname = d_path(path, tmp, PAGE_SIZE); |
3dcd25f3 JB |
1769 | len = PTR_ERR(pathname); |
1770 | if (IS_ERR(pathname)) | |
1da177e4 | 1771 | goto out; |
3dcd25f3 | 1772 | len = tmp + PAGE_SIZE - 1 - pathname; |
1da177e4 LT |
1773 | |
1774 | if (len > buflen) | |
1775 | len = buflen; | |
3dcd25f3 | 1776 | if (copy_to_user(buffer, pathname, len)) |
1da177e4 LT |
1777 | len = -EFAULT; |
1778 | out: | |
1779 | free_page((unsigned long)tmp); | |
1780 | return len; | |
1781 | } | |
1782 | ||
1783 | static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen) | |
1784 | { | |
1785 | int error = -EACCES; | |
2b0143b5 | 1786 | struct inode *inode = d_inode(dentry); |
3dcd25f3 | 1787 | struct path path; |
1da177e4 | 1788 | |
778c1144 EB |
1789 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1790 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1791 | goto out; |
1da177e4 | 1792 | |
7773fbc5 | 1793 | error = PROC_I(inode)->op.proc_get_link(dentry, &path); |
1da177e4 LT |
1794 | if (error) |
1795 | goto out; | |
1796 | ||
3dcd25f3 JB |
1797 | error = do_proc_readlink(&path, buffer, buflen); |
1798 | path_put(&path); | |
1da177e4 | 1799 | out: |
1da177e4 LT |
1800 | return error; |
1801 | } | |
1802 | ||
faf60af1 | 1803 | const struct inode_operations proc_pid_link_inode_operations = { |
1da177e4 | 1804 | .readlink = proc_pid_readlink, |
6b255391 | 1805 | .get_link = proc_pid_get_link, |
6d76fa58 | 1806 | .setattr = proc_setattr, |
1da177e4 LT |
1807 | }; |
1808 | ||
28a6d671 EB |
1809 | |
1810 | /* building an inode */ | |
1811 | ||
c6eb50d2 | 1812 | void task_dump_owner(struct task_struct *task, umode_t mode, |
68eb94f1 EB |
1813 | kuid_t *ruid, kgid_t *rgid) |
1814 | { | |
1815 | /* Depending on the state of dumpable compute who should own a | |
1816 | * proc file for a task. | |
1817 | */ | |
1818 | const struct cred *cred; | |
1819 | kuid_t uid; | |
1820 | kgid_t gid; | |
1821 | ||
2e0ad552 AD |
1822 | if (unlikely(task->flags & PF_KTHREAD)) { |
1823 | *ruid = GLOBAL_ROOT_UID; | |
1824 | *rgid = GLOBAL_ROOT_GID; | |
1825 | return; | |
1826 | } | |
1827 | ||
68eb94f1 EB |
1828 | /* Default to the tasks effective ownership */ |
1829 | rcu_read_lock(); | |
1830 | cred = __task_cred(task); | |
1831 | uid = cred->euid; | |
1832 | gid = cred->egid; | |
1833 | rcu_read_unlock(); | |
1834 | ||
1835 | /* | |
1836 | * Before the /proc/pid/status file was created the only way to read | |
1837 | * the effective uid of a /process was to stat /proc/pid. Reading | |
1838 | * /proc/pid/status is slow enough that procps and other packages | |
1839 | * kept stating /proc/pid. To keep the rules in /proc simple I have | |
1840 | * made this apply to all per process world readable and executable | |
1841 | * directories. | |
1842 | */ | |
1843 | if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) { | |
1844 | struct mm_struct *mm; | |
1845 | task_lock(task); | |
1846 | mm = task->mm; | |
1847 | /* Make non-dumpable tasks owned by some root */ | |
1848 | if (mm) { | |
1849 | if (get_dumpable(mm) != SUID_DUMP_USER) { | |
1850 | struct user_namespace *user_ns = mm->user_ns; | |
1851 | ||
1852 | uid = make_kuid(user_ns, 0); | |
1853 | if (!uid_valid(uid)) | |
1854 | uid = GLOBAL_ROOT_UID; | |
1855 | ||
1856 | gid = make_kgid(user_ns, 0); | |
1857 | if (!gid_valid(gid)) | |
1858 | gid = GLOBAL_ROOT_GID; | |
1859 | } | |
1860 | } else { | |
1861 | uid = GLOBAL_ROOT_UID; | |
1862 | gid = GLOBAL_ROOT_GID; | |
1863 | } | |
1864 | task_unlock(task); | |
1865 | } | |
1866 | *ruid = uid; | |
1867 | *rgid = gid; | |
1868 | } | |
1869 | ||
7bc3e6e5 EB |
1870 | void proc_pid_evict_inode(struct proc_inode *ei) |
1871 | { | |
1872 | struct pid *pid = ei->pid; | |
1873 | ||
1874 | if (S_ISDIR(ei->vfs_inode.i_mode)) { | |
63f818f4 | 1875 | spin_lock(&pid->lock); |
7bc3e6e5 | 1876 | hlist_del_init_rcu(&ei->sibling_inodes); |
63f818f4 | 1877 | spin_unlock(&pid->lock); |
7bc3e6e5 EB |
1878 | } |
1879 | ||
1880 | put_pid(pid); | |
1881 | } | |
1882 | ||
db978da8 AG |
1883 | struct inode *proc_pid_make_inode(struct super_block * sb, |
1884 | struct task_struct *task, umode_t mode) | |
28a6d671 EB |
1885 | { |
1886 | struct inode * inode; | |
1887 | struct proc_inode *ei; | |
7bc3e6e5 | 1888 | struct pid *pid; |
1da177e4 | 1889 | |
28a6d671 | 1890 | /* We need a new inode */ |
1da177e4 | 1891 | |
28a6d671 EB |
1892 | inode = new_inode(sb); |
1893 | if (!inode) | |
1894 | goto out; | |
1895 | ||
1896 | /* Common stuff */ | |
1897 | ei = PROC_I(inode); | |
db978da8 | 1898 | inode->i_mode = mode; |
85fe4025 | 1899 | inode->i_ino = get_next_ino(); |
078cd827 | 1900 | inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode); |
28a6d671 EB |
1901 | inode->i_op = &proc_def_inode_operations; |
1902 | ||
1903 | /* | |
1904 | * grab the reference to task. | |
1905 | */ | |
7bc3e6e5 EB |
1906 | pid = get_task_pid(task, PIDTYPE_PID); |
1907 | if (!pid) | |
28a6d671 EB |
1908 | goto out_unlock; |
1909 | ||
7bc3e6e5 EB |
1910 | /* Let the pid remember us for quick removal */ |
1911 | ei->pid = pid; | |
1912 | if (S_ISDIR(mode)) { | |
63f818f4 | 1913 | spin_lock(&pid->lock); |
7bc3e6e5 | 1914 | hlist_add_head_rcu(&ei->sibling_inodes, &pid->inodes); |
63f818f4 | 1915 | spin_unlock(&pid->lock); |
7bc3e6e5 EB |
1916 | } |
1917 | ||
68eb94f1 | 1918 | task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid); |
28a6d671 EB |
1919 | security_task_to_inode(task, inode); |
1920 | ||
1da177e4 | 1921 | out: |
28a6d671 EB |
1922 | return inode; |
1923 | ||
1924 | out_unlock: | |
1925 | iput(inode); | |
1926 | return NULL; | |
1da177e4 LT |
1927 | } |
1928 | ||
549c7297 CB |
1929 | int pid_getattr(struct user_namespace *mnt_userns, const struct path *path, |
1930 | struct kstat *stat, u32 request_mask, unsigned int query_flags) | |
1da177e4 | 1931 | { |
a528d35e | 1932 | struct inode *inode = d_inode(path->dentry); |
fa10fed3 | 1933 | struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb); |
28a6d671 | 1934 | struct task_struct *task; |
c69e8d9c | 1935 | |
0d56a451 | 1936 | generic_fillattr(&init_user_ns, inode, stat); |
1da177e4 | 1937 | |
dcb0f222 EB |
1938 | stat->uid = GLOBAL_ROOT_UID; |
1939 | stat->gid = GLOBAL_ROOT_GID; | |
94116929 | 1940 | rcu_read_lock(); |
28a6d671 EB |
1941 | task = pid_task(proc_pid(inode), PIDTYPE_PID); |
1942 | if (task) { | |
fa10fed3 | 1943 | if (!has_pid_permissions(fs_info, task, HIDEPID_INVISIBLE)) { |
0499680a VK |
1944 | rcu_read_unlock(); |
1945 | /* | |
1946 | * This doesn't prevent learning whether PID exists, | |
1947 | * it only makes getattr() consistent with readdir(). | |
1948 | */ | |
1949 | return -ENOENT; | |
1950 | } | |
68eb94f1 | 1951 | task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid); |
1da177e4 | 1952 | } |
28a6d671 | 1953 | rcu_read_unlock(); |
d6e71144 | 1954 | return 0; |
1da177e4 LT |
1955 | } |
1956 | ||
1da177e4 LT |
1957 | /* dentry stuff */ |
1958 | ||
1959 | /* | |
1bbc5513 AV |
1960 | * Set <pid>/... inode ownership (can change due to setuid(), etc.) |
1961 | */ | |
1962 | void pid_update_inode(struct task_struct *task, struct inode *inode) | |
1963 | { | |
1964 | task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid); | |
1965 | ||
1966 | inode->i_mode &= ~(S_ISUID | S_ISGID); | |
1967 | security_task_to_inode(task, inode); | |
1968 | } | |
1969 | ||
1970 | /* | |
1da177e4 LT |
1971 | * Rewrite the inode's ownerships here because the owning task may have |
1972 | * performed a setuid(), etc. | |
99f89551 | 1973 | * |
1da177e4 | 1974 | */ |
1bbc5513 | 1975 | static int pid_revalidate(struct dentry *dentry, unsigned int flags) |
1da177e4 | 1976 | { |
34286d66 NP |
1977 | struct inode *inode; |
1978 | struct task_struct *task; | |
c69e8d9c | 1979 | |
0b728e19 | 1980 | if (flags & LOOKUP_RCU) |
34286d66 NP |
1981 | return -ECHILD; |
1982 | ||
2b0143b5 | 1983 | inode = d_inode(dentry); |
34286d66 NP |
1984 | task = get_proc_task(inode); |
1985 | ||
99f89551 | 1986 | if (task) { |
1bbc5513 | 1987 | pid_update_inode(task, inode); |
99f89551 | 1988 | put_task_struct(task); |
1da177e4 LT |
1989 | return 1; |
1990 | } | |
1da177e4 LT |
1991 | return 0; |
1992 | } | |
1993 | ||
d855a4b7 ON |
1994 | static inline bool proc_inode_is_dead(struct inode *inode) |
1995 | { | |
1996 | return !proc_pid(inode)->tasks[PIDTYPE_PID].first; | |
1997 | } | |
1998 | ||
1dd704b6 DH |
1999 | int pid_delete_dentry(const struct dentry *dentry) |
2000 | { | |
2001 | /* Is the task we represent dead? | |
2002 | * If so, then don't put the dentry on the lru list, | |
2003 | * kill it immediately. | |
2004 | */ | |
2b0143b5 | 2005 | return proc_inode_is_dead(d_inode(dentry)); |
1dd704b6 DH |
2006 | } |
2007 | ||
6b4e306a | 2008 | const struct dentry_operations pid_dentry_operations = |
28a6d671 EB |
2009 | { |
2010 | .d_revalidate = pid_revalidate, | |
2011 | .d_delete = pid_delete_dentry, | |
2012 | }; | |
2013 | ||
2014 | /* Lookups */ | |
2015 | ||
1c0d04c9 EB |
2016 | /* |
2017 | * Fill a directory entry. | |
2018 | * | |
2019 | * If possible create the dcache entry and derive our inode number and | |
2020 | * file type from dcache entry. | |
2021 | * | |
2022 | * Since all of the proc inode numbers are dynamically generated, the inode | |
d2928e85 | 2023 | * numbers do not exist until the inode is cache. This means creating |
1c0d04c9 EB |
2024 | * the dcache entry in readdir is necessary to keep the inode numbers |
2025 | * reported by readdir in sync with the inode numbers reported | |
2026 | * by stat. | |
2027 | */ | |
f0c3b509 | 2028 | bool proc_fill_cache(struct file *file, struct dir_context *ctx, |
a4ef3895 | 2029 | const char *name, unsigned int len, |
c5141e6d | 2030 | instantiate_t instantiate, struct task_struct *task, const void *ptr) |
61a28784 | 2031 | { |
f0c3b509 | 2032 | struct dentry *child, *dir = file->f_path.dentry; |
1df98b8b | 2033 | struct qstr qname = QSTR_INIT(name, len); |
61a28784 | 2034 | struct inode *inode; |
0168b9e3 AV |
2035 | unsigned type = DT_UNKNOWN; |
2036 | ino_t ino = 1; | |
61a28784 | 2037 | |
1df98b8b | 2038 | child = d_hash_and_lookup(dir, &qname); |
61a28784 | 2039 | if (!child) { |
3781764b AV |
2040 | DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq); |
2041 | child = d_alloc_parallel(dir, &qname, &wq); | |
2042 | if (IS_ERR(child)) | |
1df98b8b | 2043 | goto end_instantiate; |
3781764b | 2044 | if (d_in_lookup(child)) { |
0168b9e3 AV |
2045 | struct dentry *res; |
2046 | res = instantiate(child, task, ptr); | |
3781764b | 2047 | d_lookup_done(child); |
0168b9e3 AV |
2048 | if (unlikely(res)) { |
2049 | dput(child); | |
2050 | child = res; | |
d85b399b AV |
2051 | if (IS_ERR(child)) |
2052 | goto end_instantiate; | |
3781764b | 2053 | } |
61a28784 EB |
2054 | } |
2055 | } | |
2b0143b5 | 2056 | inode = d_inode(child); |
147ce699 AV |
2057 | ino = inode->i_ino; |
2058 | type = inode->i_mode >> 12; | |
61a28784 | 2059 | dput(child); |
d85b399b | 2060 | end_instantiate: |
f0c3b509 | 2061 | return dir_emit(ctx, name, len, ino, type); |
61a28784 EB |
2062 | } |
2063 | ||
640708a2 PE |
2064 | /* |
2065 | * dname_to_vma_addr - maps a dentry name into two unsigned longs | |
2066 | * which represent vma start and end addresses. | |
2067 | */ | |
2068 | static int dname_to_vma_addr(struct dentry *dentry, | |
2069 | unsigned long *start, unsigned long *end) | |
2070 | { | |
ac7f1061 AD |
2071 | const char *str = dentry->d_name.name; |
2072 | unsigned long long sval, eval; | |
2073 | unsigned int len; | |
2074 | ||
35318db5 AD |
2075 | if (str[0] == '0' && str[1] != '-') |
2076 | return -EINVAL; | |
ac7f1061 AD |
2077 | len = _parse_integer(str, 16, &sval); |
2078 | if (len & KSTRTOX_OVERFLOW) | |
2079 | return -EINVAL; | |
2080 | if (sval != (unsigned long)sval) | |
2081 | return -EINVAL; | |
2082 | str += len; | |
2083 | ||
2084 | if (*str != '-') | |
640708a2 | 2085 | return -EINVAL; |
ac7f1061 AD |
2086 | str++; |
2087 | ||
35318db5 AD |
2088 | if (str[0] == '0' && str[1]) |
2089 | return -EINVAL; | |
ac7f1061 AD |
2090 | len = _parse_integer(str, 16, &eval); |
2091 | if (len & KSTRTOX_OVERFLOW) | |
2092 | return -EINVAL; | |
2093 | if (eval != (unsigned long)eval) | |
2094 | return -EINVAL; | |
2095 | str += len; | |
2096 | ||
2097 | if (*str != '\0') | |
2098 | return -EINVAL; | |
2099 | ||
2100 | *start = sval; | |
2101 | *end = eval; | |
640708a2 PE |
2102 | |
2103 | return 0; | |
2104 | } | |
2105 | ||
0b728e19 | 2106 | static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags) |
640708a2 PE |
2107 | { |
2108 | unsigned long vm_start, vm_end; | |
2109 | bool exact_vma_exists = false; | |
2110 | struct mm_struct *mm = NULL; | |
2111 | struct task_struct *task; | |
640708a2 PE |
2112 | struct inode *inode; |
2113 | int status = 0; | |
2114 | ||
0b728e19 | 2115 | if (flags & LOOKUP_RCU) |
640708a2 PE |
2116 | return -ECHILD; |
2117 | ||
2b0143b5 | 2118 | inode = d_inode(dentry); |
640708a2 PE |
2119 | task = get_proc_task(inode); |
2120 | if (!task) | |
2121 | goto out_notask; | |
2122 | ||
caaee623 | 2123 | mm = mm_access(task, PTRACE_MODE_READ_FSCREDS); |
2344bec7 | 2124 | if (IS_ERR_OR_NULL(mm)) |
640708a2 PE |
2125 | goto out; |
2126 | ||
2127 | if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) { | |
d8ed45c5 | 2128 | status = mmap_read_lock_killable(mm); |
cd9e2bb8 KK |
2129 | if (!status) { |
2130 | exact_vma_exists = !!find_exact_vma(mm, vm_start, | |
2131 | vm_end); | |
d8ed45c5 | 2132 | mmap_read_unlock(mm); |
cd9e2bb8 | 2133 | } |
640708a2 PE |
2134 | } |
2135 | ||
2136 | mmput(mm); | |
2137 | ||
2138 | if (exact_vma_exists) { | |
68eb94f1 EB |
2139 | task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid); |
2140 | ||
640708a2 PE |
2141 | security_task_to_inode(task, inode); |
2142 | status = 1; | |
2143 | } | |
2144 | ||
2145 | out: | |
2146 | put_task_struct(task); | |
2147 | ||
2148 | out_notask: | |
640708a2 PE |
2149 | return status; |
2150 | } | |
2151 | ||
2152 | static const struct dentry_operations tid_map_files_dentry_operations = { | |
2153 | .d_revalidate = map_files_d_revalidate, | |
2154 | .d_delete = pid_delete_dentry, | |
2155 | }; | |
2156 | ||
6b255391 | 2157 | static int map_files_get_link(struct dentry *dentry, struct path *path) |
640708a2 PE |
2158 | { |
2159 | unsigned long vm_start, vm_end; | |
2160 | struct vm_area_struct *vma; | |
2161 | struct task_struct *task; | |
2162 | struct mm_struct *mm; | |
2163 | int rc; | |
2164 | ||
2165 | rc = -ENOENT; | |
2b0143b5 | 2166 | task = get_proc_task(d_inode(dentry)); |
640708a2 PE |
2167 | if (!task) |
2168 | goto out; | |
2169 | ||
2170 | mm = get_task_mm(task); | |
2171 | put_task_struct(task); | |
2172 | if (!mm) | |
2173 | goto out; | |
2174 | ||
2175 | rc = dname_to_vma_addr(dentry, &vm_start, &vm_end); | |
2176 | if (rc) | |
2177 | goto out_mmput; | |
2178 | ||
d8ed45c5 | 2179 | rc = mmap_read_lock_killable(mm); |
cd9e2bb8 KK |
2180 | if (rc) |
2181 | goto out_mmput; | |
2182 | ||
70335abb | 2183 | rc = -ENOENT; |
640708a2 PE |
2184 | vma = find_exact_vma(mm, vm_start, vm_end); |
2185 | if (vma && vma->vm_file) { | |
2186 | *path = vma->vm_file->f_path; | |
2187 | path_get(path); | |
2188 | rc = 0; | |
2189 | } | |
d8ed45c5 | 2190 | mmap_read_unlock(mm); |
640708a2 PE |
2191 | |
2192 | out_mmput: | |
2193 | mmput(mm); | |
2194 | out: | |
2195 | return rc; | |
2196 | } | |
2197 | ||
2198 | struct map_files_info { | |
20d28cde AD |
2199 | unsigned long start; |
2200 | unsigned long end; | |
7b540d06 | 2201 | fmode_t mode; |
640708a2 PE |
2202 | }; |
2203 | ||
bdb4d100 | 2204 | /* |
12886f8a AR |
2205 | * Only allow CAP_SYS_ADMIN and CAP_CHECKPOINT_RESTORE to follow the links, due |
2206 | * to concerns about how the symlinks may be used to bypass permissions on | |
2207 | * ancestor directories in the path to the file in question. | |
bdb4d100 CO |
2208 | */ |
2209 | static const char * | |
6b255391 | 2210 | proc_map_files_get_link(struct dentry *dentry, |
fceef393 AV |
2211 | struct inode *inode, |
2212 | struct delayed_call *done) | |
bdb4d100 | 2213 | { |
12886f8a | 2214 | if (!checkpoint_restore_ns_capable(&init_user_ns)) |
bdb4d100 CO |
2215 | return ERR_PTR(-EPERM); |
2216 | ||
fceef393 | 2217 | return proc_pid_get_link(dentry, inode, done); |
bdb4d100 CO |
2218 | } |
2219 | ||
2220 | /* | |
6b255391 | 2221 | * Identical to proc_pid_link_inode_operations except for get_link() |
bdb4d100 CO |
2222 | */ |
2223 | static const struct inode_operations proc_map_files_link_inode_operations = { | |
2224 | .readlink = proc_pid_readlink, | |
6b255391 | 2225 | .get_link = proc_map_files_get_link, |
bdb4d100 CO |
2226 | .setattr = proc_setattr, |
2227 | }; | |
2228 | ||
0168b9e3 AV |
2229 | static struct dentry * |
2230 | proc_map_files_instantiate(struct dentry *dentry, | |
640708a2 PE |
2231 | struct task_struct *task, const void *ptr) |
2232 | { | |
7b540d06 | 2233 | fmode_t mode = (fmode_t)(unsigned long)ptr; |
640708a2 PE |
2234 | struct proc_inode *ei; |
2235 | struct inode *inode; | |
2236 | ||
0168b9e3 | 2237 | inode = proc_pid_make_inode(dentry->d_sb, task, S_IFLNK | |
db978da8 AG |
2238 | ((mode & FMODE_READ ) ? S_IRUSR : 0) | |
2239 | ((mode & FMODE_WRITE) ? S_IWUSR : 0)); | |
640708a2 | 2240 | if (!inode) |
0168b9e3 | 2241 | return ERR_PTR(-ENOENT); |
640708a2 PE |
2242 | |
2243 | ei = PROC_I(inode); | |
6b255391 | 2244 | ei->op.proc_get_link = map_files_get_link; |
640708a2 | 2245 | |
bdb4d100 | 2246 | inode->i_op = &proc_map_files_link_inode_operations; |
640708a2 | 2247 | inode->i_size = 64; |
640708a2 PE |
2248 | |
2249 | d_set_d_op(dentry, &tid_map_files_dentry_operations); | |
0168b9e3 | 2250 | return d_splice_alias(inode, dentry); |
640708a2 PE |
2251 | } |
2252 | ||
2253 | static struct dentry *proc_map_files_lookup(struct inode *dir, | |
00cd8dd3 | 2254 | struct dentry *dentry, unsigned int flags) |
640708a2 PE |
2255 | { |
2256 | unsigned long vm_start, vm_end; | |
2257 | struct vm_area_struct *vma; | |
2258 | struct task_struct *task; | |
0168b9e3 | 2259 | struct dentry *result; |
640708a2 PE |
2260 | struct mm_struct *mm; |
2261 | ||
0168b9e3 | 2262 | result = ERR_PTR(-ENOENT); |
640708a2 PE |
2263 | task = get_proc_task(dir); |
2264 | if (!task) | |
2265 | goto out; | |
2266 | ||
0168b9e3 | 2267 | result = ERR_PTR(-EACCES); |
caaee623 | 2268 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
640708a2 PE |
2269 | goto out_put_task; |
2270 | ||
0168b9e3 | 2271 | result = ERR_PTR(-ENOENT); |
640708a2 | 2272 | if (dname_to_vma_addr(dentry, &vm_start, &vm_end)) |
eb94cd96 | 2273 | goto out_put_task; |
640708a2 PE |
2274 | |
2275 | mm = get_task_mm(task); | |
2276 | if (!mm) | |
eb94cd96 | 2277 | goto out_put_task; |
640708a2 | 2278 | |
cd9e2bb8 | 2279 | result = ERR_PTR(-EINTR); |
d8ed45c5 | 2280 | if (mmap_read_lock_killable(mm)) |
cd9e2bb8 KK |
2281 | goto out_put_mm; |
2282 | ||
2283 | result = ERR_PTR(-ENOENT); | |
640708a2 PE |
2284 | vma = find_exact_vma(mm, vm_start, vm_end); |
2285 | if (!vma) | |
2286 | goto out_no_vma; | |
2287 | ||
05f56484 | 2288 | if (vma->vm_file) |
0168b9e3 | 2289 | result = proc_map_files_instantiate(dentry, task, |
05f56484 | 2290 | (void *)(unsigned long)vma->vm_file->f_mode); |
640708a2 PE |
2291 | |
2292 | out_no_vma: | |
d8ed45c5 | 2293 | mmap_read_unlock(mm); |
cd9e2bb8 | 2294 | out_put_mm: |
640708a2 | 2295 | mmput(mm); |
640708a2 PE |
2296 | out_put_task: |
2297 | put_task_struct(task); | |
2298 | out: | |
0168b9e3 | 2299 | return result; |
640708a2 PE |
2300 | } |
2301 | ||
2302 | static const struct inode_operations proc_map_files_inode_operations = { | |
2303 | .lookup = proc_map_files_lookup, | |
2304 | .permission = proc_fd_permission, | |
2305 | .setattr = proc_setattr, | |
2306 | }; | |
2307 | ||
2308 | static int | |
f0c3b509 | 2309 | proc_map_files_readdir(struct file *file, struct dir_context *ctx) |
640708a2 | 2310 | { |
640708a2 PE |
2311 | struct vm_area_struct *vma; |
2312 | struct task_struct *task; | |
2313 | struct mm_struct *mm; | |
f0c3b509 | 2314 | unsigned long nr_files, pos, i; |
94f8f3b0 | 2315 | GENRADIX(struct map_files_info) fa; |
f0c3b509 | 2316 | struct map_files_info *p; |
640708a2 PE |
2317 | int ret; |
2318 | ||
94f8f3b0 KO |
2319 | genradix_init(&fa); |
2320 | ||
640708a2 | 2321 | ret = -ENOENT; |
f0c3b509 | 2322 | task = get_proc_task(file_inode(file)); |
640708a2 PE |
2323 | if (!task) |
2324 | goto out; | |
2325 | ||
2326 | ret = -EACCES; | |
caaee623 | 2327 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
640708a2 PE |
2328 | goto out_put_task; |
2329 | ||
2330 | ret = 0; | |
f0c3b509 AV |
2331 | if (!dir_emit_dots(file, ctx)) |
2332 | goto out_put_task; | |
640708a2 | 2333 | |
f0c3b509 AV |
2334 | mm = get_task_mm(task); |
2335 | if (!mm) | |
2336 | goto out_put_task; | |
cd9e2bb8 | 2337 | |
89154dd5 | 2338 | ret = mmap_read_lock_killable(mm); |
cd9e2bb8 KK |
2339 | if (ret) { |
2340 | mmput(mm); | |
2341 | goto out_put_task; | |
2342 | } | |
640708a2 | 2343 | |
f0c3b509 | 2344 | nr_files = 0; |
640708a2 | 2345 | |
f0c3b509 AV |
2346 | /* |
2347 | * We need two passes here: | |
2348 | * | |
c1e8d7c6 ML |
2349 | * 1) Collect vmas of mapped files with mmap_lock taken |
2350 | * 2) Release mmap_lock and instantiate entries | |
f0c3b509 AV |
2351 | * |
2352 | * otherwise we get lockdep complained, since filldir() | |
c1e8d7c6 | 2353 | * routine might require mmap_lock taken in might_fault(). |
f0c3b509 | 2354 | */ |
640708a2 | 2355 | |
f0c3b509 | 2356 | for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) { |
94f8f3b0 KO |
2357 | if (!vma->vm_file) |
2358 | continue; | |
2359 | if (++pos <= ctx->pos) | |
2360 | continue; | |
f0c3b509 | 2361 | |
94f8f3b0 KO |
2362 | p = genradix_ptr_alloc(&fa, nr_files++, GFP_KERNEL); |
2363 | if (!p) { | |
f0c3b509 | 2364 | ret = -ENOMEM; |
89154dd5 | 2365 | mmap_read_unlock(mm); |
f0c3b509 AV |
2366 | mmput(mm); |
2367 | goto out_put_task; | |
640708a2 | 2368 | } |
f0c3b509 | 2369 | |
94f8f3b0 KO |
2370 | p->start = vma->vm_start; |
2371 | p->end = vma->vm_end; | |
2372 | p->mode = vma->vm_file->f_mode; | |
640708a2 | 2373 | } |
89154dd5 | 2374 | mmap_read_unlock(mm); |
fe079a5e | 2375 | mmput(mm); |
f0c3b509 AV |
2376 | |
2377 | for (i = 0; i < nr_files; i++) { | |
20d28cde AD |
2378 | char buf[4 * sizeof(long) + 2]; /* max: %lx-%lx\0 */ |
2379 | unsigned int len; | |
2380 | ||
94f8f3b0 | 2381 | p = genradix_ptr(&fa, i); |
20d28cde | 2382 | len = snprintf(buf, sizeof(buf), "%lx-%lx", p->start, p->end); |
f0c3b509 | 2383 | if (!proc_fill_cache(file, ctx, |
20d28cde | 2384 | buf, len, |
f0c3b509 AV |
2385 | proc_map_files_instantiate, |
2386 | task, | |
2387 | (void *)(unsigned long)p->mode)) | |
2388 | break; | |
2389 | ctx->pos++; | |
640708a2 PE |
2390 | } |
2391 | ||
640708a2 PE |
2392 | out_put_task: |
2393 | put_task_struct(task); | |
2394 | out: | |
94f8f3b0 | 2395 | genradix_free(&fa); |
640708a2 PE |
2396 | return ret; |
2397 | } | |
2398 | ||
2399 | static const struct file_operations proc_map_files_operations = { | |
2400 | .read = generic_read_dir, | |
f50752ea AV |
2401 | .iterate_shared = proc_map_files_readdir, |
2402 | .llseek = generic_file_llseek, | |
640708a2 PE |
2403 | }; |
2404 | ||
b18b6a9c | 2405 | #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS) |
48f6a7a5 PE |
2406 | struct timers_private { |
2407 | struct pid *pid; | |
2408 | struct task_struct *task; | |
2409 | struct sighand_struct *sighand; | |
57b8015e | 2410 | struct pid_namespace *ns; |
48f6a7a5 PE |
2411 | unsigned long flags; |
2412 | }; | |
2413 | ||
2414 | static void *timers_start(struct seq_file *m, loff_t *pos) | |
2415 | { | |
2416 | struct timers_private *tp = m->private; | |
2417 | ||
2418 | tp->task = get_pid_task(tp->pid, PIDTYPE_PID); | |
2419 | if (!tp->task) | |
2420 | return ERR_PTR(-ESRCH); | |
2421 | ||
2422 | tp->sighand = lock_task_sighand(tp->task, &tp->flags); | |
2423 | if (!tp->sighand) | |
2424 | return ERR_PTR(-ESRCH); | |
2425 | ||
2426 | return seq_list_start(&tp->task->signal->posix_timers, *pos); | |
2427 | } | |
2428 | ||
2429 | static void *timers_next(struct seq_file *m, void *v, loff_t *pos) | |
2430 | { | |
2431 | struct timers_private *tp = m->private; | |
2432 | return seq_list_next(v, &tp->task->signal->posix_timers, pos); | |
2433 | } | |
2434 | ||
2435 | static void timers_stop(struct seq_file *m, void *v) | |
2436 | { | |
2437 | struct timers_private *tp = m->private; | |
2438 | ||
2439 | if (tp->sighand) { | |
2440 | unlock_task_sighand(tp->task, &tp->flags); | |
2441 | tp->sighand = NULL; | |
2442 | } | |
2443 | ||
2444 | if (tp->task) { | |
2445 | put_task_struct(tp->task); | |
2446 | tp->task = NULL; | |
2447 | } | |
2448 | } | |
2449 | ||
2450 | static int show_timer(struct seq_file *m, void *v) | |
2451 | { | |
2452 | struct k_itimer *timer; | |
57b8015e PE |
2453 | struct timers_private *tp = m->private; |
2454 | int notify; | |
cedbccab | 2455 | static const char * const nstr[] = { |
57b8015e PE |
2456 | [SIGEV_SIGNAL] = "signal", |
2457 | [SIGEV_NONE] = "none", | |
2458 | [SIGEV_THREAD] = "thread", | |
2459 | }; | |
48f6a7a5 PE |
2460 | |
2461 | timer = list_entry((struct list_head *)v, struct k_itimer, list); | |
57b8015e PE |
2462 | notify = timer->it_sigev_notify; |
2463 | ||
48f6a7a5 | 2464 | seq_printf(m, "ID: %d\n", timer->it_id); |
ba3edf1f | 2465 | seq_printf(m, "signal: %d/%px\n", |
25ce3191 JP |
2466 | timer->sigq->info.si_signo, |
2467 | timer->sigq->info.si_value.sival_ptr); | |
57b8015e | 2468 | seq_printf(m, "notify: %s/%s.%d\n", |
25ce3191 JP |
2469 | nstr[notify & ~SIGEV_THREAD_ID], |
2470 | (notify & SIGEV_THREAD_ID) ? "tid" : "pid", | |
2471 | pid_nr_ns(timer->it_pid, tp->ns)); | |
15ef0298 | 2472 | seq_printf(m, "ClockID: %d\n", timer->it_clock); |
48f6a7a5 PE |
2473 | |
2474 | return 0; | |
2475 | } | |
2476 | ||
2477 | static const struct seq_operations proc_timers_seq_ops = { | |
2478 | .start = timers_start, | |
2479 | .next = timers_next, | |
2480 | .stop = timers_stop, | |
2481 | .show = show_timer, | |
2482 | }; | |
2483 | ||
2484 | static int proc_timers_open(struct inode *inode, struct file *file) | |
2485 | { | |
2486 | struct timers_private *tp; | |
2487 | ||
2488 | tp = __seq_open_private(file, &proc_timers_seq_ops, | |
2489 | sizeof(struct timers_private)); | |
2490 | if (!tp) | |
2491 | return -ENOMEM; | |
2492 | ||
2493 | tp->pid = proc_pid(inode); | |
9d78edea | 2494 | tp->ns = proc_pid_ns(inode->i_sb); |
48f6a7a5 PE |
2495 | return 0; |
2496 | } | |
2497 | ||
2498 | static const struct file_operations proc_timers_operations = { | |
2499 | .open = proc_timers_open, | |
2500 | .read = seq_read, | |
2501 | .llseek = seq_lseek, | |
2502 | .release = seq_release_private, | |
2503 | }; | |
b5946bea | 2504 | #endif |
640708a2 | 2505 | |
5de23d43 JS |
2506 | static ssize_t timerslack_ns_write(struct file *file, const char __user *buf, |
2507 | size_t count, loff_t *offset) | |
2508 | { | |
2509 | struct inode *inode = file_inode(file); | |
2510 | struct task_struct *p; | |
2511 | u64 slack_ns; | |
2512 | int err; | |
2513 | ||
2514 | err = kstrtoull_from_user(buf, count, 10, &slack_ns); | |
2515 | if (err < 0) | |
2516 | return err; | |
2517 | ||
2518 | p = get_proc_task(inode); | |
2519 | if (!p) | |
2520 | return -ESRCH; | |
2521 | ||
4b2bd5fe | 2522 | if (p != current) { |
8da0b4f6 BG |
2523 | rcu_read_lock(); |
2524 | if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) { | |
2525 | rcu_read_unlock(); | |
4b2bd5fe JS |
2526 | count = -EPERM; |
2527 | goto out; | |
2528 | } | |
8da0b4f6 | 2529 | rcu_read_unlock(); |
5de23d43 | 2530 | |
4b2bd5fe JS |
2531 | err = security_task_setscheduler(p); |
2532 | if (err) { | |
2533 | count = err; | |
2534 | goto out; | |
2535 | } | |
904763e1 JS |
2536 | } |
2537 | ||
7abbaf94 JS |
2538 | task_lock(p); |
2539 | if (slack_ns == 0) | |
2540 | p->timer_slack_ns = p->default_timer_slack_ns; | |
2541 | else | |
2542 | p->timer_slack_ns = slack_ns; | |
2543 | task_unlock(p); | |
2544 | ||
2545 | out: | |
5de23d43 JS |
2546 | put_task_struct(p); |
2547 | ||
2548 | return count; | |
2549 | } | |
2550 | ||
2551 | static int timerslack_ns_show(struct seq_file *m, void *v) | |
2552 | { | |
2553 | struct inode *inode = m->private; | |
2554 | struct task_struct *p; | |
7abbaf94 | 2555 | int err = 0; |
5de23d43 JS |
2556 | |
2557 | p = get_proc_task(inode); | |
2558 | if (!p) | |
2559 | return -ESRCH; | |
2560 | ||
4b2bd5fe | 2561 | if (p != current) { |
8da0b4f6 BG |
2562 | rcu_read_lock(); |
2563 | if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) { | |
2564 | rcu_read_unlock(); | |
4b2bd5fe JS |
2565 | err = -EPERM; |
2566 | goto out; | |
2567 | } | |
8da0b4f6 BG |
2568 | rcu_read_unlock(); |
2569 | ||
4b2bd5fe JS |
2570 | err = security_task_getscheduler(p); |
2571 | if (err) | |
2572 | goto out; | |
2573 | } | |
904763e1 | 2574 | |
7abbaf94 JS |
2575 | task_lock(p); |
2576 | seq_printf(m, "%llu\n", p->timer_slack_ns); | |
2577 | task_unlock(p); | |
2578 | ||
2579 | out: | |
5de23d43 JS |
2580 | put_task_struct(p); |
2581 | ||
2582 | return err; | |
2583 | } | |
2584 | ||
2585 | static int timerslack_ns_open(struct inode *inode, struct file *filp) | |
2586 | { | |
2587 | return single_open(filp, timerslack_ns_show, inode); | |
2588 | } | |
2589 | ||
2590 | static const struct file_operations proc_pid_set_timerslack_ns_operations = { | |
2591 | .open = timerslack_ns_open, | |
2592 | .read = seq_read, | |
2593 | .write = timerslack_ns_write, | |
2594 | .llseek = seq_lseek, | |
2595 | .release = single_release, | |
2596 | }; | |
2597 | ||
0168b9e3 AV |
2598 | static struct dentry *proc_pident_instantiate(struct dentry *dentry, |
2599 | struct task_struct *task, const void *ptr) | |
444ceed8 | 2600 | { |
c5141e6d | 2601 | const struct pid_entry *p = ptr; |
444ceed8 EB |
2602 | struct inode *inode; |
2603 | struct proc_inode *ei; | |
444ceed8 | 2604 | |
0168b9e3 | 2605 | inode = proc_pid_make_inode(dentry->d_sb, task, p->mode); |
444ceed8 | 2606 | if (!inode) |
0168b9e3 | 2607 | return ERR_PTR(-ENOENT); |
444ceed8 EB |
2608 | |
2609 | ei = PROC_I(inode); | |
444ceed8 | 2610 | if (S_ISDIR(inode->i_mode)) |
bfe86848 | 2611 | set_nlink(inode, 2); /* Use getattr to fix if necessary */ |
444ceed8 EB |
2612 | if (p->iop) |
2613 | inode->i_op = p->iop; | |
2614 | if (p->fop) | |
2615 | inode->i_fop = p->fop; | |
2616 | ei->op = p->op; | |
1bbc5513 | 2617 | pid_update_inode(task, inode); |
fb045adb | 2618 | d_set_d_op(dentry, &pid_dentry_operations); |
0168b9e3 | 2619 | return d_splice_alias(inode, dentry); |
444ceed8 EB |
2620 | } |
2621 | ||
1da177e4 LT |
2622 | static struct dentry *proc_pident_lookup(struct inode *dir, |
2623 | struct dentry *dentry, | |
d5a572a4 AD |
2624 | const struct pid_entry *p, |
2625 | const struct pid_entry *end) | |
1da177e4 | 2626 | { |
99f89551 | 2627 | struct task_struct *task = get_proc_task(dir); |
0168b9e3 | 2628 | struct dentry *res = ERR_PTR(-ENOENT); |
1da177e4 | 2629 | |
99f89551 EB |
2630 | if (!task) |
2631 | goto out_no_task; | |
1da177e4 | 2632 | |
20cdc894 EB |
2633 | /* |
2634 | * Yes, it does not scale. And it should not. Don't add | |
2635 | * new entries into /proc/<tgid>/ without very good reasons. | |
2636 | */ | |
d5a572a4 | 2637 | for (; p < end; p++) { |
1da177e4 LT |
2638 | if (p->len != dentry->d_name.len) |
2639 | continue; | |
26b95137 AD |
2640 | if (!memcmp(dentry->d_name.name, p->name, p->len)) { |
2641 | res = proc_pident_instantiate(dentry, task, p); | |
1da177e4 | 2642 | break; |
26b95137 | 2643 | } |
1da177e4 | 2644 | } |
99f89551 EB |
2645 | put_task_struct(task); |
2646 | out_no_task: | |
0168b9e3 | 2647 | return res; |
1da177e4 LT |
2648 | } |
2649 | ||
f0c3b509 | 2650 | static int proc_pident_readdir(struct file *file, struct dir_context *ctx, |
c5141e6d | 2651 | const struct pid_entry *ents, unsigned int nents) |
28a6d671 | 2652 | { |
f0c3b509 AV |
2653 | struct task_struct *task = get_proc_task(file_inode(file)); |
2654 | const struct pid_entry *p; | |
28a6d671 | 2655 | |
28a6d671 | 2656 | if (!task) |
f0c3b509 | 2657 | return -ENOENT; |
28a6d671 | 2658 | |
f0c3b509 AV |
2659 | if (!dir_emit_dots(file, ctx)) |
2660 | goto out; | |
2661 | ||
2662 | if (ctx->pos >= nents + 2) | |
2663 | goto out; | |
28a6d671 | 2664 | |
bac5f5d5 | 2665 | for (p = ents + (ctx->pos - 2); p < ents + nents; p++) { |
f0c3b509 AV |
2666 | if (!proc_fill_cache(file, ctx, p->name, p->len, |
2667 | proc_pident_instantiate, task, p)) | |
2668 | break; | |
2669 | ctx->pos++; | |
2670 | } | |
28a6d671 | 2671 | out: |
61a28784 | 2672 | put_task_struct(task); |
f0c3b509 | 2673 | return 0; |
1da177e4 LT |
2674 | } |
2675 | ||
28a6d671 | 2676 | #ifdef CONFIG_SECURITY |
591a22c1 KC |
2677 | static int proc_pid_attr_open(struct inode *inode, struct file *file) |
2678 | { | |
2679 | return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS); | |
2680 | } | |
2681 | ||
28a6d671 EB |
2682 | static ssize_t proc_pid_attr_read(struct file * file, char __user * buf, |
2683 | size_t count, loff_t *ppos) | |
2684 | { | |
496ad9aa | 2685 | struct inode * inode = file_inode(file); |
04ff9708 | 2686 | char *p = NULL; |
28a6d671 EB |
2687 | ssize_t length; |
2688 | struct task_struct *task = get_proc_task(inode); | |
2689 | ||
28a6d671 | 2690 | if (!task) |
04ff9708 | 2691 | return -ESRCH; |
28a6d671 | 2692 | |
6d9c939d | 2693 | length = security_getprocattr(task, PROC_I(inode)->op.lsm, |
2fddfeef | 2694 | (char*)file->f_path.dentry->d_name.name, |
04ff9708 | 2695 | &p); |
28a6d671 | 2696 | put_task_struct(task); |
04ff9708 AV |
2697 | if (length > 0) |
2698 | length = simple_read_from_buffer(buf, count, ppos, p, length); | |
2699 | kfree(p); | |
28a6d671 | 2700 | return length; |
1da177e4 LT |
2701 | } |
2702 | ||
28a6d671 EB |
2703 | static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf, |
2704 | size_t count, loff_t *ppos) | |
2705 | { | |
496ad9aa | 2706 | struct inode * inode = file_inode(file); |
41089b6d | 2707 | struct task_struct *task; |
bb646cdb | 2708 | void *page; |
41089b6d | 2709 | int rv; |
b21507e2 | 2710 | |
bfb819ea | 2711 | /* A task may only write when it was the opener. */ |
591a22c1 | 2712 | if (file->private_data != current->mm) |
bfb819ea KC |
2713 | return -EPERM; |
2714 | ||
41089b6d AD |
2715 | rcu_read_lock(); |
2716 | task = pid_task(proc_pid(inode), PIDTYPE_PID); | |
2717 | if (!task) { | |
2718 | rcu_read_unlock(); | |
2719 | return -ESRCH; | |
2720 | } | |
b21507e2 | 2721 | /* A task may only write its own attributes. */ |
41089b6d AD |
2722 | if (current != task) { |
2723 | rcu_read_unlock(); | |
2724 | return -EACCES; | |
2725 | } | |
35a196be PM |
2726 | /* Prevent changes to overridden credentials. */ |
2727 | if (current_cred() != current_real_cred()) { | |
2728 | rcu_read_unlock(); | |
2729 | return -EBUSY; | |
2730 | } | |
41089b6d | 2731 | rcu_read_unlock(); |
b21507e2 | 2732 | |
28a6d671 EB |
2733 | if (count > PAGE_SIZE) |
2734 | count = PAGE_SIZE; | |
2735 | ||
2736 | /* No partial writes. */ | |
28a6d671 | 2737 | if (*ppos != 0) |
41089b6d | 2738 | return -EINVAL; |
28a6d671 | 2739 | |
bb646cdb AV |
2740 | page = memdup_user(buf, count); |
2741 | if (IS_ERR(page)) { | |
41089b6d | 2742 | rv = PTR_ERR(page); |
28a6d671 | 2743 | goto out; |
bb646cdb | 2744 | } |
28a6d671 | 2745 | |
107db7c7 | 2746 | /* Guard against adverse ptrace interaction */ |
41089b6d AD |
2747 | rv = mutex_lock_interruptible(¤t->signal->cred_guard_mutex); |
2748 | if (rv < 0) | |
107db7c7 DH |
2749 | goto out_free; |
2750 | ||
6d9c939d CS |
2751 | rv = security_setprocattr(PROC_I(inode)->op.lsm, |
2752 | file->f_path.dentry->d_name.name, page, | |
2753 | count); | |
b21507e2 | 2754 | mutex_unlock(¤t->signal->cred_guard_mutex); |
28a6d671 | 2755 | out_free: |
bb646cdb | 2756 | kfree(page); |
28a6d671 | 2757 | out: |
41089b6d | 2758 | return rv; |
28a6d671 EB |
2759 | } |
2760 | ||
00977a59 | 2761 | static const struct file_operations proc_pid_attr_operations = { |
591a22c1 | 2762 | .open = proc_pid_attr_open, |
28a6d671 EB |
2763 | .read = proc_pid_attr_read, |
2764 | .write = proc_pid_attr_write, | |
87df8424 | 2765 | .llseek = generic_file_llseek, |
591a22c1 | 2766 | .release = mem_release, |
28a6d671 EB |
2767 | }; |
2768 | ||
6d9c939d CS |
2769 | #define LSM_DIR_OPS(LSM) \ |
2770 | static int proc_##LSM##_attr_dir_iterate(struct file *filp, \ | |
2771 | struct dir_context *ctx) \ | |
2772 | { \ | |
2773 | return proc_pident_readdir(filp, ctx, \ | |
2774 | LSM##_attr_dir_stuff, \ | |
2775 | ARRAY_SIZE(LSM##_attr_dir_stuff)); \ | |
2776 | } \ | |
2777 | \ | |
2778 | static const struct file_operations proc_##LSM##_attr_dir_ops = { \ | |
2779 | .read = generic_read_dir, \ | |
2780 | .iterate = proc_##LSM##_attr_dir_iterate, \ | |
2781 | .llseek = default_llseek, \ | |
2782 | }; \ | |
2783 | \ | |
2784 | static struct dentry *proc_##LSM##_attr_dir_lookup(struct inode *dir, \ | |
2785 | struct dentry *dentry, unsigned int flags) \ | |
2786 | { \ | |
2787 | return proc_pident_lookup(dir, dentry, \ | |
2788 | LSM##_attr_dir_stuff, \ | |
d5a572a4 | 2789 | LSM##_attr_dir_stuff + ARRAY_SIZE(LSM##_attr_dir_stuff)); \ |
6d9c939d CS |
2790 | } \ |
2791 | \ | |
2792 | static const struct inode_operations proc_##LSM##_attr_dir_inode_ops = { \ | |
2793 | .lookup = proc_##LSM##_attr_dir_lookup, \ | |
2794 | .getattr = pid_getattr, \ | |
2795 | .setattr = proc_setattr, \ | |
2796 | } | |
2797 | ||
2798 | #ifdef CONFIG_SECURITY_SMACK | |
2799 | static const struct pid_entry smack_attr_dir_stuff[] = { | |
2800 | ATTR("smack", "current", 0666), | |
2801 | }; | |
2802 | LSM_DIR_OPS(smack); | |
2803 | #endif | |
2804 | ||
6413f852 JJ |
2805 | #ifdef CONFIG_SECURITY_APPARMOR |
2806 | static const struct pid_entry apparmor_attr_dir_stuff[] = { | |
2807 | ATTR("apparmor", "current", 0666), | |
2808 | ATTR("apparmor", "prev", 0444), | |
2809 | ATTR("apparmor", "exec", 0666), | |
2810 | }; | |
2811 | LSM_DIR_OPS(apparmor); | |
2812 | #endif | |
2813 | ||
c5141e6d | 2814 | static const struct pid_entry attr_dir_stuff[] = { |
6d9c939d CS |
2815 | ATTR(NULL, "current", 0666), |
2816 | ATTR(NULL, "prev", 0444), | |
2817 | ATTR(NULL, "exec", 0666), | |
2818 | ATTR(NULL, "fscreate", 0666), | |
2819 | ATTR(NULL, "keycreate", 0666), | |
2820 | ATTR(NULL, "sockcreate", 0666), | |
2821 | #ifdef CONFIG_SECURITY_SMACK | |
2822 | DIR("smack", 0555, | |
2823 | proc_smack_attr_dir_inode_ops, proc_smack_attr_dir_ops), | |
2824 | #endif | |
6413f852 JJ |
2825 | #ifdef CONFIG_SECURITY_APPARMOR |
2826 | DIR("apparmor", 0555, | |
2827 | proc_apparmor_attr_dir_inode_ops, proc_apparmor_attr_dir_ops), | |
2828 | #endif | |
28a6d671 EB |
2829 | }; |
2830 | ||
f0c3b509 | 2831 | static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx) |
28a6d671 | 2832 | { |
f0c3b509 AV |
2833 | return proc_pident_readdir(file, ctx, |
2834 | attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2835 | } |
2836 | ||
00977a59 | 2837 | static const struct file_operations proc_attr_dir_operations = { |
1da177e4 | 2838 | .read = generic_read_dir, |
f50752ea AV |
2839 | .iterate_shared = proc_attr_dir_readdir, |
2840 | .llseek = generic_file_llseek, | |
1da177e4 LT |
2841 | }; |
2842 | ||
72d9dcfc | 2843 | static struct dentry *proc_attr_dir_lookup(struct inode *dir, |
00cd8dd3 | 2844 | struct dentry *dentry, unsigned int flags) |
28a6d671 | 2845 | { |
7bcd6b0e | 2846 | return proc_pident_lookup(dir, dentry, |
d5a572a4 AD |
2847 | attr_dir_stuff, |
2848 | attr_dir_stuff + ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2849 | } |
2850 | ||
c5ef1c42 | 2851 | static const struct inode_operations proc_attr_dir_inode_operations = { |
72d9dcfc | 2852 | .lookup = proc_attr_dir_lookup, |
99f89551 | 2853 | .getattr = pid_getattr, |
6d76fa58 | 2854 | .setattr = proc_setattr, |
1da177e4 LT |
2855 | }; |
2856 | ||
28a6d671 EB |
2857 | #endif |
2858 | ||
698ba7b5 | 2859 | #ifdef CONFIG_ELF_CORE |
3cb4a0bb KH |
2860 | static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf, |
2861 | size_t count, loff_t *ppos) | |
2862 | { | |
496ad9aa | 2863 | struct task_struct *task = get_proc_task(file_inode(file)); |
3cb4a0bb KH |
2864 | struct mm_struct *mm; |
2865 | char buffer[PROC_NUMBUF]; | |
2866 | size_t len; | |
2867 | int ret; | |
2868 | ||
2869 | if (!task) | |
2870 | return -ESRCH; | |
2871 | ||
2872 | ret = 0; | |
2873 | mm = get_task_mm(task); | |
2874 | if (mm) { | |
2875 | len = snprintf(buffer, sizeof(buffer), "%08lx\n", | |
2876 | ((mm->flags & MMF_DUMP_FILTER_MASK) >> | |
2877 | MMF_DUMP_FILTER_SHIFT)); | |
2878 | mmput(mm); | |
2879 | ret = simple_read_from_buffer(buf, count, ppos, buffer, len); | |
2880 | } | |
2881 | ||
2882 | put_task_struct(task); | |
2883 | ||
2884 | return ret; | |
2885 | } | |
2886 | ||
2887 | static ssize_t proc_coredump_filter_write(struct file *file, | |
2888 | const char __user *buf, | |
2889 | size_t count, | |
2890 | loff_t *ppos) | |
2891 | { | |
2892 | struct task_struct *task; | |
2893 | struct mm_struct *mm; | |
3cb4a0bb KH |
2894 | unsigned int val; |
2895 | int ret; | |
2896 | int i; | |
2897 | unsigned long mask; | |
2898 | ||
774636e1 AD |
2899 | ret = kstrtouint_from_user(buf, count, 0, &val); |
2900 | if (ret < 0) | |
2901 | return ret; | |
3cb4a0bb KH |
2902 | |
2903 | ret = -ESRCH; | |
496ad9aa | 2904 | task = get_proc_task(file_inode(file)); |
3cb4a0bb KH |
2905 | if (!task) |
2906 | goto out_no_task; | |
2907 | ||
3cb4a0bb KH |
2908 | mm = get_task_mm(task); |
2909 | if (!mm) | |
2910 | goto out_no_mm; | |
41a0c249 | 2911 | ret = 0; |
3cb4a0bb KH |
2912 | |
2913 | for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) { | |
2914 | if (val & mask) | |
2915 | set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2916 | else | |
2917 | clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2918 | } | |
2919 | ||
2920 | mmput(mm); | |
2921 | out_no_mm: | |
2922 | put_task_struct(task); | |
2923 | out_no_task: | |
774636e1 AD |
2924 | if (ret < 0) |
2925 | return ret; | |
2926 | return count; | |
3cb4a0bb KH |
2927 | } |
2928 | ||
2929 | static const struct file_operations proc_coredump_filter_operations = { | |
2930 | .read = proc_coredump_filter_read, | |
2931 | .write = proc_coredump_filter_write, | |
87df8424 | 2932 | .llseek = generic_file_llseek, |
3cb4a0bb KH |
2933 | }; |
2934 | #endif | |
2935 | ||
aba76fdb | 2936 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 2937 | static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole) |
297c5d92 | 2938 | { |
940389b8 | 2939 | struct task_io_accounting acct = task->ioac; |
5995477a | 2940 | unsigned long flags; |
293eb1e7 | 2941 | int result; |
5995477a | 2942 | |
f7cfd871 | 2943 | result = down_read_killable(&task->signal->exec_update_lock); |
293eb1e7 VK |
2944 | if (result) |
2945 | return result; | |
2946 | ||
caaee623 | 2947 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) { |
293eb1e7 VK |
2948 | result = -EACCES; |
2949 | goto out_unlock; | |
2950 | } | |
1d1221f3 | 2951 | |
5995477a AR |
2952 | if (whole && lock_task_sighand(task, &flags)) { |
2953 | struct task_struct *t = task; | |
2954 | ||
2955 | task_io_accounting_add(&acct, &task->signal->ioac); | |
2956 | while_each_thread(task, t) | |
2957 | task_io_accounting_add(&acct, &t->ioac); | |
2958 | ||
2959 | unlock_task_sighand(task, &flags); | |
297c5d92 | 2960 | } |
25ce3191 JP |
2961 | seq_printf(m, |
2962 | "rchar: %llu\n" | |
2963 | "wchar: %llu\n" | |
2964 | "syscr: %llu\n" | |
2965 | "syscw: %llu\n" | |
2966 | "read_bytes: %llu\n" | |
2967 | "write_bytes: %llu\n" | |
2968 | "cancelled_write_bytes: %llu\n", | |
2969 | (unsigned long long)acct.rchar, | |
2970 | (unsigned long long)acct.wchar, | |
2971 | (unsigned long long)acct.syscr, | |
2972 | (unsigned long long)acct.syscw, | |
2973 | (unsigned long long)acct.read_bytes, | |
2974 | (unsigned long long)acct.write_bytes, | |
2975 | (unsigned long long)acct.cancelled_write_bytes); | |
2976 | result = 0; | |
2977 | ||
293eb1e7 | 2978 | out_unlock: |
f7cfd871 | 2979 | up_read(&task->signal->exec_update_lock); |
293eb1e7 | 2980 | return result; |
297c5d92 AR |
2981 | } |
2982 | ||
19aadc98 AD |
2983 | static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns, |
2984 | struct pid *pid, struct task_struct *task) | |
297c5d92 | 2985 | { |
19aadc98 | 2986 | return do_io_accounting(task, m, 0); |
aba76fdb | 2987 | } |
297c5d92 | 2988 | |
19aadc98 AD |
2989 | static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns, |
2990 | struct pid *pid, struct task_struct *task) | |
297c5d92 | 2991 | { |
19aadc98 | 2992 | return do_io_accounting(task, m, 1); |
297c5d92 AR |
2993 | } |
2994 | #endif /* CONFIG_TASK_IO_ACCOUNTING */ | |
aba76fdb | 2995 | |
22d917d8 EB |
2996 | #ifdef CONFIG_USER_NS |
2997 | static int proc_id_map_open(struct inode *inode, struct file *file, | |
ccf94f1b | 2998 | const struct seq_operations *seq_ops) |
22d917d8 EB |
2999 | { |
3000 | struct user_namespace *ns = NULL; | |
3001 | struct task_struct *task; | |
3002 | struct seq_file *seq; | |
3003 | int ret = -EINVAL; | |
3004 | ||
3005 | task = get_proc_task(inode); | |
3006 | if (task) { | |
3007 | rcu_read_lock(); | |
3008 | ns = get_user_ns(task_cred_xxx(task, user_ns)); | |
3009 | rcu_read_unlock(); | |
3010 | put_task_struct(task); | |
3011 | } | |
3012 | if (!ns) | |
3013 | goto err; | |
3014 | ||
3015 | ret = seq_open(file, seq_ops); | |
3016 | if (ret) | |
3017 | goto err_put_ns; | |
3018 | ||
3019 | seq = file->private_data; | |
3020 | seq->private = ns; | |
3021 | ||
3022 | return 0; | |
3023 | err_put_ns: | |
3024 | put_user_ns(ns); | |
3025 | err: | |
3026 | return ret; | |
3027 | } | |
3028 | ||
3029 | static int proc_id_map_release(struct inode *inode, struct file *file) | |
3030 | { | |
3031 | struct seq_file *seq = file->private_data; | |
3032 | struct user_namespace *ns = seq->private; | |
3033 | put_user_ns(ns); | |
3034 | return seq_release(inode, file); | |
3035 | } | |
3036 | ||
3037 | static int proc_uid_map_open(struct inode *inode, struct file *file) | |
3038 | { | |
3039 | return proc_id_map_open(inode, file, &proc_uid_seq_operations); | |
3040 | } | |
3041 | ||
3042 | static int proc_gid_map_open(struct inode *inode, struct file *file) | |
3043 | { | |
3044 | return proc_id_map_open(inode, file, &proc_gid_seq_operations); | |
3045 | } | |
3046 | ||
f76d207a EB |
3047 | static int proc_projid_map_open(struct inode *inode, struct file *file) |
3048 | { | |
3049 | return proc_id_map_open(inode, file, &proc_projid_seq_operations); | |
3050 | } | |
3051 | ||
22d917d8 EB |
3052 | static const struct file_operations proc_uid_map_operations = { |
3053 | .open = proc_uid_map_open, | |
3054 | .write = proc_uid_map_write, | |
3055 | .read = seq_read, | |
3056 | .llseek = seq_lseek, | |
3057 | .release = proc_id_map_release, | |
3058 | }; | |
3059 | ||
3060 | static const struct file_operations proc_gid_map_operations = { | |
3061 | .open = proc_gid_map_open, | |
3062 | .write = proc_gid_map_write, | |
3063 | .read = seq_read, | |
3064 | .llseek = seq_lseek, | |
3065 | .release = proc_id_map_release, | |
3066 | }; | |
f76d207a EB |
3067 | |
3068 | static const struct file_operations proc_projid_map_operations = { | |
3069 | .open = proc_projid_map_open, | |
3070 | .write = proc_projid_map_write, | |
3071 | .read = seq_read, | |
3072 | .llseek = seq_lseek, | |
3073 | .release = proc_id_map_release, | |
3074 | }; | |
9cc46516 EB |
3075 | |
3076 | static int proc_setgroups_open(struct inode *inode, struct file *file) | |
3077 | { | |
3078 | struct user_namespace *ns = NULL; | |
3079 | struct task_struct *task; | |
3080 | int ret; | |
3081 | ||
3082 | ret = -ESRCH; | |
3083 | task = get_proc_task(inode); | |
3084 | if (task) { | |
3085 | rcu_read_lock(); | |
3086 | ns = get_user_ns(task_cred_xxx(task, user_ns)); | |
3087 | rcu_read_unlock(); | |
3088 | put_task_struct(task); | |
3089 | } | |
3090 | if (!ns) | |
3091 | goto err; | |
3092 | ||
3093 | if (file->f_mode & FMODE_WRITE) { | |
3094 | ret = -EACCES; | |
3095 | if (!ns_capable(ns, CAP_SYS_ADMIN)) | |
3096 | goto err_put_ns; | |
3097 | } | |
3098 | ||
3099 | ret = single_open(file, &proc_setgroups_show, ns); | |
3100 | if (ret) | |
3101 | goto err_put_ns; | |
3102 | ||
3103 | return 0; | |
3104 | err_put_ns: | |
3105 | put_user_ns(ns); | |
3106 | err: | |
3107 | return ret; | |
3108 | } | |
3109 | ||
3110 | static int proc_setgroups_release(struct inode *inode, struct file *file) | |
3111 | { | |
3112 | struct seq_file *seq = file->private_data; | |
3113 | struct user_namespace *ns = seq->private; | |
3114 | int ret = single_release(inode, file); | |
3115 | put_user_ns(ns); | |
3116 | return ret; | |
3117 | } | |
3118 | ||
3119 | static const struct file_operations proc_setgroups_operations = { | |
3120 | .open = proc_setgroups_open, | |
3121 | .write = proc_setgroups_write, | |
3122 | .read = seq_read, | |
3123 | .llseek = seq_lseek, | |
3124 | .release = proc_setgroups_release, | |
3125 | }; | |
22d917d8 EB |
3126 | #endif /* CONFIG_USER_NS */ |
3127 | ||
47830723 KC |
3128 | static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns, |
3129 | struct pid *pid, struct task_struct *task) | |
3130 | { | |
a9712bc1 AV |
3131 | int err = lock_trace(task); |
3132 | if (!err) { | |
3133 | seq_printf(m, "%08x\n", task->personality); | |
3134 | unlock_trace(task); | |
3135 | } | |
3136 | return err; | |
47830723 KC |
3137 | } |
3138 | ||
7c23b330 JP |
3139 | #ifdef CONFIG_LIVEPATCH |
3140 | static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns, | |
3141 | struct pid *pid, struct task_struct *task) | |
3142 | { | |
3143 | seq_printf(m, "%d\n", task->patch_state); | |
3144 | return 0; | |
3145 | } | |
3146 | #endif /* CONFIG_LIVEPATCH */ | |
3147 | ||
c8d12627 AP |
3148 | #ifdef CONFIG_STACKLEAK_METRICS |
3149 | static int proc_stack_depth(struct seq_file *m, struct pid_namespace *ns, | |
3150 | struct pid *pid, struct task_struct *task) | |
3151 | { | |
3152 | unsigned long prev_depth = THREAD_SIZE - | |
3153 | (task->prev_lowest_stack & (THREAD_SIZE - 1)); | |
3154 | unsigned long depth = THREAD_SIZE - | |
3155 | (task->lowest_stack & (THREAD_SIZE - 1)); | |
3156 | ||
3157 | seq_printf(m, "previous stack depth: %lu\nstack depth: %lu\n", | |
3158 | prev_depth, depth); | |
3159 | return 0; | |
3160 | } | |
3161 | #endif /* CONFIG_STACKLEAK_METRICS */ | |
3162 | ||
28a6d671 EB |
3163 | /* |
3164 | * Thread groups | |
3165 | */ | |
00977a59 | 3166 | static const struct file_operations proc_task_operations; |
c5ef1c42 | 3167 | static const struct inode_operations proc_task_inode_operations; |
20cdc894 | 3168 | |
c5141e6d | 3169 | static const struct pid_entry tgid_base_stuff[] = { |
631f9c18 AD |
3170 | DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations), |
3171 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), | |
640708a2 | 3172 | DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations), |
631f9c18 | 3173 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
6b4e306a | 3174 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
b2211a36 | 3175 | #ifdef CONFIG_NET |
631f9c18 | 3176 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), |
b2211a36 | 3177 | #endif |
631f9c18 | 3178 | REG("environ", S_IRUSR, proc_environ_operations), |
c5317167 | 3179 | REG("auxv", S_IRUSR, proc_auxv_operations), |
631f9c18 | 3180 | ONE("status", S_IRUGO, proc_pid_status), |
35a35046 | 3181 | ONE("personality", S_IRUSR, proc_pid_personality), |
1c963eb1 | 3182 | ONE("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 3183 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 3184 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
5091faa4 MG |
3185 | #endif |
3186 | #ifdef CONFIG_SCHED_AUTOGROUP | |
3187 | REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations), | |
04a8682a AV |
3188 | #endif |
3189 | #ifdef CONFIG_TIME_NS | |
3190 | REG("timens_offsets", S_IRUGO|S_IWUSR, proc_timens_offsets_operations), | |
ebcb6734 | 3191 | #endif |
4614a696 | 3192 | REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations), |
ebcb6734 | 3193 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 | 3194 | ONE("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 3195 | #endif |
c2c0bb44 | 3196 | REG("cmdline", S_IRUGO, proc_pid_cmdline_ops), |
631f9c18 AD |
3197 | ONE("stat", S_IRUGO, proc_tgid_stat), |
3198 | ONE("statm", S_IRUGO, proc_pid_statm), | |
b7643757 | 3199 | REG("maps", S_IRUGO, proc_pid_maps_operations), |
28a6d671 | 3200 | #ifdef CONFIG_NUMA |
b7643757 | 3201 | REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations), |
28a6d671 | 3202 | #endif |
631f9c18 AD |
3203 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
3204 | LNK("cwd", proc_cwd_link), | |
3205 | LNK("root", proc_root_link), | |
3206 | LNK("exe", proc_exe_link), | |
3207 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
3208 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
3209 | REG("mountstats", S_IRUSR, proc_mountstats_operations), | |
1e883281 | 3210 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 | 3211 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
b7643757 | 3212 | REG("smaps", S_IRUGO, proc_pid_smaps_operations), |
493b0e9d | 3213 | REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations), |
32ed74a4 | 3214 | REG("pagemap", S_IRUSR, proc_pagemap_operations), |
28a6d671 EB |
3215 | #endif |
3216 | #ifdef CONFIG_SECURITY | |
631f9c18 | 3217 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
3218 | #endif |
3219 | #ifdef CONFIG_KALLSYMS | |
edfcd606 | 3220 | ONE("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 3221 | #endif |
2ec220e2 | 3222 | #ifdef CONFIG_STACKTRACE |
35a35046 | 3223 | ONE("stack", S_IRUSR, proc_pid_stack), |
28a6d671 | 3224 | #endif |
5968cece | 3225 | #ifdef CONFIG_SCHED_INFO |
f6e826ca | 3226 | ONE("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 3227 | #endif |
9745512c | 3228 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 3229 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 3230 | #endif |
8793d854 | 3231 | #ifdef CONFIG_PROC_PID_CPUSET |
52de4779 | 3232 | ONE("cpuset", S_IRUGO, proc_cpuset_show), |
a424316c PM |
3233 | #endif |
3234 | #ifdef CONFIG_CGROUPS | |
006f4ac4 | 3235 | ONE("cgroup", S_IRUGO, proc_cgroup_show), |
e79f15a4 CY |
3236 | #endif |
3237 | #ifdef CONFIG_PROC_CPU_RESCTRL | |
3238 | ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show), | |
28a6d671 | 3239 | #endif |
6ba51e37 | 3240 | ONE("oom_score", S_IRUGO, proc_oom_score), |
fa0cbbf1 | 3241 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations), |
a63d83f4 | 3242 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
4b7d248b | 3243 | #ifdef CONFIG_AUDIT |
631f9c18 AD |
3244 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
3245 | REG("sessionid", S_IRUGO, proc_sessionid_operations), | |
28a6d671 | 3246 | #endif |
f4f154fd | 3247 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 3248 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
168c42bc | 3249 | REG("fail-nth", 0644, proc_fail_nth_operations), |
f4f154fd | 3250 | #endif |
698ba7b5 | 3251 | #ifdef CONFIG_ELF_CORE |
631f9c18 | 3252 | REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations), |
3cb4a0bb | 3253 | #endif |
aba76fdb | 3254 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 3255 | ONE("io", S_IRUSR, proc_tgid_io_accounting), |
aba76fdb | 3256 | #endif |
22d917d8 EB |
3257 | #ifdef CONFIG_USER_NS |
3258 | REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations), | |
3259 | REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations), | |
f76d207a | 3260 | REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations), |
9cc46516 | 3261 | REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations), |
22d917d8 | 3262 | #endif |
b18b6a9c | 3263 | #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS) |
48f6a7a5 PE |
3264 | REG("timers", S_IRUGO, proc_timers_operations), |
3265 | #endif | |
5de23d43 | 3266 | REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations), |
7c23b330 JP |
3267 | #ifdef CONFIG_LIVEPATCH |
3268 | ONE("patch_state", S_IRUSR, proc_pid_patch_state), | |
3269 | #endif | |
c8d12627 AP |
3270 | #ifdef CONFIG_STACKLEAK_METRICS |
3271 | ONE("stack_depth", S_IRUGO, proc_stack_depth), | |
3272 | #endif | |
68bc30bb AL |
3273 | #ifdef CONFIG_PROC_PID_ARCH_STATUS |
3274 | ONE("arch_status", S_IRUGO, proc_pid_arch_status), | |
3275 | #endif | |
0d8315dd YZ |
3276 | #ifdef CONFIG_SECCOMP_CACHE_DEBUG |
3277 | ONE("seccomp_cache", S_IRUSR, proc_pid_seccomp_cache), | |
3278 | #endif | |
28a6d671 | 3279 | }; |
1da177e4 | 3280 | |
f0c3b509 | 3281 | static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3282 | { |
f0c3b509 AV |
3283 | return proc_pident_readdir(file, ctx, |
3284 | tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
3285 | } |
3286 | ||
00977a59 | 3287 | static const struct file_operations proc_tgid_base_operations = { |
1da177e4 | 3288 | .read = generic_read_dir, |
f50752ea AV |
3289 | .iterate_shared = proc_tgid_base_readdir, |
3290 | .llseek = generic_file_llseek, | |
1da177e4 LT |
3291 | }; |
3292 | ||
3eb39f47 CB |
3293 | struct pid *tgid_pidfd_to_pid(const struct file *file) |
3294 | { | |
30d158b1 | 3295 | if (file->f_op != &proc_tgid_base_operations) |
3eb39f47 CB |
3296 | return ERR_PTR(-EBADF); |
3297 | ||
3298 | return proc_pid(file_inode(file)); | |
3299 | } | |
3300 | ||
00cd8dd3 AV |
3301 | static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
3302 | { | |
7bcd6b0e | 3303 | return proc_pident_lookup(dir, dentry, |
d5a572a4 AD |
3304 | tgid_base_stuff, |
3305 | tgid_base_stuff + ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
3306 | } |
3307 | ||
c5ef1c42 | 3308 | static const struct inode_operations proc_tgid_base_inode_operations = { |
28a6d671 | 3309 | .lookup = proc_tgid_base_lookup, |
99f89551 | 3310 | .getattr = pid_getattr, |
6d76fa58 | 3311 | .setattr = proc_setattr, |
0499680a | 3312 | .permission = proc_pid_permission, |
1da177e4 | 3313 | }; |
1da177e4 | 3314 | |
0895e91d | 3315 | /** |
7bc3e6e5 EB |
3316 | * proc_flush_pid - Remove dcache entries for @pid from the /proc dcache. |
3317 | * @pid: pid that should be flushed. | |
0895e91d | 3318 | * |
7bc3e6e5 EB |
3319 | * This function walks a list of inodes (that belong to any proc |
3320 | * filesystem) that are attached to the pid and flushes them from | |
3321 | * the dentry cache. | |
0895e91d RD |
3322 | * |
3323 | * It is safe and reasonable to cache /proc entries for a task until | |
3324 | * that task exits. After that they just clog up the dcache with | |
3325 | * useless entries, possibly causing useful dcache entries to be | |
7bc3e6e5 EB |
3326 | * flushed instead. This routine is provided to flush those useless |
3327 | * dcache entries when a process is reaped. | |
0895e91d RD |
3328 | * |
3329 | * NOTE: This routine is just an optimization so it does not guarantee | |
7bc3e6e5 EB |
3330 | * that no dcache entries will exist after a process is reaped |
3331 | * it just makes it very unlikely that any will persist. | |
60347f67 PE |
3332 | */ |
3333 | ||
7bc3e6e5 | 3334 | void proc_flush_pid(struct pid *pid) |
60347f67 | 3335 | { |
63f818f4 | 3336 | proc_invalidate_siblings_dcache(&pid->inodes, &pid->lock); |
60347f67 PE |
3337 | } |
3338 | ||
0168b9e3 | 3339 | static struct dentry *proc_pid_instantiate(struct dentry * dentry, |
c52a47ac | 3340 | struct task_struct *task, const void *ptr) |
444ceed8 | 3341 | { |
444ceed8 EB |
3342 | struct inode *inode; |
3343 | ||
0168b9e3 | 3344 | inode = proc_pid_make_inode(dentry->d_sb, task, S_IFDIR | S_IRUGO | S_IXUGO); |
444ceed8 | 3345 | if (!inode) |
0168b9e3 | 3346 | return ERR_PTR(-ENOENT); |
444ceed8 | 3347 | |
444ceed8 EB |
3348 | inode->i_op = &proc_tgid_base_inode_operations; |
3349 | inode->i_fop = &proc_tgid_base_operations; | |
3350 | inode->i_flags|=S_IMMUTABLE; | |
aed54175 | 3351 | |
1270dd8d | 3352 | set_nlink(inode, nlink_tgid); |
1bbc5513 | 3353 | pid_update_inode(task, inode); |
444ceed8 | 3354 | |
fb045adb | 3355 | d_set_d_op(dentry, &pid_dentry_operations); |
0168b9e3 | 3356 | return d_splice_alias(inode, dentry); |
444ceed8 EB |
3357 | } |
3358 | ||
867aaccf | 3359 | struct dentry *proc_pid_lookup(struct dentry *dentry, unsigned int flags) |
1da177e4 LT |
3360 | { |
3361 | struct task_struct *task; | |
1da177e4 | 3362 | unsigned tgid; |
fa10fed3 | 3363 | struct proc_fs_info *fs_info; |
b488893a | 3364 | struct pid_namespace *ns; |
0168b9e3 | 3365 | struct dentry *result = ERR_PTR(-ENOENT); |
1da177e4 | 3366 | |
dbcdb504 | 3367 | tgid = name_to_int(&dentry->d_name); |
1da177e4 LT |
3368 | if (tgid == ~0U) |
3369 | goto out; | |
3370 | ||
fa10fed3 AG |
3371 | fs_info = proc_sb_info(dentry->d_sb); |
3372 | ns = fs_info->pid_ns; | |
de758734 | 3373 | rcu_read_lock(); |
b488893a | 3374 | task = find_task_by_pid_ns(tgid, ns); |
1da177e4 LT |
3375 | if (task) |
3376 | get_task_struct(task); | |
de758734 | 3377 | rcu_read_unlock(); |
1da177e4 LT |
3378 | if (!task) |
3379 | goto out; | |
3380 | ||
24a71ce5 AG |
3381 | /* Limit procfs to only ptraceable tasks */ |
3382 | if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE) { | |
3383 | if (!has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS)) | |
3384 | goto out_put_task; | |
3385 | } | |
3386 | ||
0168b9e3 | 3387 | result = proc_pid_instantiate(dentry, task, NULL); |
24a71ce5 | 3388 | out_put_task: |
1da177e4 | 3389 | put_task_struct(task); |
1da177e4 | 3390 | out: |
0168b9e3 | 3391 | return result; |
1da177e4 LT |
3392 | } |
3393 | ||
1da177e4 | 3394 | /* |
0804ef4b | 3395 | * Find the first task with tgid >= tgid |
0bc58a91 | 3396 | * |
1da177e4 | 3397 | */ |
19fd4bb2 EB |
3398 | struct tgid_iter { |
3399 | unsigned int tgid; | |
0804ef4b | 3400 | struct task_struct *task; |
19fd4bb2 EB |
3401 | }; |
3402 | static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter) | |
3403 | { | |
0804ef4b | 3404 | struct pid *pid; |
1da177e4 | 3405 | |
19fd4bb2 EB |
3406 | if (iter.task) |
3407 | put_task_struct(iter.task); | |
454cc105 | 3408 | rcu_read_lock(); |
0804ef4b | 3409 | retry: |
19fd4bb2 EB |
3410 | iter.task = NULL; |
3411 | pid = find_ge_pid(iter.tgid, ns); | |
0804ef4b | 3412 | if (pid) { |
19fd4bb2 | 3413 | iter.tgid = pid_nr_ns(pid, ns); |
3147d8aa EB |
3414 | iter.task = pid_task(pid, PIDTYPE_TGID); |
3415 | if (!iter.task) { | |
19fd4bb2 | 3416 | iter.tgid += 1; |
0804ef4b | 3417 | goto retry; |
19fd4bb2 EB |
3418 | } |
3419 | get_task_struct(iter.task); | |
0bc58a91 | 3420 | } |
454cc105 | 3421 | rcu_read_unlock(); |
19fd4bb2 | 3422 | return iter; |
1da177e4 LT |
3423 | } |
3424 | ||
0097875b | 3425 | #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2) |
0804ef4b | 3426 | |
1da177e4 | 3427 | /* for the /proc/ directory itself, after non-process stuff has been done */ |
f0c3b509 | 3428 | int proc_pid_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3429 | { |
19fd4bb2 | 3430 | struct tgid_iter iter; |
fa10fed3 | 3431 | struct proc_fs_info *fs_info = proc_sb_info(file_inode(file)->i_sb); |
9d78edea | 3432 | struct pid_namespace *ns = proc_pid_ns(file_inode(file)->i_sb); |
f0c3b509 | 3433 | loff_t pos = ctx->pos; |
1da177e4 | 3434 | |
021ada7d | 3435 | if (pos >= PID_MAX_LIMIT + TGID_OFFSET) |
f0c3b509 | 3436 | return 0; |
1da177e4 | 3437 | |
0097875b | 3438 | if (pos == TGID_OFFSET - 2) { |
fa10fed3 | 3439 | struct inode *inode = d_inode(fs_info->proc_self); |
db963164 | 3440 | if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK)) |
f0c3b509 | 3441 | return 0; |
0097875b EB |
3442 | ctx->pos = pos = pos + 1; |
3443 | } | |
3444 | if (pos == TGID_OFFSET - 1) { | |
fa10fed3 | 3445 | struct inode *inode = d_inode(fs_info->proc_thread_self); |
0097875b EB |
3446 | if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK)) |
3447 | return 0; | |
3448 | ctx->pos = pos = pos + 1; | |
021ada7d | 3449 | } |
0097875b | 3450 | iter.tgid = pos - TGID_OFFSET; |
19fd4bb2 | 3451 | iter.task = NULL; |
19fd4bb2 EB |
3452 | for (iter = next_tgid(ns, iter); |
3453 | iter.task; | |
3454 | iter.tgid += 1, iter = next_tgid(ns, iter)) { | |
e3912ac3 | 3455 | char name[10 + 1]; |
a4ef3895 | 3456 | unsigned int len; |
3ba4bcee ED |
3457 | |
3458 | cond_resched(); | |
fa10fed3 | 3459 | if (!has_pid_permissions(fs_info, iter.task, HIDEPID_INVISIBLE)) |
f0c3b509 | 3460 | continue; |
0499680a | 3461 | |
e3912ac3 | 3462 | len = snprintf(name, sizeof(name), "%u", iter.tgid); |
f0c3b509 AV |
3463 | ctx->pos = iter.tgid + TGID_OFFSET; |
3464 | if (!proc_fill_cache(file, ctx, name, len, | |
3465 | proc_pid_instantiate, iter.task, NULL)) { | |
19fd4bb2 | 3466 | put_task_struct(iter.task); |
f0c3b509 | 3467 | return 0; |
1da177e4 | 3468 | } |
0bc58a91 | 3469 | } |
f0c3b509 | 3470 | ctx->pos = PID_MAX_LIMIT + TGID_OFFSET; |
0bc58a91 EB |
3471 | return 0; |
3472 | } | |
1da177e4 | 3473 | |
1b3044e3 JD |
3474 | /* |
3475 | * proc_tid_comm_permission is a special permission function exclusively | |
3476 | * used for the node /proc/<pid>/task/<tid>/comm. | |
3477 | * It bypasses generic permission checks in the case where a task of the same | |
3478 | * task group attempts to access the node. | |
3479 | * The rationale behind this is that glibc and bionic access this node for | |
3480 | * cross thread naming (pthread_set/getname_np(!self)). However, if | |
3481 | * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0, | |
3482 | * which locks out the cross thread naming implementation. | |
3483 | * This function makes sure that the node is always accessible for members of | |
3484 | * same thread group. | |
3485 | */ | |
549c7297 CB |
3486 | static int proc_tid_comm_permission(struct user_namespace *mnt_userns, |
3487 | struct inode *inode, int mask) | |
1b3044e3 JD |
3488 | { |
3489 | bool is_same_tgroup; | |
3490 | struct task_struct *task; | |
3491 | ||
3492 | task = get_proc_task(inode); | |
3493 | if (!task) | |
3494 | return -ESRCH; | |
3495 | is_same_tgroup = same_thread_group(current, task); | |
3496 | put_task_struct(task); | |
3497 | ||
3498 | if (likely(is_same_tgroup && !(mask & MAY_EXEC))) { | |
3499 | /* This file (/proc/<pid>/task/<tid>/comm) can always be | |
3500 | * read or written by the members of the corresponding | |
3501 | * thread group. | |
3502 | */ | |
3503 | return 0; | |
3504 | } | |
3505 | ||
47291baa | 3506 | return generic_permission(&init_user_ns, inode, mask); |
1b3044e3 JD |
3507 | } |
3508 | ||
3509 | static const struct inode_operations proc_tid_comm_inode_operations = { | |
3510 | .permission = proc_tid_comm_permission, | |
3511 | }; | |
3512 | ||
28a6d671 EB |
3513 | /* |
3514 | * Tasks | |
3515 | */ | |
c5141e6d | 3516 | static const struct pid_entry tid_base_stuff[] = { |
631f9c18 | 3517 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), |
3835541d | 3518 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
6b4e306a | 3519 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
6ba8ed79 EB |
3520 | #ifdef CONFIG_NET |
3521 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), | |
3522 | #endif | |
631f9c18 | 3523 | REG("environ", S_IRUSR, proc_environ_operations), |
c5317167 | 3524 | REG("auxv", S_IRUSR, proc_auxv_operations), |
631f9c18 | 3525 | ONE("status", S_IRUGO, proc_pid_status), |
35a35046 | 3526 | ONE("personality", S_IRUSR, proc_pid_personality), |
1c963eb1 | 3527 | ONE("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 3528 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 3529 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
ebcb6734 | 3530 | #endif |
1b3044e3 JD |
3531 | NOD("comm", S_IFREG|S_IRUGO|S_IWUSR, |
3532 | &proc_tid_comm_inode_operations, | |
3533 | &proc_pid_set_comm_operations, {}), | |
ebcb6734 | 3534 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 | 3535 | ONE("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 3536 | #endif |
c2c0bb44 | 3537 | REG("cmdline", S_IRUGO, proc_pid_cmdline_ops), |
631f9c18 AD |
3538 | ONE("stat", S_IRUGO, proc_tid_stat), |
3539 | ONE("statm", S_IRUGO, proc_pid_statm), | |
871305bb | 3540 | REG("maps", S_IRUGO, proc_pid_maps_operations), |
2e13ba54 | 3541 | #ifdef CONFIG_PROC_CHILDREN |
81841161 CG |
3542 | REG("children", S_IRUGO, proc_tid_children_operations), |
3543 | #endif | |
28a6d671 | 3544 | #ifdef CONFIG_NUMA |
871305bb | 3545 | REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations), |
28a6d671 | 3546 | #endif |
631f9c18 AD |
3547 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
3548 | LNK("cwd", proc_cwd_link), | |
3549 | LNK("root", proc_root_link), | |
3550 | LNK("exe", proc_exe_link), | |
3551 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
3552 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
1e883281 | 3553 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 | 3554 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
871305bb | 3555 | REG("smaps", S_IRUGO, proc_pid_smaps_operations), |
493b0e9d | 3556 | REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations), |
32ed74a4 | 3557 | REG("pagemap", S_IRUSR, proc_pagemap_operations), |
28a6d671 EB |
3558 | #endif |
3559 | #ifdef CONFIG_SECURITY | |
631f9c18 | 3560 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
3561 | #endif |
3562 | #ifdef CONFIG_KALLSYMS | |
edfcd606 | 3563 | ONE("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 3564 | #endif |
2ec220e2 | 3565 | #ifdef CONFIG_STACKTRACE |
35a35046 | 3566 | ONE("stack", S_IRUSR, proc_pid_stack), |
28a6d671 | 3567 | #endif |
5968cece | 3568 | #ifdef CONFIG_SCHED_INFO |
f6e826ca | 3569 | ONE("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 3570 | #endif |
9745512c | 3571 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 3572 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 3573 | #endif |
8793d854 | 3574 | #ifdef CONFIG_PROC_PID_CPUSET |
52de4779 | 3575 | ONE("cpuset", S_IRUGO, proc_cpuset_show), |
a424316c PM |
3576 | #endif |
3577 | #ifdef CONFIG_CGROUPS | |
006f4ac4 | 3578 | ONE("cgroup", S_IRUGO, proc_cgroup_show), |
e79f15a4 CY |
3579 | #endif |
3580 | #ifdef CONFIG_PROC_CPU_RESCTRL | |
3581 | ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show), | |
28a6d671 | 3582 | #endif |
6ba51e37 | 3583 | ONE("oom_score", S_IRUGO, proc_oom_score), |
fa0cbbf1 | 3584 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations), |
a63d83f4 | 3585 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
4b7d248b | 3586 | #ifdef CONFIG_AUDIT |
631f9c18 | 3587 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
26ec3c64 | 3588 | REG("sessionid", S_IRUGO, proc_sessionid_operations), |
28a6d671 | 3589 | #endif |
f4f154fd | 3590 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 3591 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
1203c8e6 | 3592 | REG("fail-nth", 0644, proc_fail_nth_operations), |
f4f154fd | 3593 | #endif |
297c5d92 | 3594 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 3595 | ONE("io", S_IRUSR, proc_tid_io_accounting), |
297c5d92 | 3596 | #endif |
22d917d8 EB |
3597 | #ifdef CONFIG_USER_NS |
3598 | REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations), | |
3599 | REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations), | |
f76d207a | 3600 | REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations), |
9cc46516 | 3601 | REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations), |
22d917d8 | 3602 | #endif |
7c23b330 JP |
3603 | #ifdef CONFIG_LIVEPATCH |
3604 | ONE("patch_state", S_IRUSR, proc_pid_patch_state), | |
3605 | #endif | |
68bc30bb AL |
3606 | #ifdef CONFIG_PROC_PID_ARCH_STATUS |
3607 | ONE("arch_status", S_IRUGO, proc_pid_arch_status), | |
3608 | #endif | |
0d8315dd YZ |
3609 | #ifdef CONFIG_SECCOMP_CACHE_DEBUG |
3610 | ONE("seccomp_cache", S_IRUSR, proc_pid_seccomp_cache), | |
3611 | #endif | |
28a6d671 EB |
3612 | }; |
3613 | ||
f0c3b509 | 3614 | static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx) |
28a6d671 | 3615 | { |
f0c3b509 AV |
3616 | return proc_pident_readdir(file, ctx, |
3617 | tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
3618 | } |
3619 | ||
00cd8dd3 AV |
3620 | static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
3621 | { | |
7bcd6b0e | 3622 | return proc_pident_lookup(dir, dentry, |
d5a572a4 AD |
3623 | tid_base_stuff, |
3624 | tid_base_stuff + ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
3625 | } |
3626 | ||
00977a59 | 3627 | static const struct file_operations proc_tid_base_operations = { |
28a6d671 | 3628 | .read = generic_read_dir, |
f50752ea AV |
3629 | .iterate_shared = proc_tid_base_readdir, |
3630 | .llseek = generic_file_llseek, | |
28a6d671 EB |
3631 | }; |
3632 | ||
c5ef1c42 | 3633 | static const struct inode_operations proc_tid_base_inode_operations = { |
28a6d671 EB |
3634 | .lookup = proc_tid_base_lookup, |
3635 | .getattr = pid_getattr, | |
3636 | .setattr = proc_setattr, | |
3637 | }; | |
3638 | ||
0168b9e3 AV |
3639 | static struct dentry *proc_task_instantiate(struct dentry *dentry, |
3640 | struct task_struct *task, const void *ptr) | |
444ceed8 | 3641 | { |
444ceed8 | 3642 | struct inode *inode; |
0168b9e3 | 3643 | inode = proc_pid_make_inode(dentry->d_sb, task, S_IFDIR | S_IRUGO | S_IXUGO); |
444ceed8 | 3644 | if (!inode) |
0168b9e3 | 3645 | return ERR_PTR(-ENOENT); |
1bbc5513 | 3646 | |
444ceed8 EB |
3647 | inode->i_op = &proc_tid_base_inode_operations; |
3648 | inode->i_fop = &proc_tid_base_operations; | |
1bbc5513 | 3649 | inode->i_flags |= S_IMMUTABLE; |
aed54175 | 3650 | |
1270dd8d | 3651 | set_nlink(inode, nlink_tid); |
1bbc5513 | 3652 | pid_update_inode(task, inode); |
444ceed8 | 3653 | |
fb045adb | 3654 | d_set_d_op(dentry, &pid_dentry_operations); |
0168b9e3 | 3655 | return d_splice_alias(inode, dentry); |
444ceed8 EB |
3656 | } |
3657 | ||
00cd8dd3 | 3658 | static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags) |
28a6d671 | 3659 | { |
28a6d671 EB |
3660 | struct task_struct *task; |
3661 | struct task_struct *leader = get_proc_task(dir); | |
28a6d671 | 3662 | unsigned tid; |
fa10fed3 | 3663 | struct proc_fs_info *fs_info; |
b488893a | 3664 | struct pid_namespace *ns; |
0168b9e3 | 3665 | struct dentry *result = ERR_PTR(-ENOENT); |
28a6d671 EB |
3666 | |
3667 | if (!leader) | |
3668 | goto out_no_task; | |
3669 | ||
dbcdb504 | 3670 | tid = name_to_int(&dentry->d_name); |
28a6d671 EB |
3671 | if (tid == ~0U) |
3672 | goto out; | |
3673 | ||
fa10fed3 AG |
3674 | fs_info = proc_sb_info(dentry->d_sb); |
3675 | ns = fs_info->pid_ns; | |
28a6d671 | 3676 | rcu_read_lock(); |
b488893a | 3677 | task = find_task_by_pid_ns(tid, ns); |
28a6d671 EB |
3678 | if (task) |
3679 | get_task_struct(task); | |
3680 | rcu_read_unlock(); | |
3681 | if (!task) | |
3682 | goto out; | |
bac0abd6 | 3683 | if (!same_thread_group(leader, task)) |
28a6d671 EB |
3684 | goto out_drop_task; |
3685 | ||
0168b9e3 | 3686 | result = proc_task_instantiate(dentry, task, NULL); |
28a6d671 EB |
3687 | out_drop_task: |
3688 | put_task_struct(task); | |
3689 | out: | |
3690 | put_task_struct(leader); | |
3691 | out_no_task: | |
0168b9e3 | 3692 | return result; |
28a6d671 EB |
3693 | } |
3694 | ||
0bc58a91 EB |
3695 | /* |
3696 | * Find the first tid of a thread group to return to user space. | |
3697 | * | |
3698 | * Usually this is just the thread group leader, but if the users | |
3699 | * buffer was too small or there was a seek into the middle of the | |
3700 | * directory we have more work todo. | |
3701 | * | |
3702 | * In the case of a short read we start with find_task_by_pid. | |
3703 | * | |
3704 | * In the case of a seek we start with the leader and walk nr | |
3705 | * threads past it. | |
3706 | */ | |
9f6e963f ON |
3707 | static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos, |
3708 | struct pid_namespace *ns) | |
0bc58a91 | 3709 | { |
d855a4b7 | 3710 | struct task_struct *pos, *task; |
9f6e963f ON |
3711 | unsigned long nr = f_pos; |
3712 | ||
3713 | if (nr != f_pos) /* 32bit overflow? */ | |
3714 | return NULL; | |
1da177e4 | 3715 | |
cc288738 | 3716 | rcu_read_lock(); |
d855a4b7 ON |
3717 | task = pid_task(pid, PIDTYPE_PID); |
3718 | if (!task) | |
3719 | goto fail; | |
3720 | ||
3721 | /* Attempt to start with the tid of a thread */ | |
9f6e963f | 3722 | if (tid && nr) { |
b488893a | 3723 | pos = find_task_by_pid_ns(tid, ns); |
d855a4b7 | 3724 | if (pos && same_thread_group(pos, task)) |
a872ff0c | 3725 | goto found; |
0bc58a91 | 3726 | } |
1da177e4 | 3727 | |
0bc58a91 | 3728 | /* If nr exceeds the number of threads there is nothing todo */ |
9f6e963f | 3729 | if (nr >= get_nr_threads(task)) |
c986c14a | 3730 | goto fail; |
1da177e4 | 3731 | |
a872ff0c ON |
3732 | /* If we haven't found our starting place yet start |
3733 | * with the leader and walk nr threads forward. | |
0bc58a91 | 3734 | */ |
d855a4b7 | 3735 | pos = task = task->group_leader; |
c986c14a | 3736 | do { |
9f6e963f | 3737 | if (!nr--) |
c986c14a | 3738 | goto found; |
d855a4b7 | 3739 | } while_each_thread(task, pos); |
c986c14a ON |
3740 | fail: |
3741 | pos = NULL; | |
3742 | goto out; | |
a872ff0c ON |
3743 | found: |
3744 | get_task_struct(pos); | |
3745 | out: | |
cc288738 | 3746 | rcu_read_unlock(); |
0bc58a91 EB |
3747 | return pos; |
3748 | } | |
3749 | ||
3750 | /* | |
3751 | * Find the next thread in the thread list. | |
3752 | * Return NULL if there is an error or no next thread. | |
3753 | * | |
3754 | * The reference to the input task_struct is released. | |
3755 | */ | |
3756 | static struct task_struct *next_tid(struct task_struct *start) | |
3757 | { | |
c1df7fb8 | 3758 | struct task_struct *pos = NULL; |
cc288738 | 3759 | rcu_read_lock(); |
c1df7fb8 | 3760 | if (pid_alive(start)) { |
0bc58a91 | 3761 | pos = next_thread(start); |
c1df7fb8 ON |
3762 | if (thread_group_leader(pos)) |
3763 | pos = NULL; | |
3764 | else | |
3765 | get_task_struct(pos); | |
3766 | } | |
cc288738 | 3767 | rcu_read_unlock(); |
0bc58a91 EB |
3768 | put_task_struct(start); |
3769 | return pos; | |
1da177e4 LT |
3770 | } |
3771 | ||
3772 | /* for the /proc/TGID/task/ directories */ | |
f0c3b509 | 3773 | static int proc_task_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3774 | { |
d855a4b7 ON |
3775 | struct inode *inode = file_inode(file); |
3776 | struct task_struct *task; | |
b488893a | 3777 | struct pid_namespace *ns; |
f0c3b509 | 3778 | int tid; |
1da177e4 | 3779 | |
d855a4b7 | 3780 | if (proc_inode_is_dead(inode)) |
f0c3b509 | 3781 | return -ENOENT; |
1da177e4 | 3782 | |
f0c3b509 | 3783 | if (!dir_emit_dots(file, ctx)) |
d855a4b7 | 3784 | return 0; |
1da177e4 | 3785 | |
0bc58a91 EB |
3786 | /* f_version caches the tgid value that the last readdir call couldn't |
3787 | * return. lseek aka telldir automagically resets f_version to 0. | |
3788 | */ | |
9d78edea | 3789 | ns = proc_pid_ns(inode->i_sb); |
f0c3b509 AV |
3790 | tid = (int)file->f_version; |
3791 | file->f_version = 0; | |
d855a4b7 | 3792 | for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns); |
0bc58a91 | 3793 | task; |
f0c3b509 | 3794 | task = next_tid(task), ctx->pos++) { |
e3912ac3 | 3795 | char name[10 + 1]; |
a4ef3895 | 3796 | unsigned int len; |
b488893a | 3797 | tid = task_pid_nr_ns(task, ns); |
e3912ac3 | 3798 | len = snprintf(name, sizeof(name), "%u", tid); |
f0c3b509 AV |
3799 | if (!proc_fill_cache(file, ctx, name, len, |
3800 | proc_task_instantiate, task, NULL)) { | |
0bc58a91 EB |
3801 | /* returning this tgid failed, save it as the first |
3802 | * pid for the next readir call */ | |
f0c3b509 | 3803 | file->f_version = (u64)tid; |
0bc58a91 | 3804 | put_task_struct(task); |
1da177e4 | 3805 | break; |
0bc58a91 | 3806 | } |
1da177e4 | 3807 | } |
d855a4b7 | 3808 | |
f0c3b509 | 3809 | return 0; |
1da177e4 | 3810 | } |
6e66b52b | 3811 | |
549c7297 CB |
3812 | static int proc_task_getattr(struct user_namespace *mnt_userns, |
3813 | const struct path *path, struct kstat *stat, | |
a528d35e | 3814 | u32 request_mask, unsigned int query_flags) |
6e66b52b | 3815 | { |
a528d35e | 3816 | struct inode *inode = d_inode(path->dentry); |
99f89551 | 3817 | struct task_struct *p = get_proc_task(inode); |
0d56a451 | 3818 | generic_fillattr(&init_user_ns, inode, stat); |
6e66b52b | 3819 | |
99f89551 | 3820 | if (p) { |
99f89551 | 3821 | stat->nlink += get_nr_threads(p); |
99f89551 | 3822 | put_task_struct(p); |
6e66b52b EB |
3823 | } |
3824 | ||
3825 | return 0; | |
3826 | } | |
28a6d671 | 3827 | |
c5ef1c42 | 3828 | static const struct inode_operations proc_task_inode_operations = { |
28a6d671 EB |
3829 | .lookup = proc_task_lookup, |
3830 | .getattr = proc_task_getattr, | |
3831 | .setattr = proc_setattr, | |
0499680a | 3832 | .permission = proc_pid_permission, |
28a6d671 EB |
3833 | }; |
3834 | ||
00977a59 | 3835 | static const struct file_operations proc_task_operations = { |
28a6d671 | 3836 | .read = generic_read_dir, |
f50752ea AV |
3837 | .iterate_shared = proc_task_readdir, |
3838 | .llseek = generic_file_llseek, | |
28a6d671 | 3839 | }; |
1270dd8d AD |
3840 | |
3841 | void __init set_proc_pid_nlink(void) | |
3842 | { | |
3843 | nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
3844 | nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
3845 | } |