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> |
54354c6a | 70 | #include <linux/kallsyms.h> |
2ec220e2 | 71 | #include <linux/stacktrace.h> |
d85f50d5 | 72 | #include <linux/resource.h> |
5096add8 | 73 | #include <linux/module.h> |
1da177e4 LT |
74 | #include <linux/mount.h> |
75 | #include <linux/security.h> | |
76 | #include <linux/ptrace.h> | |
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> |
c2f273eb | 98 | #include <linux/cn_proc.h> |
43d2b113 | 99 | #include <trace/events/oom.h> |
1da177e4 | 100 | #include "internal.h" |
faf60af1 | 101 | #include "fd.h" |
1da177e4 | 102 | |
ac7f1061 AD |
103 | #include "../../lib/kstrtox.h" |
104 | ||
0f2fe20f EB |
105 | /* NOTE: |
106 | * Implementing inode permission operations in /proc is almost | |
107 | * certainly an error. Permission checks need to happen during | |
108 | * each system call not at open time. The reason is that most of | |
109 | * what we wish to check for permissions in /proc varies at runtime. | |
110 | * | |
111 | * The classic example of a problem is opening file descriptors | |
112 | * in /proc for a task before it execs a suid executable. | |
113 | */ | |
114 | ||
efb1a57d AD |
115 | static u8 nlink_tid __ro_after_init; |
116 | static u8 nlink_tgid __ro_after_init; | |
1270dd8d | 117 | |
1da177e4 | 118 | struct pid_entry { |
cedbccab | 119 | const char *name; |
623f594e | 120 | unsigned int len; |
d161a13f | 121 | umode_t mode; |
c5ef1c42 | 122 | const struct inode_operations *iop; |
00977a59 | 123 | const struct file_operations *fop; |
20cdc894 | 124 | union proc_op op; |
1da177e4 LT |
125 | }; |
126 | ||
61a28784 | 127 | #define NOD(NAME, MODE, IOP, FOP, OP) { \ |
20cdc894 | 128 | .name = (NAME), \ |
c5141e6d | 129 | .len = sizeof(NAME) - 1, \ |
20cdc894 EB |
130 | .mode = MODE, \ |
131 | .iop = IOP, \ | |
132 | .fop = FOP, \ | |
133 | .op = OP, \ | |
134 | } | |
135 | ||
631f9c18 AD |
136 | #define DIR(NAME, MODE, iops, fops) \ |
137 | NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} ) | |
138 | #define LNK(NAME, get_link) \ | |
61a28784 | 139 | NOD(NAME, (S_IFLNK|S_IRWXUGO), \ |
20cdc894 | 140 | &proc_pid_link_inode_operations, NULL, \ |
631f9c18 AD |
141 | { .proc_get_link = get_link } ) |
142 | #define REG(NAME, MODE, fops) \ | |
143 | NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {}) | |
631f9c18 | 144 | #define ONE(NAME, MODE, show) \ |
6d9c939d | 145 | NOD(NAME, (S_IFREG|(MODE)), \ |
be614086 | 146 | NULL, &proc_single_file_operations, \ |
631f9c18 | 147 | { .proc_show = show } ) |
6d9c939d CS |
148 | #define ATTR(LSM, NAME, MODE) \ |
149 | NOD(NAME, (S_IFREG|(MODE)), \ | |
150 | NULL, &proc_pid_attr_operations, \ | |
151 | { .lsm = LSM }) | |
1da177e4 | 152 | |
aed54175 VN |
153 | /* |
154 | * Count the number of hardlinks for the pid_entry table, excluding the . | |
155 | * and .. links. | |
156 | */ | |
1270dd8d | 157 | static unsigned int __init pid_entry_nlink(const struct pid_entry *entries, |
aed54175 VN |
158 | unsigned int n) |
159 | { | |
160 | unsigned int i; | |
161 | unsigned int count; | |
162 | ||
1270dd8d | 163 | count = 2; |
aed54175 VN |
164 | for (i = 0; i < n; ++i) { |
165 | if (S_ISDIR(entries[i].mode)) | |
166 | ++count; | |
167 | } | |
168 | ||
169 | return count; | |
170 | } | |
171 | ||
f7ad3c6b | 172 | static int get_task_root(struct task_struct *task, struct path *root) |
1da177e4 | 173 | { |
7c2c7d99 HD |
174 | int result = -ENOENT; |
175 | ||
0494f6ec | 176 | task_lock(task); |
f7ad3c6b MS |
177 | if (task->fs) { |
178 | get_fs_root(task->fs, root); | |
7c2c7d99 HD |
179 | result = 0; |
180 | } | |
0494f6ec | 181 | task_unlock(task); |
7c2c7d99 | 182 | return result; |
0494f6ec MS |
183 | } |
184 | ||
7773fbc5 | 185 | static int proc_cwd_link(struct dentry *dentry, struct path *path) |
0494f6ec | 186 | { |
2b0143b5 | 187 | struct task_struct *task = get_proc_task(d_inode(dentry)); |
0494f6ec | 188 | int result = -ENOENT; |
99f89551 EB |
189 | |
190 | if (task) { | |
f7ad3c6b MS |
191 | task_lock(task); |
192 | if (task->fs) { | |
193 | get_fs_pwd(task->fs, path); | |
194 | result = 0; | |
195 | } | |
196 | task_unlock(task); | |
99f89551 EB |
197 | put_task_struct(task); |
198 | } | |
1da177e4 LT |
199 | return result; |
200 | } | |
201 | ||
7773fbc5 | 202 | static int proc_root_link(struct dentry *dentry, struct path *path) |
1da177e4 | 203 | { |
2b0143b5 | 204 | struct task_struct *task = get_proc_task(d_inode(dentry)); |
1da177e4 | 205 | int result = -ENOENT; |
99f89551 EB |
206 | |
207 | if (task) { | |
f7ad3c6b | 208 | result = get_task_root(task, path); |
99f89551 EB |
209 | put_task_struct(task); |
210 | } | |
1da177e4 LT |
211 | return result; |
212 | } | |
213 | ||
d26d0cd9 LT |
214 | /* |
215 | * If the user used setproctitle(), we just get the string from | |
216 | * user space at arg_start, and limit it to a maximum of one page. | |
217 | */ | |
218 | static ssize_t get_mm_proctitle(struct mm_struct *mm, char __user *buf, | |
219 | size_t count, unsigned long pos, | |
220 | unsigned long arg_start) | |
221 | { | |
222 | char *page; | |
223 | int ret, got; | |
224 | ||
225 | if (pos >= PAGE_SIZE) | |
226 | return 0; | |
227 | ||
228 | page = (char *)__get_free_page(GFP_KERNEL); | |
229 | if (!page) | |
230 | return -ENOMEM; | |
231 | ||
232 | ret = 0; | |
233 | got = access_remote_vm(mm, arg_start, page, PAGE_SIZE, FOLL_ANON); | |
234 | if (got > 0) { | |
235 | int len = strnlen(page, got); | |
236 | ||
237 | /* Include the NUL character if it was found */ | |
238 | if (len < got) | |
239 | len++; | |
240 | ||
241 | if (len > pos) { | |
242 | len -= pos; | |
243 | if (len > count) | |
244 | len = count; | |
245 | len -= copy_to_user(buf, page+pos, len); | |
246 | if (!len) | |
247 | len = -EFAULT; | |
248 | ret = len; | |
249 | } | |
250 | } | |
251 | free_page((unsigned long)page); | |
252 | return ret; | |
253 | } | |
254 | ||
e4b4e441 | 255 | static ssize_t get_mm_cmdline(struct mm_struct *mm, char __user *buf, |
5ab82718 | 256 | size_t count, loff_t *ppos) |
1da177e4 | 257 | { |
c2c0bb44 | 258 | unsigned long arg_start, arg_end, env_start, env_end; |
5ab82718 | 259 | unsigned long pos, len; |
d26d0cd9 | 260 | char *page, c; |
c2c0bb44 | 261 | |
c2c0bb44 | 262 | /* Check if process spawned far enough to have cmdline. */ |
e4b4e441 LT |
263 | if (!mm->env_end) |
264 | return 0; | |
c2c0bb44 | 265 | |
88aa7cc6 | 266 | spin_lock(&mm->arg_lock); |
c2c0bb44 AD |
267 | arg_start = mm->arg_start; |
268 | arg_end = mm->arg_end; | |
269 | env_start = mm->env_start; | |
270 | env_end = mm->env_end; | |
88aa7cc6 | 271 | spin_unlock(&mm->arg_lock); |
c2c0bb44 | 272 | |
5ab82718 LT |
273 | if (arg_start >= arg_end) |
274 | return 0; | |
6a6cbe75 | 275 | |
2ca66ff7 | 276 | /* |
d26d0cd9 LT |
277 | * We allow setproctitle() to overwrite the argument |
278 | * strings, and overflow past the original end. But | |
279 | * only when it overflows into the environment area. | |
2ca66ff7 | 280 | */ |
d26d0cd9 | 281 | if (env_start != arg_end || env_end < env_start) |
5ab82718 | 282 | env_start = env_end = arg_end; |
d26d0cd9 | 283 | len = env_end - arg_start; |
f5b65348 | 284 | |
5ab82718 | 285 | /* We're not going to care if "*ppos" has high bits set */ |
d26d0cd9 LT |
286 | pos = *ppos; |
287 | if (pos >= len) | |
5ab82718 | 288 | return 0; |
d26d0cd9 LT |
289 | if (count > len - pos) |
290 | count = len - pos; | |
291 | if (!count) | |
292 | return 0; | |
293 | ||
294 | /* | |
295 | * Magical special case: if the argv[] end byte is not | |
296 | * zero, the user has overwritten it with setproctitle(3). | |
297 | * | |
298 | * Possible future enhancement: do this only once when | |
299 | * pos is 0, and set a flag in the 'struct file'. | |
300 | */ | |
301 | if (access_remote_vm(mm, arg_end-1, &c, 1, FOLL_ANON) == 1 && c) | |
302 | return get_mm_proctitle(mm, buf, count, pos, arg_start); | |
a0a07b87 | 303 | |
d26d0cd9 LT |
304 | /* |
305 | * For the non-setproctitle() case we limit things strictly | |
306 | * to the [arg_start, arg_end[ range. | |
307 | */ | |
308 | pos += arg_start; | |
3d712546 | 309 | if (pos < arg_start || pos >= arg_end) |
5ab82718 | 310 | return 0; |
3d712546 LT |
311 | if (count > arg_end - pos) |
312 | count = arg_end - pos; | |
5ab82718 LT |
313 | |
314 | page = (char *)__get_free_page(GFP_KERNEL); | |
315 | if (!page) | |
316 | return -ENOMEM; | |
317 | ||
318 | len = 0; | |
319 | while (count) { | |
320 | int got; | |
321 | size_t size = min_t(size_t, PAGE_SIZE, count); | |
322 | ||
3d712546 LT |
323 | got = access_remote_vm(mm, pos, page, size, FOLL_ANON); |
324 | if (got <= 0) | |
5ab82718 | 325 | break; |
3d712546 | 326 | got -= copy_to_user(buf, page, got); |
5ab82718 LT |
327 | if (unlikely(!got)) { |
328 | if (!len) | |
329 | len = -EFAULT; | |
330 | break; | |
c2c0bb44 | 331 | } |
5ab82718 LT |
332 | pos += got; |
333 | buf += got; | |
334 | len += got; | |
335 | count -= got; | |
c2c0bb44 AD |
336 | } |
337 | ||
c2c0bb44 | 338 | free_page((unsigned long)page); |
5ab82718 | 339 | return len; |
1da177e4 LT |
340 | } |
341 | ||
e4b4e441 LT |
342 | static ssize_t get_task_cmdline(struct task_struct *tsk, char __user *buf, |
343 | size_t count, loff_t *pos) | |
344 | { | |
345 | struct mm_struct *mm; | |
346 | ssize_t ret; | |
347 | ||
348 | mm = get_task_mm(tsk); | |
349 | if (!mm) | |
350 | return 0; | |
351 | ||
352 | ret = get_mm_cmdline(mm, buf, count, pos); | |
c2c0bb44 | 353 | mmput(mm); |
e4b4e441 LT |
354 | return ret; |
355 | } | |
356 | ||
357 | static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf, | |
358 | size_t count, loff_t *pos) | |
359 | { | |
360 | struct task_struct *tsk; | |
361 | ssize_t ret; | |
362 | ||
363 | BUG_ON(*pos < 0); | |
364 | ||
365 | tsk = get_proc_task(file_inode(file)); | |
366 | if (!tsk) | |
367 | return -ESRCH; | |
368 | ret = get_task_cmdline(tsk, buf, count, pos); | |
369 | put_task_struct(tsk); | |
370 | if (ret > 0) | |
371 | *pos += ret; | |
372 | return ret; | |
1da177e4 LT |
373 | } |
374 | ||
c2c0bb44 AD |
375 | static const struct file_operations proc_pid_cmdline_ops = { |
376 | .read = proc_pid_cmdline_read, | |
377 | .llseek = generic_file_llseek, | |
378 | }; | |
379 | ||
1da177e4 LT |
380 | #ifdef CONFIG_KALLSYMS |
381 | /* | |
382 | * Provides a wchan file via kallsyms in a proper one-value-per-file format. | |
383 | * Returns the resolved symbol. If that fails, simply return the address. | |
384 | */ | |
edfcd606 AD |
385 | static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns, |
386 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 387 | { |
ffb45122 | 388 | unsigned long wchan; |
54354c6a | 389 | char symname[KSYM_NAME_LEN]; |
1da177e4 | 390 | |
54354c6a KC |
391 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
392 | goto print0; | |
1da177e4 | 393 | |
54354c6a KC |
394 | wchan = get_wchan(task); |
395 | if (wchan && !lookup_symbol_name(wchan, symname)) { | |
396 | seq_puts(m, symname); | |
397 | return 0; | |
398 | } | |
25ce3191 | 399 | |
54354c6a KC |
400 | print0: |
401 | seq_putc(m, '0'); | |
25ce3191 | 402 | return 0; |
1da177e4 LT |
403 | } |
404 | #endif /* CONFIG_KALLSYMS */ | |
405 | ||
a9712bc1 AV |
406 | static int lock_trace(struct task_struct *task) |
407 | { | |
f7cfd871 | 408 | int err = down_read_killable(&task->signal->exec_update_lock); |
a9712bc1 AV |
409 | if (err) |
410 | return err; | |
caaee623 | 411 | if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) { |
f7cfd871 | 412 | up_read(&task->signal->exec_update_lock); |
a9712bc1 AV |
413 | return -EPERM; |
414 | } | |
415 | return 0; | |
416 | } | |
417 | ||
418 | static void unlock_trace(struct task_struct *task) | |
419 | { | |
f7cfd871 | 420 | up_read(&task->signal->exec_update_lock); |
a9712bc1 AV |
421 | } |
422 | ||
2ec220e2 KC |
423 | #ifdef CONFIG_STACKTRACE |
424 | ||
425 | #define MAX_STACK_TRACE_DEPTH 64 | |
426 | ||
427 | static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns, | |
428 | struct pid *pid, struct task_struct *task) | |
429 | { | |
2ec220e2 | 430 | unsigned long *entries; |
a9712bc1 | 431 | int err; |
2ec220e2 | 432 | |
f8a00cef JH |
433 | /* |
434 | * The ability to racily run the kernel stack unwinder on a running task | |
435 | * and then observe the unwinder output is scary; while it is useful for | |
436 | * debugging kernel issues, it can also allow an attacker to leak kernel | |
437 | * stack contents. | |
438 | * Doing this in a manner that is at least safe from races would require | |
439 | * some work to ensure that the remote task can not be scheduled; and | |
440 | * even then, this would still expose the unwinder as local attack | |
441 | * surface. | |
442 | * Therefore, this interface is restricted to root. | |
443 | */ | |
444 | if (!file_ns_capable(m->file, &init_user_ns, CAP_SYS_ADMIN)) | |
445 | return -EACCES; | |
446 | ||
6da2ec56 KC |
447 | entries = kmalloc_array(MAX_STACK_TRACE_DEPTH, sizeof(*entries), |
448 | GFP_KERNEL); | |
2ec220e2 KC |
449 | if (!entries) |
450 | return -ENOMEM; | |
451 | ||
a9712bc1 AV |
452 | err = lock_trace(task); |
453 | if (!err) { | |
e988e5ec | 454 | unsigned int i, nr_entries; |
5d008fb4 | 455 | |
e988e5ec TG |
456 | nr_entries = stack_trace_save_tsk(task, entries, |
457 | MAX_STACK_TRACE_DEPTH, 0); | |
a9712bc1 | 458 | |
e988e5ec | 459 | for (i = 0; i < nr_entries; i++) { |
8f5abe84 | 460 | seq_printf(m, "[<0>] %pB\n", (void *)entries[i]); |
a9712bc1 | 461 | } |
e988e5ec | 462 | |
a9712bc1 | 463 | unlock_trace(task); |
2ec220e2 KC |
464 | } |
465 | kfree(entries); | |
466 | ||
a9712bc1 | 467 | return err; |
2ec220e2 KC |
468 | } |
469 | #endif | |
470 | ||
5968cece | 471 | #ifdef CONFIG_SCHED_INFO |
1da177e4 LT |
472 | /* |
473 | * Provides /proc/PID/schedstat | |
474 | */ | |
f6e826ca AD |
475 | static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns, |
476 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 477 | { |
5968cece | 478 | if (unlikely(!sched_info_on())) |
08b55775 | 479 | seq_puts(m, "0 0 0\n"); |
5968cece NR |
480 | else |
481 | seq_printf(m, "%llu %llu %lu\n", | |
25ce3191 JP |
482 | (unsigned long long)task->se.sum_exec_runtime, |
483 | (unsigned long long)task->sched_info.run_delay, | |
484 | task->sched_info.pcount); | |
485 | ||
486 | return 0; | |
1da177e4 LT |
487 | } |
488 | #endif | |
489 | ||
9745512c AV |
490 | #ifdef CONFIG_LATENCYTOP |
491 | static int lstats_show_proc(struct seq_file *m, void *v) | |
492 | { | |
493 | int i; | |
13d77c37 HS |
494 | struct inode *inode = m->private; |
495 | struct task_struct *task = get_proc_task(inode); | |
9745512c | 496 | |
13d77c37 HS |
497 | if (!task) |
498 | return -ESRCH; | |
499 | seq_puts(m, "Latency Top version : v0.1\n"); | |
f6d2f584 | 500 | for (i = 0; i < LT_SAVECOUNT; i++) { |
34e49d4f JP |
501 | struct latency_record *lr = &task->latency_record[i]; |
502 | if (lr->backtrace[0]) { | |
9745512c | 503 | int q; |
34e49d4f JP |
504 | seq_printf(m, "%i %li %li", |
505 | lr->count, lr->time, lr->max); | |
9745512c | 506 | for (q = 0; q < LT_BACKTRACEDEPTH; q++) { |
34e49d4f | 507 | unsigned long bt = lr->backtrace[q]; |
accddc41 | 508 | |
34e49d4f | 509 | if (!bt) |
9745512c | 510 | break; |
34e49d4f | 511 | seq_printf(m, " %ps", (void *)bt); |
9745512c | 512 | } |
9d6de12f | 513 | seq_putc(m, '\n'); |
9745512c AV |
514 | } |
515 | ||
516 | } | |
13d77c37 | 517 | put_task_struct(task); |
9745512c AV |
518 | return 0; |
519 | } | |
520 | ||
521 | static int lstats_open(struct inode *inode, struct file *file) | |
522 | { | |
13d77c37 | 523 | return single_open(file, lstats_show_proc, inode); |
d6643d12 HS |
524 | } |
525 | ||
9745512c AV |
526 | static ssize_t lstats_write(struct file *file, const char __user *buf, |
527 | size_t count, loff_t *offs) | |
528 | { | |
496ad9aa | 529 | struct task_struct *task = get_proc_task(file_inode(file)); |
9745512c | 530 | |
13d77c37 HS |
531 | if (!task) |
532 | return -ESRCH; | |
e02c9b0d | 533 | clear_tsk_latency_tracing(task); |
13d77c37 | 534 | put_task_struct(task); |
9745512c AV |
535 | |
536 | return count; | |
537 | } | |
538 | ||
539 | static const struct file_operations proc_lstats_operations = { | |
540 | .open = lstats_open, | |
541 | .read = seq_read, | |
542 | .write = lstats_write, | |
543 | .llseek = seq_lseek, | |
13d77c37 | 544 | .release = single_release, |
9745512c AV |
545 | }; |
546 | ||
547 | #endif | |
548 | ||
6ba51e37 AD |
549 | static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns, |
550 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 551 | { |
ca79b0c2 | 552 | unsigned long totalpages = totalram_pages() + total_swap_pages; |
b95c35e7 | 553 | unsigned long points = 0; |
9066e5cf YS |
554 | long badness; |
555 | ||
556 | badness = oom_badness(task, totalpages); | |
557 | /* | |
558 | * Special case OOM_SCORE_ADJ_MIN for all others scale the | |
559 | * badness value into [0, 2000] range which we have been | |
560 | * exporting for a long time so userspace might depend on it. | |
561 | */ | |
562 | if (badness != LONG_MIN) | |
563 | points = (1000 + badness * 1000 / (long)totalpages) * 2 / 3; | |
1da177e4 | 564 | |
25ce3191 JP |
565 | seq_printf(m, "%lu\n", points); |
566 | ||
567 | return 0; | |
1da177e4 LT |
568 | } |
569 | ||
d85f50d5 | 570 | struct limit_names { |
cedbccab AD |
571 | const char *name; |
572 | const char *unit; | |
d85f50d5 NH |
573 | }; |
574 | ||
575 | static const struct limit_names lnames[RLIM_NLIMITS] = { | |
cff4edb5 | 576 | [RLIMIT_CPU] = {"Max cpu time", "seconds"}, |
d85f50d5 NH |
577 | [RLIMIT_FSIZE] = {"Max file size", "bytes"}, |
578 | [RLIMIT_DATA] = {"Max data size", "bytes"}, | |
579 | [RLIMIT_STACK] = {"Max stack size", "bytes"}, | |
580 | [RLIMIT_CORE] = {"Max core file size", "bytes"}, | |
581 | [RLIMIT_RSS] = {"Max resident set", "bytes"}, | |
582 | [RLIMIT_NPROC] = {"Max processes", "processes"}, | |
583 | [RLIMIT_NOFILE] = {"Max open files", "files"}, | |
584 | [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"}, | |
585 | [RLIMIT_AS] = {"Max address space", "bytes"}, | |
586 | [RLIMIT_LOCKS] = {"Max file locks", "locks"}, | |
587 | [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"}, | |
588 | [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"}, | |
589 | [RLIMIT_NICE] = {"Max nice priority", NULL}, | |
590 | [RLIMIT_RTPRIO] = {"Max realtime priority", NULL}, | |
8808117c | 591 | [RLIMIT_RTTIME] = {"Max realtime timeout", "us"}, |
d85f50d5 NH |
592 | }; |
593 | ||
594 | /* Display limits for a process */ | |
1c963eb1 AD |
595 | static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns, |
596 | struct pid *pid, struct task_struct *task) | |
d85f50d5 NH |
597 | { |
598 | unsigned int i; | |
d85f50d5 | 599 | unsigned long flags; |
d85f50d5 NH |
600 | |
601 | struct rlimit rlim[RLIM_NLIMITS]; | |
602 | ||
a6bebbc8 | 603 | if (!lock_task_sighand(task, &flags)) |
d85f50d5 | 604 | return 0; |
d85f50d5 NH |
605 | memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS); |
606 | unlock_task_sighand(task, &flags); | |
d85f50d5 NH |
607 | |
608 | /* | |
609 | * print the file header | |
610 | */ | |
afe922c2 AD |
611 | seq_puts(m, "Limit " |
612 | "Soft Limit " | |
613 | "Hard Limit " | |
614 | "Units \n"); | |
d85f50d5 NH |
615 | |
616 | for (i = 0; i < RLIM_NLIMITS; i++) { | |
617 | if (rlim[i].rlim_cur == RLIM_INFINITY) | |
1c963eb1 | 618 | seq_printf(m, "%-25s %-20s ", |
25ce3191 | 619 | lnames[i].name, "unlimited"); |
d85f50d5 | 620 | else |
1c963eb1 | 621 | seq_printf(m, "%-25s %-20lu ", |
25ce3191 | 622 | lnames[i].name, rlim[i].rlim_cur); |
d85f50d5 NH |
623 | |
624 | if (rlim[i].rlim_max == RLIM_INFINITY) | |
1c963eb1 | 625 | seq_printf(m, "%-20s ", "unlimited"); |
d85f50d5 | 626 | else |
1c963eb1 | 627 | seq_printf(m, "%-20lu ", rlim[i].rlim_max); |
d85f50d5 NH |
628 | |
629 | if (lnames[i].unit) | |
1c963eb1 | 630 | seq_printf(m, "%-10s\n", lnames[i].unit); |
d85f50d5 | 631 | else |
1c963eb1 | 632 | seq_putc(m, '\n'); |
d85f50d5 NH |
633 | } |
634 | ||
1c963eb1 | 635 | return 0; |
d85f50d5 NH |
636 | } |
637 | ||
ebcb6734 | 638 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 AD |
639 | static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns, |
640 | struct pid *pid, struct task_struct *task) | |
ebcb6734 | 641 | { |
631b7aba SRRH |
642 | struct syscall_info info; |
643 | u64 *args = &info.data.args[0]; | |
25ce3191 JP |
644 | int res; |
645 | ||
646 | res = lock_trace(task); | |
a9712bc1 AV |
647 | if (res) |
648 | return res; | |
ebcb6734 | 649 | |
631b7aba | 650 | if (task_current_syscall(task, &info)) |
09d93bd6 | 651 | seq_puts(m, "running\n"); |
631b7aba SRRH |
652 | else if (info.data.nr < 0) |
653 | seq_printf(m, "%d 0x%llx 0x%llx\n", | |
654 | info.data.nr, info.sp, info.data.instruction_pointer); | |
a9712bc1 | 655 | else |
09d93bd6 | 656 | seq_printf(m, |
631b7aba SRRH |
657 | "%d 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx\n", |
658 | info.data.nr, | |
ebcb6734 | 659 | args[0], args[1], args[2], args[3], args[4], args[5], |
631b7aba | 660 | info.sp, info.data.instruction_pointer); |
a9712bc1 | 661 | unlock_trace(task); |
25ce3191 JP |
662 | |
663 | return 0; | |
ebcb6734 RM |
664 | } |
665 | #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */ | |
666 | ||
1da177e4 LT |
667 | /************************************************************************/ |
668 | /* Here the fs part begins */ | |
669 | /************************************************************************/ | |
670 | ||
671 | /* permission checks */ | |
51a18734 | 672 | static bool proc_fd_access_allowed(struct inode *inode) |
1da177e4 | 673 | { |
778c1144 | 674 | struct task_struct *task; |
51a18734 | 675 | bool allowed = false; |
df26c40e EB |
676 | /* Allow access to a task's file descriptors if it is us or we |
677 | * may use ptrace attach to the process and find out that | |
678 | * information. | |
778c1144 EB |
679 | */ |
680 | task = get_proc_task(inode); | |
df26c40e | 681 | if (task) { |
caaee623 | 682 | allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS); |
778c1144 | 683 | put_task_struct(task); |
df26c40e | 684 | } |
778c1144 | 685 | return allowed; |
1da177e4 LT |
686 | } |
687 | ||
549c7297 CB |
688 | int proc_setattr(struct user_namespace *mnt_userns, struct dentry *dentry, |
689 | struct iattr *attr) | |
6d76fa58 LT |
690 | { |
691 | int error; | |
2b0143b5 | 692 | struct inode *inode = d_inode(dentry); |
6d76fa58 LT |
693 | |
694 | if (attr->ia_valid & ATTR_MODE) | |
695 | return -EPERM; | |
696 | ||
2f221d6f | 697 | error = setattr_prepare(&init_user_ns, dentry, attr); |
1025774c CH |
698 | if (error) |
699 | return error; | |
700 | ||
2f221d6f | 701 | setattr_copy(&init_user_ns, inode, attr); |
1025774c CH |
702 | mark_inode_dirty(inode); |
703 | return 0; | |
6d76fa58 LT |
704 | } |
705 | ||
0499680a VK |
706 | /* |
707 | * May current process learn task's sched/cmdline info (for hide_pid_min=1) | |
708 | * or euid/egid (for hide_pid_min=2)? | |
709 | */ | |
fa10fed3 | 710 | static bool has_pid_permissions(struct proc_fs_info *fs_info, |
0499680a | 711 | struct task_struct *task, |
e61bb8b3 | 712 | enum proc_hidepid hide_pid_min) |
0499680a | 713 | { |
24a71ce5 AG |
714 | /* |
715 | * If 'hidpid' mount option is set force a ptrace check, | |
716 | * we indicate that we are using a filesystem syscall | |
717 | * by passing PTRACE_MODE_READ_FSCREDS | |
718 | */ | |
719 | if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE) | |
720 | return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS); | |
721 | ||
fa10fed3 | 722 | if (fs_info->hide_pid < hide_pid_min) |
0499680a | 723 | return true; |
fa10fed3 | 724 | if (in_group_p(fs_info->pid_gid)) |
0499680a | 725 | return true; |
caaee623 | 726 | return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS); |
0499680a VK |
727 | } |
728 | ||
729 | ||
549c7297 CB |
730 | static int proc_pid_permission(struct user_namespace *mnt_userns, |
731 | struct inode *inode, int mask) | |
0499680a | 732 | { |
fa10fed3 | 733 | struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb); |
0499680a VK |
734 | struct task_struct *task; |
735 | bool has_perms; | |
736 | ||
737 | task = get_proc_task(inode); | |
a2ef990a XF |
738 | if (!task) |
739 | return -ESRCH; | |
fa10fed3 | 740 | has_perms = has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS); |
0499680a VK |
741 | put_task_struct(task); |
742 | ||
743 | if (!has_perms) { | |
fa10fed3 | 744 | if (fs_info->hide_pid == HIDEPID_INVISIBLE) { |
0499680a VK |
745 | /* |
746 | * Let's make getdents(), stat(), and open() | |
747 | * consistent with each other. If a process | |
748 | * may not stat() a file, it shouldn't be seen | |
749 | * in procfs at all. | |
750 | */ | |
751 | return -ENOENT; | |
752 | } | |
753 | ||
754 | return -EPERM; | |
755 | } | |
47291baa | 756 | return generic_permission(&init_user_ns, inode, mask); |
0499680a VK |
757 | } |
758 | ||
759 | ||
760 | ||
c5ef1c42 | 761 | static const struct inode_operations proc_def_inode_operations = { |
6d76fa58 LT |
762 | .setattr = proc_setattr, |
763 | }; | |
764 | ||
be614086 EB |
765 | static int proc_single_show(struct seq_file *m, void *v) |
766 | { | |
767 | struct inode *inode = m->private; | |
9d78edea | 768 | struct pid_namespace *ns = proc_pid_ns(inode->i_sb); |
76f668be | 769 | struct pid *pid = proc_pid(inode); |
be614086 EB |
770 | struct task_struct *task; |
771 | int ret; | |
772 | ||
be614086 EB |
773 | task = get_pid_task(pid, PIDTYPE_PID); |
774 | if (!task) | |
775 | return -ESRCH; | |
776 | ||
777 | ret = PROC_I(inode)->op.proc_show(m, ns, pid, task); | |
778 | ||
779 | put_task_struct(task); | |
780 | return ret; | |
781 | } | |
782 | ||
783 | static int proc_single_open(struct inode *inode, struct file *filp) | |
784 | { | |
c6a34058 | 785 | return single_open(filp, proc_single_show, inode); |
be614086 EB |
786 | } |
787 | ||
788 | static const struct file_operations proc_single_file_operations = { | |
789 | .open = proc_single_open, | |
790 | .read = seq_read, | |
791 | .llseek = seq_lseek, | |
792 | .release = single_release, | |
793 | }; | |
794 | ||
5381e169 ON |
795 | |
796 | struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode) | |
1da177e4 | 797 | { |
5381e169 ON |
798 | struct task_struct *task = get_proc_task(inode); |
799 | struct mm_struct *mm = ERR_PTR(-ESRCH); | |
e268337d | 800 | |
5381e169 | 801 | if (task) { |
caaee623 | 802 | mm = mm_access(task, mode | PTRACE_MODE_FSCREDS); |
5381e169 | 803 | put_task_struct(task); |
e268337d | 804 | |
5381e169 ON |
805 | if (!IS_ERR_OR_NULL(mm)) { |
806 | /* ensure this mm_struct can't be freed */ | |
f1f10076 | 807 | mmgrab(mm); |
5381e169 ON |
808 | /* but do not pin its memory */ |
809 | mmput(mm); | |
810 | } | |
811 | } | |
812 | ||
813 | return mm; | |
814 | } | |
815 | ||
816 | static int __mem_open(struct inode *inode, struct file *file, unsigned int mode) | |
817 | { | |
818 | struct mm_struct *mm = proc_mem_open(inode, mode); | |
e268337d LT |
819 | |
820 | if (IS_ERR(mm)) | |
821 | return PTR_ERR(mm); | |
822 | ||
e268337d | 823 | file->private_data = mm; |
1da177e4 LT |
824 | return 0; |
825 | } | |
826 | ||
b409e578 CW |
827 | static int mem_open(struct inode *inode, struct file *file) |
828 | { | |
bc452b4b DH |
829 | int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH); |
830 | ||
831 | /* OK to pass negative loff_t, we can catch out-of-range */ | |
832 | file->f_mode |= FMODE_UNSIGNED_OFFSET; | |
833 | ||
834 | return ret; | |
b409e578 CW |
835 | } |
836 | ||
572d34b9 ON |
837 | static ssize_t mem_rw(struct file *file, char __user *buf, |
838 | size_t count, loff_t *ppos, int write) | |
1da177e4 | 839 | { |
e268337d | 840 | struct mm_struct *mm = file->private_data; |
572d34b9 ON |
841 | unsigned long addr = *ppos; |
842 | ssize_t copied; | |
1da177e4 | 843 | char *page; |
272ddc8b | 844 | unsigned int flags; |
1da177e4 | 845 | |
e268337d LT |
846 | if (!mm) |
847 | return 0; | |
99f89551 | 848 | |
0ee931c4 | 849 | page = (char *)__get_free_page(GFP_KERNEL); |
30cd8903 | 850 | if (!page) |
e268337d | 851 | return -ENOMEM; |
1da177e4 | 852 | |
f7ca54f4 | 853 | copied = 0; |
388f7934 | 854 | if (!mmget_not_zero(mm)) |
6d08f2c7 ON |
855 | goto free; |
856 | ||
f511c0b1 | 857 | flags = FOLL_FORCE | (write ? FOLL_WRITE : 0); |
6347e8d5 | 858 | |
1da177e4 | 859 | while (count > 0) { |
d238692b | 860 | size_t this_len = min_t(size_t, count, PAGE_SIZE); |
1da177e4 | 861 | |
572d34b9 | 862 | if (write && copy_from_user(page, buf, this_len)) { |
1da177e4 LT |
863 | copied = -EFAULT; |
864 | break; | |
865 | } | |
572d34b9 | 866 | |
6347e8d5 | 867 | this_len = access_remote_vm(mm, addr, page, this_len, flags); |
572d34b9 | 868 | if (!this_len) { |
1da177e4 LT |
869 | if (!copied) |
870 | copied = -EIO; | |
871 | break; | |
872 | } | |
572d34b9 ON |
873 | |
874 | if (!write && copy_to_user(buf, page, this_len)) { | |
875 | copied = -EFAULT; | |
876 | break; | |
877 | } | |
878 | ||
879 | buf += this_len; | |
880 | addr += this_len; | |
881 | copied += this_len; | |
882 | count -= this_len; | |
1da177e4 | 883 | } |
572d34b9 | 884 | *ppos = addr; |
30cd8903 | 885 | |
6d08f2c7 ON |
886 | mmput(mm); |
887 | free: | |
30cd8903 | 888 | free_page((unsigned long) page); |
1da177e4 LT |
889 | return copied; |
890 | } | |
1da177e4 | 891 | |
572d34b9 ON |
892 | static ssize_t mem_read(struct file *file, char __user *buf, |
893 | size_t count, loff_t *ppos) | |
894 | { | |
895 | return mem_rw(file, buf, count, ppos, 0); | |
896 | } | |
897 | ||
898 | static ssize_t mem_write(struct file *file, const char __user *buf, | |
899 | size_t count, loff_t *ppos) | |
900 | { | |
901 | return mem_rw(file, (char __user*)buf, count, ppos, 1); | |
902 | } | |
903 | ||
85863e47 | 904 | loff_t mem_lseek(struct file *file, loff_t offset, int orig) |
1da177e4 LT |
905 | { |
906 | switch (orig) { | |
907 | case 0: | |
908 | file->f_pos = offset; | |
909 | break; | |
910 | case 1: | |
911 | file->f_pos += offset; | |
912 | break; | |
913 | default: | |
914 | return -EINVAL; | |
915 | } | |
916 | force_successful_syscall_return(); | |
917 | return file->f_pos; | |
918 | } | |
919 | ||
e268337d LT |
920 | static int mem_release(struct inode *inode, struct file *file) |
921 | { | |
922 | struct mm_struct *mm = file->private_data; | |
71879d3c | 923 | if (mm) |
6d08f2c7 | 924 | mmdrop(mm); |
e268337d LT |
925 | return 0; |
926 | } | |
927 | ||
00977a59 | 928 | static const struct file_operations proc_mem_operations = { |
1da177e4 LT |
929 | .llseek = mem_lseek, |
930 | .read = mem_read, | |
931 | .write = mem_write, | |
932 | .open = mem_open, | |
e268337d | 933 | .release = mem_release, |
1da177e4 LT |
934 | }; |
935 | ||
b409e578 CW |
936 | static int environ_open(struct inode *inode, struct file *file) |
937 | { | |
938 | return __mem_open(inode, file, PTRACE_MODE_READ); | |
939 | } | |
940 | ||
315e28c8 JP |
941 | static ssize_t environ_read(struct file *file, char __user *buf, |
942 | size_t count, loff_t *ppos) | |
943 | { | |
315e28c8 JP |
944 | char *page; |
945 | unsigned long src = *ppos; | |
b409e578 CW |
946 | int ret = 0; |
947 | struct mm_struct *mm = file->private_data; | |
a3b609ef | 948 | unsigned long env_start, env_end; |
315e28c8 | 949 | |
8148a73c MK |
950 | /* Ensure the process spawned far enough to have an environment. */ |
951 | if (!mm || !mm->env_end) | |
b409e578 | 952 | return 0; |
315e28c8 | 953 | |
0ee931c4 | 954 | page = (char *)__get_free_page(GFP_KERNEL); |
315e28c8 | 955 | if (!page) |
b409e578 | 956 | return -ENOMEM; |
315e28c8 | 957 | |
d6f64b89 | 958 | ret = 0; |
388f7934 | 959 | if (!mmget_not_zero(mm)) |
b409e578 | 960 | goto free; |
a3b609ef | 961 | |
88aa7cc6 | 962 | spin_lock(&mm->arg_lock); |
a3b609ef MG |
963 | env_start = mm->env_start; |
964 | env_end = mm->env_end; | |
88aa7cc6 | 965 | spin_unlock(&mm->arg_lock); |
a3b609ef | 966 | |
315e28c8 | 967 | while (count > 0) { |
e8905ec2 DH |
968 | size_t this_len, max_len; |
969 | int retval; | |
315e28c8 | 970 | |
a3b609ef | 971 | if (src >= (env_end - env_start)) |
315e28c8 JP |
972 | break; |
973 | ||
a3b609ef | 974 | this_len = env_end - (env_start + src); |
e8905ec2 DH |
975 | |
976 | max_len = min_t(size_t, PAGE_SIZE, count); | |
977 | this_len = min(max_len, this_len); | |
315e28c8 | 978 | |
7f7ccc2c | 979 | retval = access_remote_vm(mm, (env_start + src), page, this_len, FOLL_ANON); |
315e28c8 JP |
980 | |
981 | if (retval <= 0) { | |
982 | ret = retval; | |
983 | break; | |
984 | } | |
985 | ||
986 | if (copy_to_user(buf, page, retval)) { | |
987 | ret = -EFAULT; | |
988 | break; | |
989 | } | |
990 | ||
991 | ret += retval; | |
992 | src += retval; | |
993 | buf += retval; | |
994 | count -= retval; | |
995 | } | |
996 | *ppos = src; | |
315e28c8 | 997 | mmput(mm); |
b409e578 CW |
998 | |
999 | free: | |
315e28c8 | 1000 | free_page((unsigned long) page); |
315e28c8 JP |
1001 | return ret; |
1002 | } | |
1003 | ||
1004 | static const struct file_operations proc_environ_operations = { | |
b409e578 | 1005 | .open = environ_open, |
315e28c8 | 1006 | .read = environ_read, |
87df8424 | 1007 | .llseek = generic_file_llseek, |
b409e578 | 1008 | .release = mem_release, |
315e28c8 JP |
1009 | }; |
1010 | ||
c5317167 AV |
1011 | static int auxv_open(struct inode *inode, struct file *file) |
1012 | { | |
1013 | return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS); | |
1014 | } | |
1015 | ||
1016 | static ssize_t auxv_read(struct file *file, char __user *buf, | |
1017 | size_t count, loff_t *ppos) | |
1018 | { | |
1019 | struct mm_struct *mm = file->private_data; | |
1020 | unsigned int nwords = 0; | |
06b2849d LY |
1021 | |
1022 | if (!mm) | |
1023 | return 0; | |
c5317167 AV |
1024 | do { |
1025 | nwords += 2; | |
1026 | } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */ | |
1027 | return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv, | |
1028 | nwords * sizeof(mm->saved_auxv[0])); | |
1029 | } | |
1030 | ||
1031 | static const struct file_operations proc_auxv_operations = { | |
1032 | .open = auxv_open, | |
1033 | .read = auxv_read, | |
1034 | .llseek = generic_file_llseek, | |
1035 | .release = mem_release, | |
1036 | }; | |
1037 | ||
fa0cbbf1 DR |
1038 | static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count, |
1039 | loff_t *ppos) | |
1040 | { | |
496ad9aa | 1041 | struct task_struct *task = get_proc_task(file_inode(file)); |
fa0cbbf1 DR |
1042 | char buffer[PROC_NUMBUF]; |
1043 | int oom_adj = OOM_ADJUST_MIN; | |
1044 | size_t len; | |
fa0cbbf1 DR |
1045 | |
1046 | if (!task) | |
1047 | return -ESRCH; | |
f913da59 MH |
1048 | if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX) |
1049 | oom_adj = OOM_ADJUST_MAX; | |
1050 | else | |
1051 | oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) / | |
1052 | OOM_SCORE_ADJ_MAX; | |
fa0cbbf1 | 1053 | put_task_struct(task); |
66606567 CH |
1054 | if (oom_adj > OOM_ADJUST_MAX) |
1055 | oom_adj = OOM_ADJUST_MAX; | |
fa0cbbf1 DR |
1056 | len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj); |
1057 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
1058 | } | |
1059 | ||
1d5f0acb MH |
1060 | static int __set_oom_adj(struct file *file, int oom_adj, bool legacy) |
1061 | { | |
44a70ade | 1062 | struct mm_struct *mm = NULL; |
1d5f0acb MH |
1063 | struct task_struct *task; |
1064 | int err = 0; | |
1065 | ||
1066 | task = get_proc_task(file_inode(file)); | |
1067 | if (!task) | |
1068 | return -ESRCH; | |
1069 | ||
1070 | mutex_lock(&oom_adj_mutex); | |
1071 | if (legacy) { | |
1072 | if (oom_adj < task->signal->oom_score_adj && | |
1073 | !capable(CAP_SYS_RESOURCE)) { | |
1074 | err = -EACCES; | |
1075 | goto err_unlock; | |
1076 | } | |
1077 | /* | |
1078 | * /proc/pid/oom_adj is provided for legacy purposes, ask users to use | |
1079 | * /proc/pid/oom_score_adj instead. | |
1080 | */ | |
1081 | pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n", | |
1082 | current->comm, task_pid_nr(current), task_pid_nr(task), | |
1083 | task_pid_nr(task)); | |
1084 | } else { | |
1085 | if ((short)oom_adj < task->signal->oom_score_adj_min && | |
1086 | !capable(CAP_SYS_RESOURCE)) { | |
1087 | err = -EACCES; | |
1088 | goto err_unlock; | |
1089 | } | |
1090 | } | |
1091 | ||
44a70ade MH |
1092 | /* |
1093 | * Make sure we will check other processes sharing the mm if this is | |
1094 | * not vfrok which wants its own oom_score_adj. | |
1095 | * pin the mm so it doesn't go away and get reused after task_unlock | |
1096 | */ | |
1097 | if (!task->vfork_done) { | |
1098 | struct task_struct *p = find_lock_task_mm(task); | |
1099 | ||
1100 | if (p) { | |
67197a4f | 1101 | if (test_bit(MMF_MULTIPROCESS, &p->mm->flags)) { |
44a70ade | 1102 | mm = p->mm; |
f1f10076 | 1103 | mmgrab(mm); |
44a70ade MH |
1104 | } |
1105 | task_unlock(p); | |
1106 | } | |
1107 | } | |
1108 | ||
1d5f0acb MH |
1109 | task->signal->oom_score_adj = oom_adj; |
1110 | if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE)) | |
1111 | task->signal->oom_score_adj_min = (short)oom_adj; | |
1112 | trace_oom_score_adj_update(task); | |
44a70ade MH |
1113 | |
1114 | if (mm) { | |
1115 | struct task_struct *p; | |
1116 | ||
1117 | rcu_read_lock(); | |
1118 | for_each_process(p) { | |
1119 | if (same_thread_group(task, p)) | |
1120 | continue; | |
1121 | ||
1122 | /* do not touch kernel threads or the global init */ | |
1123 | if (p->flags & PF_KTHREAD || is_global_init(p)) | |
1124 | continue; | |
1125 | ||
1126 | task_lock(p); | |
1127 | if (!p->vfork_done && process_shares_mm(p, mm)) { | |
44a70ade MH |
1128 | p->signal->oom_score_adj = oom_adj; |
1129 | if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE)) | |
1130 | p->signal->oom_score_adj_min = (short)oom_adj; | |
1131 | } | |
1132 | task_unlock(p); | |
1133 | } | |
1134 | rcu_read_unlock(); | |
1135 | mmdrop(mm); | |
1136 | } | |
1d5f0acb MH |
1137 | err_unlock: |
1138 | mutex_unlock(&oom_adj_mutex); | |
1139 | put_task_struct(task); | |
1140 | return err; | |
1141 | } | |
f913da59 | 1142 | |
b72bdfa7 DR |
1143 | /* |
1144 | * /proc/pid/oom_adj exists solely for backwards compatibility with previous | |
1145 | * kernels. The effective policy is defined by oom_score_adj, which has a | |
1146 | * different scale: oom_adj grew exponentially and oom_score_adj grows linearly. | |
1147 | * Values written to oom_adj are simply mapped linearly to oom_score_adj. | |
1148 | * Processes that become oom disabled via oom_adj will still be oom disabled | |
1149 | * with this implementation. | |
1150 | * | |
1151 | * oom_adj cannot be removed since existing userspace binaries use it. | |
1152 | */ | |
fa0cbbf1 DR |
1153 | static ssize_t oom_adj_write(struct file *file, const char __user *buf, |
1154 | size_t count, loff_t *ppos) | |
1155 | { | |
fa0cbbf1 DR |
1156 | char buffer[PROC_NUMBUF]; |
1157 | int oom_adj; | |
fa0cbbf1 DR |
1158 | int err; |
1159 | ||
1160 | memset(buffer, 0, sizeof(buffer)); | |
1161 | if (count > sizeof(buffer) - 1) | |
1162 | count = sizeof(buffer) - 1; | |
1163 | if (copy_from_user(buffer, buf, count)) { | |
1164 | err = -EFAULT; | |
1165 | goto out; | |
1166 | } | |
1167 | ||
1168 | err = kstrtoint(strstrip(buffer), 0, &oom_adj); | |
1169 | if (err) | |
1170 | goto out; | |
1171 | if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) && | |
1172 | oom_adj != OOM_DISABLE) { | |
1173 | err = -EINVAL; | |
1174 | goto out; | |
1175 | } | |
1176 | ||
fa0cbbf1 DR |
1177 | /* |
1178 | * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum | |
1179 | * value is always attainable. | |
1180 | */ | |
1181 | if (oom_adj == OOM_ADJUST_MAX) | |
1182 | oom_adj = OOM_SCORE_ADJ_MAX; | |
1183 | else | |
1184 | oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE; | |
1185 | ||
1d5f0acb | 1186 | err = __set_oom_adj(file, oom_adj, true); |
fa0cbbf1 DR |
1187 | out: |
1188 | return err < 0 ? err : count; | |
1189 | } | |
1190 | ||
1191 | static const struct file_operations proc_oom_adj_operations = { | |
1192 | .read = oom_adj_read, | |
1193 | .write = oom_adj_write, | |
1194 | .llseek = generic_file_llseek, | |
1195 | }; | |
1196 | ||
a63d83f4 DR |
1197 | static ssize_t oom_score_adj_read(struct file *file, char __user *buf, |
1198 | size_t count, loff_t *ppos) | |
1199 | { | |
496ad9aa | 1200 | struct task_struct *task = get_proc_task(file_inode(file)); |
a63d83f4 | 1201 | char buffer[PROC_NUMBUF]; |
a9c58b90 | 1202 | short oom_score_adj = OOM_SCORE_ADJ_MIN; |
a63d83f4 DR |
1203 | size_t len; |
1204 | ||
1205 | if (!task) | |
1206 | return -ESRCH; | |
f913da59 | 1207 | oom_score_adj = task->signal->oom_score_adj; |
a63d83f4 | 1208 | put_task_struct(task); |
a9c58b90 | 1209 | len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj); |
a63d83f4 DR |
1210 | return simple_read_from_buffer(buf, count, ppos, buffer, len); |
1211 | } | |
1212 | ||
1213 | static ssize_t oom_score_adj_write(struct file *file, const char __user *buf, | |
1214 | size_t count, loff_t *ppos) | |
1215 | { | |
a63d83f4 | 1216 | char buffer[PROC_NUMBUF]; |
0a8cb8e3 | 1217 | int oom_score_adj; |
a63d83f4 DR |
1218 | int err; |
1219 | ||
1220 | memset(buffer, 0, sizeof(buffer)); | |
1221 | if (count > sizeof(buffer) - 1) | |
1222 | count = sizeof(buffer) - 1; | |
723548bf DR |
1223 | if (copy_from_user(buffer, buf, count)) { |
1224 | err = -EFAULT; | |
1225 | goto out; | |
1226 | } | |
a63d83f4 | 1227 | |
0a8cb8e3 | 1228 | err = kstrtoint(strstrip(buffer), 0, &oom_score_adj); |
a63d83f4 | 1229 | if (err) |
723548bf | 1230 | goto out; |
a63d83f4 | 1231 | if (oom_score_adj < OOM_SCORE_ADJ_MIN || |
723548bf DR |
1232 | oom_score_adj > OOM_SCORE_ADJ_MAX) { |
1233 | err = -EINVAL; | |
1234 | goto out; | |
1235 | } | |
a63d83f4 | 1236 | |
1d5f0acb | 1237 | err = __set_oom_adj(file, oom_score_adj, false); |
723548bf DR |
1238 | out: |
1239 | return err < 0 ? err : count; | |
a63d83f4 DR |
1240 | } |
1241 | ||
1242 | static const struct file_operations proc_oom_score_adj_operations = { | |
1243 | .read = oom_score_adj_read, | |
1244 | .write = oom_score_adj_write, | |
6038f373 | 1245 | .llseek = default_llseek, |
a63d83f4 DR |
1246 | }; |
1247 | ||
4b7d248b | 1248 | #ifdef CONFIG_AUDIT |
b4eb4f7f | 1249 | #define TMPBUFLEN 11 |
1da177e4 LT |
1250 | static ssize_t proc_loginuid_read(struct file * file, char __user * buf, |
1251 | size_t count, loff_t *ppos) | |
1252 | { | |
496ad9aa | 1253 | struct inode * inode = file_inode(file); |
99f89551 | 1254 | struct task_struct *task = get_proc_task(inode); |
1da177e4 LT |
1255 | ssize_t length; |
1256 | char tmpbuf[TMPBUFLEN]; | |
1257 | ||
99f89551 EB |
1258 | if (!task) |
1259 | return -ESRCH; | |
1da177e4 | 1260 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", |
e1760bd5 EB |
1261 | from_kuid(file->f_cred->user_ns, |
1262 | audit_get_loginuid(task))); | |
99f89551 | 1263 | put_task_struct(task); |
1da177e4 LT |
1264 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); |
1265 | } | |
1266 | ||
1267 | static ssize_t proc_loginuid_write(struct file * file, const char __user * buf, | |
1268 | size_t count, loff_t *ppos) | |
1269 | { | |
496ad9aa | 1270 | struct inode * inode = file_inode(file); |
1da177e4 | 1271 | uid_t loginuid; |
e1760bd5 | 1272 | kuid_t kloginuid; |
774636e1 | 1273 | int rv; |
1da177e4 | 1274 | |
4ea33a97 JA |
1275 | /* Don't let kthreads write their own loginuid */ |
1276 | if (current->flags & PF_KTHREAD) | |
1277 | return -EPERM; | |
1278 | ||
7dc52157 PM |
1279 | rcu_read_lock(); |
1280 | if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) { | |
1281 | rcu_read_unlock(); | |
1da177e4 | 1282 | return -EPERM; |
7dc52157 PM |
1283 | } |
1284 | rcu_read_unlock(); | |
1da177e4 | 1285 | |
1da177e4 LT |
1286 | if (*ppos != 0) { |
1287 | /* No partial writes. */ | |
1288 | return -EINVAL; | |
1289 | } | |
1da177e4 | 1290 | |
774636e1 AD |
1291 | rv = kstrtou32_from_user(buf, count, 10, &loginuid); |
1292 | if (rv < 0) | |
1293 | return rv; | |
81407c84 EP |
1294 | |
1295 | /* is userspace tring to explicitly UNSET the loginuid? */ | |
1296 | if (loginuid == AUDIT_UID_UNSET) { | |
1297 | kloginuid = INVALID_UID; | |
1298 | } else { | |
1299 | kloginuid = make_kuid(file->f_cred->user_ns, loginuid); | |
774636e1 AD |
1300 | if (!uid_valid(kloginuid)) |
1301 | return -EINVAL; | |
e1760bd5 EB |
1302 | } |
1303 | ||
774636e1 AD |
1304 | rv = audit_set_loginuid(kloginuid); |
1305 | if (rv < 0) | |
1306 | return rv; | |
1307 | return count; | |
1da177e4 LT |
1308 | } |
1309 | ||
00977a59 | 1310 | static const struct file_operations proc_loginuid_operations = { |
1da177e4 LT |
1311 | .read = proc_loginuid_read, |
1312 | .write = proc_loginuid_write, | |
87df8424 | 1313 | .llseek = generic_file_llseek, |
1da177e4 | 1314 | }; |
1e0bd755 EP |
1315 | |
1316 | static ssize_t proc_sessionid_read(struct file * file, char __user * buf, | |
1317 | size_t count, loff_t *ppos) | |
1318 | { | |
496ad9aa | 1319 | struct inode * inode = file_inode(file); |
1e0bd755 EP |
1320 | struct task_struct *task = get_proc_task(inode); |
1321 | ssize_t length; | |
1322 | char tmpbuf[TMPBUFLEN]; | |
1323 | ||
1324 | if (!task) | |
1325 | return -ESRCH; | |
1326 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", | |
1327 | audit_get_sessionid(task)); | |
1328 | put_task_struct(task); | |
1329 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); | |
1330 | } | |
1331 | ||
1332 | static const struct file_operations proc_sessionid_operations = { | |
1333 | .read = proc_sessionid_read, | |
87df8424 | 1334 | .llseek = generic_file_llseek, |
1e0bd755 | 1335 | }; |
1da177e4 LT |
1336 | #endif |
1337 | ||
f4f154fd AM |
1338 | #ifdef CONFIG_FAULT_INJECTION |
1339 | static ssize_t proc_fault_inject_read(struct file * file, char __user * buf, | |
1340 | size_t count, loff_t *ppos) | |
1341 | { | |
496ad9aa | 1342 | struct task_struct *task = get_proc_task(file_inode(file)); |
f4f154fd AM |
1343 | char buffer[PROC_NUMBUF]; |
1344 | size_t len; | |
1345 | int make_it_fail; | |
f4f154fd AM |
1346 | |
1347 | if (!task) | |
1348 | return -ESRCH; | |
1349 | make_it_fail = task->make_it_fail; | |
1350 | put_task_struct(task); | |
1351 | ||
1352 | len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail); | |
0c28f287 AM |
1353 | |
1354 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
f4f154fd AM |
1355 | } |
1356 | ||
1357 | static ssize_t proc_fault_inject_write(struct file * file, | |
1358 | const char __user * buf, size_t count, loff_t *ppos) | |
1359 | { | |
1360 | struct task_struct *task; | |
774636e1 | 1361 | char buffer[PROC_NUMBUF]; |
f4f154fd | 1362 | int make_it_fail; |
774636e1 | 1363 | int rv; |
f4f154fd AM |
1364 | |
1365 | if (!capable(CAP_SYS_RESOURCE)) | |
1366 | return -EPERM; | |
1367 | memset(buffer, 0, sizeof(buffer)); | |
1368 | if (count > sizeof(buffer) - 1) | |
1369 | count = sizeof(buffer) - 1; | |
1370 | if (copy_from_user(buffer, buf, count)) | |
1371 | return -EFAULT; | |
774636e1 AD |
1372 | rv = kstrtoint(strstrip(buffer), 0, &make_it_fail); |
1373 | if (rv < 0) | |
1374 | return rv; | |
16caed31 DJ |
1375 | if (make_it_fail < 0 || make_it_fail > 1) |
1376 | return -EINVAL; | |
1377 | ||
496ad9aa | 1378 | task = get_proc_task(file_inode(file)); |
f4f154fd AM |
1379 | if (!task) |
1380 | return -ESRCH; | |
1381 | task->make_it_fail = make_it_fail; | |
1382 | put_task_struct(task); | |
cba8aafe VL |
1383 | |
1384 | return count; | |
f4f154fd AM |
1385 | } |
1386 | ||
00977a59 | 1387 | static const struct file_operations proc_fault_inject_operations = { |
f4f154fd AM |
1388 | .read = proc_fault_inject_read, |
1389 | .write = proc_fault_inject_write, | |
87df8424 | 1390 | .llseek = generic_file_llseek, |
f4f154fd | 1391 | }; |
e41d5818 DV |
1392 | |
1393 | static ssize_t proc_fail_nth_write(struct file *file, const char __user *buf, | |
1394 | size_t count, loff_t *ppos) | |
1395 | { | |
1396 | struct task_struct *task; | |
9049f2f6 AM |
1397 | int err; |
1398 | unsigned int n; | |
e41d5818 | 1399 | |
1203c8e6 AM |
1400 | err = kstrtouint_from_user(buf, count, 0, &n); |
1401 | if (err) | |
1402 | return err; | |
1403 | ||
e41d5818 DV |
1404 | task = get_proc_task(file_inode(file)); |
1405 | if (!task) | |
1406 | return -ESRCH; | |
9f7118b2 | 1407 | task->fail_nth = n; |
e41d5818 | 1408 | put_task_struct(task); |
1203c8e6 | 1409 | |
e41d5818 DV |
1410 | return count; |
1411 | } | |
1412 | ||
1413 | static ssize_t proc_fail_nth_read(struct file *file, char __user *buf, | |
1414 | size_t count, loff_t *ppos) | |
1415 | { | |
1416 | struct task_struct *task; | |
bfc74093 AM |
1417 | char numbuf[PROC_NUMBUF]; |
1418 | ssize_t len; | |
e41d5818 DV |
1419 | |
1420 | task = get_proc_task(file_inode(file)); | |
1421 | if (!task) | |
1422 | return -ESRCH; | |
9f7118b2 | 1423 | len = snprintf(numbuf, sizeof(numbuf), "%u\n", task->fail_nth); |
1203c8e6 | 1424 | put_task_struct(task); |
a44937fe | 1425 | return simple_read_from_buffer(buf, count, ppos, numbuf, len); |
e41d5818 DV |
1426 | } |
1427 | ||
1428 | static const struct file_operations proc_fail_nth_operations = { | |
1429 | .read = proc_fail_nth_read, | |
1430 | .write = proc_fail_nth_write, | |
1431 | }; | |
f4f154fd AM |
1432 | #endif |
1433 | ||
9745512c | 1434 | |
43ae34cb IM |
1435 | #ifdef CONFIG_SCHED_DEBUG |
1436 | /* | |
1437 | * Print out various scheduling related per-task fields: | |
1438 | */ | |
1439 | static int sched_show(struct seq_file *m, void *v) | |
1440 | { | |
1441 | struct inode *inode = m->private; | |
9d78edea | 1442 | struct pid_namespace *ns = proc_pid_ns(inode->i_sb); |
43ae34cb IM |
1443 | struct task_struct *p; |
1444 | ||
43ae34cb IM |
1445 | p = get_proc_task(inode); |
1446 | if (!p) | |
1447 | return -ESRCH; | |
74dc3384 | 1448 | proc_sched_show_task(p, ns, m); |
43ae34cb IM |
1449 | |
1450 | put_task_struct(p); | |
1451 | ||
1452 | return 0; | |
1453 | } | |
1454 | ||
1455 | static ssize_t | |
1456 | sched_write(struct file *file, const char __user *buf, | |
1457 | size_t count, loff_t *offset) | |
1458 | { | |
496ad9aa | 1459 | struct inode *inode = file_inode(file); |
43ae34cb IM |
1460 | struct task_struct *p; |
1461 | ||
43ae34cb IM |
1462 | p = get_proc_task(inode); |
1463 | if (!p) | |
1464 | return -ESRCH; | |
1465 | proc_sched_set_task(p); | |
1466 | ||
1467 | put_task_struct(p); | |
1468 | ||
1469 | return count; | |
1470 | } | |
1471 | ||
1472 | static int sched_open(struct inode *inode, struct file *filp) | |
1473 | { | |
c6a34058 | 1474 | return single_open(filp, sched_show, inode); |
43ae34cb IM |
1475 | } |
1476 | ||
1477 | static const struct file_operations proc_pid_sched_operations = { | |
1478 | .open = sched_open, | |
1479 | .read = seq_read, | |
1480 | .write = sched_write, | |
1481 | .llseek = seq_lseek, | |
5ea473a1 | 1482 | .release = single_release, |
43ae34cb IM |
1483 | }; |
1484 | ||
1485 | #endif | |
1486 | ||
5091faa4 MG |
1487 | #ifdef CONFIG_SCHED_AUTOGROUP |
1488 | /* | |
1489 | * Print out autogroup related information: | |
1490 | */ | |
1491 | static int sched_autogroup_show(struct seq_file *m, void *v) | |
1492 | { | |
1493 | struct inode *inode = m->private; | |
1494 | struct task_struct *p; | |
1495 | ||
1496 | p = get_proc_task(inode); | |
1497 | if (!p) | |
1498 | return -ESRCH; | |
1499 | proc_sched_autogroup_show_task(p, m); | |
1500 | ||
1501 | put_task_struct(p); | |
1502 | ||
1503 | return 0; | |
1504 | } | |
1505 | ||
1506 | static ssize_t | |
1507 | sched_autogroup_write(struct file *file, const char __user *buf, | |
1508 | size_t count, loff_t *offset) | |
1509 | { | |
496ad9aa | 1510 | struct inode *inode = file_inode(file); |
5091faa4 MG |
1511 | struct task_struct *p; |
1512 | char buffer[PROC_NUMBUF]; | |
0a8cb8e3 | 1513 | int nice; |
5091faa4 MG |
1514 | int err; |
1515 | ||
1516 | memset(buffer, 0, sizeof(buffer)); | |
1517 | if (count > sizeof(buffer) - 1) | |
1518 | count = sizeof(buffer) - 1; | |
1519 | if (copy_from_user(buffer, buf, count)) | |
1520 | return -EFAULT; | |
1521 | ||
0a8cb8e3 AD |
1522 | err = kstrtoint(strstrip(buffer), 0, &nice); |
1523 | if (err < 0) | |
1524 | return err; | |
5091faa4 MG |
1525 | |
1526 | p = get_proc_task(inode); | |
1527 | if (!p) | |
1528 | return -ESRCH; | |
1529 | ||
2e5b5b3a | 1530 | err = proc_sched_autogroup_set_nice(p, nice); |
5091faa4 MG |
1531 | if (err) |
1532 | count = err; | |
1533 | ||
1534 | put_task_struct(p); | |
1535 | ||
1536 | return count; | |
1537 | } | |
1538 | ||
1539 | static int sched_autogroup_open(struct inode *inode, struct file *filp) | |
1540 | { | |
1541 | int ret; | |
1542 | ||
1543 | ret = single_open(filp, sched_autogroup_show, NULL); | |
1544 | if (!ret) { | |
1545 | struct seq_file *m = filp->private_data; | |
1546 | ||
1547 | m->private = inode; | |
1548 | } | |
1549 | return ret; | |
1550 | } | |
1551 | ||
1552 | static const struct file_operations proc_pid_sched_autogroup_operations = { | |
1553 | .open = sched_autogroup_open, | |
1554 | .read = seq_read, | |
1555 | .write = sched_autogroup_write, | |
1556 | .llseek = seq_lseek, | |
1557 | .release = single_release, | |
1558 | }; | |
1559 | ||
1560 | #endif /* CONFIG_SCHED_AUTOGROUP */ | |
1561 | ||
04a8682a AV |
1562 | #ifdef CONFIG_TIME_NS |
1563 | static int timens_offsets_show(struct seq_file *m, void *v) | |
1564 | { | |
1565 | struct task_struct *p; | |
1566 | ||
1567 | p = get_proc_task(file_inode(m->file)); | |
1568 | if (!p) | |
1569 | return -ESRCH; | |
1570 | proc_timens_show_offsets(p, m); | |
1571 | ||
1572 | put_task_struct(p); | |
1573 | ||
1574 | return 0; | |
1575 | } | |
1576 | ||
1577 | static ssize_t timens_offsets_write(struct file *file, const char __user *buf, | |
1578 | size_t count, loff_t *ppos) | |
1579 | { | |
1580 | struct inode *inode = file_inode(file); | |
1581 | struct proc_timens_offset offsets[2]; | |
1582 | char *kbuf = NULL, *pos, *next_line; | |
1583 | struct task_struct *p; | |
1584 | int ret, noffsets; | |
1585 | ||
1586 | /* Only allow < page size writes at the beginning of the file */ | |
1587 | if ((*ppos != 0) || (count >= PAGE_SIZE)) | |
1588 | return -EINVAL; | |
1589 | ||
1590 | /* Slurp in the user data */ | |
1591 | kbuf = memdup_user_nul(buf, count); | |
1592 | if (IS_ERR(kbuf)) | |
1593 | return PTR_ERR(kbuf); | |
1594 | ||
1595 | /* Parse the user data */ | |
1596 | ret = -EINVAL; | |
1597 | noffsets = 0; | |
1598 | for (pos = kbuf; pos; pos = next_line) { | |
1599 | struct proc_timens_offset *off = &offsets[noffsets]; | |
94d440d6 | 1600 | char clock[10]; |
04a8682a AV |
1601 | int err; |
1602 | ||
1603 | /* Find the end of line and ensure we don't look past it */ | |
1604 | next_line = strchr(pos, '\n'); | |
1605 | if (next_line) { | |
1606 | *next_line = '\0'; | |
1607 | next_line++; | |
1608 | if (*next_line == '\0') | |
1609 | next_line = NULL; | |
1610 | } | |
1611 | ||
94d440d6 | 1612 | err = sscanf(pos, "%9s %lld %lu", clock, |
04a8682a AV |
1613 | &off->val.tv_sec, &off->val.tv_nsec); |
1614 | if (err != 3 || off->val.tv_nsec >= NSEC_PER_SEC) | |
1615 | goto out; | |
94d440d6 AV |
1616 | |
1617 | clock[sizeof(clock) - 1] = 0; | |
1618 | if (strcmp(clock, "monotonic") == 0 || | |
1619 | strcmp(clock, __stringify(CLOCK_MONOTONIC)) == 0) | |
1620 | off->clockid = CLOCK_MONOTONIC; | |
1621 | else if (strcmp(clock, "boottime") == 0 || | |
1622 | strcmp(clock, __stringify(CLOCK_BOOTTIME)) == 0) | |
1623 | off->clockid = CLOCK_BOOTTIME; | |
1624 | else | |
1625 | goto out; | |
1626 | ||
04a8682a AV |
1627 | noffsets++; |
1628 | if (noffsets == ARRAY_SIZE(offsets)) { | |
1629 | if (next_line) | |
1630 | count = next_line - kbuf; | |
1631 | break; | |
1632 | } | |
1633 | } | |
1634 | ||
1635 | ret = -ESRCH; | |
1636 | p = get_proc_task(inode); | |
1637 | if (!p) | |
1638 | goto out; | |
1639 | ret = proc_timens_set_offset(file, p, offsets, noffsets); | |
1640 | put_task_struct(p); | |
1641 | if (ret) | |
1642 | goto out; | |
1643 | ||
1644 | ret = count; | |
1645 | out: | |
1646 | kfree(kbuf); | |
1647 | return ret; | |
1648 | } | |
1649 | ||
1650 | static int timens_offsets_open(struct inode *inode, struct file *filp) | |
1651 | { | |
1652 | return single_open(filp, timens_offsets_show, inode); | |
1653 | } | |
1654 | ||
1655 | static const struct file_operations proc_timens_offsets_operations = { | |
1656 | .open = timens_offsets_open, | |
1657 | .read = seq_read, | |
1658 | .write = timens_offsets_write, | |
1659 | .llseek = seq_lseek, | |
1660 | .release = single_release, | |
1661 | }; | |
1662 | #endif /* CONFIG_TIME_NS */ | |
1663 | ||
4614a696 | 1664 | static ssize_t comm_write(struct file *file, const char __user *buf, |
1665 | size_t count, loff_t *offset) | |
1666 | { | |
496ad9aa | 1667 | struct inode *inode = file_inode(file); |
4614a696 | 1668 | struct task_struct *p; |
1669 | char buffer[TASK_COMM_LEN]; | |
830e0fc9 | 1670 | const size_t maxlen = sizeof(buffer) - 1; |
4614a696 | 1671 | |
1672 | memset(buffer, 0, sizeof(buffer)); | |
830e0fc9 | 1673 | if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count)) |
4614a696 | 1674 | return -EFAULT; |
1675 | ||
1676 | p = get_proc_task(inode); | |
1677 | if (!p) | |
1678 | return -ESRCH; | |
1679 | ||
c2f273eb | 1680 | if (same_thread_group(current, p)) { |
4614a696 | 1681 | set_task_comm(p, buffer); |
c2f273eb OK |
1682 | proc_comm_connector(p); |
1683 | } | |
4614a696 | 1684 | else |
1685 | count = -EINVAL; | |
1686 | ||
1687 | put_task_struct(p); | |
1688 | ||
1689 | return count; | |
1690 | } | |
1691 | ||
1692 | static int comm_show(struct seq_file *m, void *v) | |
1693 | { | |
1694 | struct inode *inode = m->private; | |
1695 | struct task_struct *p; | |
1696 | ||
1697 | p = get_proc_task(inode); | |
1698 | if (!p) | |
1699 | return -ESRCH; | |
1700 | ||
88b72b31 TH |
1701 | proc_task_name(m, p, false); |
1702 | seq_putc(m, '\n'); | |
4614a696 | 1703 | |
1704 | put_task_struct(p); | |
1705 | ||
1706 | return 0; | |
1707 | } | |
1708 | ||
1709 | static int comm_open(struct inode *inode, struct file *filp) | |
1710 | { | |
c6a34058 | 1711 | return single_open(filp, comm_show, inode); |
4614a696 | 1712 | } |
1713 | ||
1714 | static const struct file_operations proc_pid_set_comm_operations = { | |
1715 | .open = comm_open, | |
1716 | .read = seq_read, | |
1717 | .write = comm_write, | |
1718 | .llseek = seq_lseek, | |
1719 | .release = single_release, | |
1720 | }; | |
1721 | ||
7773fbc5 | 1722 | static int proc_exe_link(struct dentry *dentry, struct path *exe_path) |
925d1c40 MH |
1723 | { |
1724 | struct task_struct *task; | |
925d1c40 MH |
1725 | struct file *exe_file; |
1726 | ||
2b0143b5 | 1727 | task = get_proc_task(d_inode(dentry)); |
925d1c40 MH |
1728 | if (!task) |
1729 | return -ENOENT; | |
cd81a917 | 1730 | exe_file = get_task_exe_file(task); |
925d1c40 | 1731 | put_task_struct(task); |
925d1c40 MH |
1732 | if (exe_file) { |
1733 | *exe_path = exe_file->f_path; | |
1734 | path_get(&exe_file->f_path); | |
1735 | fput(exe_file); | |
1736 | return 0; | |
1737 | } else | |
1738 | return -ENOENT; | |
1739 | } | |
1740 | ||
6b255391 | 1741 | static const char *proc_pid_get_link(struct dentry *dentry, |
fceef393 AV |
1742 | struct inode *inode, |
1743 | struct delayed_call *done) | |
1da177e4 | 1744 | { |
408ef013 | 1745 | struct path path; |
1da177e4 LT |
1746 | int error = -EACCES; |
1747 | ||
6b255391 AV |
1748 | if (!dentry) |
1749 | return ERR_PTR(-ECHILD); | |
1750 | ||
778c1144 EB |
1751 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1752 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1753 | goto out; |
1da177e4 | 1754 | |
408ef013 CH |
1755 | error = PROC_I(inode)->op.proc_get_link(dentry, &path); |
1756 | if (error) | |
1757 | goto out; | |
1758 | ||
1bc82070 | 1759 | error = nd_jump_link(&path); |
1da177e4 | 1760 | out: |
008b150a | 1761 | return ERR_PTR(error); |
1da177e4 LT |
1762 | } |
1763 | ||
3dcd25f3 | 1764 | static int do_proc_readlink(struct path *path, char __user *buffer, int buflen) |
1da177e4 | 1765 | { |
3a72917c | 1766 | char *tmp = kmalloc(PATH_MAX, GFP_KERNEL); |
3dcd25f3 | 1767 | char *pathname; |
1da177e4 LT |
1768 | int len; |
1769 | ||
1770 | if (!tmp) | |
1771 | return -ENOMEM; | |
0c28f287 | 1772 | |
3a72917c | 1773 | pathname = d_path(path, tmp, PATH_MAX); |
3dcd25f3 JB |
1774 | len = PTR_ERR(pathname); |
1775 | if (IS_ERR(pathname)) | |
1da177e4 | 1776 | goto out; |
3a72917c | 1777 | len = tmp + PATH_MAX - 1 - pathname; |
1da177e4 LT |
1778 | |
1779 | if (len > buflen) | |
1780 | len = buflen; | |
3dcd25f3 | 1781 | if (copy_to_user(buffer, pathname, len)) |
1da177e4 LT |
1782 | len = -EFAULT; |
1783 | out: | |
3a72917c | 1784 | kfree(tmp); |
1da177e4 LT |
1785 | return len; |
1786 | } | |
1787 | ||
1788 | static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen) | |
1789 | { | |
1790 | int error = -EACCES; | |
2b0143b5 | 1791 | struct inode *inode = d_inode(dentry); |
3dcd25f3 | 1792 | struct path path; |
1da177e4 | 1793 | |
778c1144 EB |
1794 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1795 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1796 | goto out; |
1da177e4 | 1797 | |
7773fbc5 | 1798 | error = PROC_I(inode)->op.proc_get_link(dentry, &path); |
1da177e4 LT |
1799 | if (error) |
1800 | goto out; | |
1801 | ||
3dcd25f3 JB |
1802 | error = do_proc_readlink(&path, buffer, buflen); |
1803 | path_put(&path); | |
1da177e4 | 1804 | out: |
1da177e4 LT |
1805 | return error; |
1806 | } | |
1807 | ||
faf60af1 | 1808 | const struct inode_operations proc_pid_link_inode_operations = { |
1da177e4 | 1809 | .readlink = proc_pid_readlink, |
6b255391 | 1810 | .get_link = proc_pid_get_link, |
6d76fa58 | 1811 | .setattr = proc_setattr, |
1da177e4 LT |
1812 | }; |
1813 | ||
28a6d671 EB |
1814 | |
1815 | /* building an inode */ | |
1816 | ||
c6eb50d2 | 1817 | void task_dump_owner(struct task_struct *task, umode_t mode, |
68eb94f1 EB |
1818 | kuid_t *ruid, kgid_t *rgid) |
1819 | { | |
1820 | /* Depending on the state of dumpable compute who should own a | |
1821 | * proc file for a task. | |
1822 | */ | |
1823 | const struct cred *cred; | |
1824 | kuid_t uid; | |
1825 | kgid_t gid; | |
1826 | ||
2e0ad552 AD |
1827 | if (unlikely(task->flags & PF_KTHREAD)) { |
1828 | *ruid = GLOBAL_ROOT_UID; | |
1829 | *rgid = GLOBAL_ROOT_GID; | |
1830 | return; | |
1831 | } | |
1832 | ||
68eb94f1 EB |
1833 | /* Default to the tasks effective ownership */ |
1834 | rcu_read_lock(); | |
1835 | cred = __task_cred(task); | |
1836 | uid = cred->euid; | |
1837 | gid = cred->egid; | |
1838 | rcu_read_unlock(); | |
1839 | ||
1840 | /* | |
1841 | * Before the /proc/pid/status file was created the only way to read | |
1842 | * the effective uid of a /process was to stat /proc/pid. Reading | |
1843 | * /proc/pid/status is slow enough that procps and other packages | |
1844 | * kept stating /proc/pid. To keep the rules in /proc simple I have | |
1845 | * made this apply to all per process world readable and executable | |
1846 | * directories. | |
1847 | */ | |
1848 | if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) { | |
1849 | struct mm_struct *mm; | |
1850 | task_lock(task); | |
1851 | mm = task->mm; | |
1852 | /* Make non-dumpable tasks owned by some root */ | |
1853 | if (mm) { | |
1854 | if (get_dumpable(mm) != SUID_DUMP_USER) { | |
1855 | struct user_namespace *user_ns = mm->user_ns; | |
1856 | ||
1857 | uid = make_kuid(user_ns, 0); | |
1858 | if (!uid_valid(uid)) | |
1859 | uid = GLOBAL_ROOT_UID; | |
1860 | ||
1861 | gid = make_kgid(user_ns, 0); | |
1862 | if (!gid_valid(gid)) | |
1863 | gid = GLOBAL_ROOT_GID; | |
1864 | } | |
1865 | } else { | |
1866 | uid = GLOBAL_ROOT_UID; | |
1867 | gid = GLOBAL_ROOT_GID; | |
1868 | } | |
1869 | task_unlock(task); | |
1870 | } | |
1871 | *ruid = uid; | |
1872 | *rgid = gid; | |
1873 | } | |
1874 | ||
7bc3e6e5 EB |
1875 | void proc_pid_evict_inode(struct proc_inode *ei) |
1876 | { | |
1877 | struct pid *pid = ei->pid; | |
1878 | ||
1879 | if (S_ISDIR(ei->vfs_inode.i_mode)) { | |
63f818f4 | 1880 | spin_lock(&pid->lock); |
7bc3e6e5 | 1881 | hlist_del_init_rcu(&ei->sibling_inodes); |
63f818f4 | 1882 | spin_unlock(&pid->lock); |
7bc3e6e5 EB |
1883 | } |
1884 | ||
1885 | put_pid(pid); | |
1886 | } | |
1887 | ||
d919a1e7 | 1888 | struct inode *proc_pid_make_inode(struct super_block *sb, |
db978da8 | 1889 | struct task_struct *task, umode_t mode) |
28a6d671 EB |
1890 | { |
1891 | struct inode * inode; | |
1892 | struct proc_inode *ei; | |
7bc3e6e5 | 1893 | struct pid *pid; |
1da177e4 | 1894 | |
28a6d671 | 1895 | /* We need a new inode */ |
1da177e4 | 1896 | |
28a6d671 EB |
1897 | inode = new_inode(sb); |
1898 | if (!inode) | |
1899 | goto out; | |
1900 | ||
1901 | /* Common stuff */ | |
1902 | ei = PROC_I(inode); | |
db978da8 | 1903 | inode->i_mode = mode; |
85fe4025 | 1904 | inode->i_ino = get_next_ino(); |
078cd827 | 1905 | inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode); |
28a6d671 EB |
1906 | inode->i_op = &proc_def_inode_operations; |
1907 | ||
1908 | /* | |
1909 | * grab the reference to task. | |
1910 | */ | |
7bc3e6e5 EB |
1911 | pid = get_task_pid(task, PIDTYPE_PID); |
1912 | if (!pid) | |
28a6d671 EB |
1913 | goto out_unlock; |
1914 | ||
7bc3e6e5 EB |
1915 | /* Let the pid remember us for quick removal */ |
1916 | ei->pid = pid; | |
7bc3e6e5 | 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 | ||
d919a1e7 ZC |
1929 | /* |
1930 | * Generating an inode and adding it into @pid->inodes, so that task will | |
1931 | * invalidate inode's dentry before being released. | |
1932 | * | |
1933 | * This helper is used for creating dir-type entries under '/proc' and | |
1934 | * '/proc/<tgid>/task'. Other entries(eg. fd, stat) under '/proc/<tgid>' | |
1935 | * can be released by invalidating '/proc/<tgid>' dentry. | |
1936 | * In theory, dentries under '/proc/<tgid>/task' can also be released by | |
1937 | * invalidating '/proc/<tgid>' dentry, we reserve it to handle single | |
1938 | * thread exiting situation: Any one of threads should invalidate its | |
1939 | * '/proc/<tgid>/task/<pid>' dentry before released. | |
1940 | */ | |
1941 | static struct inode *proc_pid_make_base_inode(struct super_block *sb, | |
1942 | struct task_struct *task, umode_t mode) | |
1943 | { | |
1944 | struct inode *inode; | |
1945 | struct proc_inode *ei; | |
1946 | struct pid *pid; | |
1947 | ||
1948 | inode = proc_pid_make_inode(sb, task, mode); | |
1949 | if (!inode) | |
1950 | return NULL; | |
1951 | ||
1952 | /* Let proc_flush_pid find this directory inode */ | |
1953 | ei = PROC_I(inode); | |
1954 | pid = ei->pid; | |
1955 | spin_lock(&pid->lock); | |
1956 | hlist_add_head_rcu(&ei->sibling_inodes, &pid->inodes); | |
1957 | spin_unlock(&pid->lock); | |
1958 | ||
1959 | return inode; | |
1960 | } | |
1961 | ||
549c7297 CB |
1962 | int pid_getattr(struct user_namespace *mnt_userns, const struct path *path, |
1963 | struct kstat *stat, u32 request_mask, unsigned int query_flags) | |
1da177e4 | 1964 | { |
a528d35e | 1965 | struct inode *inode = d_inode(path->dentry); |
fa10fed3 | 1966 | struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb); |
28a6d671 | 1967 | struct task_struct *task; |
c69e8d9c | 1968 | |
0d56a451 | 1969 | generic_fillattr(&init_user_ns, inode, stat); |
1da177e4 | 1970 | |
dcb0f222 EB |
1971 | stat->uid = GLOBAL_ROOT_UID; |
1972 | stat->gid = GLOBAL_ROOT_GID; | |
94116929 | 1973 | rcu_read_lock(); |
28a6d671 EB |
1974 | task = pid_task(proc_pid(inode), PIDTYPE_PID); |
1975 | if (task) { | |
fa10fed3 | 1976 | if (!has_pid_permissions(fs_info, task, HIDEPID_INVISIBLE)) { |
0499680a VK |
1977 | rcu_read_unlock(); |
1978 | /* | |
1979 | * This doesn't prevent learning whether PID exists, | |
1980 | * it only makes getattr() consistent with readdir(). | |
1981 | */ | |
1982 | return -ENOENT; | |
1983 | } | |
68eb94f1 | 1984 | task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid); |
1da177e4 | 1985 | } |
28a6d671 | 1986 | rcu_read_unlock(); |
d6e71144 | 1987 | return 0; |
1da177e4 LT |
1988 | } |
1989 | ||
1da177e4 LT |
1990 | /* dentry stuff */ |
1991 | ||
1992 | /* | |
1bbc5513 AV |
1993 | * Set <pid>/... inode ownership (can change due to setuid(), etc.) |
1994 | */ | |
1995 | void pid_update_inode(struct task_struct *task, struct inode *inode) | |
1996 | { | |
1997 | task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid); | |
1998 | ||
1999 | inode->i_mode &= ~(S_ISUID | S_ISGID); | |
2000 | security_task_to_inode(task, inode); | |
2001 | } | |
2002 | ||
2003 | /* | |
1da177e4 LT |
2004 | * Rewrite the inode's ownerships here because the owning task may have |
2005 | * performed a setuid(), etc. | |
99f89551 | 2006 | * |
1da177e4 | 2007 | */ |
1bbc5513 | 2008 | static int pid_revalidate(struct dentry *dentry, unsigned int flags) |
1da177e4 | 2009 | { |
34286d66 NP |
2010 | struct inode *inode; |
2011 | struct task_struct *task; | |
da4d6b9c | 2012 | int ret = 0; |
c69e8d9c | 2013 | |
da4d6b9c SB |
2014 | rcu_read_lock(); |
2015 | inode = d_inode_rcu(dentry); | |
2016 | if (!inode) | |
2017 | goto out; | |
2018 | task = pid_task(proc_pid(inode), PIDTYPE_PID); | |
34286d66 | 2019 | |
99f89551 | 2020 | if (task) { |
1bbc5513 | 2021 | pid_update_inode(task, inode); |
da4d6b9c | 2022 | ret = 1; |
1da177e4 | 2023 | } |
da4d6b9c SB |
2024 | out: |
2025 | rcu_read_unlock(); | |
2026 | return ret; | |
1da177e4 LT |
2027 | } |
2028 | ||
d855a4b7 ON |
2029 | static inline bool proc_inode_is_dead(struct inode *inode) |
2030 | { | |
2031 | return !proc_pid(inode)->tasks[PIDTYPE_PID].first; | |
2032 | } | |
2033 | ||
1dd704b6 DH |
2034 | int pid_delete_dentry(const struct dentry *dentry) |
2035 | { | |
2036 | /* Is the task we represent dead? | |
2037 | * If so, then don't put the dentry on the lru list, | |
2038 | * kill it immediately. | |
2039 | */ | |
2b0143b5 | 2040 | return proc_inode_is_dead(d_inode(dentry)); |
1dd704b6 DH |
2041 | } |
2042 | ||
6b4e306a | 2043 | const struct dentry_operations pid_dentry_operations = |
28a6d671 EB |
2044 | { |
2045 | .d_revalidate = pid_revalidate, | |
2046 | .d_delete = pid_delete_dentry, | |
2047 | }; | |
2048 | ||
2049 | /* Lookups */ | |
2050 | ||
1c0d04c9 EB |
2051 | /* |
2052 | * Fill a directory entry. | |
2053 | * | |
2054 | * If possible create the dcache entry and derive our inode number and | |
2055 | * file type from dcache entry. | |
2056 | * | |
2057 | * Since all of the proc inode numbers are dynamically generated, the inode | |
d2928e85 | 2058 | * numbers do not exist until the inode is cache. This means creating |
1c0d04c9 EB |
2059 | * the dcache entry in readdir is necessary to keep the inode numbers |
2060 | * reported by readdir in sync with the inode numbers reported | |
2061 | * by stat. | |
2062 | */ | |
f0c3b509 | 2063 | bool proc_fill_cache(struct file *file, struct dir_context *ctx, |
a4ef3895 | 2064 | const char *name, unsigned int len, |
c5141e6d | 2065 | instantiate_t instantiate, struct task_struct *task, const void *ptr) |
61a28784 | 2066 | { |
f0c3b509 | 2067 | struct dentry *child, *dir = file->f_path.dentry; |
1df98b8b | 2068 | struct qstr qname = QSTR_INIT(name, len); |
61a28784 | 2069 | struct inode *inode; |
0168b9e3 AV |
2070 | unsigned type = DT_UNKNOWN; |
2071 | ino_t ino = 1; | |
61a28784 | 2072 | |
1df98b8b | 2073 | child = d_hash_and_lookup(dir, &qname); |
61a28784 | 2074 | if (!child) { |
3781764b AV |
2075 | DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq); |
2076 | child = d_alloc_parallel(dir, &qname, &wq); | |
2077 | if (IS_ERR(child)) | |
1df98b8b | 2078 | goto end_instantiate; |
3781764b | 2079 | if (d_in_lookup(child)) { |
0168b9e3 AV |
2080 | struct dentry *res; |
2081 | res = instantiate(child, task, ptr); | |
3781764b | 2082 | d_lookup_done(child); |
0168b9e3 AV |
2083 | if (unlikely(res)) { |
2084 | dput(child); | |
2085 | child = res; | |
d85b399b AV |
2086 | if (IS_ERR(child)) |
2087 | goto end_instantiate; | |
3781764b | 2088 | } |
61a28784 EB |
2089 | } |
2090 | } | |
2b0143b5 | 2091 | inode = d_inode(child); |
147ce699 AV |
2092 | ino = inode->i_ino; |
2093 | type = inode->i_mode >> 12; | |
61a28784 | 2094 | dput(child); |
d85b399b | 2095 | end_instantiate: |
f0c3b509 | 2096 | return dir_emit(ctx, name, len, ino, type); |
61a28784 EB |
2097 | } |
2098 | ||
640708a2 PE |
2099 | /* |
2100 | * dname_to_vma_addr - maps a dentry name into two unsigned longs | |
2101 | * which represent vma start and end addresses. | |
2102 | */ | |
2103 | static int dname_to_vma_addr(struct dentry *dentry, | |
2104 | unsigned long *start, unsigned long *end) | |
2105 | { | |
ac7f1061 AD |
2106 | const char *str = dentry->d_name.name; |
2107 | unsigned long long sval, eval; | |
2108 | unsigned int len; | |
2109 | ||
35318db5 AD |
2110 | if (str[0] == '0' && str[1] != '-') |
2111 | return -EINVAL; | |
ac7f1061 AD |
2112 | len = _parse_integer(str, 16, &sval); |
2113 | if (len & KSTRTOX_OVERFLOW) | |
2114 | return -EINVAL; | |
2115 | if (sval != (unsigned long)sval) | |
2116 | return -EINVAL; | |
2117 | str += len; | |
2118 | ||
2119 | if (*str != '-') | |
640708a2 | 2120 | return -EINVAL; |
ac7f1061 AD |
2121 | str++; |
2122 | ||
35318db5 AD |
2123 | if (str[0] == '0' && str[1]) |
2124 | return -EINVAL; | |
ac7f1061 AD |
2125 | len = _parse_integer(str, 16, &eval); |
2126 | if (len & KSTRTOX_OVERFLOW) | |
2127 | return -EINVAL; | |
2128 | if (eval != (unsigned long)eval) | |
2129 | return -EINVAL; | |
2130 | str += len; | |
2131 | ||
2132 | if (*str != '\0') | |
2133 | return -EINVAL; | |
2134 | ||
2135 | *start = sval; | |
2136 | *end = eval; | |
640708a2 PE |
2137 | |
2138 | return 0; | |
2139 | } | |
2140 | ||
0b728e19 | 2141 | static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags) |
640708a2 PE |
2142 | { |
2143 | unsigned long vm_start, vm_end; | |
2144 | bool exact_vma_exists = false; | |
2145 | struct mm_struct *mm = NULL; | |
2146 | struct task_struct *task; | |
640708a2 PE |
2147 | struct inode *inode; |
2148 | int status = 0; | |
2149 | ||
0b728e19 | 2150 | if (flags & LOOKUP_RCU) |
640708a2 PE |
2151 | return -ECHILD; |
2152 | ||
2b0143b5 | 2153 | inode = d_inode(dentry); |
640708a2 PE |
2154 | task = get_proc_task(inode); |
2155 | if (!task) | |
2156 | goto out_notask; | |
2157 | ||
caaee623 | 2158 | mm = mm_access(task, PTRACE_MODE_READ_FSCREDS); |
2344bec7 | 2159 | if (IS_ERR_OR_NULL(mm)) |
640708a2 PE |
2160 | goto out; |
2161 | ||
2162 | if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) { | |
d8ed45c5 | 2163 | status = mmap_read_lock_killable(mm); |
cd9e2bb8 KK |
2164 | if (!status) { |
2165 | exact_vma_exists = !!find_exact_vma(mm, vm_start, | |
2166 | vm_end); | |
d8ed45c5 | 2167 | mmap_read_unlock(mm); |
cd9e2bb8 | 2168 | } |
640708a2 PE |
2169 | } |
2170 | ||
2171 | mmput(mm); | |
2172 | ||
2173 | if (exact_vma_exists) { | |
68eb94f1 EB |
2174 | task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid); |
2175 | ||
640708a2 PE |
2176 | security_task_to_inode(task, inode); |
2177 | status = 1; | |
2178 | } | |
2179 | ||
2180 | out: | |
2181 | put_task_struct(task); | |
2182 | ||
2183 | out_notask: | |
640708a2 PE |
2184 | return status; |
2185 | } | |
2186 | ||
2187 | static const struct dentry_operations tid_map_files_dentry_operations = { | |
2188 | .d_revalidate = map_files_d_revalidate, | |
2189 | .d_delete = pid_delete_dentry, | |
2190 | }; | |
2191 | ||
6b255391 | 2192 | static int map_files_get_link(struct dentry *dentry, struct path *path) |
640708a2 PE |
2193 | { |
2194 | unsigned long vm_start, vm_end; | |
2195 | struct vm_area_struct *vma; | |
2196 | struct task_struct *task; | |
2197 | struct mm_struct *mm; | |
2198 | int rc; | |
2199 | ||
2200 | rc = -ENOENT; | |
2b0143b5 | 2201 | task = get_proc_task(d_inode(dentry)); |
640708a2 PE |
2202 | if (!task) |
2203 | goto out; | |
2204 | ||
2205 | mm = get_task_mm(task); | |
2206 | put_task_struct(task); | |
2207 | if (!mm) | |
2208 | goto out; | |
2209 | ||
2210 | rc = dname_to_vma_addr(dentry, &vm_start, &vm_end); | |
2211 | if (rc) | |
2212 | goto out_mmput; | |
2213 | ||
d8ed45c5 | 2214 | rc = mmap_read_lock_killable(mm); |
cd9e2bb8 KK |
2215 | if (rc) |
2216 | goto out_mmput; | |
2217 | ||
70335abb | 2218 | rc = -ENOENT; |
640708a2 PE |
2219 | vma = find_exact_vma(mm, vm_start, vm_end); |
2220 | if (vma && vma->vm_file) { | |
2221 | *path = vma->vm_file->f_path; | |
2222 | path_get(path); | |
2223 | rc = 0; | |
2224 | } | |
d8ed45c5 | 2225 | mmap_read_unlock(mm); |
640708a2 PE |
2226 | |
2227 | out_mmput: | |
2228 | mmput(mm); | |
2229 | out: | |
2230 | return rc; | |
2231 | } | |
2232 | ||
2233 | struct map_files_info { | |
20d28cde AD |
2234 | unsigned long start; |
2235 | unsigned long end; | |
7b540d06 | 2236 | fmode_t mode; |
640708a2 PE |
2237 | }; |
2238 | ||
bdb4d100 | 2239 | /* |
12886f8a AR |
2240 | * Only allow CAP_SYS_ADMIN and CAP_CHECKPOINT_RESTORE to follow the links, due |
2241 | * to concerns about how the symlinks may be used to bypass permissions on | |
2242 | * ancestor directories in the path to the file in question. | |
bdb4d100 CO |
2243 | */ |
2244 | static const char * | |
6b255391 | 2245 | proc_map_files_get_link(struct dentry *dentry, |
fceef393 AV |
2246 | struct inode *inode, |
2247 | struct delayed_call *done) | |
bdb4d100 | 2248 | { |
12886f8a | 2249 | if (!checkpoint_restore_ns_capable(&init_user_ns)) |
bdb4d100 CO |
2250 | return ERR_PTR(-EPERM); |
2251 | ||
fceef393 | 2252 | return proc_pid_get_link(dentry, inode, done); |
bdb4d100 CO |
2253 | } |
2254 | ||
2255 | /* | |
6b255391 | 2256 | * Identical to proc_pid_link_inode_operations except for get_link() |
bdb4d100 CO |
2257 | */ |
2258 | static const struct inode_operations proc_map_files_link_inode_operations = { | |
2259 | .readlink = proc_pid_readlink, | |
6b255391 | 2260 | .get_link = proc_map_files_get_link, |
bdb4d100 CO |
2261 | .setattr = proc_setattr, |
2262 | }; | |
2263 | ||
0168b9e3 AV |
2264 | static struct dentry * |
2265 | proc_map_files_instantiate(struct dentry *dentry, | |
640708a2 PE |
2266 | struct task_struct *task, const void *ptr) |
2267 | { | |
7b540d06 | 2268 | fmode_t mode = (fmode_t)(unsigned long)ptr; |
640708a2 PE |
2269 | struct proc_inode *ei; |
2270 | struct inode *inode; | |
2271 | ||
0168b9e3 | 2272 | inode = proc_pid_make_inode(dentry->d_sb, task, S_IFLNK | |
db978da8 AG |
2273 | ((mode & FMODE_READ ) ? S_IRUSR : 0) | |
2274 | ((mode & FMODE_WRITE) ? S_IWUSR : 0)); | |
640708a2 | 2275 | if (!inode) |
0168b9e3 | 2276 | return ERR_PTR(-ENOENT); |
640708a2 PE |
2277 | |
2278 | ei = PROC_I(inode); | |
6b255391 | 2279 | ei->op.proc_get_link = map_files_get_link; |
640708a2 | 2280 | |
bdb4d100 | 2281 | inode->i_op = &proc_map_files_link_inode_operations; |
640708a2 | 2282 | inode->i_size = 64; |
640708a2 PE |
2283 | |
2284 | d_set_d_op(dentry, &tid_map_files_dentry_operations); | |
0168b9e3 | 2285 | return d_splice_alias(inode, dentry); |
640708a2 PE |
2286 | } |
2287 | ||
2288 | static struct dentry *proc_map_files_lookup(struct inode *dir, | |
00cd8dd3 | 2289 | struct dentry *dentry, unsigned int flags) |
640708a2 PE |
2290 | { |
2291 | unsigned long vm_start, vm_end; | |
2292 | struct vm_area_struct *vma; | |
2293 | struct task_struct *task; | |
0168b9e3 | 2294 | struct dentry *result; |
640708a2 PE |
2295 | struct mm_struct *mm; |
2296 | ||
0168b9e3 | 2297 | result = ERR_PTR(-ENOENT); |
640708a2 PE |
2298 | task = get_proc_task(dir); |
2299 | if (!task) | |
2300 | goto out; | |
2301 | ||
0168b9e3 | 2302 | result = ERR_PTR(-EACCES); |
caaee623 | 2303 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
640708a2 PE |
2304 | goto out_put_task; |
2305 | ||
0168b9e3 | 2306 | result = ERR_PTR(-ENOENT); |
640708a2 | 2307 | if (dname_to_vma_addr(dentry, &vm_start, &vm_end)) |
eb94cd96 | 2308 | goto out_put_task; |
640708a2 PE |
2309 | |
2310 | mm = get_task_mm(task); | |
2311 | if (!mm) | |
eb94cd96 | 2312 | goto out_put_task; |
640708a2 | 2313 | |
cd9e2bb8 | 2314 | result = ERR_PTR(-EINTR); |
d8ed45c5 | 2315 | if (mmap_read_lock_killable(mm)) |
cd9e2bb8 KK |
2316 | goto out_put_mm; |
2317 | ||
2318 | result = ERR_PTR(-ENOENT); | |
640708a2 PE |
2319 | vma = find_exact_vma(mm, vm_start, vm_end); |
2320 | if (!vma) | |
2321 | goto out_no_vma; | |
2322 | ||
05f56484 | 2323 | if (vma->vm_file) |
0168b9e3 | 2324 | result = proc_map_files_instantiate(dentry, task, |
05f56484 | 2325 | (void *)(unsigned long)vma->vm_file->f_mode); |
640708a2 PE |
2326 | |
2327 | out_no_vma: | |
d8ed45c5 | 2328 | mmap_read_unlock(mm); |
cd9e2bb8 | 2329 | out_put_mm: |
640708a2 | 2330 | mmput(mm); |
640708a2 PE |
2331 | out_put_task: |
2332 | put_task_struct(task); | |
2333 | out: | |
0168b9e3 | 2334 | return result; |
640708a2 PE |
2335 | } |
2336 | ||
2337 | static const struct inode_operations proc_map_files_inode_operations = { | |
2338 | .lookup = proc_map_files_lookup, | |
2339 | .permission = proc_fd_permission, | |
2340 | .setattr = proc_setattr, | |
2341 | }; | |
2342 | ||
2343 | static int | |
f0c3b509 | 2344 | proc_map_files_readdir(struct file *file, struct dir_context *ctx) |
640708a2 | 2345 | { |
640708a2 PE |
2346 | struct vm_area_struct *vma; |
2347 | struct task_struct *task; | |
2348 | struct mm_struct *mm; | |
f0c3b509 | 2349 | unsigned long nr_files, pos, i; |
94f8f3b0 | 2350 | GENRADIX(struct map_files_info) fa; |
f0c3b509 | 2351 | struct map_files_info *p; |
640708a2 PE |
2352 | int ret; |
2353 | ||
94f8f3b0 KO |
2354 | genradix_init(&fa); |
2355 | ||
640708a2 | 2356 | ret = -ENOENT; |
f0c3b509 | 2357 | task = get_proc_task(file_inode(file)); |
640708a2 PE |
2358 | if (!task) |
2359 | goto out; | |
2360 | ||
2361 | ret = -EACCES; | |
caaee623 | 2362 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
640708a2 PE |
2363 | goto out_put_task; |
2364 | ||
2365 | ret = 0; | |
f0c3b509 AV |
2366 | if (!dir_emit_dots(file, ctx)) |
2367 | goto out_put_task; | |
640708a2 | 2368 | |
f0c3b509 AV |
2369 | mm = get_task_mm(task); |
2370 | if (!mm) | |
2371 | goto out_put_task; | |
cd9e2bb8 | 2372 | |
89154dd5 | 2373 | ret = mmap_read_lock_killable(mm); |
cd9e2bb8 KK |
2374 | if (ret) { |
2375 | mmput(mm); | |
2376 | goto out_put_task; | |
2377 | } | |
640708a2 | 2378 | |
f0c3b509 | 2379 | nr_files = 0; |
640708a2 | 2380 | |
f0c3b509 AV |
2381 | /* |
2382 | * We need two passes here: | |
2383 | * | |
c1e8d7c6 ML |
2384 | * 1) Collect vmas of mapped files with mmap_lock taken |
2385 | * 2) Release mmap_lock and instantiate entries | |
f0c3b509 AV |
2386 | * |
2387 | * otherwise we get lockdep complained, since filldir() | |
c1e8d7c6 | 2388 | * routine might require mmap_lock taken in might_fault(). |
f0c3b509 | 2389 | */ |
640708a2 | 2390 | |
f0c3b509 | 2391 | for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) { |
94f8f3b0 KO |
2392 | if (!vma->vm_file) |
2393 | continue; | |
2394 | if (++pos <= ctx->pos) | |
2395 | continue; | |
f0c3b509 | 2396 | |
94f8f3b0 KO |
2397 | p = genradix_ptr_alloc(&fa, nr_files++, GFP_KERNEL); |
2398 | if (!p) { | |
f0c3b509 | 2399 | ret = -ENOMEM; |
89154dd5 | 2400 | mmap_read_unlock(mm); |
f0c3b509 AV |
2401 | mmput(mm); |
2402 | goto out_put_task; | |
640708a2 | 2403 | } |
f0c3b509 | 2404 | |
94f8f3b0 KO |
2405 | p->start = vma->vm_start; |
2406 | p->end = vma->vm_end; | |
2407 | p->mode = vma->vm_file->f_mode; | |
640708a2 | 2408 | } |
89154dd5 | 2409 | mmap_read_unlock(mm); |
fe079a5e | 2410 | mmput(mm); |
f0c3b509 AV |
2411 | |
2412 | for (i = 0; i < nr_files; i++) { | |
20d28cde AD |
2413 | char buf[4 * sizeof(long) + 2]; /* max: %lx-%lx\0 */ |
2414 | unsigned int len; | |
2415 | ||
94f8f3b0 | 2416 | p = genradix_ptr(&fa, i); |
20d28cde | 2417 | len = snprintf(buf, sizeof(buf), "%lx-%lx", p->start, p->end); |
f0c3b509 | 2418 | if (!proc_fill_cache(file, ctx, |
20d28cde | 2419 | buf, len, |
f0c3b509 AV |
2420 | proc_map_files_instantiate, |
2421 | task, | |
2422 | (void *)(unsigned long)p->mode)) | |
2423 | break; | |
2424 | ctx->pos++; | |
640708a2 PE |
2425 | } |
2426 | ||
640708a2 PE |
2427 | out_put_task: |
2428 | put_task_struct(task); | |
2429 | out: | |
94f8f3b0 | 2430 | genradix_free(&fa); |
640708a2 PE |
2431 | return ret; |
2432 | } | |
2433 | ||
2434 | static const struct file_operations proc_map_files_operations = { | |
2435 | .read = generic_read_dir, | |
f50752ea AV |
2436 | .iterate_shared = proc_map_files_readdir, |
2437 | .llseek = generic_file_llseek, | |
640708a2 PE |
2438 | }; |
2439 | ||
b18b6a9c | 2440 | #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS) |
48f6a7a5 PE |
2441 | struct timers_private { |
2442 | struct pid *pid; | |
2443 | struct task_struct *task; | |
2444 | struct sighand_struct *sighand; | |
57b8015e | 2445 | struct pid_namespace *ns; |
48f6a7a5 PE |
2446 | unsigned long flags; |
2447 | }; | |
2448 | ||
2449 | static void *timers_start(struct seq_file *m, loff_t *pos) | |
2450 | { | |
2451 | struct timers_private *tp = m->private; | |
2452 | ||
2453 | tp->task = get_pid_task(tp->pid, PIDTYPE_PID); | |
2454 | if (!tp->task) | |
2455 | return ERR_PTR(-ESRCH); | |
2456 | ||
2457 | tp->sighand = lock_task_sighand(tp->task, &tp->flags); | |
2458 | if (!tp->sighand) | |
2459 | return ERR_PTR(-ESRCH); | |
2460 | ||
2461 | return seq_list_start(&tp->task->signal->posix_timers, *pos); | |
2462 | } | |
2463 | ||
2464 | static void *timers_next(struct seq_file *m, void *v, loff_t *pos) | |
2465 | { | |
2466 | struct timers_private *tp = m->private; | |
2467 | return seq_list_next(v, &tp->task->signal->posix_timers, pos); | |
2468 | } | |
2469 | ||
2470 | static void timers_stop(struct seq_file *m, void *v) | |
2471 | { | |
2472 | struct timers_private *tp = m->private; | |
2473 | ||
2474 | if (tp->sighand) { | |
2475 | unlock_task_sighand(tp->task, &tp->flags); | |
2476 | tp->sighand = NULL; | |
2477 | } | |
2478 | ||
2479 | if (tp->task) { | |
2480 | put_task_struct(tp->task); | |
2481 | tp->task = NULL; | |
2482 | } | |
2483 | } | |
2484 | ||
2485 | static int show_timer(struct seq_file *m, void *v) | |
2486 | { | |
2487 | struct k_itimer *timer; | |
57b8015e PE |
2488 | struct timers_private *tp = m->private; |
2489 | int notify; | |
cedbccab | 2490 | static const char * const nstr[] = { |
57b8015e PE |
2491 | [SIGEV_SIGNAL] = "signal", |
2492 | [SIGEV_NONE] = "none", | |
2493 | [SIGEV_THREAD] = "thread", | |
2494 | }; | |
48f6a7a5 PE |
2495 | |
2496 | timer = list_entry((struct list_head *)v, struct k_itimer, list); | |
57b8015e PE |
2497 | notify = timer->it_sigev_notify; |
2498 | ||
48f6a7a5 | 2499 | seq_printf(m, "ID: %d\n", timer->it_id); |
ba3edf1f | 2500 | seq_printf(m, "signal: %d/%px\n", |
25ce3191 JP |
2501 | timer->sigq->info.si_signo, |
2502 | timer->sigq->info.si_value.sival_ptr); | |
57b8015e | 2503 | seq_printf(m, "notify: %s/%s.%d\n", |
25ce3191 JP |
2504 | nstr[notify & ~SIGEV_THREAD_ID], |
2505 | (notify & SIGEV_THREAD_ID) ? "tid" : "pid", | |
2506 | pid_nr_ns(timer->it_pid, tp->ns)); | |
15ef0298 | 2507 | seq_printf(m, "ClockID: %d\n", timer->it_clock); |
48f6a7a5 PE |
2508 | |
2509 | return 0; | |
2510 | } | |
2511 | ||
2512 | static const struct seq_operations proc_timers_seq_ops = { | |
2513 | .start = timers_start, | |
2514 | .next = timers_next, | |
2515 | .stop = timers_stop, | |
2516 | .show = show_timer, | |
2517 | }; | |
2518 | ||
2519 | static int proc_timers_open(struct inode *inode, struct file *file) | |
2520 | { | |
2521 | struct timers_private *tp; | |
2522 | ||
2523 | tp = __seq_open_private(file, &proc_timers_seq_ops, | |
2524 | sizeof(struct timers_private)); | |
2525 | if (!tp) | |
2526 | return -ENOMEM; | |
2527 | ||
2528 | tp->pid = proc_pid(inode); | |
9d78edea | 2529 | tp->ns = proc_pid_ns(inode->i_sb); |
48f6a7a5 PE |
2530 | return 0; |
2531 | } | |
2532 | ||
2533 | static const struct file_operations proc_timers_operations = { | |
2534 | .open = proc_timers_open, | |
2535 | .read = seq_read, | |
2536 | .llseek = seq_lseek, | |
2537 | .release = seq_release_private, | |
2538 | }; | |
b5946bea | 2539 | #endif |
640708a2 | 2540 | |
5de23d43 JS |
2541 | static ssize_t timerslack_ns_write(struct file *file, const char __user *buf, |
2542 | size_t count, loff_t *offset) | |
2543 | { | |
2544 | struct inode *inode = file_inode(file); | |
2545 | struct task_struct *p; | |
2546 | u64 slack_ns; | |
2547 | int err; | |
2548 | ||
2549 | err = kstrtoull_from_user(buf, count, 10, &slack_ns); | |
2550 | if (err < 0) | |
2551 | return err; | |
2552 | ||
2553 | p = get_proc_task(inode); | |
2554 | if (!p) | |
2555 | return -ESRCH; | |
2556 | ||
4b2bd5fe | 2557 | if (p != current) { |
8da0b4f6 BG |
2558 | rcu_read_lock(); |
2559 | if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) { | |
2560 | rcu_read_unlock(); | |
4b2bd5fe JS |
2561 | count = -EPERM; |
2562 | goto out; | |
2563 | } | |
8da0b4f6 | 2564 | rcu_read_unlock(); |
5de23d43 | 2565 | |
4b2bd5fe JS |
2566 | err = security_task_setscheduler(p); |
2567 | if (err) { | |
2568 | count = err; | |
2569 | goto out; | |
2570 | } | |
904763e1 JS |
2571 | } |
2572 | ||
7abbaf94 JS |
2573 | task_lock(p); |
2574 | if (slack_ns == 0) | |
2575 | p->timer_slack_ns = p->default_timer_slack_ns; | |
2576 | else | |
2577 | p->timer_slack_ns = slack_ns; | |
2578 | task_unlock(p); | |
2579 | ||
2580 | out: | |
5de23d43 JS |
2581 | put_task_struct(p); |
2582 | ||
2583 | return count; | |
2584 | } | |
2585 | ||
2586 | static int timerslack_ns_show(struct seq_file *m, void *v) | |
2587 | { | |
2588 | struct inode *inode = m->private; | |
2589 | struct task_struct *p; | |
7abbaf94 | 2590 | int err = 0; |
5de23d43 JS |
2591 | |
2592 | p = get_proc_task(inode); | |
2593 | if (!p) | |
2594 | return -ESRCH; | |
2595 | ||
4b2bd5fe | 2596 | if (p != current) { |
8da0b4f6 BG |
2597 | rcu_read_lock(); |
2598 | if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) { | |
2599 | rcu_read_unlock(); | |
4b2bd5fe JS |
2600 | err = -EPERM; |
2601 | goto out; | |
2602 | } | |
8da0b4f6 BG |
2603 | rcu_read_unlock(); |
2604 | ||
4b2bd5fe JS |
2605 | err = security_task_getscheduler(p); |
2606 | if (err) | |
2607 | goto out; | |
2608 | } | |
904763e1 | 2609 | |
7abbaf94 JS |
2610 | task_lock(p); |
2611 | seq_printf(m, "%llu\n", p->timer_slack_ns); | |
2612 | task_unlock(p); | |
2613 | ||
2614 | out: | |
5de23d43 JS |
2615 | put_task_struct(p); |
2616 | ||
2617 | return err; | |
2618 | } | |
2619 | ||
2620 | static int timerslack_ns_open(struct inode *inode, struct file *filp) | |
2621 | { | |
2622 | return single_open(filp, timerslack_ns_show, inode); | |
2623 | } | |
2624 | ||
2625 | static const struct file_operations proc_pid_set_timerslack_ns_operations = { | |
2626 | .open = timerslack_ns_open, | |
2627 | .read = seq_read, | |
2628 | .write = timerslack_ns_write, | |
2629 | .llseek = seq_lseek, | |
2630 | .release = single_release, | |
2631 | }; | |
2632 | ||
0168b9e3 AV |
2633 | static struct dentry *proc_pident_instantiate(struct dentry *dentry, |
2634 | struct task_struct *task, const void *ptr) | |
444ceed8 | 2635 | { |
c5141e6d | 2636 | const struct pid_entry *p = ptr; |
444ceed8 EB |
2637 | struct inode *inode; |
2638 | struct proc_inode *ei; | |
444ceed8 | 2639 | |
0168b9e3 | 2640 | inode = proc_pid_make_inode(dentry->d_sb, task, p->mode); |
444ceed8 | 2641 | if (!inode) |
0168b9e3 | 2642 | return ERR_PTR(-ENOENT); |
444ceed8 EB |
2643 | |
2644 | ei = PROC_I(inode); | |
444ceed8 | 2645 | if (S_ISDIR(inode->i_mode)) |
bfe86848 | 2646 | set_nlink(inode, 2); /* Use getattr to fix if necessary */ |
444ceed8 EB |
2647 | if (p->iop) |
2648 | inode->i_op = p->iop; | |
2649 | if (p->fop) | |
2650 | inode->i_fop = p->fop; | |
2651 | ei->op = p->op; | |
1bbc5513 | 2652 | pid_update_inode(task, inode); |
fb045adb | 2653 | d_set_d_op(dentry, &pid_dentry_operations); |
0168b9e3 | 2654 | return d_splice_alias(inode, dentry); |
444ceed8 EB |
2655 | } |
2656 | ||
1da177e4 LT |
2657 | static struct dentry *proc_pident_lookup(struct inode *dir, |
2658 | struct dentry *dentry, | |
d5a572a4 AD |
2659 | const struct pid_entry *p, |
2660 | const struct pid_entry *end) | |
1da177e4 | 2661 | { |
99f89551 | 2662 | struct task_struct *task = get_proc_task(dir); |
0168b9e3 | 2663 | struct dentry *res = ERR_PTR(-ENOENT); |
1da177e4 | 2664 | |
99f89551 EB |
2665 | if (!task) |
2666 | goto out_no_task; | |
1da177e4 | 2667 | |
20cdc894 EB |
2668 | /* |
2669 | * Yes, it does not scale. And it should not. Don't add | |
2670 | * new entries into /proc/<tgid>/ without very good reasons. | |
2671 | */ | |
d5a572a4 | 2672 | for (; p < end; p++) { |
1da177e4 LT |
2673 | if (p->len != dentry->d_name.len) |
2674 | continue; | |
26b95137 AD |
2675 | if (!memcmp(dentry->d_name.name, p->name, p->len)) { |
2676 | res = proc_pident_instantiate(dentry, task, p); | |
1da177e4 | 2677 | break; |
26b95137 | 2678 | } |
1da177e4 | 2679 | } |
99f89551 EB |
2680 | put_task_struct(task); |
2681 | out_no_task: | |
0168b9e3 | 2682 | return res; |
1da177e4 LT |
2683 | } |
2684 | ||
f0c3b509 | 2685 | static int proc_pident_readdir(struct file *file, struct dir_context *ctx, |
c5141e6d | 2686 | const struct pid_entry *ents, unsigned int nents) |
28a6d671 | 2687 | { |
f0c3b509 AV |
2688 | struct task_struct *task = get_proc_task(file_inode(file)); |
2689 | const struct pid_entry *p; | |
28a6d671 | 2690 | |
28a6d671 | 2691 | if (!task) |
f0c3b509 | 2692 | return -ENOENT; |
28a6d671 | 2693 | |
f0c3b509 AV |
2694 | if (!dir_emit_dots(file, ctx)) |
2695 | goto out; | |
2696 | ||
2697 | if (ctx->pos >= nents + 2) | |
2698 | goto out; | |
28a6d671 | 2699 | |
bac5f5d5 | 2700 | for (p = ents + (ctx->pos - 2); p < ents + nents; p++) { |
f0c3b509 AV |
2701 | if (!proc_fill_cache(file, ctx, p->name, p->len, |
2702 | proc_pident_instantiate, task, p)) | |
2703 | break; | |
2704 | ctx->pos++; | |
2705 | } | |
28a6d671 | 2706 | out: |
61a28784 | 2707 | put_task_struct(task); |
f0c3b509 | 2708 | return 0; |
1da177e4 LT |
2709 | } |
2710 | ||
28a6d671 | 2711 | #ifdef CONFIG_SECURITY |
591a22c1 KC |
2712 | static int proc_pid_attr_open(struct inode *inode, struct file *file) |
2713 | { | |
94f0b2d4 LT |
2714 | file->private_data = NULL; |
2715 | __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS); | |
2716 | return 0; | |
591a22c1 KC |
2717 | } |
2718 | ||
28a6d671 EB |
2719 | static ssize_t proc_pid_attr_read(struct file * file, char __user * buf, |
2720 | size_t count, loff_t *ppos) | |
2721 | { | |
496ad9aa | 2722 | struct inode * inode = file_inode(file); |
04ff9708 | 2723 | char *p = NULL; |
28a6d671 EB |
2724 | ssize_t length; |
2725 | struct task_struct *task = get_proc_task(inode); | |
2726 | ||
28a6d671 | 2727 | if (!task) |
04ff9708 | 2728 | return -ESRCH; |
28a6d671 | 2729 | |
6d9c939d | 2730 | length = security_getprocattr(task, PROC_I(inode)->op.lsm, |
c8e477c6 | 2731 | file->f_path.dentry->d_name.name, |
04ff9708 | 2732 | &p); |
28a6d671 | 2733 | put_task_struct(task); |
04ff9708 AV |
2734 | if (length > 0) |
2735 | length = simple_read_from_buffer(buf, count, ppos, p, length); | |
2736 | kfree(p); | |
28a6d671 | 2737 | return length; |
1da177e4 LT |
2738 | } |
2739 | ||
28a6d671 EB |
2740 | static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf, |
2741 | size_t count, loff_t *ppos) | |
2742 | { | |
496ad9aa | 2743 | struct inode * inode = file_inode(file); |
41089b6d | 2744 | struct task_struct *task; |
bb646cdb | 2745 | void *page; |
41089b6d | 2746 | int rv; |
b21507e2 | 2747 | |
bfb819ea | 2748 | /* A task may only write when it was the opener. */ |
591a22c1 | 2749 | if (file->private_data != current->mm) |
bfb819ea KC |
2750 | return -EPERM; |
2751 | ||
41089b6d AD |
2752 | rcu_read_lock(); |
2753 | task = pid_task(proc_pid(inode), PIDTYPE_PID); | |
2754 | if (!task) { | |
2755 | rcu_read_unlock(); | |
2756 | return -ESRCH; | |
2757 | } | |
b21507e2 | 2758 | /* A task may only write its own attributes. */ |
41089b6d AD |
2759 | if (current != task) { |
2760 | rcu_read_unlock(); | |
2761 | return -EACCES; | |
2762 | } | |
35a196be PM |
2763 | /* Prevent changes to overridden credentials. */ |
2764 | if (current_cred() != current_real_cred()) { | |
2765 | rcu_read_unlock(); | |
2766 | return -EBUSY; | |
2767 | } | |
41089b6d | 2768 | rcu_read_unlock(); |
b21507e2 | 2769 | |
28a6d671 EB |
2770 | if (count > PAGE_SIZE) |
2771 | count = PAGE_SIZE; | |
2772 | ||
2773 | /* No partial writes. */ | |
28a6d671 | 2774 | if (*ppos != 0) |
41089b6d | 2775 | return -EINVAL; |
28a6d671 | 2776 | |
bb646cdb AV |
2777 | page = memdup_user(buf, count); |
2778 | if (IS_ERR(page)) { | |
41089b6d | 2779 | rv = PTR_ERR(page); |
28a6d671 | 2780 | goto out; |
bb646cdb | 2781 | } |
28a6d671 | 2782 | |
107db7c7 | 2783 | /* Guard against adverse ptrace interaction */ |
41089b6d AD |
2784 | rv = mutex_lock_interruptible(¤t->signal->cred_guard_mutex); |
2785 | if (rv < 0) | |
107db7c7 DH |
2786 | goto out_free; |
2787 | ||
6d9c939d CS |
2788 | rv = security_setprocattr(PROC_I(inode)->op.lsm, |
2789 | file->f_path.dentry->d_name.name, page, | |
2790 | count); | |
b21507e2 | 2791 | mutex_unlock(¤t->signal->cred_guard_mutex); |
28a6d671 | 2792 | out_free: |
bb646cdb | 2793 | kfree(page); |
28a6d671 | 2794 | out: |
41089b6d | 2795 | return rv; |
28a6d671 EB |
2796 | } |
2797 | ||
00977a59 | 2798 | static const struct file_operations proc_pid_attr_operations = { |
591a22c1 | 2799 | .open = proc_pid_attr_open, |
28a6d671 EB |
2800 | .read = proc_pid_attr_read, |
2801 | .write = proc_pid_attr_write, | |
87df8424 | 2802 | .llseek = generic_file_llseek, |
591a22c1 | 2803 | .release = mem_release, |
28a6d671 EB |
2804 | }; |
2805 | ||
6d9c939d CS |
2806 | #define LSM_DIR_OPS(LSM) \ |
2807 | static int proc_##LSM##_attr_dir_iterate(struct file *filp, \ | |
2808 | struct dir_context *ctx) \ | |
2809 | { \ | |
2810 | return proc_pident_readdir(filp, ctx, \ | |
2811 | LSM##_attr_dir_stuff, \ | |
2812 | ARRAY_SIZE(LSM##_attr_dir_stuff)); \ | |
2813 | } \ | |
2814 | \ | |
2815 | static const struct file_operations proc_##LSM##_attr_dir_ops = { \ | |
2816 | .read = generic_read_dir, \ | |
2817 | .iterate = proc_##LSM##_attr_dir_iterate, \ | |
2818 | .llseek = default_llseek, \ | |
2819 | }; \ | |
2820 | \ | |
2821 | static struct dentry *proc_##LSM##_attr_dir_lookup(struct inode *dir, \ | |
2822 | struct dentry *dentry, unsigned int flags) \ | |
2823 | { \ | |
2824 | return proc_pident_lookup(dir, dentry, \ | |
2825 | LSM##_attr_dir_stuff, \ | |
d5a572a4 | 2826 | LSM##_attr_dir_stuff + ARRAY_SIZE(LSM##_attr_dir_stuff)); \ |
6d9c939d CS |
2827 | } \ |
2828 | \ | |
2829 | static const struct inode_operations proc_##LSM##_attr_dir_inode_ops = { \ | |
2830 | .lookup = proc_##LSM##_attr_dir_lookup, \ | |
2831 | .getattr = pid_getattr, \ | |
2832 | .setattr = proc_setattr, \ | |
2833 | } | |
2834 | ||
2835 | #ifdef CONFIG_SECURITY_SMACK | |
2836 | static const struct pid_entry smack_attr_dir_stuff[] = { | |
2837 | ATTR("smack", "current", 0666), | |
2838 | }; | |
2839 | LSM_DIR_OPS(smack); | |
2840 | #endif | |
2841 | ||
6413f852 JJ |
2842 | #ifdef CONFIG_SECURITY_APPARMOR |
2843 | static const struct pid_entry apparmor_attr_dir_stuff[] = { | |
2844 | ATTR("apparmor", "current", 0666), | |
2845 | ATTR("apparmor", "prev", 0444), | |
2846 | ATTR("apparmor", "exec", 0666), | |
2847 | }; | |
2848 | LSM_DIR_OPS(apparmor); | |
2849 | #endif | |
2850 | ||
c5141e6d | 2851 | static const struct pid_entry attr_dir_stuff[] = { |
6d9c939d CS |
2852 | ATTR(NULL, "current", 0666), |
2853 | ATTR(NULL, "prev", 0444), | |
2854 | ATTR(NULL, "exec", 0666), | |
2855 | ATTR(NULL, "fscreate", 0666), | |
2856 | ATTR(NULL, "keycreate", 0666), | |
2857 | ATTR(NULL, "sockcreate", 0666), | |
2858 | #ifdef CONFIG_SECURITY_SMACK | |
2859 | DIR("smack", 0555, | |
2860 | proc_smack_attr_dir_inode_ops, proc_smack_attr_dir_ops), | |
2861 | #endif | |
6413f852 JJ |
2862 | #ifdef CONFIG_SECURITY_APPARMOR |
2863 | DIR("apparmor", 0555, | |
2864 | proc_apparmor_attr_dir_inode_ops, proc_apparmor_attr_dir_ops), | |
2865 | #endif | |
28a6d671 EB |
2866 | }; |
2867 | ||
f0c3b509 | 2868 | static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx) |
28a6d671 | 2869 | { |
f0c3b509 AV |
2870 | return proc_pident_readdir(file, ctx, |
2871 | attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2872 | } |
2873 | ||
00977a59 | 2874 | static const struct file_operations proc_attr_dir_operations = { |
1da177e4 | 2875 | .read = generic_read_dir, |
f50752ea AV |
2876 | .iterate_shared = proc_attr_dir_readdir, |
2877 | .llseek = generic_file_llseek, | |
1da177e4 LT |
2878 | }; |
2879 | ||
72d9dcfc | 2880 | static struct dentry *proc_attr_dir_lookup(struct inode *dir, |
00cd8dd3 | 2881 | struct dentry *dentry, unsigned int flags) |
28a6d671 | 2882 | { |
7bcd6b0e | 2883 | return proc_pident_lookup(dir, dentry, |
d5a572a4 AD |
2884 | attr_dir_stuff, |
2885 | attr_dir_stuff + ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2886 | } |
2887 | ||
c5ef1c42 | 2888 | static const struct inode_operations proc_attr_dir_inode_operations = { |
72d9dcfc | 2889 | .lookup = proc_attr_dir_lookup, |
99f89551 | 2890 | .getattr = pid_getattr, |
6d76fa58 | 2891 | .setattr = proc_setattr, |
1da177e4 LT |
2892 | }; |
2893 | ||
28a6d671 EB |
2894 | #endif |
2895 | ||
698ba7b5 | 2896 | #ifdef CONFIG_ELF_CORE |
3cb4a0bb KH |
2897 | static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf, |
2898 | size_t count, loff_t *ppos) | |
2899 | { | |
496ad9aa | 2900 | struct task_struct *task = get_proc_task(file_inode(file)); |
3cb4a0bb KH |
2901 | struct mm_struct *mm; |
2902 | char buffer[PROC_NUMBUF]; | |
2903 | size_t len; | |
2904 | int ret; | |
2905 | ||
2906 | if (!task) | |
2907 | return -ESRCH; | |
2908 | ||
2909 | ret = 0; | |
2910 | mm = get_task_mm(task); | |
2911 | if (mm) { | |
2912 | len = snprintf(buffer, sizeof(buffer), "%08lx\n", | |
2913 | ((mm->flags & MMF_DUMP_FILTER_MASK) >> | |
2914 | MMF_DUMP_FILTER_SHIFT)); | |
2915 | mmput(mm); | |
2916 | ret = simple_read_from_buffer(buf, count, ppos, buffer, len); | |
2917 | } | |
2918 | ||
2919 | put_task_struct(task); | |
2920 | ||
2921 | return ret; | |
2922 | } | |
2923 | ||
2924 | static ssize_t proc_coredump_filter_write(struct file *file, | |
2925 | const char __user *buf, | |
2926 | size_t count, | |
2927 | loff_t *ppos) | |
2928 | { | |
2929 | struct task_struct *task; | |
2930 | struct mm_struct *mm; | |
3cb4a0bb KH |
2931 | unsigned int val; |
2932 | int ret; | |
2933 | int i; | |
2934 | unsigned long mask; | |
2935 | ||
774636e1 AD |
2936 | ret = kstrtouint_from_user(buf, count, 0, &val); |
2937 | if (ret < 0) | |
2938 | return ret; | |
3cb4a0bb KH |
2939 | |
2940 | ret = -ESRCH; | |
496ad9aa | 2941 | task = get_proc_task(file_inode(file)); |
3cb4a0bb KH |
2942 | if (!task) |
2943 | goto out_no_task; | |
2944 | ||
3cb4a0bb KH |
2945 | mm = get_task_mm(task); |
2946 | if (!mm) | |
2947 | goto out_no_mm; | |
41a0c249 | 2948 | ret = 0; |
3cb4a0bb KH |
2949 | |
2950 | for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) { | |
2951 | if (val & mask) | |
2952 | set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2953 | else | |
2954 | clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2955 | } | |
2956 | ||
2957 | mmput(mm); | |
2958 | out_no_mm: | |
2959 | put_task_struct(task); | |
2960 | out_no_task: | |
774636e1 AD |
2961 | if (ret < 0) |
2962 | return ret; | |
2963 | return count; | |
3cb4a0bb KH |
2964 | } |
2965 | ||
2966 | static const struct file_operations proc_coredump_filter_operations = { | |
2967 | .read = proc_coredump_filter_read, | |
2968 | .write = proc_coredump_filter_write, | |
87df8424 | 2969 | .llseek = generic_file_llseek, |
3cb4a0bb KH |
2970 | }; |
2971 | #endif | |
2972 | ||
aba76fdb | 2973 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 2974 | static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole) |
297c5d92 | 2975 | { |
940389b8 | 2976 | struct task_io_accounting acct = task->ioac; |
5995477a | 2977 | unsigned long flags; |
293eb1e7 | 2978 | int result; |
5995477a | 2979 | |
f7cfd871 | 2980 | result = down_read_killable(&task->signal->exec_update_lock); |
293eb1e7 VK |
2981 | if (result) |
2982 | return result; | |
2983 | ||
caaee623 | 2984 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) { |
293eb1e7 VK |
2985 | result = -EACCES; |
2986 | goto out_unlock; | |
2987 | } | |
1d1221f3 | 2988 | |
5995477a AR |
2989 | if (whole && lock_task_sighand(task, &flags)) { |
2990 | struct task_struct *t = task; | |
2991 | ||
2992 | task_io_accounting_add(&acct, &task->signal->ioac); | |
2993 | while_each_thread(task, t) | |
2994 | task_io_accounting_add(&acct, &t->ioac); | |
2995 | ||
2996 | unlock_task_sighand(task, &flags); | |
297c5d92 | 2997 | } |
25ce3191 JP |
2998 | seq_printf(m, |
2999 | "rchar: %llu\n" | |
3000 | "wchar: %llu\n" | |
3001 | "syscr: %llu\n" | |
3002 | "syscw: %llu\n" | |
3003 | "read_bytes: %llu\n" | |
3004 | "write_bytes: %llu\n" | |
3005 | "cancelled_write_bytes: %llu\n", | |
3006 | (unsigned long long)acct.rchar, | |
3007 | (unsigned long long)acct.wchar, | |
3008 | (unsigned long long)acct.syscr, | |
3009 | (unsigned long long)acct.syscw, | |
3010 | (unsigned long long)acct.read_bytes, | |
3011 | (unsigned long long)acct.write_bytes, | |
3012 | (unsigned long long)acct.cancelled_write_bytes); | |
3013 | result = 0; | |
3014 | ||
293eb1e7 | 3015 | out_unlock: |
f7cfd871 | 3016 | up_read(&task->signal->exec_update_lock); |
293eb1e7 | 3017 | return result; |
297c5d92 AR |
3018 | } |
3019 | ||
19aadc98 AD |
3020 | static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns, |
3021 | struct pid *pid, struct task_struct *task) | |
297c5d92 | 3022 | { |
19aadc98 | 3023 | return do_io_accounting(task, m, 0); |
aba76fdb | 3024 | } |
297c5d92 | 3025 | |
19aadc98 AD |
3026 | static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns, |
3027 | struct pid *pid, struct task_struct *task) | |
297c5d92 | 3028 | { |
19aadc98 | 3029 | return do_io_accounting(task, m, 1); |
297c5d92 AR |
3030 | } |
3031 | #endif /* CONFIG_TASK_IO_ACCOUNTING */ | |
aba76fdb | 3032 | |
22d917d8 EB |
3033 | #ifdef CONFIG_USER_NS |
3034 | static int proc_id_map_open(struct inode *inode, struct file *file, | |
ccf94f1b | 3035 | const struct seq_operations *seq_ops) |
22d917d8 EB |
3036 | { |
3037 | struct user_namespace *ns = NULL; | |
3038 | struct task_struct *task; | |
3039 | struct seq_file *seq; | |
3040 | int ret = -EINVAL; | |
3041 | ||
3042 | task = get_proc_task(inode); | |
3043 | if (task) { | |
3044 | rcu_read_lock(); | |
3045 | ns = get_user_ns(task_cred_xxx(task, user_ns)); | |
3046 | rcu_read_unlock(); | |
3047 | put_task_struct(task); | |
3048 | } | |
3049 | if (!ns) | |
3050 | goto err; | |
3051 | ||
3052 | ret = seq_open(file, seq_ops); | |
3053 | if (ret) | |
3054 | goto err_put_ns; | |
3055 | ||
3056 | seq = file->private_data; | |
3057 | seq->private = ns; | |
3058 | ||
3059 | return 0; | |
3060 | err_put_ns: | |
3061 | put_user_ns(ns); | |
3062 | err: | |
3063 | return ret; | |
3064 | } | |
3065 | ||
3066 | static int proc_id_map_release(struct inode *inode, struct file *file) | |
3067 | { | |
3068 | struct seq_file *seq = file->private_data; | |
3069 | struct user_namespace *ns = seq->private; | |
3070 | put_user_ns(ns); | |
3071 | return seq_release(inode, file); | |
3072 | } | |
3073 | ||
3074 | static int proc_uid_map_open(struct inode *inode, struct file *file) | |
3075 | { | |
3076 | return proc_id_map_open(inode, file, &proc_uid_seq_operations); | |
3077 | } | |
3078 | ||
3079 | static int proc_gid_map_open(struct inode *inode, struct file *file) | |
3080 | { | |
3081 | return proc_id_map_open(inode, file, &proc_gid_seq_operations); | |
3082 | } | |
3083 | ||
f76d207a EB |
3084 | static int proc_projid_map_open(struct inode *inode, struct file *file) |
3085 | { | |
3086 | return proc_id_map_open(inode, file, &proc_projid_seq_operations); | |
3087 | } | |
3088 | ||
22d917d8 EB |
3089 | static const struct file_operations proc_uid_map_operations = { |
3090 | .open = proc_uid_map_open, | |
3091 | .write = proc_uid_map_write, | |
3092 | .read = seq_read, | |
3093 | .llseek = seq_lseek, | |
3094 | .release = proc_id_map_release, | |
3095 | }; | |
3096 | ||
3097 | static const struct file_operations proc_gid_map_operations = { | |
3098 | .open = proc_gid_map_open, | |
3099 | .write = proc_gid_map_write, | |
3100 | .read = seq_read, | |
3101 | .llseek = seq_lseek, | |
3102 | .release = proc_id_map_release, | |
3103 | }; | |
f76d207a EB |
3104 | |
3105 | static const struct file_operations proc_projid_map_operations = { | |
3106 | .open = proc_projid_map_open, | |
3107 | .write = proc_projid_map_write, | |
3108 | .read = seq_read, | |
3109 | .llseek = seq_lseek, | |
3110 | .release = proc_id_map_release, | |
3111 | }; | |
9cc46516 EB |
3112 | |
3113 | static int proc_setgroups_open(struct inode *inode, struct file *file) | |
3114 | { | |
3115 | struct user_namespace *ns = NULL; | |
3116 | struct task_struct *task; | |
3117 | int ret; | |
3118 | ||
3119 | ret = -ESRCH; | |
3120 | task = get_proc_task(inode); | |
3121 | if (task) { | |
3122 | rcu_read_lock(); | |
3123 | ns = get_user_ns(task_cred_xxx(task, user_ns)); | |
3124 | rcu_read_unlock(); | |
3125 | put_task_struct(task); | |
3126 | } | |
3127 | if (!ns) | |
3128 | goto err; | |
3129 | ||
3130 | if (file->f_mode & FMODE_WRITE) { | |
3131 | ret = -EACCES; | |
3132 | if (!ns_capable(ns, CAP_SYS_ADMIN)) | |
3133 | goto err_put_ns; | |
3134 | } | |
3135 | ||
3136 | ret = single_open(file, &proc_setgroups_show, ns); | |
3137 | if (ret) | |
3138 | goto err_put_ns; | |
3139 | ||
3140 | return 0; | |
3141 | err_put_ns: | |
3142 | put_user_ns(ns); | |
3143 | err: | |
3144 | return ret; | |
3145 | } | |
3146 | ||
3147 | static int proc_setgroups_release(struct inode *inode, struct file *file) | |
3148 | { | |
3149 | struct seq_file *seq = file->private_data; | |
3150 | struct user_namespace *ns = seq->private; | |
3151 | int ret = single_release(inode, file); | |
3152 | put_user_ns(ns); | |
3153 | return ret; | |
3154 | } | |
3155 | ||
3156 | static const struct file_operations proc_setgroups_operations = { | |
3157 | .open = proc_setgroups_open, | |
3158 | .write = proc_setgroups_write, | |
3159 | .read = seq_read, | |
3160 | .llseek = seq_lseek, | |
3161 | .release = proc_setgroups_release, | |
3162 | }; | |
22d917d8 EB |
3163 | #endif /* CONFIG_USER_NS */ |
3164 | ||
47830723 KC |
3165 | static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns, |
3166 | struct pid *pid, struct task_struct *task) | |
3167 | { | |
a9712bc1 AV |
3168 | int err = lock_trace(task); |
3169 | if (!err) { | |
3170 | seq_printf(m, "%08x\n", task->personality); | |
3171 | unlock_trace(task); | |
3172 | } | |
3173 | return err; | |
47830723 KC |
3174 | } |
3175 | ||
7c23b330 JP |
3176 | #ifdef CONFIG_LIVEPATCH |
3177 | static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns, | |
3178 | struct pid *pid, struct task_struct *task) | |
3179 | { | |
3180 | seq_printf(m, "%d\n", task->patch_state); | |
3181 | return 0; | |
3182 | } | |
3183 | #endif /* CONFIG_LIVEPATCH */ | |
3184 | ||
76093853 | 3185 | #ifdef CONFIG_KSM |
3186 | static int proc_pid_ksm_merging_pages(struct seq_file *m, struct pid_namespace *ns, | |
3187 | struct pid *pid, struct task_struct *task) | |
3188 | { | |
3189 | struct mm_struct *mm; | |
3190 | ||
3191 | mm = get_task_mm(task); | |
3192 | if (mm) { | |
3193 | seq_printf(m, "%lu\n", mm->ksm_merging_pages); | |
3194 | mmput(mm); | |
3195 | } | |
3196 | ||
3197 | return 0; | |
3198 | } | |
3199 | #endif /* CONFIG_KSM */ | |
3200 | ||
c8d12627 AP |
3201 | #ifdef CONFIG_STACKLEAK_METRICS |
3202 | static int proc_stack_depth(struct seq_file *m, struct pid_namespace *ns, | |
3203 | struct pid *pid, struct task_struct *task) | |
3204 | { | |
3205 | unsigned long prev_depth = THREAD_SIZE - | |
3206 | (task->prev_lowest_stack & (THREAD_SIZE - 1)); | |
3207 | unsigned long depth = THREAD_SIZE - | |
3208 | (task->lowest_stack & (THREAD_SIZE - 1)); | |
3209 | ||
3210 | seq_printf(m, "previous stack depth: %lu\nstack depth: %lu\n", | |
3211 | prev_depth, depth); | |
3212 | return 0; | |
3213 | } | |
3214 | #endif /* CONFIG_STACKLEAK_METRICS */ | |
3215 | ||
28a6d671 EB |
3216 | /* |
3217 | * Thread groups | |
3218 | */ | |
00977a59 | 3219 | static const struct file_operations proc_task_operations; |
c5ef1c42 | 3220 | static const struct inode_operations proc_task_inode_operations; |
20cdc894 | 3221 | |
c5141e6d | 3222 | static const struct pid_entry tgid_base_stuff[] = { |
631f9c18 AD |
3223 | DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations), |
3224 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), | |
640708a2 | 3225 | DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations), |
7bc3fa01 | 3226 | DIR("fdinfo", S_IRUGO|S_IXUGO, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
6b4e306a | 3227 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
b2211a36 | 3228 | #ifdef CONFIG_NET |
631f9c18 | 3229 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), |
b2211a36 | 3230 | #endif |
631f9c18 | 3231 | REG("environ", S_IRUSR, proc_environ_operations), |
c5317167 | 3232 | REG("auxv", S_IRUSR, proc_auxv_operations), |
631f9c18 | 3233 | ONE("status", S_IRUGO, proc_pid_status), |
35a35046 | 3234 | ONE("personality", S_IRUSR, proc_pid_personality), |
1c963eb1 | 3235 | ONE("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 3236 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 3237 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
5091faa4 MG |
3238 | #endif |
3239 | #ifdef CONFIG_SCHED_AUTOGROUP | |
3240 | REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations), | |
04a8682a AV |
3241 | #endif |
3242 | #ifdef CONFIG_TIME_NS | |
3243 | REG("timens_offsets", S_IRUGO|S_IWUSR, proc_timens_offsets_operations), | |
ebcb6734 | 3244 | #endif |
4614a696 | 3245 | REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations), |
ebcb6734 | 3246 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 | 3247 | ONE("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 3248 | #endif |
c2c0bb44 | 3249 | REG("cmdline", S_IRUGO, proc_pid_cmdline_ops), |
631f9c18 AD |
3250 | ONE("stat", S_IRUGO, proc_tgid_stat), |
3251 | ONE("statm", S_IRUGO, proc_pid_statm), | |
b7643757 | 3252 | REG("maps", S_IRUGO, proc_pid_maps_operations), |
28a6d671 | 3253 | #ifdef CONFIG_NUMA |
b7643757 | 3254 | REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations), |
28a6d671 | 3255 | #endif |
631f9c18 AD |
3256 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
3257 | LNK("cwd", proc_cwd_link), | |
3258 | LNK("root", proc_root_link), | |
3259 | LNK("exe", proc_exe_link), | |
3260 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
3261 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
3262 | REG("mountstats", S_IRUSR, proc_mountstats_operations), | |
1e883281 | 3263 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 | 3264 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
b7643757 | 3265 | REG("smaps", S_IRUGO, proc_pid_smaps_operations), |
493b0e9d | 3266 | REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations), |
32ed74a4 | 3267 | REG("pagemap", S_IRUSR, proc_pagemap_operations), |
28a6d671 EB |
3268 | #endif |
3269 | #ifdef CONFIG_SECURITY | |
631f9c18 | 3270 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
3271 | #endif |
3272 | #ifdef CONFIG_KALLSYMS | |
edfcd606 | 3273 | ONE("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 3274 | #endif |
2ec220e2 | 3275 | #ifdef CONFIG_STACKTRACE |
35a35046 | 3276 | ONE("stack", S_IRUSR, proc_pid_stack), |
28a6d671 | 3277 | #endif |
5968cece | 3278 | #ifdef CONFIG_SCHED_INFO |
f6e826ca | 3279 | ONE("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 3280 | #endif |
9745512c | 3281 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 3282 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 3283 | #endif |
8793d854 | 3284 | #ifdef CONFIG_PROC_PID_CPUSET |
52de4779 | 3285 | ONE("cpuset", S_IRUGO, proc_cpuset_show), |
a424316c PM |
3286 | #endif |
3287 | #ifdef CONFIG_CGROUPS | |
006f4ac4 | 3288 | ONE("cgroup", S_IRUGO, proc_cgroup_show), |
e79f15a4 CY |
3289 | #endif |
3290 | #ifdef CONFIG_PROC_CPU_RESCTRL | |
3291 | ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show), | |
28a6d671 | 3292 | #endif |
6ba51e37 | 3293 | ONE("oom_score", S_IRUGO, proc_oom_score), |
fa0cbbf1 | 3294 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations), |
a63d83f4 | 3295 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
4b7d248b | 3296 | #ifdef CONFIG_AUDIT |
631f9c18 AD |
3297 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
3298 | REG("sessionid", S_IRUGO, proc_sessionid_operations), | |
28a6d671 | 3299 | #endif |
f4f154fd | 3300 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 3301 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
168c42bc | 3302 | REG("fail-nth", 0644, proc_fail_nth_operations), |
f4f154fd | 3303 | #endif |
698ba7b5 | 3304 | #ifdef CONFIG_ELF_CORE |
631f9c18 | 3305 | REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations), |
3cb4a0bb | 3306 | #endif |
aba76fdb | 3307 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 3308 | ONE("io", S_IRUSR, proc_tgid_io_accounting), |
aba76fdb | 3309 | #endif |
22d917d8 EB |
3310 | #ifdef CONFIG_USER_NS |
3311 | REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations), | |
3312 | REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations), | |
f76d207a | 3313 | REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations), |
9cc46516 | 3314 | REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations), |
22d917d8 | 3315 | #endif |
b18b6a9c | 3316 | #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS) |
48f6a7a5 PE |
3317 | REG("timers", S_IRUGO, proc_timers_operations), |
3318 | #endif | |
5de23d43 | 3319 | REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations), |
7c23b330 JP |
3320 | #ifdef CONFIG_LIVEPATCH |
3321 | ONE("patch_state", S_IRUSR, proc_pid_patch_state), | |
3322 | #endif | |
c8d12627 AP |
3323 | #ifdef CONFIG_STACKLEAK_METRICS |
3324 | ONE("stack_depth", S_IRUGO, proc_stack_depth), | |
3325 | #endif | |
68bc30bb AL |
3326 | #ifdef CONFIG_PROC_PID_ARCH_STATUS |
3327 | ONE("arch_status", S_IRUGO, proc_pid_arch_status), | |
3328 | #endif | |
0d8315dd YZ |
3329 | #ifdef CONFIG_SECCOMP_CACHE_DEBUG |
3330 | ONE("seccomp_cache", S_IRUSR, proc_pid_seccomp_cache), | |
3331 | #endif | |
76093853 | 3332 | #ifdef CONFIG_KSM |
3333 | ONE("ksm_merging_pages", S_IRUSR, proc_pid_ksm_merging_pages), | |
3334 | #endif | |
28a6d671 | 3335 | }; |
1da177e4 | 3336 | |
f0c3b509 | 3337 | static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3338 | { |
f0c3b509 AV |
3339 | return proc_pident_readdir(file, ctx, |
3340 | tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
3341 | } |
3342 | ||
00977a59 | 3343 | static const struct file_operations proc_tgid_base_operations = { |
1da177e4 | 3344 | .read = generic_read_dir, |
f50752ea AV |
3345 | .iterate_shared = proc_tgid_base_readdir, |
3346 | .llseek = generic_file_llseek, | |
1da177e4 LT |
3347 | }; |
3348 | ||
3eb39f47 CB |
3349 | struct pid *tgid_pidfd_to_pid(const struct file *file) |
3350 | { | |
30d158b1 | 3351 | if (file->f_op != &proc_tgid_base_operations) |
3eb39f47 CB |
3352 | return ERR_PTR(-EBADF); |
3353 | ||
3354 | return proc_pid(file_inode(file)); | |
3355 | } | |
3356 | ||
00cd8dd3 AV |
3357 | static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
3358 | { | |
7bcd6b0e | 3359 | return proc_pident_lookup(dir, dentry, |
d5a572a4 AD |
3360 | tgid_base_stuff, |
3361 | tgid_base_stuff + ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
3362 | } |
3363 | ||
c5ef1c42 | 3364 | static const struct inode_operations proc_tgid_base_inode_operations = { |
28a6d671 | 3365 | .lookup = proc_tgid_base_lookup, |
99f89551 | 3366 | .getattr = pid_getattr, |
6d76fa58 | 3367 | .setattr = proc_setattr, |
0499680a | 3368 | .permission = proc_pid_permission, |
1da177e4 | 3369 | }; |
1da177e4 | 3370 | |
0895e91d | 3371 | /** |
7bc3e6e5 EB |
3372 | * proc_flush_pid - Remove dcache entries for @pid from the /proc dcache. |
3373 | * @pid: pid that should be flushed. | |
0895e91d | 3374 | * |
7bc3e6e5 EB |
3375 | * This function walks a list of inodes (that belong to any proc |
3376 | * filesystem) that are attached to the pid and flushes them from | |
3377 | * the dentry cache. | |
0895e91d RD |
3378 | * |
3379 | * It is safe and reasonable to cache /proc entries for a task until | |
3380 | * that task exits. After that they just clog up the dcache with | |
3381 | * useless entries, possibly causing useful dcache entries to be | |
7bc3e6e5 EB |
3382 | * flushed instead. This routine is provided to flush those useless |
3383 | * dcache entries when a process is reaped. | |
0895e91d RD |
3384 | * |
3385 | * NOTE: This routine is just an optimization so it does not guarantee | |
7bc3e6e5 EB |
3386 | * that no dcache entries will exist after a process is reaped |
3387 | * it just makes it very unlikely that any will persist. | |
60347f67 PE |
3388 | */ |
3389 | ||
7bc3e6e5 | 3390 | void proc_flush_pid(struct pid *pid) |
60347f67 | 3391 | { |
63f818f4 | 3392 | proc_invalidate_siblings_dcache(&pid->inodes, &pid->lock); |
60347f67 PE |
3393 | } |
3394 | ||
0168b9e3 | 3395 | static struct dentry *proc_pid_instantiate(struct dentry * dentry, |
c52a47ac | 3396 | struct task_struct *task, const void *ptr) |
444ceed8 | 3397 | { |
444ceed8 EB |
3398 | struct inode *inode; |
3399 | ||
d919a1e7 ZC |
3400 | inode = proc_pid_make_base_inode(dentry->d_sb, task, |
3401 | S_IFDIR | S_IRUGO | S_IXUGO); | |
444ceed8 | 3402 | if (!inode) |
0168b9e3 | 3403 | return ERR_PTR(-ENOENT); |
444ceed8 | 3404 | |
444ceed8 EB |
3405 | inode->i_op = &proc_tgid_base_inode_operations; |
3406 | inode->i_fop = &proc_tgid_base_operations; | |
3407 | inode->i_flags|=S_IMMUTABLE; | |
aed54175 | 3408 | |
1270dd8d | 3409 | set_nlink(inode, nlink_tgid); |
1bbc5513 | 3410 | pid_update_inode(task, inode); |
444ceed8 | 3411 | |
fb045adb | 3412 | d_set_d_op(dentry, &pid_dentry_operations); |
0168b9e3 | 3413 | return d_splice_alias(inode, dentry); |
444ceed8 EB |
3414 | } |
3415 | ||
867aaccf | 3416 | struct dentry *proc_pid_lookup(struct dentry *dentry, unsigned int flags) |
1da177e4 LT |
3417 | { |
3418 | struct task_struct *task; | |
1da177e4 | 3419 | unsigned tgid; |
fa10fed3 | 3420 | struct proc_fs_info *fs_info; |
b488893a | 3421 | struct pid_namespace *ns; |
0168b9e3 | 3422 | struct dentry *result = ERR_PTR(-ENOENT); |
1da177e4 | 3423 | |
dbcdb504 | 3424 | tgid = name_to_int(&dentry->d_name); |
1da177e4 LT |
3425 | if (tgid == ~0U) |
3426 | goto out; | |
3427 | ||
fa10fed3 AG |
3428 | fs_info = proc_sb_info(dentry->d_sb); |
3429 | ns = fs_info->pid_ns; | |
de758734 | 3430 | rcu_read_lock(); |
b488893a | 3431 | task = find_task_by_pid_ns(tgid, ns); |
1da177e4 LT |
3432 | if (task) |
3433 | get_task_struct(task); | |
de758734 | 3434 | rcu_read_unlock(); |
1da177e4 LT |
3435 | if (!task) |
3436 | goto out; | |
3437 | ||
24a71ce5 AG |
3438 | /* Limit procfs to only ptraceable tasks */ |
3439 | if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE) { | |
3440 | if (!has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS)) | |
3441 | goto out_put_task; | |
3442 | } | |
3443 | ||
0168b9e3 | 3444 | result = proc_pid_instantiate(dentry, task, NULL); |
24a71ce5 | 3445 | out_put_task: |
1da177e4 | 3446 | put_task_struct(task); |
1da177e4 | 3447 | out: |
0168b9e3 | 3448 | return result; |
1da177e4 LT |
3449 | } |
3450 | ||
1da177e4 | 3451 | /* |
0804ef4b | 3452 | * Find the first task with tgid >= tgid |
0bc58a91 | 3453 | * |
1da177e4 | 3454 | */ |
19fd4bb2 EB |
3455 | struct tgid_iter { |
3456 | unsigned int tgid; | |
0804ef4b | 3457 | struct task_struct *task; |
19fd4bb2 EB |
3458 | }; |
3459 | static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter) | |
3460 | { | |
0804ef4b | 3461 | struct pid *pid; |
1da177e4 | 3462 | |
19fd4bb2 EB |
3463 | if (iter.task) |
3464 | put_task_struct(iter.task); | |
454cc105 | 3465 | rcu_read_lock(); |
0804ef4b | 3466 | retry: |
19fd4bb2 EB |
3467 | iter.task = NULL; |
3468 | pid = find_ge_pid(iter.tgid, ns); | |
0804ef4b | 3469 | if (pid) { |
19fd4bb2 | 3470 | iter.tgid = pid_nr_ns(pid, ns); |
3147d8aa EB |
3471 | iter.task = pid_task(pid, PIDTYPE_TGID); |
3472 | if (!iter.task) { | |
19fd4bb2 | 3473 | iter.tgid += 1; |
0804ef4b | 3474 | goto retry; |
19fd4bb2 EB |
3475 | } |
3476 | get_task_struct(iter.task); | |
0bc58a91 | 3477 | } |
454cc105 | 3478 | rcu_read_unlock(); |
19fd4bb2 | 3479 | return iter; |
1da177e4 LT |
3480 | } |
3481 | ||
0097875b | 3482 | #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2) |
0804ef4b | 3483 | |
1da177e4 | 3484 | /* for the /proc/ directory itself, after non-process stuff has been done */ |
f0c3b509 | 3485 | int proc_pid_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3486 | { |
19fd4bb2 | 3487 | struct tgid_iter iter; |
fa10fed3 | 3488 | struct proc_fs_info *fs_info = proc_sb_info(file_inode(file)->i_sb); |
9d78edea | 3489 | struct pid_namespace *ns = proc_pid_ns(file_inode(file)->i_sb); |
f0c3b509 | 3490 | loff_t pos = ctx->pos; |
1da177e4 | 3491 | |
021ada7d | 3492 | if (pos >= PID_MAX_LIMIT + TGID_OFFSET) |
f0c3b509 | 3493 | return 0; |
1da177e4 | 3494 | |
0097875b | 3495 | if (pos == TGID_OFFSET - 2) { |
fa10fed3 | 3496 | struct inode *inode = d_inode(fs_info->proc_self); |
db963164 | 3497 | if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK)) |
f0c3b509 | 3498 | return 0; |
0097875b EB |
3499 | ctx->pos = pos = pos + 1; |
3500 | } | |
3501 | if (pos == TGID_OFFSET - 1) { | |
fa10fed3 | 3502 | struct inode *inode = d_inode(fs_info->proc_thread_self); |
0097875b EB |
3503 | if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK)) |
3504 | return 0; | |
3505 | ctx->pos = pos = pos + 1; | |
021ada7d | 3506 | } |
0097875b | 3507 | iter.tgid = pos - TGID_OFFSET; |
19fd4bb2 | 3508 | iter.task = NULL; |
19fd4bb2 EB |
3509 | for (iter = next_tgid(ns, iter); |
3510 | iter.task; | |
3511 | iter.tgid += 1, iter = next_tgid(ns, iter)) { | |
e3912ac3 | 3512 | char name[10 + 1]; |
a4ef3895 | 3513 | unsigned int len; |
3ba4bcee ED |
3514 | |
3515 | cond_resched(); | |
fa10fed3 | 3516 | if (!has_pid_permissions(fs_info, iter.task, HIDEPID_INVISIBLE)) |
f0c3b509 | 3517 | continue; |
0499680a | 3518 | |
e3912ac3 | 3519 | len = snprintf(name, sizeof(name), "%u", iter.tgid); |
f0c3b509 AV |
3520 | ctx->pos = iter.tgid + TGID_OFFSET; |
3521 | if (!proc_fill_cache(file, ctx, name, len, | |
3522 | proc_pid_instantiate, iter.task, NULL)) { | |
19fd4bb2 | 3523 | put_task_struct(iter.task); |
f0c3b509 | 3524 | return 0; |
1da177e4 | 3525 | } |
0bc58a91 | 3526 | } |
f0c3b509 | 3527 | ctx->pos = PID_MAX_LIMIT + TGID_OFFSET; |
0bc58a91 EB |
3528 | return 0; |
3529 | } | |
1da177e4 | 3530 | |
1b3044e3 JD |
3531 | /* |
3532 | * proc_tid_comm_permission is a special permission function exclusively | |
3533 | * used for the node /proc/<pid>/task/<tid>/comm. | |
3534 | * It bypasses generic permission checks in the case where a task of the same | |
3535 | * task group attempts to access the node. | |
3536 | * The rationale behind this is that glibc and bionic access this node for | |
3537 | * cross thread naming (pthread_set/getname_np(!self)). However, if | |
3538 | * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0, | |
3539 | * which locks out the cross thread naming implementation. | |
3540 | * This function makes sure that the node is always accessible for members of | |
3541 | * same thread group. | |
3542 | */ | |
549c7297 CB |
3543 | static int proc_tid_comm_permission(struct user_namespace *mnt_userns, |
3544 | struct inode *inode, int mask) | |
1b3044e3 JD |
3545 | { |
3546 | bool is_same_tgroup; | |
3547 | struct task_struct *task; | |
3548 | ||
3549 | task = get_proc_task(inode); | |
3550 | if (!task) | |
3551 | return -ESRCH; | |
3552 | is_same_tgroup = same_thread_group(current, task); | |
3553 | put_task_struct(task); | |
3554 | ||
3555 | if (likely(is_same_tgroup && !(mask & MAY_EXEC))) { | |
3556 | /* This file (/proc/<pid>/task/<tid>/comm) can always be | |
3557 | * read or written by the members of the corresponding | |
3558 | * thread group. | |
3559 | */ | |
3560 | return 0; | |
3561 | } | |
3562 | ||
47291baa | 3563 | return generic_permission(&init_user_ns, inode, mask); |
1b3044e3 JD |
3564 | } |
3565 | ||
3566 | static const struct inode_operations proc_tid_comm_inode_operations = { | |
3567 | .permission = proc_tid_comm_permission, | |
3568 | }; | |
3569 | ||
28a6d671 EB |
3570 | /* |
3571 | * Tasks | |
3572 | */ | |
c5141e6d | 3573 | static const struct pid_entry tid_base_stuff[] = { |
631f9c18 | 3574 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), |
7bc3fa01 | 3575 | DIR("fdinfo", S_IRUGO|S_IXUGO, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
6b4e306a | 3576 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
6ba8ed79 EB |
3577 | #ifdef CONFIG_NET |
3578 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), | |
3579 | #endif | |
631f9c18 | 3580 | REG("environ", S_IRUSR, proc_environ_operations), |
c5317167 | 3581 | REG("auxv", S_IRUSR, proc_auxv_operations), |
631f9c18 | 3582 | ONE("status", S_IRUGO, proc_pid_status), |
35a35046 | 3583 | ONE("personality", S_IRUSR, proc_pid_personality), |
1c963eb1 | 3584 | ONE("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 3585 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 3586 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
ebcb6734 | 3587 | #endif |
1b3044e3 JD |
3588 | NOD("comm", S_IFREG|S_IRUGO|S_IWUSR, |
3589 | &proc_tid_comm_inode_operations, | |
3590 | &proc_pid_set_comm_operations, {}), | |
ebcb6734 | 3591 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 | 3592 | ONE("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 3593 | #endif |
c2c0bb44 | 3594 | REG("cmdline", S_IRUGO, proc_pid_cmdline_ops), |
631f9c18 AD |
3595 | ONE("stat", S_IRUGO, proc_tid_stat), |
3596 | ONE("statm", S_IRUGO, proc_pid_statm), | |
871305bb | 3597 | REG("maps", S_IRUGO, proc_pid_maps_operations), |
2e13ba54 | 3598 | #ifdef CONFIG_PROC_CHILDREN |
81841161 CG |
3599 | REG("children", S_IRUGO, proc_tid_children_operations), |
3600 | #endif | |
28a6d671 | 3601 | #ifdef CONFIG_NUMA |
871305bb | 3602 | REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations), |
28a6d671 | 3603 | #endif |
631f9c18 AD |
3604 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
3605 | LNK("cwd", proc_cwd_link), | |
3606 | LNK("root", proc_root_link), | |
3607 | LNK("exe", proc_exe_link), | |
3608 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
3609 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
1e883281 | 3610 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 | 3611 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
871305bb | 3612 | REG("smaps", S_IRUGO, proc_pid_smaps_operations), |
493b0e9d | 3613 | REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations), |
32ed74a4 | 3614 | REG("pagemap", S_IRUSR, proc_pagemap_operations), |
28a6d671 EB |
3615 | #endif |
3616 | #ifdef CONFIG_SECURITY | |
631f9c18 | 3617 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
3618 | #endif |
3619 | #ifdef CONFIG_KALLSYMS | |
edfcd606 | 3620 | ONE("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 3621 | #endif |
2ec220e2 | 3622 | #ifdef CONFIG_STACKTRACE |
35a35046 | 3623 | ONE("stack", S_IRUSR, proc_pid_stack), |
28a6d671 | 3624 | #endif |
5968cece | 3625 | #ifdef CONFIG_SCHED_INFO |
f6e826ca | 3626 | ONE("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 3627 | #endif |
9745512c | 3628 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 3629 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 3630 | #endif |
8793d854 | 3631 | #ifdef CONFIG_PROC_PID_CPUSET |
52de4779 | 3632 | ONE("cpuset", S_IRUGO, proc_cpuset_show), |
a424316c PM |
3633 | #endif |
3634 | #ifdef CONFIG_CGROUPS | |
006f4ac4 | 3635 | ONE("cgroup", S_IRUGO, proc_cgroup_show), |
e79f15a4 CY |
3636 | #endif |
3637 | #ifdef CONFIG_PROC_CPU_RESCTRL | |
3638 | ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show), | |
28a6d671 | 3639 | #endif |
6ba51e37 | 3640 | ONE("oom_score", S_IRUGO, proc_oom_score), |
fa0cbbf1 | 3641 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations), |
a63d83f4 | 3642 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
4b7d248b | 3643 | #ifdef CONFIG_AUDIT |
631f9c18 | 3644 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
26ec3c64 | 3645 | REG("sessionid", S_IRUGO, proc_sessionid_operations), |
28a6d671 | 3646 | #endif |
f4f154fd | 3647 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 3648 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
1203c8e6 | 3649 | REG("fail-nth", 0644, proc_fail_nth_operations), |
f4f154fd | 3650 | #endif |
297c5d92 | 3651 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 3652 | ONE("io", S_IRUSR, proc_tid_io_accounting), |
297c5d92 | 3653 | #endif |
22d917d8 EB |
3654 | #ifdef CONFIG_USER_NS |
3655 | REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations), | |
3656 | REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations), | |
f76d207a | 3657 | REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations), |
9cc46516 | 3658 | REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations), |
22d917d8 | 3659 | #endif |
7c23b330 JP |
3660 | #ifdef CONFIG_LIVEPATCH |
3661 | ONE("patch_state", S_IRUSR, proc_pid_patch_state), | |
3662 | #endif | |
68bc30bb AL |
3663 | #ifdef CONFIG_PROC_PID_ARCH_STATUS |
3664 | ONE("arch_status", S_IRUGO, proc_pid_arch_status), | |
3665 | #endif | |
0d8315dd YZ |
3666 | #ifdef CONFIG_SECCOMP_CACHE_DEBUG |
3667 | ONE("seccomp_cache", S_IRUSR, proc_pid_seccomp_cache), | |
3668 | #endif | |
76093853 | 3669 | #ifdef CONFIG_KSM |
3670 | ONE("ksm_merging_pages", S_IRUSR, proc_pid_ksm_merging_pages), | |
3671 | #endif | |
28a6d671 EB |
3672 | }; |
3673 | ||
f0c3b509 | 3674 | static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx) |
28a6d671 | 3675 | { |
f0c3b509 AV |
3676 | return proc_pident_readdir(file, ctx, |
3677 | tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
3678 | } |
3679 | ||
00cd8dd3 AV |
3680 | static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
3681 | { | |
7bcd6b0e | 3682 | return proc_pident_lookup(dir, dentry, |
d5a572a4 AD |
3683 | tid_base_stuff, |
3684 | tid_base_stuff + ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
3685 | } |
3686 | ||
00977a59 | 3687 | static const struct file_operations proc_tid_base_operations = { |
28a6d671 | 3688 | .read = generic_read_dir, |
f50752ea AV |
3689 | .iterate_shared = proc_tid_base_readdir, |
3690 | .llseek = generic_file_llseek, | |
28a6d671 EB |
3691 | }; |
3692 | ||
c5ef1c42 | 3693 | static const struct inode_operations proc_tid_base_inode_operations = { |
28a6d671 EB |
3694 | .lookup = proc_tid_base_lookup, |
3695 | .getattr = pid_getattr, | |
3696 | .setattr = proc_setattr, | |
3697 | }; | |
3698 | ||
0168b9e3 AV |
3699 | static struct dentry *proc_task_instantiate(struct dentry *dentry, |
3700 | struct task_struct *task, const void *ptr) | |
444ceed8 | 3701 | { |
444ceed8 | 3702 | struct inode *inode; |
d919a1e7 ZC |
3703 | inode = proc_pid_make_base_inode(dentry->d_sb, task, |
3704 | S_IFDIR | S_IRUGO | S_IXUGO); | |
444ceed8 | 3705 | if (!inode) |
0168b9e3 | 3706 | return ERR_PTR(-ENOENT); |
1bbc5513 | 3707 | |
444ceed8 EB |
3708 | inode->i_op = &proc_tid_base_inode_operations; |
3709 | inode->i_fop = &proc_tid_base_operations; | |
1bbc5513 | 3710 | inode->i_flags |= S_IMMUTABLE; |
aed54175 | 3711 | |
1270dd8d | 3712 | set_nlink(inode, nlink_tid); |
1bbc5513 | 3713 | pid_update_inode(task, inode); |
444ceed8 | 3714 | |
fb045adb | 3715 | d_set_d_op(dentry, &pid_dentry_operations); |
0168b9e3 | 3716 | return d_splice_alias(inode, dentry); |
444ceed8 EB |
3717 | } |
3718 | ||
00cd8dd3 | 3719 | static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags) |
28a6d671 | 3720 | { |
28a6d671 EB |
3721 | struct task_struct *task; |
3722 | struct task_struct *leader = get_proc_task(dir); | |
28a6d671 | 3723 | unsigned tid; |
fa10fed3 | 3724 | struct proc_fs_info *fs_info; |
b488893a | 3725 | struct pid_namespace *ns; |
0168b9e3 | 3726 | struct dentry *result = ERR_PTR(-ENOENT); |
28a6d671 EB |
3727 | |
3728 | if (!leader) | |
3729 | goto out_no_task; | |
3730 | ||
dbcdb504 | 3731 | tid = name_to_int(&dentry->d_name); |
28a6d671 EB |
3732 | if (tid == ~0U) |
3733 | goto out; | |
3734 | ||
fa10fed3 AG |
3735 | fs_info = proc_sb_info(dentry->d_sb); |
3736 | ns = fs_info->pid_ns; | |
28a6d671 | 3737 | rcu_read_lock(); |
b488893a | 3738 | task = find_task_by_pid_ns(tid, ns); |
28a6d671 EB |
3739 | if (task) |
3740 | get_task_struct(task); | |
3741 | rcu_read_unlock(); | |
3742 | if (!task) | |
3743 | goto out; | |
bac0abd6 | 3744 | if (!same_thread_group(leader, task)) |
28a6d671 EB |
3745 | goto out_drop_task; |
3746 | ||
0168b9e3 | 3747 | result = proc_task_instantiate(dentry, task, NULL); |
28a6d671 EB |
3748 | out_drop_task: |
3749 | put_task_struct(task); | |
3750 | out: | |
3751 | put_task_struct(leader); | |
3752 | out_no_task: | |
0168b9e3 | 3753 | return result; |
28a6d671 EB |
3754 | } |
3755 | ||
0bc58a91 EB |
3756 | /* |
3757 | * Find the first tid of a thread group to return to user space. | |
3758 | * | |
3759 | * Usually this is just the thread group leader, but if the users | |
3760 | * buffer was too small or there was a seek into the middle of the | |
3761 | * directory we have more work todo. | |
3762 | * | |
3763 | * In the case of a short read we start with find_task_by_pid. | |
3764 | * | |
3765 | * In the case of a seek we start with the leader and walk nr | |
3766 | * threads past it. | |
3767 | */ | |
9f6e963f ON |
3768 | static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos, |
3769 | struct pid_namespace *ns) | |
0bc58a91 | 3770 | { |
d855a4b7 | 3771 | struct task_struct *pos, *task; |
9f6e963f ON |
3772 | unsigned long nr = f_pos; |
3773 | ||
3774 | if (nr != f_pos) /* 32bit overflow? */ | |
3775 | return NULL; | |
1da177e4 | 3776 | |
cc288738 | 3777 | rcu_read_lock(); |
d855a4b7 ON |
3778 | task = pid_task(pid, PIDTYPE_PID); |
3779 | if (!task) | |
3780 | goto fail; | |
3781 | ||
3782 | /* Attempt to start with the tid of a thread */ | |
9f6e963f | 3783 | if (tid && nr) { |
b488893a | 3784 | pos = find_task_by_pid_ns(tid, ns); |
d855a4b7 | 3785 | if (pos && same_thread_group(pos, task)) |
a872ff0c | 3786 | goto found; |
0bc58a91 | 3787 | } |
1da177e4 | 3788 | |
0bc58a91 | 3789 | /* If nr exceeds the number of threads there is nothing todo */ |
9f6e963f | 3790 | if (nr >= get_nr_threads(task)) |
c986c14a | 3791 | goto fail; |
1da177e4 | 3792 | |
a872ff0c ON |
3793 | /* If we haven't found our starting place yet start |
3794 | * with the leader and walk nr threads forward. | |
0bc58a91 | 3795 | */ |
d855a4b7 | 3796 | pos = task = task->group_leader; |
c986c14a | 3797 | do { |
9f6e963f | 3798 | if (!nr--) |
c986c14a | 3799 | goto found; |
d855a4b7 | 3800 | } while_each_thread(task, pos); |
c986c14a ON |
3801 | fail: |
3802 | pos = NULL; | |
3803 | goto out; | |
a872ff0c ON |
3804 | found: |
3805 | get_task_struct(pos); | |
3806 | out: | |
cc288738 | 3807 | rcu_read_unlock(); |
0bc58a91 EB |
3808 | return pos; |
3809 | } | |
3810 | ||
3811 | /* | |
3812 | * Find the next thread in the thread list. | |
3813 | * Return NULL if there is an error or no next thread. | |
3814 | * | |
3815 | * The reference to the input task_struct is released. | |
3816 | */ | |
3817 | static struct task_struct *next_tid(struct task_struct *start) | |
3818 | { | |
c1df7fb8 | 3819 | struct task_struct *pos = NULL; |
cc288738 | 3820 | rcu_read_lock(); |
c1df7fb8 | 3821 | if (pid_alive(start)) { |
0bc58a91 | 3822 | pos = next_thread(start); |
c1df7fb8 ON |
3823 | if (thread_group_leader(pos)) |
3824 | pos = NULL; | |
3825 | else | |
3826 | get_task_struct(pos); | |
3827 | } | |
cc288738 | 3828 | rcu_read_unlock(); |
0bc58a91 EB |
3829 | put_task_struct(start); |
3830 | return pos; | |
1da177e4 LT |
3831 | } |
3832 | ||
3833 | /* for the /proc/TGID/task/ directories */ | |
f0c3b509 | 3834 | static int proc_task_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3835 | { |
d855a4b7 ON |
3836 | struct inode *inode = file_inode(file); |
3837 | struct task_struct *task; | |
b488893a | 3838 | struct pid_namespace *ns; |
f0c3b509 | 3839 | int tid; |
1da177e4 | 3840 | |
d855a4b7 | 3841 | if (proc_inode_is_dead(inode)) |
f0c3b509 | 3842 | return -ENOENT; |
1da177e4 | 3843 | |
f0c3b509 | 3844 | if (!dir_emit_dots(file, ctx)) |
d855a4b7 | 3845 | return 0; |
1da177e4 | 3846 | |
0bc58a91 EB |
3847 | /* f_version caches the tgid value that the last readdir call couldn't |
3848 | * return. lseek aka telldir automagically resets f_version to 0. | |
3849 | */ | |
9d78edea | 3850 | ns = proc_pid_ns(inode->i_sb); |
f0c3b509 AV |
3851 | tid = (int)file->f_version; |
3852 | file->f_version = 0; | |
d855a4b7 | 3853 | for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns); |
0bc58a91 | 3854 | task; |
f0c3b509 | 3855 | task = next_tid(task), ctx->pos++) { |
e3912ac3 | 3856 | char name[10 + 1]; |
a4ef3895 | 3857 | unsigned int len; |
0658a096 | 3858 | |
b488893a | 3859 | tid = task_pid_nr_ns(task, ns); |
0658a096 FW |
3860 | if (!tid) |
3861 | continue; /* The task has just exited. */ | |
e3912ac3 | 3862 | len = snprintf(name, sizeof(name), "%u", tid); |
f0c3b509 AV |
3863 | if (!proc_fill_cache(file, ctx, name, len, |
3864 | proc_task_instantiate, task, NULL)) { | |
0bc58a91 EB |
3865 | /* returning this tgid failed, save it as the first |
3866 | * pid for the next readir call */ | |
f0c3b509 | 3867 | file->f_version = (u64)tid; |
0bc58a91 | 3868 | put_task_struct(task); |
1da177e4 | 3869 | break; |
0bc58a91 | 3870 | } |
1da177e4 | 3871 | } |
d855a4b7 | 3872 | |
f0c3b509 | 3873 | return 0; |
1da177e4 | 3874 | } |
6e66b52b | 3875 | |
549c7297 CB |
3876 | static int proc_task_getattr(struct user_namespace *mnt_userns, |
3877 | const struct path *path, struct kstat *stat, | |
a528d35e | 3878 | u32 request_mask, unsigned int query_flags) |
6e66b52b | 3879 | { |
a528d35e | 3880 | struct inode *inode = d_inode(path->dentry); |
99f89551 | 3881 | struct task_struct *p = get_proc_task(inode); |
0d56a451 | 3882 | generic_fillattr(&init_user_ns, inode, stat); |
6e66b52b | 3883 | |
99f89551 | 3884 | if (p) { |
99f89551 | 3885 | stat->nlink += get_nr_threads(p); |
99f89551 | 3886 | put_task_struct(p); |
6e66b52b EB |
3887 | } |
3888 | ||
3889 | return 0; | |
3890 | } | |
28a6d671 | 3891 | |
c5ef1c42 | 3892 | static const struct inode_operations proc_task_inode_operations = { |
28a6d671 EB |
3893 | .lookup = proc_task_lookup, |
3894 | .getattr = proc_task_getattr, | |
3895 | .setattr = proc_setattr, | |
0499680a | 3896 | .permission = proc_pid_permission, |
28a6d671 EB |
3897 | }; |
3898 | ||
00977a59 | 3899 | static const struct file_operations proc_task_operations = { |
28a6d671 | 3900 | .read = generic_read_dir, |
f50752ea AV |
3901 | .iterate_shared = proc_task_readdir, |
3902 | .llseek = generic_file_llseek, | |
28a6d671 | 3903 | }; |
1270dd8d AD |
3904 | |
3905 | void __init set_proc_pid_nlink(void) | |
3906 | { | |
3907 | nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
3908 | nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
3909 | } |