]>
Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 LT |
2 | /* |
3 | * linux/fs/proc/base.c | |
4 | * | |
5 | * Copyright (C) 1991, 1992 Linus Torvalds | |
6 | * | |
7 | * proc base directory handling functions | |
8 | * | |
9 | * 1999, Al Viro. Rewritten. Now it covers the whole per-process part. | |
10 | * Instead of using magical inumbers to determine the kind of object | |
11 | * we allocate and fill in-core inodes upon lookup. They don't even | |
12 | * go into icache. We cache the reference to task_struct upon lookup too. | |
13 | * Eventually it should become a filesystem in its own. We don't use the | |
14 | * rest of procfs anymore. | |
e070ad49 ML |
15 | * |
16 | * | |
17 | * Changelog: | |
18 | * 17-Jan-2005 | |
19 | * Allan Bezerra | |
20 | * Bruna Moreira <bruna.moreira@indt.org.br> | |
21 | * Edjard Mota <edjard.mota@indt.org.br> | |
22 | * Ilias Biris <ilias.biris@indt.org.br> | |
23 | * Mauricio Lin <mauricio.lin@indt.org.br> | |
24 | * | |
25 | * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT | |
26 | * | |
27 | * A new process specific entry (smaps) included in /proc. It shows the | |
28 | * size of rss for each memory area. The maps entry lacks information | |
29 | * about physical memory size (rss) for each mapped file, i.e., | |
30 | * rss information for executables and library files. | |
31 | * This additional information is useful for any tools that need to know | |
32 | * about physical memory consumption for a process specific library. | |
33 | * | |
34 | * Changelog: | |
35 | * 21-Feb-2005 | |
36 | * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT | |
37 | * Pud inclusion in the page table walking. | |
38 | * | |
39 | * ChangeLog: | |
40 | * 10-Mar-2005 | |
41 | * 10LE Instituto Nokia de Tecnologia - INdT: | |
42 | * A better way to walks through the page table as suggested by Hugh Dickins. | |
43 | * | |
44 | * Simo Piiroinen <simo.piiroinen@nokia.com>: | |
45 | * Smaps information related to shared, private, clean and dirty pages. | |
46 | * | |
47 | * Paul Mundt <paul.mundt@nokia.com>: | |
48 | * Overall revision about smaps. | |
1da177e4 LT |
49 | */ |
50 | ||
7c0f6ba6 | 51 | #include <linux/uaccess.h> |
1da177e4 | 52 | |
1da177e4 LT |
53 | #include <linux/errno.h> |
54 | #include <linux/time.h> | |
55 | #include <linux/proc_fs.h> | |
56 | #include <linux/stat.h> | |
5995477a | 57 | #include <linux/task_io_accounting_ops.h> |
1da177e4 | 58 | #include <linux/init.h> |
16f7e0fe | 59 | #include <linux/capability.h> |
1da177e4 | 60 | #include <linux/file.h> |
9f3acc31 | 61 | #include <linux/fdtable.h> |
94f8f3b0 | 62 | #include <linux/generic-radix-tree.h> |
1da177e4 LT |
63 | #include <linux/string.h> |
64 | #include <linux/seq_file.h> | |
65 | #include <linux/namei.h> | |
6b3286ed | 66 | #include <linux/mnt_namespace.h> |
1da177e4 | 67 | #include <linux/mm.h> |
a63d83f4 | 68 | #include <linux/swap.h> |
b835996f | 69 | #include <linux/rcupdate.h> |
1da177e4 | 70 | #include <linux/kallsyms.h> |
2ec220e2 | 71 | #include <linux/stacktrace.h> |
d85f50d5 | 72 | #include <linux/resource.h> |
5096add8 | 73 | #include <linux/module.h> |
1da177e4 LT |
74 | #include <linux/mount.h> |
75 | #include <linux/security.h> | |
76 | #include <linux/ptrace.h> | |
0d094efe | 77 | #include <linux/tracehook.h> |
87ebdc00 | 78 | #include <linux/printk.h> |
efb1a57d | 79 | #include <linux/cache.h> |
a424316c | 80 | #include <linux/cgroup.h> |
1da177e4 LT |
81 | #include <linux/cpuset.h> |
82 | #include <linux/audit.h> | |
5addc5dd | 83 | #include <linux/poll.h> |
1651e14e | 84 | #include <linux/nsproxy.h> |
8ac773b4 | 85 | #include <linux/oom.h> |
3cb4a0bb | 86 | #include <linux/elf.h> |
60347f67 | 87 | #include <linux/pid_namespace.h> |
22d917d8 | 88 | #include <linux/user_namespace.h> |
5ad4e53b | 89 | #include <linux/fs_struct.h> |
5a0e3ad6 | 90 | #include <linux/slab.h> |
4eb5aaa3 | 91 | #include <linux/sched/autogroup.h> |
6e84f315 | 92 | #include <linux/sched/mm.h> |
f7ccbae4 | 93 | #include <linux/sched/coredump.h> |
b17b0153 | 94 | #include <linux/sched/debug.h> |
3905f9ad | 95 | #include <linux/sched/stat.h> |
48f6a7a5 | 96 | #include <linux/posix-timers.h> |
04a8682a | 97 | #include <linux/time_namespace.h> |
e79f15a4 | 98 | #include <linux/resctrl.h> |
43d2b113 | 99 | #include <trace/events/oom.h> |
1da177e4 | 100 | #include "internal.h" |
faf60af1 | 101 | #include "fd.h" |
1da177e4 | 102 | |
ac7f1061 AD |
103 | #include "../../lib/kstrtox.h" |
104 | ||
0f2fe20f EB |
105 | /* NOTE: |
106 | * Implementing inode permission operations in /proc is almost | |
107 | * certainly an error. Permission checks need to happen during | |
108 | * each system call not at open time. The reason is that most of | |
109 | * what we wish to check for permissions in /proc varies at runtime. | |
110 | * | |
111 | * The classic example of a problem is opening file descriptors | |
112 | * in /proc for a task before it execs a suid executable. | |
113 | */ | |
114 | ||
efb1a57d AD |
115 | static u8 nlink_tid __ro_after_init; |
116 | static u8 nlink_tgid __ro_after_init; | |
1270dd8d | 117 | |
1da177e4 | 118 | struct pid_entry { |
cedbccab | 119 | const char *name; |
623f594e | 120 | unsigned int len; |
d161a13f | 121 | umode_t mode; |
c5ef1c42 | 122 | const struct inode_operations *iop; |
00977a59 | 123 | const struct file_operations *fop; |
20cdc894 | 124 | union proc_op op; |
1da177e4 LT |
125 | }; |
126 | ||
61a28784 | 127 | #define NOD(NAME, MODE, IOP, FOP, OP) { \ |
20cdc894 | 128 | .name = (NAME), \ |
c5141e6d | 129 | .len = sizeof(NAME) - 1, \ |
20cdc894 EB |
130 | .mode = MODE, \ |
131 | .iop = IOP, \ | |
132 | .fop = FOP, \ | |
133 | .op = OP, \ | |
134 | } | |
135 | ||
631f9c18 AD |
136 | #define DIR(NAME, MODE, iops, fops) \ |
137 | NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} ) | |
138 | #define LNK(NAME, get_link) \ | |
61a28784 | 139 | NOD(NAME, (S_IFLNK|S_IRWXUGO), \ |
20cdc894 | 140 | &proc_pid_link_inode_operations, NULL, \ |
631f9c18 AD |
141 | { .proc_get_link = get_link } ) |
142 | #define REG(NAME, MODE, fops) \ | |
143 | NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {}) | |
631f9c18 | 144 | #define ONE(NAME, MODE, show) \ |
6d9c939d | 145 | NOD(NAME, (S_IFREG|(MODE)), \ |
be614086 | 146 | NULL, &proc_single_file_operations, \ |
631f9c18 | 147 | { .proc_show = show } ) |
6d9c939d CS |
148 | #define ATTR(LSM, NAME, MODE) \ |
149 | NOD(NAME, (S_IFREG|(MODE)), \ | |
150 | NULL, &proc_pid_attr_operations, \ | |
151 | { .lsm = LSM }) | |
1da177e4 | 152 | |
aed54175 VN |
153 | /* |
154 | * Count the number of hardlinks for the pid_entry table, excluding the . | |
155 | * and .. links. | |
156 | */ | |
1270dd8d | 157 | static unsigned int __init pid_entry_nlink(const struct pid_entry *entries, |
aed54175 VN |
158 | unsigned int n) |
159 | { | |
160 | unsigned int i; | |
161 | unsigned int count; | |
162 | ||
1270dd8d | 163 | count = 2; |
aed54175 VN |
164 | for (i = 0; i < n; ++i) { |
165 | if (S_ISDIR(entries[i].mode)) | |
166 | ++count; | |
167 | } | |
168 | ||
169 | return count; | |
170 | } | |
171 | ||
f7ad3c6b | 172 | static int get_task_root(struct task_struct *task, struct path *root) |
1da177e4 | 173 | { |
7c2c7d99 HD |
174 | int result = -ENOENT; |
175 | ||
0494f6ec | 176 | task_lock(task); |
f7ad3c6b MS |
177 | if (task->fs) { |
178 | get_fs_root(task->fs, root); | |
7c2c7d99 HD |
179 | result = 0; |
180 | } | |
0494f6ec | 181 | task_unlock(task); |
7c2c7d99 | 182 | return result; |
0494f6ec MS |
183 | } |
184 | ||
7773fbc5 | 185 | static int proc_cwd_link(struct dentry *dentry, struct path *path) |
0494f6ec | 186 | { |
2b0143b5 | 187 | struct task_struct *task = get_proc_task(d_inode(dentry)); |
0494f6ec | 188 | int result = -ENOENT; |
99f89551 EB |
189 | |
190 | if (task) { | |
f7ad3c6b MS |
191 | task_lock(task); |
192 | if (task->fs) { | |
193 | get_fs_pwd(task->fs, path); | |
194 | result = 0; | |
195 | } | |
196 | task_unlock(task); | |
99f89551 EB |
197 | put_task_struct(task); |
198 | } | |
1da177e4 LT |
199 | return result; |
200 | } | |
201 | ||
7773fbc5 | 202 | static int proc_root_link(struct dentry *dentry, struct path *path) |
1da177e4 | 203 | { |
2b0143b5 | 204 | struct task_struct *task = get_proc_task(d_inode(dentry)); |
1da177e4 | 205 | int result = -ENOENT; |
99f89551 EB |
206 | |
207 | if (task) { | |
f7ad3c6b | 208 | result = get_task_root(task, path); |
99f89551 EB |
209 | put_task_struct(task); |
210 | } | |
1da177e4 LT |
211 | return result; |
212 | } | |
213 | ||
d26d0cd9 LT |
214 | /* |
215 | * If the user used setproctitle(), we just get the string from | |
216 | * user space at arg_start, and limit it to a maximum of one page. | |
217 | */ | |
218 | static ssize_t get_mm_proctitle(struct mm_struct *mm, char __user *buf, | |
219 | size_t count, unsigned long pos, | |
220 | unsigned long arg_start) | |
221 | { | |
222 | char *page; | |
223 | int ret, got; | |
224 | ||
225 | if (pos >= PAGE_SIZE) | |
226 | return 0; | |
227 | ||
228 | page = (char *)__get_free_page(GFP_KERNEL); | |
229 | if (!page) | |
230 | return -ENOMEM; | |
231 | ||
232 | ret = 0; | |
233 | got = access_remote_vm(mm, arg_start, page, PAGE_SIZE, FOLL_ANON); | |
234 | if (got > 0) { | |
235 | int len = strnlen(page, got); | |
236 | ||
237 | /* Include the NUL character if it was found */ | |
238 | if (len < got) | |
239 | len++; | |
240 | ||
241 | if (len > pos) { | |
242 | len -= pos; | |
243 | if (len > count) | |
244 | len = count; | |
245 | len -= copy_to_user(buf, page+pos, len); | |
246 | if (!len) | |
247 | len = -EFAULT; | |
248 | ret = len; | |
249 | } | |
250 | } | |
251 | free_page((unsigned long)page); | |
252 | return ret; | |
253 | } | |
254 | ||
e4b4e441 | 255 | static ssize_t get_mm_cmdline(struct mm_struct *mm, char __user *buf, |
5ab82718 | 256 | size_t count, loff_t *ppos) |
1da177e4 | 257 | { |
c2c0bb44 | 258 | unsigned long arg_start, arg_end, env_start, env_end; |
5ab82718 | 259 | unsigned long pos, len; |
d26d0cd9 | 260 | char *page, c; |
c2c0bb44 | 261 | |
c2c0bb44 | 262 | /* Check if process spawned far enough to have cmdline. */ |
e4b4e441 LT |
263 | if (!mm->env_end) |
264 | return 0; | |
c2c0bb44 | 265 | |
88aa7cc6 | 266 | spin_lock(&mm->arg_lock); |
c2c0bb44 AD |
267 | arg_start = mm->arg_start; |
268 | arg_end = mm->arg_end; | |
269 | env_start = mm->env_start; | |
270 | env_end = mm->env_end; | |
88aa7cc6 | 271 | spin_unlock(&mm->arg_lock); |
c2c0bb44 | 272 | |
5ab82718 LT |
273 | if (arg_start >= arg_end) |
274 | return 0; | |
6a6cbe75 | 275 | |
2ca66ff7 | 276 | /* |
d26d0cd9 LT |
277 | * We allow setproctitle() to overwrite the argument |
278 | * strings, and overflow past the original end. But | |
279 | * only when it overflows into the environment area. | |
2ca66ff7 | 280 | */ |
d26d0cd9 | 281 | if (env_start != arg_end || env_end < env_start) |
5ab82718 | 282 | env_start = env_end = arg_end; |
d26d0cd9 | 283 | len = env_end - arg_start; |
f5b65348 | 284 | |
5ab82718 | 285 | /* We're not going to care if "*ppos" has high bits set */ |
d26d0cd9 LT |
286 | pos = *ppos; |
287 | if (pos >= len) | |
5ab82718 | 288 | return 0; |
d26d0cd9 LT |
289 | if (count > len - pos) |
290 | count = len - pos; | |
291 | if (!count) | |
292 | return 0; | |
293 | ||
294 | /* | |
295 | * Magical special case: if the argv[] end byte is not | |
296 | * zero, the user has overwritten it with setproctitle(3). | |
297 | * | |
298 | * Possible future enhancement: do this only once when | |
299 | * pos is 0, and set a flag in the 'struct file'. | |
300 | */ | |
301 | if (access_remote_vm(mm, arg_end-1, &c, 1, FOLL_ANON) == 1 && c) | |
302 | return get_mm_proctitle(mm, buf, count, pos, arg_start); | |
a0a07b87 | 303 | |
d26d0cd9 LT |
304 | /* |
305 | * For the non-setproctitle() case we limit things strictly | |
306 | * to the [arg_start, arg_end[ range. | |
307 | */ | |
308 | pos += arg_start; | |
3d712546 | 309 | if (pos < arg_start || pos >= arg_end) |
5ab82718 | 310 | return 0; |
3d712546 LT |
311 | if (count > arg_end - pos) |
312 | count = arg_end - pos; | |
5ab82718 LT |
313 | |
314 | page = (char *)__get_free_page(GFP_KERNEL); | |
315 | if (!page) | |
316 | return -ENOMEM; | |
317 | ||
318 | len = 0; | |
319 | while (count) { | |
320 | int got; | |
321 | size_t size = min_t(size_t, PAGE_SIZE, count); | |
322 | ||
3d712546 LT |
323 | got = access_remote_vm(mm, pos, page, size, FOLL_ANON); |
324 | if (got <= 0) | |
5ab82718 | 325 | break; |
3d712546 | 326 | got -= copy_to_user(buf, page, got); |
5ab82718 LT |
327 | if (unlikely(!got)) { |
328 | if (!len) | |
329 | len = -EFAULT; | |
330 | break; | |
c2c0bb44 | 331 | } |
5ab82718 LT |
332 | pos += got; |
333 | buf += got; | |
334 | len += got; | |
335 | count -= got; | |
c2c0bb44 AD |
336 | } |
337 | ||
c2c0bb44 | 338 | free_page((unsigned long)page); |
5ab82718 | 339 | return len; |
1da177e4 LT |
340 | } |
341 | ||
e4b4e441 LT |
342 | static ssize_t get_task_cmdline(struct task_struct *tsk, char __user *buf, |
343 | size_t count, loff_t *pos) | |
344 | { | |
345 | struct mm_struct *mm; | |
346 | ssize_t ret; | |
347 | ||
348 | mm = get_task_mm(tsk); | |
349 | if (!mm) | |
350 | return 0; | |
351 | ||
352 | ret = get_mm_cmdline(mm, buf, count, pos); | |
c2c0bb44 | 353 | mmput(mm); |
e4b4e441 LT |
354 | return ret; |
355 | } | |
356 | ||
357 | static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf, | |
358 | size_t count, loff_t *pos) | |
359 | { | |
360 | struct task_struct *tsk; | |
361 | ssize_t ret; | |
362 | ||
363 | BUG_ON(*pos < 0); | |
364 | ||
365 | tsk = get_proc_task(file_inode(file)); | |
366 | if (!tsk) | |
367 | return -ESRCH; | |
368 | ret = get_task_cmdline(tsk, buf, count, pos); | |
369 | put_task_struct(tsk); | |
370 | if (ret > 0) | |
371 | *pos += ret; | |
372 | return ret; | |
1da177e4 LT |
373 | } |
374 | ||
c2c0bb44 AD |
375 | static const struct file_operations proc_pid_cmdline_ops = { |
376 | .read = proc_pid_cmdline_read, | |
377 | .llseek = generic_file_llseek, | |
378 | }; | |
379 | ||
1da177e4 LT |
380 | #ifdef CONFIG_KALLSYMS |
381 | /* | |
382 | * Provides a wchan file via kallsyms in a proper one-value-per-file format. | |
383 | * Returns the resolved symbol. If that fails, simply return the address. | |
384 | */ | |
edfcd606 AD |
385 | static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns, |
386 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 387 | { |
ffb45122 | 388 | unsigned long wchan; |
9281acea | 389 | char symname[KSYM_NAME_LEN]; |
1da177e4 | 390 | |
24b2ec21 AD |
391 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
392 | goto print0; | |
1da177e4 | 393 | |
24b2ec21 AD |
394 | wchan = get_wchan(task); |
395 | if (wchan && !lookup_symbol_name(wchan, symname)) { | |
21dae0ad | 396 | seq_puts(m, symname); |
24b2ec21 AD |
397 | return 0; |
398 | } | |
25ce3191 | 399 | |
24b2ec21 AD |
400 | print0: |
401 | seq_putc(m, '0'); | |
25ce3191 | 402 | return 0; |
1da177e4 LT |
403 | } |
404 | #endif /* CONFIG_KALLSYMS */ | |
405 | ||
a9712bc1 AV |
406 | static int lock_trace(struct task_struct *task) |
407 | { | |
2db9dbf7 | 408 | int err = mutex_lock_killable(&task->signal->exec_update_mutex); |
a9712bc1 AV |
409 | if (err) |
410 | return err; | |
caaee623 | 411 | if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) { |
2db9dbf7 | 412 | mutex_unlock(&task->signal->exec_update_mutex); |
a9712bc1 AV |
413 | return -EPERM; |
414 | } | |
415 | return 0; | |
416 | } | |
417 | ||
418 | static void unlock_trace(struct task_struct *task) | |
419 | { | |
2db9dbf7 | 420 | mutex_unlock(&task->signal->exec_update_mutex); |
a9712bc1 AV |
421 | } |
422 | ||
2ec220e2 KC |
423 | #ifdef CONFIG_STACKTRACE |
424 | ||
425 | #define MAX_STACK_TRACE_DEPTH 64 | |
426 | ||
427 | static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns, | |
428 | struct pid *pid, struct task_struct *task) | |
429 | { | |
2ec220e2 | 430 | unsigned long *entries; |
a9712bc1 | 431 | int err; |
2ec220e2 | 432 | |
f8a00cef JH |
433 | /* |
434 | * The ability to racily run the kernel stack unwinder on a running task | |
435 | * and then observe the unwinder output is scary; while it is useful for | |
436 | * debugging kernel issues, it can also allow an attacker to leak kernel | |
437 | * stack contents. | |
438 | * Doing this in a manner that is at least safe from races would require | |
439 | * some work to ensure that the remote task can not be scheduled; and | |
440 | * even then, this would still expose the unwinder as local attack | |
441 | * surface. | |
442 | * Therefore, this interface is restricted to root. | |
443 | */ | |
444 | if (!file_ns_capable(m->file, &init_user_ns, CAP_SYS_ADMIN)) | |
445 | return -EACCES; | |
446 | ||
6da2ec56 KC |
447 | entries = kmalloc_array(MAX_STACK_TRACE_DEPTH, sizeof(*entries), |
448 | GFP_KERNEL); | |
2ec220e2 KC |
449 | if (!entries) |
450 | return -ENOMEM; | |
451 | ||
a9712bc1 AV |
452 | err = lock_trace(task); |
453 | if (!err) { | |
e988e5ec | 454 | unsigned int i, nr_entries; |
5d008fb4 | 455 | |
e988e5ec TG |
456 | nr_entries = stack_trace_save_tsk(task, entries, |
457 | MAX_STACK_TRACE_DEPTH, 0); | |
a9712bc1 | 458 | |
e988e5ec | 459 | for (i = 0; i < nr_entries; i++) { |
8f5abe84 | 460 | seq_printf(m, "[<0>] %pB\n", (void *)entries[i]); |
a9712bc1 | 461 | } |
e988e5ec | 462 | |
a9712bc1 | 463 | unlock_trace(task); |
2ec220e2 KC |
464 | } |
465 | kfree(entries); | |
466 | ||
a9712bc1 | 467 | return err; |
2ec220e2 KC |
468 | } |
469 | #endif | |
470 | ||
5968cece | 471 | #ifdef CONFIG_SCHED_INFO |
1da177e4 LT |
472 | /* |
473 | * Provides /proc/PID/schedstat | |
474 | */ | |
f6e826ca AD |
475 | static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns, |
476 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 477 | { |
5968cece | 478 | if (unlikely(!sched_info_on())) |
08b55775 | 479 | seq_puts(m, "0 0 0\n"); |
5968cece NR |
480 | else |
481 | seq_printf(m, "%llu %llu %lu\n", | |
25ce3191 JP |
482 | (unsigned long long)task->se.sum_exec_runtime, |
483 | (unsigned long long)task->sched_info.run_delay, | |
484 | task->sched_info.pcount); | |
485 | ||
486 | return 0; | |
1da177e4 LT |
487 | } |
488 | #endif | |
489 | ||
9745512c AV |
490 | #ifdef CONFIG_LATENCYTOP |
491 | static int lstats_show_proc(struct seq_file *m, void *v) | |
492 | { | |
493 | int i; | |
13d77c37 HS |
494 | struct inode *inode = m->private; |
495 | struct task_struct *task = get_proc_task(inode); | |
9745512c | 496 | |
13d77c37 HS |
497 | if (!task) |
498 | return -ESRCH; | |
499 | seq_puts(m, "Latency Top version : v0.1\n"); | |
f6d2f584 | 500 | for (i = 0; i < LT_SAVECOUNT; i++) { |
34e49d4f JP |
501 | struct latency_record *lr = &task->latency_record[i]; |
502 | if (lr->backtrace[0]) { | |
9745512c | 503 | int q; |
34e49d4f JP |
504 | seq_printf(m, "%i %li %li", |
505 | lr->count, lr->time, lr->max); | |
9745512c | 506 | for (q = 0; q < LT_BACKTRACEDEPTH; q++) { |
34e49d4f | 507 | unsigned long bt = lr->backtrace[q]; |
accddc41 | 508 | |
34e49d4f | 509 | if (!bt) |
9745512c | 510 | break; |
34e49d4f | 511 | seq_printf(m, " %ps", (void *)bt); |
9745512c | 512 | } |
9d6de12f | 513 | seq_putc(m, '\n'); |
9745512c AV |
514 | } |
515 | ||
516 | } | |
13d77c37 | 517 | put_task_struct(task); |
9745512c AV |
518 | return 0; |
519 | } | |
520 | ||
521 | static int lstats_open(struct inode *inode, struct file *file) | |
522 | { | |
13d77c37 | 523 | return single_open(file, lstats_show_proc, inode); |
d6643d12 HS |
524 | } |
525 | ||
9745512c AV |
526 | static ssize_t lstats_write(struct file *file, const char __user *buf, |
527 | size_t count, loff_t *offs) | |
528 | { | |
496ad9aa | 529 | struct task_struct *task = get_proc_task(file_inode(file)); |
9745512c | 530 | |
13d77c37 HS |
531 | if (!task) |
532 | return -ESRCH; | |
e02c9b0d | 533 | clear_tsk_latency_tracing(task); |
13d77c37 | 534 | put_task_struct(task); |
9745512c AV |
535 | |
536 | return count; | |
537 | } | |
538 | ||
539 | static const struct file_operations proc_lstats_operations = { | |
540 | .open = lstats_open, | |
541 | .read = seq_read, | |
542 | .write = lstats_write, | |
543 | .llseek = seq_lseek, | |
13d77c37 | 544 | .release = single_release, |
9745512c AV |
545 | }; |
546 | ||
547 | #endif | |
548 | ||
6ba51e37 AD |
549 | static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns, |
550 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 551 | { |
ca79b0c2 | 552 | unsigned long totalpages = totalram_pages() + total_swap_pages; |
b95c35e7 | 553 | unsigned long points = 0; |
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 */ | |
778c1144 | 672 | static int proc_fd_access_allowed(struct inode *inode) |
1da177e4 | 673 | { |
778c1144 EB |
674 | struct task_struct *task; |
675 | int allowed = 0; | |
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 | ||
6b4e306a | 688 | int proc_setattr(struct dentry *dentry, struct iattr *attr) |
6d76fa58 LT |
689 | { |
690 | int error; | |
2b0143b5 | 691 | struct inode *inode = d_inode(dentry); |
6d76fa58 LT |
692 | |
693 | if (attr->ia_valid & ATTR_MODE) | |
694 | return -EPERM; | |
695 | ||
31051c85 | 696 | error = setattr_prepare(dentry, attr); |
1025774c CH |
697 | if (error) |
698 | return error; | |
699 | ||
1025774c CH |
700 | setattr_copy(inode, attr); |
701 | mark_inode_dirty(inode); | |
702 | return 0; | |
6d76fa58 LT |
703 | } |
704 | ||
0499680a VK |
705 | /* |
706 | * May current process learn task's sched/cmdline info (for hide_pid_min=1) | |
707 | * or euid/egid (for hide_pid_min=2)? | |
708 | */ | |
fa10fed3 | 709 | static bool has_pid_permissions(struct proc_fs_info *fs_info, |
0499680a | 710 | struct task_struct *task, |
e61bb8b3 | 711 | enum proc_hidepid hide_pid_min) |
0499680a | 712 | { |
24a71ce5 AG |
713 | /* |
714 | * If 'hidpid' mount option is set force a ptrace check, | |
715 | * we indicate that we are using a filesystem syscall | |
716 | * by passing PTRACE_MODE_READ_FSCREDS | |
717 | */ | |
718 | if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE) | |
719 | return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS); | |
720 | ||
fa10fed3 | 721 | if (fs_info->hide_pid < hide_pid_min) |
0499680a | 722 | return true; |
fa10fed3 | 723 | if (in_group_p(fs_info->pid_gid)) |
0499680a | 724 | return true; |
caaee623 | 725 | return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS); |
0499680a VK |
726 | } |
727 | ||
728 | ||
729 | static int proc_pid_permission(struct inode *inode, int mask) | |
730 | { | |
fa10fed3 | 731 | struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb); |
0499680a VK |
732 | struct task_struct *task; |
733 | bool has_perms; | |
734 | ||
735 | task = get_proc_task(inode); | |
a2ef990a XF |
736 | if (!task) |
737 | return -ESRCH; | |
fa10fed3 | 738 | has_perms = has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS); |
0499680a VK |
739 | put_task_struct(task); |
740 | ||
741 | if (!has_perms) { | |
fa10fed3 | 742 | if (fs_info->hide_pid == HIDEPID_INVISIBLE) { |
0499680a VK |
743 | /* |
744 | * Let's make getdents(), stat(), and open() | |
745 | * consistent with each other. If a process | |
746 | * may not stat() a file, it shouldn't be seen | |
747 | * in procfs at all. | |
748 | */ | |
749 | return -ENOENT; | |
750 | } | |
751 | ||
752 | return -EPERM; | |
753 | } | |
754 | return generic_permission(inode, mask); | |
755 | } | |
756 | ||
757 | ||
758 | ||
c5ef1c42 | 759 | static const struct inode_operations proc_def_inode_operations = { |
6d76fa58 LT |
760 | .setattr = proc_setattr, |
761 | }; | |
762 | ||
be614086 EB |
763 | static int proc_single_show(struct seq_file *m, void *v) |
764 | { | |
765 | struct inode *inode = m->private; | |
9d78edea | 766 | struct pid_namespace *ns = proc_pid_ns(inode->i_sb); |
76f668be | 767 | struct pid *pid = proc_pid(inode); |
be614086 EB |
768 | struct task_struct *task; |
769 | int ret; | |
770 | ||
be614086 EB |
771 | task = get_pid_task(pid, PIDTYPE_PID); |
772 | if (!task) | |
773 | return -ESRCH; | |
774 | ||
775 | ret = PROC_I(inode)->op.proc_show(m, ns, pid, task); | |
776 | ||
777 | put_task_struct(task); | |
778 | return ret; | |
779 | } | |
780 | ||
781 | static int proc_single_open(struct inode *inode, struct file *filp) | |
782 | { | |
c6a34058 | 783 | return single_open(filp, proc_single_show, inode); |
be614086 EB |
784 | } |
785 | ||
786 | static const struct file_operations proc_single_file_operations = { | |
787 | .open = proc_single_open, | |
788 | .read = seq_read, | |
789 | .llseek = seq_lseek, | |
790 | .release = single_release, | |
791 | }; | |
792 | ||
5381e169 ON |
793 | |
794 | struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode) | |
1da177e4 | 795 | { |
5381e169 ON |
796 | struct task_struct *task = get_proc_task(inode); |
797 | struct mm_struct *mm = ERR_PTR(-ESRCH); | |
e268337d | 798 | |
5381e169 | 799 | if (task) { |
caaee623 | 800 | mm = mm_access(task, mode | PTRACE_MODE_FSCREDS); |
5381e169 | 801 | put_task_struct(task); |
e268337d | 802 | |
5381e169 ON |
803 | if (!IS_ERR_OR_NULL(mm)) { |
804 | /* ensure this mm_struct can't be freed */ | |
f1f10076 | 805 | mmgrab(mm); |
5381e169 ON |
806 | /* but do not pin its memory */ |
807 | mmput(mm); | |
808 | } | |
809 | } | |
810 | ||
811 | return mm; | |
812 | } | |
813 | ||
814 | static int __mem_open(struct inode *inode, struct file *file, unsigned int mode) | |
815 | { | |
816 | struct mm_struct *mm = proc_mem_open(inode, mode); | |
e268337d LT |
817 | |
818 | if (IS_ERR(mm)) | |
819 | return PTR_ERR(mm); | |
820 | ||
e268337d | 821 | file->private_data = mm; |
1da177e4 LT |
822 | return 0; |
823 | } | |
824 | ||
b409e578 CW |
825 | static int mem_open(struct inode *inode, struct file *file) |
826 | { | |
bc452b4b DH |
827 | int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH); |
828 | ||
829 | /* OK to pass negative loff_t, we can catch out-of-range */ | |
830 | file->f_mode |= FMODE_UNSIGNED_OFFSET; | |
831 | ||
832 | return ret; | |
b409e578 CW |
833 | } |
834 | ||
572d34b9 ON |
835 | static ssize_t mem_rw(struct file *file, char __user *buf, |
836 | size_t count, loff_t *ppos, int write) | |
1da177e4 | 837 | { |
e268337d | 838 | struct mm_struct *mm = file->private_data; |
572d34b9 ON |
839 | unsigned long addr = *ppos; |
840 | ssize_t copied; | |
1da177e4 | 841 | char *page; |
272ddc8b | 842 | unsigned int flags; |
1da177e4 | 843 | |
e268337d LT |
844 | if (!mm) |
845 | return 0; | |
99f89551 | 846 | |
0ee931c4 | 847 | page = (char *)__get_free_page(GFP_KERNEL); |
30cd8903 | 848 | if (!page) |
e268337d | 849 | return -ENOMEM; |
1da177e4 | 850 | |
f7ca54f4 | 851 | copied = 0; |
388f7934 | 852 | if (!mmget_not_zero(mm)) |
6d08f2c7 ON |
853 | goto free; |
854 | ||
f511c0b1 | 855 | flags = FOLL_FORCE | (write ? FOLL_WRITE : 0); |
6347e8d5 | 856 | |
1da177e4 | 857 | while (count > 0) { |
572d34b9 | 858 | int this_len = min_t(int, count, PAGE_SIZE); |
1da177e4 | 859 | |
572d34b9 | 860 | if (write && copy_from_user(page, buf, this_len)) { |
1da177e4 LT |
861 | copied = -EFAULT; |
862 | break; | |
863 | } | |
572d34b9 | 864 | |
6347e8d5 | 865 | this_len = access_remote_vm(mm, addr, page, this_len, flags); |
572d34b9 | 866 | if (!this_len) { |
1da177e4 LT |
867 | if (!copied) |
868 | copied = -EIO; | |
869 | break; | |
870 | } | |
572d34b9 ON |
871 | |
872 | if (!write && copy_to_user(buf, page, this_len)) { | |
873 | copied = -EFAULT; | |
874 | break; | |
875 | } | |
876 | ||
877 | buf += this_len; | |
878 | addr += this_len; | |
879 | copied += this_len; | |
880 | count -= this_len; | |
1da177e4 | 881 | } |
572d34b9 | 882 | *ppos = addr; |
30cd8903 | 883 | |
6d08f2c7 ON |
884 | mmput(mm); |
885 | free: | |
30cd8903 | 886 | free_page((unsigned long) page); |
1da177e4 LT |
887 | return copied; |
888 | } | |
1da177e4 | 889 | |
572d34b9 ON |
890 | static ssize_t mem_read(struct file *file, char __user *buf, |
891 | size_t count, loff_t *ppos) | |
892 | { | |
893 | return mem_rw(file, buf, count, ppos, 0); | |
894 | } | |
895 | ||
896 | static ssize_t mem_write(struct file *file, const char __user *buf, | |
897 | size_t count, loff_t *ppos) | |
898 | { | |
899 | return mem_rw(file, (char __user*)buf, count, ppos, 1); | |
900 | } | |
901 | ||
85863e47 | 902 | loff_t mem_lseek(struct file *file, loff_t offset, int orig) |
1da177e4 LT |
903 | { |
904 | switch (orig) { | |
905 | case 0: | |
906 | file->f_pos = offset; | |
907 | break; | |
908 | case 1: | |
909 | file->f_pos += offset; | |
910 | break; | |
911 | default: | |
912 | return -EINVAL; | |
913 | } | |
914 | force_successful_syscall_return(); | |
915 | return file->f_pos; | |
916 | } | |
917 | ||
e268337d LT |
918 | static int mem_release(struct inode *inode, struct file *file) |
919 | { | |
920 | struct mm_struct *mm = file->private_data; | |
71879d3c | 921 | if (mm) |
6d08f2c7 | 922 | mmdrop(mm); |
e268337d LT |
923 | return 0; |
924 | } | |
925 | ||
00977a59 | 926 | static const struct file_operations proc_mem_operations = { |
1da177e4 LT |
927 | .llseek = mem_lseek, |
928 | .read = mem_read, | |
929 | .write = mem_write, | |
930 | .open = mem_open, | |
e268337d | 931 | .release = mem_release, |
1da177e4 LT |
932 | }; |
933 | ||
b409e578 CW |
934 | static int environ_open(struct inode *inode, struct file *file) |
935 | { | |
936 | return __mem_open(inode, file, PTRACE_MODE_READ); | |
937 | } | |
938 | ||
315e28c8 JP |
939 | static ssize_t environ_read(struct file *file, char __user *buf, |
940 | size_t count, loff_t *ppos) | |
941 | { | |
315e28c8 JP |
942 | char *page; |
943 | unsigned long src = *ppos; | |
b409e578 CW |
944 | int ret = 0; |
945 | struct mm_struct *mm = file->private_data; | |
a3b609ef | 946 | unsigned long env_start, env_end; |
315e28c8 | 947 | |
8148a73c MK |
948 | /* Ensure the process spawned far enough to have an environment. */ |
949 | if (!mm || !mm->env_end) | |
b409e578 | 950 | return 0; |
315e28c8 | 951 | |
0ee931c4 | 952 | page = (char *)__get_free_page(GFP_KERNEL); |
315e28c8 | 953 | if (!page) |
b409e578 | 954 | return -ENOMEM; |
315e28c8 | 955 | |
d6f64b89 | 956 | ret = 0; |
388f7934 | 957 | if (!mmget_not_zero(mm)) |
b409e578 | 958 | goto free; |
a3b609ef | 959 | |
88aa7cc6 | 960 | spin_lock(&mm->arg_lock); |
a3b609ef MG |
961 | env_start = mm->env_start; |
962 | env_end = mm->env_end; | |
88aa7cc6 | 963 | spin_unlock(&mm->arg_lock); |
a3b609ef | 964 | |
315e28c8 | 965 | while (count > 0) { |
e8905ec2 DH |
966 | size_t this_len, max_len; |
967 | int retval; | |
315e28c8 | 968 | |
a3b609ef | 969 | if (src >= (env_end - env_start)) |
315e28c8 JP |
970 | break; |
971 | ||
a3b609ef | 972 | this_len = env_end - (env_start + src); |
e8905ec2 DH |
973 | |
974 | max_len = min_t(size_t, PAGE_SIZE, count); | |
975 | this_len = min(max_len, this_len); | |
315e28c8 | 976 | |
7f7ccc2c | 977 | retval = access_remote_vm(mm, (env_start + src), page, this_len, FOLL_ANON); |
315e28c8 JP |
978 | |
979 | if (retval <= 0) { | |
980 | ret = retval; | |
981 | break; | |
982 | } | |
983 | ||
984 | if (copy_to_user(buf, page, retval)) { | |
985 | ret = -EFAULT; | |
986 | break; | |
987 | } | |
988 | ||
989 | ret += retval; | |
990 | src += retval; | |
991 | buf += retval; | |
992 | count -= retval; | |
993 | } | |
994 | *ppos = src; | |
315e28c8 | 995 | mmput(mm); |
b409e578 CW |
996 | |
997 | free: | |
315e28c8 | 998 | free_page((unsigned long) page); |
315e28c8 JP |
999 | return ret; |
1000 | } | |
1001 | ||
1002 | static const struct file_operations proc_environ_operations = { | |
b409e578 | 1003 | .open = environ_open, |
315e28c8 | 1004 | .read = environ_read, |
87df8424 | 1005 | .llseek = generic_file_llseek, |
b409e578 | 1006 | .release = mem_release, |
315e28c8 JP |
1007 | }; |
1008 | ||
c5317167 AV |
1009 | static int auxv_open(struct inode *inode, struct file *file) |
1010 | { | |
1011 | return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS); | |
1012 | } | |
1013 | ||
1014 | static ssize_t auxv_read(struct file *file, char __user *buf, | |
1015 | size_t count, loff_t *ppos) | |
1016 | { | |
1017 | struct mm_struct *mm = file->private_data; | |
1018 | unsigned int nwords = 0; | |
06b2849d LY |
1019 | |
1020 | if (!mm) | |
1021 | return 0; | |
c5317167 AV |
1022 | do { |
1023 | nwords += 2; | |
1024 | } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */ | |
1025 | return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv, | |
1026 | nwords * sizeof(mm->saved_auxv[0])); | |
1027 | } | |
1028 | ||
1029 | static const struct file_operations proc_auxv_operations = { | |
1030 | .open = auxv_open, | |
1031 | .read = auxv_read, | |
1032 | .llseek = generic_file_llseek, | |
1033 | .release = mem_release, | |
1034 | }; | |
1035 | ||
fa0cbbf1 DR |
1036 | static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count, |
1037 | loff_t *ppos) | |
1038 | { | |
496ad9aa | 1039 | struct task_struct *task = get_proc_task(file_inode(file)); |
fa0cbbf1 DR |
1040 | char buffer[PROC_NUMBUF]; |
1041 | int oom_adj = OOM_ADJUST_MIN; | |
1042 | size_t len; | |
fa0cbbf1 DR |
1043 | |
1044 | if (!task) | |
1045 | return -ESRCH; | |
f913da59 MH |
1046 | if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX) |
1047 | oom_adj = OOM_ADJUST_MAX; | |
1048 | else | |
1049 | oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) / | |
1050 | OOM_SCORE_ADJ_MAX; | |
fa0cbbf1 DR |
1051 | put_task_struct(task); |
1052 | len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj); | |
1053 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
1054 | } | |
1055 | ||
1d5f0acb MH |
1056 | static int __set_oom_adj(struct file *file, int oom_adj, bool legacy) |
1057 | { | |
1058 | static DEFINE_MUTEX(oom_adj_mutex); | |
44a70ade | 1059 | struct mm_struct *mm = NULL; |
1d5f0acb MH |
1060 | struct task_struct *task; |
1061 | int err = 0; | |
1062 | ||
1063 | task = get_proc_task(file_inode(file)); | |
1064 | if (!task) | |
1065 | return -ESRCH; | |
1066 | ||
1067 | mutex_lock(&oom_adj_mutex); | |
1068 | if (legacy) { | |
1069 | if (oom_adj < task->signal->oom_score_adj && | |
1070 | !capable(CAP_SYS_RESOURCE)) { | |
1071 | err = -EACCES; | |
1072 | goto err_unlock; | |
1073 | } | |
1074 | /* | |
1075 | * /proc/pid/oom_adj is provided for legacy purposes, ask users to use | |
1076 | * /proc/pid/oom_score_adj instead. | |
1077 | */ | |
1078 | pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n", | |
1079 | current->comm, task_pid_nr(current), task_pid_nr(task), | |
1080 | task_pid_nr(task)); | |
1081 | } else { | |
1082 | if ((short)oom_adj < task->signal->oom_score_adj_min && | |
1083 | !capable(CAP_SYS_RESOURCE)) { | |
1084 | err = -EACCES; | |
1085 | goto err_unlock; | |
1086 | } | |
1087 | } | |
1088 | ||
44a70ade MH |
1089 | /* |
1090 | * Make sure we will check other processes sharing the mm if this is | |
1091 | * not vfrok which wants its own oom_score_adj. | |
1092 | * pin the mm so it doesn't go away and get reused after task_unlock | |
1093 | */ | |
1094 | if (!task->vfork_done) { | |
1095 | struct task_struct *p = find_lock_task_mm(task); | |
1096 | ||
1097 | if (p) { | |
1098 | if (atomic_read(&p->mm->mm_users) > 1) { | |
1099 | mm = p->mm; | |
f1f10076 | 1100 | mmgrab(mm); |
44a70ade MH |
1101 | } |
1102 | task_unlock(p); | |
1103 | } | |
1104 | } | |
1105 | ||
1d5f0acb MH |
1106 | task->signal->oom_score_adj = oom_adj; |
1107 | if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE)) | |
1108 | task->signal->oom_score_adj_min = (short)oom_adj; | |
1109 | trace_oom_score_adj_update(task); | |
44a70ade MH |
1110 | |
1111 | if (mm) { | |
1112 | struct task_struct *p; | |
1113 | ||
1114 | rcu_read_lock(); | |
1115 | for_each_process(p) { | |
1116 | if (same_thread_group(task, p)) | |
1117 | continue; | |
1118 | ||
1119 | /* do not touch kernel threads or the global init */ | |
1120 | if (p->flags & PF_KTHREAD || is_global_init(p)) | |
1121 | continue; | |
1122 | ||
1123 | task_lock(p); | |
1124 | if (!p->vfork_done && process_shares_mm(p, mm)) { | |
44a70ade MH |
1125 | p->signal->oom_score_adj = oom_adj; |
1126 | if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE)) | |
1127 | p->signal->oom_score_adj_min = (short)oom_adj; | |
1128 | } | |
1129 | task_unlock(p); | |
1130 | } | |
1131 | rcu_read_unlock(); | |
1132 | mmdrop(mm); | |
1133 | } | |
1d5f0acb MH |
1134 | err_unlock: |
1135 | mutex_unlock(&oom_adj_mutex); | |
1136 | put_task_struct(task); | |
1137 | return err; | |
1138 | } | |
f913da59 | 1139 | |
b72bdfa7 DR |
1140 | /* |
1141 | * /proc/pid/oom_adj exists solely for backwards compatibility with previous | |
1142 | * kernels. The effective policy is defined by oom_score_adj, which has a | |
1143 | * different scale: oom_adj grew exponentially and oom_score_adj grows linearly. | |
1144 | * Values written to oom_adj are simply mapped linearly to oom_score_adj. | |
1145 | * Processes that become oom disabled via oom_adj will still be oom disabled | |
1146 | * with this implementation. | |
1147 | * | |
1148 | * oom_adj cannot be removed since existing userspace binaries use it. | |
1149 | */ | |
fa0cbbf1 DR |
1150 | static ssize_t oom_adj_write(struct file *file, const char __user *buf, |
1151 | size_t count, loff_t *ppos) | |
1152 | { | |
fa0cbbf1 DR |
1153 | char buffer[PROC_NUMBUF]; |
1154 | int oom_adj; | |
fa0cbbf1 DR |
1155 | int err; |
1156 | ||
1157 | memset(buffer, 0, sizeof(buffer)); | |
1158 | if (count > sizeof(buffer) - 1) | |
1159 | count = sizeof(buffer) - 1; | |
1160 | if (copy_from_user(buffer, buf, count)) { | |
1161 | err = -EFAULT; | |
1162 | goto out; | |
1163 | } | |
1164 | ||
1165 | err = kstrtoint(strstrip(buffer), 0, &oom_adj); | |
1166 | if (err) | |
1167 | goto out; | |
1168 | if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) && | |
1169 | oom_adj != OOM_DISABLE) { | |
1170 | err = -EINVAL; | |
1171 | goto out; | |
1172 | } | |
1173 | ||
fa0cbbf1 DR |
1174 | /* |
1175 | * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum | |
1176 | * value is always attainable. | |
1177 | */ | |
1178 | if (oom_adj == OOM_ADJUST_MAX) | |
1179 | oom_adj = OOM_SCORE_ADJ_MAX; | |
1180 | else | |
1181 | oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE; | |
1182 | ||
1d5f0acb | 1183 | err = __set_oom_adj(file, oom_adj, true); |
fa0cbbf1 DR |
1184 | out: |
1185 | return err < 0 ? err : count; | |
1186 | } | |
1187 | ||
1188 | static const struct file_operations proc_oom_adj_operations = { | |
1189 | .read = oom_adj_read, | |
1190 | .write = oom_adj_write, | |
1191 | .llseek = generic_file_llseek, | |
1192 | }; | |
1193 | ||
a63d83f4 DR |
1194 | static ssize_t oom_score_adj_read(struct file *file, char __user *buf, |
1195 | size_t count, loff_t *ppos) | |
1196 | { | |
496ad9aa | 1197 | struct task_struct *task = get_proc_task(file_inode(file)); |
a63d83f4 | 1198 | char buffer[PROC_NUMBUF]; |
a9c58b90 | 1199 | short oom_score_adj = OOM_SCORE_ADJ_MIN; |
a63d83f4 DR |
1200 | size_t len; |
1201 | ||
1202 | if (!task) | |
1203 | return -ESRCH; | |
f913da59 | 1204 | oom_score_adj = task->signal->oom_score_adj; |
a63d83f4 | 1205 | put_task_struct(task); |
a9c58b90 | 1206 | len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj); |
a63d83f4 DR |
1207 | return simple_read_from_buffer(buf, count, ppos, buffer, len); |
1208 | } | |
1209 | ||
1210 | static ssize_t oom_score_adj_write(struct file *file, const char __user *buf, | |
1211 | size_t count, loff_t *ppos) | |
1212 | { | |
a63d83f4 | 1213 | char buffer[PROC_NUMBUF]; |
0a8cb8e3 | 1214 | int oom_score_adj; |
a63d83f4 DR |
1215 | int err; |
1216 | ||
1217 | memset(buffer, 0, sizeof(buffer)); | |
1218 | if (count > sizeof(buffer) - 1) | |
1219 | count = sizeof(buffer) - 1; | |
723548bf DR |
1220 | if (copy_from_user(buffer, buf, count)) { |
1221 | err = -EFAULT; | |
1222 | goto out; | |
1223 | } | |
a63d83f4 | 1224 | |
0a8cb8e3 | 1225 | err = kstrtoint(strstrip(buffer), 0, &oom_score_adj); |
a63d83f4 | 1226 | if (err) |
723548bf | 1227 | goto out; |
a63d83f4 | 1228 | if (oom_score_adj < OOM_SCORE_ADJ_MIN || |
723548bf DR |
1229 | oom_score_adj > OOM_SCORE_ADJ_MAX) { |
1230 | err = -EINVAL; | |
1231 | goto out; | |
1232 | } | |
a63d83f4 | 1233 | |
1d5f0acb | 1234 | err = __set_oom_adj(file, oom_score_adj, false); |
723548bf DR |
1235 | out: |
1236 | return err < 0 ? err : count; | |
a63d83f4 DR |
1237 | } |
1238 | ||
1239 | static const struct file_operations proc_oom_score_adj_operations = { | |
1240 | .read = oom_score_adj_read, | |
1241 | .write = oom_score_adj_write, | |
6038f373 | 1242 | .llseek = default_llseek, |
a63d83f4 DR |
1243 | }; |
1244 | ||
4b7d248b | 1245 | #ifdef CONFIG_AUDIT |
b4eb4f7f | 1246 | #define TMPBUFLEN 11 |
1da177e4 LT |
1247 | static ssize_t proc_loginuid_read(struct file * file, char __user * buf, |
1248 | size_t count, loff_t *ppos) | |
1249 | { | |
496ad9aa | 1250 | struct inode * inode = file_inode(file); |
99f89551 | 1251 | struct task_struct *task = get_proc_task(inode); |
1da177e4 LT |
1252 | ssize_t length; |
1253 | char tmpbuf[TMPBUFLEN]; | |
1254 | ||
99f89551 EB |
1255 | if (!task) |
1256 | return -ESRCH; | |
1da177e4 | 1257 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", |
e1760bd5 EB |
1258 | from_kuid(file->f_cred->user_ns, |
1259 | audit_get_loginuid(task))); | |
99f89551 | 1260 | put_task_struct(task); |
1da177e4 LT |
1261 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); |
1262 | } | |
1263 | ||
1264 | static ssize_t proc_loginuid_write(struct file * file, const char __user * buf, | |
1265 | size_t count, loff_t *ppos) | |
1266 | { | |
496ad9aa | 1267 | struct inode * inode = file_inode(file); |
1da177e4 | 1268 | uid_t loginuid; |
e1760bd5 | 1269 | kuid_t kloginuid; |
774636e1 | 1270 | int rv; |
1da177e4 | 1271 | |
7dc52157 PM |
1272 | rcu_read_lock(); |
1273 | if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) { | |
1274 | rcu_read_unlock(); | |
1da177e4 | 1275 | return -EPERM; |
7dc52157 PM |
1276 | } |
1277 | rcu_read_unlock(); | |
1da177e4 | 1278 | |
1da177e4 LT |
1279 | if (*ppos != 0) { |
1280 | /* No partial writes. */ | |
1281 | return -EINVAL; | |
1282 | } | |
1da177e4 | 1283 | |
774636e1 AD |
1284 | rv = kstrtou32_from_user(buf, count, 10, &loginuid); |
1285 | if (rv < 0) | |
1286 | return rv; | |
81407c84 EP |
1287 | |
1288 | /* is userspace tring to explicitly UNSET the loginuid? */ | |
1289 | if (loginuid == AUDIT_UID_UNSET) { | |
1290 | kloginuid = INVALID_UID; | |
1291 | } else { | |
1292 | kloginuid = make_kuid(file->f_cred->user_ns, loginuid); | |
774636e1 AD |
1293 | if (!uid_valid(kloginuid)) |
1294 | return -EINVAL; | |
e1760bd5 EB |
1295 | } |
1296 | ||
774636e1 AD |
1297 | rv = audit_set_loginuid(kloginuid); |
1298 | if (rv < 0) | |
1299 | return rv; | |
1300 | return count; | |
1da177e4 LT |
1301 | } |
1302 | ||
00977a59 | 1303 | static const struct file_operations proc_loginuid_operations = { |
1da177e4 LT |
1304 | .read = proc_loginuid_read, |
1305 | .write = proc_loginuid_write, | |
87df8424 | 1306 | .llseek = generic_file_llseek, |
1da177e4 | 1307 | }; |
1e0bd755 EP |
1308 | |
1309 | static ssize_t proc_sessionid_read(struct file * file, char __user * buf, | |
1310 | size_t count, loff_t *ppos) | |
1311 | { | |
496ad9aa | 1312 | struct inode * inode = file_inode(file); |
1e0bd755 EP |
1313 | struct task_struct *task = get_proc_task(inode); |
1314 | ssize_t length; | |
1315 | char tmpbuf[TMPBUFLEN]; | |
1316 | ||
1317 | if (!task) | |
1318 | return -ESRCH; | |
1319 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", | |
1320 | audit_get_sessionid(task)); | |
1321 | put_task_struct(task); | |
1322 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); | |
1323 | } | |
1324 | ||
1325 | static const struct file_operations proc_sessionid_operations = { | |
1326 | .read = proc_sessionid_read, | |
87df8424 | 1327 | .llseek = generic_file_llseek, |
1e0bd755 | 1328 | }; |
1da177e4 LT |
1329 | #endif |
1330 | ||
f4f154fd AM |
1331 | #ifdef CONFIG_FAULT_INJECTION |
1332 | static ssize_t proc_fault_inject_read(struct file * file, char __user * buf, | |
1333 | size_t count, loff_t *ppos) | |
1334 | { | |
496ad9aa | 1335 | struct task_struct *task = get_proc_task(file_inode(file)); |
f4f154fd AM |
1336 | char buffer[PROC_NUMBUF]; |
1337 | size_t len; | |
1338 | int make_it_fail; | |
f4f154fd AM |
1339 | |
1340 | if (!task) | |
1341 | return -ESRCH; | |
1342 | make_it_fail = task->make_it_fail; | |
1343 | put_task_struct(task); | |
1344 | ||
1345 | len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail); | |
0c28f287 AM |
1346 | |
1347 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
f4f154fd AM |
1348 | } |
1349 | ||
1350 | static ssize_t proc_fault_inject_write(struct file * file, | |
1351 | const char __user * buf, size_t count, loff_t *ppos) | |
1352 | { | |
1353 | struct task_struct *task; | |
774636e1 | 1354 | char buffer[PROC_NUMBUF]; |
f4f154fd | 1355 | int make_it_fail; |
774636e1 | 1356 | int rv; |
f4f154fd AM |
1357 | |
1358 | if (!capable(CAP_SYS_RESOURCE)) | |
1359 | return -EPERM; | |
1360 | memset(buffer, 0, sizeof(buffer)); | |
1361 | if (count > sizeof(buffer) - 1) | |
1362 | count = sizeof(buffer) - 1; | |
1363 | if (copy_from_user(buffer, buf, count)) | |
1364 | return -EFAULT; | |
774636e1 AD |
1365 | rv = kstrtoint(strstrip(buffer), 0, &make_it_fail); |
1366 | if (rv < 0) | |
1367 | return rv; | |
16caed31 DJ |
1368 | if (make_it_fail < 0 || make_it_fail > 1) |
1369 | return -EINVAL; | |
1370 | ||
496ad9aa | 1371 | task = get_proc_task(file_inode(file)); |
f4f154fd AM |
1372 | if (!task) |
1373 | return -ESRCH; | |
1374 | task->make_it_fail = make_it_fail; | |
1375 | put_task_struct(task); | |
cba8aafe VL |
1376 | |
1377 | return count; | |
f4f154fd AM |
1378 | } |
1379 | ||
00977a59 | 1380 | static const struct file_operations proc_fault_inject_operations = { |
f4f154fd AM |
1381 | .read = proc_fault_inject_read, |
1382 | .write = proc_fault_inject_write, | |
87df8424 | 1383 | .llseek = generic_file_llseek, |
f4f154fd | 1384 | }; |
e41d5818 DV |
1385 | |
1386 | static ssize_t proc_fail_nth_write(struct file *file, const char __user *buf, | |
1387 | size_t count, loff_t *ppos) | |
1388 | { | |
1389 | struct task_struct *task; | |
9049f2f6 AM |
1390 | int err; |
1391 | unsigned int n; | |
e41d5818 | 1392 | |
1203c8e6 AM |
1393 | err = kstrtouint_from_user(buf, count, 0, &n); |
1394 | if (err) | |
1395 | return err; | |
1396 | ||
e41d5818 DV |
1397 | task = get_proc_task(file_inode(file)); |
1398 | if (!task) | |
1399 | return -ESRCH; | |
9f7118b2 | 1400 | task->fail_nth = n; |
e41d5818 | 1401 | put_task_struct(task); |
1203c8e6 | 1402 | |
e41d5818 DV |
1403 | return count; |
1404 | } | |
1405 | ||
1406 | static ssize_t proc_fail_nth_read(struct file *file, char __user *buf, | |
1407 | size_t count, loff_t *ppos) | |
1408 | { | |
1409 | struct task_struct *task; | |
bfc74093 AM |
1410 | char numbuf[PROC_NUMBUF]; |
1411 | ssize_t len; | |
e41d5818 DV |
1412 | |
1413 | task = get_proc_task(file_inode(file)); | |
1414 | if (!task) | |
1415 | return -ESRCH; | |
9f7118b2 | 1416 | len = snprintf(numbuf, sizeof(numbuf), "%u\n", task->fail_nth); |
1203c8e6 | 1417 | put_task_struct(task); |
a44937fe | 1418 | return simple_read_from_buffer(buf, count, ppos, numbuf, len); |
e41d5818 DV |
1419 | } |
1420 | ||
1421 | static const struct file_operations proc_fail_nth_operations = { | |
1422 | .read = proc_fail_nth_read, | |
1423 | .write = proc_fail_nth_write, | |
1424 | }; | |
f4f154fd AM |
1425 | #endif |
1426 | ||
9745512c | 1427 | |
43ae34cb IM |
1428 | #ifdef CONFIG_SCHED_DEBUG |
1429 | /* | |
1430 | * Print out various scheduling related per-task fields: | |
1431 | */ | |
1432 | static int sched_show(struct seq_file *m, void *v) | |
1433 | { | |
1434 | struct inode *inode = m->private; | |
9d78edea | 1435 | struct pid_namespace *ns = proc_pid_ns(inode->i_sb); |
43ae34cb IM |
1436 | struct task_struct *p; |
1437 | ||
43ae34cb IM |
1438 | p = get_proc_task(inode); |
1439 | if (!p) | |
1440 | return -ESRCH; | |
74dc3384 | 1441 | proc_sched_show_task(p, ns, m); |
43ae34cb IM |
1442 | |
1443 | put_task_struct(p); | |
1444 | ||
1445 | return 0; | |
1446 | } | |
1447 | ||
1448 | static ssize_t | |
1449 | sched_write(struct file *file, const char __user *buf, | |
1450 | size_t count, loff_t *offset) | |
1451 | { | |
496ad9aa | 1452 | struct inode *inode = file_inode(file); |
43ae34cb IM |
1453 | struct task_struct *p; |
1454 | ||
43ae34cb IM |
1455 | p = get_proc_task(inode); |
1456 | if (!p) | |
1457 | return -ESRCH; | |
1458 | proc_sched_set_task(p); | |
1459 | ||
1460 | put_task_struct(p); | |
1461 | ||
1462 | return count; | |
1463 | } | |
1464 | ||
1465 | static int sched_open(struct inode *inode, struct file *filp) | |
1466 | { | |
c6a34058 | 1467 | return single_open(filp, sched_show, inode); |
43ae34cb IM |
1468 | } |
1469 | ||
1470 | static const struct file_operations proc_pid_sched_operations = { | |
1471 | .open = sched_open, | |
1472 | .read = seq_read, | |
1473 | .write = sched_write, | |
1474 | .llseek = seq_lseek, | |
5ea473a1 | 1475 | .release = single_release, |
43ae34cb IM |
1476 | }; |
1477 | ||
1478 | #endif | |
1479 | ||
5091faa4 MG |
1480 | #ifdef CONFIG_SCHED_AUTOGROUP |
1481 | /* | |
1482 | * Print out autogroup related information: | |
1483 | */ | |
1484 | static int sched_autogroup_show(struct seq_file *m, void *v) | |
1485 | { | |
1486 | struct inode *inode = m->private; | |
1487 | struct task_struct *p; | |
1488 | ||
1489 | p = get_proc_task(inode); | |
1490 | if (!p) | |
1491 | return -ESRCH; | |
1492 | proc_sched_autogroup_show_task(p, m); | |
1493 | ||
1494 | put_task_struct(p); | |
1495 | ||
1496 | return 0; | |
1497 | } | |
1498 | ||
1499 | static ssize_t | |
1500 | sched_autogroup_write(struct file *file, const char __user *buf, | |
1501 | size_t count, loff_t *offset) | |
1502 | { | |
496ad9aa | 1503 | struct inode *inode = file_inode(file); |
5091faa4 MG |
1504 | struct task_struct *p; |
1505 | char buffer[PROC_NUMBUF]; | |
0a8cb8e3 | 1506 | int nice; |
5091faa4 MG |
1507 | int err; |
1508 | ||
1509 | memset(buffer, 0, sizeof(buffer)); | |
1510 | if (count > sizeof(buffer) - 1) | |
1511 | count = sizeof(buffer) - 1; | |
1512 | if (copy_from_user(buffer, buf, count)) | |
1513 | return -EFAULT; | |
1514 | ||
0a8cb8e3 AD |
1515 | err = kstrtoint(strstrip(buffer), 0, &nice); |
1516 | if (err < 0) | |
1517 | return err; | |
5091faa4 MG |
1518 | |
1519 | p = get_proc_task(inode); | |
1520 | if (!p) | |
1521 | return -ESRCH; | |
1522 | ||
2e5b5b3a | 1523 | err = proc_sched_autogroup_set_nice(p, nice); |
5091faa4 MG |
1524 | if (err) |
1525 | count = err; | |
1526 | ||
1527 | put_task_struct(p); | |
1528 | ||
1529 | return count; | |
1530 | } | |
1531 | ||
1532 | static int sched_autogroup_open(struct inode *inode, struct file *filp) | |
1533 | { | |
1534 | int ret; | |
1535 | ||
1536 | ret = single_open(filp, sched_autogroup_show, NULL); | |
1537 | if (!ret) { | |
1538 | struct seq_file *m = filp->private_data; | |
1539 | ||
1540 | m->private = inode; | |
1541 | } | |
1542 | return ret; | |
1543 | } | |
1544 | ||
1545 | static const struct file_operations proc_pid_sched_autogroup_operations = { | |
1546 | .open = sched_autogroup_open, | |
1547 | .read = seq_read, | |
1548 | .write = sched_autogroup_write, | |
1549 | .llseek = seq_lseek, | |
1550 | .release = single_release, | |
1551 | }; | |
1552 | ||
1553 | #endif /* CONFIG_SCHED_AUTOGROUP */ | |
1554 | ||
04a8682a AV |
1555 | #ifdef CONFIG_TIME_NS |
1556 | static int timens_offsets_show(struct seq_file *m, void *v) | |
1557 | { | |
1558 | struct task_struct *p; | |
1559 | ||
1560 | p = get_proc_task(file_inode(m->file)); | |
1561 | if (!p) | |
1562 | return -ESRCH; | |
1563 | proc_timens_show_offsets(p, m); | |
1564 | ||
1565 | put_task_struct(p); | |
1566 | ||
1567 | return 0; | |
1568 | } | |
1569 | ||
1570 | static ssize_t timens_offsets_write(struct file *file, const char __user *buf, | |
1571 | size_t count, loff_t *ppos) | |
1572 | { | |
1573 | struct inode *inode = file_inode(file); | |
1574 | struct proc_timens_offset offsets[2]; | |
1575 | char *kbuf = NULL, *pos, *next_line; | |
1576 | struct task_struct *p; | |
1577 | int ret, noffsets; | |
1578 | ||
1579 | /* Only allow < page size writes at the beginning of the file */ | |
1580 | if ((*ppos != 0) || (count >= PAGE_SIZE)) | |
1581 | return -EINVAL; | |
1582 | ||
1583 | /* Slurp in the user data */ | |
1584 | kbuf = memdup_user_nul(buf, count); | |
1585 | if (IS_ERR(kbuf)) | |
1586 | return PTR_ERR(kbuf); | |
1587 | ||
1588 | /* Parse the user data */ | |
1589 | ret = -EINVAL; | |
1590 | noffsets = 0; | |
1591 | for (pos = kbuf; pos; pos = next_line) { | |
1592 | struct proc_timens_offset *off = &offsets[noffsets]; | |
94d440d6 | 1593 | char clock[10]; |
04a8682a AV |
1594 | int err; |
1595 | ||
1596 | /* Find the end of line and ensure we don't look past it */ | |
1597 | next_line = strchr(pos, '\n'); | |
1598 | if (next_line) { | |
1599 | *next_line = '\0'; | |
1600 | next_line++; | |
1601 | if (*next_line == '\0') | |
1602 | next_line = NULL; | |
1603 | } | |
1604 | ||
94d440d6 | 1605 | err = sscanf(pos, "%9s %lld %lu", clock, |
04a8682a AV |
1606 | &off->val.tv_sec, &off->val.tv_nsec); |
1607 | if (err != 3 || off->val.tv_nsec >= NSEC_PER_SEC) | |
1608 | goto out; | |
94d440d6 AV |
1609 | |
1610 | clock[sizeof(clock) - 1] = 0; | |
1611 | if (strcmp(clock, "monotonic") == 0 || | |
1612 | strcmp(clock, __stringify(CLOCK_MONOTONIC)) == 0) | |
1613 | off->clockid = CLOCK_MONOTONIC; | |
1614 | else if (strcmp(clock, "boottime") == 0 || | |
1615 | strcmp(clock, __stringify(CLOCK_BOOTTIME)) == 0) | |
1616 | off->clockid = CLOCK_BOOTTIME; | |
1617 | else | |
1618 | goto out; | |
1619 | ||
04a8682a AV |
1620 | noffsets++; |
1621 | if (noffsets == ARRAY_SIZE(offsets)) { | |
1622 | if (next_line) | |
1623 | count = next_line - kbuf; | |
1624 | break; | |
1625 | } | |
1626 | } | |
1627 | ||
1628 | ret = -ESRCH; | |
1629 | p = get_proc_task(inode); | |
1630 | if (!p) | |
1631 | goto out; | |
1632 | ret = proc_timens_set_offset(file, p, offsets, noffsets); | |
1633 | put_task_struct(p); | |
1634 | if (ret) | |
1635 | goto out; | |
1636 | ||
1637 | ret = count; | |
1638 | out: | |
1639 | kfree(kbuf); | |
1640 | return ret; | |
1641 | } | |
1642 | ||
1643 | static int timens_offsets_open(struct inode *inode, struct file *filp) | |
1644 | { | |
1645 | return single_open(filp, timens_offsets_show, inode); | |
1646 | } | |
1647 | ||
1648 | static const struct file_operations proc_timens_offsets_operations = { | |
1649 | .open = timens_offsets_open, | |
1650 | .read = seq_read, | |
1651 | .write = timens_offsets_write, | |
1652 | .llseek = seq_lseek, | |
1653 | .release = single_release, | |
1654 | }; | |
1655 | #endif /* CONFIG_TIME_NS */ | |
1656 | ||
4614a696 JS |
1657 | static ssize_t comm_write(struct file *file, const char __user *buf, |
1658 | size_t count, loff_t *offset) | |
1659 | { | |
496ad9aa | 1660 | struct inode *inode = file_inode(file); |
4614a696 JS |
1661 | struct task_struct *p; |
1662 | char buffer[TASK_COMM_LEN]; | |
830e0fc9 | 1663 | const size_t maxlen = sizeof(buffer) - 1; |
4614a696 JS |
1664 | |
1665 | memset(buffer, 0, sizeof(buffer)); | |
830e0fc9 | 1666 | if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count)) |
4614a696 JS |
1667 | return -EFAULT; |
1668 | ||
1669 | p = get_proc_task(inode); | |
1670 | if (!p) | |
1671 | return -ESRCH; | |
1672 | ||
1673 | if (same_thread_group(current, p)) | |
1674 | set_task_comm(p, buffer); | |
1675 | else | |
1676 | count = -EINVAL; | |
1677 | ||
1678 | put_task_struct(p); | |
1679 | ||
1680 | return count; | |
1681 | } | |
1682 | ||
1683 | static int comm_show(struct seq_file *m, void *v) | |
1684 | { | |
1685 | struct inode *inode = m->private; | |
1686 | struct task_struct *p; | |
1687 | ||
1688 | p = get_proc_task(inode); | |
1689 | if (!p) | |
1690 | return -ESRCH; | |
1691 | ||
88b72b31 TH |
1692 | proc_task_name(m, p, false); |
1693 | seq_putc(m, '\n'); | |
4614a696 JS |
1694 | |
1695 | put_task_struct(p); | |
1696 | ||
1697 | return 0; | |
1698 | } | |
1699 | ||
1700 | static int comm_open(struct inode *inode, struct file *filp) | |
1701 | { | |
c6a34058 | 1702 | return single_open(filp, comm_show, inode); |
4614a696 JS |
1703 | } |
1704 | ||
1705 | static const struct file_operations proc_pid_set_comm_operations = { | |
1706 | .open = comm_open, | |
1707 | .read = seq_read, | |
1708 | .write = comm_write, | |
1709 | .llseek = seq_lseek, | |
1710 | .release = single_release, | |
1711 | }; | |
1712 | ||
7773fbc5 | 1713 | static int proc_exe_link(struct dentry *dentry, struct path *exe_path) |
925d1c40 MH |
1714 | { |
1715 | struct task_struct *task; | |
925d1c40 MH |
1716 | struct file *exe_file; |
1717 | ||
2b0143b5 | 1718 | task = get_proc_task(d_inode(dentry)); |
925d1c40 MH |
1719 | if (!task) |
1720 | return -ENOENT; | |
cd81a917 | 1721 | exe_file = get_task_exe_file(task); |
925d1c40 | 1722 | put_task_struct(task); |
925d1c40 MH |
1723 | if (exe_file) { |
1724 | *exe_path = exe_file->f_path; | |
1725 | path_get(&exe_file->f_path); | |
1726 | fput(exe_file); | |
1727 | return 0; | |
1728 | } else | |
1729 | return -ENOENT; | |
1730 | } | |
1731 | ||
6b255391 | 1732 | static const char *proc_pid_get_link(struct dentry *dentry, |
fceef393 AV |
1733 | struct inode *inode, |
1734 | struct delayed_call *done) | |
1da177e4 | 1735 | { |
408ef013 | 1736 | struct path path; |
1da177e4 LT |
1737 | int error = -EACCES; |
1738 | ||
6b255391 AV |
1739 | if (!dentry) |
1740 | return ERR_PTR(-ECHILD); | |
1741 | ||
778c1144 EB |
1742 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1743 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1744 | goto out; |
1da177e4 | 1745 | |
408ef013 CH |
1746 | error = PROC_I(inode)->op.proc_get_link(dentry, &path); |
1747 | if (error) | |
1748 | goto out; | |
1749 | ||
1bc82070 | 1750 | error = nd_jump_link(&path); |
1da177e4 | 1751 | out: |
008b150a | 1752 | return ERR_PTR(error); |
1da177e4 LT |
1753 | } |
1754 | ||
3dcd25f3 | 1755 | static int do_proc_readlink(struct path *path, char __user *buffer, int buflen) |
1da177e4 | 1756 | { |
0ee931c4 | 1757 | char *tmp = (char *)__get_free_page(GFP_KERNEL); |
3dcd25f3 | 1758 | char *pathname; |
1da177e4 LT |
1759 | int len; |
1760 | ||
1761 | if (!tmp) | |
1762 | return -ENOMEM; | |
0c28f287 | 1763 | |
7b2a69ba | 1764 | pathname = d_path(path, tmp, PAGE_SIZE); |
3dcd25f3 JB |
1765 | len = PTR_ERR(pathname); |
1766 | if (IS_ERR(pathname)) | |
1da177e4 | 1767 | goto out; |
3dcd25f3 | 1768 | len = tmp + PAGE_SIZE - 1 - pathname; |
1da177e4 LT |
1769 | |
1770 | if (len > buflen) | |
1771 | len = buflen; | |
3dcd25f3 | 1772 | if (copy_to_user(buffer, pathname, len)) |
1da177e4 LT |
1773 | len = -EFAULT; |
1774 | out: | |
1775 | free_page((unsigned long)tmp); | |
1776 | return len; | |
1777 | } | |
1778 | ||
1779 | static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen) | |
1780 | { | |
1781 | int error = -EACCES; | |
2b0143b5 | 1782 | struct inode *inode = d_inode(dentry); |
3dcd25f3 | 1783 | struct path path; |
1da177e4 | 1784 | |
778c1144 EB |
1785 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1786 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1787 | goto out; |
1da177e4 | 1788 | |
7773fbc5 | 1789 | error = PROC_I(inode)->op.proc_get_link(dentry, &path); |
1da177e4 LT |
1790 | if (error) |
1791 | goto out; | |
1792 | ||
3dcd25f3 JB |
1793 | error = do_proc_readlink(&path, buffer, buflen); |
1794 | path_put(&path); | |
1da177e4 | 1795 | out: |
1da177e4 LT |
1796 | return error; |
1797 | } | |
1798 | ||
faf60af1 | 1799 | const struct inode_operations proc_pid_link_inode_operations = { |
1da177e4 | 1800 | .readlink = proc_pid_readlink, |
6b255391 | 1801 | .get_link = proc_pid_get_link, |
6d76fa58 | 1802 | .setattr = proc_setattr, |
1da177e4 LT |
1803 | }; |
1804 | ||
28a6d671 EB |
1805 | |
1806 | /* building an inode */ | |
1807 | ||
c6eb50d2 | 1808 | void task_dump_owner(struct task_struct *task, umode_t mode, |
68eb94f1 EB |
1809 | kuid_t *ruid, kgid_t *rgid) |
1810 | { | |
1811 | /* Depending on the state of dumpable compute who should own a | |
1812 | * proc file for a task. | |
1813 | */ | |
1814 | const struct cred *cred; | |
1815 | kuid_t uid; | |
1816 | kgid_t gid; | |
1817 | ||
2e0ad552 AD |
1818 | if (unlikely(task->flags & PF_KTHREAD)) { |
1819 | *ruid = GLOBAL_ROOT_UID; | |
1820 | *rgid = GLOBAL_ROOT_GID; | |
1821 | return; | |
1822 | } | |
1823 | ||
68eb94f1 EB |
1824 | /* Default to the tasks effective ownership */ |
1825 | rcu_read_lock(); | |
1826 | cred = __task_cred(task); | |
1827 | uid = cred->euid; | |
1828 | gid = cred->egid; | |
1829 | rcu_read_unlock(); | |
1830 | ||
1831 | /* | |
1832 | * Before the /proc/pid/status file was created the only way to read | |
1833 | * the effective uid of a /process was to stat /proc/pid. Reading | |
1834 | * /proc/pid/status is slow enough that procps and other packages | |
1835 | * kept stating /proc/pid. To keep the rules in /proc simple I have | |
1836 | * made this apply to all per process world readable and executable | |
1837 | * directories. | |
1838 | */ | |
1839 | if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) { | |
1840 | struct mm_struct *mm; | |
1841 | task_lock(task); | |
1842 | mm = task->mm; | |
1843 | /* Make non-dumpable tasks owned by some root */ | |
1844 | if (mm) { | |
1845 | if (get_dumpable(mm) != SUID_DUMP_USER) { | |
1846 | struct user_namespace *user_ns = mm->user_ns; | |
1847 | ||
1848 | uid = make_kuid(user_ns, 0); | |
1849 | if (!uid_valid(uid)) | |
1850 | uid = GLOBAL_ROOT_UID; | |
1851 | ||
1852 | gid = make_kgid(user_ns, 0); | |
1853 | if (!gid_valid(gid)) | |
1854 | gid = GLOBAL_ROOT_GID; | |
1855 | } | |
1856 | } else { | |
1857 | uid = GLOBAL_ROOT_UID; | |
1858 | gid = GLOBAL_ROOT_GID; | |
1859 | } | |
1860 | task_unlock(task); | |
1861 | } | |
1862 | *ruid = uid; | |
1863 | *rgid = gid; | |
1864 | } | |
1865 | ||
7bc3e6e5 EB |
1866 | void proc_pid_evict_inode(struct proc_inode *ei) |
1867 | { | |
1868 | struct pid *pid = ei->pid; | |
1869 | ||
1870 | if (S_ISDIR(ei->vfs_inode.i_mode)) { | |
63f818f4 | 1871 | spin_lock(&pid->lock); |
7bc3e6e5 | 1872 | hlist_del_init_rcu(&ei->sibling_inodes); |
63f818f4 | 1873 | spin_unlock(&pid->lock); |
7bc3e6e5 EB |
1874 | } |
1875 | ||
1876 | put_pid(pid); | |
1877 | } | |
1878 | ||
db978da8 AG |
1879 | struct inode *proc_pid_make_inode(struct super_block * sb, |
1880 | struct task_struct *task, umode_t mode) | |
28a6d671 EB |
1881 | { |
1882 | struct inode * inode; | |
1883 | struct proc_inode *ei; | |
7bc3e6e5 | 1884 | struct pid *pid; |
1da177e4 | 1885 | |
28a6d671 | 1886 | /* We need a new inode */ |
1da177e4 | 1887 | |
28a6d671 EB |
1888 | inode = new_inode(sb); |
1889 | if (!inode) | |
1890 | goto out; | |
1891 | ||
1892 | /* Common stuff */ | |
1893 | ei = PROC_I(inode); | |
db978da8 | 1894 | inode->i_mode = mode; |
85fe4025 | 1895 | inode->i_ino = get_next_ino(); |
078cd827 | 1896 | inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode); |
28a6d671 EB |
1897 | inode->i_op = &proc_def_inode_operations; |
1898 | ||
1899 | /* | |
1900 | * grab the reference to task. | |
1901 | */ | |
7bc3e6e5 EB |
1902 | pid = get_task_pid(task, PIDTYPE_PID); |
1903 | if (!pid) | |
28a6d671 EB |
1904 | goto out_unlock; |
1905 | ||
7bc3e6e5 EB |
1906 | /* Let the pid remember us for quick removal */ |
1907 | ei->pid = pid; | |
1908 | if (S_ISDIR(mode)) { | |
63f818f4 | 1909 | spin_lock(&pid->lock); |
7bc3e6e5 | 1910 | hlist_add_head_rcu(&ei->sibling_inodes, &pid->inodes); |
63f818f4 | 1911 | spin_unlock(&pid->lock); |
7bc3e6e5 EB |
1912 | } |
1913 | ||
68eb94f1 | 1914 | task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid); |
28a6d671 EB |
1915 | security_task_to_inode(task, inode); |
1916 | ||
1da177e4 | 1917 | out: |
28a6d671 EB |
1918 | return inode; |
1919 | ||
1920 | out_unlock: | |
1921 | iput(inode); | |
1922 | return NULL; | |
1da177e4 LT |
1923 | } |
1924 | ||
a528d35e DH |
1925 | int pid_getattr(const struct path *path, struct kstat *stat, |
1926 | u32 request_mask, unsigned int query_flags) | |
1da177e4 | 1927 | { |
a528d35e | 1928 | struct inode *inode = d_inode(path->dentry); |
fa10fed3 | 1929 | struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb); |
28a6d671 | 1930 | struct task_struct *task; |
c69e8d9c | 1931 | |
28a6d671 | 1932 | generic_fillattr(inode, stat); |
1da177e4 | 1933 | |
dcb0f222 EB |
1934 | stat->uid = GLOBAL_ROOT_UID; |
1935 | stat->gid = GLOBAL_ROOT_GID; | |
94116929 | 1936 | rcu_read_lock(); |
28a6d671 EB |
1937 | task = pid_task(proc_pid(inode), PIDTYPE_PID); |
1938 | if (task) { | |
fa10fed3 | 1939 | if (!has_pid_permissions(fs_info, task, HIDEPID_INVISIBLE)) { |
0499680a VK |
1940 | rcu_read_unlock(); |
1941 | /* | |
1942 | * This doesn't prevent learning whether PID exists, | |
1943 | * it only makes getattr() consistent with readdir(). | |
1944 | */ | |
1945 | return -ENOENT; | |
1946 | } | |
68eb94f1 | 1947 | task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid); |
1da177e4 | 1948 | } |
28a6d671 | 1949 | rcu_read_unlock(); |
d6e71144 | 1950 | return 0; |
1da177e4 LT |
1951 | } |
1952 | ||
1da177e4 LT |
1953 | /* dentry stuff */ |
1954 | ||
1955 | /* | |
1bbc5513 AV |
1956 | * Set <pid>/... inode ownership (can change due to setuid(), etc.) |
1957 | */ | |
1958 | void pid_update_inode(struct task_struct *task, struct inode *inode) | |
1959 | { | |
1960 | task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid); | |
1961 | ||
1962 | inode->i_mode &= ~(S_ISUID | S_ISGID); | |
1963 | security_task_to_inode(task, inode); | |
1964 | } | |
1965 | ||
1966 | /* | |
1da177e4 LT |
1967 | * Rewrite the inode's ownerships here because the owning task may have |
1968 | * performed a setuid(), etc. | |
99f89551 | 1969 | * |
1da177e4 | 1970 | */ |
1bbc5513 | 1971 | static int pid_revalidate(struct dentry *dentry, unsigned int flags) |
1da177e4 | 1972 | { |
34286d66 NP |
1973 | struct inode *inode; |
1974 | struct task_struct *task; | |
c69e8d9c | 1975 | |
0b728e19 | 1976 | if (flags & LOOKUP_RCU) |
34286d66 NP |
1977 | return -ECHILD; |
1978 | ||
2b0143b5 | 1979 | inode = d_inode(dentry); |
34286d66 NP |
1980 | task = get_proc_task(inode); |
1981 | ||
99f89551 | 1982 | if (task) { |
1bbc5513 | 1983 | pid_update_inode(task, inode); |
99f89551 | 1984 | put_task_struct(task); |
1da177e4 LT |
1985 | return 1; |
1986 | } | |
1da177e4 LT |
1987 | return 0; |
1988 | } | |
1989 | ||
d855a4b7 ON |
1990 | static inline bool proc_inode_is_dead(struct inode *inode) |
1991 | { | |
1992 | return !proc_pid(inode)->tasks[PIDTYPE_PID].first; | |
1993 | } | |
1994 | ||
1dd704b6 DH |
1995 | int pid_delete_dentry(const struct dentry *dentry) |
1996 | { | |
1997 | /* Is the task we represent dead? | |
1998 | * If so, then don't put the dentry on the lru list, | |
1999 | * kill it immediately. | |
2000 | */ | |
2b0143b5 | 2001 | return proc_inode_is_dead(d_inode(dentry)); |
1dd704b6 DH |
2002 | } |
2003 | ||
6b4e306a | 2004 | const struct dentry_operations pid_dentry_operations = |
28a6d671 EB |
2005 | { |
2006 | .d_revalidate = pid_revalidate, | |
2007 | .d_delete = pid_delete_dentry, | |
2008 | }; | |
2009 | ||
2010 | /* Lookups */ | |
2011 | ||
1c0d04c9 EB |
2012 | /* |
2013 | * Fill a directory entry. | |
2014 | * | |
2015 | * If possible create the dcache entry and derive our inode number and | |
2016 | * file type from dcache entry. | |
2017 | * | |
2018 | * Since all of the proc inode numbers are dynamically generated, the inode | |
2019 | * numbers do not exist until the inode is cache. This means creating the | |
2020 | * the dcache entry in readdir is necessary to keep the inode numbers | |
2021 | * reported by readdir in sync with the inode numbers reported | |
2022 | * by stat. | |
2023 | */ | |
f0c3b509 | 2024 | bool proc_fill_cache(struct file *file, struct dir_context *ctx, |
a4ef3895 | 2025 | const char *name, unsigned int len, |
c5141e6d | 2026 | instantiate_t instantiate, struct task_struct *task, const void *ptr) |
61a28784 | 2027 | { |
f0c3b509 | 2028 | struct dentry *child, *dir = file->f_path.dentry; |
1df98b8b | 2029 | struct qstr qname = QSTR_INIT(name, len); |
61a28784 | 2030 | struct inode *inode; |
0168b9e3 AV |
2031 | unsigned type = DT_UNKNOWN; |
2032 | ino_t ino = 1; | |
61a28784 | 2033 | |
1df98b8b | 2034 | child = d_hash_and_lookup(dir, &qname); |
61a28784 | 2035 | if (!child) { |
3781764b AV |
2036 | DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq); |
2037 | child = d_alloc_parallel(dir, &qname, &wq); | |
2038 | if (IS_ERR(child)) | |
1df98b8b | 2039 | goto end_instantiate; |
3781764b | 2040 | if (d_in_lookup(child)) { |
0168b9e3 AV |
2041 | struct dentry *res; |
2042 | res = instantiate(child, task, ptr); | |
3781764b | 2043 | d_lookup_done(child); |
0168b9e3 AV |
2044 | if (unlikely(res)) { |
2045 | dput(child); | |
2046 | child = res; | |
d85b399b AV |
2047 | if (IS_ERR(child)) |
2048 | goto end_instantiate; | |
3781764b | 2049 | } |
61a28784 EB |
2050 | } |
2051 | } | |
2b0143b5 | 2052 | inode = d_inode(child); |
147ce699 AV |
2053 | ino = inode->i_ino; |
2054 | type = inode->i_mode >> 12; | |
61a28784 | 2055 | dput(child); |
d85b399b | 2056 | end_instantiate: |
f0c3b509 | 2057 | return dir_emit(ctx, name, len, ino, type); |
61a28784 EB |
2058 | } |
2059 | ||
640708a2 PE |
2060 | /* |
2061 | * dname_to_vma_addr - maps a dentry name into two unsigned longs | |
2062 | * which represent vma start and end addresses. | |
2063 | */ | |
2064 | static int dname_to_vma_addr(struct dentry *dentry, | |
2065 | unsigned long *start, unsigned long *end) | |
2066 | { | |
ac7f1061 AD |
2067 | const char *str = dentry->d_name.name; |
2068 | unsigned long long sval, eval; | |
2069 | unsigned int len; | |
2070 | ||
35318db5 AD |
2071 | if (str[0] == '0' && str[1] != '-') |
2072 | return -EINVAL; | |
ac7f1061 AD |
2073 | len = _parse_integer(str, 16, &sval); |
2074 | if (len & KSTRTOX_OVERFLOW) | |
2075 | return -EINVAL; | |
2076 | if (sval != (unsigned long)sval) | |
2077 | return -EINVAL; | |
2078 | str += len; | |
2079 | ||
2080 | if (*str != '-') | |
640708a2 | 2081 | return -EINVAL; |
ac7f1061 AD |
2082 | str++; |
2083 | ||
35318db5 AD |
2084 | if (str[0] == '0' && str[1]) |
2085 | return -EINVAL; | |
ac7f1061 AD |
2086 | len = _parse_integer(str, 16, &eval); |
2087 | if (len & KSTRTOX_OVERFLOW) | |
2088 | return -EINVAL; | |
2089 | if (eval != (unsigned long)eval) | |
2090 | return -EINVAL; | |
2091 | str += len; | |
2092 | ||
2093 | if (*str != '\0') | |
2094 | return -EINVAL; | |
2095 | ||
2096 | *start = sval; | |
2097 | *end = eval; | |
640708a2 PE |
2098 | |
2099 | return 0; | |
2100 | } | |
2101 | ||
0b728e19 | 2102 | static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags) |
640708a2 PE |
2103 | { |
2104 | unsigned long vm_start, vm_end; | |
2105 | bool exact_vma_exists = false; | |
2106 | struct mm_struct *mm = NULL; | |
2107 | struct task_struct *task; | |
640708a2 PE |
2108 | struct inode *inode; |
2109 | int status = 0; | |
2110 | ||
0b728e19 | 2111 | if (flags & LOOKUP_RCU) |
640708a2 PE |
2112 | return -ECHILD; |
2113 | ||
2b0143b5 | 2114 | inode = d_inode(dentry); |
640708a2 PE |
2115 | task = get_proc_task(inode); |
2116 | if (!task) | |
2117 | goto out_notask; | |
2118 | ||
caaee623 | 2119 | mm = mm_access(task, PTRACE_MODE_READ_FSCREDS); |
2344bec7 | 2120 | if (IS_ERR_OR_NULL(mm)) |
640708a2 PE |
2121 | goto out; |
2122 | ||
2123 | if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) { | |
d8ed45c5 | 2124 | status = mmap_read_lock_killable(mm); |
cd9e2bb8 KK |
2125 | if (!status) { |
2126 | exact_vma_exists = !!find_exact_vma(mm, vm_start, | |
2127 | vm_end); | |
d8ed45c5 | 2128 | mmap_read_unlock(mm); |
cd9e2bb8 | 2129 | } |
640708a2 PE |
2130 | } |
2131 | ||
2132 | mmput(mm); | |
2133 | ||
2134 | if (exact_vma_exists) { | |
68eb94f1 EB |
2135 | task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid); |
2136 | ||
640708a2 PE |
2137 | security_task_to_inode(task, inode); |
2138 | status = 1; | |
2139 | } | |
2140 | ||
2141 | out: | |
2142 | put_task_struct(task); | |
2143 | ||
2144 | out_notask: | |
640708a2 PE |
2145 | return status; |
2146 | } | |
2147 | ||
2148 | static const struct dentry_operations tid_map_files_dentry_operations = { | |
2149 | .d_revalidate = map_files_d_revalidate, | |
2150 | .d_delete = pid_delete_dentry, | |
2151 | }; | |
2152 | ||
6b255391 | 2153 | static int map_files_get_link(struct dentry *dentry, struct path *path) |
640708a2 PE |
2154 | { |
2155 | unsigned long vm_start, vm_end; | |
2156 | struct vm_area_struct *vma; | |
2157 | struct task_struct *task; | |
2158 | struct mm_struct *mm; | |
2159 | int rc; | |
2160 | ||
2161 | rc = -ENOENT; | |
2b0143b5 | 2162 | task = get_proc_task(d_inode(dentry)); |
640708a2 PE |
2163 | if (!task) |
2164 | goto out; | |
2165 | ||
2166 | mm = get_task_mm(task); | |
2167 | put_task_struct(task); | |
2168 | if (!mm) | |
2169 | goto out; | |
2170 | ||
2171 | rc = dname_to_vma_addr(dentry, &vm_start, &vm_end); | |
2172 | if (rc) | |
2173 | goto out_mmput; | |
2174 | ||
d8ed45c5 | 2175 | rc = mmap_read_lock_killable(mm); |
cd9e2bb8 KK |
2176 | if (rc) |
2177 | goto out_mmput; | |
2178 | ||
70335abb | 2179 | rc = -ENOENT; |
640708a2 PE |
2180 | vma = find_exact_vma(mm, vm_start, vm_end); |
2181 | if (vma && vma->vm_file) { | |
2182 | *path = vma->vm_file->f_path; | |
2183 | path_get(path); | |
2184 | rc = 0; | |
2185 | } | |
d8ed45c5 | 2186 | mmap_read_unlock(mm); |
640708a2 PE |
2187 | |
2188 | out_mmput: | |
2189 | mmput(mm); | |
2190 | out: | |
2191 | return rc; | |
2192 | } | |
2193 | ||
2194 | struct map_files_info { | |
20d28cde AD |
2195 | unsigned long start; |
2196 | unsigned long end; | |
7b540d06 | 2197 | fmode_t mode; |
640708a2 PE |
2198 | }; |
2199 | ||
bdb4d100 | 2200 | /* |
12886f8a AR |
2201 | * Only allow CAP_SYS_ADMIN and CAP_CHECKPOINT_RESTORE to follow the links, due |
2202 | * to concerns about how the symlinks may be used to bypass permissions on | |
2203 | * ancestor directories in the path to the file in question. | |
bdb4d100 CO |
2204 | */ |
2205 | static const char * | |
6b255391 | 2206 | proc_map_files_get_link(struct dentry *dentry, |
fceef393 AV |
2207 | struct inode *inode, |
2208 | struct delayed_call *done) | |
bdb4d100 | 2209 | { |
12886f8a | 2210 | if (!checkpoint_restore_ns_capable(&init_user_ns)) |
bdb4d100 CO |
2211 | return ERR_PTR(-EPERM); |
2212 | ||
fceef393 | 2213 | return proc_pid_get_link(dentry, inode, done); |
bdb4d100 CO |
2214 | } |
2215 | ||
2216 | /* | |
6b255391 | 2217 | * Identical to proc_pid_link_inode_operations except for get_link() |
bdb4d100 CO |
2218 | */ |
2219 | static const struct inode_operations proc_map_files_link_inode_operations = { | |
2220 | .readlink = proc_pid_readlink, | |
6b255391 | 2221 | .get_link = proc_map_files_get_link, |
bdb4d100 CO |
2222 | .setattr = proc_setattr, |
2223 | }; | |
2224 | ||
0168b9e3 AV |
2225 | static struct dentry * |
2226 | proc_map_files_instantiate(struct dentry *dentry, | |
640708a2 PE |
2227 | struct task_struct *task, const void *ptr) |
2228 | { | |
7b540d06 | 2229 | fmode_t mode = (fmode_t)(unsigned long)ptr; |
640708a2 PE |
2230 | struct proc_inode *ei; |
2231 | struct inode *inode; | |
2232 | ||
0168b9e3 | 2233 | inode = proc_pid_make_inode(dentry->d_sb, task, S_IFLNK | |
db978da8 AG |
2234 | ((mode & FMODE_READ ) ? S_IRUSR : 0) | |
2235 | ((mode & FMODE_WRITE) ? S_IWUSR : 0)); | |
640708a2 | 2236 | if (!inode) |
0168b9e3 | 2237 | return ERR_PTR(-ENOENT); |
640708a2 PE |
2238 | |
2239 | ei = PROC_I(inode); | |
6b255391 | 2240 | ei->op.proc_get_link = map_files_get_link; |
640708a2 | 2241 | |
bdb4d100 | 2242 | inode->i_op = &proc_map_files_link_inode_operations; |
640708a2 | 2243 | inode->i_size = 64; |
640708a2 PE |
2244 | |
2245 | d_set_d_op(dentry, &tid_map_files_dentry_operations); | |
0168b9e3 | 2246 | return d_splice_alias(inode, dentry); |
640708a2 PE |
2247 | } |
2248 | ||
2249 | static struct dentry *proc_map_files_lookup(struct inode *dir, | |
00cd8dd3 | 2250 | struct dentry *dentry, unsigned int flags) |
640708a2 PE |
2251 | { |
2252 | unsigned long vm_start, vm_end; | |
2253 | struct vm_area_struct *vma; | |
2254 | struct task_struct *task; | |
0168b9e3 | 2255 | struct dentry *result; |
640708a2 PE |
2256 | struct mm_struct *mm; |
2257 | ||
0168b9e3 | 2258 | result = ERR_PTR(-ENOENT); |
640708a2 PE |
2259 | task = get_proc_task(dir); |
2260 | if (!task) | |
2261 | goto out; | |
2262 | ||
0168b9e3 | 2263 | result = ERR_PTR(-EACCES); |
caaee623 | 2264 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
640708a2 PE |
2265 | goto out_put_task; |
2266 | ||
0168b9e3 | 2267 | result = ERR_PTR(-ENOENT); |
640708a2 | 2268 | if (dname_to_vma_addr(dentry, &vm_start, &vm_end)) |
eb94cd96 | 2269 | goto out_put_task; |
640708a2 PE |
2270 | |
2271 | mm = get_task_mm(task); | |
2272 | if (!mm) | |
eb94cd96 | 2273 | goto out_put_task; |
640708a2 | 2274 | |
cd9e2bb8 | 2275 | result = ERR_PTR(-EINTR); |
d8ed45c5 | 2276 | if (mmap_read_lock_killable(mm)) |
cd9e2bb8 KK |
2277 | goto out_put_mm; |
2278 | ||
2279 | result = ERR_PTR(-ENOENT); | |
640708a2 PE |
2280 | vma = find_exact_vma(mm, vm_start, vm_end); |
2281 | if (!vma) | |
2282 | goto out_no_vma; | |
2283 | ||
05f56484 | 2284 | if (vma->vm_file) |
0168b9e3 | 2285 | result = proc_map_files_instantiate(dentry, task, |
05f56484 | 2286 | (void *)(unsigned long)vma->vm_file->f_mode); |
640708a2 PE |
2287 | |
2288 | out_no_vma: | |
d8ed45c5 | 2289 | mmap_read_unlock(mm); |
cd9e2bb8 | 2290 | out_put_mm: |
640708a2 | 2291 | mmput(mm); |
640708a2 PE |
2292 | out_put_task: |
2293 | put_task_struct(task); | |
2294 | out: | |
0168b9e3 | 2295 | return result; |
640708a2 PE |
2296 | } |
2297 | ||
2298 | static const struct inode_operations proc_map_files_inode_operations = { | |
2299 | .lookup = proc_map_files_lookup, | |
2300 | .permission = proc_fd_permission, | |
2301 | .setattr = proc_setattr, | |
2302 | }; | |
2303 | ||
2304 | static int | |
f0c3b509 | 2305 | proc_map_files_readdir(struct file *file, struct dir_context *ctx) |
640708a2 | 2306 | { |
640708a2 PE |
2307 | struct vm_area_struct *vma; |
2308 | struct task_struct *task; | |
2309 | struct mm_struct *mm; | |
f0c3b509 | 2310 | unsigned long nr_files, pos, i; |
94f8f3b0 | 2311 | GENRADIX(struct map_files_info) fa; |
f0c3b509 | 2312 | struct map_files_info *p; |
640708a2 PE |
2313 | int ret; |
2314 | ||
94f8f3b0 KO |
2315 | genradix_init(&fa); |
2316 | ||
640708a2 | 2317 | ret = -ENOENT; |
f0c3b509 | 2318 | task = get_proc_task(file_inode(file)); |
640708a2 PE |
2319 | if (!task) |
2320 | goto out; | |
2321 | ||
2322 | ret = -EACCES; | |
caaee623 | 2323 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
640708a2 PE |
2324 | goto out_put_task; |
2325 | ||
2326 | ret = 0; | |
f0c3b509 AV |
2327 | if (!dir_emit_dots(file, ctx)) |
2328 | goto out_put_task; | |
640708a2 | 2329 | |
f0c3b509 AV |
2330 | mm = get_task_mm(task); |
2331 | if (!mm) | |
2332 | goto out_put_task; | |
cd9e2bb8 | 2333 | |
89154dd5 | 2334 | ret = mmap_read_lock_killable(mm); |
cd9e2bb8 KK |
2335 | if (ret) { |
2336 | mmput(mm); | |
2337 | goto out_put_task; | |
2338 | } | |
640708a2 | 2339 | |
f0c3b509 | 2340 | nr_files = 0; |
640708a2 | 2341 | |
f0c3b509 AV |
2342 | /* |
2343 | * We need two passes here: | |
2344 | * | |
c1e8d7c6 ML |
2345 | * 1) Collect vmas of mapped files with mmap_lock taken |
2346 | * 2) Release mmap_lock and instantiate entries | |
f0c3b509 AV |
2347 | * |
2348 | * otherwise we get lockdep complained, since filldir() | |
c1e8d7c6 | 2349 | * routine might require mmap_lock taken in might_fault(). |
f0c3b509 | 2350 | */ |
640708a2 | 2351 | |
f0c3b509 | 2352 | for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) { |
94f8f3b0 KO |
2353 | if (!vma->vm_file) |
2354 | continue; | |
2355 | if (++pos <= ctx->pos) | |
2356 | continue; | |
f0c3b509 | 2357 | |
94f8f3b0 KO |
2358 | p = genradix_ptr_alloc(&fa, nr_files++, GFP_KERNEL); |
2359 | if (!p) { | |
f0c3b509 | 2360 | ret = -ENOMEM; |
89154dd5 | 2361 | mmap_read_unlock(mm); |
f0c3b509 AV |
2362 | mmput(mm); |
2363 | goto out_put_task; | |
640708a2 | 2364 | } |
f0c3b509 | 2365 | |
94f8f3b0 KO |
2366 | p->start = vma->vm_start; |
2367 | p->end = vma->vm_end; | |
2368 | p->mode = vma->vm_file->f_mode; | |
640708a2 | 2369 | } |
89154dd5 | 2370 | mmap_read_unlock(mm); |
fe079a5e | 2371 | mmput(mm); |
f0c3b509 AV |
2372 | |
2373 | for (i = 0; i < nr_files; i++) { | |
20d28cde AD |
2374 | char buf[4 * sizeof(long) + 2]; /* max: %lx-%lx\0 */ |
2375 | unsigned int len; | |
2376 | ||
94f8f3b0 | 2377 | p = genradix_ptr(&fa, i); |
20d28cde | 2378 | len = snprintf(buf, sizeof(buf), "%lx-%lx", p->start, p->end); |
f0c3b509 | 2379 | if (!proc_fill_cache(file, ctx, |
20d28cde | 2380 | buf, len, |
f0c3b509 AV |
2381 | proc_map_files_instantiate, |
2382 | task, | |
2383 | (void *)(unsigned long)p->mode)) | |
2384 | break; | |
2385 | ctx->pos++; | |
640708a2 PE |
2386 | } |
2387 | ||
640708a2 PE |
2388 | out_put_task: |
2389 | put_task_struct(task); | |
2390 | out: | |
94f8f3b0 | 2391 | genradix_free(&fa); |
640708a2 PE |
2392 | return ret; |
2393 | } | |
2394 | ||
2395 | static const struct file_operations proc_map_files_operations = { | |
2396 | .read = generic_read_dir, | |
f50752ea AV |
2397 | .iterate_shared = proc_map_files_readdir, |
2398 | .llseek = generic_file_llseek, | |
640708a2 PE |
2399 | }; |
2400 | ||
b18b6a9c | 2401 | #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS) |
48f6a7a5 PE |
2402 | struct timers_private { |
2403 | struct pid *pid; | |
2404 | struct task_struct *task; | |
2405 | struct sighand_struct *sighand; | |
57b8015e | 2406 | struct pid_namespace *ns; |
48f6a7a5 PE |
2407 | unsigned long flags; |
2408 | }; | |
2409 | ||
2410 | static void *timers_start(struct seq_file *m, loff_t *pos) | |
2411 | { | |
2412 | struct timers_private *tp = m->private; | |
2413 | ||
2414 | tp->task = get_pid_task(tp->pid, PIDTYPE_PID); | |
2415 | if (!tp->task) | |
2416 | return ERR_PTR(-ESRCH); | |
2417 | ||
2418 | tp->sighand = lock_task_sighand(tp->task, &tp->flags); | |
2419 | if (!tp->sighand) | |
2420 | return ERR_PTR(-ESRCH); | |
2421 | ||
2422 | return seq_list_start(&tp->task->signal->posix_timers, *pos); | |
2423 | } | |
2424 | ||
2425 | static void *timers_next(struct seq_file *m, void *v, loff_t *pos) | |
2426 | { | |
2427 | struct timers_private *tp = m->private; | |
2428 | return seq_list_next(v, &tp->task->signal->posix_timers, pos); | |
2429 | } | |
2430 | ||
2431 | static void timers_stop(struct seq_file *m, void *v) | |
2432 | { | |
2433 | struct timers_private *tp = m->private; | |
2434 | ||
2435 | if (tp->sighand) { | |
2436 | unlock_task_sighand(tp->task, &tp->flags); | |
2437 | tp->sighand = NULL; | |
2438 | } | |
2439 | ||
2440 | if (tp->task) { | |
2441 | put_task_struct(tp->task); | |
2442 | tp->task = NULL; | |
2443 | } | |
2444 | } | |
2445 | ||
2446 | static int show_timer(struct seq_file *m, void *v) | |
2447 | { | |
2448 | struct k_itimer *timer; | |
57b8015e PE |
2449 | struct timers_private *tp = m->private; |
2450 | int notify; | |
cedbccab | 2451 | static const char * const nstr[] = { |
57b8015e PE |
2452 | [SIGEV_SIGNAL] = "signal", |
2453 | [SIGEV_NONE] = "none", | |
2454 | [SIGEV_THREAD] = "thread", | |
2455 | }; | |
48f6a7a5 PE |
2456 | |
2457 | timer = list_entry((struct list_head *)v, struct k_itimer, list); | |
57b8015e PE |
2458 | notify = timer->it_sigev_notify; |
2459 | ||
48f6a7a5 | 2460 | seq_printf(m, "ID: %d\n", timer->it_id); |
ba3edf1f | 2461 | seq_printf(m, "signal: %d/%px\n", |
25ce3191 JP |
2462 | timer->sigq->info.si_signo, |
2463 | timer->sigq->info.si_value.sival_ptr); | |
57b8015e | 2464 | seq_printf(m, "notify: %s/%s.%d\n", |
25ce3191 JP |
2465 | nstr[notify & ~SIGEV_THREAD_ID], |
2466 | (notify & SIGEV_THREAD_ID) ? "tid" : "pid", | |
2467 | pid_nr_ns(timer->it_pid, tp->ns)); | |
15ef0298 | 2468 | seq_printf(m, "ClockID: %d\n", timer->it_clock); |
48f6a7a5 PE |
2469 | |
2470 | return 0; | |
2471 | } | |
2472 | ||
2473 | static const struct seq_operations proc_timers_seq_ops = { | |
2474 | .start = timers_start, | |
2475 | .next = timers_next, | |
2476 | .stop = timers_stop, | |
2477 | .show = show_timer, | |
2478 | }; | |
2479 | ||
2480 | static int proc_timers_open(struct inode *inode, struct file *file) | |
2481 | { | |
2482 | struct timers_private *tp; | |
2483 | ||
2484 | tp = __seq_open_private(file, &proc_timers_seq_ops, | |
2485 | sizeof(struct timers_private)); | |
2486 | if (!tp) | |
2487 | return -ENOMEM; | |
2488 | ||
2489 | tp->pid = proc_pid(inode); | |
9d78edea | 2490 | tp->ns = proc_pid_ns(inode->i_sb); |
48f6a7a5 PE |
2491 | return 0; |
2492 | } | |
2493 | ||
2494 | static const struct file_operations proc_timers_operations = { | |
2495 | .open = proc_timers_open, | |
2496 | .read = seq_read, | |
2497 | .llseek = seq_lseek, | |
2498 | .release = seq_release_private, | |
2499 | }; | |
b5946bea | 2500 | #endif |
640708a2 | 2501 | |
5de23d43 JS |
2502 | static ssize_t timerslack_ns_write(struct file *file, const char __user *buf, |
2503 | size_t count, loff_t *offset) | |
2504 | { | |
2505 | struct inode *inode = file_inode(file); | |
2506 | struct task_struct *p; | |
2507 | u64 slack_ns; | |
2508 | int err; | |
2509 | ||
2510 | err = kstrtoull_from_user(buf, count, 10, &slack_ns); | |
2511 | if (err < 0) | |
2512 | return err; | |
2513 | ||
2514 | p = get_proc_task(inode); | |
2515 | if (!p) | |
2516 | return -ESRCH; | |
2517 | ||
4b2bd5fe | 2518 | if (p != current) { |
8da0b4f6 BG |
2519 | rcu_read_lock(); |
2520 | if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) { | |
2521 | rcu_read_unlock(); | |
4b2bd5fe JS |
2522 | count = -EPERM; |
2523 | goto out; | |
2524 | } | |
8da0b4f6 | 2525 | rcu_read_unlock(); |
5de23d43 | 2526 | |
4b2bd5fe JS |
2527 | err = security_task_setscheduler(p); |
2528 | if (err) { | |
2529 | count = err; | |
2530 | goto out; | |
2531 | } | |
904763e1 JS |
2532 | } |
2533 | ||
7abbaf94 JS |
2534 | task_lock(p); |
2535 | if (slack_ns == 0) | |
2536 | p->timer_slack_ns = p->default_timer_slack_ns; | |
2537 | else | |
2538 | p->timer_slack_ns = slack_ns; | |
2539 | task_unlock(p); | |
2540 | ||
2541 | out: | |
5de23d43 JS |
2542 | put_task_struct(p); |
2543 | ||
2544 | return count; | |
2545 | } | |
2546 | ||
2547 | static int timerslack_ns_show(struct seq_file *m, void *v) | |
2548 | { | |
2549 | struct inode *inode = m->private; | |
2550 | struct task_struct *p; | |
7abbaf94 | 2551 | int err = 0; |
5de23d43 JS |
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 | err = -EPERM; |
2562 | goto out; | |
2563 | } | |
8da0b4f6 BG |
2564 | rcu_read_unlock(); |
2565 | ||
4b2bd5fe JS |
2566 | err = security_task_getscheduler(p); |
2567 | if (err) | |
2568 | goto out; | |
2569 | } | |
904763e1 | 2570 | |
7abbaf94 JS |
2571 | task_lock(p); |
2572 | seq_printf(m, "%llu\n", p->timer_slack_ns); | |
2573 | task_unlock(p); | |
2574 | ||
2575 | out: | |
5de23d43 JS |
2576 | put_task_struct(p); |
2577 | ||
2578 | return err; | |
2579 | } | |
2580 | ||
2581 | static int timerslack_ns_open(struct inode *inode, struct file *filp) | |
2582 | { | |
2583 | return single_open(filp, timerslack_ns_show, inode); | |
2584 | } | |
2585 | ||
2586 | static const struct file_operations proc_pid_set_timerslack_ns_operations = { | |
2587 | .open = timerslack_ns_open, | |
2588 | .read = seq_read, | |
2589 | .write = timerslack_ns_write, | |
2590 | .llseek = seq_lseek, | |
2591 | .release = single_release, | |
2592 | }; | |
2593 | ||
0168b9e3 AV |
2594 | static struct dentry *proc_pident_instantiate(struct dentry *dentry, |
2595 | struct task_struct *task, const void *ptr) | |
444ceed8 | 2596 | { |
c5141e6d | 2597 | const struct pid_entry *p = ptr; |
444ceed8 EB |
2598 | struct inode *inode; |
2599 | struct proc_inode *ei; | |
444ceed8 | 2600 | |
0168b9e3 | 2601 | inode = proc_pid_make_inode(dentry->d_sb, task, p->mode); |
444ceed8 | 2602 | if (!inode) |
0168b9e3 | 2603 | return ERR_PTR(-ENOENT); |
444ceed8 EB |
2604 | |
2605 | ei = PROC_I(inode); | |
444ceed8 | 2606 | if (S_ISDIR(inode->i_mode)) |
bfe86848 | 2607 | set_nlink(inode, 2); /* Use getattr to fix if necessary */ |
444ceed8 EB |
2608 | if (p->iop) |
2609 | inode->i_op = p->iop; | |
2610 | if (p->fop) | |
2611 | inode->i_fop = p->fop; | |
2612 | ei->op = p->op; | |
1bbc5513 | 2613 | pid_update_inode(task, inode); |
fb045adb | 2614 | d_set_d_op(dentry, &pid_dentry_operations); |
0168b9e3 | 2615 | return d_splice_alias(inode, dentry); |
444ceed8 EB |
2616 | } |
2617 | ||
1da177e4 LT |
2618 | static struct dentry *proc_pident_lookup(struct inode *dir, |
2619 | struct dentry *dentry, | |
d5a572a4 AD |
2620 | const struct pid_entry *p, |
2621 | const struct pid_entry *end) | |
1da177e4 | 2622 | { |
99f89551 | 2623 | struct task_struct *task = get_proc_task(dir); |
0168b9e3 | 2624 | struct dentry *res = ERR_PTR(-ENOENT); |
1da177e4 | 2625 | |
99f89551 EB |
2626 | if (!task) |
2627 | goto out_no_task; | |
1da177e4 | 2628 | |
20cdc894 EB |
2629 | /* |
2630 | * Yes, it does not scale. And it should not. Don't add | |
2631 | * new entries into /proc/<tgid>/ without very good reasons. | |
2632 | */ | |
d5a572a4 | 2633 | for (; p < end; p++) { |
1da177e4 LT |
2634 | if (p->len != dentry->d_name.len) |
2635 | continue; | |
26b95137 AD |
2636 | if (!memcmp(dentry->d_name.name, p->name, p->len)) { |
2637 | res = proc_pident_instantiate(dentry, task, p); | |
1da177e4 | 2638 | break; |
26b95137 | 2639 | } |
1da177e4 | 2640 | } |
99f89551 EB |
2641 | put_task_struct(task); |
2642 | out_no_task: | |
0168b9e3 | 2643 | return res; |
1da177e4 LT |
2644 | } |
2645 | ||
f0c3b509 | 2646 | static int proc_pident_readdir(struct file *file, struct dir_context *ctx, |
c5141e6d | 2647 | const struct pid_entry *ents, unsigned int nents) |
28a6d671 | 2648 | { |
f0c3b509 AV |
2649 | struct task_struct *task = get_proc_task(file_inode(file)); |
2650 | const struct pid_entry *p; | |
28a6d671 | 2651 | |
28a6d671 | 2652 | if (!task) |
f0c3b509 | 2653 | return -ENOENT; |
28a6d671 | 2654 | |
f0c3b509 AV |
2655 | if (!dir_emit_dots(file, ctx)) |
2656 | goto out; | |
2657 | ||
2658 | if (ctx->pos >= nents + 2) | |
2659 | goto out; | |
28a6d671 | 2660 | |
bac5f5d5 | 2661 | for (p = ents + (ctx->pos - 2); p < ents + nents; p++) { |
f0c3b509 AV |
2662 | if (!proc_fill_cache(file, ctx, p->name, p->len, |
2663 | proc_pident_instantiate, task, p)) | |
2664 | break; | |
2665 | ctx->pos++; | |
2666 | } | |
28a6d671 | 2667 | out: |
61a28784 | 2668 | put_task_struct(task); |
f0c3b509 | 2669 | return 0; |
1da177e4 LT |
2670 | } |
2671 | ||
28a6d671 EB |
2672 | #ifdef CONFIG_SECURITY |
2673 | static ssize_t proc_pid_attr_read(struct file * file, char __user * buf, | |
2674 | size_t count, loff_t *ppos) | |
2675 | { | |
496ad9aa | 2676 | struct inode * inode = file_inode(file); |
04ff9708 | 2677 | char *p = NULL; |
28a6d671 EB |
2678 | ssize_t length; |
2679 | struct task_struct *task = get_proc_task(inode); | |
2680 | ||
28a6d671 | 2681 | if (!task) |
04ff9708 | 2682 | return -ESRCH; |
28a6d671 | 2683 | |
6d9c939d | 2684 | length = security_getprocattr(task, PROC_I(inode)->op.lsm, |
2fddfeef | 2685 | (char*)file->f_path.dentry->d_name.name, |
04ff9708 | 2686 | &p); |
28a6d671 | 2687 | put_task_struct(task); |
04ff9708 AV |
2688 | if (length > 0) |
2689 | length = simple_read_from_buffer(buf, count, ppos, p, length); | |
2690 | kfree(p); | |
28a6d671 | 2691 | return length; |
1da177e4 LT |
2692 | } |
2693 | ||
28a6d671 EB |
2694 | static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf, |
2695 | size_t count, loff_t *ppos) | |
2696 | { | |
496ad9aa | 2697 | struct inode * inode = file_inode(file); |
41089b6d | 2698 | struct task_struct *task; |
bb646cdb | 2699 | void *page; |
41089b6d | 2700 | int rv; |
b21507e2 | 2701 | |
41089b6d AD |
2702 | rcu_read_lock(); |
2703 | task = pid_task(proc_pid(inode), PIDTYPE_PID); | |
2704 | if (!task) { | |
2705 | rcu_read_unlock(); | |
2706 | return -ESRCH; | |
2707 | } | |
b21507e2 | 2708 | /* A task may only write its own attributes. */ |
41089b6d AD |
2709 | if (current != task) { |
2710 | rcu_read_unlock(); | |
2711 | return -EACCES; | |
2712 | } | |
35a196be PM |
2713 | /* Prevent changes to overridden credentials. */ |
2714 | if (current_cred() != current_real_cred()) { | |
2715 | rcu_read_unlock(); | |
2716 | return -EBUSY; | |
2717 | } | |
41089b6d | 2718 | rcu_read_unlock(); |
b21507e2 | 2719 | |
28a6d671 EB |
2720 | if (count > PAGE_SIZE) |
2721 | count = PAGE_SIZE; | |
2722 | ||
2723 | /* No partial writes. */ | |
28a6d671 | 2724 | if (*ppos != 0) |
41089b6d | 2725 | return -EINVAL; |
28a6d671 | 2726 | |
bb646cdb AV |
2727 | page = memdup_user(buf, count); |
2728 | if (IS_ERR(page)) { | |
41089b6d | 2729 | rv = PTR_ERR(page); |
28a6d671 | 2730 | goto out; |
bb646cdb | 2731 | } |
28a6d671 | 2732 | |
107db7c7 | 2733 | /* Guard against adverse ptrace interaction */ |
41089b6d AD |
2734 | rv = mutex_lock_interruptible(¤t->signal->cred_guard_mutex); |
2735 | if (rv < 0) | |
107db7c7 DH |
2736 | goto out_free; |
2737 | ||
6d9c939d CS |
2738 | rv = security_setprocattr(PROC_I(inode)->op.lsm, |
2739 | file->f_path.dentry->d_name.name, page, | |
2740 | count); | |
b21507e2 | 2741 | mutex_unlock(¤t->signal->cred_guard_mutex); |
28a6d671 | 2742 | out_free: |
bb646cdb | 2743 | kfree(page); |
28a6d671 | 2744 | out: |
41089b6d | 2745 | return rv; |
28a6d671 EB |
2746 | } |
2747 | ||
00977a59 | 2748 | static const struct file_operations proc_pid_attr_operations = { |
28a6d671 EB |
2749 | .read = proc_pid_attr_read, |
2750 | .write = proc_pid_attr_write, | |
87df8424 | 2751 | .llseek = generic_file_llseek, |
28a6d671 EB |
2752 | }; |
2753 | ||
6d9c939d CS |
2754 | #define LSM_DIR_OPS(LSM) \ |
2755 | static int proc_##LSM##_attr_dir_iterate(struct file *filp, \ | |
2756 | struct dir_context *ctx) \ | |
2757 | { \ | |
2758 | return proc_pident_readdir(filp, ctx, \ | |
2759 | LSM##_attr_dir_stuff, \ | |
2760 | ARRAY_SIZE(LSM##_attr_dir_stuff)); \ | |
2761 | } \ | |
2762 | \ | |
2763 | static const struct file_operations proc_##LSM##_attr_dir_ops = { \ | |
2764 | .read = generic_read_dir, \ | |
2765 | .iterate = proc_##LSM##_attr_dir_iterate, \ | |
2766 | .llseek = default_llseek, \ | |
2767 | }; \ | |
2768 | \ | |
2769 | static struct dentry *proc_##LSM##_attr_dir_lookup(struct inode *dir, \ | |
2770 | struct dentry *dentry, unsigned int flags) \ | |
2771 | { \ | |
2772 | return proc_pident_lookup(dir, dentry, \ | |
2773 | LSM##_attr_dir_stuff, \ | |
d5a572a4 | 2774 | LSM##_attr_dir_stuff + ARRAY_SIZE(LSM##_attr_dir_stuff)); \ |
6d9c939d CS |
2775 | } \ |
2776 | \ | |
2777 | static const struct inode_operations proc_##LSM##_attr_dir_inode_ops = { \ | |
2778 | .lookup = proc_##LSM##_attr_dir_lookup, \ | |
2779 | .getattr = pid_getattr, \ | |
2780 | .setattr = proc_setattr, \ | |
2781 | } | |
2782 | ||
2783 | #ifdef CONFIG_SECURITY_SMACK | |
2784 | static const struct pid_entry smack_attr_dir_stuff[] = { | |
2785 | ATTR("smack", "current", 0666), | |
2786 | }; | |
2787 | LSM_DIR_OPS(smack); | |
2788 | #endif | |
2789 | ||
6413f852 JJ |
2790 | #ifdef CONFIG_SECURITY_APPARMOR |
2791 | static const struct pid_entry apparmor_attr_dir_stuff[] = { | |
2792 | ATTR("apparmor", "current", 0666), | |
2793 | ATTR("apparmor", "prev", 0444), | |
2794 | ATTR("apparmor", "exec", 0666), | |
2795 | }; | |
2796 | LSM_DIR_OPS(apparmor); | |
2797 | #endif | |
2798 | ||
c5141e6d | 2799 | static const struct pid_entry attr_dir_stuff[] = { |
6d9c939d CS |
2800 | ATTR(NULL, "current", 0666), |
2801 | ATTR(NULL, "prev", 0444), | |
2802 | ATTR(NULL, "exec", 0666), | |
2803 | ATTR(NULL, "fscreate", 0666), | |
2804 | ATTR(NULL, "keycreate", 0666), | |
2805 | ATTR(NULL, "sockcreate", 0666), | |
2806 | #ifdef CONFIG_SECURITY_SMACK | |
2807 | DIR("smack", 0555, | |
2808 | proc_smack_attr_dir_inode_ops, proc_smack_attr_dir_ops), | |
2809 | #endif | |
6413f852 JJ |
2810 | #ifdef CONFIG_SECURITY_APPARMOR |
2811 | DIR("apparmor", 0555, | |
2812 | proc_apparmor_attr_dir_inode_ops, proc_apparmor_attr_dir_ops), | |
2813 | #endif | |
28a6d671 EB |
2814 | }; |
2815 | ||
f0c3b509 | 2816 | static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx) |
28a6d671 | 2817 | { |
f0c3b509 AV |
2818 | return proc_pident_readdir(file, ctx, |
2819 | attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2820 | } |
2821 | ||
00977a59 | 2822 | static const struct file_operations proc_attr_dir_operations = { |
1da177e4 | 2823 | .read = generic_read_dir, |
f50752ea AV |
2824 | .iterate_shared = proc_attr_dir_readdir, |
2825 | .llseek = generic_file_llseek, | |
1da177e4 LT |
2826 | }; |
2827 | ||
72d9dcfc | 2828 | static struct dentry *proc_attr_dir_lookup(struct inode *dir, |
00cd8dd3 | 2829 | struct dentry *dentry, unsigned int flags) |
28a6d671 | 2830 | { |
7bcd6b0e | 2831 | return proc_pident_lookup(dir, dentry, |
d5a572a4 AD |
2832 | attr_dir_stuff, |
2833 | attr_dir_stuff + ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2834 | } |
2835 | ||
c5ef1c42 | 2836 | static const struct inode_operations proc_attr_dir_inode_operations = { |
72d9dcfc | 2837 | .lookup = proc_attr_dir_lookup, |
99f89551 | 2838 | .getattr = pid_getattr, |
6d76fa58 | 2839 | .setattr = proc_setattr, |
1da177e4 LT |
2840 | }; |
2841 | ||
28a6d671 EB |
2842 | #endif |
2843 | ||
698ba7b5 | 2844 | #ifdef CONFIG_ELF_CORE |
3cb4a0bb KH |
2845 | static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf, |
2846 | size_t count, loff_t *ppos) | |
2847 | { | |
496ad9aa | 2848 | struct task_struct *task = get_proc_task(file_inode(file)); |
3cb4a0bb KH |
2849 | struct mm_struct *mm; |
2850 | char buffer[PROC_NUMBUF]; | |
2851 | size_t len; | |
2852 | int ret; | |
2853 | ||
2854 | if (!task) | |
2855 | return -ESRCH; | |
2856 | ||
2857 | ret = 0; | |
2858 | mm = get_task_mm(task); | |
2859 | if (mm) { | |
2860 | len = snprintf(buffer, sizeof(buffer), "%08lx\n", | |
2861 | ((mm->flags & MMF_DUMP_FILTER_MASK) >> | |
2862 | MMF_DUMP_FILTER_SHIFT)); | |
2863 | mmput(mm); | |
2864 | ret = simple_read_from_buffer(buf, count, ppos, buffer, len); | |
2865 | } | |
2866 | ||
2867 | put_task_struct(task); | |
2868 | ||
2869 | return ret; | |
2870 | } | |
2871 | ||
2872 | static ssize_t proc_coredump_filter_write(struct file *file, | |
2873 | const char __user *buf, | |
2874 | size_t count, | |
2875 | loff_t *ppos) | |
2876 | { | |
2877 | struct task_struct *task; | |
2878 | struct mm_struct *mm; | |
3cb4a0bb KH |
2879 | unsigned int val; |
2880 | int ret; | |
2881 | int i; | |
2882 | unsigned long mask; | |
2883 | ||
774636e1 AD |
2884 | ret = kstrtouint_from_user(buf, count, 0, &val); |
2885 | if (ret < 0) | |
2886 | return ret; | |
3cb4a0bb KH |
2887 | |
2888 | ret = -ESRCH; | |
496ad9aa | 2889 | task = get_proc_task(file_inode(file)); |
3cb4a0bb KH |
2890 | if (!task) |
2891 | goto out_no_task; | |
2892 | ||
3cb4a0bb KH |
2893 | mm = get_task_mm(task); |
2894 | if (!mm) | |
2895 | goto out_no_mm; | |
41a0c249 | 2896 | ret = 0; |
3cb4a0bb KH |
2897 | |
2898 | for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) { | |
2899 | if (val & mask) | |
2900 | set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2901 | else | |
2902 | clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2903 | } | |
2904 | ||
2905 | mmput(mm); | |
2906 | out_no_mm: | |
2907 | put_task_struct(task); | |
2908 | out_no_task: | |
774636e1 AD |
2909 | if (ret < 0) |
2910 | return ret; | |
2911 | return count; | |
3cb4a0bb KH |
2912 | } |
2913 | ||
2914 | static const struct file_operations proc_coredump_filter_operations = { | |
2915 | .read = proc_coredump_filter_read, | |
2916 | .write = proc_coredump_filter_write, | |
87df8424 | 2917 | .llseek = generic_file_llseek, |
3cb4a0bb KH |
2918 | }; |
2919 | #endif | |
2920 | ||
aba76fdb | 2921 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 2922 | static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole) |
297c5d92 | 2923 | { |
940389b8 | 2924 | struct task_io_accounting acct = task->ioac; |
5995477a | 2925 | unsigned long flags; |
293eb1e7 | 2926 | int result; |
5995477a | 2927 | |
76518d37 | 2928 | result = mutex_lock_killable(&task->signal->exec_update_mutex); |
293eb1e7 VK |
2929 | if (result) |
2930 | return result; | |
2931 | ||
caaee623 | 2932 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) { |
293eb1e7 VK |
2933 | result = -EACCES; |
2934 | goto out_unlock; | |
2935 | } | |
1d1221f3 | 2936 | |
5995477a AR |
2937 | if (whole && lock_task_sighand(task, &flags)) { |
2938 | struct task_struct *t = task; | |
2939 | ||
2940 | task_io_accounting_add(&acct, &task->signal->ioac); | |
2941 | while_each_thread(task, t) | |
2942 | task_io_accounting_add(&acct, &t->ioac); | |
2943 | ||
2944 | unlock_task_sighand(task, &flags); | |
297c5d92 | 2945 | } |
25ce3191 JP |
2946 | seq_printf(m, |
2947 | "rchar: %llu\n" | |
2948 | "wchar: %llu\n" | |
2949 | "syscr: %llu\n" | |
2950 | "syscw: %llu\n" | |
2951 | "read_bytes: %llu\n" | |
2952 | "write_bytes: %llu\n" | |
2953 | "cancelled_write_bytes: %llu\n", | |
2954 | (unsigned long long)acct.rchar, | |
2955 | (unsigned long long)acct.wchar, | |
2956 | (unsigned long long)acct.syscr, | |
2957 | (unsigned long long)acct.syscw, | |
2958 | (unsigned long long)acct.read_bytes, | |
2959 | (unsigned long long)acct.write_bytes, | |
2960 | (unsigned long long)acct.cancelled_write_bytes); | |
2961 | result = 0; | |
2962 | ||
293eb1e7 | 2963 | out_unlock: |
76518d37 | 2964 | mutex_unlock(&task->signal->exec_update_mutex); |
293eb1e7 | 2965 | return result; |
297c5d92 AR |
2966 | } |
2967 | ||
19aadc98 AD |
2968 | static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns, |
2969 | struct pid *pid, struct task_struct *task) | |
297c5d92 | 2970 | { |
19aadc98 | 2971 | return do_io_accounting(task, m, 0); |
aba76fdb | 2972 | } |
297c5d92 | 2973 | |
19aadc98 AD |
2974 | static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns, |
2975 | struct pid *pid, struct task_struct *task) | |
297c5d92 | 2976 | { |
19aadc98 | 2977 | return do_io_accounting(task, m, 1); |
297c5d92 AR |
2978 | } |
2979 | #endif /* CONFIG_TASK_IO_ACCOUNTING */ | |
aba76fdb | 2980 | |
22d917d8 EB |
2981 | #ifdef CONFIG_USER_NS |
2982 | static int proc_id_map_open(struct inode *inode, struct file *file, | |
ccf94f1b | 2983 | const struct seq_operations *seq_ops) |
22d917d8 EB |
2984 | { |
2985 | struct user_namespace *ns = NULL; | |
2986 | struct task_struct *task; | |
2987 | struct seq_file *seq; | |
2988 | int ret = -EINVAL; | |
2989 | ||
2990 | task = get_proc_task(inode); | |
2991 | if (task) { | |
2992 | rcu_read_lock(); | |
2993 | ns = get_user_ns(task_cred_xxx(task, user_ns)); | |
2994 | rcu_read_unlock(); | |
2995 | put_task_struct(task); | |
2996 | } | |
2997 | if (!ns) | |
2998 | goto err; | |
2999 | ||
3000 | ret = seq_open(file, seq_ops); | |
3001 | if (ret) | |
3002 | goto err_put_ns; | |
3003 | ||
3004 | seq = file->private_data; | |
3005 | seq->private = ns; | |
3006 | ||
3007 | return 0; | |
3008 | err_put_ns: | |
3009 | put_user_ns(ns); | |
3010 | err: | |
3011 | return ret; | |
3012 | } | |
3013 | ||
3014 | static int proc_id_map_release(struct inode *inode, struct file *file) | |
3015 | { | |
3016 | struct seq_file *seq = file->private_data; | |
3017 | struct user_namespace *ns = seq->private; | |
3018 | put_user_ns(ns); | |
3019 | return seq_release(inode, file); | |
3020 | } | |
3021 | ||
3022 | static int proc_uid_map_open(struct inode *inode, struct file *file) | |
3023 | { | |
3024 | return proc_id_map_open(inode, file, &proc_uid_seq_operations); | |
3025 | } | |
3026 | ||
3027 | static int proc_gid_map_open(struct inode *inode, struct file *file) | |
3028 | { | |
3029 | return proc_id_map_open(inode, file, &proc_gid_seq_operations); | |
3030 | } | |
3031 | ||
f76d207a EB |
3032 | static int proc_projid_map_open(struct inode *inode, struct file *file) |
3033 | { | |
3034 | return proc_id_map_open(inode, file, &proc_projid_seq_operations); | |
3035 | } | |
3036 | ||
22d917d8 EB |
3037 | static const struct file_operations proc_uid_map_operations = { |
3038 | .open = proc_uid_map_open, | |
3039 | .write = proc_uid_map_write, | |
3040 | .read = seq_read, | |
3041 | .llseek = seq_lseek, | |
3042 | .release = proc_id_map_release, | |
3043 | }; | |
3044 | ||
3045 | static const struct file_operations proc_gid_map_operations = { | |
3046 | .open = proc_gid_map_open, | |
3047 | .write = proc_gid_map_write, | |
3048 | .read = seq_read, | |
3049 | .llseek = seq_lseek, | |
3050 | .release = proc_id_map_release, | |
3051 | }; | |
f76d207a EB |
3052 | |
3053 | static const struct file_operations proc_projid_map_operations = { | |
3054 | .open = proc_projid_map_open, | |
3055 | .write = proc_projid_map_write, | |
3056 | .read = seq_read, | |
3057 | .llseek = seq_lseek, | |
3058 | .release = proc_id_map_release, | |
3059 | }; | |
9cc46516 EB |
3060 | |
3061 | static int proc_setgroups_open(struct inode *inode, struct file *file) | |
3062 | { | |
3063 | struct user_namespace *ns = NULL; | |
3064 | struct task_struct *task; | |
3065 | int ret; | |
3066 | ||
3067 | ret = -ESRCH; | |
3068 | task = get_proc_task(inode); | |
3069 | if (task) { | |
3070 | rcu_read_lock(); | |
3071 | ns = get_user_ns(task_cred_xxx(task, user_ns)); | |
3072 | rcu_read_unlock(); | |
3073 | put_task_struct(task); | |
3074 | } | |
3075 | if (!ns) | |
3076 | goto err; | |
3077 | ||
3078 | if (file->f_mode & FMODE_WRITE) { | |
3079 | ret = -EACCES; | |
3080 | if (!ns_capable(ns, CAP_SYS_ADMIN)) | |
3081 | goto err_put_ns; | |
3082 | } | |
3083 | ||
3084 | ret = single_open(file, &proc_setgroups_show, ns); | |
3085 | if (ret) | |
3086 | goto err_put_ns; | |
3087 | ||
3088 | return 0; | |
3089 | err_put_ns: | |
3090 | put_user_ns(ns); | |
3091 | err: | |
3092 | return ret; | |
3093 | } | |
3094 | ||
3095 | static int proc_setgroups_release(struct inode *inode, struct file *file) | |
3096 | { | |
3097 | struct seq_file *seq = file->private_data; | |
3098 | struct user_namespace *ns = seq->private; | |
3099 | int ret = single_release(inode, file); | |
3100 | put_user_ns(ns); | |
3101 | return ret; | |
3102 | } | |
3103 | ||
3104 | static const struct file_operations proc_setgroups_operations = { | |
3105 | .open = proc_setgroups_open, | |
3106 | .write = proc_setgroups_write, | |
3107 | .read = seq_read, | |
3108 | .llseek = seq_lseek, | |
3109 | .release = proc_setgroups_release, | |
3110 | }; | |
22d917d8 EB |
3111 | #endif /* CONFIG_USER_NS */ |
3112 | ||
47830723 KC |
3113 | static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns, |
3114 | struct pid *pid, struct task_struct *task) | |
3115 | { | |
a9712bc1 AV |
3116 | int err = lock_trace(task); |
3117 | if (!err) { | |
3118 | seq_printf(m, "%08x\n", task->personality); | |
3119 | unlock_trace(task); | |
3120 | } | |
3121 | return err; | |
47830723 KC |
3122 | } |
3123 | ||
7c23b330 JP |
3124 | #ifdef CONFIG_LIVEPATCH |
3125 | static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns, | |
3126 | struct pid *pid, struct task_struct *task) | |
3127 | { | |
3128 | seq_printf(m, "%d\n", task->patch_state); | |
3129 | return 0; | |
3130 | } | |
3131 | #endif /* CONFIG_LIVEPATCH */ | |
3132 | ||
c8d12627 AP |
3133 | #ifdef CONFIG_STACKLEAK_METRICS |
3134 | static int proc_stack_depth(struct seq_file *m, struct pid_namespace *ns, | |
3135 | struct pid *pid, struct task_struct *task) | |
3136 | { | |
3137 | unsigned long prev_depth = THREAD_SIZE - | |
3138 | (task->prev_lowest_stack & (THREAD_SIZE - 1)); | |
3139 | unsigned long depth = THREAD_SIZE - | |
3140 | (task->lowest_stack & (THREAD_SIZE - 1)); | |
3141 | ||
3142 | seq_printf(m, "previous stack depth: %lu\nstack depth: %lu\n", | |
3143 | prev_depth, depth); | |
3144 | return 0; | |
3145 | } | |
3146 | #endif /* CONFIG_STACKLEAK_METRICS */ | |
3147 | ||
28a6d671 EB |
3148 | /* |
3149 | * Thread groups | |
3150 | */ | |
00977a59 | 3151 | static const struct file_operations proc_task_operations; |
c5ef1c42 | 3152 | static const struct inode_operations proc_task_inode_operations; |
20cdc894 | 3153 | |
c5141e6d | 3154 | static const struct pid_entry tgid_base_stuff[] = { |
631f9c18 AD |
3155 | DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations), |
3156 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), | |
640708a2 | 3157 | DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations), |
631f9c18 | 3158 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
6b4e306a | 3159 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
b2211a36 | 3160 | #ifdef CONFIG_NET |
631f9c18 | 3161 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), |
b2211a36 | 3162 | #endif |
631f9c18 | 3163 | REG("environ", S_IRUSR, proc_environ_operations), |
c5317167 | 3164 | REG("auxv", S_IRUSR, proc_auxv_operations), |
631f9c18 | 3165 | ONE("status", S_IRUGO, proc_pid_status), |
35a35046 | 3166 | ONE("personality", S_IRUSR, proc_pid_personality), |
1c963eb1 | 3167 | ONE("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 3168 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 3169 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
5091faa4 MG |
3170 | #endif |
3171 | #ifdef CONFIG_SCHED_AUTOGROUP | |
3172 | REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations), | |
04a8682a AV |
3173 | #endif |
3174 | #ifdef CONFIG_TIME_NS | |
3175 | REG("timens_offsets", S_IRUGO|S_IWUSR, proc_timens_offsets_operations), | |
ebcb6734 | 3176 | #endif |
4614a696 | 3177 | REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations), |
ebcb6734 | 3178 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 | 3179 | ONE("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 3180 | #endif |
c2c0bb44 | 3181 | REG("cmdline", S_IRUGO, proc_pid_cmdline_ops), |
631f9c18 AD |
3182 | ONE("stat", S_IRUGO, proc_tgid_stat), |
3183 | ONE("statm", S_IRUGO, proc_pid_statm), | |
b7643757 | 3184 | REG("maps", S_IRUGO, proc_pid_maps_operations), |
28a6d671 | 3185 | #ifdef CONFIG_NUMA |
b7643757 | 3186 | REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations), |
28a6d671 | 3187 | #endif |
631f9c18 AD |
3188 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
3189 | LNK("cwd", proc_cwd_link), | |
3190 | LNK("root", proc_root_link), | |
3191 | LNK("exe", proc_exe_link), | |
3192 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
3193 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
3194 | REG("mountstats", S_IRUSR, proc_mountstats_operations), | |
1e883281 | 3195 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 | 3196 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
b7643757 | 3197 | REG("smaps", S_IRUGO, proc_pid_smaps_operations), |
493b0e9d | 3198 | REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations), |
32ed74a4 | 3199 | REG("pagemap", S_IRUSR, proc_pagemap_operations), |
28a6d671 EB |
3200 | #endif |
3201 | #ifdef CONFIG_SECURITY | |
631f9c18 | 3202 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
3203 | #endif |
3204 | #ifdef CONFIG_KALLSYMS | |
edfcd606 | 3205 | ONE("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 3206 | #endif |
2ec220e2 | 3207 | #ifdef CONFIG_STACKTRACE |
35a35046 | 3208 | ONE("stack", S_IRUSR, proc_pid_stack), |
28a6d671 | 3209 | #endif |
5968cece | 3210 | #ifdef CONFIG_SCHED_INFO |
f6e826ca | 3211 | ONE("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 3212 | #endif |
9745512c | 3213 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 3214 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 3215 | #endif |
8793d854 | 3216 | #ifdef CONFIG_PROC_PID_CPUSET |
52de4779 | 3217 | ONE("cpuset", S_IRUGO, proc_cpuset_show), |
a424316c PM |
3218 | #endif |
3219 | #ifdef CONFIG_CGROUPS | |
006f4ac4 | 3220 | ONE("cgroup", S_IRUGO, proc_cgroup_show), |
e79f15a4 CY |
3221 | #endif |
3222 | #ifdef CONFIG_PROC_CPU_RESCTRL | |
3223 | ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show), | |
28a6d671 | 3224 | #endif |
6ba51e37 | 3225 | ONE("oom_score", S_IRUGO, proc_oom_score), |
fa0cbbf1 | 3226 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations), |
a63d83f4 | 3227 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
4b7d248b | 3228 | #ifdef CONFIG_AUDIT |
631f9c18 AD |
3229 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
3230 | REG("sessionid", S_IRUGO, proc_sessionid_operations), | |
28a6d671 | 3231 | #endif |
f4f154fd | 3232 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 3233 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
168c42bc | 3234 | REG("fail-nth", 0644, proc_fail_nth_operations), |
f4f154fd | 3235 | #endif |
698ba7b5 | 3236 | #ifdef CONFIG_ELF_CORE |
631f9c18 | 3237 | REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations), |
3cb4a0bb | 3238 | #endif |
aba76fdb | 3239 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 3240 | ONE("io", S_IRUSR, proc_tgid_io_accounting), |
aba76fdb | 3241 | #endif |
22d917d8 EB |
3242 | #ifdef CONFIG_USER_NS |
3243 | REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations), | |
3244 | REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations), | |
f76d207a | 3245 | REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations), |
9cc46516 | 3246 | REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations), |
22d917d8 | 3247 | #endif |
b18b6a9c | 3248 | #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS) |
48f6a7a5 PE |
3249 | REG("timers", S_IRUGO, proc_timers_operations), |
3250 | #endif | |
5de23d43 | 3251 | REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations), |
7c23b330 JP |
3252 | #ifdef CONFIG_LIVEPATCH |
3253 | ONE("patch_state", S_IRUSR, proc_pid_patch_state), | |
3254 | #endif | |
c8d12627 AP |
3255 | #ifdef CONFIG_STACKLEAK_METRICS |
3256 | ONE("stack_depth", S_IRUGO, proc_stack_depth), | |
3257 | #endif | |
68bc30bb AL |
3258 | #ifdef CONFIG_PROC_PID_ARCH_STATUS |
3259 | ONE("arch_status", S_IRUGO, proc_pid_arch_status), | |
3260 | #endif | |
28a6d671 | 3261 | }; |
1da177e4 | 3262 | |
f0c3b509 | 3263 | static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3264 | { |
f0c3b509 AV |
3265 | return proc_pident_readdir(file, ctx, |
3266 | tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
3267 | } |
3268 | ||
00977a59 | 3269 | static const struct file_operations proc_tgid_base_operations = { |
1da177e4 | 3270 | .read = generic_read_dir, |
f50752ea AV |
3271 | .iterate_shared = proc_tgid_base_readdir, |
3272 | .llseek = generic_file_llseek, | |
1da177e4 LT |
3273 | }; |
3274 | ||
3eb39f47 CB |
3275 | struct pid *tgid_pidfd_to_pid(const struct file *file) |
3276 | { | |
30d158b1 | 3277 | if (file->f_op != &proc_tgid_base_operations) |
3eb39f47 CB |
3278 | return ERR_PTR(-EBADF); |
3279 | ||
3280 | return proc_pid(file_inode(file)); | |
3281 | } | |
3282 | ||
00cd8dd3 AV |
3283 | static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
3284 | { | |
7bcd6b0e | 3285 | return proc_pident_lookup(dir, dentry, |
d5a572a4 AD |
3286 | tgid_base_stuff, |
3287 | tgid_base_stuff + ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
3288 | } |
3289 | ||
c5ef1c42 | 3290 | static const struct inode_operations proc_tgid_base_inode_operations = { |
28a6d671 | 3291 | .lookup = proc_tgid_base_lookup, |
99f89551 | 3292 | .getattr = pid_getattr, |
6d76fa58 | 3293 | .setattr = proc_setattr, |
0499680a | 3294 | .permission = proc_pid_permission, |
1da177e4 | 3295 | }; |
1da177e4 | 3296 | |
0895e91d | 3297 | /** |
7bc3e6e5 EB |
3298 | * proc_flush_pid - Remove dcache entries for @pid from the /proc dcache. |
3299 | * @pid: pid that should be flushed. | |
0895e91d | 3300 | * |
7bc3e6e5 EB |
3301 | * This function walks a list of inodes (that belong to any proc |
3302 | * filesystem) that are attached to the pid and flushes them from | |
3303 | * the dentry cache. | |
0895e91d RD |
3304 | * |
3305 | * It is safe and reasonable to cache /proc entries for a task until | |
3306 | * that task exits. After that they just clog up the dcache with | |
3307 | * useless entries, possibly causing useful dcache entries to be | |
7bc3e6e5 EB |
3308 | * flushed instead. This routine is provided to flush those useless |
3309 | * dcache entries when a process is reaped. | |
0895e91d RD |
3310 | * |
3311 | * NOTE: This routine is just an optimization so it does not guarantee | |
7bc3e6e5 EB |
3312 | * that no dcache entries will exist after a process is reaped |
3313 | * it just makes it very unlikely that any will persist. | |
60347f67 PE |
3314 | */ |
3315 | ||
7bc3e6e5 | 3316 | void proc_flush_pid(struct pid *pid) |
60347f67 | 3317 | { |
63f818f4 | 3318 | proc_invalidate_siblings_dcache(&pid->inodes, &pid->lock); |
60347f67 PE |
3319 | } |
3320 | ||
0168b9e3 | 3321 | static struct dentry *proc_pid_instantiate(struct dentry * dentry, |
c52a47ac | 3322 | struct task_struct *task, const void *ptr) |
444ceed8 | 3323 | { |
444ceed8 EB |
3324 | struct inode *inode; |
3325 | ||
0168b9e3 | 3326 | inode = proc_pid_make_inode(dentry->d_sb, task, S_IFDIR | S_IRUGO | S_IXUGO); |
444ceed8 | 3327 | if (!inode) |
0168b9e3 | 3328 | return ERR_PTR(-ENOENT); |
444ceed8 | 3329 | |
444ceed8 EB |
3330 | inode->i_op = &proc_tgid_base_inode_operations; |
3331 | inode->i_fop = &proc_tgid_base_operations; | |
3332 | inode->i_flags|=S_IMMUTABLE; | |
aed54175 | 3333 | |
1270dd8d | 3334 | set_nlink(inode, nlink_tgid); |
1bbc5513 | 3335 | pid_update_inode(task, inode); |
444ceed8 | 3336 | |
fb045adb | 3337 | d_set_d_op(dentry, &pid_dentry_operations); |
0168b9e3 | 3338 | return d_splice_alias(inode, dentry); |
444ceed8 EB |
3339 | } |
3340 | ||
867aaccf | 3341 | struct dentry *proc_pid_lookup(struct dentry *dentry, unsigned int flags) |
1da177e4 LT |
3342 | { |
3343 | struct task_struct *task; | |
1da177e4 | 3344 | unsigned tgid; |
fa10fed3 | 3345 | struct proc_fs_info *fs_info; |
b488893a | 3346 | struct pid_namespace *ns; |
0168b9e3 | 3347 | struct dentry *result = ERR_PTR(-ENOENT); |
1da177e4 | 3348 | |
dbcdb504 | 3349 | tgid = name_to_int(&dentry->d_name); |
1da177e4 LT |
3350 | if (tgid == ~0U) |
3351 | goto out; | |
3352 | ||
fa10fed3 AG |
3353 | fs_info = proc_sb_info(dentry->d_sb); |
3354 | ns = fs_info->pid_ns; | |
de758734 | 3355 | rcu_read_lock(); |
b488893a | 3356 | task = find_task_by_pid_ns(tgid, ns); |
1da177e4 LT |
3357 | if (task) |
3358 | get_task_struct(task); | |
de758734 | 3359 | rcu_read_unlock(); |
1da177e4 LT |
3360 | if (!task) |
3361 | goto out; | |
3362 | ||
24a71ce5 AG |
3363 | /* Limit procfs to only ptraceable tasks */ |
3364 | if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE) { | |
3365 | if (!has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS)) | |
3366 | goto out_put_task; | |
3367 | } | |
3368 | ||
0168b9e3 | 3369 | result = proc_pid_instantiate(dentry, task, NULL); |
24a71ce5 | 3370 | out_put_task: |
1da177e4 | 3371 | put_task_struct(task); |
1da177e4 | 3372 | out: |
0168b9e3 | 3373 | return result; |
1da177e4 LT |
3374 | } |
3375 | ||
1da177e4 | 3376 | /* |
0804ef4b | 3377 | * Find the first task with tgid >= tgid |
0bc58a91 | 3378 | * |
1da177e4 | 3379 | */ |
19fd4bb2 EB |
3380 | struct tgid_iter { |
3381 | unsigned int tgid; | |
0804ef4b | 3382 | struct task_struct *task; |
19fd4bb2 EB |
3383 | }; |
3384 | static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter) | |
3385 | { | |
0804ef4b | 3386 | struct pid *pid; |
1da177e4 | 3387 | |
19fd4bb2 EB |
3388 | if (iter.task) |
3389 | put_task_struct(iter.task); | |
454cc105 | 3390 | rcu_read_lock(); |
0804ef4b | 3391 | retry: |
19fd4bb2 EB |
3392 | iter.task = NULL; |
3393 | pid = find_ge_pid(iter.tgid, ns); | |
0804ef4b | 3394 | if (pid) { |
19fd4bb2 | 3395 | iter.tgid = pid_nr_ns(pid, ns); |
3147d8aa EB |
3396 | iter.task = pid_task(pid, PIDTYPE_TGID); |
3397 | if (!iter.task) { | |
19fd4bb2 | 3398 | iter.tgid += 1; |
0804ef4b | 3399 | goto retry; |
19fd4bb2 EB |
3400 | } |
3401 | get_task_struct(iter.task); | |
0bc58a91 | 3402 | } |
454cc105 | 3403 | rcu_read_unlock(); |
19fd4bb2 | 3404 | return iter; |
1da177e4 LT |
3405 | } |
3406 | ||
0097875b | 3407 | #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2) |
0804ef4b | 3408 | |
1da177e4 | 3409 | /* for the /proc/ directory itself, after non-process stuff has been done */ |
f0c3b509 | 3410 | int proc_pid_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3411 | { |
19fd4bb2 | 3412 | struct tgid_iter iter; |
fa10fed3 | 3413 | struct proc_fs_info *fs_info = proc_sb_info(file_inode(file)->i_sb); |
9d78edea | 3414 | struct pid_namespace *ns = proc_pid_ns(file_inode(file)->i_sb); |
f0c3b509 | 3415 | loff_t pos = ctx->pos; |
1da177e4 | 3416 | |
021ada7d | 3417 | if (pos >= PID_MAX_LIMIT + TGID_OFFSET) |
f0c3b509 | 3418 | return 0; |
1da177e4 | 3419 | |
0097875b | 3420 | if (pos == TGID_OFFSET - 2) { |
fa10fed3 | 3421 | struct inode *inode = d_inode(fs_info->proc_self); |
db963164 | 3422 | if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK)) |
f0c3b509 | 3423 | return 0; |
0097875b EB |
3424 | ctx->pos = pos = pos + 1; |
3425 | } | |
3426 | if (pos == TGID_OFFSET - 1) { | |
fa10fed3 | 3427 | struct inode *inode = d_inode(fs_info->proc_thread_self); |
0097875b EB |
3428 | if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK)) |
3429 | return 0; | |
3430 | ctx->pos = pos = pos + 1; | |
021ada7d | 3431 | } |
0097875b | 3432 | iter.tgid = pos - TGID_OFFSET; |
19fd4bb2 | 3433 | iter.task = NULL; |
19fd4bb2 EB |
3434 | for (iter = next_tgid(ns, iter); |
3435 | iter.task; | |
3436 | iter.tgid += 1, iter = next_tgid(ns, iter)) { | |
e3912ac3 | 3437 | char name[10 + 1]; |
a4ef3895 | 3438 | unsigned int len; |
3ba4bcee ED |
3439 | |
3440 | cond_resched(); | |
fa10fed3 | 3441 | if (!has_pid_permissions(fs_info, iter.task, HIDEPID_INVISIBLE)) |
f0c3b509 | 3442 | continue; |
0499680a | 3443 | |
e3912ac3 | 3444 | len = snprintf(name, sizeof(name), "%u", iter.tgid); |
f0c3b509 AV |
3445 | ctx->pos = iter.tgid + TGID_OFFSET; |
3446 | if (!proc_fill_cache(file, ctx, name, len, | |
3447 | proc_pid_instantiate, iter.task, NULL)) { | |
19fd4bb2 | 3448 | put_task_struct(iter.task); |
f0c3b509 | 3449 | return 0; |
1da177e4 | 3450 | } |
0bc58a91 | 3451 | } |
f0c3b509 | 3452 | ctx->pos = PID_MAX_LIMIT + TGID_OFFSET; |
0bc58a91 EB |
3453 | return 0; |
3454 | } | |
1da177e4 | 3455 | |
1b3044e3 JD |
3456 | /* |
3457 | * proc_tid_comm_permission is a special permission function exclusively | |
3458 | * used for the node /proc/<pid>/task/<tid>/comm. | |
3459 | * It bypasses generic permission checks in the case where a task of the same | |
3460 | * task group attempts to access the node. | |
3461 | * The rationale behind this is that glibc and bionic access this node for | |
3462 | * cross thread naming (pthread_set/getname_np(!self)). However, if | |
3463 | * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0, | |
3464 | * which locks out the cross thread naming implementation. | |
3465 | * This function makes sure that the node is always accessible for members of | |
3466 | * same thread group. | |
3467 | */ | |
3468 | static int proc_tid_comm_permission(struct inode *inode, int mask) | |
3469 | { | |
3470 | bool is_same_tgroup; | |
3471 | struct task_struct *task; | |
3472 | ||
3473 | task = get_proc_task(inode); | |
3474 | if (!task) | |
3475 | return -ESRCH; | |
3476 | is_same_tgroup = same_thread_group(current, task); | |
3477 | put_task_struct(task); | |
3478 | ||
3479 | if (likely(is_same_tgroup && !(mask & MAY_EXEC))) { | |
3480 | /* This file (/proc/<pid>/task/<tid>/comm) can always be | |
3481 | * read or written by the members of the corresponding | |
3482 | * thread group. | |
3483 | */ | |
3484 | return 0; | |
3485 | } | |
3486 | ||
3487 | return generic_permission(inode, mask); | |
3488 | } | |
3489 | ||
3490 | static const struct inode_operations proc_tid_comm_inode_operations = { | |
3491 | .permission = proc_tid_comm_permission, | |
3492 | }; | |
3493 | ||
28a6d671 EB |
3494 | /* |
3495 | * Tasks | |
3496 | */ | |
c5141e6d | 3497 | static const struct pid_entry tid_base_stuff[] = { |
631f9c18 | 3498 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), |
3835541d | 3499 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
6b4e306a | 3500 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
6ba8ed79 EB |
3501 | #ifdef CONFIG_NET |
3502 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), | |
3503 | #endif | |
631f9c18 | 3504 | REG("environ", S_IRUSR, proc_environ_operations), |
c5317167 | 3505 | REG("auxv", S_IRUSR, proc_auxv_operations), |
631f9c18 | 3506 | ONE("status", S_IRUGO, proc_pid_status), |
35a35046 | 3507 | ONE("personality", S_IRUSR, proc_pid_personality), |
1c963eb1 | 3508 | ONE("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 3509 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 3510 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
ebcb6734 | 3511 | #endif |
1b3044e3 JD |
3512 | NOD("comm", S_IFREG|S_IRUGO|S_IWUSR, |
3513 | &proc_tid_comm_inode_operations, | |
3514 | &proc_pid_set_comm_operations, {}), | |
ebcb6734 | 3515 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 | 3516 | ONE("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 3517 | #endif |
c2c0bb44 | 3518 | REG("cmdline", S_IRUGO, proc_pid_cmdline_ops), |
631f9c18 AD |
3519 | ONE("stat", S_IRUGO, proc_tid_stat), |
3520 | ONE("statm", S_IRUGO, proc_pid_statm), | |
871305bb | 3521 | REG("maps", S_IRUGO, proc_pid_maps_operations), |
2e13ba54 | 3522 | #ifdef CONFIG_PROC_CHILDREN |
81841161 CG |
3523 | REG("children", S_IRUGO, proc_tid_children_operations), |
3524 | #endif | |
28a6d671 | 3525 | #ifdef CONFIG_NUMA |
871305bb | 3526 | REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations), |
28a6d671 | 3527 | #endif |
631f9c18 AD |
3528 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
3529 | LNK("cwd", proc_cwd_link), | |
3530 | LNK("root", proc_root_link), | |
3531 | LNK("exe", proc_exe_link), | |
3532 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
3533 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
1e883281 | 3534 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 | 3535 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
871305bb | 3536 | REG("smaps", S_IRUGO, proc_pid_smaps_operations), |
493b0e9d | 3537 | REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations), |
32ed74a4 | 3538 | REG("pagemap", S_IRUSR, proc_pagemap_operations), |
28a6d671 EB |
3539 | #endif |
3540 | #ifdef CONFIG_SECURITY | |
631f9c18 | 3541 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
3542 | #endif |
3543 | #ifdef CONFIG_KALLSYMS | |
edfcd606 | 3544 | ONE("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 3545 | #endif |
2ec220e2 | 3546 | #ifdef CONFIG_STACKTRACE |
35a35046 | 3547 | ONE("stack", S_IRUSR, proc_pid_stack), |
28a6d671 | 3548 | #endif |
5968cece | 3549 | #ifdef CONFIG_SCHED_INFO |
f6e826ca | 3550 | ONE("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 3551 | #endif |
9745512c | 3552 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 3553 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 3554 | #endif |
8793d854 | 3555 | #ifdef CONFIG_PROC_PID_CPUSET |
52de4779 | 3556 | ONE("cpuset", S_IRUGO, proc_cpuset_show), |
a424316c PM |
3557 | #endif |
3558 | #ifdef CONFIG_CGROUPS | |
006f4ac4 | 3559 | ONE("cgroup", S_IRUGO, proc_cgroup_show), |
e79f15a4 CY |
3560 | #endif |
3561 | #ifdef CONFIG_PROC_CPU_RESCTRL | |
3562 | ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show), | |
28a6d671 | 3563 | #endif |
6ba51e37 | 3564 | ONE("oom_score", S_IRUGO, proc_oom_score), |
fa0cbbf1 | 3565 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations), |
a63d83f4 | 3566 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
4b7d248b | 3567 | #ifdef CONFIG_AUDIT |
631f9c18 | 3568 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
26ec3c64 | 3569 | REG("sessionid", S_IRUGO, proc_sessionid_operations), |
28a6d671 | 3570 | #endif |
f4f154fd | 3571 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 3572 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
1203c8e6 | 3573 | REG("fail-nth", 0644, proc_fail_nth_operations), |
f4f154fd | 3574 | #endif |
297c5d92 | 3575 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 3576 | ONE("io", S_IRUSR, proc_tid_io_accounting), |
297c5d92 | 3577 | #endif |
22d917d8 EB |
3578 | #ifdef CONFIG_USER_NS |
3579 | REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations), | |
3580 | REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations), | |
f76d207a | 3581 | REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations), |
9cc46516 | 3582 | REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations), |
22d917d8 | 3583 | #endif |
7c23b330 JP |
3584 | #ifdef CONFIG_LIVEPATCH |
3585 | ONE("patch_state", S_IRUSR, proc_pid_patch_state), | |
3586 | #endif | |
68bc30bb AL |
3587 | #ifdef CONFIG_PROC_PID_ARCH_STATUS |
3588 | ONE("arch_status", S_IRUGO, proc_pid_arch_status), | |
3589 | #endif | |
28a6d671 EB |
3590 | }; |
3591 | ||
f0c3b509 | 3592 | static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx) |
28a6d671 | 3593 | { |
f0c3b509 AV |
3594 | return proc_pident_readdir(file, ctx, |
3595 | tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
3596 | } |
3597 | ||
00cd8dd3 AV |
3598 | static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
3599 | { | |
7bcd6b0e | 3600 | return proc_pident_lookup(dir, dentry, |
d5a572a4 AD |
3601 | tid_base_stuff, |
3602 | tid_base_stuff + ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
3603 | } |
3604 | ||
00977a59 | 3605 | static const struct file_operations proc_tid_base_operations = { |
28a6d671 | 3606 | .read = generic_read_dir, |
f50752ea AV |
3607 | .iterate_shared = proc_tid_base_readdir, |
3608 | .llseek = generic_file_llseek, | |
28a6d671 EB |
3609 | }; |
3610 | ||
c5ef1c42 | 3611 | static const struct inode_operations proc_tid_base_inode_operations = { |
28a6d671 EB |
3612 | .lookup = proc_tid_base_lookup, |
3613 | .getattr = pid_getattr, | |
3614 | .setattr = proc_setattr, | |
3615 | }; | |
3616 | ||
0168b9e3 AV |
3617 | static struct dentry *proc_task_instantiate(struct dentry *dentry, |
3618 | struct task_struct *task, const void *ptr) | |
444ceed8 | 3619 | { |
444ceed8 | 3620 | struct inode *inode; |
0168b9e3 | 3621 | inode = proc_pid_make_inode(dentry->d_sb, task, S_IFDIR | S_IRUGO | S_IXUGO); |
444ceed8 | 3622 | if (!inode) |
0168b9e3 | 3623 | return ERR_PTR(-ENOENT); |
1bbc5513 | 3624 | |
444ceed8 EB |
3625 | inode->i_op = &proc_tid_base_inode_operations; |
3626 | inode->i_fop = &proc_tid_base_operations; | |
1bbc5513 | 3627 | inode->i_flags |= S_IMMUTABLE; |
aed54175 | 3628 | |
1270dd8d | 3629 | set_nlink(inode, nlink_tid); |
1bbc5513 | 3630 | pid_update_inode(task, inode); |
444ceed8 | 3631 | |
fb045adb | 3632 | d_set_d_op(dentry, &pid_dentry_operations); |
0168b9e3 | 3633 | return d_splice_alias(inode, dentry); |
444ceed8 EB |
3634 | } |
3635 | ||
00cd8dd3 | 3636 | static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags) |
28a6d671 | 3637 | { |
28a6d671 EB |
3638 | struct task_struct *task; |
3639 | struct task_struct *leader = get_proc_task(dir); | |
28a6d671 | 3640 | unsigned tid; |
fa10fed3 | 3641 | struct proc_fs_info *fs_info; |
b488893a | 3642 | struct pid_namespace *ns; |
0168b9e3 | 3643 | struct dentry *result = ERR_PTR(-ENOENT); |
28a6d671 EB |
3644 | |
3645 | if (!leader) | |
3646 | goto out_no_task; | |
3647 | ||
dbcdb504 | 3648 | tid = name_to_int(&dentry->d_name); |
28a6d671 EB |
3649 | if (tid == ~0U) |
3650 | goto out; | |
3651 | ||
fa10fed3 AG |
3652 | fs_info = proc_sb_info(dentry->d_sb); |
3653 | ns = fs_info->pid_ns; | |
28a6d671 | 3654 | rcu_read_lock(); |
b488893a | 3655 | task = find_task_by_pid_ns(tid, ns); |
28a6d671 EB |
3656 | if (task) |
3657 | get_task_struct(task); | |
3658 | rcu_read_unlock(); | |
3659 | if (!task) | |
3660 | goto out; | |
bac0abd6 | 3661 | if (!same_thread_group(leader, task)) |
28a6d671 EB |
3662 | goto out_drop_task; |
3663 | ||
0168b9e3 | 3664 | result = proc_task_instantiate(dentry, task, NULL); |
28a6d671 EB |
3665 | out_drop_task: |
3666 | put_task_struct(task); | |
3667 | out: | |
3668 | put_task_struct(leader); | |
3669 | out_no_task: | |
0168b9e3 | 3670 | return result; |
28a6d671 EB |
3671 | } |
3672 | ||
0bc58a91 EB |
3673 | /* |
3674 | * Find the first tid of a thread group to return to user space. | |
3675 | * | |
3676 | * Usually this is just the thread group leader, but if the users | |
3677 | * buffer was too small or there was a seek into the middle of the | |
3678 | * directory we have more work todo. | |
3679 | * | |
3680 | * In the case of a short read we start with find_task_by_pid. | |
3681 | * | |
3682 | * In the case of a seek we start with the leader and walk nr | |
3683 | * threads past it. | |
3684 | */ | |
9f6e963f ON |
3685 | static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos, |
3686 | struct pid_namespace *ns) | |
0bc58a91 | 3687 | { |
d855a4b7 | 3688 | struct task_struct *pos, *task; |
9f6e963f ON |
3689 | unsigned long nr = f_pos; |
3690 | ||
3691 | if (nr != f_pos) /* 32bit overflow? */ | |
3692 | return NULL; | |
1da177e4 | 3693 | |
cc288738 | 3694 | rcu_read_lock(); |
d855a4b7 ON |
3695 | task = pid_task(pid, PIDTYPE_PID); |
3696 | if (!task) | |
3697 | goto fail; | |
3698 | ||
3699 | /* Attempt to start with the tid of a thread */ | |
9f6e963f | 3700 | if (tid && nr) { |
b488893a | 3701 | pos = find_task_by_pid_ns(tid, ns); |
d855a4b7 | 3702 | if (pos && same_thread_group(pos, task)) |
a872ff0c | 3703 | goto found; |
0bc58a91 | 3704 | } |
1da177e4 | 3705 | |
0bc58a91 | 3706 | /* If nr exceeds the number of threads there is nothing todo */ |
9f6e963f | 3707 | if (nr >= get_nr_threads(task)) |
c986c14a | 3708 | goto fail; |
1da177e4 | 3709 | |
a872ff0c ON |
3710 | /* If we haven't found our starting place yet start |
3711 | * with the leader and walk nr threads forward. | |
0bc58a91 | 3712 | */ |
d855a4b7 | 3713 | pos = task = task->group_leader; |
c986c14a | 3714 | do { |
9f6e963f | 3715 | if (!nr--) |
c986c14a | 3716 | goto found; |
d855a4b7 | 3717 | } while_each_thread(task, pos); |
c986c14a ON |
3718 | fail: |
3719 | pos = NULL; | |
3720 | goto out; | |
a872ff0c ON |
3721 | found: |
3722 | get_task_struct(pos); | |
3723 | out: | |
cc288738 | 3724 | rcu_read_unlock(); |
0bc58a91 EB |
3725 | return pos; |
3726 | } | |
3727 | ||
3728 | /* | |
3729 | * Find the next thread in the thread list. | |
3730 | * Return NULL if there is an error or no next thread. | |
3731 | * | |
3732 | * The reference to the input task_struct is released. | |
3733 | */ | |
3734 | static struct task_struct *next_tid(struct task_struct *start) | |
3735 | { | |
c1df7fb8 | 3736 | struct task_struct *pos = NULL; |
cc288738 | 3737 | rcu_read_lock(); |
c1df7fb8 | 3738 | if (pid_alive(start)) { |
0bc58a91 | 3739 | pos = next_thread(start); |
c1df7fb8 ON |
3740 | if (thread_group_leader(pos)) |
3741 | pos = NULL; | |
3742 | else | |
3743 | get_task_struct(pos); | |
3744 | } | |
cc288738 | 3745 | rcu_read_unlock(); |
0bc58a91 EB |
3746 | put_task_struct(start); |
3747 | return pos; | |
1da177e4 LT |
3748 | } |
3749 | ||
3750 | /* for the /proc/TGID/task/ directories */ | |
f0c3b509 | 3751 | static int proc_task_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3752 | { |
d855a4b7 ON |
3753 | struct inode *inode = file_inode(file); |
3754 | struct task_struct *task; | |
b488893a | 3755 | struct pid_namespace *ns; |
f0c3b509 | 3756 | int tid; |
1da177e4 | 3757 | |
d855a4b7 | 3758 | if (proc_inode_is_dead(inode)) |
f0c3b509 | 3759 | return -ENOENT; |
1da177e4 | 3760 | |
f0c3b509 | 3761 | if (!dir_emit_dots(file, ctx)) |
d855a4b7 | 3762 | return 0; |
1da177e4 | 3763 | |
0bc58a91 EB |
3764 | /* f_version caches the tgid value that the last readdir call couldn't |
3765 | * return. lseek aka telldir automagically resets f_version to 0. | |
3766 | */ | |
9d78edea | 3767 | ns = proc_pid_ns(inode->i_sb); |
f0c3b509 AV |
3768 | tid = (int)file->f_version; |
3769 | file->f_version = 0; | |
d855a4b7 | 3770 | for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns); |
0bc58a91 | 3771 | task; |
f0c3b509 | 3772 | task = next_tid(task), ctx->pos++) { |
e3912ac3 | 3773 | char name[10 + 1]; |
a4ef3895 | 3774 | unsigned int len; |
b488893a | 3775 | tid = task_pid_nr_ns(task, ns); |
e3912ac3 | 3776 | len = snprintf(name, sizeof(name), "%u", tid); |
f0c3b509 AV |
3777 | if (!proc_fill_cache(file, ctx, name, len, |
3778 | proc_task_instantiate, task, NULL)) { | |
0bc58a91 EB |
3779 | /* returning this tgid failed, save it as the first |
3780 | * pid for the next readir call */ | |
f0c3b509 | 3781 | file->f_version = (u64)tid; |
0bc58a91 | 3782 | put_task_struct(task); |
1da177e4 | 3783 | break; |
0bc58a91 | 3784 | } |
1da177e4 | 3785 | } |
d855a4b7 | 3786 | |
f0c3b509 | 3787 | return 0; |
1da177e4 | 3788 | } |
6e66b52b | 3789 | |
a528d35e DH |
3790 | static int proc_task_getattr(const struct path *path, struct kstat *stat, |
3791 | u32 request_mask, unsigned int query_flags) | |
6e66b52b | 3792 | { |
a528d35e | 3793 | struct inode *inode = d_inode(path->dentry); |
99f89551 | 3794 | struct task_struct *p = get_proc_task(inode); |
6e66b52b EB |
3795 | generic_fillattr(inode, stat); |
3796 | ||
99f89551 | 3797 | if (p) { |
99f89551 | 3798 | stat->nlink += get_nr_threads(p); |
99f89551 | 3799 | put_task_struct(p); |
6e66b52b EB |
3800 | } |
3801 | ||
3802 | return 0; | |
3803 | } | |
28a6d671 | 3804 | |
c5ef1c42 | 3805 | static const struct inode_operations proc_task_inode_operations = { |
28a6d671 EB |
3806 | .lookup = proc_task_lookup, |
3807 | .getattr = proc_task_getattr, | |
3808 | .setattr = proc_setattr, | |
0499680a | 3809 | .permission = proc_pid_permission, |
28a6d671 EB |
3810 | }; |
3811 | ||
00977a59 | 3812 | static const struct file_operations proc_task_operations = { |
28a6d671 | 3813 | .read = generic_read_dir, |
f50752ea AV |
3814 | .iterate_shared = proc_task_readdir, |
3815 | .llseek = generic_file_llseek, | |
28a6d671 | 3816 | }; |
1270dd8d AD |
3817 | |
3818 | void __init set_proc_pid_nlink(void) | |
3819 | { | |
3820 | nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
3821 | nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
3822 | } |