]>
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> |
43d2b113 | 97 | #include <trace/events/oom.h> |
1da177e4 | 98 | #include "internal.h" |
faf60af1 | 99 | #include "fd.h" |
1da177e4 | 100 | |
ac7f1061 AD |
101 | #include "../../lib/kstrtox.h" |
102 | ||
0f2fe20f EB |
103 | /* NOTE: |
104 | * Implementing inode permission operations in /proc is almost | |
105 | * certainly an error. Permission checks need to happen during | |
106 | * each system call not at open time. The reason is that most of | |
107 | * what we wish to check for permissions in /proc varies at runtime. | |
108 | * | |
109 | * The classic example of a problem is opening file descriptors | |
110 | * in /proc for a task before it execs a suid executable. | |
111 | */ | |
112 | ||
efb1a57d AD |
113 | static u8 nlink_tid __ro_after_init; |
114 | static u8 nlink_tgid __ro_after_init; | |
1270dd8d | 115 | |
1da177e4 | 116 | struct pid_entry { |
cedbccab | 117 | const char *name; |
623f594e | 118 | unsigned int len; |
d161a13f | 119 | umode_t mode; |
c5ef1c42 | 120 | const struct inode_operations *iop; |
00977a59 | 121 | const struct file_operations *fop; |
20cdc894 | 122 | union proc_op op; |
1da177e4 LT |
123 | }; |
124 | ||
61a28784 | 125 | #define NOD(NAME, MODE, IOP, FOP, OP) { \ |
20cdc894 | 126 | .name = (NAME), \ |
c5141e6d | 127 | .len = sizeof(NAME) - 1, \ |
20cdc894 EB |
128 | .mode = MODE, \ |
129 | .iop = IOP, \ | |
130 | .fop = FOP, \ | |
131 | .op = OP, \ | |
132 | } | |
133 | ||
631f9c18 AD |
134 | #define DIR(NAME, MODE, iops, fops) \ |
135 | NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} ) | |
136 | #define LNK(NAME, get_link) \ | |
61a28784 | 137 | NOD(NAME, (S_IFLNK|S_IRWXUGO), \ |
20cdc894 | 138 | &proc_pid_link_inode_operations, NULL, \ |
631f9c18 AD |
139 | { .proc_get_link = get_link } ) |
140 | #define REG(NAME, MODE, fops) \ | |
141 | NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {}) | |
631f9c18 | 142 | #define ONE(NAME, MODE, show) \ |
6d9c939d | 143 | NOD(NAME, (S_IFREG|(MODE)), \ |
be614086 | 144 | NULL, &proc_single_file_operations, \ |
631f9c18 | 145 | { .proc_show = show } ) |
6d9c939d CS |
146 | #define ATTR(LSM, NAME, MODE) \ |
147 | NOD(NAME, (S_IFREG|(MODE)), \ | |
148 | NULL, &proc_pid_attr_operations, \ | |
149 | { .lsm = LSM }) | |
1da177e4 | 150 | |
aed54175 VN |
151 | /* |
152 | * Count the number of hardlinks for the pid_entry table, excluding the . | |
153 | * and .. links. | |
154 | */ | |
1270dd8d | 155 | static unsigned int __init pid_entry_nlink(const struct pid_entry *entries, |
aed54175 VN |
156 | unsigned int n) |
157 | { | |
158 | unsigned int i; | |
159 | unsigned int count; | |
160 | ||
1270dd8d | 161 | count = 2; |
aed54175 VN |
162 | for (i = 0; i < n; ++i) { |
163 | if (S_ISDIR(entries[i].mode)) | |
164 | ++count; | |
165 | } | |
166 | ||
167 | return count; | |
168 | } | |
169 | ||
f7ad3c6b | 170 | static int get_task_root(struct task_struct *task, struct path *root) |
1da177e4 | 171 | { |
7c2c7d99 HD |
172 | int result = -ENOENT; |
173 | ||
0494f6ec | 174 | task_lock(task); |
f7ad3c6b MS |
175 | if (task->fs) { |
176 | get_fs_root(task->fs, root); | |
7c2c7d99 HD |
177 | result = 0; |
178 | } | |
0494f6ec | 179 | task_unlock(task); |
7c2c7d99 | 180 | return result; |
0494f6ec MS |
181 | } |
182 | ||
7773fbc5 | 183 | static int proc_cwd_link(struct dentry *dentry, struct path *path) |
0494f6ec | 184 | { |
2b0143b5 | 185 | struct task_struct *task = get_proc_task(d_inode(dentry)); |
0494f6ec | 186 | int result = -ENOENT; |
99f89551 EB |
187 | |
188 | if (task) { | |
f7ad3c6b MS |
189 | task_lock(task); |
190 | if (task->fs) { | |
191 | get_fs_pwd(task->fs, path); | |
192 | result = 0; | |
193 | } | |
194 | task_unlock(task); | |
99f89551 EB |
195 | put_task_struct(task); |
196 | } | |
1da177e4 LT |
197 | return result; |
198 | } | |
199 | ||
7773fbc5 | 200 | static int proc_root_link(struct dentry *dentry, struct path *path) |
1da177e4 | 201 | { |
2b0143b5 | 202 | struct task_struct *task = get_proc_task(d_inode(dentry)); |
1da177e4 | 203 | int result = -ENOENT; |
99f89551 EB |
204 | |
205 | if (task) { | |
f7ad3c6b | 206 | result = get_task_root(task, path); |
99f89551 EB |
207 | put_task_struct(task); |
208 | } | |
1da177e4 LT |
209 | return result; |
210 | } | |
211 | ||
d26d0cd9 LT |
212 | /* |
213 | * If the user used setproctitle(), we just get the string from | |
214 | * user space at arg_start, and limit it to a maximum of one page. | |
215 | */ | |
216 | static ssize_t get_mm_proctitle(struct mm_struct *mm, char __user *buf, | |
217 | size_t count, unsigned long pos, | |
218 | unsigned long arg_start) | |
219 | { | |
220 | char *page; | |
221 | int ret, got; | |
222 | ||
223 | if (pos >= PAGE_SIZE) | |
224 | return 0; | |
225 | ||
226 | page = (char *)__get_free_page(GFP_KERNEL); | |
227 | if (!page) | |
228 | return -ENOMEM; | |
229 | ||
230 | ret = 0; | |
231 | got = access_remote_vm(mm, arg_start, page, PAGE_SIZE, FOLL_ANON); | |
232 | if (got > 0) { | |
233 | int len = strnlen(page, got); | |
234 | ||
235 | /* Include the NUL character if it was found */ | |
236 | if (len < got) | |
237 | len++; | |
238 | ||
239 | if (len > pos) { | |
240 | len -= pos; | |
241 | if (len > count) | |
242 | len = count; | |
243 | len -= copy_to_user(buf, page+pos, len); | |
244 | if (!len) | |
245 | len = -EFAULT; | |
246 | ret = len; | |
247 | } | |
248 | } | |
249 | free_page((unsigned long)page); | |
250 | return ret; | |
251 | } | |
252 | ||
e4b4e441 | 253 | static ssize_t get_mm_cmdline(struct mm_struct *mm, char __user *buf, |
5ab82718 | 254 | size_t count, loff_t *ppos) |
1da177e4 | 255 | { |
c2c0bb44 | 256 | unsigned long arg_start, arg_end, env_start, env_end; |
5ab82718 | 257 | unsigned long pos, len; |
d26d0cd9 | 258 | char *page, c; |
c2c0bb44 | 259 | |
c2c0bb44 | 260 | /* Check if process spawned far enough to have cmdline. */ |
e4b4e441 LT |
261 | if (!mm->env_end) |
262 | return 0; | |
c2c0bb44 | 263 | |
88aa7cc6 | 264 | spin_lock(&mm->arg_lock); |
c2c0bb44 AD |
265 | arg_start = mm->arg_start; |
266 | arg_end = mm->arg_end; | |
267 | env_start = mm->env_start; | |
268 | env_end = mm->env_end; | |
88aa7cc6 | 269 | spin_unlock(&mm->arg_lock); |
c2c0bb44 | 270 | |
5ab82718 LT |
271 | if (arg_start >= arg_end) |
272 | return 0; | |
6a6cbe75 | 273 | |
2ca66ff7 | 274 | /* |
d26d0cd9 LT |
275 | * We allow setproctitle() to overwrite the argument |
276 | * strings, and overflow past the original end. But | |
277 | * only when it overflows into the environment area. | |
2ca66ff7 | 278 | */ |
d26d0cd9 | 279 | if (env_start != arg_end || env_end < env_start) |
5ab82718 | 280 | env_start = env_end = arg_end; |
d26d0cd9 | 281 | len = env_end - arg_start; |
f5b65348 | 282 | |
5ab82718 | 283 | /* We're not going to care if "*ppos" has high bits set */ |
d26d0cd9 LT |
284 | pos = *ppos; |
285 | if (pos >= len) | |
5ab82718 | 286 | return 0; |
d26d0cd9 LT |
287 | if (count > len - pos) |
288 | count = len - pos; | |
289 | if (!count) | |
290 | return 0; | |
291 | ||
292 | /* | |
293 | * Magical special case: if the argv[] end byte is not | |
294 | * zero, the user has overwritten it with setproctitle(3). | |
295 | * | |
296 | * Possible future enhancement: do this only once when | |
297 | * pos is 0, and set a flag in the 'struct file'. | |
298 | */ | |
299 | if (access_remote_vm(mm, arg_end-1, &c, 1, FOLL_ANON) == 1 && c) | |
300 | return get_mm_proctitle(mm, buf, count, pos, arg_start); | |
a0a07b87 | 301 | |
d26d0cd9 LT |
302 | /* |
303 | * For the non-setproctitle() case we limit things strictly | |
304 | * to the [arg_start, arg_end[ range. | |
305 | */ | |
306 | pos += arg_start; | |
3d712546 | 307 | if (pos < arg_start || pos >= arg_end) |
5ab82718 | 308 | return 0; |
3d712546 LT |
309 | if (count > arg_end - pos) |
310 | count = arg_end - pos; | |
5ab82718 LT |
311 | |
312 | page = (char *)__get_free_page(GFP_KERNEL); | |
313 | if (!page) | |
314 | return -ENOMEM; | |
315 | ||
316 | len = 0; | |
317 | while (count) { | |
318 | int got; | |
319 | size_t size = min_t(size_t, PAGE_SIZE, count); | |
320 | ||
3d712546 LT |
321 | got = access_remote_vm(mm, pos, page, size, FOLL_ANON); |
322 | if (got <= 0) | |
5ab82718 | 323 | break; |
3d712546 | 324 | got -= copy_to_user(buf, page, got); |
5ab82718 LT |
325 | if (unlikely(!got)) { |
326 | if (!len) | |
327 | len = -EFAULT; | |
328 | break; | |
c2c0bb44 | 329 | } |
5ab82718 LT |
330 | pos += got; |
331 | buf += got; | |
332 | len += got; | |
333 | count -= got; | |
c2c0bb44 AD |
334 | } |
335 | ||
c2c0bb44 | 336 | free_page((unsigned long)page); |
5ab82718 | 337 | return len; |
1da177e4 LT |
338 | } |
339 | ||
e4b4e441 LT |
340 | static ssize_t get_task_cmdline(struct task_struct *tsk, char __user *buf, |
341 | size_t count, loff_t *pos) | |
342 | { | |
343 | struct mm_struct *mm; | |
344 | ssize_t ret; | |
345 | ||
346 | mm = get_task_mm(tsk); | |
347 | if (!mm) | |
348 | return 0; | |
349 | ||
350 | ret = get_mm_cmdline(mm, buf, count, pos); | |
c2c0bb44 | 351 | mmput(mm); |
e4b4e441 LT |
352 | return ret; |
353 | } | |
354 | ||
355 | static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf, | |
356 | size_t count, loff_t *pos) | |
357 | { | |
358 | struct task_struct *tsk; | |
359 | ssize_t ret; | |
360 | ||
361 | BUG_ON(*pos < 0); | |
362 | ||
363 | tsk = get_proc_task(file_inode(file)); | |
364 | if (!tsk) | |
365 | return -ESRCH; | |
366 | ret = get_task_cmdline(tsk, buf, count, pos); | |
367 | put_task_struct(tsk); | |
368 | if (ret > 0) | |
369 | *pos += ret; | |
370 | return ret; | |
1da177e4 LT |
371 | } |
372 | ||
c2c0bb44 AD |
373 | static const struct file_operations proc_pid_cmdline_ops = { |
374 | .read = proc_pid_cmdline_read, | |
375 | .llseek = generic_file_llseek, | |
376 | }; | |
377 | ||
1da177e4 LT |
378 | #ifdef CONFIG_KALLSYMS |
379 | /* | |
380 | * Provides a wchan file via kallsyms in a proper one-value-per-file format. | |
381 | * Returns the resolved symbol. If that fails, simply return the address. | |
382 | */ | |
edfcd606 AD |
383 | static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns, |
384 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 385 | { |
ffb45122 | 386 | unsigned long wchan; |
9281acea | 387 | char symname[KSYM_NAME_LEN]; |
1da177e4 | 388 | |
24b2ec21 AD |
389 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
390 | goto print0; | |
1da177e4 | 391 | |
24b2ec21 AD |
392 | wchan = get_wchan(task); |
393 | if (wchan && !lookup_symbol_name(wchan, symname)) { | |
21dae0ad | 394 | seq_puts(m, symname); |
24b2ec21 AD |
395 | return 0; |
396 | } | |
25ce3191 | 397 | |
24b2ec21 AD |
398 | print0: |
399 | seq_putc(m, '0'); | |
25ce3191 | 400 | return 0; |
1da177e4 LT |
401 | } |
402 | #endif /* CONFIG_KALLSYMS */ | |
403 | ||
a9712bc1 AV |
404 | static int lock_trace(struct task_struct *task) |
405 | { | |
406 | int err = mutex_lock_killable(&task->signal->cred_guard_mutex); | |
407 | if (err) | |
408 | return err; | |
caaee623 | 409 | if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) { |
a9712bc1 AV |
410 | mutex_unlock(&task->signal->cred_guard_mutex); |
411 | return -EPERM; | |
412 | } | |
413 | return 0; | |
414 | } | |
415 | ||
416 | static void unlock_trace(struct task_struct *task) | |
417 | { | |
418 | mutex_unlock(&task->signal->cred_guard_mutex); | |
419 | } | |
420 | ||
2ec220e2 KC |
421 | #ifdef CONFIG_STACKTRACE |
422 | ||
423 | #define MAX_STACK_TRACE_DEPTH 64 | |
424 | ||
425 | static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns, | |
426 | struct pid *pid, struct task_struct *task) | |
427 | { | |
2ec220e2 | 428 | unsigned long *entries; |
a9712bc1 | 429 | int err; |
2ec220e2 | 430 | |
f8a00cef JH |
431 | /* |
432 | * The ability to racily run the kernel stack unwinder on a running task | |
433 | * and then observe the unwinder output is scary; while it is useful for | |
434 | * debugging kernel issues, it can also allow an attacker to leak kernel | |
435 | * stack contents. | |
436 | * Doing this in a manner that is at least safe from races would require | |
437 | * some work to ensure that the remote task can not be scheduled; and | |
438 | * even then, this would still expose the unwinder as local attack | |
439 | * surface. | |
440 | * Therefore, this interface is restricted to root. | |
441 | */ | |
442 | if (!file_ns_capable(m->file, &init_user_ns, CAP_SYS_ADMIN)) | |
443 | return -EACCES; | |
444 | ||
6da2ec56 KC |
445 | entries = kmalloc_array(MAX_STACK_TRACE_DEPTH, sizeof(*entries), |
446 | GFP_KERNEL); | |
2ec220e2 KC |
447 | if (!entries) |
448 | return -ENOMEM; | |
449 | ||
a9712bc1 AV |
450 | err = lock_trace(task); |
451 | if (!err) { | |
e988e5ec | 452 | unsigned int i, nr_entries; |
5d008fb4 | 453 | |
e988e5ec TG |
454 | nr_entries = stack_trace_save_tsk(task, entries, |
455 | MAX_STACK_TRACE_DEPTH, 0); | |
a9712bc1 | 456 | |
e988e5ec | 457 | for (i = 0; i < nr_entries; i++) { |
8f5abe84 | 458 | seq_printf(m, "[<0>] %pB\n", (void *)entries[i]); |
a9712bc1 | 459 | } |
e988e5ec | 460 | |
a9712bc1 | 461 | unlock_trace(task); |
2ec220e2 KC |
462 | } |
463 | kfree(entries); | |
464 | ||
a9712bc1 | 465 | return err; |
2ec220e2 KC |
466 | } |
467 | #endif | |
468 | ||
5968cece | 469 | #ifdef CONFIG_SCHED_INFO |
1da177e4 LT |
470 | /* |
471 | * Provides /proc/PID/schedstat | |
472 | */ | |
f6e826ca AD |
473 | static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns, |
474 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 475 | { |
5968cece | 476 | if (unlikely(!sched_info_on())) |
08b55775 | 477 | seq_puts(m, "0 0 0\n"); |
5968cece NR |
478 | else |
479 | seq_printf(m, "%llu %llu %lu\n", | |
25ce3191 JP |
480 | (unsigned long long)task->se.sum_exec_runtime, |
481 | (unsigned long long)task->sched_info.run_delay, | |
482 | task->sched_info.pcount); | |
483 | ||
484 | return 0; | |
1da177e4 LT |
485 | } |
486 | #endif | |
487 | ||
9745512c AV |
488 | #ifdef CONFIG_LATENCYTOP |
489 | static int lstats_show_proc(struct seq_file *m, void *v) | |
490 | { | |
491 | int i; | |
13d77c37 HS |
492 | struct inode *inode = m->private; |
493 | struct task_struct *task = get_proc_task(inode); | |
9745512c | 494 | |
13d77c37 HS |
495 | if (!task) |
496 | return -ESRCH; | |
497 | seq_puts(m, "Latency Top version : v0.1\n"); | |
f6d2f584 | 498 | for (i = 0; i < LT_SAVECOUNT; i++) { |
34e49d4f JP |
499 | struct latency_record *lr = &task->latency_record[i]; |
500 | if (lr->backtrace[0]) { | |
9745512c | 501 | int q; |
34e49d4f JP |
502 | seq_printf(m, "%i %li %li", |
503 | lr->count, lr->time, lr->max); | |
9745512c | 504 | for (q = 0; q < LT_BACKTRACEDEPTH; q++) { |
34e49d4f | 505 | unsigned long bt = lr->backtrace[q]; |
accddc41 | 506 | |
34e49d4f | 507 | if (!bt) |
9745512c | 508 | break; |
34e49d4f | 509 | seq_printf(m, " %ps", (void *)bt); |
9745512c | 510 | } |
9d6de12f | 511 | seq_putc(m, '\n'); |
9745512c AV |
512 | } |
513 | ||
514 | } | |
13d77c37 | 515 | put_task_struct(task); |
9745512c AV |
516 | return 0; |
517 | } | |
518 | ||
519 | static int lstats_open(struct inode *inode, struct file *file) | |
520 | { | |
13d77c37 | 521 | return single_open(file, lstats_show_proc, inode); |
d6643d12 HS |
522 | } |
523 | ||
9745512c AV |
524 | static ssize_t lstats_write(struct file *file, const char __user *buf, |
525 | size_t count, loff_t *offs) | |
526 | { | |
496ad9aa | 527 | struct task_struct *task = get_proc_task(file_inode(file)); |
9745512c | 528 | |
13d77c37 HS |
529 | if (!task) |
530 | return -ESRCH; | |
e02c9b0d | 531 | clear_tsk_latency_tracing(task); |
13d77c37 | 532 | put_task_struct(task); |
9745512c AV |
533 | |
534 | return count; | |
535 | } | |
536 | ||
537 | static const struct file_operations proc_lstats_operations = { | |
538 | .open = lstats_open, | |
539 | .read = seq_read, | |
540 | .write = lstats_write, | |
541 | .llseek = seq_lseek, | |
13d77c37 | 542 | .release = single_release, |
9745512c AV |
543 | }; |
544 | ||
545 | #endif | |
546 | ||
6ba51e37 AD |
547 | static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns, |
548 | struct pid *pid, struct task_struct *task) | |
1da177e4 | 549 | { |
ca79b0c2 | 550 | unsigned long totalpages = totalram_pages() + total_swap_pages; |
b95c35e7 | 551 | unsigned long points = 0; |
1da177e4 | 552 | |
ac311a14 | 553 | points = oom_badness(task, totalpages) * 1000 / totalpages; |
25ce3191 JP |
554 | seq_printf(m, "%lu\n", points); |
555 | ||
556 | return 0; | |
1da177e4 LT |
557 | } |
558 | ||
d85f50d5 | 559 | struct limit_names { |
cedbccab AD |
560 | const char *name; |
561 | const char *unit; | |
d85f50d5 NH |
562 | }; |
563 | ||
564 | static const struct limit_names lnames[RLIM_NLIMITS] = { | |
cff4edb5 | 565 | [RLIMIT_CPU] = {"Max cpu time", "seconds"}, |
d85f50d5 NH |
566 | [RLIMIT_FSIZE] = {"Max file size", "bytes"}, |
567 | [RLIMIT_DATA] = {"Max data size", "bytes"}, | |
568 | [RLIMIT_STACK] = {"Max stack size", "bytes"}, | |
569 | [RLIMIT_CORE] = {"Max core file size", "bytes"}, | |
570 | [RLIMIT_RSS] = {"Max resident set", "bytes"}, | |
571 | [RLIMIT_NPROC] = {"Max processes", "processes"}, | |
572 | [RLIMIT_NOFILE] = {"Max open files", "files"}, | |
573 | [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"}, | |
574 | [RLIMIT_AS] = {"Max address space", "bytes"}, | |
575 | [RLIMIT_LOCKS] = {"Max file locks", "locks"}, | |
576 | [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"}, | |
577 | [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"}, | |
578 | [RLIMIT_NICE] = {"Max nice priority", NULL}, | |
579 | [RLIMIT_RTPRIO] = {"Max realtime priority", NULL}, | |
8808117c | 580 | [RLIMIT_RTTIME] = {"Max realtime timeout", "us"}, |
d85f50d5 NH |
581 | }; |
582 | ||
583 | /* Display limits for a process */ | |
1c963eb1 AD |
584 | static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns, |
585 | struct pid *pid, struct task_struct *task) | |
d85f50d5 NH |
586 | { |
587 | unsigned int i; | |
d85f50d5 | 588 | unsigned long flags; |
d85f50d5 NH |
589 | |
590 | struct rlimit rlim[RLIM_NLIMITS]; | |
591 | ||
a6bebbc8 | 592 | if (!lock_task_sighand(task, &flags)) |
d85f50d5 | 593 | return 0; |
d85f50d5 NH |
594 | memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS); |
595 | unlock_task_sighand(task, &flags); | |
d85f50d5 NH |
596 | |
597 | /* | |
598 | * print the file header | |
599 | */ | |
afe922c2 AD |
600 | seq_puts(m, "Limit " |
601 | "Soft Limit " | |
602 | "Hard Limit " | |
603 | "Units \n"); | |
d85f50d5 NH |
604 | |
605 | for (i = 0; i < RLIM_NLIMITS; i++) { | |
606 | if (rlim[i].rlim_cur == RLIM_INFINITY) | |
1c963eb1 | 607 | seq_printf(m, "%-25s %-20s ", |
25ce3191 | 608 | lnames[i].name, "unlimited"); |
d85f50d5 | 609 | else |
1c963eb1 | 610 | seq_printf(m, "%-25s %-20lu ", |
25ce3191 | 611 | lnames[i].name, rlim[i].rlim_cur); |
d85f50d5 NH |
612 | |
613 | if (rlim[i].rlim_max == RLIM_INFINITY) | |
1c963eb1 | 614 | seq_printf(m, "%-20s ", "unlimited"); |
d85f50d5 | 615 | else |
1c963eb1 | 616 | seq_printf(m, "%-20lu ", rlim[i].rlim_max); |
d85f50d5 NH |
617 | |
618 | if (lnames[i].unit) | |
1c963eb1 | 619 | seq_printf(m, "%-10s\n", lnames[i].unit); |
d85f50d5 | 620 | else |
1c963eb1 | 621 | seq_putc(m, '\n'); |
d85f50d5 NH |
622 | } |
623 | ||
1c963eb1 | 624 | return 0; |
d85f50d5 NH |
625 | } |
626 | ||
ebcb6734 | 627 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 AD |
628 | static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns, |
629 | struct pid *pid, struct task_struct *task) | |
ebcb6734 | 630 | { |
631b7aba SRRH |
631 | struct syscall_info info; |
632 | u64 *args = &info.data.args[0]; | |
25ce3191 JP |
633 | int res; |
634 | ||
635 | res = lock_trace(task); | |
a9712bc1 AV |
636 | if (res) |
637 | return res; | |
ebcb6734 | 638 | |
631b7aba | 639 | if (task_current_syscall(task, &info)) |
09d93bd6 | 640 | seq_puts(m, "running\n"); |
631b7aba SRRH |
641 | else if (info.data.nr < 0) |
642 | seq_printf(m, "%d 0x%llx 0x%llx\n", | |
643 | info.data.nr, info.sp, info.data.instruction_pointer); | |
a9712bc1 | 644 | else |
09d93bd6 | 645 | seq_printf(m, |
631b7aba SRRH |
646 | "%d 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx\n", |
647 | info.data.nr, | |
ebcb6734 | 648 | args[0], args[1], args[2], args[3], args[4], args[5], |
631b7aba | 649 | info.sp, info.data.instruction_pointer); |
a9712bc1 | 650 | unlock_trace(task); |
25ce3191 JP |
651 | |
652 | return 0; | |
ebcb6734 RM |
653 | } |
654 | #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */ | |
655 | ||
1da177e4 LT |
656 | /************************************************************************/ |
657 | /* Here the fs part begins */ | |
658 | /************************************************************************/ | |
659 | ||
660 | /* permission checks */ | |
778c1144 | 661 | static int proc_fd_access_allowed(struct inode *inode) |
1da177e4 | 662 | { |
778c1144 EB |
663 | struct task_struct *task; |
664 | int allowed = 0; | |
df26c40e EB |
665 | /* Allow access to a task's file descriptors if it is us or we |
666 | * may use ptrace attach to the process and find out that | |
667 | * information. | |
778c1144 EB |
668 | */ |
669 | task = get_proc_task(inode); | |
df26c40e | 670 | if (task) { |
caaee623 | 671 | allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS); |
778c1144 | 672 | put_task_struct(task); |
df26c40e | 673 | } |
778c1144 | 674 | return allowed; |
1da177e4 LT |
675 | } |
676 | ||
6b4e306a | 677 | int proc_setattr(struct dentry *dentry, struct iattr *attr) |
6d76fa58 LT |
678 | { |
679 | int error; | |
2b0143b5 | 680 | struct inode *inode = d_inode(dentry); |
6d76fa58 LT |
681 | |
682 | if (attr->ia_valid & ATTR_MODE) | |
683 | return -EPERM; | |
684 | ||
31051c85 | 685 | error = setattr_prepare(dentry, attr); |
1025774c CH |
686 | if (error) |
687 | return error; | |
688 | ||
1025774c CH |
689 | setattr_copy(inode, attr); |
690 | mark_inode_dirty(inode); | |
691 | return 0; | |
6d76fa58 LT |
692 | } |
693 | ||
0499680a VK |
694 | /* |
695 | * May current process learn task's sched/cmdline info (for hide_pid_min=1) | |
696 | * or euid/egid (for hide_pid_min=2)? | |
697 | */ | |
698 | static bool has_pid_permissions(struct pid_namespace *pid, | |
699 | struct task_struct *task, | |
700 | int hide_pid_min) | |
701 | { | |
702 | if (pid->hide_pid < hide_pid_min) | |
703 | return true; | |
704 | if (in_group_p(pid->pid_gid)) | |
705 | return true; | |
caaee623 | 706 | return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS); |
0499680a VK |
707 | } |
708 | ||
709 | ||
710 | static int proc_pid_permission(struct inode *inode, int mask) | |
711 | { | |
76f668be | 712 | struct pid_namespace *pid = proc_pid_ns(inode); |
0499680a VK |
713 | struct task_struct *task; |
714 | bool has_perms; | |
715 | ||
716 | task = get_proc_task(inode); | |
a2ef990a XF |
717 | if (!task) |
718 | return -ESRCH; | |
796f571b | 719 | has_perms = has_pid_permissions(pid, task, HIDEPID_NO_ACCESS); |
0499680a VK |
720 | put_task_struct(task); |
721 | ||
722 | if (!has_perms) { | |
796f571b | 723 | if (pid->hide_pid == HIDEPID_INVISIBLE) { |
0499680a VK |
724 | /* |
725 | * Let's make getdents(), stat(), and open() | |
726 | * consistent with each other. If a process | |
727 | * may not stat() a file, it shouldn't be seen | |
728 | * in procfs at all. | |
729 | */ | |
730 | return -ENOENT; | |
731 | } | |
732 | ||
733 | return -EPERM; | |
734 | } | |
735 | return generic_permission(inode, mask); | |
736 | } | |
737 | ||
738 | ||
739 | ||
c5ef1c42 | 740 | static const struct inode_operations proc_def_inode_operations = { |
6d76fa58 LT |
741 | .setattr = proc_setattr, |
742 | }; | |
743 | ||
be614086 EB |
744 | static int proc_single_show(struct seq_file *m, void *v) |
745 | { | |
746 | struct inode *inode = m->private; | |
76f668be CH |
747 | struct pid_namespace *ns = proc_pid_ns(inode); |
748 | struct pid *pid = proc_pid(inode); | |
be614086 EB |
749 | struct task_struct *task; |
750 | int ret; | |
751 | ||
be614086 EB |
752 | task = get_pid_task(pid, PIDTYPE_PID); |
753 | if (!task) | |
754 | return -ESRCH; | |
755 | ||
756 | ret = PROC_I(inode)->op.proc_show(m, ns, pid, task); | |
757 | ||
758 | put_task_struct(task); | |
759 | return ret; | |
760 | } | |
761 | ||
762 | static int proc_single_open(struct inode *inode, struct file *filp) | |
763 | { | |
c6a34058 | 764 | return single_open(filp, proc_single_show, inode); |
be614086 EB |
765 | } |
766 | ||
767 | static const struct file_operations proc_single_file_operations = { | |
768 | .open = proc_single_open, | |
769 | .read = seq_read, | |
770 | .llseek = seq_lseek, | |
771 | .release = single_release, | |
772 | }; | |
773 | ||
5381e169 ON |
774 | |
775 | struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode) | |
1da177e4 | 776 | { |
5381e169 ON |
777 | struct task_struct *task = get_proc_task(inode); |
778 | struct mm_struct *mm = ERR_PTR(-ESRCH); | |
e268337d | 779 | |
5381e169 | 780 | if (task) { |
caaee623 | 781 | mm = mm_access(task, mode | PTRACE_MODE_FSCREDS); |
5381e169 | 782 | put_task_struct(task); |
e268337d | 783 | |
5381e169 ON |
784 | if (!IS_ERR_OR_NULL(mm)) { |
785 | /* ensure this mm_struct can't be freed */ | |
f1f10076 | 786 | mmgrab(mm); |
5381e169 ON |
787 | /* but do not pin its memory */ |
788 | mmput(mm); | |
789 | } | |
790 | } | |
791 | ||
792 | return mm; | |
793 | } | |
794 | ||
795 | static int __mem_open(struct inode *inode, struct file *file, unsigned int mode) | |
796 | { | |
797 | struct mm_struct *mm = proc_mem_open(inode, mode); | |
e268337d LT |
798 | |
799 | if (IS_ERR(mm)) | |
800 | return PTR_ERR(mm); | |
801 | ||
e268337d | 802 | file->private_data = mm; |
1da177e4 LT |
803 | return 0; |
804 | } | |
805 | ||
b409e578 CW |
806 | static int mem_open(struct inode *inode, struct file *file) |
807 | { | |
bc452b4b DH |
808 | int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH); |
809 | ||
810 | /* OK to pass negative loff_t, we can catch out-of-range */ | |
811 | file->f_mode |= FMODE_UNSIGNED_OFFSET; | |
812 | ||
813 | return ret; | |
b409e578 CW |
814 | } |
815 | ||
572d34b9 ON |
816 | static ssize_t mem_rw(struct file *file, char __user *buf, |
817 | size_t count, loff_t *ppos, int write) | |
1da177e4 | 818 | { |
e268337d | 819 | struct mm_struct *mm = file->private_data; |
572d34b9 ON |
820 | unsigned long addr = *ppos; |
821 | ssize_t copied; | |
1da177e4 | 822 | char *page; |
272ddc8b | 823 | unsigned int flags; |
1da177e4 | 824 | |
e268337d LT |
825 | if (!mm) |
826 | return 0; | |
99f89551 | 827 | |
0ee931c4 | 828 | page = (char *)__get_free_page(GFP_KERNEL); |
30cd8903 | 829 | if (!page) |
e268337d | 830 | return -ENOMEM; |
1da177e4 | 831 | |
f7ca54f4 | 832 | copied = 0; |
388f7934 | 833 | if (!mmget_not_zero(mm)) |
6d08f2c7 ON |
834 | goto free; |
835 | ||
f511c0b1 | 836 | flags = FOLL_FORCE | (write ? FOLL_WRITE : 0); |
6347e8d5 | 837 | |
1da177e4 | 838 | while (count > 0) { |
572d34b9 | 839 | int this_len = min_t(int, count, PAGE_SIZE); |
1da177e4 | 840 | |
572d34b9 | 841 | if (write && copy_from_user(page, buf, this_len)) { |
1da177e4 LT |
842 | copied = -EFAULT; |
843 | break; | |
844 | } | |
572d34b9 | 845 | |
6347e8d5 | 846 | this_len = access_remote_vm(mm, addr, page, this_len, flags); |
572d34b9 | 847 | if (!this_len) { |
1da177e4 LT |
848 | if (!copied) |
849 | copied = -EIO; | |
850 | break; | |
851 | } | |
572d34b9 ON |
852 | |
853 | if (!write && copy_to_user(buf, page, this_len)) { | |
854 | copied = -EFAULT; | |
855 | break; | |
856 | } | |
857 | ||
858 | buf += this_len; | |
859 | addr += this_len; | |
860 | copied += this_len; | |
861 | count -= this_len; | |
1da177e4 | 862 | } |
572d34b9 | 863 | *ppos = addr; |
30cd8903 | 864 | |
6d08f2c7 ON |
865 | mmput(mm); |
866 | free: | |
30cd8903 | 867 | free_page((unsigned long) page); |
1da177e4 LT |
868 | return copied; |
869 | } | |
1da177e4 | 870 | |
572d34b9 ON |
871 | static ssize_t mem_read(struct file *file, char __user *buf, |
872 | size_t count, loff_t *ppos) | |
873 | { | |
874 | return mem_rw(file, buf, count, ppos, 0); | |
875 | } | |
876 | ||
877 | static ssize_t mem_write(struct file *file, const char __user *buf, | |
878 | size_t count, loff_t *ppos) | |
879 | { | |
880 | return mem_rw(file, (char __user*)buf, count, ppos, 1); | |
881 | } | |
882 | ||
85863e47 | 883 | loff_t mem_lseek(struct file *file, loff_t offset, int orig) |
1da177e4 LT |
884 | { |
885 | switch (orig) { | |
886 | case 0: | |
887 | file->f_pos = offset; | |
888 | break; | |
889 | case 1: | |
890 | file->f_pos += offset; | |
891 | break; | |
892 | default: | |
893 | return -EINVAL; | |
894 | } | |
895 | force_successful_syscall_return(); | |
896 | return file->f_pos; | |
897 | } | |
898 | ||
e268337d LT |
899 | static int mem_release(struct inode *inode, struct file *file) |
900 | { | |
901 | struct mm_struct *mm = file->private_data; | |
71879d3c | 902 | if (mm) |
6d08f2c7 | 903 | mmdrop(mm); |
e268337d LT |
904 | return 0; |
905 | } | |
906 | ||
00977a59 | 907 | static const struct file_operations proc_mem_operations = { |
1da177e4 LT |
908 | .llseek = mem_lseek, |
909 | .read = mem_read, | |
910 | .write = mem_write, | |
911 | .open = mem_open, | |
e268337d | 912 | .release = mem_release, |
1da177e4 LT |
913 | }; |
914 | ||
b409e578 CW |
915 | static int environ_open(struct inode *inode, struct file *file) |
916 | { | |
917 | return __mem_open(inode, file, PTRACE_MODE_READ); | |
918 | } | |
919 | ||
315e28c8 JP |
920 | static ssize_t environ_read(struct file *file, char __user *buf, |
921 | size_t count, loff_t *ppos) | |
922 | { | |
315e28c8 JP |
923 | char *page; |
924 | unsigned long src = *ppos; | |
b409e578 CW |
925 | int ret = 0; |
926 | struct mm_struct *mm = file->private_data; | |
a3b609ef | 927 | unsigned long env_start, env_end; |
315e28c8 | 928 | |
8148a73c MK |
929 | /* Ensure the process spawned far enough to have an environment. */ |
930 | if (!mm || !mm->env_end) | |
b409e578 | 931 | return 0; |
315e28c8 | 932 | |
0ee931c4 | 933 | page = (char *)__get_free_page(GFP_KERNEL); |
315e28c8 | 934 | if (!page) |
b409e578 | 935 | return -ENOMEM; |
315e28c8 | 936 | |
d6f64b89 | 937 | ret = 0; |
388f7934 | 938 | if (!mmget_not_zero(mm)) |
b409e578 | 939 | goto free; |
a3b609ef | 940 | |
88aa7cc6 | 941 | spin_lock(&mm->arg_lock); |
a3b609ef MG |
942 | env_start = mm->env_start; |
943 | env_end = mm->env_end; | |
88aa7cc6 | 944 | spin_unlock(&mm->arg_lock); |
a3b609ef | 945 | |
315e28c8 | 946 | while (count > 0) { |
e8905ec2 DH |
947 | size_t this_len, max_len; |
948 | int retval; | |
315e28c8 | 949 | |
a3b609ef | 950 | if (src >= (env_end - env_start)) |
315e28c8 JP |
951 | break; |
952 | ||
a3b609ef | 953 | this_len = env_end - (env_start + src); |
e8905ec2 DH |
954 | |
955 | max_len = min_t(size_t, PAGE_SIZE, count); | |
956 | this_len = min(max_len, this_len); | |
315e28c8 | 957 | |
7f7ccc2c | 958 | retval = access_remote_vm(mm, (env_start + src), page, this_len, FOLL_ANON); |
315e28c8 JP |
959 | |
960 | if (retval <= 0) { | |
961 | ret = retval; | |
962 | break; | |
963 | } | |
964 | ||
965 | if (copy_to_user(buf, page, retval)) { | |
966 | ret = -EFAULT; | |
967 | break; | |
968 | } | |
969 | ||
970 | ret += retval; | |
971 | src += retval; | |
972 | buf += retval; | |
973 | count -= retval; | |
974 | } | |
975 | *ppos = src; | |
315e28c8 | 976 | mmput(mm); |
b409e578 CW |
977 | |
978 | free: | |
315e28c8 | 979 | free_page((unsigned long) page); |
315e28c8 JP |
980 | return ret; |
981 | } | |
982 | ||
983 | static const struct file_operations proc_environ_operations = { | |
b409e578 | 984 | .open = environ_open, |
315e28c8 | 985 | .read = environ_read, |
87df8424 | 986 | .llseek = generic_file_llseek, |
b409e578 | 987 | .release = mem_release, |
315e28c8 JP |
988 | }; |
989 | ||
c5317167 AV |
990 | static int auxv_open(struct inode *inode, struct file *file) |
991 | { | |
992 | return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS); | |
993 | } | |
994 | ||
995 | static ssize_t auxv_read(struct file *file, char __user *buf, | |
996 | size_t count, loff_t *ppos) | |
997 | { | |
998 | struct mm_struct *mm = file->private_data; | |
999 | unsigned int nwords = 0; | |
06b2849d LY |
1000 | |
1001 | if (!mm) | |
1002 | return 0; | |
c5317167 AV |
1003 | do { |
1004 | nwords += 2; | |
1005 | } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */ | |
1006 | return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv, | |
1007 | nwords * sizeof(mm->saved_auxv[0])); | |
1008 | } | |
1009 | ||
1010 | static const struct file_operations proc_auxv_operations = { | |
1011 | .open = auxv_open, | |
1012 | .read = auxv_read, | |
1013 | .llseek = generic_file_llseek, | |
1014 | .release = mem_release, | |
1015 | }; | |
1016 | ||
fa0cbbf1 DR |
1017 | static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count, |
1018 | loff_t *ppos) | |
1019 | { | |
496ad9aa | 1020 | struct task_struct *task = get_proc_task(file_inode(file)); |
fa0cbbf1 DR |
1021 | char buffer[PROC_NUMBUF]; |
1022 | int oom_adj = OOM_ADJUST_MIN; | |
1023 | size_t len; | |
fa0cbbf1 DR |
1024 | |
1025 | if (!task) | |
1026 | return -ESRCH; | |
f913da59 MH |
1027 | if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX) |
1028 | oom_adj = OOM_ADJUST_MAX; | |
1029 | else | |
1030 | oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) / | |
1031 | OOM_SCORE_ADJ_MAX; | |
fa0cbbf1 DR |
1032 | put_task_struct(task); |
1033 | len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj); | |
1034 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
1035 | } | |
1036 | ||
1d5f0acb MH |
1037 | static int __set_oom_adj(struct file *file, int oom_adj, bool legacy) |
1038 | { | |
1039 | static DEFINE_MUTEX(oom_adj_mutex); | |
44a70ade | 1040 | struct mm_struct *mm = NULL; |
1d5f0acb MH |
1041 | struct task_struct *task; |
1042 | int err = 0; | |
1043 | ||
1044 | task = get_proc_task(file_inode(file)); | |
1045 | if (!task) | |
1046 | return -ESRCH; | |
1047 | ||
1048 | mutex_lock(&oom_adj_mutex); | |
1049 | if (legacy) { | |
1050 | if (oom_adj < task->signal->oom_score_adj && | |
1051 | !capable(CAP_SYS_RESOURCE)) { | |
1052 | err = -EACCES; | |
1053 | goto err_unlock; | |
1054 | } | |
1055 | /* | |
1056 | * /proc/pid/oom_adj is provided for legacy purposes, ask users to use | |
1057 | * /proc/pid/oom_score_adj instead. | |
1058 | */ | |
1059 | pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n", | |
1060 | current->comm, task_pid_nr(current), task_pid_nr(task), | |
1061 | task_pid_nr(task)); | |
1062 | } else { | |
1063 | if ((short)oom_adj < task->signal->oom_score_adj_min && | |
1064 | !capable(CAP_SYS_RESOURCE)) { | |
1065 | err = -EACCES; | |
1066 | goto err_unlock; | |
1067 | } | |
1068 | } | |
1069 | ||
44a70ade MH |
1070 | /* |
1071 | * Make sure we will check other processes sharing the mm if this is | |
1072 | * not vfrok which wants its own oom_score_adj. | |
1073 | * pin the mm so it doesn't go away and get reused after task_unlock | |
1074 | */ | |
1075 | if (!task->vfork_done) { | |
1076 | struct task_struct *p = find_lock_task_mm(task); | |
1077 | ||
1078 | if (p) { | |
1079 | if (atomic_read(&p->mm->mm_users) > 1) { | |
1080 | mm = p->mm; | |
f1f10076 | 1081 | mmgrab(mm); |
44a70ade MH |
1082 | } |
1083 | task_unlock(p); | |
1084 | } | |
1085 | } | |
1086 | ||
1d5f0acb MH |
1087 | task->signal->oom_score_adj = oom_adj; |
1088 | if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE)) | |
1089 | task->signal->oom_score_adj_min = (short)oom_adj; | |
1090 | trace_oom_score_adj_update(task); | |
44a70ade MH |
1091 | |
1092 | if (mm) { | |
1093 | struct task_struct *p; | |
1094 | ||
1095 | rcu_read_lock(); | |
1096 | for_each_process(p) { | |
1097 | if (same_thread_group(task, p)) | |
1098 | continue; | |
1099 | ||
1100 | /* do not touch kernel threads or the global init */ | |
1101 | if (p->flags & PF_KTHREAD || is_global_init(p)) | |
1102 | continue; | |
1103 | ||
1104 | task_lock(p); | |
1105 | if (!p->vfork_done && process_shares_mm(p, mm)) { | |
44a70ade MH |
1106 | p->signal->oom_score_adj = oom_adj; |
1107 | if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE)) | |
1108 | p->signal->oom_score_adj_min = (short)oom_adj; | |
1109 | } | |
1110 | task_unlock(p); | |
1111 | } | |
1112 | rcu_read_unlock(); | |
1113 | mmdrop(mm); | |
1114 | } | |
1d5f0acb MH |
1115 | err_unlock: |
1116 | mutex_unlock(&oom_adj_mutex); | |
1117 | put_task_struct(task); | |
1118 | return err; | |
1119 | } | |
f913da59 | 1120 | |
b72bdfa7 DR |
1121 | /* |
1122 | * /proc/pid/oom_adj exists solely for backwards compatibility with previous | |
1123 | * kernels. The effective policy is defined by oom_score_adj, which has a | |
1124 | * different scale: oom_adj grew exponentially and oom_score_adj grows linearly. | |
1125 | * Values written to oom_adj are simply mapped linearly to oom_score_adj. | |
1126 | * Processes that become oom disabled via oom_adj will still be oom disabled | |
1127 | * with this implementation. | |
1128 | * | |
1129 | * oom_adj cannot be removed since existing userspace binaries use it. | |
1130 | */ | |
fa0cbbf1 DR |
1131 | static ssize_t oom_adj_write(struct file *file, const char __user *buf, |
1132 | size_t count, loff_t *ppos) | |
1133 | { | |
fa0cbbf1 DR |
1134 | char buffer[PROC_NUMBUF]; |
1135 | int oom_adj; | |
fa0cbbf1 DR |
1136 | int err; |
1137 | ||
1138 | memset(buffer, 0, sizeof(buffer)); | |
1139 | if (count > sizeof(buffer) - 1) | |
1140 | count = sizeof(buffer) - 1; | |
1141 | if (copy_from_user(buffer, buf, count)) { | |
1142 | err = -EFAULT; | |
1143 | goto out; | |
1144 | } | |
1145 | ||
1146 | err = kstrtoint(strstrip(buffer), 0, &oom_adj); | |
1147 | if (err) | |
1148 | goto out; | |
1149 | if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) && | |
1150 | oom_adj != OOM_DISABLE) { | |
1151 | err = -EINVAL; | |
1152 | goto out; | |
1153 | } | |
1154 | ||
fa0cbbf1 DR |
1155 | /* |
1156 | * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum | |
1157 | * value is always attainable. | |
1158 | */ | |
1159 | if (oom_adj == OOM_ADJUST_MAX) | |
1160 | oom_adj = OOM_SCORE_ADJ_MAX; | |
1161 | else | |
1162 | oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE; | |
1163 | ||
1d5f0acb | 1164 | err = __set_oom_adj(file, oom_adj, true); |
fa0cbbf1 DR |
1165 | out: |
1166 | return err < 0 ? err : count; | |
1167 | } | |
1168 | ||
1169 | static const struct file_operations proc_oom_adj_operations = { | |
1170 | .read = oom_adj_read, | |
1171 | .write = oom_adj_write, | |
1172 | .llseek = generic_file_llseek, | |
1173 | }; | |
1174 | ||
a63d83f4 DR |
1175 | static ssize_t oom_score_adj_read(struct file *file, char __user *buf, |
1176 | size_t count, loff_t *ppos) | |
1177 | { | |
496ad9aa | 1178 | struct task_struct *task = get_proc_task(file_inode(file)); |
a63d83f4 | 1179 | char buffer[PROC_NUMBUF]; |
a9c58b90 | 1180 | short oom_score_adj = OOM_SCORE_ADJ_MIN; |
a63d83f4 DR |
1181 | size_t len; |
1182 | ||
1183 | if (!task) | |
1184 | return -ESRCH; | |
f913da59 | 1185 | oom_score_adj = task->signal->oom_score_adj; |
a63d83f4 | 1186 | put_task_struct(task); |
a9c58b90 | 1187 | len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj); |
a63d83f4 DR |
1188 | return simple_read_from_buffer(buf, count, ppos, buffer, len); |
1189 | } | |
1190 | ||
1191 | static ssize_t oom_score_adj_write(struct file *file, const char __user *buf, | |
1192 | size_t count, loff_t *ppos) | |
1193 | { | |
a63d83f4 | 1194 | char buffer[PROC_NUMBUF]; |
0a8cb8e3 | 1195 | int oom_score_adj; |
a63d83f4 DR |
1196 | int err; |
1197 | ||
1198 | memset(buffer, 0, sizeof(buffer)); | |
1199 | if (count > sizeof(buffer) - 1) | |
1200 | count = sizeof(buffer) - 1; | |
723548bf DR |
1201 | if (copy_from_user(buffer, buf, count)) { |
1202 | err = -EFAULT; | |
1203 | goto out; | |
1204 | } | |
a63d83f4 | 1205 | |
0a8cb8e3 | 1206 | err = kstrtoint(strstrip(buffer), 0, &oom_score_adj); |
a63d83f4 | 1207 | if (err) |
723548bf | 1208 | goto out; |
a63d83f4 | 1209 | if (oom_score_adj < OOM_SCORE_ADJ_MIN || |
723548bf DR |
1210 | oom_score_adj > OOM_SCORE_ADJ_MAX) { |
1211 | err = -EINVAL; | |
1212 | goto out; | |
1213 | } | |
a63d83f4 | 1214 | |
1d5f0acb | 1215 | err = __set_oom_adj(file, oom_score_adj, false); |
723548bf DR |
1216 | out: |
1217 | return err < 0 ? err : count; | |
a63d83f4 DR |
1218 | } |
1219 | ||
1220 | static const struct file_operations proc_oom_score_adj_operations = { | |
1221 | .read = oom_score_adj_read, | |
1222 | .write = oom_score_adj_write, | |
6038f373 | 1223 | .llseek = default_llseek, |
a63d83f4 DR |
1224 | }; |
1225 | ||
4b7d248b | 1226 | #ifdef CONFIG_AUDIT |
b4eb4f7f | 1227 | #define TMPBUFLEN 11 |
1da177e4 LT |
1228 | static ssize_t proc_loginuid_read(struct file * file, char __user * buf, |
1229 | size_t count, loff_t *ppos) | |
1230 | { | |
496ad9aa | 1231 | struct inode * inode = file_inode(file); |
99f89551 | 1232 | struct task_struct *task = get_proc_task(inode); |
1da177e4 LT |
1233 | ssize_t length; |
1234 | char tmpbuf[TMPBUFLEN]; | |
1235 | ||
99f89551 EB |
1236 | if (!task) |
1237 | return -ESRCH; | |
1da177e4 | 1238 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", |
e1760bd5 EB |
1239 | from_kuid(file->f_cred->user_ns, |
1240 | audit_get_loginuid(task))); | |
99f89551 | 1241 | put_task_struct(task); |
1da177e4 LT |
1242 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); |
1243 | } | |
1244 | ||
1245 | static ssize_t proc_loginuid_write(struct file * file, const char __user * buf, | |
1246 | size_t count, loff_t *ppos) | |
1247 | { | |
496ad9aa | 1248 | struct inode * inode = file_inode(file); |
1da177e4 | 1249 | uid_t loginuid; |
e1760bd5 | 1250 | kuid_t kloginuid; |
774636e1 | 1251 | int rv; |
1da177e4 | 1252 | |
7dc52157 PM |
1253 | rcu_read_lock(); |
1254 | if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) { | |
1255 | rcu_read_unlock(); | |
1da177e4 | 1256 | return -EPERM; |
7dc52157 PM |
1257 | } |
1258 | rcu_read_unlock(); | |
1da177e4 | 1259 | |
1da177e4 LT |
1260 | if (*ppos != 0) { |
1261 | /* No partial writes. */ | |
1262 | return -EINVAL; | |
1263 | } | |
1da177e4 | 1264 | |
774636e1 AD |
1265 | rv = kstrtou32_from_user(buf, count, 10, &loginuid); |
1266 | if (rv < 0) | |
1267 | return rv; | |
81407c84 EP |
1268 | |
1269 | /* is userspace tring to explicitly UNSET the loginuid? */ | |
1270 | if (loginuid == AUDIT_UID_UNSET) { | |
1271 | kloginuid = INVALID_UID; | |
1272 | } else { | |
1273 | kloginuid = make_kuid(file->f_cred->user_ns, loginuid); | |
774636e1 AD |
1274 | if (!uid_valid(kloginuid)) |
1275 | return -EINVAL; | |
e1760bd5 EB |
1276 | } |
1277 | ||
774636e1 AD |
1278 | rv = audit_set_loginuid(kloginuid); |
1279 | if (rv < 0) | |
1280 | return rv; | |
1281 | return count; | |
1da177e4 LT |
1282 | } |
1283 | ||
00977a59 | 1284 | static const struct file_operations proc_loginuid_operations = { |
1da177e4 LT |
1285 | .read = proc_loginuid_read, |
1286 | .write = proc_loginuid_write, | |
87df8424 | 1287 | .llseek = generic_file_llseek, |
1da177e4 | 1288 | }; |
1e0bd755 EP |
1289 | |
1290 | static ssize_t proc_sessionid_read(struct file * file, char __user * buf, | |
1291 | size_t count, loff_t *ppos) | |
1292 | { | |
496ad9aa | 1293 | struct inode * inode = file_inode(file); |
1e0bd755 EP |
1294 | struct task_struct *task = get_proc_task(inode); |
1295 | ssize_t length; | |
1296 | char tmpbuf[TMPBUFLEN]; | |
1297 | ||
1298 | if (!task) | |
1299 | return -ESRCH; | |
1300 | length = scnprintf(tmpbuf, TMPBUFLEN, "%u", | |
1301 | audit_get_sessionid(task)); | |
1302 | put_task_struct(task); | |
1303 | return simple_read_from_buffer(buf, count, ppos, tmpbuf, length); | |
1304 | } | |
1305 | ||
1306 | static const struct file_operations proc_sessionid_operations = { | |
1307 | .read = proc_sessionid_read, | |
87df8424 | 1308 | .llseek = generic_file_llseek, |
1e0bd755 | 1309 | }; |
1da177e4 LT |
1310 | #endif |
1311 | ||
f4f154fd AM |
1312 | #ifdef CONFIG_FAULT_INJECTION |
1313 | static ssize_t proc_fault_inject_read(struct file * file, char __user * buf, | |
1314 | size_t count, loff_t *ppos) | |
1315 | { | |
496ad9aa | 1316 | struct task_struct *task = get_proc_task(file_inode(file)); |
f4f154fd AM |
1317 | char buffer[PROC_NUMBUF]; |
1318 | size_t len; | |
1319 | int make_it_fail; | |
f4f154fd AM |
1320 | |
1321 | if (!task) | |
1322 | return -ESRCH; | |
1323 | make_it_fail = task->make_it_fail; | |
1324 | put_task_struct(task); | |
1325 | ||
1326 | len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail); | |
0c28f287 AM |
1327 | |
1328 | return simple_read_from_buffer(buf, count, ppos, buffer, len); | |
f4f154fd AM |
1329 | } |
1330 | ||
1331 | static ssize_t proc_fault_inject_write(struct file * file, | |
1332 | const char __user * buf, size_t count, loff_t *ppos) | |
1333 | { | |
1334 | struct task_struct *task; | |
774636e1 | 1335 | char buffer[PROC_NUMBUF]; |
f4f154fd | 1336 | int make_it_fail; |
774636e1 | 1337 | int rv; |
f4f154fd AM |
1338 | |
1339 | if (!capable(CAP_SYS_RESOURCE)) | |
1340 | return -EPERM; | |
1341 | memset(buffer, 0, sizeof(buffer)); | |
1342 | if (count > sizeof(buffer) - 1) | |
1343 | count = sizeof(buffer) - 1; | |
1344 | if (copy_from_user(buffer, buf, count)) | |
1345 | return -EFAULT; | |
774636e1 AD |
1346 | rv = kstrtoint(strstrip(buffer), 0, &make_it_fail); |
1347 | if (rv < 0) | |
1348 | return rv; | |
16caed31 DJ |
1349 | if (make_it_fail < 0 || make_it_fail > 1) |
1350 | return -EINVAL; | |
1351 | ||
496ad9aa | 1352 | task = get_proc_task(file_inode(file)); |
f4f154fd AM |
1353 | if (!task) |
1354 | return -ESRCH; | |
1355 | task->make_it_fail = make_it_fail; | |
1356 | put_task_struct(task); | |
cba8aafe VL |
1357 | |
1358 | return count; | |
f4f154fd AM |
1359 | } |
1360 | ||
00977a59 | 1361 | static const struct file_operations proc_fault_inject_operations = { |
f4f154fd AM |
1362 | .read = proc_fault_inject_read, |
1363 | .write = proc_fault_inject_write, | |
87df8424 | 1364 | .llseek = generic_file_llseek, |
f4f154fd | 1365 | }; |
e41d5818 DV |
1366 | |
1367 | static ssize_t proc_fail_nth_write(struct file *file, const char __user *buf, | |
1368 | size_t count, loff_t *ppos) | |
1369 | { | |
1370 | struct task_struct *task; | |
9049f2f6 AM |
1371 | int err; |
1372 | unsigned int n; | |
e41d5818 | 1373 | |
1203c8e6 AM |
1374 | err = kstrtouint_from_user(buf, count, 0, &n); |
1375 | if (err) | |
1376 | return err; | |
1377 | ||
e41d5818 DV |
1378 | task = get_proc_task(file_inode(file)); |
1379 | if (!task) | |
1380 | return -ESRCH; | |
9f7118b2 | 1381 | task->fail_nth = n; |
e41d5818 | 1382 | put_task_struct(task); |
1203c8e6 | 1383 | |
e41d5818 DV |
1384 | return count; |
1385 | } | |
1386 | ||
1387 | static ssize_t proc_fail_nth_read(struct file *file, char __user *buf, | |
1388 | size_t count, loff_t *ppos) | |
1389 | { | |
1390 | struct task_struct *task; | |
bfc74093 AM |
1391 | char numbuf[PROC_NUMBUF]; |
1392 | ssize_t len; | |
e41d5818 DV |
1393 | |
1394 | task = get_proc_task(file_inode(file)); | |
1395 | if (!task) | |
1396 | return -ESRCH; | |
9f7118b2 | 1397 | len = snprintf(numbuf, sizeof(numbuf), "%u\n", task->fail_nth); |
1203c8e6 | 1398 | put_task_struct(task); |
a44937fe | 1399 | return simple_read_from_buffer(buf, count, ppos, numbuf, len); |
e41d5818 DV |
1400 | } |
1401 | ||
1402 | static const struct file_operations proc_fail_nth_operations = { | |
1403 | .read = proc_fail_nth_read, | |
1404 | .write = proc_fail_nth_write, | |
1405 | }; | |
f4f154fd AM |
1406 | #endif |
1407 | ||
9745512c | 1408 | |
43ae34cb IM |
1409 | #ifdef CONFIG_SCHED_DEBUG |
1410 | /* | |
1411 | * Print out various scheduling related per-task fields: | |
1412 | */ | |
1413 | static int sched_show(struct seq_file *m, void *v) | |
1414 | { | |
1415 | struct inode *inode = m->private; | |
76f668be | 1416 | struct pid_namespace *ns = proc_pid_ns(inode); |
43ae34cb IM |
1417 | struct task_struct *p; |
1418 | ||
43ae34cb IM |
1419 | p = get_proc_task(inode); |
1420 | if (!p) | |
1421 | return -ESRCH; | |
74dc3384 | 1422 | proc_sched_show_task(p, ns, m); |
43ae34cb IM |
1423 | |
1424 | put_task_struct(p); | |
1425 | ||
1426 | return 0; | |
1427 | } | |
1428 | ||
1429 | static ssize_t | |
1430 | sched_write(struct file *file, const char __user *buf, | |
1431 | size_t count, loff_t *offset) | |
1432 | { | |
496ad9aa | 1433 | struct inode *inode = file_inode(file); |
43ae34cb IM |
1434 | struct task_struct *p; |
1435 | ||
43ae34cb IM |
1436 | p = get_proc_task(inode); |
1437 | if (!p) | |
1438 | return -ESRCH; | |
1439 | proc_sched_set_task(p); | |
1440 | ||
1441 | put_task_struct(p); | |
1442 | ||
1443 | return count; | |
1444 | } | |
1445 | ||
1446 | static int sched_open(struct inode *inode, struct file *filp) | |
1447 | { | |
c6a34058 | 1448 | return single_open(filp, sched_show, inode); |
43ae34cb IM |
1449 | } |
1450 | ||
1451 | static const struct file_operations proc_pid_sched_operations = { | |
1452 | .open = sched_open, | |
1453 | .read = seq_read, | |
1454 | .write = sched_write, | |
1455 | .llseek = seq_lseek, | |
5ea473a1 | 1456 | .release = single_release, |
43ae34cb IM |
1457 | }; |
1458 | ||
1459 | #endif | |
1460 | ||
5091faa4 MG |
1461 | #ifdef CONFIG_SCHED_AUTOGROUP |
1462 | /* | |
1463 | * Print out autogroup related information: | |
1464 | */ | |
1465 | static int sched_autogroup_show(struct seq_file *m, void *v) | |
1466 | { | |
1467 | struct inode *inode = m->private; | |
1468 | struct task_struct *p; | |
1469 | ||
1470 | p = get_proc_task(inode); | |
1471 | if (!p) | |
1472 | return -ESRCH; | |
1473 | proc_sched_autogroup_show_task(p, m); | |
1474 | ||
1475 | put_task_struct(p); | |
1476 | ||
1477 | return 0; | |
1478 | } | |
1479 | ||
1480 | static ssize_t | |
1481 | sched_autogroup_write(struct file *file, const char __user *buf, | |
1482 | size_t count, loff_t *offset) | |
1483 | { | |
496ad9aa | 1484 | struct inode *inode = file_inode(file); |
5091faa4 MG |
1485 | struct task_struct *p; |
1486 | char buffer[PROC_NUMBUF]; | |
0a8cb8e3 | 1487 | int nice; |
5091faa4 MG |
1488 | int err; |
1489 | ||
1490 | memset(buffer, 0, sizeof(buffer)); | |
1491 | if (count > sizeof(buffer) - 1) | |
1492 | count = sizeof(buffer) - 1; | |
1493 | if (copy_from_user(buffer, buf, count)) | |
1494 | return -EFAULT; | |
1495 | ||
0a8cb8e3 AD |
1496 | err = kstrtoint(strstrip(buffer), 0, &nice); |
1497 | if (err < 0) | |
1498 | return err; | |
5091faa4 MG |
1499 | |
1500 | p = get_proc_task(inode); | |
1501 | if (!p) | |
1502 | return -ESRCH; | |
1503 | ||
2e5b5b3a | 1504 | err = proc_sched_autogroup_set_nice(p, nice); |
5091faa4 MG |
1505 | if (err) |
1506 | count = err; | |
1507 | ||
1508 | put_task_struct(p); | |
1509 | ||
1510 | return count; | |
1511 | } | |
1512 | ||
1513 | static int sched_autogroup_open(struct inode *inode, struct file *filp) | |
1514 | { | |
1515 | int ret; | |
1516 | ||
1517 | ret = single_open(filp, sched_autogroup_show, NULL); | |
1518 | if (!ret) { | |
1519 | struct seq_file *m = filp->private_data; | |
1520 | ||
1521 | m->private = inode; | |
1522 | } | |
1523 | return ret; | |
1524 | } | |
1525 | ||
1526 | static const struct file_operations proc_pid_sched_autogroup_operations = { | |
1527 | .open = sched_autogroup_open, | |
1528 | .read = seq_read, | |
1529 | .write = sched_autogroup_write, | |
1530 | .llseek = seq_lseek, | |
1531 | .release = single_release, | |
1532 | }; | |
1533 | ||
1534 | #endif /* CONFIG_SCHED_AUTOGROUP */ | |
1535 | ||
4614a696 JS |
1536 | static ssize_t comm_write(struct file *file, const char __user *buf, |
1537 | size_t count, loff_t *offset) | |
1538 | { | |
496ad9aa | 1539 | struct inode *inode = file_inode(file); |
4614a696 JS |
1540 | struct task_struct *p; |
1541 | char buffer[TASK_COMM_LEN]; | |
830e0fc9 | 1542 | const size_t maxlen = sizeof(buffer) - 1; |
4614a696 JS |
1543 | |
1544 | memset(buffer, 0, sizeof(buffer)); | |
830e0fc9 | 1545 | if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count)) |
4614a696 JS |
1546 | return -EFAULT; |
1547 | ||
1548 | p = get_proc_task(inode); | |
1549 | if (!p) | |
1550 | return -ESRCH; | |
1551 | ||
1552 | if (same_thread_group(current, p)) | |
1553 | set_task_comm(p, buffer); | |
1554 | else | |
1555 | count = -EINVAL; | |
1556 | ||
1557 | put_task_struct(p); | |
1558 | ||
1559 | return count; | |
1560 | } | |
1561 | ||
1562 | static int comm_show(struct seq_file *m, void *v) | |
1563 | { | |
1564 | struct inode *inode = m->private; | |
1565 | struct task_struct *p; | |
1566 | ||
1567 | p = get_proc_task(inode); | |
1568 | if (!p) | |
1569 | return -ESRCH; | |
1570 | ||
88b72b31 TH |
1571 | proc_task_name(m, p, false); |
1572 | seq_putc(m, '\n'); | |
4614a696 JS |
1573 | |
1574 | put_task_struct(p); | |
1575 | ||
1576 | return 0; | |
1577 | } | |
1578 | ||
1579 | static int comm_open(struct inode *inode, struct file *filp) | |
1580 | { | |
c6a34058 | 1581 | return single_open(filp, comm_show, inode); |
4614a696 JS |
1582 | } |
1583 | ||
1584 | static const struct file_operations proc_pid_set_comm_operations = { | |
1585 | .open = comm_open, | |
1586 | .read = seq_read, | |
1587 | .write = comm_write, | |
1588 | .llseek = seq_lseek, | |
1589 | .release = single_release, | |
1590 | }; | |
1591 | ||
7773fbc5 | 1592 | static int proc_exe_link(struct dentry *dentry, struct path *exe_path) |
925d1c40 MH |
1593 | { |
1594 | struct task_struct *task; | |
925d1c40 MH |
1595 | struct file *exe_file; |
1596 | ||
2b0143b5 | 1597 | task = get_proc_task(d_inode(dentry)); |
925d1c40 MH |
1598 | if (!task) |
1599 | return -ENOENT; | |
cd81a917 | 1600 | exe_file = get_task_exe_file(task); |
925d1c40 | 1601 | put_task_struct(task); |
925d1c40 MH |
1602 | if (exe_file) { |
1603 | *exe_path = exe_file->f_path; | |
1604 | path_get(&exe_file->f_path); | |
1605 | fput(exe_file); | |
1606 | return 0; | |
1607 | } else | |
1608 | return -ENOENT; | |
1609 | } | |
1610 | ||
6b255391 | 1611 | static const char *proc_pid_get_link(struct dentry *dentry, |
fceef393 AV |
1612 | struct inode *inode, |
1613 | struct delayed_call *done) | |
1da177e4 | 1614 | { |
408ef013 | 1615 | struct path path; |
1da177e4 LT |
1616 | int error = -EACCES; |
1617 | ||
6b255391 AV |
1618 | if (!dentry) |
1619 | return ERR_PTR(-ECHILD); | |
1620 | ||
778c1144 EB |
1621 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1622 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1623 | goto out; |
1da177e4 | 1624 | |
408ef013 CH |
1625 | error = PROC_I(inode)->op.proc_get_link(dentry, &path); |
1626 | if (error) | |
1627 | goto out; | |
1628 | ||
6e77137b | 1629 | nd_jump_link(&path); |
408ef013 | 1630 | return NULL; |
1da177e4 | 1631 | out: |
008b150a | 1632 | return ERR_PTR(error); |
1da177e4 LT |
1633 | } |
1634 | ||
3dcd25f3 | 1635 | static int do_proc_readlink(struct path *path, char __user *buffer, int buflen) |
1da177e4 | 1636 | { |
0ee931c4 | 1637 | char *tmp = (char *)__get_free_page(GFP_KERNEL); |
3dcd25f3 | 1638 | char *pathname; |
1da177e4 LT |
1639 | int len; |
1640 | ||
1641 | if (!tmp) | |
1642 | return -ENOMEM; | |
0c28f287 | 1643 | |
7b2a69ba | 1644 | pathname = d_path(path, tmp, PAGE_SIZE); |
3dcd25f3 JB |
1645 | len = PTR_ERR(pathname); |
1646 | if (IS_ERR(pathname)) | |
1da177e4 | 1647 | goto out; |
3dcd25f3 | 1648 | len = tmp + PAGE_SIZE - 1 - pathname; |
1da177e4 LT |
1649 | |
1650 | if (len > buflen) | |
1651 | len = buflen; | |
3dcd25f3 | 1652 | if (copy_to_user(buffer, pathname, len)) |
1da177e4 LT |
1653 | len = -EFAULT; |
1654 | out: | |
1655 | free_page((unsigned long)tmp); | |
1656 | return len; | |
1657 | } | |
1658 | ||
1659 | static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen) | |
1660 | { | |
1661 | int error = -EACCES; | |
2b0143b5 | 1662 | struct inode *inode = d_inode(dentry); |
3dcd25f3 | 1663 | struct path path; |
1da177e4 | 1664 | |
778c1144 EB |
1665 | /* Are we allowed to snoop on the tasks file descriptors? */ |
1666 | if (!proc_fd_access_allowed(inode)) | |
1da177e4 | 1667 | goto out; |
1da177e4 | 1668 | |
7773fbc5 | 1669 | error = PROC_I(inode)->op.proc_get_link(dentry, &path); |
1da177e4 LT |
1670 | if (error) |
1671 | goto out; | |
1672 | ||
3dcd25f3 JB |
1673 | error = do_proc_readlink(&path, buffer, buflen); |
1674 | path_put(&path); | |
1da177e4 | 1675 | out: |
1da177e4 LT |
1676 | return error; |
1677 | } | |
1678 | ||
faf60af1 | 1679 | const struct inode_operations proc_pid_link_inode_operations = { |
1da177e4 | 1680 | .readlink = proc_pid_readlink, |
6b255391 | 1681 | .get_link = proc_pid_get_link, |
6d76fa58 | 1682 | .setattr = proc_setattr, |
1da177e4 LT |
1683 | }; |
1684 | ||
28a6d671 EB |
1685 | |
1686 | /* building an inode */ | |
1687 | ||
c6eb50d2 | 1688 | void task_dump_owner(struct task_struct *task, umode_t mode, |
68eb94f1 EB |
1689 | kuid_t *ruid, kgid_t *rgid) |
1690 | { | |
1691 | /* Depending on the state of dumpable compute who should own a | |
1692 | * proc file for a task. | |
1693 | */ | |
1694 | const struct cred *cred; | |
1695 | kuid_t uid; | |
1696 | kgid_t gid; | |
1697 | ||
2e0ad552 AD |
1698 | if (unlikely(task->flags & PF_KTHREAD)) { |
1699 | *ruid = GLOBAL_ROOT_UID; | |
1700 | *rgid = GLOBAL_ROOT_GID; | |
1701 | return; | |
1702 | } | |
1703 | ||
68eb94f1 EB |
1704 | /* Default to the tasks effective ownership */ |
1705 | rcu_read_lock(); | |
1706 | cred = __task_cred(task); | |
1707 | uid = cred->euid; | |
1708 | gid = cred->egid; | |
1709 | rcu_read_unlock(); | |
1710 | ||
1711 | /* | |
1712 | * Before the /proc/pid/status file was created the only way to read | |
1713 | * the effective uid of a /process was to stat /proc/pid. Reading | |
1714 | * /proc/pid/status is slow enough that procps and other packages | |
1715 | * kept stating /proc/pid. To keep the rules in /proc simple I have | |
1716 | * made this apply to all per process world readable and executable | |
1717 | * directories. | |
1718 | */ | |
1719 | if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) { | |
1720 | struct mm_struct *mm; | |
1721 | task_lock(task); | |
1722 | mm = task->mm; | |
1723 | /* Make non-dumpable tasks owned by some root */ | |
1724 | if (mm) { | |
1725 | if (get_dumpable(mm) != SUID_DUMP_USER) { | |
1726 | struct user_namespace *user_ns = mm->user_ns; | |
1727 | ||
1728 | uid = make_kuid(user_ns, 0); | |
1729 | if (!uid_valid(uid)) | |
1730 | uid = GLOBAL_ROOT_UID; | |
1731 | ||
1732 | gid = make_kgid(user_ns, 0); | |
1733 | if (!gid_valid(gid)) | |
1734 | gid = GLOBAL_ROOT_GID; | |
1735 | } | |
1736 | } else { | |
1737 | uid = GLOBAL_ROOT_UID; | |
1738 | gid = GLOBAL_ROOT_GID; | |
1739 | } | |
1740 | task_unlock(task); | |
1741 | } | |
1742 | *ruid = uid; | |
1743 | *rgid = gid; | |
1744 | } | |
1745 | ||
db978da8 AG |
1746 | struct inode *proc_pid_make_inode(struct super_block * sb, |
1747 | struct task_struct *task, umode_t mode) | |
28a6d671 EB |
1748 | { |
1749 | struct inode * inode; | |
1750 | struct proc_inode *ei; | |
1da177e4 | 1751 | |
28a6d671 | 1752 | /* We need a new inode */ |
1da177e4 | 1753 | |
28a6d671 EB |
1754 | inode = new_inode(sb); |
1755 | if (!inode) | |
1756 | goto out; | |
1757 | ||
1758 | /* Common stuff */ | |
1759 | ei = PROC_I(inode); | |
db978da8 | 1760 | inode->i_mode = mode; |
85fe4025 | 1761 | inode->i_ino = get_next_ino(); |
078cd827 | 1762 | inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode); |
28a6d671 EB |
1763 | inode->i_op = &proc_def_inode_operations; |
1764 | ||
1765 | /* | |
1766 | * grab the reference to task. | |
1767 | */ | |
1a657f78 | 1768 | ei->pid = get_task_pid(task, PIDTYPE_PID); |
28a6d671 EB |
1769 | if (!ei->pid) |
1770 | goto out_unlock; | |
1771 | ||
68eb94f1 | 1772 | task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid); |
28a6d671 EB |
1773 | security_task_to_inode(task, inode); |
1774 | ||
1da177e4 | 1775 | out: |
28a6d671 EB |
1776 | return inode; |
1777 | ||
1778 | out_unlock: | |
1779 | iput(inode); | |
1780 | return NULL; | |
1da177e4 LT |
1781 | } |
1782 | ||
a528d35e DH |
1783 | int pid_getattr(const struct path *path, struct kstat *stat, |
1784 | u32 request_mask, unsigned int query_flags) | |
1da177e4 | 1785 | { |
a528d35e | 1786 | struct inode *inode = d_inode(path->dentry); |
76f668be | 1787 | struct pid_namespace *pid = proc_pid_ns(inode); |
28a6d671 | 1788 | struct task_struct *task; |
c69e8d9c | 1789 | |
28a6d671 | 1790 | generic_fillattr(inode, stat); |
1da177e4 | 1791 | |
dcb0f222 EB |
1792 | stat->uid = GLOBAL_ROOT_UID; |
1793 | stat->gid = GLOBAL_ROOT_GID; | |
94116929 | 1794 | rcu_read_lock(); |
28a6d671 EB |
1795 | task = pid_task(proc_pid(inode), PIDTYPE_PID); |
1796 | if (task) { | |
796f571b | 1797 | if (!has_pid_permissions(pid, task, HIDEPID_INVISIBLE)) { |
0499680a VK |
1798 | rcu_read_unlock(); |
1799 | /* | |
1800 | * This doesn't prevent learning whether PID exists, | |
1801 | * it only makes getattr() consistent with readdir(). | |
1802 | */ | |
1803 | return -ENOENT; | |
1804 | } | |
68eb94f1 | 1805 | task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid); |
1da177e4 | 1806 | } |
28a6d671 | 1807 | rcu_read_unlock(); |
d6e71144 | 1808 | return 0; |
1da177e4 LT |
1809 | } |
1810 | ||
1da177e4 LT |
1811 | /* dentry stuff */ |
1812 | ||
1813 | /* | |
1bbc5513 AV |
1814 | * Set <pid>/... inode ownership (can change due to setuid(), etc.) |
1815 | */ | |
1816 | void pid_update_inode(struct task_struct *task, struct inode *inode) | |
1817 | { | |
1818 | task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid); | |
1819 | ||
1820 | inode->i_mode &= ~(S_ISUID | S_ISGID); | |
1821 | security_task_to_inode(task, inode); | |
1822 | } | |
1823 | ||
1824 | /* | |
1da177e4 LT |
1825 | * Rewrite the inode's ownerships here because the owning task may have |
1826 | * performed a setuid(), etc. | |
99f89551 | 1827 | * |
1da177e4 | 1828 | */ |
1bbc5513 | 1829 | static int pid_revalidate(struct dentry *dentry, unsigned int flags) |
1da177e4 | 1830 | { |
34286d66 NP |
1831 | struct inode *inode; |
1832 | struct task_struct *task; | |
c69e8d9c | 1833 | |
0b728e19 | 1834 | if (flags & LOOKUP_RCU) |
34286d66 NP |
1835 | return -ECHILD; |
1836 | ||
2b0143b5 | 1837 | inode = d_inode(dentry); |
34286d66 NP |
1838 | task = get_proc_task(inode); |
1839 | ||
99f89551 | 1840 | if (task) { |
1bbc5513 | 1841 | pid_update_inode(task, inode); |
99f89551 | 1842 | put_task_struct(task); |
1da177e4 LT |
1843 | return 1; |
1844 | } | |
1da177e4 LT |
1845 | return 0; |
1846 | } | |
1847 | ||
d855a4b7 ON |
1848 | static inline bool proc_inode_is_dead(struct inode *inode) |
1849 | { | |
1850 | return !proc_pid(inode)->tasks[PIDTYPE_PID].first; | |
1851 | } | |
1852 | ||
1dd704b6 DH |
1853 | int pid_delete_dentry(const struct dentry *dentry) |
1854 | { | |
1855 | /* Is the task we represent dead? | |
1856 | * If so, then don't put the dentry on the lru list, | |
1857 | * kill it immediately. | |
1858 | */ | |
2b0143b5 | 1859 | return proc_inode_is_dead(d_inode(dentry)); |
1dd704b6 DH |
1860 | } |
1861 | ||
6b4e306a | 1862 | const struct dentry_operations pid_dentry_operations = |
28a6d671 EB |
1863 | { |
1864 | .d_revalidate = pid_revalidate, | |
1865 | .d_delete = pid_delete_dentry, | |
1866 | }; | |
1867 | ||
1868 | /* Lookups */ | |
1869 | ||
1c0d04c9 EB |
1870 | /* |
1871 | * Fill a directory entry. | |
1872 | * | |
1873 | * If possible create the dcache entry and derive our inode number and | |
1874 | * file type from dcache entry. | |
1875 | * | |
1876 | * Since all of the proc inode numbers are dynamically generated, the inode | |
1877 | * numbers do not exist until the inode is cache. This means creating the | |
1878 | * the dcache entry in readdir is necessary to keep the inode numbers | |
1879 | * reported by readdir in sync with the inode numbers reported | |
1880 | * by stat. | |
1881 | */ | |
f0c3b509 | 1882 | bool proc_fill_cache(struct file *file, struct dir_context *ctx, |
a4ef3895 | 1883 | const char *name, unsigned int len, |
c5141e6d | 1884 | instantiate_t instantiate, struct task_struct *task, const void *ptr) |
61a28784 | 1885 | { |
f0c3b509 | 1886 | struct dentry *child, *dir = file->f_path.dentry; |
1df98b8b | 1887 | struct qstr qname = QSTR_INIT(name, len); |
61a28784 | 1888 | struct inode *inode; |
0168b9e3 AV |
1889 | unsigned type = DT_UNKNOWN; |
1890 | ino_t ino = 1; | |
61a28784 | 1891 | |
1df98b8b | 1892 | child = d_hash_and_lookup(dir, &qname); |
61a28784 | 1893 | if (!child) { |
3781764b AV |
1894 | DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq); |
1895 | child = d_alloc_parallel(dir, &qname, &wq); | |
1896 | if (IS_ERR(child)) | |
1df98b8b | 1897 | goto end_instantiate; |
3781764b | 1898 | if (d_in_lookup(child)) { |
0168b9e3 AV |
1899 | struct dentry *res; |
1900 | res = instantiate(child, task, ptr); | |
3781764b | 1901 | d_lookup_done(child); |
0168b9e3 AV |
1902 | if (unlikely(res)) { |
1903 | dput(child); | |
1904 | child = res; | |
d85b399b AV |
1905 | if (IS_ERR(child)) |
1906 | goto end_instantiate; | |
3781764b | 1907 | } |
61a28784 EB |
1908 | } |
1909 | } | |
2b0143b5 | 1910 | inode = d_inode(child); |
147ce699 AV |
1911 | ino = inode->i_ino; |
1912 | type = inode->i_mode >> 12; | |
61a28784 | 1913 | dput(child); |
d85b399b | 1914 | end_instantiate: |
f0c3b509 | 1915 | return dir_emit(ctx, name, len, ino, type); |
61a28784 EB |
1916 | } |
1917 | ||
640708a2 PE |
1918 | /* |
1919 | * dname_to_vma_addr - maps a dentry name into two unsigned longs | |
1920 | * which represent vma start and end addresses. | |
1921 | */ | |
1922 | static int dname_to_vma_addr(struct dentry *dentry, | |
1923 | unsigned long *start, unsigned long *end) | |
1924 | { | |
ac7f1061 AD |
1925 | const char *str = dentry->d_name.name; |
1926 | unsigned long long sval, eval; | |
1927 | unsigned int len; | |
1928 | ||
35318db5 AD |
1929 | if (str[0] == '0' && str[1] != '-') |
1930 | return -EINVAL; | |
ac7f1061 AD |
1931 | len = _parse_integer(str, 16, &sval); |
1932 | if (len & KSTRTOX_OVERFLOW) | |
1933 | return -EINVAL; | |
1934 | if (sval != (unsigned long)sval) | |
1935 | return -EINVAL; | |
1936 | str += len; | |
1937 | ||
1938 | if (*str != '-') | |
640708a2 | 1939 | return -EINVAL; |
ac7f1061 AD |
1940 | str++; |
1941 | ||
35318db5 AD |
1942 | if (str[0] == '0' && str[1]) |
1943 | return -EINVAL; | |
ac7f1061 AD |
1944 | len = _parse_integer(str, 16, &eval); |
1945 | if (len & KSTRTOX_OVERFLOW) | |
1946 | return -EINVAL; | |
1947 | if (eval != (unsigned long)eval) | |
1948 | return -EINVAL; | |
1949 | str += len; | |
1950 | ||
1951 | if (*str != '\0') | |
1952 | return -EINVAL; | |
1953 | ||
1954 | *start = sval; | |
1955 | *end = eval; | |
640708a2 PE |
1956 | |
1957 | return 0; | |
1958 | } | |
1959 | ||
0b728e19 | 1960 | static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags) |
640708a2 PE |
1961 | { |
1962 | unsigned long vm_start, vm_end; | |
1963 | bool exact_vma_exists = false; | |
1964 | struct mm_struct *mm = NULL; | |
1965 | struct task_struct *task; | |
640708a2 PE |
1966 | struct inode *inode; |
1967 | int status = 0; | |
1968 | ||
0b728e19 | 1969 | if (flags & LOOKUP_RCU) |
640708a2 PE |
1970 | return -ECHILD; |
1971 | ||
2b0143b5 | 1972 | inode = d_inode(dentry); |
640708a2 PE |
1973 | task = get_proc_task(inode); |
1974 | if (!task) | |
1975 | goto out_notask; | |
1976 | ||
caaee623 | 1977 | mm = mm_access(task, PTRACE_MODE_READ_FSCREDS); |
2344bec7 | 1978 | if (IS_ERR_OR_NULL(mm)) |
640708a2 PE |
1979 | goto out; |
1980 | ||
1981 | if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) { | |
cd9e2bb8 KK |
1982 | status = down_read_killable(&mm->mmap_sem); |
1983 | if (!status) { | |
1984 | exact_vma_exists = !!find_exact_vma(mm, vm_start, | |
1985 | vm_end); | |
1986 | up_read(&mm->mmap_sem); | |
1987 | } | |
640708a2 PE |
1988 | } |
1989 | ||
1990 | mmput(mm); | |
1991 | ||
1992 | if (exact_vma_exists) { | |
68eb94f1 EB |
1993 | task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid); |
1994 | ||
640708a2 PE |
1995 | security_task_to_inode(task, inode); |
1996 | status = 1; | |
1997 | } | |
1998 | ||
1999 | out: | |
2000 | put_task_struct(task); | |
2001 | ||
2002 | out_notask: | |
640708a2 PE |
2003 | return status; |
2004 | } | |
2005 | ||
2006 | static const struct dentry_operations tid_map_files_dentry_operations = { | |
2007 | .d_revalidate = map_files_d_revalidate, | |
2008 | .d_delete = pid_delete_dentry, | |
2009 | }; | |
2010 | ||
6b255391 | 2011 | static int map_files_get_link(struct dentry *dentry, struct path *path) |
640708a2 PE |
2012 | { |
2013 | unsigned long vm_start, vm_end; | |
2014 | struct vm_area_struct *vma; | |
2015 | struct task_struct *task; | |
2016 | struct mm_struct *mm; | |
2017 | int rc; | |
2018 | ||
2019 | rc = -ENOENT; | |
2b0143b5 | 2020 | task = get_proc_task(d_inode(dentry)); |
640708a2 PE |
2021 | if (!task) |
2022 | goto out; | |
2023 | ||
2024 | mm = get_task_mm(task); | |
2025 | put_task_struct(task); | |
2026 | if (!mm) | |
2027 | goto out; | |
2028 | ||
2029 | rc = dname_to_vma_addr(dentry, &vm_start, &vm_end); | |
2030 | if (rc) | |
2031 | goto out_mmput; | |
2032 | ||
cd9e2bb8 KK |
2033 | rc = down_read_killable(&mm->mmap_sem); |
2034 | if (rc) | |
2035 | goto out_mmput; | |
2036 | ||
70335abb | 2037 | rc = -ENOENT; |
640708a2 PE |
2038 | vma = find_exact_vma(mm, vm_start, vm_end); |
2039 | if (vma && vma->vm_file) { | |
8994cce9 | 2040 | *path = vma_pr_or_file(vma)->f_path; |
640708a2 PE |
2041 | path_get(path); |
2042 | rc = 0; | |
2043 | } | |
2044 | up_read(&mm->mmap_sem); | |
2045 | ||
2046 | out_mmput: | |
2047 | mmput(mm); | |
2048 | out: | |
2049 | return rc; | |
2050 | } | |
2051 | ||
2052 | struct map_files_info { | |
20d28cde AD |
2053 | unsigned long start; |
2054 | unsigned long end; | |
7b540d06 | 2055 | fmode_t mode; |
640708a2 PE |
2056 | }; |
2057 | ||
bdb4d100 CO |
2058 | /* |
2059 | * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the | |
2060 | * symlinks may be used to bypass permissions on ancestor directories in the | |
2061 | * path to the file in question. | |
2062 | */ | |
2063 | static const char * | |
6b255391 | 2064 | proc_map_files_get_link(struct dentry *dentry, |
fceef393 AV |
2065 | struct inode *inode, |
2066 | struct delayed_call *done) | |
bdb4d100 CO |
2067 | { |
2068 | if (!capable(CAP_SYS_ADMIN)) | |
2069 | return ERR_PTR(-EPERM); | |
2070 | ||
fceef393 | 2071 | return proc_pid_get_link(dentry, inode, done); |
bdb4d100 CO |
2072 | } |
2073 | ||
2074 | /* | |
6b255391 | 2075 | * Identical to proc_pid_link_inode_operations except for get_link() |
bdb4d100 CO |
2076 | */ |
2077 | static const struct inode_operations proc_map_files_link_inode_operations = { | |
2078 | .readlink = proc_pid_readlink, | |
6b255391 | 2079 | .get_link = proc_map_files_get_link, |
bdb4d100 CO |
2080 | .setattr = proc_setattr, |
2081 | }; | |
2082 | ||
0168b9e3 AV |
2083 | static struct dentry * |
2084 | proc_map_files_instantiate(struct dentry *dentry, | |
640708a2 PE |
2085 | struct task_struct *task, const void *ptr) |
2086 | { | |
7b540d06 | 2087 | fmode_t mode = (fmode_t)(unsigned long)ptr; |
640708a2 PE |
2088 | struct proc_inode *ei; |
2089 | struct inode *inode; | |
2090 | ||
0168b9e3 | 2091 | inode = proc_pid_make_inode(dentry->d_sb, task, S_IFLNK | |
db978da8 AG |
2092 | ((mode & FMODE_READ ) ? S_IRUSR : 0) | |
2093 | ((mode & FMODE_WRITE) ? S_IWUSR : 0)); | |
640708a2 | 2094 | if (!inode) |
0168b9e3 | 2095 | return ERR_PTR(-ENOENT); |
640708a2 PE |
2096 | |
2097 | ei = PROC_I(inode); | |
6b255391 | 2098 | ei->op.proc_get_link = map_files_get_link; |
640708a2 | 2099 | |
bdb4d100 | 2100 | inode->i_op = &proc_map_files_link_inode_operations; |
640708a2 | 2101 | inode->i_size = 64; |
640708a2 PE |
2102 | |
2103 | d_set_d_op(dentry, &tid_map_files_dentry_operations); | |
0168b9e3 | 2104 | return d_splice_alias(inode, dentry); |
640708a2 PE |
2105 | } |
2106 | ||
2107 | static struct dentry *proc_map_files_lookup(struct inode *dir, | |
00cd8dd3 | 2108 | struct dentry *dentry, unsigned int flags) |
640708a2 PE |
2109 | { |
2110 | unsigned long vm_start, vm_end; | |
2111 | struct vm_area_struct *vma; | |
2112 | struct task_struct *task; | |
0168b9e3 | 2113 | struct dentry *result; |
640708a2 PE |
2114 | struct mm_struct *mm; |
2115 | ||
0168b9e3 | 2116 | result = ERR_PTR(-ENOENT); |
640708a2 PE |
2117 | task = get_proc_task(dir); |
2118 | if (!task) | |
2119 | goto out; | |
2120 | ||
0168b9e3 | 2121 | result = ERR_PTR(-EACCES); |
caaee623 | 2122 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
640708a2 PE |
2123 | goto out_put_task; |
2124 | ||
0168b9e3 | 2125 | result = ERR_PTR(-ENOENT); |
640708a2 | 2126 | if (dname_to_vma_addr(dentry, &vm_start, &vm_end)) |
eb94cd96 | 2127 | goto out_put_task; |
640708a2 PE |
2128 | |
2129 | mm = get_task_mm(task); | |
2130 | if (!mm) | |
eb94cd96 | 2131 | goto out_put_task; |
640708a2 | 2132 | |
cd9e2bb8 KK |
2133 | result = ERR_PTR(-EINTR); |
2134 | if (down_read_killable(&mm->mmap_sem)) | |
2135 | goto out_put_mm; | |
2136 | ||
2137 | result = ERR_PTR(-ENOENT); | |
640708a2 PE |
2138 | vma = find_exact_vma(mm, vm_start, vm_end); |
2139 | if (!vma) | |
2140 | goto out_no_vma; | |
2141 | ||
05f56484 | 2142 | if (vma->vm_file) |
0168b9e3 | 2143 | result = proc_map_files_instantiate(dentry, task, |
05f56484 | 2144 | (void *)(unsigned long)vma->vm_file->f_mode); |
640708a2 PE |
2145 | |
2146 | out_no_vma: | |
2147 | up_read(&mm->mmap_sem); | |
cd9e2bb8 | 2148 | out_put_mm: |
640708a2 | 2149 | mmput(mm); |
640708a2 PE |
2150 | out_put_task: |
2151 | put_task_struct(task); | |
2152 | out: | |
0168b9e3 | 2153 | return result; |
640708a2 PE |
2154 | } |
2155 | ||
2156 | static const struct inode_operations proc_map_files_inode_operations = { | |
2157 | .lookup = proc_map_files_lookup, | |
2158 | .permission = proc_fd_permission, | |
2159 | .setattr = proc_setattr, | |
2160 | }; | |
2161 | ||
2162 | static int | |
f0c3b509 | 2163 | proc_map_files_readdir(struct file *file, struct dir_context *ctx) |
640708a2 | 2164 | { |
640708a2 PE |
2165 | struct vm_area_struct *vma; |
2166 | struct task_struct *task; | |
2167 | struct mm_struct *mm; | |
f0c3b509 | 2168 | unsigned long nr_files, pos, i; |
94f8f3b0 | 2169 | GENRADIX(struct map_files_info) fa; |
f0c3b509 | 2170 | struct map_files_info *p; |
640708a2 PE |
2171 | int ret; |
2172 | ||
94f8f3b0 KO |
2173 | genradix_init(&fa); |
2174 | ||
640708a2 | 2175 | ret = -ENOENT; |
f0c3b509 | 2176 | task = get_proc_task(file_inode(file)); |
640708a2 PE |
2177 | if (!task) |
2178 | goto out; | |
2179 | ||
2180 | ret = -EACCES; | |
caaee623 | 2181 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) |
640708a2 PE |
2182 | goto out_put_task; |
2183 | ||
2184 | ret = 0; | |
f0c3b509 AV |
2185 | if (!dir_emit_dots(file, ctx)) |
2186 | goto out_put_task; | |
640708a2 | 2187 | |
f0c3b509 AV |
2188 | mm = get_task_mm(task); |
2189 | if (!mm) | |
2190 | goto out_put_task; | |
cd9e2bb8 KK |
2191 | |
2192 | ret = down_read_killable(&mm->mmap_sem); | |
2193 | if (ret) { | |
2194 | mmput(mm); | |
2195 | goto out_put_task; | |
2196 | } | |
640708a2 | 2197 | |
f0c3b509 | 2198 | nr_files = 0; |
640708a2 | 2199 | |
f0c3b509 AV |
2200 | /* |
2201 | * We need two passes here: | |
2202 | * | |
2203 | * 1) Collect vmas of mapped files with mmap_sem taken | |
2204 | * 2) Release mmap_sem and instantiate entries | |
2205 | * | |
2206 | * otherwise we get lockdep complained, since filldir() | |
2207 | * routine might require mmap_sem taken in might_fault(). | |
2208 | */ | |
640708a2 | 2209 | |
f0c3b509 | 2210 | for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) { |
94f8f3b0 KO |
2211 | if (!vma->vm_file) |
2212 | continue; | |
2213 | if (++pos <= ctx->pos) | |
2214 | continue; | |
f0c3b509 | 2215 | |
94f8f3b0 KO |
2216 | p = genradix_ptr_alloc(&fa, nr_files++, GFP_KERNEL); |
2217 | if (!p) { | |
f0c3b509 | 2218 | ret = -ENOMEM; |
f0c3b509 AV |
2219 | up_read(&mm->mmap_sem); |
2220 | mmput(mm); | |
2221 | goto out_put_task; | |
640708a2 | 2222 | } |
f0c3b509 | 2223 | |
94f8f3b0 KO |
2224 | p->start = vma->vm_start; |
2225 | p->end = vma->vm_end; | |
2226 | p->mode = vma->vm_file->f_mode; | |
640708a2 | 2227 | } |
f0c3b509 | 2228 | up_read(&mm->mmap_sem); |
fe079a5e | 2229 | mmput(mm); |
f0c3b509 AV |
2230 | |
2231 | for (i = 0; i < nr_files; i++) { | |
20d28cde AD |
2232 | char buf[4 * sizeof(long) + 2]; /* max: %lx-%lx\0 */ |
2233 | unsigned int len; | |
2234 | ||
94f8f3b0 | 2235 | p = genradix_ptr(&fa, i); |
20d28cde | 2236 | len = snprintf(buf, sizeof(buf), "%lx-%lx", p->start, p->end); |
f0c3b509 | 2237 | if (!proc_fill_cache(file, ctx, |
20d28cde | 2238 | buf, len, |
f0c3b509 AV |
2239 | proc_map_files_instantiate, |
2240 | task, | |
2241 | (void *)(unsigned long)p->mode)) | |
2242 | break; | |
2243 | ctx->pos++; | |
640708a2 PE |
2244 | } |
2245 | ||
640708a2 PE |
2246 | out_put_task: |
2247 | put_task_struct(task); | |
2248 | out: | |
94f8f3b0 | 2249 | genradix_free(&fa); |
640708a2 PE |
2250 | return ret; |
2251 | } | |
2252 | ||
2253 | static const struct file_operations proc_map_files_operations = { | |
2254 | .read = generic_read_dir, | |
f50752ea AV |
2255 | .iterate_shared = proc_map_files_readdir, |
2256 | .llseek = generic_file_llseek, | |
640708a2 PE |
2257 | }; |
2258 | ||
b18b6a9c | 2259 | #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS) |
48f6a7a5 PE |
2260 | struct timers_private { |
2261 | struct pid *pid; | |
2262 | struct task_struct *task; | |
2263 | struct sighand_struct *sighand; | |
57b8015e | 2264 | struct pid_namespace *ns; |
48f6a7a5 PE |
2265 | unsigned long flags; |
2266 | }; | |
2267 | ||
2268 | static void *timers_start(struct seq_file *m, loff_t *pos) | |
2269 | { | |
2270 | struct timers_private *tp = m->private; | |
2271 | ||
2272 | tp->task = get_pid_task(tp->pid, PIDTYPE_PID); | |
2273 | if (!tp->task) | |
2274 | return ERR_PTR(-ESRCH); | |
2275 | ||
2276 | tp->sighand = lock_task_sighand(tp->task, &tp->flags); | |
2277 | if (!tp->sighand) | |
2278 | return ERR_PTR(-ESRCH); | |
2279 | ||
2280 | return seq_list_start(&tp->task->signal->posix_timers, *pos); | |
2281 | } | |
2282 | ||
2283 | static void *timers_next(struct seq_file *m, void *v, loff_t *pos) | |
2284 | { | |
2285 | struct timers_private *tp = m->private; | |
2286 | return seq_list_next(v, &tp->task->signal->posix_timers, pos); | |
2287 | } | |
2288 | ||
2289 | static void timers_stop(struct seq_file *m, void *v) | |
2290 | { | |
2291 | struct timers_private *tp = m->private; | |
2292 | ||
2293 | if (tp->sighand) { | |
2294 | unlock_task_sighand(tp->task, &tp->flags); | |
2295 | tp->sighand = NULL; | |
2296 | } | |
2297 | ||
2298 | if (tp->task) { | |
2299 | put_task_struct(tp->task); | |
2300 | tp->task = NULL; | |
2301 | } | |
2302 | } | |
2303 | ||
2304 | static int show_timer(struct seq_file *m, void *v) | |
2305 | { | |
2306 | struct k_itimer *timer; | |
57b8015e PE |
2307 | struct timers_private *tp = m->private; |
2308 | int notify; | |
cedbccab | 2309 | static const char * const nstr[] = { |
57b8015e PE |
2310 | [SIGEV_SIGNAL] = "signal", |
2311 | [SIGEV_NONE] = "none", | |
2312 | [SIGEV_THREAD] = "thread", | |
2313 | }; | |
48f6a7a5 PE |
2314 | |
2315 | timer = list_entry((struct list_head *)v, struct k_itimer, list); | |
57b8015e PE |
2316 | notify = timer->it_sigev_notify; |
2317 | ||
48f6a7a5 | 2318 | seq_printf(m, "ID: %d\n", timer->it_id); |
ba3edf1f | 2319 | seq_printf(m, "signal: %d/%px\n", |
25ce3191 JP |
2320 | timer->sigq->info.si_signo, |
2321 | timer->sigq->info.si_value.sival_ptr); | |
57b8015e | 2322 | seq_printf(m, "notify: %s/%s.%d\n", |
25ce3191 JP |
2323 | nstr[notify & ~SIGEV_THREAD_ID], |
2324 | (notify & SIGEV_THREAD_ID) ? "tid" : "pid", | |
2325 | pid_nr_ns(timer->it_pid, tp->ns)); | |
15ef0298 | 2326 | seq_printf(m, "ClockID: %d\n", timer->it_clock); |
48f6a7a5 PE |
2327 | |
2328 | return 0; | |
2329 | } | |
2330 | ||
2331 | static const struct seq_operations proc_timers_seq_ops = { | |
2332 | .start = timers_start, | |
2333 | .next = timers_next, | |
2334 | .stop = timers_stop, | |
2335 | .show = show_timer, | |
2336 | }; | |
2337 | ||
2338 | static int proc_timers_open(struct inode *inode, struct file *file) | |
2339 | { | |
2340 | struct timers_private *tp; | |
2341 | ||
2342 | tp = __seq_open_private(file, &proc_timers_seq_ops, | |
2343 | sizeof(struct timers_private)); | |
2344 | if (!tp) | |
2345 | return -ENOMEM; | |
2346 | ||
2347 | tp->pid = proc_pid(inode); | |
76f668be | 2348 | tp->ns = proc_pid_ns(inode); |
48f6a7a5 PE |
2349 | return 0; |
2350 | } | |
2351 | ||
2352 | static const struct file_operations proc_timers_operations = { | |
2353 | .open = proc_timers_open, | |
2354 | .read = seq_read, | |
2355 | .llseek = seq_lseek, | |
2356 | .release = seq_release_private, | |
2357 | }; | |
b5946bea | 2358 | #endif |
640708a2 | 2359 | |
5de23d43 JS |
2360 | static ssize_t timerslack_ns_write(struct file *file, const char __user *buf, |
2361 | size_t count, loff_t *offset) | |
2362 | { | |
2363 | struct inode *inode = file_inode(file); | |
2364 | struct task_struct *p; | |
2365 | u64 slack_ns; | |
2366 | int err; | |
2367 | ||
2368 | err = kstrtoull_from_user(buf, count, 10, &slack_ns); | |
2369 | if (err < 0) | |
2370 | return err; | |
2371 | ||
2372 | p = get_proc_task(inode); | |
2373 | if (!p) | |
2374 | return -ESRCH; | |
2375 | ||
4b2bd5fe | 2376 | if (p != current) { |
8da0b4f6 BG |
2377 | rcu_read_lock(); |
2378 | if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) { | |
2379 | rcu_read_unlock(); | |
4b2bd5fe JS |
2380 | count = -EPERM; |
2381 | goto out; | |
2382 | } | |
8da0b4f6 | 2383 | rcu_read_unlock(); |
5de23d43 | 2384 | |
4b2bd5fe JS |
2385 | err = security_task_setscheduler(p); |
2386 | if (err) { | |
2387 | count = err; | |
2388 | goto out; | |
2389 | } | |
904763e1 JS |
2390 | } |
2391 | ||
7abbaf94 JS |
2392 | task_lock(p); |
2393 | if (slack_ns == 0) | |
2394 | p->timer_slack_ns = p->default_timer_slack_ns; | |
2395 | else | |
2396 | p->timer_slack_ns = slack_ns; | |
2397 | task_unlock(p); | |
2398 | ||
2399 | out: | |
5de23d43 JS |
2400 | put_task_struct(p); |
2401 | ||
2402 | return count; | |
2403 | } | |
2404 | ||
2405 | static int timerslack_ns_show(struct seq_file *m, void *v) | |
2406 | { | |
2407 | struct inode *inode = m->private; | |
2408 | struct task_struct *p; | |
7abbaf94 | 2409 | int err = 0; |
5de23d43 JS |
2410 | |
2411 | p = get_proc_task(inode); | |
2412 | if (!p) | |
2413 | return -ESRCH; | |
2414 | ||
4b2bd5fe | 2415 | if (p != current) { |
8da0b4f6 BG |
2416 | rcu_read_lock(); |
2417 | if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) { | |
2418 | rcu_read_unlock(); | |
4b2bd5fe JS |
2419 | err = -EPERM; |
2420 | goto out; | |
2421 | } | |
8da0b4f6 BG |
2422 | rcu_read_unlock(); |
2423 | ||
4b2bd5fe JS |
2424 | err = security_task_getscheduler(p); |
2425 | if (err) | |
2426 | goto out; | |
2427 | } | |
904763e1 | 2428 | |
7abbaf94 JS |
2429 | task_lock(p); |
2430 | seq_printf(m, "%llu\n", p->timer_slack_ns); | |
2431 | task_unlock(p); | |
2432 | ||
2433 | out: | |
5de23d43 JS |
2434 | put_task_struct(p); |
2435 | ||
2436 | return err; | |
2437 | } | |
2438 | ||
2439 | static int timerslack_ns_open(struct inode *inode, struct file *filp) | |
2440 | { | |
2441 | return single_open(filp, timerslack_ns_show, inode); | |
2442 | } | |
2443 | ||
2444 | static const struct file_operations proc_pid_set_timerslack_ns_operations = { | |
2445 | .open = timerslack_ns_open, | |
2446 | .read = seq_read, | |
2447 | .write = timerslack_ns_write, | |
2448 | .llseek = seq_lseek, | |
2449 | .release = single_release, | |
2450 | }; | |
2451 | ||
0168b9e3 AV |
2452 | static struct dentry *proc_pident_instantiate(struct dentry *dentry, |
2453 | struct task_struct *task, const void *ptr) | |
444ceed8 | 2454 | { |
c5141e6d | 2455 | const struct pid_entry *p = ptr; |
444ceed8 EB |
2456 | struct inode *inode; |
2457 | struct proc_inode *ei; | |
444ceed8 | 2458 | |
0168b9e3 | 2459 | inode = proc_pid_make_inode(dentry->d_sb, task, p->mode); |
444ceed8 | 2460 | if (!inode) |
0168b9e3 | 2461 | return ERR_PTR(-ENOENT); |
444ceed8 EB |
2462 | |
2463 | ei = PROC_I(inode); | |
444ceed8 | 2464 | if (S_ISDIR(inode->i_mode)) |
bfe86848 | 2465 | set_nlink(inode, 2); /* Use getattr to fix if necessary */ |
444ceed8 EB |
2466 | if (p->iop) |
2467 | inode->i_op = p->iop; | |
2468 | if (p->fop) | |
2469 | inode->i_fop = p->fop; | |
2470 | ei->op = p->op; | |
1bbc5513 | 2471 | pid_update_inode(task, inode); |
fb045adb | 2472 | d_set_d_op(dentry, &pid_dentry_operations); |
0168b9e3 | 2473 | return d_splice_alias(inode, dentry); |
444ceed8 EB |
2474 | } |
2475 | ||
1da177e4 LT |
2476 | static struct dentry *proc_pident_lookup(struct inode *dir, |
2477 | struct dentry *dentry, | |
d5a572a4 AD |
2478 | const struct pid_entry *p, |
2479 | const struct pid_entry *end) | |
1da177e4 | 2480 | { |
99f89551 | 2481 | struct task_struct *task = get_proc_task(dir); |
0168b9e3 | 2482 | struct dentry *res = ERR_PTR(-ENOENT); |
1da177e4 | 2483 | |
99f89551 EB |
2484 | if (!task) |
2485 | goto out_no_task; | |
1da177e4 | 2486 | |
20cdc894 EB |
2487 | /* |
2488 | * Yes, it does not scale. And it should not. Don't add | |
2489 | * new entries into /proc/<tgid>/ without very good reasons. | |
2490 | */ | |
d5a572a4 | 2491 | for (; p < end; p++) { |
1da177e4 LT |
2492 | if (p->len != dentry->d_name.len) |
2493 | continue; | |
26b95137 AD |
2494 | if (!memcmp(dentry->d_name.name, p->name, p->len)) { |
2495 | res = proc_pident_instantiate(dentry, task, p); | |
1da177e4 | 2496 | break; |
26b95137 | 2497 | } |
1da177e4 | 2498 | } |
99f89551 EB |
2499 | put_task_struct(task); |
2500 | out_no_task: | |
0168b9e3 | 2501 | return res; |
1da177e4 LT |
2502 | } |
2503 | ||
f0c3b509 | 2504 | static int proc_pident_readdir(struct file *file, struct dir_context *ctx, |
c5141e6d | 2505 | const struct pid_entry *ents, unsigned int nents) |
28a6d671 | 2506 | { |
f0c3b509 AV |
2507 | struct task_struct *task = get_proc_task(file_inode(file)); |
2508 | const struct pid_entry *p; | |
28a6d671 | 2509 | |
28a6d671 | 2510 | if (!task) |
f0c3b509 | 2511 | return -ENOENT; |
28a6d671 | 2512 | |
f0c3b509 AV |
2513 | if (!dir_emit_dots(file, ctx)) |
2514 | goto out; | |
2515 | ||
2516 | if (ctx->pos >= nents + 2) | |
2517 | goto out; | |
28a6d671 | 2518 | |
bac5f5d5 | 2519 | for (p = ents + (ctx->pos - 2); p < ents + nents; p++) { |
f0c3b509 AV |
2520 | if (!proc_fill_cache(file, ctx, p->name, p->len, |
2521 | proc_pident_instantiate, task, p)) | |
2522 | break; | |
2523 | ctx->pos++; | |
2524 | } | |
28a6d671 | 2525 | out: |
61a28784 | 2526 | put_task_struct(task); |
f0c3b509 | 2527 | return 0; |
1da177e4 LT |
2528 | } |
2529 | ||
28a6d671 EB |
2530 | #ifdef CONFIG_SECURITY |
2531 | static ssize_t proc_pid_attr_read(struct file * file, char __user * buf, | |
2532 | size_t count, loff_t *ppos) | |
2533 | { | |
496ad9aa | 2534 | struct inode * inode = file_inode(file); |
04ff9708 | 2535 | char *p = NULL; |
28a6d671 EB |
2536 | ssize_t length; |
2537 | struct task_struct *task = get_proc_task(inode); | |
2538 | ||
28a6d671 | 2539 | if (!task) |
04ff9708 | 2540 | return -ESRCH; |
28a6d671 | 2541 | |
6d9c939d | 2542 | length = security_getprocattr(task, PROC_I(inode)->op.lsm, |
2fddfeef | 2543 | (char*)file->f_path.dentry->d_name.name, |
04ff9708 | 2544 | &p); |
28a6d671 | 2545 | put_task_struct(task); |
04ff9708 AV |
2546 | if (length > 0) |
2547 | length = simple_read_from_buffer(buf, count, ppos, p, length); | |
2548 | kfree(p); | |
28a6d671 | 2549 | return length; |
1da177e4 LT |
2550 | } |
2551 | ||
28a6d671 EB |
2552 | static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf, |
2553 | size_t count, loff_t *ppos) | |
2554 | { | |
496ad9aa | 2555 | struct inode * inode = file_inode(file); |
41089b6d | 2556 | struct task_struct *task; |
bb646cdb | 2557 | void *page; |
41089b6d | 2558 | int rv; |
b21507e2 | 2559 | |
41089b6d AD |
2560 | rcu_read_lock(); |
2561 | task = pid_task(proc_pid(inode), PIDTYPE_PID); | |
2562 | if (!task) { | |
2563 | rcu_read_unlock(); | |
2564 | return -ESRCH; | |
2565 | } | |
b21507e2 | 2566 | /* A task may only write its own attributes. */ |
41089b6d AD |
2567 | if (current != task) { |
2568 | rcu_read_unlock(); | |
2569 | return -EACCES; | |
2570 | } | |
35a196be PM |
2571 | /* Prevent changes to overridden credentials. */ |
2572 | if (current_cred() != current_real_cred()) { | |
2573 | rcu_read_unlock(); | |
2574 | return -EBUSY; | |
2575 | } | |
41089b6d | 2576 | rcu_read_unlock(); |
b21507e2 | 2577 | |
28a6d671 EB |
2578 | if (count > PAGE_SIZE) |
2579 | count = PAGE_SIZE; | |
2580 | ||
2581 | /* No partial writes. */ | |
28a6d671 | 2582 | if (*ppos != 0) |
41089b6d | 2583 | return -EINVAL; |
28a6d671 | 2584 | |
bb646cdb AV |
2585 | page = memdup_user(buf, count); |
2586 | if (IS_ERR(page)) { | |
41089b6d | 2587 | rv = PTR_ERR(page); |
28a6d671 | 2588 | goto out; |
bb646cdb | 2589 | } |
28a6d671 | 2590 | |
107db7c7 | 2591 | /* Guard against adverse ptrace interaction */ |
41089b6d AD |
2592 | rv = mutex_lock_interruptible(¤t->signal->cred_guard_mutex); |
2593 | if (rv < 0) | |
107db7c7 DH |
2594 | goto out_free; |
2595 | ||
6d9c939d CS |
2596 | rv = security_setprocattr(PROC_I(inode)->op.lsm, |
2597 | file->f_path.dentry->d_name.name, page, | |
2598 | count); | |
b21507e2 | 2599 | mutex_unlock(¤t->signal->cred_guard_mutex); |
28a6d671 | 2600 | out_free: |
bb646cdb | 2601 | kfree(page); |
28a6d671 | 2602 | out: |
41089b6d | 2603 | return rv; |
28a6d671 EB |
2604 | } |
2605 | ||
00977a59 | 2606 | static const struct file_operations proc_pid_attr_operations = { |
28a6d671 EB |
2607 | .read = proc_pid_attr_read, |
2608 | .write = proc_pid_attr_write, | |
87df8424 | 2609 | .llseek = generic_file_llseek, |
28a6d671 EB |
2610 | }; |
2611 | ||
6d9c939d CS |
2612 | #define LSM_DIR_OPS(LSM) \ |
2613 | static int proc_##LSM##_attr_dir_iterate(struct file *filp, \ | |
2614 | struct dir_context *ctx) \ | |
2615 | { \ | |
2616 | return proc_pident_readdir(filp, ctx, \ | |
2617 | LSM##_attr_dir_stuff, \ | |
2618 | ARRAY_SIZE(LSM##_attr_dir_stuff)); \ | |
2619 | } \ | |
2620 | \ | |
2621 | static const struct file_operations proc_##LSM##_attr_dir_ops = { \ | |
2622 | .read = generic_read_dir, \ | |
2623 | .iterate = proc_##LSM##_attr_dir_iterate, \ | |
2624 | .llseek = default_llseek, \ | |
2625 | }; \ | |
2626 | \ | |
2627 | static struct dentry *proc_##LSM##_attr_dir_lookup(struct inode *dir, \ | |
2628 | struct dentry *dentry, unsigned int flags) \ | |
2629 | { \ | |
2630 | return proc_pident_lookup(dir, dentry, \ | |
2631 | LSM##_attr_dir_stuff, \ | |
d5a572a4 | 2632 | LSM##_attr_dir_stuff + ARRAY_SIZE(LSM##_attr_dir_stuff)); \ |
6d9c939d CS |
2633 | } \ |
2634 | \ | |
2635 | static const struct inode_operations proc_##LSM##_attr_dir_inode_ops = { \ | |
2636 | .lookup = proc_##LSM##_attr_dir_lookup, \ | |
2637 | .getattr = pid_getattr, \ | |
2638 | .setattr = proc_setattr, \ | |
2639 | } | |
2640 | ||
2641 | #ifdef CONFIG_SECURITY_SMACK | |
2642 | static const struct pid_entry smack_attr_dir_stuff[] = { | |
2643 | ATTR("smack", "current", 0666), | |
2644 | }; | |
2645 | LSM_DIR_OPS(smack); | |
2646 | #endif | |
2647 | ||
2b892677 JJ |
2648 | #ifdef CONFIG_SECURITY_APPARMOR |
2649 | static const struct pid_entry apparmor_attr_dir_stuff[] = { | |
2650 | ATTR("apparmor", "current", 0666), | |
2651 | ATTR("apparmor", "prev", 0444), | |
2652 | ATTR("apparmor", "exec", 0666), | |
2653 | }; | |
2654 | LSM_DIR_OPS(apparmor); | |
2655 | #endif | |
2656 | ||
c5141e6d | 2657 | static const struct pid_entry attr_dir_stuff[] = { |
6d9c939d CS |
2658 | ATTR(NULL, "current", 0666), |
2659 | ATTR(NULL, "prev", 0444), | |
2660 | ATTR(NULL, "exec", 0666), | |
2661 | ATTR(NULL, "fscreate", 0666), | |
2662 | ATTR(NULL, "keycreate", 0666), | |
2663 | ATTR(NULL, "sockcreate", 0666), | |
502fb1b1 | 2664 | ATTR(NULL, "display", 0666), |
6d9c939d CS |
2665 | #ifdef CONFIG_SECURITY_SMACK |
2666 | DIR("smack", 0555, | |
2667 | proc_smack_attr_dir_inode_ops, proc_smack_attr_dir_ops), | |
2668 | #endif | |
2b892677 JJ |
2669 | #ifdef CONFIG_SECURITY_APPARMOR |
2670 | DIR("apparmor", 0555, | |
2671 | proc_apparmor_attr_dir_inode_ops, proc_apparmor_attr_dir_ops), | |
2672 | #endif | |
28a6d671 EB |
2673 | }; |
2674 | ||
f0c3b509 | 2675 | static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx) |
28a6d671 | 2676 | { |
f0c3b509 AV |
2677 | return proc_pident_readdir(file, ctx, |
2678 | attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2679 | } |
2680 | ||
00977a59 | 2681 | static const struct file_operations proc_attr_dir_operations = { |
1da177e4 | 2682 | .read = generic_read_dir, |
f50752ea AV |
2683 | .iterate_shared = proc_attr_dir_readdir, |
2684 | .llseek = generic_file_llseek, | |
1da177e4 LT |
2685 | }; |
2686 | ||
72d9dcfc | 2687 | static struct dentry *proc_attr_dir_lookup(struct inode *dir, |
00cd8dd3 | 2688 | struct dentry *dentry, unsigned int flags) |
28a6d671 | 2689 | { |
7bcd6b0e | 2690 | return proc_pident_lookup(dir, dentry, |
d5a572a4 AD |
2691 | attr_dir_stuff, |
2692 | attr_dir_stuff + ARRAY_SIZE(attr_dir_stuff)); | |
28a6d671 EB |
2693 | } |
2694 | ||
c5ef1c42 | 2695 | static const struct inode_operations proc_attr_dir_inode_operations = { |
72d9dcfc | 2696 | .lookup = proc_attr_dir_lookup, |
99f89551 | 2697 | .getattr = pid_getattr, |
6d76fa58 | 2698 | .setattr = proc_setattr, |
1da177e4 LT |
2699 | }; |
2700 | ||
28a6d671 EB |
2701 | #endif |
2702 | ||
698ba7b5 | 2703 | #ifdef CONFIG_ELF_CORE |
3cb4a0bb KH |
2704 | static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf, |
2705 | size_t count, loff_t *ppos) | |
2706 | { | |
496ad9aa | 2707 | struct task_struct *task = get_proc_task(file_inode(file)); |
3cb4a0bb KH |
2708 | struct mm_struct *mm; |
2709 | char buffer[PROC_NUMBUF]; | |
2710 | size_t len; | |
2711 | int ret; | |
2712 | ||
2713 | if (!task) | |
2714 | return -ESRCH; | |
2715 | ||
2716 | ret = 0; | |
2717 | mm = get_task_mm(task); | |
2718 | if (mm) { | |
2719 | len = snprintf(buffer, sizeof(buffer), "%08lx\n", | |
2720 | ((mm->flags & MMF_DUMP_FILTER_MASK) >> | |
2721 | MMF_DUMP_FILTER_SHIFT)); | |
2722 | mmput(mm); | |
2723 | ret = simple_read_from_buffer(buf, count, ppos, buffer, len); | |
2724 | } | |
2725 | ||
2726 | put_task_struct(task); | |
2727 | ||
2728 | return ret; | |
2729 | } | |
2730 | ||
2731 | static ssize_t proc_coredump_filter_write(struct file *file, | |
2732 | const char __user *buf, | |
2733 | size_t count, | |
2734 | loff_t *ppos) | |
2735 | { | |
2736 | struct task_struct *task; | |
2737 | struct mm_struct *mm; | |
3cb4a0bb KH |
2738 | unsigned int val; |
2739 | int ret; | |
2740 | int i; | |
2741 | unsigned long mask; | |
2742 | ||
774636e1 AD |
2743 | ret = kstrtouint_from_user(buf, count, 0, &val); |
2744 | if (ret < 0) | |
2745 | return ret; | |
3cb4a0bb KH |
2746 | |
2747 | ret = -ESRCH; | |
496ad9aa | 2748 | task = get_proc_task(file_inode(file)); |
3cb4a0bb KH |
2749 | if (!task) |
2750 | goto out_no_task; | |
2751 | ||
3cb4a0bb KH |
2752 | mm = get_task_mm(task); |
2753 | if (!mm) | |
2754 | goto out_no_mm; | |
41a0c249 | 2755 | ret = 0; |
3cb4a0bb KH |
2756 | |
2757 | for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) { | |
2758 | if (val & mask) | |
2759 | set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2760 | else | |
2761 | clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags); | |
2762 | } | |
2763 | ||
2764 | mmput(mm); | |
2765 | out_no_mm: | |
2766 | put_task_struct(task); | |
2767 | out_no_task: | |
774636e1 AD |
2768 | if (ret < 0) |
2769 | return ret; | |
2770 | return count; | |
3cb4a0bb KH |
2771 | } |
2772 | ||
2773 | static const struct file_operations proc_coredump_filter_operations = { | |
2774 | .read = proc_coredump_filter_read, | |
2775 | .write = proc_coredump_filter_write, | |
87df8424 | 2776 | .llseek = generic_file_llseek, |
3cb4a0bb KH |
2777 | }; |
2778 | #endif | |
2779 | ||
aba76fdb | 2780 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 2781 | static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole) |
297c5d92 | 2782 | { |
940389b8 | 2783 | struct task_io_accounting acct = task->ioac; |
5995477a | 2784 | unsigned long flags; |
293eb1e7 | 2785 | int result; |
5995477a | 2786 | |
293eb1e7 VK |
2787 | result = mutex_lock_killable(&task->signal->cred_guard_mutex); |
2788 | if (result) | |
2789 | return result; | |
2790 | ||
caaee623 | 2791 | if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) { |
293eb1e7 VK |
2792 | result = -EACCES; |
2793 | goto out_unlock; | |
2794 | } | |
1d1221f3 | 2795 | |
5995477a AR |
2796 | if (whole && lock_task_sighand(task, &flags)) { |
2797 | struct task_struct *t = task; | |
2798 | ||
2799 | task_io_accounting_add(&acct, &task->signal->ioac); | |
2800 | while_each_thread(task, t) | |
2801 | task_io_accounting_add(&acct, &t->ioac); | |
2802 | ||
2803 | unlock_task_sighand(task, &flags); | |
297c5d92 | 2804 | } |
25ce3191 JP |
2805 | seq_printf(m, |
2806 | "rchar: %llu\n" | |
2807 | "wchar: %llu\n" | |
2808 | "syscr: %llu\n" | |
2809 | "syscw: %llu\n" | |
2810 | "read_bytes: %llu\n" | |
2811 | "write_bytes: %llu\n" | |
2812 | "cancelled_write_bytes: %llu\n", | |
2813 | (unsigned long long)acct.rchar, | |
2814 | (unsigned long long)acct.wchar, | |
2815 | (unsigned long long)acct.syscr, | |
2816 | (unsigned long long)acct.syscw, | |
2817 | (unsigned long long)acct.read_bytes, | |
2818 | (unsigned long long)acct.write_bytes, | |
2819 | (unsigned long long)acct.cancelled_write_bytes); | |
2820 | result = 0; | |
2821 | ||
293eb1e7 VK |
2822 | out_unlock: |
2823 | mutex_unlock(&task->signal->cred_guard_mutex); | |
2824 | return result; | |
297c5d92 AR |
2825 | } |
2826 | ||
19aadc98 AD |
2827 | static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns, |
2828 | struct pid *pid, struct task_struct *task) | |
297c5d92 | 2829 | { |
19aadc98 | 2830 | return do_io_accounting(task, m, 0); |
aba76fdb | 2831 | } |
297c5d92 | 2832 | |
19aadc98 AD |
2833 | static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns, |
2834 | struct pid *pid, struct task_struct *task) | |
297c5d92 | 2835 | { |
19aadc98 | 2836 | return do_io_accounting(task, m, 1); |
297c5d92 AR |
2837 | } |
2838 | #endif /* CONFIG_TASK_IO_ACCOUNTING */ | |
aba76fdb | 2839 | |
22d917d8 EB |
2840 | #ifdef CONFIG_USER_NS |
2841 | static int proc_id_map_open(struct inode *inode, struct file *file, | |
ccf94f1b | 2842 | const struct seq_operations *seq_ops) |
22d917d8 EB |
2843 | { |
2844 | struct user_namespace *ns = NULL; | |
2845 | struct task_struct *task; | |
2846 | struct seq_file *seq; | |
2847 | int ret = -EINVAL; | |
2848 | ||
2849 | task = get_proc_task(inode); | |
2850 | if (task) { | |
2851 | rcu_read_lock(); | |
2852 | ns = get_user_ns(task_cred_xxx(task, user_ns)); | |
2853 | rcu_read_unlock(); | |
2854 | put_task_struct(task); | |
2855 | } | |
2856 | if (!ns) | |
2857 | goto err; | |
2858 | ||
2859 | ret = seq_open(file, seq_ops); | |
2860 | if (ret) | |
2861 | goto err_put_ns; | |
2862 | ||
2863 | seq = file->private_data; | |
2864 | seq->private = ns; | |
2865 | ||
2866 | return 0; | |
2867 | err_put_ns: | |
2868 | put_user_ns(ns); | |
2869 | err: | |
2870 | return ret; | |
2871 | } | |
2872 | ||
2873 | static int proc_id_map_release(struct inode *inode, struct file *file) | |
2874 | { | |
2875 | struct seq_file *seq = file->private_data; | |
2876 | struct user_namespace *ns = seq->private; | |
2877 | put_user_ns(ns); | |
2878 | return seq_release(inode, file); | |
2879 | } | |
2880 | ||
2881 | static int proc_uid_map_open(struct inode *inode, struct file *file) | |
2882 | { | |
2883 | return proc_id_map_open(inode, file, &proc_uid_seq_operations); | |
2884 | } | |
2885 | ||
2886 | static int proc_gid_map_open(struct inode *inode, struct file *file) | |
2887 | { | |
2888 | return proc_id_map_open(inode, file, &proc_gid_seq_operations); | |
2889 | } | |
2890 | ||
f76d207a EB |
2891 | static int proc_projid_map_open(struct inode *inode, struct file *file) |
2892 | { | |
2893 | return proc_id_map_open(inode, file, &proc_projid_seq_operations); | |
2894 | } | |
2895 | ||
22d917d8 EB |
2896 | static const struct file_operations proc_uid_map_operations = { |
2897 | .open = proc_uid_map_open, | |
2898 | .write = proc_uid_map_write, | |
2899 | .read = seq_read, | |
2900 | .llseek = seq_lseek, | |
2901 | .release = proc_id_map_release, | |
2902 | }; | |
2903 | ||
2904 | static const struct file_operations proc_gid_map_operations = { | |
2905 | .open = proc_gid_map_open, | |
2906 | .write = proc_gid_map_write, | |
2907 | .read = seq_read, | |
2908 | .llseek = seq_lseek, | |
2909 | .release = proc_id_map_release, | |
2910 | }; | |
f76d207a EB |
2911 | |
2912 | static const struct file_operations proc_projid_map_operations = { | |
2913 | .open = proc_projid_map_open, | |
2914 | .write = proc_projid_map_write, | |
2915 | .read = seq_read, | |
2916 | .llseek = seq_lseek, | |
2917 | .release = proc_id_map_release, | |
2918 | }; | |
9cc46516 EB |
2919 | |
2920 | static int proc_setgroups_open(struct inode *inode, struct file *file) | |
2921 | { | |
2922 | struct user_namespace *ns = NULL; | |
2923 | struct task_struct *task; | |
2924 | int ret; | |
2925 | ||
2926 | ret = -ESRCH; | |
2927 | task = get_proc_task(inode); | |
2928 | if (task) { | |
2929 | rcu_read_lock(); | |
2930 | ns = get_user_ns(task_cred_xxx(task, user_ns)); | |
2931 | rcu_read_unlock(); | |
2932 | put_task_struct(task); | |
2933 | } | |
2934 | if (!ns) | |
2935 | goto err; | |
2936 | ||
2937 | if (file->f_mode & FMODE_WRITE) { | |
2938 | ret = -EACCES; | |
2939 | if (!ns_capable(ns, CAP_SYS_ADMIN)) | |
2940 | goto err_put_ns; | |
2941 | } | |
2942 | ||
2943 | ret = single_open(file, &proc_setgroups_show, ns); | |
2944 | if (ret) | |
2945 | goto err_put_ns; | |
2946 | ||
2947 | return 0; | |
2948 | err_put_ns: | |
2949 | put_user_ns(ns); | |
2950 | err: | |
2951 | return ret; | |
2952 | } | |
2953 | ||
2954 | static int proc_setgroups_release(struct inode *inode, struct file *file) | |
2955 | { | |
2956 | struct seq_file *seq = file->private_data; | |
2957 | struct user_namespace *ns = seq->private; | |
2958 | int ret = single_release(inode, file); | |
2959 | put_user_ns(ns); | |
2960 | return ret; | |
2961 | } | |
2962 | ||
2963 | static const struct file_operations proc_setgroups_operations = { | |
2964 | .open = proc_setgroups_open, | |
2965 | .write = proc_setgroups_write, | |
2966 | .read = seq_read, | |
2967 | .llseek = seq_lseek, | |
2968 | .release = proc_setgroups_release, | |
2969 | }; | |
22d917d8 EB |
2970 | #endif /* CONFIG_USER_NS */ |
2971 | ||
47830723 KC |
2972 | static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns, |
2973 | struct pid *pid, struct task_struct *task) | |
2974 | { | |
a9712bc1 AV |
2975 | int err = lock_trace(task); |
2976 | if (!err) { | |
2977 | seq_printf(m, "%08x\n", task->personality); | |
2978 | unlock_trace(task); | |
2979 | } | |
2980 | return err; | |
47830723 KC |
2981 | } |
2982 | ||
7c23b330 JP |
2983 | #ifdef CONFIG_LIVEPATCH |
2984 | static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns, | |
2985 | struct pid *pid, struct task_struct *task) | |
2986 | { | |
2987 | seq_printf(m, "%d\n", task->patch_state); | |
2988 | return 0; | |
2989 | } | |
2990 | #endif /* CONFIG_LIVEPATCH */ | |
2991 | ||
c8d12627 AP |
2992 | #ifdef CONFIG_STACKLEAK_METRICS |
2993 | static int proc_stack_depth(struct seq_file *m, struct pid_namespace *ns, | |
2994 | struct pid *pid, struct task_struct *task) | |
2995 | { | |
2996 | unsigned long prev_depth = THREAD_SIZE - | |
2997 | (task->prev_lowest_stack & (THREAD_SIZE - 1)); | |
2998 | unsigned long depth = THREAD_SIZE - | |
2999 | (task->lowest_stack & (THREAD_SIZE - 1)); | |
3000 | ||
3001 | seq_printf(m, "previous stack depth: %lu\nstack depth: %lu\n", | |
3002 | prev_depth, depth); | |
3003 | return 0; | |
3004 | } | |
3005 | #endif /* CONFIG_STACKLEAK_METRICS */ | |
3006 | ||
28a6d671 EB |
3007 | /* |
3008 | * Thread groups | |
3009 | */ | |
00977a59 | 3010 | static const struct file_operations proc_task_operations; |
c5ef1c42 | 3011 | static const struct inode_operations proc_task_inode_operations; |
20cdc894 | 3012 | |
c5141e6d | 3013 | static const struct pid_entry tgid_base_stuff[] = { |
631f9c18 AD |
3014 | DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations), |
3015 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), | |
640708a2 | 3016 | DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations), |
631f9c18 | 3017 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
6b4e306a | 3018 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
b2211a36 | 3019 | #ifdef CONFIG_NET |
631f9c18 | 3020 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), |
b2211a36 | 3021 | #endif |
631f9c18 | 3022 | REG("environ", S_IRUSR, proc_environ_operations), |
c5317167 | 3023 | REG("auxv", S_IRUSR, proc_auxv_operations), |
631f9c18 | 3024 | ONE("status", S_IRUGO, proc_pid_status), |
35a35046 | 3025 | ONE("personality", S_IRUSR, proc_pid_personality), |
1c963eb1 | 3026 | ONE("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 3027 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 3028 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
5091faa4 MG |
3029 | #endif |
3030 | #ifdef CONFIG_SCHED_AUTOGROUP | |
3031 | REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations), | |
ebcb6734 | 3032 | #endif |
4614a696 | 3033 | REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations), |
ebcb6734 | 3034 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 | 3035 | ONE("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 3036 | #endif |
c2c0bb44 | 3037 | REG("cmdline", S_IRUGO, proc_pid_cmdline_ops), |
631f9c18 AD |
3038 | ONE("stat", S_IRUGO, proc_tgid_stat), |
3039 | ONE("statm", S_IRUGO, proc_pid_statm), | |
b7643757 | 3040 | REG("maps", S_IRUGO, proc_pid_maps_operations), |
28a6d671 | 3041 | #ifdef CONFIG_NUMA |
b7643757 | 3042 | REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations), |
28a6d671 | 3043 | #endif |
631f9c18 AD |
3044 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
3045 | LNK("cwd", proc_cwd_link), | |
3046 | LNK("root", proc_root_link), | |
3047 | LNK("exe", proc_exe_link), | |
3048 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
3049 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
3050 | REG("mountstats", S_IRUSR, proc_mountstats_operations), | |
1e883281 | 3051 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 | 3052 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
b7643757 | 3053 | REG("smaps", S_IRUGO, proc_pid_smaps_operations), |
493b0e9d | 3054 | REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations), |
32ed74a4 | 3055 | REG("pagemap", S_IRUSR, proc_pagemap_operations), |
28a6d671 EB |
3056 | #endif |
3057 | #ifdef CONFIG_SECURITY | |
631f9c18 | 3058 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
3059 | #endif |
3060 | #ifdef CONFIG_KALLSYMS | |
edfcd606 | 3061 | ONE("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 3062 | #endif |
2ec220e2 | 3063 | #ifdef CONFIG_STACKTRACE |
35a35046 | 3064 | ONE("stack", S_IRUSR, proc_pid_stack), |
28a6d671 | 3065 | #endif |
5968cece | 3066 | #ifdef CONFIG_SCHED_INFO |
f6e826ca | 3067 | ONE("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 3068 | #endif |
9745512c | 3069 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 3070 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 3071 | #endif |
8793d854 | 3072 | #ifdef CONFIG_PROC_PID_CPUSET |
52de4779 | 3073 | ONE("cpuset", S_IRUGO, proc_cpuset_show), |
a424316c PM |
3074 | #endif |
3075 | #ifdef CONFIG_CGROUPS | |
006f4ac4 | 3076 | ONE("cgroup", S_IRUGO, proc_cgroup_show), |
28a6d671 | 3077 | #endif |
6ba51e37 | 3078 | ONE("oom_score", S_IRUGO, proc_oom_score), |
fa0cbbf1 | 3079 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations), |
a63d83f4 | 3080 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
4b7d248b | 3081 | #ifdef CONFIG_AUDIT |
631f9c18 AD |
3082 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
3083 | REG("sessionid", S_IRUGO, proc_sessionid_operations), | |
28a6d671 | 3084 | #endif |
f4f154fd | 3085 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 3086 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
168c42bc | 3087 | REG("fail-nth", 0644, proc_fail_nth_operations), |
f4f154fd | 3088 | #endif |
698ba7b5 | 3089 | #ifdef CONFIG_ELF_CORE |
631f9c18 | 3090 | REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations), |
3cb4a0bb | 3091 | #endif |
aba76fdb | 3092 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 3093 | ONE("io", S_IRUSR, proc_tgid_io_accounting), |
aba76fdb | 3094 | #endif |
22d917d8 EB |
3095 | #ifdef CONFIG_USER_NS |
3096 | REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations), | |
3097 | REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations), | |
f76d207a | 3098 | REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations), |
9cc46516 | 3099 | REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations), |
22d917d8 | 3100 | #endif |
b18b6a9c | 3101 | #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS) |
48f6a7a5 PE |
3102 | REG("timers", S_IRUGO, proc_timers_operations), |
3103 | #endif | |
5de23d43 | 3104 | REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations), |
7c23b330 JP |
3105 | #ifdef CONFIG_LIVEPATCH |
3106 | ONE("patch_state", S_IRUSR, proc_pid_patch_state), | |
3107 | #endif | |
c8d12627 AP |
3108 | #ifdef CONFIG_STACKLEAK_METRICS |
3109 | ONE("stack_depth", S_IRUGO, proc_stack_depth), | |
3110 | #endif | |
68bc30bb AL |
3111 | #ifdef CONFIG_PROC_PID_ARCH_STATUS |
3112 | ONE("arch_status", S_IRUGO, proc_pid_arch_status), | |
3113 | #endif | |
28a6d671 | 3114 | }; |
1da177e4 | 3115 | |
f0c3b509 | 3116 | static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3117 | { |
f0c3b509 AV |
3118 | return proc_pident_readdir(file, ctx, |
3119 | tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
3120 | } |
3121 | ||
00977a59 | 3122 | static const struct file_operations proc_tgid_base_operations = { |
1da177e4 | 3123 | .read = generic_read_dir, |
f50752ea AV |
3124 | .iterate_shared = proc_tgid_base_readdir, |
3125 | .llseek = generic_file_llseek, | |
1da177e4 LT |
3126 | }; |
3127 | ||
3eb39f47 CB |
3128 | struct pid *tgid_pidfd_to_pid(const struct file *file) |
3129 | { | |
30d158b1 | 3130 | if (file->f_op != &proc_tgid_base_operations) |
3eb39f47 CB |
3131 | return ERR_PTR(-EBADF); |
3132 | ||
3133 | return proc_pid(file_inode(file)); | |
3134 | } | |
3135 | ||
00cd8dd3 AV |
3136 | static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
3137 | { | |
7bcd6b0e | 3138 | return proc_pident_lookup(dir, dentry, |
d5a572a4 AD |
3139 | tgid_base_stuff, |
3140 | tgid_base_stuff + ARRAY_SIZE(tgid_base_stuff)); | |
1da177e4 LT |
3141 | } |
3142 | ||
c5ef1c42 | 3143 | static const struct inode_operations proc_tgid_base_inode_operations = { |
28a6d671 | 3144 | .lookup = proc_tgid_base_lookup, |
99f89551 | 3145 | .getattr = pid_getattr, |
6d76fa58 | 3146 | .setattr = proc_setattr, |
0499680a | 3147 | .permission = proc_pid_permission, |
1da177e4 | 3148 | }; |
1da177e4 | 3149 | |
60347f67 | 3150 | static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid) |
1da177e4 | 3151 | { |
48e6484d | 3152 | struct dentry *dentry, *leader, *dir; |
e3912ac3 | 3153 | char buf[10 + 1]; |
48e6484d EB |
3154 | struct qstr name; |
3155 | ||
3156 | name.name = buf; | |
e3912ac3 | 3157 | name.len = snprintf(buf, sizeof(buf), "%u", pid); |
4f522a24 | 3158 | /* no ->d_hash() rejects on procfs */ |
60347f67 | 3159 | dentry = d_hash_and_lookup(mnt->mnt_root, &name); |
48e6484d | 3160 | if (dentry) { |
bbd51924 | 3161 | d_invalidate(dentry); |
48e6484d EB |
3162 | dput(dentry); |
3163 | } | |
1da177e4 | 3164 | |
c35a7f18 ON |
3165 | if (pid == tgid) |
3166 | return; | |
3167 | ||
48e6484d | 3168 | name.name = buf; |
e3912ac3 | 3169 | name.len = snprintf(buf, sizeof(buf), "%u", tgid); |
60347f67 | 3170 | leader = d_hash_and_lookup(mnt->mnt_root, &name); |
48e6484d EB |
3171 | if (!leader) |
3172 | goto out; | |
1da177e4 | 3173 | |
48e6484d EB |
3174 | name.name = "task"; |
3175 | name.len = strlen(name.name); | |
3176 | dir = d_hash_and_lookup(leader, &name); | |
3177 | if (!dir) | |
3178 | goto out_put_leader; | |
3179 | ||
3180 | name.name = buf; | |
e3912ac3 | 3181 | name.len = snprintf(buf, sizeof(buf), "%u", pid); |
48e6484d EB |
3182 | dentry = d_hash_and_lookup(dir, &name); |
3183 | if (dentry) { | |
bbd51924 | 3184 | d_invalidate(dentry); |
48e6484d | 3185 | dput(dentry); |
1da177e4 | 3186 | } |
48e6484d EB |
3187 | |
3188 | dput(dir); | |
3189 | out_put_leader: | |
3190 | dput(leader); | |
3191 | out: | |
3192 | return; | |
1da177e4 LT |
3193 | } |
3194 | ||
0895e91d RD |
3195 | /** |
3196 | * proc_flush_task - Remove dcache entries for @task from the /proc dcache. | |
3197 | * @task: task that should be flushed. | |
3198 | * | |
3199 | * When flushing dentries from proc, one needs to flush them from global | |
60347f67 | 3200 | * proc (proc_mnt) and from all the namespaces' procs this task was seen |
0895e91d RD |
3201 | * in. This call is supposed to do all of this job. |
3202 | * | |
3203 | * Looks in the dcache for | |
3204 | * /proc/@pid | |
3205 | * /proc/@tgid/task/@pid | |
3206 | * if either directory is present flushes it and all of it'ts children | |
3207 | * from the dcache. | |
3208 | * | |
3209 | * It is safe and reasonable to cache /proc entries for a task until | |
3210 | * that task exits. After that they just clog up the dcache with | |
3211 | * useless entries, possibly causing useful dcache entries to be | |
3212 | * flushed instead. This routine is proved to flush those useless | |
3213 | * dcache entries at process exit time. | |
3214 | * | |
3215 | * NOTE: This routine is just an optimization so it does not guarantee | |
3216 | * that no dcache entries will exist at process exit time it | |
3217 | * just makes it very unlikely that any will persist. | |
60347f67 PE |
3218 | */ |
3219 | ||
3220 | void proc_flush_task(struct task_struct *task) | |
3221 | { | |
9fcc2d15 | 3222 | int i; |
9b4d1cbe | 3223 | struct pid *pid, *tgid; |
130f77ec PE |
3224 | struct upid *upid; |
3225 | ||
130f77ec | 3226 | pid = task_pid(task); |
9b4d1cbe | 3227 | tgid = task_tgid(task); |
130f77ec | 3228 | |
9fcc2d15 | 3229 | for (i = 0; i <= pid->level; i++) { |
130f77ec PE |
3230 | upid = &pid->numbers[i]; |
3231 | proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr, | |
9b4d1cbe | 3232 | tgid->numbers[i].nr); |
130f77ec | 3233 | } |
60347f67 PE |
3234 | } |
3235 | ||
0168b9e3 | 3236 | static struct dentry *proc_pid_instantiate(struct dentry * dentry, |
c52a47ac | 3237 | struct task_struct *task, const void *ptr) |
444ceed8 | 3238 | { |
444ceed8 EB |
3239 | struct inode *inode; |
3240 | ||
0168b9e3 | 3241 | inode = proc_pid_make_inode(dentry->d_sb, task, S_IFDIR | S_IRUGO | S_IXUGO); |
444ceed8 | 3242 | if (!inode) |
0168b9e3 | 3243 | return ERR_PTR(-ENOENT); |
444ceed8 | 3244 | |
444ceed8 EB |
3245 | inode->i_op = &proc_tgid_base_inode_operations; |
3246 | inode->i_fop = &proc_tgid_base_operations; | |
3247 | inode->i_flags|=S_IMMUTABLE; | |
aed54175 | 3248 | |
1270dd8d | 3249 | set_nlink(inode, nlink_tgid); |
1bbc5513 | 3250 | pid_update_inode(task, inode); |
444ceed8 | 3251 | |
fb045adb | 3252 | d_set_d_op(dentry, &pid_dentry_operations); |
0168b9e3 | 3253 | return d_splice_alias(inode, dentry); |
444ceed8 EB |
3254 | } |
3255 | ||
867aaccf | 3256 | struct dentry *proc_pid_lookup(struct dentry *dentry, unsigned int flags) |
1da177e4 LT |
3257 | { |
3258 | struct task_struct *task; | |
1da177e4 | 3259 | unsigned tgid; |
b488893a | 3260 | struct pid_namespace *ns; |
0168b9e3 | 3261 | struct dentry *result = ERR_PTR(-ENOENT); |
1da177e4 | 3262 | |
dbcdb504 | 3263 | tgid = name_to_int(&dentry->d_name); |
1da177e4 LT |
3264 | if (tgid == ~0U) |
3265 | goto out; | |
3266 | ||
b488893a | 3267 | ns = dentry->d_sb->s_fs_info; |
de758734 | 3268 | rcu_read_lock(); |
b488893a | 3269 | task = find_task_by_pid_ns(tgid, ns); |
1da177e4 LT |
3270 | if (task) |
3271 | get_task_struct(task); | |
de758734 | 3272 | rcu_read_unlock(); |
1da177e4 LT |
3273 | if (!task) |
3274 | goto out; | |
3275 | ||
0168b9e3 | 3276 | result = proc_pid_instantiate(dentry, task, NULL); |
1da177e4 | 3277 | put_task_struct(task); |
1da177e4 | 3278 | out: |
0168b9e3 | 3279 | return result; |
1da177e4 LT |
3280 | } |
3281 | ||
1da177e4 | 3282 | /* |
0804ef4b | 3283 | * Find the first task with tgid >= tgid |
0bc58a91 | 3284 | * |
1da177e4 | 3285 | */ |
19fd4bb2 EB |
3286 | struct tgid_iter { |
3287 | unsigned int tgid; | |
0804ef4b | 3288 | struct task_struct *task; |
19fd4bb2 EB |
3289 | }; |
3290 | static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter) | |
3291 | { | |
0804ef4b | 3292 | struct pid *pid; |
1da177e4 | 3293 | |
19fd4bb2 EB |
3294 | if (iter.task) |
3295 | put_task_struct(iter.task); | |
454cc105 | 3296 | rcu_read_lock(); |
0804ef4b | 3297 | retry: |
19fd4bb2 EB |
3298 | iter.task = NULL; |
3299 | pid = find_ge_pid(iter.tgid, ns); | |
0804ef4b | 3300 | if (pid) { |
19fd4bb2 EB |
3301 | iter.tgid = pid_nr_ns(pid, ns); |
3302 | iter.task = pid_task(pid, PIDTYPE_PID); | |
0804ef4b EB |
3303 | /* What we to know is if the pid we have find is the |
3304 | * pid of a thread_group_leader. Testing for task | |
3305 | * being a thread_group_leader is the obvious thing | |
3306 | * todo but there is a window when it fails, due to | |
3307 | * the pid transfer logic in de_thread. | |
3308 | * | |
3309 | * So we perform the straight forward test of seeing | |
3310 | * if the pid we have found is the pid of a thread | |
3311 | * group leader, and don't worry if the task we have | |
3312 | * found doesn't happen to be a thread group leader. | |
3313 | * As we don't care in the case of readdir. | |
3314 | */ | |
19fd4bb2 EB |
3315 | if (!iter.task || !has_group_leader_pid(iter.task)) { |
3316 | iter.tgid += 1; | |
0804ef4b | 3317 | goto retry; |
19fd4bb2 EB |
3318 | } |
3319 | get_task_struct(iter.task); | |
0bc58a91 | 3320 | } |
454cc105 | 3321 | rcu_read_unlock(); |
19fd4bb2 | 3322 | return iter; |
1da177e4 LT |
3323 | } |
3324 | ||
0097875b | 3325 | #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2) |
0804ef4b | 3326 | |
1da177e4 | 3327 | /* for the /proc/ directory itself, after non-process stuff has been done */ |
f0c3b509 | 3328 | int proc_pid_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3329 | { |
19fd4bb2 | 3330 | struct tgid_iter iter; |
76f668be | 3331 | struct pid_namespace *ns = proc_pid_ns(file_inode(file)); |
f0c3b509 | 3332 | loff_t pos = ctx->pos; |
1da177e4 | 3333 | |
021ada7d | 3334 | if (pos >= PID_MAX_LIMIT + TGID_OFFSET) |
f0c3b509 | 3335 | return 0; |
1da177e4 | 3336 | |
0097875b | 3337 | if (pos == TGID_OFFSET - 2) { |
2b0143b5 | 3338 | struct inode *inode = d_inode(ns->proc_self); |
db963164 | 3339 | if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK)) |
f0c3b509 | 3340 | return 0; |
0097875b EB |
3341 | ctx->pos = pos = pos + 1; |
3342 | } | |
3343 | if (pos == TGID_OFFSET - 1) { | |
2b0143b5 | 3344 | struct inode *inode = d_inode(ns->proc_thread_self); |
0097875b EB |
3345 | if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK)) |
3346 | return 0; | |
3347 | ctx->pos = pos = pos + 1; | |
021ada7d | 3348 | } |
0097875b | 3349 | iter.tgid = pos - TGID_OFFSET; |
19fd4bb2 | 3350 | iter.task = NULL; |
19fd4bb2 EB |
3351 | for (iter = next_tgid(ns, iter); |
3352 | iter.task; | |
3353 | iter.tgid += 1, iter = next_tgid(ns, iter)) { | |
e3912ac3 | 3354 | char name[10 + 1]; |
a4ef3895 | 3355 | unsigned int len; |
3ba4bcee ED |
3356 | |
3357 | cond_resched(); | |
796f571b | 3358 | if (!has_pid_permissions(ns, iter.task, HIDEPID_INVISIBLE)) |
f0c3b509 | 3359 | continue; |
0499680a | 3360 | |
e3912ac3 | 3361 | len = snprintf(name, sizeof(name), "%u", iter.tgid); |
f0c3b509 AV |
3362 | ctx->pos = iter.tgid + TGID_OFFSET; |
3363 | if (!proc_fill_cache(file, ctx, name, len, | |
3364 | proc_pid_instantiate, iter.task, NULL)) { | |
19fd4bb2 | 3365 | put_task_struct(iter.task); |
f0c3b509 | 3366 | return 0; |
1da177e4 | 3367 | } |
0bc58a91 | 3368 | } |
f0c3b509 | 3369 | ctx->pos = PID_MAX_LIMIT + TGID_OFFSET; |
0bc58a91 EB |
3370 | return 0; |
3371 | } | |
1da177e4 | 3372 | |
1b3044e3 JD |
3373 | /* |
3374 | * proc_tid_comm_permission is a special permission function exclusively | |
3375 | * used for the node /proc/<pid>/task/<tid>/comm. | |
3376 | * It bypasses generic permission checks in the case where a task of the same | |
3377 | * task group attempts to access the node. | |
3378 | * The rationale behind this is that glibc and bionic access this node for | |
3379 | * cross thread naming (pthread_set/getname_np(!self)). However, if | |
3380 | * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0, | |
3381 | * which locks out the cross thread naming implementation. | |
3382 | * This function makes sure that the node is always accessible for members of | |
3383 | * same thread group. | |
3384 | */ | |
3385 | static int proc_tid_comm_permission(struct inode *inode, int mask) | |
3386 | { | |
3387 | bool is_same_tgroup; | |
3388 | struct task_struct *task; | |
3389 | ||
3390 | task = get_proc_task(inode); | |
3391 | if (!task) | |
3392 | return -ESRCH; | |
3393 | is_same_tgroup = same_thread_group(current, task); | |
3394 | put_task_struct(task); | |
3395 | ||
3396 | if (likely(is_same_tgroup && !(mask & MAY_EXEC))) { | |
3397 | /* This file (/proc/<pid>/task/<tid>/comm) can always be | |
3398 | * read or written by the members of the corresponding | |
3399 | * thread group. | |
3400 | */ | |
3401 | return 0; | |
3402 | } | |
3403 | ||
3404 | return generic_permission(inode, mask); | |
3405 | } | |
3406 | ||
3407 | static const struct inode_operations proc_tid_comm_inode_operations = { | |
3408 | .permission = proc_tid_comm_permission, | |
3409 | }; | |
3410 | ||
28a6d671 EB |
3411 | /* |
3412 | * Tasks | |
3413 | */ | |
c5141e6d | 3414 | static const struct pid_entry tid_base_stuff[] = { |
631f9c18 | 3415 | DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations), |
3835541d | 3416 | DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations), |
6b4e306a | 3417 | DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations), |
6ba8ed79 EB |
3418 | #ifdef CONFIG_NET |
3419 | DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations), | |
3420 | #endif | |
631f9c18 | 3421 | REG("environ", S_IRUSR, proc_environ_operations), |
c5317167 | 3422 | REG("auxv", S_IRUSR, proc_auxv_operations), |
631f9c18 | 3423 | ONE("status", S_IRUGO, proc_pid_status), |
35a35046 | 3424 | ONE("personality", S_IRUSR, proc_pid_personality), |
1c963eb1 | 3425 | ONE("limits", S_IRUGO, proc_pid_limits), |
43ae34cb | 3426 | #ifdef CONFIG_SCHED_DEBUG |
631f9c18 | 3427 | REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations), |
ebcb6734 | 3428 | #endif |
1b3044e3 JD |
3429 | NOD("comm", S_IFREG|S_IRUGO|S_IWUSR, |
3430 | &proc_tid_comm_inode_operations, | |
3431 | &proc_pid_set_comm_operations, {}), | |
ebcb6734 | 3432 | #ifdef CONFIG_HAVE_ARCH_TRACEHOOK |
09d93bd6 | 3433 | ONE("syscall", S_IRUSR, proc_pid_syscall), |
43ae34cb | 3434 | #endif |
c2c0bb44 | 3435 | REG("cmdline", S_IRUGO, proc_pid_cmdline_ops), |
631f9c18 AD |
3436 | ONE("stat", S_IRUGO, proc_tid_stat), |
3437 | ONE("statm", S_IRUGO, proc_pid_statm), | |
871305bb | 3438 | REG("maps", S_IRUGO, proc_pid_maps_operations), |
2e13ba54 | 3439 | #ifdef CONFIG_PROC_CHILDREN |
81841161 CG |
3440 | REG("children", S_IRUGO, proc_tid_children_operations), |
3441 | #endif | |
28a6d671 | 3442 | #ifdef CONFIG_NUMA |
871305bb | 3443 | REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations), |
28a6d671 | 3444 | #endif |
631f9c18 AD |
3445 | REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations), |
3446 | LNK("cwd", proc_cwd_link), | |
3447 | LNK("root", proc_root_link), | |
3448 | LNK("exe", proc_exe_link), | |
3449 | REG("mounts", S_IRUGO, proc_mounts_operations), | |
3450 | REG("mountinfo", S_IRUGO, proc_mountinfo_operations), | |
1e883281 | 3451 | #ifdef CONFIG_PROC_PAGE_MONITOR |
631f9c18 | 3452 | REG("clear_refs", S_IWUSR, proc_clear_refs_operations), |
871305bb | 3453 | REG("smaps", S_IRUGO, proc_pid_smaps_operations), |
493b0e9d | 3454 | REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations), |
32ed74a4 | 3455 | REG("pagemap", S_IRUSR, proc_pagemap_operations), |
28a6d671 EB |
3456 | #endif |
3457 | #ifdef CONFIG_SECURITY | |
631f9c18 | 3458 | DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations), |
28a6d671 EB |
3459 | #endif |
3460 | #ifdef CONFIG_KALLSYMS | |
edfcd606 | 3461 | ONE("wchan", S_IRUGO, proc_pid_wchan), |
28a6d671 | 3462 | #endif |
2ec220e2 | 3463 | #ifdef CONFIG_STACKTRACE |
35a35046 | 3464 | ONE("stack", S_IRUSR, proc_pid_stack), |
28a6d671 | 3465 | #endif |
5968cece | 3466 | #ifdef CONFIG_SCHED_INFO |
f6e826ca | 3467 | ONE("schedstat", S_IRUGO, proc_pid_schedstat), |
28a6d671 | 3468 | #endif |
9745512c | 3469 | #ifdef CONFIG_LATENCYTOP |
631f9c18 | 3470 | REG("latency", S_IRUGO, proc_lstats_operations), |
9745512c | 3471 | #endif |
8793d854 | 3472 | #ifdef CONFIG_PROC_PID_CPUSET |
52de4779 | 3473 | ONE("cpuset", S_IRUGO, proc_cpuset_show), |
a424316c PM |
3474 | #endif |
3475 | #ifdef CONFIG_CGROUPS | |
006f4ac4 | 3476 | ONE("cgroup", S_IRUGO, proc_cgroup_show), |
28a6d671 | 3477 | #endif |
6ba51e37 | 3478 | ONE("oom_score", S_IRUGO, proc_oom_score), |
fa0cbbf1 | 3479 | REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations), |
a63d83f4 | 3480 | REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations), |
4b7d248b | 3481 | #ifdef CONFIG_AUDIT |
631f9c18 | 3482 | REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations), |
26ec3c64 | 3483 | REG("sessionid", S_IRUGO, proc_sessionid_operations), |
28a6d671 | 3484 | #endif |
f4f154fd | 3485 | #ifdef CONFIG_FAULT_INJECTION |
631f9c18 | 3486 | REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations), |
1203c8e6 | 3487 | REG("fail-nth", 0644, proc_fail_nth_operations), |
f4f154fd | 3488 | #endif |
297c5d92 | 3489 | #ifdef CONFIG_TASK_IO_ACCOUNTING |
19aadc98 | 3490 | ONE("io", S_IRUSR, proc_tid_io_accounting), |
297c5d92 | 3491 | #endif |
22d917d8 EB |
3492 | #ifdef CONFIG_USER_NS |
3493 | REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations), | |
3494 | REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations), | |
f76d207a | 3495 | REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations), |
9cc46516 | 3496 | REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations), |
22d917d8 | 3497 | #endif |
7c23b330 JP |
3498 | #ifdef CONFIG_LIVEPATCH |
3499 | ONE("patch_state", S_IRUSR, proc_pid_patch_state), | |
3500 | #endif | |
68bc30bb AL |
3501 | #ifdef CONFIG_PROC_PID_ARCH_STATUS |
3502 | ONE("arch_status", S_IRUGO, proc_pid_arch_status), | |
3503 | #endif | |
28a6d671 EB |
3504 | }; |
3505 | ||
f0c3b509 | 3506 | static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx) |
28a6d671 | 3507 | { |
f0c3b509 AV |
3508 | return proc_pident_readdir(file, ctx, |
3509 | tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
3510 | } |
3511 | ||
00cd8dd3 AV |
3512 | static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
3513 | { | |
7bcd6b0e | 3514 | return proc_pident_lookup(dir, dentry, |
d5a572a4 AD |
3515 | tid_base_stuff, |
3516 | tid_base_stuff + ARRAY_SIZE(tid_base_stuff)); | |
28a6d671 EB |
3517 | } |
3518 | ||
00977a59 | 3519 | static const struct file_operations proc_tid_base_operations = { |
28a6d671 | 3520 | .read = generic_read_dir, |
f50752ea AV |
3521 | .iterate_shared = proc_tid_base_readdir, |
3522 | .llseek = generic_file_llseek, | |
28a6d671 EB |
3523 | }; |
3524 | ||
c5ef1c42 | 3525 | static const struct inode_operations proc_tid_base_inode_operations = { |
28a6d671 EB |
3526 | .lookup = proc_tid_base_lookup, |
3527 | .getattr = pid_getattr, | |
3528 | .setattr = proc_setattr, | |
3529 | }; | |
3530 | ||
0168b9e3 AV |
3531 | static struct dentry *proc_task_instantiate(struct dentry *dentry, |
3532 | struct task_struct *task, const void *ptr) | |
444ceed8 | 3533 | { |
444ceed8 | 3534 | struct inode *inode; |
0168b9e3 | 3535 | inode = proc_pid_make_inode(dentry->d_sb, task, S_IFDIR | S_IRUGO | S_IXUGO); |
444ceed8 | 3536 | if (!inode) |
0168b9e3 | 3537 | return ERR_PTR(-ENOENT); |
1bbc5513 | 3538 | |
444ceed8 EB |
3539 | inode->i_op = &proc_tid_base_inode_operations; |
3540 | inode->i_fop = &proc_tid_base_operations; | |
1bbc5513 | 3541 | inode->i_flags |= S_IMMUTABLE; |
aed54175 | 3542 | |
1270dd8d | 3543 | set_nlink(inode, nlink_tid); |
1bbc5513 | 3544 | pid_update_inode(task, inode); |
444ceed8 | 3545 | |
fb045adb | 3546 | d_set_d_op(dentry, &pid_dentry_operations); |
0168b9e3 | 3547 | return d_splice_alias(inode, dentry); |
444ceed8 EB |
3548 | } |
3549 | ||
00cd8dd3 | 3550 | static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags) |
28a6d671 | 3551 | { |
28a6d671 EB |
3552 | struct task_struct *task; |
3553 | struct task_struct *leader = get_proc_task(dir); | |
28a6d671 | 3554 | unsigned tid; |
b488893a | 3555 | struct pid_namespace *ns; |
0168b9e3 | 3556 | struct dentry *result = ERR_PTR(-ENOENT); |
28a6d671 EB |
3557 | |
3558 | if (!leader) | |
3559 | goto out_no_task; | |
3560 | ||
dbcdb504 | 3561 | tid = name_to_int(&dentry->d_name); |
28a6d671 EB |
3562 | if (tid == ~0U) |
3563 | goto out; | |
3564 | ||
b488893a | 3565 | ns = dentry->d_sb->s_fs_info; |
28a6d671 | 3566 | rcu_read_lock(); |
b488893a | 3567 | task = find_task_by_pid_ns(tid, ns); |
28a6d671 EB |
3568 | if (task) |
3569 | get_task_struct(task); | |
3570 | rcu_read_unlock(); | |
3571 | if (!task) | |
3572 | goto out; | |
bac0abd6 | 3573 | if (!same_thread_group(leader, task)) |
28a6d671 EB |
3574 | goto out_drop_task; |
3575 | ||
0168b9e3 | 3576 | result = proc_task_instantiate(dentry, task, NULL); |
28a6d671 EB |
3577 | out_drop_task: |
3578 | put_task_struct(task); | |
3579 | out: | |
3580 | put_task_struct(leader); | |
3581 | out_no_task: | |
0168b9e3 | 3582 | return result; |
28a6d671 EB |
3583 | } |
3584 | ||
0bc58a91 EB |
3585 | /* |
3586 | * Find the first tid of a thread group to return to user space. | |
3587 | * | |
3588 | * Usually this is just the thread group leader, but if the users | |
3589 | * buffer was too small or there was a seek into the middle of the | |
3590 | * directory we have more work todo. | |
3591 | * | |
3592 | * In the case of a short read we start with find_task_by_pid. | |
3593 | * | |
3594 | * In the case of a seek we start with the leader and walk nr | |
3595 | * threads past it. | |
3596 | */ | |
9f6e963f ON |
3597 | static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos, |
3598 | struct pid_namespace *ns) | |
0bc58a91 | 3599 | { |
d855a4b7 | 3600 | struct task_struct *pos, *task; |
9f6e963f ON |
3601 | unsigned long nr = f_pos; |
3602 | ||
3603 | if (nr != f_pos) /* 32bit overflow? */ | |
3604 | return NULL; | |
1da177e4 | 3605 | |
cc288738 | 3606 | rcu_read_lock(); |
d855a4b7 ON |
3607 | task = pid_task(pid, PIDTYPE_PID); |
3608 | if (!task) | |
3609 | goto fail; | |
3610 | ||
3611 | /* Attempt to start with the tid of a thread */ | |
9f6e963f | 3612 | if (tid && nr) { |
b488893a | 3613 | pos = find_task_by_pid_ns(tid, ns); |
d855a4b7 | 3614 | if (pos && same_thread_group(pos, task)) |
a872ff0c | 3615 | goto found; |
0bc58a91 | 3616 | } |
1da177e4 | 3617 | |
0bc58a91 | 3618 | /* If nr exceeds the number of threads there is nothing todo */ |
9f6e963f | 3619 | if (nr >= get_nr_threads(task)) |
c986c14a | 3620 | goto fail; |
1da177e4 | 3621 | |
a872ff0c ON |
3622 | /* If we haven't found our starting place yet start |
3623 | * with the leader and walk nr threads forward. | |
0bc58a91 | 3624 | */ |
d855a4b7 | 3625 | pos = task = task->group_leader; |
c986c14a | 3626 | do { |
9f6e963f | 3627 | if (!nr--) |
c986c14a | 3628 | goto found; |
d855a4b7 | 3629 | } while_each_thread(task, pos); |
c986c14a ON |
3630 | fail: |
3631 | pos = NULL; | |
3632 | goto out; | |
a872ff0c ON |
3633 | found: |
3634 | get_task_struct(pos); | |
3635 | out: | |
cc288738 | 3636 | rcu_read_unlock(); |
0bc58a91 EB |
3637 | return pos; |
3638 | } | |
3639 | ||
3640 | /* | |
3641 | * Find the next thread in the thread list. | |
3642 | * Return NULL if there is an error or no next thread. | |
3643 | * | |
3644 | * The reference to the input task_struct is released. | |
3645 | */ | |
3646 | static struct task_struct *next_tid(struct task_struct *start) | |
3647 | { | |
c1df7fb8 | 3648 | struct task_struct *pos = NULL; |
cc288738 | 3649 | rcu_read_lock(); |
c1df7fb8 | 3650 | if (pid_alive(start)) { |
0bc58a91 | 3651 | pos = next_thread(start); |
c1df7fb8 ON |
3652 | if (thread_group_leader(pos)) |
3653 | pos = NULL; | |
3654 | else | |
3655 | get_task_struct(pos); | |
3656 | } | |
cc288738 | 3657 | rcu_read_unlock(); |
0bc58a91 EB |
3658 | put_task_struct(start); |
3659 | return pos; | |
1da177e4 LT |
3660 | } |
3661 | ||
3662 | /* for the /proc/TGID/task/ directories */ | |
f0c3b509 | 3663 | static int proc_task_readdir(struct file *file, struct dir_context *ctx) |
1da177e4 | 3664 | { |
d855a4b7 ON |
3665 | struct inode *inode = file_inode(file); |
3666 | struct task_struct *task; | |
b488893a | 3667 | struct pid_namespace *ns; |
f0c3b509 | 3668 | int tid; |
1da177e4 | 3669 | |
d855a4b7 | 3670 | if (proc_inode_is_dead(inode)) |
f0c3b509 | 3671 | return -ENOENT; |
1da177e4 | 3672 | |
f0c3b509 | 3673 | if (!dir_emit_dots(file, ctx)) |
d855a4b7 | 3674 | return 0; |
1da177e4 | 3675 | |
0bc58a91 EB |
3676 | /* f_version caches the tgid value that the last readdir call couldn't |
3677 | * return. lseek aka telldir automagically resets f_version to 0. | |
3678 | */ | |
76f668be | 3679 | ns = proc_pid_ns(inode); |
f0c3b509 AV |
3680 | tid = (int)file->f_version; |
3681 | file->f_version = 0; | |
d855a4b7 | 3682 | for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns); |
0bc58a91 | 3683 | task; |
f0c3b509 | 3684 | task = next_tid(task), ctx->pos++) { |
e3912ac3 | 3685 | char name[10 + 1]; |
a4ef3895 | 3686 | unsigned int len; |
b488893a | 3687 | tid = task_pid_nr_ns(task, ns); |
e3912ac3 | 3688 | len = snprintf(name, sizeof(name), "%u", tid); |
f0c3b509 AV |
3689 | if (!proc_fill_cache(file, ctx, name, len, |
3690 | proc_task_instantiate, task, NULL)) { | |
0bc58a91 EB |
3691 | /* returning this tgid failed, save it as the first |
3692 | * pid for the next readir call */ | |
f0c3b509 | 3693 | file->f_version = (u64)tid; |
0bc58a91 | 3694 | put_task_struct(task); |
1da177e4 | 3695 | break; |
0bc58a91 | 3696 | } |
1da177e4 | 3697 | } |
d855a4b7 | 3698 | |
f0c3b509 | 3699 | return 0; |
1da177e4 | 3700 | } |
6e66b52b | 3701 | |
a528d35e DH |
3702 | static int proc_task_getattr(const struct path *path, struct kstat *stat, |
3703 | u32 request_mask, unsigned int query_flags) | |
6e66b52b | 3704 | { |
a528d35e | 3705 | struct inode *inode = d_inode(path->dentry); |
99f89551 | 3706 | struct task_struct *p = get_proc_task(inode); |
6e66b52b EB |
3707 | generic_fillattr(inode, stat); |
3708 | ||
99f89551 | 3709 | if (p) { |
99f89551 | 3710 | stat->nlink += get_nr_threads(p); |
99f89551 | 3711 | put_task_struct(p); |
6e66b52b EB |
3712 | } |
3713 | ||
3714 | return 0; | |
3715 | } | |
28a6d671 | 3716 | |
c5ef1c42 | 3717 | static const struct inode_operations proc_task_inode_operations = { |
28a6d671 EB |
3718 | .lookup = proc_task_lookup, |
3719 | .getattr = proc_task_getattr, | |
3720 | .setattr = proc_setattr, | |
0499680a | 3721 | .permission = proc_pid_permission, |
28a6d671 EB |
3722 | }; |
3723 | ||
00977a59 | 3724 | static const struct file_operations proc_task_operations = { |
28a6d671 | 3725 | .read = generic_read_dir, |
f50752ea AV |
3726 | .iterate_shared = proc_task_readdir, |
3727 | .llseek = generic_file_llseek, | |
28a6d671 | 3728 | }; |
1270dd8d AD |
3729 | |
3730 | void __init set_proc_pid_nlink(void) | |
3731 | { | |
3732 | nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff)); | |
3733 | nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff)); | |
3734 | } |